December 2014

Study on the Impact of Food Information

on Consumers’ Decision Making

TNS European Behaviour Studies Consortium

FINAL REPORT
Impact of Food Information on Consumers’ Decision Making

REPORTING CONVENTIONS

This is the final report of the “study on the impact of food information on consumers’ decision
making”. It was launched under the framework contract (No EAHC/2011/CP/01) on
behavioural studies by The Consumer, Health and Food Executive Agency (CHAFEA) acting
under its mandate from the European Commission. The Directorate General Health and
Consumers commissioned the TNS European Behaviour Studies Consortium with this research
in December 2012.

This report presents main findings from the primary research conducted in September 2013
and in July 2014. The research was designed to meet the following two key objectives:
 Explore whether and how consumers currently access, understand and use specific
food information
 Provide empirical evidence on the effectiveness of alternative treatments

Detailed survey documentation will be available in a separate appendix document.

Disclaimer:
The content of this report represents the views of the TNS European Behaviour Studies
Consortium and is its sole responsibility. It can in no way be taken to reflect the views of the
European Commission and/or CHAFEA or any other body of the European Union. The European
Commission and/or CHAFEA do not guarantee the accuracy of the data included in this report,
nor do they accept responsibility for any use made by third parties thereof.

TNS Consortium and authors:

 Elke Himmelsbach - TNS Infratest Munich
 Anthony Allen - TNS opinion, Brussels
 Mark Francas - TNS Global, Perth

2
Impact of Food Information on Consumers’ Decision Making

TABLE OF CONTENTS

REPORTING CONVENTIONS ................................................................................... 2 

LIST OF FIGURES AND TABLES .............................................................................. 6 

ABBREVIATIONS AND GLOSSARY OF TERMS........................................................ 10 

1  EXECUTIVE SUMMARY ................................................................................... 12 

1.1  Background and objectives .................................................................................... 12  

1.2  Research methodology .......................................................................................... 13  
1.3  Findings and implications – Trans Fats .................................................................. 13 
1.4  Findings– Precautionary Allergen Warnings (PAWs) ............................................. 16 
1.5  Findings and implications – Alcohol Consumption ................................................. 18 
1.6  Findings and implications – Food waste................................................................. 22 

2  INTRODUCTION ............................................................................................ 23  

2.1  Background............................................................................................................ 23  
2.1.1  EU legislation on food labelling (FIC regulation) ...................................23 
2.1.2  The changing role of research in food labelling regulation .....................25 
2.1.3  Key elements of researching a purchase decision .................................26 
2.2  Research objectives ............................................................................................... 29  
2.3  Research design overview ..................................................................................... 30 

3  EXPERIMENTAL ‘LAB’ DESIGN ....................................................................... 33 

3.1  Overview ............................................................................................................... 33  

3.2  Respondent segmentation ..................................................................................... 34 
3.3  Experiments on TFA information ............................................................................ 36 
3.3.1  Product category for TFA ..................................................................36 
3.3.2  Framing of TFA experiments..............................................................37 
3.3.3  Treatment splits for TFA ...................................................................38 
3.3.4  Choice tasks and choice characteristics for TFA....................................41 
3.4  Experiments on PAW ............................................................................................. 43 
3.4.1  Allergen and product category for PAW ...............................................44 
3.4.2  Framing of PAW experiments ............................................................45 
3.4.3  Treatment splits for PAW ..................................................................46 
3.4.4  Choice tasks and choice characteristics for PAW ..................................47 
3.5  Experiments on alcohol consumption..................................................................... 49 
3.5.1  Product categories for alcohol consumption .........................................49 
3.5.2  Framing of alcohol experiments .........................................................50 

3
Impact of Food Information on Consumers’ Decision Making

3.5.3  Treatment splits for alcohol ...............................................................52 

3.5.4  Choice tasks and choice characteristics for alcohol consumption ............56 

4  EXPERIMENTAL ‘LAB’ FINDINGS ON TRANS FATS .......................................... 60 

4.1  Observed choice behaviour .................................................................................... 61 

4.1.1  Potential awareness drivers...............................................................62 
4.1.2  Potential beliefs & attitude drivers .....................................................67 
4.1.3  Experimental treatments ..................................................................70 
4.2  Findings of the multivariate analysis ..................................................................... 72 
4.2.1  Main biases and barriers of healthy TFA choices...................................73 
4.2.2  Effectiveness of tested treatments on TFA...........................................80 
4.3  Conclusions regarding TFA..................................................................................... 88  

5  EXPERIMENTAL ‘LAB’ FINDINGS ON PRECAUTIONARY ALLERGEN WARNINGS

(PAW) ................................................................................................................. 92  

5.1  Consumers’ cognitive risk assessment of PAWs ..................................................... 94 

5.2  Observed choice behaviour .................................................................................... 96 
5.2.1  Potential awareness drivers...............................................................99 
5.2.2  Potential beliefs and attitude drivers ................................................ 101 
5.2.3  Potential habitual drivers ................................................................ 103 
5.2.4  Socio-demographic variables ........................................................... 103 
5.2.5  Treatments ................................................................................... 104 
5.3  Findings of multivariate analysis ......................................................................... 107  
5.3.1  Main biases and barriers of PAW avoiding behaviour .......................... 107 
5.3.2  Effectiveness of tested PAW options ................................................. 110 
5.4  Conclusions regarding PAWs ............................................................................... 115 

6  EXPERIMENTAL ‘LAB’ FINDINGS ON ALCOHOL CONSUMPTION .................... 120 

6.1  Observed choice behaviour .................................................................................. 121 

6.1.1  Potential beliefs and attitude drivers ................................................ 124 
6.1.2  Potential habitual drivers ................................................................ 128 
6.1.3  Socio-demographic variables ........................................................... 128 
6.1.4  Treatments ................................................................................... 129 
6.2  Findings of multivariate analysis ......................................................................... 132  
6.2.1  Main biases and barriers on alcohol consumption ............................... 133 
6.2.2  Effectiveness of tested treatments on alcohol consumption ................. 138 
6.3  Conclusions regarding alcohol consumption ........................................................ 141 

7  FINDINGS ON DATE MARKING AND FOOD WASTE ........................................ 149 

7.1  Knowledge test and claimed behaviour................................................................ 149 

7.2  Findings of multivariate analysis ......................................................................... 155  

4
Impact of Food Information on Consumers’ Decision Making

8  DESIGN OF STORE EXPERIMENTS ON TRANS FAT ........................................ 160 

8.1  Updated research objectives for TFA experiments ............................................... 160 

8.2  Overview of experimental design considerations ................................................. 161 
8.3  Product categories tested .................................................................................... 163  
8.4  Treatment splits for TFA transparency ................................................................. 164 
8.5  Selection of markets and respondent screening .................................................. 167 
8.6  Framing of TFA experiments ................................................................................ 169 
8.7  Choice tasks and choice characteristics ............................................................... 169 

9  FINDINGS ON STORE EXPERIMENTS REGARDING TRANS FATS .................... 175 

9.1  Observed choice behaviour at the shelf and stated preferences .......................... 176 
9.2  Observed initial choice behaviour – disaggregated findings ................................ 187 
9.2.1  Rationale given for initial choices ..................................................... 189 
9.2.2  Potential access and awareness drivers ............................................ 190 
9.2.3  Potential goals and habitual drivers .................................................. 191 
9.2.4  Treatments ................................................................................... 192 
9.3  Observed ability to identify healthy products – disaggregated analysis ............... 195 
9.3.1  Rationale given for initial choices ..................................................... 197 
9.3.2  Potential access and awareness drivers ............................................ 198 
9.3.3  Potential goals and habitual drivers .................................................. 200 
9.3.4  Treatments ................................................................................... 200 
9.4  Findings of the multivariate analysis ................................................................... 202 
9.4.1  Drivers of initial healthy product choices in Task 1 ............................. 203 
9.4.2  Drivers of initial healthy product choices in Task 2 ............................. 204 
9.4.3  Drivers of the ability to identify healthy products in Task 1 ................. 206 
9.4.4  Drivers of the ability to identify healthy products in Task 2 ................. 207 
9.5  Conclusions regarding TFA................................................................................... 208  

APPENDIX A – TECHNICAL DETAILS OF THE SURVEY 

APPENDIX B – QUESTIONNAIRE 

APPENDIX C – STATISTICAL MODELS 

5
Impact of Food Information on Consumers’ Decision Making

LIST OF FIGURES AND TABLES

Figure 1 Research design overview ........................................................................... 13 

Figure 2 Research design overview ........................................................................... 31 
Figure 3 Questionnaire flow for respondents .............................................................. 34 
Figure 4 Example of an individualised introduction for TFA experiments ......................... 38 
Figure 5 Example without TFA and example with TFA .................................................. 39 
Figure 6 Education intervention about fat types .......................................................... 40 
Table 7 TFA condition splits ..................................................................................... 41 
Figure 8 TFA choice Task 1 – example with TFA information and zoomed labels .............. 42 
Figure 9 TFA choice Task 2 – example with TFA information and zoomed labels .............. 43 
Figure 10 Example for individualised introduction for PAW experiments.......................... 46 
Table 11 PAW treatment splits ................................................................................. 47 
Figure 12 PAW choice Task 1 – example with ‘May contain’ and zoomed labels................ 48 
Table 13 Serving sizes used for collecting drinking volumes ......................................... 51 
Figure 14 Introduction for alcohol experiments – example for beer ................................ 51 
Figure 15 Knowledge test on calories for beer ............................................................ 53 
Figure 16 Education on calories for beer .................................................................... 54 
Figure 17 Knowledge test on limits for wine ............................................................... 55 
Figure 18 Education on limits for wine ....................................................................... 55 
Table 19 Alcohol condition splits ............................................................................... 56 
Figure 20 Alcohol choice 1 – example with wine and calories treatment ......................... 57 
Figure 21 Alcohol choice 1 – example with wine and limits treatment ............................ 58 
Figure 22 Alcohol choice 2 – example with wine ......................................................... 58 
Figure 23 Guideline for analysis of TFA results............................................................ 61 
Table 24 Observed choice behaviour in TFA Tasks 1 and 2 ........................................... 62 
Figure 25 Awareness of fat types.............................................................................. 63 
Figure 26 Evaluation of fat types .............................................................................. 63 
Figure 27 Pre-experimental awareness of fat types vs. Task 1 behaviour........................ 64 
Figure 28 Pre-experimental awareness of fat types vs. Task 2 behaviour........................ 65 
Table 29 Zooming behaviour in Task 1 ...................................................................... 66 
Table 30 Zooming behaviour in Task 2 ...................................................................... 67 
Figure 31 Shopper types ......................................................................................... 68 
Table 32 Shopper type summary on TFA experiments ................................................. 70 
Table 33 Treatments in Task 1 ................................................................................. 71 
Table 34 Treatments in Task 2 ................................................................................. 72 

6
Impact of Food Information on Consumers’ Decision Making

Figure 35 Drivers of healthy choices in TFA Task 1 without treatment options ................. 74 
Figure 36 Drivers of healthy choices in TFA Task 2 without treatment options ................. 77 
Figure 37 Drivers of healthy choices in both TFA tasks without treatment options ............ 79 
Figure 38 Drivers of healthy choices in TFA Task 1 with treatment options ..................... 82 
Figure 39 Drivers of healthy choices in TFA Task 2 with treatment options ..................... 83 
Figure 40 Drivers of healthy choices in TFA tasks 1+2 with treatments .......................... 84 
Figure 41 Drivers of zooming any ingredients or nutrition facts in tasks 1+2................... 87 
Figure 42 Summary of results on TFA analysis ........................................................... 89 
Figure 43 Relevance of allergen/PAW information ....................................................... 93 
Figure 44 Guideline for analysis of results on PAW ...................................................... 94 
Figure 45 Risk assessment of PAWs .......................................................................... 95 
Figure 46 Cognitive risk assessment of PAW wordings ................................................. 96 
Figure 47 PAW choice behaviour by affected and not affected respondents ..................... 98 
Figure 48 Choice rationale for PAW experiments ......................................................... 99 
Table 49 Zooming vs. choice behaviour in Task 1 ...................................................... 100 
Table 50 Zooming vs. choice behaviour in Task 2 ...................................................... 100 
Table 51 Zooming vs. choice behaviour in Task 3 ...................................................... 101 
Figure 52 Selection of product without PAW vs. severity of allergy reaction ................... 102 
Table 53 PAW choice behaviour vs. treatments.......................................................... 104 
Figure 54 Cognitive risk assessment vs. observed avoidance behaviour ........................ 105 
Figure 55 Selection of product without PAW based on risk assessment of PAW labels ...... 106 
Figure 56 Drivers of PAW avoiding choices without treatments..................................... 109 
Figure 57 Impact of PAW wording on cautious choices ................................................ 111 
Figure 58 Drivers of PAW avoiding choices with PAW options ....................................... 112 
Figure 59 Drivers of PAW avoiding choices with PAW options and interaction effects ....... 114 
Figure 60 Summary of results on PAW analysis ......................................................... 116 
Figure 61 Guideline for analysis of results on alcohol consumption ............................... 120 
Table 62 Observed choice behaviour in alcohol tasks 1 and 2 ...................................... 122 
Table 63 Intention to drink less on specific occasion vs. any occasion in future .............. 122 
Table 64 Observed choice behaviour in Task 1 vs. alcohol category .............................. 123 
Table 65 Observed choice behaviour in Task 2 vs. alcohol category .............................. 123 
Figure 66 Intention to drink less vs. image and peer group behaviour ........................... 124 
Figure 67 Intention to drink less vs. choice preferences .............................................. 125 
Figure 68 Intention to drink less vs. motivation type .................................................. 126 
Figure 69 Intention to drink less vs. self-control and risk attitude................................. 127 
Table 70 Choice behaviour in Task 1 vs. treatments ................................................... 129 

7
Impact of Food Information on Consumers’ Decision Making

Table 71 Choice behaviour in Task 2 vs. treatments ................................................... 129 

Figure 72 Intention to drink less vs. alcohol category and treatments ........................... 129 
Figure 73 Knowledge test results vs. alcohol category and treatments .......................... 131 
Table 74 Susceptibility towards calories treatment vs. alcohol category ........................ 132 
Figure 75 Drivers of choice behaviour in alcohol Task 1 without treatments ................... 135 
Figure 76 Drivers of choice behaviour in alcohol Task 2 without treatments ................... 137 
Figure 77 Drivers of choice behaviour in alcohol Task 1 with treatments ....................... 139 
Figure 78 Drivers of choice behaviour in alcohol Task 2 with treatments ....................... 140 
Figure 79 Summary of results on alcohol consumption analysis ................................... 142 
Table 80 Knowledge test results of best before by country .......................................... 149 
Table 81 Knowledge test results of best before by sociodemographics .......................... 150 
Figure 82 Knowledge test on expiry dates................................................................. 151 
Table 83 Knowledge test results of best before vs. use by ........................................... 152 
Figure 84 Food waste consumer behaviour ............................................................... 153 
Table 85 Food waste behaviour type vs. ‘best before’ awareness ................................. 154 
Figure 86 Drivers of aversion to consume outdated food ............................................. 156 
Figure 87 Drivers of throwing away not-outdated food................................................ 158 
Figure 88 Images of the data collection process in supermarkets ................................. 162 
Figure 89 Product images ....................................................................................... 164 
Figure 90 Example without TFA (control group) and with TFA (treatments 2/3) .............. 165 
Figure 91 Education treatment for TFA field experiment .............................................. 166 
Figure 92 Store segmentation ................................................................................. 167 
Table 93 Sample sizes per split ............................................................................... 168 
Table 94 Overview of screen out process .................................................................. 168 
Figure 95 Introduction to TFA experiments ............................................................... 169 
Figure 96 Overview of TFA experiments flow for respondents ...................................... 171 
Figure 97 Choice architecture of TFA experiments ...................................................... 172 
Figure 98 Feedback with choice characteristics of yoghurt category in Task 1 ................ 173 
Figure 99 Feedback with choice characteristics of pizza category in Task 2 .................... 174 
Figure 100 Guideline for analysis of results on store experiments ................................. 175 
Table 101 Shopping and reading behaviour at the shelf .............................................. 176 
Table 102 Average duration of shopping processes at the shelf .................................... 177 
Figure 103 Type of food information read at the shelf ................................................. 178 
Figure 104 Relevance of shopping goals ................................................................... 179 
Figure 105 Relevance of health goal by product categories .......................................... 180 
Figure 106 Healthiness assessment of nutrition elements ............................................ 181 

8
Impact of Food Information on Consumers’ Decision Making

Figure 107 Surprise effect of TFA disclosure .............................................................. 182 

Figure 108 Confusion effect of TFA disclosure ............................................................ 183 
Figure 109 Strategy of ignorance on TFA disclosure ................................................... 184 
Figure 110 Strategy of avoidance on TFA disclosure ................................................... 185 
Figure 111 Preferred interventions by consumers ...................................................... 186 
Table 112 Initial choice behaviour in Tasks 1 and 2 .................................................... 187 
Table 113 Initial choice behaviour in Task 1 by product category ................................. 188 
Table 114 Initial choice behaviour in Task 2 by product category ................................. 188 
Figure 115 Rationale for initial choice Task 1 if product A or B was chosen .................... 189 
Figure 116 Rationale for initial choice Task 2 if product X or Y was chosen .................... 190 
Table 117 Initial choice behaviour in Task 1 by accessed information............................ 191 
Table 118 Initial choice behaviour in Task 2 by accessed information............................ 191 
Table 119 Initial choice behaviour in Task 1 by treatments ......................................... 193 
Table 120 Initial choice behaviour in Task 1 by treatments and products....................... 193 
Table 121 Initial choice behaviour in Task 2 by treatments ......................................... 194 
Table 122 Initial choice behaviour in Task 2 by treatments and products....................... 194 
Table 123 Identification of healthy products in Task 1 and 2........................................ 195 
Table 124 Identification of healthy product in Task 1 by product category ..................... 196 
Table 125 Identification of healthy product in Task 2 by product category ..................... 196 
Figure 126 Rationale for healthy product identification in Task 1 .................................. 197 
Figure 127 Rationale for healthy product identification in Task 2 .................................. 198 
Table 128 Healthy product identification in Task 1 by accessed information ................... 199 
Table 129 Healthy product identification in Task 2 by accessed information ................... 199 
Table 130 Healthy product identification in Task 1 by treatments ................................. 201 
Table 131 Healthy product identification in Task 1 by treatments and products .............. 201 
Table 132 Healthy product identification in Task 2 by treatments ................................. 202 
Table 133 Healthy product identification in Task 2 by treatments and products .............. 202 
Figure 134 Drivers of initial choice of healthy product in Task 1 ................................... 203 
Figure 135 Drivers of initial choice of healthy product in Task 2 ................................... 205 
Figure 136 Drivers of the ability to identify healthy products in Task 1 .......................... 206 
Figure 137 Drivers of the ability to identify healthy products in Task 2 .......................... 207 

9
Impact of Food Information on Consumers’ Decision Making

ABBREVIATIONS AND GLOSSARY OF TERMS

Alcopop
A colloquial term describing a ready-mixed drink that contains alcohol – also called flavoured
alcoholic beverages, including:
 malt beverages to which various fruit juices or other flavourings have been added
 beverages containing wine to which ingredients such as fruit juice or other flavourings
have been added (wine coolers)
 beverages containing distilled alcohol and added ingredients such as fruit juices or
other flavourings

BMI – Body Mass Index

A measure for human body shape based on an individual's mass and height

FHO – Fully hydrogenated oil

Full hydrogenation of unsaturated fats produces virtually only saturated fats and no trans
fats.

Food Information
This is all types of information concerning a food that is made available to the final consumer
by means of a label, other accompanying material, or any other means including modern
technology tools or verbal communication.

PHO – Partially hydrogenated oil

In the process of partial hydrogenation, both trans fats and saturated fats are generated in
varying proportions depending on the process conditions.

PAW – Precautionary allergen warning

A food allergen precautionary statement (e.g. may contain nuts) is a declaration on the label of a
pre-packaged food of the possible inadvertent presence of an allergen in the food.

Saturates or saturated fat

Contain no double bonds between the carbon atoms, so it is saturated with hydrogen

TFA – Trans fatty acids

‘Trans fat’ means fatty acids with at least one non-conjugated (namely interrupted by at least
one methylene group) carbon-carbon double bond in the trans configuration (Regulation (EU)
1169/2011); trans fat can occur naturally in milk and meat products from certain animals
(ruminants, like cows or sheep) or artificially as a result of food processing; the main source
of artificial or industrially produced trans fat is PHO; high intakes of trans fat has been

10
Impact of Food Information on Consumers’ Decision Making

consistently found to be associated with increased risk of coronary heart disease (CHD) and
trans fat intakes "should be as low as is possible within the context of a nutritionally adequate
diet"1 .

1
EFSA Journal. 2010;8(3):1461

11
Impact of Food Information on Consumers’ Decision Making

1 EXECUTIVE SUMMARY

1.1 Background and objectives

Never before has so much information on food and healthy diets been provided to consumers.
The wealth of health claims, quality labels, nutrition facts, apps and advice, together with
sometimes misleading or contradictory marketing information, is overwhelming for many
citizens. Moreover, it represents a contextual choice architecture that may contribute to
consumer confusion across all education levels.
Consumers develop different strategies to cope with this overload of information. These can
range from ignoring any information or delegating responsibility for decisions to family
members or brands (i.e. always buying the same products), or spending too much time and
thought on the right choice. Between these extremes, there are several ‘behavioural
segments’ of consumers who do not ignore but drastically simplify information intake based
on heuristic shortcuts.
On the one hand most stakeholders agree that there is a need for more clarity and
understanding regarding food information. On the other hand there seems to be little
consensus on solutions, despite the publication of numerous studies testing different formats
and content of food information with varying outcomes.
Therefore, this study aimed to learn from existing evidence and establish a robust
methodology that takes into account all issues relating to the following key outcome
objectives:
a) Status quo description of consumer attitudes when it comes to food information, i.e.
explaining the capacity to access, interpret and use existing food product information
specifically on allergens, trans fats, origin of primary ingredients, date labels, health
warnings and labelling of alcoholic beverages as well as further quality label information
and thereby:
 Consider individual consumer features and identify relevant biases and constraints
such as cognitive abilities, time resources, repetitive purchases, etc.;
 Understand external determinants and different framings of the choice situation
contributing to “bad consumer choices” by fostering misunderstandings, information
overload or ignorance;
b) Identifying the optimal aspects of presentation of food information for the consumer to
make better (informed) decisions and thereby:
 Consider a holistic approach, i.e. assess the interactive effect of one particular label
on one particular kind of product in the context of increasing complexity of labelling in
the marketplace;
 Look for the best compromise between exhaustiveness and clarity of the label,
exploring the possible use of symbols or other means instead of words.

12
Impact of Food Information on Consumers’ Decision Making

1.2 Research methodology

The objectives outlined above are best addressed by applying behavioural experiments that
are complemented by classical survey design.
Figure 1 Research design overview

Preparatory stage Online ‘Laboratory’ Offline ‘Field & Lab’

experiments or store experiments
 Review existing  Experiments covering  Experiments on
evidence and literature several policy areas trans fat only
 Develop hypothesis  8 countries  1 country (Germany)
 Consider relevant  N = 8076 interviews  N = 4841 observations
issues for the design of and observations  N = 1851 interviews
experiments

The preparatory stage comprised a review of existing evidence - via desk research, combined
with a series of meetings of key stakeholders at the European Commission - to scope out the
hypotheses to be tested in the experiments and decide on the overall survey design.
Online ‘Laboratory’ (‘lab’) experiments were conducted in eight member states (United
Kingdom, France, Germany, Italy, Spain, Finland, Poland, and Romania) in September 2013.
A questionnaire with online panellists included the simulation of various shopping and
consumption scenarios, to collect relevant choice observations on the following policy areas:
trans fats, precautionary allergen warning, health warnings and calorific information on
alcoholic drinks. Further information without experiments was collected on two topics: origin
of primary ingredients and food waste.
Offline ‘Field & Lab’ or store experiments were conducted in super markets in one
member state (Germany) in July 2014. With the consent of a large retailer chain, customers
were observed when shopping for specific food categories (yoghurt, frozen pizza and biscuits)
and subsequently screened for immediate participation in a ‘lab’ experiment at a stand with
mock-up products of the category they just shopped. The focus of this experiment was on
testing the reaction to different levels of transparency of trans fat information.

1.3 Findings and implications – Trans Fats

What were the key drivers of healthier choices when selecting products
containing TFA?


The evidence from the online and offline experiments confirms that being
able to identify the healthier alternative is largely driven by initially
accessing the relevant information on the food label and the familiarity
with the choice architecture.

Checking the nutrition label encouraged healthier choices in the experiments. Accessing
the nutrition label had a stronger positive effect in driving healthier choices than the

13
Impact of Food Information on Consumers’ Decision Making

ingredients list, possibly because the more structured design of the nutrition label is better
suited to show relevant food information (amounts in particular).
However irrelevant information (the country of origin, in this case) tended to have a
negative effect in these tasks. They presented a strong barrier as they either highlighted other
consumer preferences (such as sustainability) or distracted consumers from the relevant
information about fat types.
Familiarity with at least some of the choice architecture was the second most important
driver of healthy choices. The less familiar the task and the terminology employed in the
experiment, the less likely people were to choose the healthier option.
Previous knowledge about saturates and TFA being unhealthy also had a positive effect
as did consumers with an interest in looking for healthier options (e.g. people on a diet).
These were likely to both look at the relevant information and also make the right choice in
the task.

Would the display of TFA amounts on the nutrition label lead to more
healthy choices or would people misunderstand and over-react (in other
words, not choose a product containing TFA even if it is the healthier
option)?

The empirical evidence in this study suggests that showing the amount of
TFA in the nutrition information enables more healthy choices in a simple


choice context and, in particular, in a scenario where the healthier option
is the product with less TFA. However, this was not observed in a more
complex realistic context, where display of TFA can cause some
overreactions in a scenario where the healthier choice had slightly more
TFA but a significantly less healthy composition of other nutrients (salt,
sugar, saturates).

The transparency of TFA amounts had a clearly positive impact in the choices where the
healthier option is simply the product with less TFA (according to both experiments) However,
in the online experiments for the second task, where TFA needed to be balanced against
saturated fats, and the TFA containing product was still the overall healthier choice
considering the high amounts of saturated fats, the presence of TFA on the label misled some
participants in their choice.
In the field experiment, this overreaction was observed at an even lower scale. A possible
explanation is the more complex choice architecture as well as the offline field environment
in this experiment. When making complex choices in a busy supermarket, people were less
likely to be influenced by information that they are unfamiliar with (such as TFA).
The empirical evidence from the store experiments is more useful to predict real consumer
behaviour in the initial implementation phase. However, the more familiar consumers become
with the term TFA or PHO (through communication campaigns), the more consumers will
over-react to the presence of TFA, with respect to what was observed in the online survey.
It is worth stressing that the experimental choice context was overly simplified. In real life,
the choice context would be significantly more complex, as all elements in the nutritional
declaration are expected to differ among any two products. It might be expected that

14
Impact of Food Information on Consumers’ Decision Making

consumers would be less likely to make the healthier choice in more complex choice contexts
although this assumption is beyond the scope of this particular piece of research.

How can accessing food information and the clarity of the choice
architecture be improved in order to encourage more healthy choices?
Does education help consumers to identify the healthier product?

The education intervention did not help consumers changing the negative
overreaction on TFA transparency into a positive impact in the laboratory
experiment.

 In the store experiments – however, the reading of the nutrition

guidelines was not as effective. Consumers who read the information
quickly made worse decisions, while consumers who read the information
thoroughly did not perform differently from the control group who had no
information.

A simple educational intervention alone (without TFA amounts on the nutrition label) was not
effective at all in influencing healthy consumer choices. The nutrition guideline explaining the
healthiness of different fat types only had an impact if the terminology used (e.g. TFA) is also
found on the nutrition label.
As expected the education treatment in combination with TFA transparency was a strong
driver in healthy choices in task 1, where the healthier product was the one with less TFA
(according to both online and offline experiments).
The second choice scenario was designed to test any potential overreactions as the overall
healthier product was the one with slightly more TFA. In this task it was important for
consumers to check also other nutrition elements, such as saturates, sugar and salt. From an
expert point of view, one could say this was the more complex task. Interestingly, this task
consistently generated more healthy choices than the first task, which already indicates that
consumers are more able to make healthy choices, if they are more familiar with the
terminology.
However, for the impact of the education intervention, this second task generated different
results in the offline store experiments than in the online lab situation. While education did
not help consumers in the online experiment to avoid overreactions compared to the control
group without treatments, the education treatment in the field experiment actually had a
negative impact. In particular those who only briefly read the nutrition guideline tended to
make worse choices, while those who spent more time reading the leaflet were equally likely
to make healthy choices compared to the control group who were given no information.

What conclusions can be drawn from the evidence collected in the online
study and the store experiments?

15
Impact of Food Information on Consumers’ Decision Making


The introduction of TFA amounts on food labels did not consistently
enable consumers to identify the healthier choice. Providing information
was not effective in preventing consumers’ overreactions. ,

The empirical evidence in the store experiments supports the results of the online study,
though the experiments are not exactly comparable.
Making healthy food choices is predominantly driven by accessing the relevant food
information. In a supermarket field experiment less than a quarter of shoppers actually took
time to read information on the package. Of these, only a fraction of consumers read relevant
food information that informs about the healthiness of the product.
When asked directly, consumers said they would prefer to receive more education on this
topic and more transparency, i.e. the TFA amounts should be stated on the nutrition tables.
However, these two measures are only effectively driving good choices, if the products with
less TFA are overall the healthier option.
Consumers are regularly overconfident that with more transparency they will be able to make
better choices, but the task to determine which combination of nutritional elements is the
healthier option is often more difficult than expected. Therefore, many consumers simplify
their food decisions by using heuristics, such as the availability bias, habits, mental shortcuts,
etc.
The choice scenarios presented to participants showed that most consumers simply did not
recognise the less familiar TFA/PHO information. Only after the educational nudge, this
information was noticed with the outcome of an overreaction among those who only read the
leaflet superficially. This scenario was also seen in the online survey where a slightly stronger
overreaction was measured in a more controlled environment.
Overall, the objective of making healthier food choices easier for consumers is as important
as it is difficult to implement. It requires a high level of expertise, time and commitment to
correctly evaluate the presented information. The additional information about trans fats only
adds another variable to an already complex choice context.
This report focuses only on consumer reactions to TFA transparency.

1.4 Findings– Precautionary Allergen Warnings (PAWs)

Do consumers take a cautious approach when they see a PAW? What are
the main individual biases and barriers restricting consumer choice?


Cautious consumer choice (i.e. avoiding products with a PAW) is largely
driven by a high issue salience (accessing relevant allergen information)
as well as by the belief that a product without a PAW is a safer choice.

16
Impact of Food Information on Consumers’ Decision Making

Around 40% of allergy affected consumers in the tested Member States made cautious
choices (in other words, they avoided products which had a PAW). The key driver of a
cautious choice was accessing the ingredients list whether or not the product contained
a PAW – i.e. accessing the relevant information needed to make the right choice.
Checking the nutrition label made people less likely to make cautious choices as it distracted
attention from the PAW label where it was present. As in the other food experiment the
nutrition label was accessed much more frequently than the ingredients list despite the task
of shopping for someone who is allergic to a specific ingredient.
Checking the country of origin also tended to distract consumers when the choice was
between a local product with a PAW and an imported product without a PAW. In this context
consumers with a strong preference for local products were less likely to make a cautious
choice.

What level of risk is associated by consumers with different types of PAW

wordings as well as with the absence of a PAW?


The lack of any advisory allergen information was seen to represent less
risk than most PAWs by the average allergy-affected consumer. Only the
wording ‘Made in a facility’ was rated at a similar lower risk level as if no
information is given.

To identify the individual risk levels that consumers assigned to various versions of allergen
information, a cognitive risk assessment exercise was conducted after the experiments.
Around two thirds of the allergy affected consumers considered each of the presented PAWs
as definitely or probably unsafe. The risk level was substantially higher for this affected
segment, compared to consumers without personal experience of food allergies or
intolerances.
Respondents clearly rated all PAW options as less risky than the label ‘contains nuts’ and also
as more risky than ‘no information’.

What is the impact of the tested PAW labels on making cautious choices?

Overall, the impact of the PAW wording was only small.

 Only ‘May contain’ was effective in driving more cautious choices among
allergy affected consumers.
‘Cannot guarantee’ was more effective at driving less cautious choices.

17
Impact of Food Information on Consumers’ Decision Making

To identify the most effective PAW in driving cautious choices (avoiding product labelled
with a PAW), the analysis looked at the impact of each tested wording across all choice
tasks.
The ranking of the PAW options based on the cognitive risk assessment only partly matched
the ranking based on the actual observed impact of these options on actual behaviour:

The key difference was the divergent positioning of the label ‘Cannot guarantee’, which
contributed to less cautious choices in the experiment with isolated treatments, but was rated
high risk in the cognitive assessment in a simultaneous evaluation framing.
When controlling for this interaction effect of the ‘Cannot guarantee’ treatment, then its
negative impact on cautious choices increases significantly. Hence, this label represents the
PAW option that is most efficient in driving less cautious choices, while the label ‘May contain’
is most efficient in driving more cautious choices.

How can the choices be made easier for allergy affected consumers?

 Making allergen information clearer in terms of the risk level the product
represents will help to avoid misunderstandings by consumers.

The main policy goal is to identify whether allergy affected consumers have restricted choice
due to misunderstanding the risk represented by Paws. Dealing with this issue requires an
understanding of the process of individual risk assessment and the level of trust in food
information which the study shows can vary depending on the framing that is presented to
consumers.
Cautious choices are more likely to be made by consumers who consider a product without a
PAW as safer compared to a product with a PAW. This is a significant contributor to a restricted
product choice.
If there was more clarity that there was no real difference in risk between a product with or
without a PAW, then fewer consumers would restrict their choices. Alternatively, if there were
common rules when to apply a PAW, in terms of the risk they represent together with a
standardised wording and format to be used by all manufacturers then this also may help
consumers make better informed choices.

1.5 Findings and implications – Alcohol Consumption

What drives or hinders drinking intention?

18
Impact of Food Information on Consumers’ Decision Making


The evidence from this study confirms that the motivation to drink less
depends on the individual relevance of ‘healthy drinking’ as well as the
drinking habits and attitudes of peers.

The key motivation to reduce the drinking volume were healthy preferences and habits, such
as the preference for choosing drinks with lower alcoholic content and following a weight loss
programme.
The key hurdle to intending to drink less was the social acceptance and relevance of alcohol
among peers. As most drinking occasions are social (i.e. with friends or family), one’s own
drinking intentions are highly correlated to the drinking habits of the peer group.
Consumers who drank alcohol more frequently were also less likely to reduce their drinking
volume. Overall, frequency of consuming alcohol had more impact than the amount normally
consumed. The psychological and social barriers were higher for those who frequently drank
than for those who drank only occasionally high volumes.

How does the additional information and education about the CALORIFIC
CONTENT of an alcoholic drink encourage consumers to drink less?


Education and information about CALORIES affected consumers
motivated by health goals – in particular those who check for low
alcoholic content and fewer calories.

When looking at the sample split with the calorific treatment only, it was
possible to identify the drivers of the attitude change based on this new
type of food information, which nudged 16% of consumers to plan to
reduce their alcohol consumption on the specified occasion. The calorific
treatment was more effective:
 Among women
 Among consumers in France
 Among wine drinkers
 If presented with a negative framing, e.g. ‘Wine has more
calories than chocolate’
The calorific treatment was less effective among consumers:
 Who were not interested in any health issues
 Whose peers commonly drank alcohol
 Who had a generally positive view of alcohol

19
Impact of Food Information on Consumers’ Decision Making

How does the additional information and education about the

RECOMMENDED LIMITS of alcohol consumption encourage consumers to
drink less?


Education and information about ‘KNOW YOUR LIMITS’ influenced
consumers with a more general preference for low alcoholic content, and
seemed to reach a slightly wider audience than the calorific treatment.

When looking at the sample split with the ‘Know your limits’
treatment only, it was possible to identify the drivers of the attitude
change based on this type of food information, which nudged 19%
of consumers into reducing their planned alcohol consumption on
the specified occasion.
The limits treatment was more effective among consumers with a
lower level of full-time education. This indicates that the ultimate
message of ‘Know your limits’ is easily understood and motivating.
At the same time the limits treatment was less effective among
consumers:
 That were self-determined in their choices and not interested in health issues
 Whose peers commonly drank alcohol
 Who were familiar with the drinking scenario presented in the experiment
 In UK and Germany2
 In households with children

What conclusions can be drawn on the effectiveness of each treatment in

motivating consumers to drink less alcohol at the specific occasion?


While the ‘Know your limits’ treatment was slightly more effective than the
calorific treatment, the combination of a negative framing of calories for
wine achieved the highest impact. Further empirical evidence is required to
identify the optimal treatment and context combination.

2
The low impact in these two countries might be due to the fact that this campaign has already been
launched there and therefore may lack the novelty factor. However, this cannot be verified, since the
questionnaire did not include any awareness parameters of this campaign.

20
Impact of Food Information on Consumers’ Decision Making

Both treatments together had a minimal impact on motivation. Even with the forced exposure
in this experimental design, the immediate impact of these information-based measures was
still small.
Having said this, the alcohol warning ‘Know your limits’ was slightly more effective than the
calories treatment because addressing low alcohol content targeted more consumers’ intuitive
preferences than calories as pre-treatment questions showed that few participants take
calories into account, a fact likely to be influenced by the fact that currently consumers are
not usually presented with calorific information but only alcoholic content on alcoholic
beverages.
Whether the impact of calorific information would increase if this information were to become
more widely available on alcoholic drinks might be indicated by looking at a similar post-
treatment question. Respondents were asked about the ideal information to be provided on
alcoholic drinks in the future and almost half (49%) of participants with the calories treatment
wanted information on ‘calories’ for alcoholic drinks. This does indeed suggest its impact will
increase once this type of information is available in reality.
When looking at all tested treatments, the combination of a negative framing3 of the education
on calories for wine displayed the highest effectiveness with 21% of consumers intending to
drink less. On the other hand the least effective treatment was the positive framing4 of the
education on calories for beer which resulted in only 13% of consumers planning to reduce
their alcohol consumption.

What factors impact the more general and long-term motivation to drink
less alcohol?


The general motivation to drink less in future is not driven by a one-time
exposure to new information on calories or recommended limits. Key
barriers to overcome are peer group drinking and positive alcohol image.

Overall, the willingness to reduce alcohol consumption in the longer term was the same level
as for the specific occasion (17%). As expected the second decision was heavily influenced
by the first decision. Two thirds of all respondents who were motivated to drink less in the
concrete scenario of a friend’s party also intended to drink less in the future.
None of the treatment splits exerted any significant influence on this second decision. Instead
the analysis reveals that social norms and alcohol image impacted the second decision in the
same way. These two barriers are the recommended starting points to be targeted for
achieving a more effective motivation and behaviour change:
 Peer group drinking

3
Wine has more calories than chocolate.
4
Beer has less calories than pizza.

21
Impact of Food Information on Consumers’ Decision Making

 Positive image of alcohol (has more advantages than disadvantages)

1.6 Findings and implications – Food waste

Do consumer misunderstandings of the ‘best before’ label contribute to

avoidable food waste in private households or do consumers throw food
away (consciously) for quality concerns?


The empirical evidence shows that a misunderstanding of the ‘best before’
date as related to food safety contributes to consumers throwing away
outdated food when compared with the other attitudinal and socio-
demographic factors taken into account in the analysis.

A knowledge test was conducted for the two most commonly used expiry labels ‘best before’
and ‘use by’. Following this, participants were asked whether they ever consumed outdated
food and whether they threw away food before the ‘best before’ date.
The knowledge tests revealed a widespread misinterpretation of expiry dates by consumers,
which can be assumed to increase avoidable food waste.
 A majority of 54% of European household shoppers were not able to identify the
correct interpretation of ‘best before’ as a quality related date.
 The most common misunderstanding (37%) confused the ‘best before’ date with the
safety related ‘use by’ date.
Consumers, who believed that the ‘best before’ label represents a ‘safe to eat/drink’ limit were
significantly more likely to claim not to consume outdated food.
Consumers who said that they were generally cautious in their way of life were also more
likely to watch out for any indicators (like any expiry date) that may affect their safety and
well-being.
The habit of checking the expiry date – when shopping for products such as crisps or muesli
– suggests that this information is highly relevant to these consumers. Therefore it is not
surprising that these people are less likely to consume outdated food.
The most relevant factors driving food waste before its ‘best before’ date were predominantly
socio-demographical and motivational variables and only to a lesser extent awareness of the
expiry date.
Further empirical evidence is required to
 Explore further reasons why consumers throw away food before and/or after the expiry
date;
 Better understand strong country effects, which might be due to varying social norms,
differences in climate or storage equipment, cultural cooking or consumption habits,
specific food categories, etc.
 Test whether any information and/or education is effective enough to change at least
the intention to reduce food waste.

22
Impact of Food Information on Consumers’ Decision Making

2 INTRODUCTION

2.1 Background

Food labelling helps consumers to get information about their food. It is important that this
information is accurate and not confusing so consumers can make informed choices about the
food they buy and consume.
On the other hand, making the right food choices may have never been as complicated as it
is nowadays, not only because of the variety of products on offer but the volume of food
information provided on these products. Scientific research has increased our knowledge
about what is more or less healthy. In order to make healthier or more ecologically aware
food choices, or to take account of individual dietary requirements (e.g. vegan, vegetarian),
more transparency on (individualised) food information is needed. For food suppliers these
changing requirements provide new opportunities and new challenges in developing new
products and marketing them in the right way.

2.1.1 EU legislation on food labelling (FIC regulation)

Ensuring food safety for European consumers is a top priority for EU policy makers whose
ultimate goal is to achieve a high level of protection of consumers’ health and interests by
providing a basis for informed and safe food choices.
The general EU food labelling legislation has been in force since 1978,5 while the existing
nutrition labelling rules were adopted in 1990.6 Since then, the number of products as well as
the volume of food information has increased considerably and the market and consumer
requirements have changed.
To modernise and to streamline the EU legal framework on food nutrition information, the
European Parliament and the Council adopted Regulation (EU) No 1169/2011 on the provision
of ‘Food Information to Consumers’ – the “FIC Regulation”.7
The remit of the FIC Regulation is to serve the interests of the internal market by
 Simplifying the law,
 Ensuring legal certainty and
 Reducing administrative burden, and
 Benefit citizens by requiring clear, comprehensible and legible labelling of foods.

5
The existing EU food labelling provisions are laid down in Directive 2000/13/EC of the European Parliament and of
the Council of 20 March 2000 on the approximation of the laws of the Member States relating to the labelling,
presentation and advertising of foodstuffs. The majority of the provisions provided in that Directive date back to
1978: http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2000:109:0029:0042:EN:PDF
6
Directive 90/496/EEC – nutrition labelling for foodstuffs
7
Regulation (EU) No 1169/2011 of the European Parliament and of the Council of 25 October 2011 on the
provision of food information to consumers:
http://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32011R1169&from=EN

23
Impact of Food Information on Consumers’ Decision Making

The new Regulation brings EU rules on general and nutrition labelling together into a single
legal text to simplify and consolidate existing labelling legislation. This new Regulation will
apply directly in all Member States, replacing current laws after a three year transitional
period. While most requirements apply from 13 December 2014, the nutrition labelling
requirements will become mandatory as of 13 December 2016.
In a number of cases, the FIC Regulation sets general principles, leaving the role of defining
more specific rules to subsequent implementing acts. This behavioural study is part of the
process of gathering robust evidence to support the implementation of the Regulation.
The FIC Regulation improves the existing EU legislation on food and nutrition labelling. It
requires certain nutritional information to be shown, ensures legibility and prohibits any
misleading information. It also covers specific aspects of food information provision including
the following topics that have been covered by this research:
a) Trans fats:
The Commission is asked (FIC – Article 30) to assess the impact of appropriate means
that could help consumers to make healthier food and overall dietary choices, or that
could promote the provision of healthier food options to consumers including, among
others, the provision of information of trans fats to consumers or restrictions on their
use. Hereby it is important to investigate consumers' understanding of information on
trans fat itself and in the context of an overall diet.
b) Precautionary allergen labelling:
The Regulation (FIC – Article 36(3)(a)) requires the Commission to establish rules
concerning the use of voluntary food information on the possible and unintentional
presence in food of substances or products causing allergies or intolerances
(precautionary allergens labelling). There are currently various types of such voluntary
labelling, which may or may not influence consumer choices in practice.
c) Labelling of alcoholic beverages:
In the case of alcoholic drinks, the Regulation (FIC – Article 16) provides derogations
for the inclusion of an ingredients list and a nutrition declaration. There will be a future
Commission report on these derogations. The study will address whether alcoholic
beverages should in future be covered, in particular, by the requirement to provide
the information on the energy value, and the reasons justifying possible exemptions,
taking into account the need to ensure coherence with other relevant Union policies.
d) Health-related messages linked to alcohol consumption:
Alcoholic beverage labels are increasingly used by alcohol producers in the EU, mainly
on a voluntary basis. However, there is no adequate information on the extent of these
practices or on their effectiveness in terms of visibility and information value. The
purpose of this part of the study is to assess effectiveness of existing and alternative
messages in terms of attitude change. Although there is no specific request the FIC
emphasizes to support the Member States in reducing alcohol-related harm (FIC –
recital 40).
e) Date marking and food waste:
Depending on the type of food, consumers will continue to see ‘best before’ and ‘use
by’ dates on pre-packed foods. The latter will be more tightly linked to food safety (FIC
– Article 24). It is assumed that consumers often throw away food unnecessarily due
to confusion about the difference between "use by" and "best before" date. Behavioural
evidence on this issue will allow better defining policy remedies to curb the amount of
food simply wasted because of this type of misinterpretations.

24
Impact of Food Information on Consumers’ Decision Making

2.1.2 The changing role of research in food labelling regulation

In the past, food labelling policy and decision-making involved a consideration of scientific
evidence, the input of industry stakeholders such as food manufacturers and retailers, and
government and non-government stakeholders.
Over time, regulatory authorities have recognized the importance of consumer research as a
critical input to regulatory decision making. Food manufacturing continues to advance, with
more innovative and “new” food products available now than ever before. Consumers, as well
as manufacturers, are driving this food innovation.
Food nowadays is recognised by policy makers to play an important role in improving health
and quality of life, based on evidence that a healthy diet is not only in the interest of
individuals (in terms of obesity and mortality) but also influencing the cost of public health
care and productivity of the working population.
Gaining a better understanding of the impact of food labelling on consumer behaviour has
therefore become an integral part of regulatory decision making. Given the increasing cross-
border flow of food products within the European internal market, the European Commission
has initiated policy activities which aim to reduce the complexity of quality regulation in
Europe.
‘Average’ consumer vs. real consumer
The European legislation has been based on a definition of an ‘average consumer’ which is
consistent with the traditional economic model of consumer behaviour, in which consumers
are defined as reasonably well-informed and observant, and make fully rational decisions
when it comes making food choices.
However, recent work in behavioural economics has shown that real consumers do not
typically behave in that way and that, in real life, decisions are based on various short cuts
(or heuristics) and biases (e.g. as a consequence of habitual choice behaviour). It is useful to
apply Kahneman’s dual process thinking8 when trying to understand the way that consumers
access and process information on food labels.
In summary, the dual process model asserts that human behaviour comprises the interaction
between two systems:
1. Instinctive, intuitive system – wherein we do not process information in a detailed
or even conscious level. We act quickly and spontaneously because of factors such as
time constraints, information overload, etc. which characterise day-to-day life.
Deliberative, reflective system – wherein our behaviour is rational or planned in
that we are consciously aware of the behaviour we are performing and information we
are considering. The ‘deliberative’ brain takes time and thinks about things.
In the context of reading and processing food labels, consumers often (but not always) engage
their ‘intuitive’ brain during the in-store purchase situation – i.e. they glance fairly cursorily
at the food label (if at all), and focus only on one of two key pieces of information. If at all, it
is only later in the home (e.g. when cooking with or consuming the food at the table) that
they engage their ‘deliberative’ brain and read and consider the labelling information in some
detail.

8
Daniel Kahneman 2011: Thinking fast and slow

25
Impact of Food Information on Consumers’ Decision Making

Hence, our approach to investigate consumers’ behaviour when it comes to food label
information combines the differentiation between “instinctive, intuitive system” and the
“deliberative, reflective system” with an empirical “bottom up” design to understand their role
in the typical food purchase context.
Consumer understanding of food labels
There is a key distinction to be made between consumer information and understanding. This
study aims to assess the way in which consumers’ access, interpret and use food information.
It is both the provision of information and the level of consumer understanding which are
under consideration. Consumer policy based on the ‘average consumer’ standard has been
brought into question by behavioural theory which highlights the other aspects that govern
consumer behaviour. Providing information is not enough. There is also the gap between
information provision and information comprehension which underpins the definition of food
product literacy. As Janis Pappalardo points out:9
“Product literacy requires accurate information and a means of evaluating that information.
Obviously, product literacy is thwarted when consumers do not have access to the quality
and type of information necessary to evaluate product choices. Thus, truthful information
is necessary for product literacy. In addition, truthful information alone may not be
sufficient to achieve product literacy, because information is useless if one does not
understand how to use it and transform it into practical knowledge.”
Accurate and clear information is a prerequisite to the concept of the literate consumer. But
the consumer also faces other constraints than just a presumed lack of knowledge. The
information provision needs to be understood within the real-life context of people’s shopping
behaviour (a context which includes income constraints and strongly established habitual
patterns) if it is to be truly effective. In addition, providing accurate and clear information is
only the first step. The next step for consumers is to weigh up the pros and cons of reading
that information. Do they notice the information, do they realise the relevance or importance
of the information, is the benefit of reading this information seen to outweigh the “cost” of
the time and effort it will require to read it?
Similarly the idea of misinformation needs to be viewed in the same way. Misleading
information is not only a matter of fraudulent health claims or other deliberately inaccurate
information. It is a matter of how consumers interpret the information they are given. As
Pappalardo mentions in the same article, “What matters for deception is not simply what
words are uttered or written on paper, but consumer interpretation”.
For this behavioural study, it is not just about assessing the optimal presentation of key food
information. It is about fully understanding how this information is or will be used and
interpreted by consumers within the broader context of the many constraints faced in the real
world.

2.1.3 Key elements of researching a purchase decision

Understanding what shoppers respond to before purchase cannot be explained by a simple

decision tree that loosely refers to assumed shopper priorities. This would oversimplify the

9
« Product Literacy and the Economics of Consumer Protection Policy » (Pappalardo, Journal of Consumer Affairs,
2012)

26
Impact of Food Information on Consumers’ Decision Making

research issue, which would lead to inappropriate methodologies and/or misleading or flawed
results.
In our view it is crucial to disentangle different facets of the purchase decision and avoid
oversimplifying what are complex, multi-level (and sometimes sub-conscious) processes. The
recommended approach is to systematically examine how shoppers plan, search, select and
buy their food.
Let’s look at each of these steps in more detail:

Planning refers to when the decision is made about what to buy rather than whether to buy.
It relates to product rather than category planning. Sometimes shoppers actively plan what
product to buy; sometimes they plan certain aspects of a product. Often, this takes the form
of previous knowledge, i.e. shoppers simply know what they are going to buy without active
planning (habitual shopping).
Searching relates to how the decision is made regarding “what to buy?” Here, shoppers try
to locate an item or an acceptable set of items from which to make a selection. When in
search mode, the focus is on de-selection since shoppers scan many products and filter out
irrelevant products.
Selection refers to why the product choice decision was made. When in selection mode,
shoppers make a choice from available options, sometimes many, sometimes just a few. At
some stage, the selection involves trade-offs based on personal preferences. This may not
always happen in-store because the choice was made beforehand.
Buying completes the purchase process and all other elements of the purchase decision
should be considered in the context of the product bought as a result of the decision process.
Sometimes shoppers will simply act on impulse and here they effectively skip the first three
phases.
Our knowledge of shoppers, based on many years of behavioural research in the consumer
sector and comprising many different techniques (including filming, eye tracking, shopping
trip tracking, shopping simulations, neuroscience, and virtual reality testing) has shown us
that the reality in-store is that shoppers often decide what they want in advance or else they
don’t make an active decision. In fact, more time is spent navigating through the shop than
engaging with products since most products are irrelevant to the shopper.
The following principles – a selection of general insights about the shopping behaviour and
process based on the research we have conducted – are useful for scoping the design of the
experiments.
Principle 1: Clarity at the shelf improves ease of shopping

27
Impact of Food Information on Consumers’ Decision Making

We know that the vast majority of consumers shop under time constraints. Shoppers have a
finite amount of time they allocate to shopping for a particular product. So, how can shopping
be made easier, i.e. more efficient and beneficial for the shopper?
Principle 2: Improve ease of shopping by reflecting the shopper’s search process
When in search mode, interaction tends to be at a category level not at a product level.
Shoppers scan many items with very rapid eye movements, looking for visual cues to help
make sense of the product category allocation and shelf layout, and narrow the options.
So they look at colours, shapes, sizes and other characteristics, resulting in an at least
partially sub-conscious process that filters out irrelevant products. Thus, the primary task is
de-selection, not selection. While the visual process may be rapid and sub-conscious,
shoppers do know what they are looking for (the screening criteria can be anything: a specific
product/brand or a category (generic) “cereal” or a specific consumer need “something quick
for lunch”, etc.).
A shopper scanning 100 products will do so to reject 95 of them. We know from empirical
investigations of the shopping process in different markets by combining eye tracking with
neuroscience measures, that positive emotions are only experienced when the shopper finds
relevant products. Finding is a positive but searching is a negative experience.
Since shoppers focus on de-selection, it is really important to identify the product features
they are looking for and then organise the shelf on that basis. This is the stage where shelf
layout really matters. The first goal should be to reduce searching time so that the shopper
can move on to the selecting mode.
Principle 3: Minimise searching to increase time for browsing, category engagement
and relevant food information
As mentioned earlier, when in “search mode”, shoppers focus on what they want to buy and
filter out anything not perceived to be relevant to that task. The result is that other products
(and point of sale communication measures) become invisible.
For policy makers, the ultimate goal is to improve the purchase decision process so that
consumers can make better informed choices (and ideally healthier, more sustainable choices)
in relation to a product category. This will only be possible if shoppers are moved quickly into
“selection mode”.
As we know from our extensive experience with behavioural shopper research, when
searching time is reduced, rather than leaving the category, shoppers use their time more
productively. Thus they spend more time browsing the different products and interacting with
products and labels, which increases the likelihood of looking at relevant food information.
Principle 4: Focus on organising the shelf for shoppers who have made up their mind
before
This principle relates to the fact that different shoppers have different tasks when shopping.
When shoppers’ minds are made up before they go to the shop, they are there to find not to
browse. Decided shoppers represent the majority and organising the shelf for them will
increase efficiency and thus improve the shopping experience. In addition, ’easier’ shopping
increases the chances of changing the shopper´s mind set, converting them from ‘searchers’
to ‘selectors’. In turn, this will increase the number of shoppers who may consider healthier
alternatives or become open to other choice criteria.
Principle 5: Focus on clear and motivating messages for shoppers who have not
made up their mind yet

28
Impact of Food Information on Consumers’ Decision Making

The fact that different shoppers have different tasks at the shelf has implications for how to
communicate to shoppers in store. In order to make a better connection in store, it is
recommended to clearly target the messages.
We know that typically point-of-sale material and packs are looked at only very briefly and
that most information is discarded so it is vital for messages to be relevant and motivating.
But the question is – relevant and motivating for whom?
Messages, particularly those designed to prompt action, will be much more effective if they
target the minority of shoppers who have not made up their mind what to buy and are open
to influence in store. If one can identify what motivates these shoppers, this will help
improving the presentation of food information leading to better informed purchase decisions
in store.
So what do these principles mean for the approach to this research?
Firstly, our key focus for investigating the access, comprehension and evaluation of food
information, is the “selection” phase, the main aspect of the purchase process which will be
actively influenced by policy measures. A common problem with purchase decision research
has been the use of inappropriate attributes. Separating search from selection overcomes this
by allowing us to create specific attributes which relate to the research goals at hand.
 Search attributes should be capable of being used for shelf organisation. This means
that it must be possible to group products on the basis of that attribute and that the
groups will be recognisable to shoppers on the shelf.
 Selection attributes should reflect consumer needs/motivations which are known to
drive product/brand choice in the category.
Also, we take account of the shoppers’ task when shopping the category. Decided shoppers
come to the shelf to search and their needs are very different from those who are open and
in selection mode.
Secondly, our emphasis on shopper behaviour underlines the importance of undertaking
purchase decision research also in store where we are dealing with a real field experiment
with real stakes.
Changes to packaging or point of sale layout often confuse habitual shoppers, and new ideas
that work in a lab environment often don´t work in a busy, real-life store. Testing new
initiatives in a realistic retail context before roll-out is therefore critical in order to select the
ideas that work best on shelf, gain buy-in from retail partners and avoid costly mistakes.

2.2 Research objectives

Never before has so much information on food and healthy diets been provided to consumers.
The wealth of health claims, quality labels, nutrition facts, apps and advice, together with
sometimes misleading or contradictory marketing information, is overwhelming for many
citizens. Moreover, it represents a contextual choice architecture that contributes to consumer
confusion across all education levels.
Although the situation varies across the EU, mainly because of differences between existing
national regulatory policies, the convergent theme across Europe in this area is characterised
by an increasing quantity of information and labels shown on food products. At the same time,
there are still issues not yet fully understood relating to the quality and presentation of
information in a “consumer friendly” way.

29
Impact of Food Information on Consumers’ Decision Making

Consumers develop different strategies to cope with this overload of information. These can
range from ignoring any information or delegating responsibility for decisions to family
members or brands (i.e. always buying the same products), or spending too much time and
thought on the right choice. Between these extremes, there are several ‘behavioural
segments’ of consumers who do not ignore but drastically simplify information intake based
on heuristic shortcuts.
As a consequence of this increasing consumer confusion, the food industry is observing a
general decrease of trust in brands and higher product dissatisfaction levels as well as
reducing product and brand loyalty. 10
On the one hand most stakeholders agree that there is a need for more clarity and
understanding regarding food information. On the other hand there seems to be little
consensus on solutions, despite the publication of numerous studies testing different formats
and content of food information with varying outcomes.
Therefore, it is critical for the success of this study to learn from existing evidence and
establish a robust methodology that takes into account all issues relating to the following key
outcome objectives:
c) Status quo description of consumer understanding and confusion, i.e. explaining the
capacity to access, interpret and use existing food product information specifically on
allergens, trans fats, origin of primary ingredients, durability, health warnings and
labelling of alcoholic beverages as well as further quality label information and thereby:
 Consider individual consumer features and identify relevant biases and constraints
such as cognitive abilities, time resources, repetitive purchases, etc.
Understand external determinants and different framings of the choice situation
contributing to “bad consumer choices” by fostering misunderstandings, information
overload or ignorance
d) Identifying the optimal aspects of presentation of food information for the consumer to
make better (informed) decisions and thereby:
 Consider a holistic approach, i.e. assess the interactive effect of one particular label
on one particular kind of product in the context of increasing complexity of labelling in
the marketplace
 Look for the best compromise between exhaustiveness and clarity of the label,
exploring the possible use of symbols or other means instead of words.

2.3 Research design overview

The objectives outlined above are best addressed by applying behavioural experiments that
are complemented by classical survey design. As described in Chapter 2.1 food choices can
be habitual, impulsive (less conscious), considered (more conscious) or a mix of both. While
a classical survey design is useful for collecting relevant cognitive information (e.g. on
previous experiences and shopping goals), more indirect tools (e.g. observation) are required
to capture less conscious elements of the choice behaviour.

10
Factbook on consumer confusion 2012 by TNS Infratest (http://nestle-zukunftsforum.de)

30
Impact of Food Information on Consumers’ Decision Making

Behavioural experiments also were the requested design by the European Commission, who
has started to test consumer understanding and policy alternatives using behavioural
experiments since 2012. As explained in a recently published Joint-Research-Centre report,
this methodology can help policy makers improve their policy with evidence going beyond
stated preferences as well as the often unrealistic assumption of rational consumer
behaviour.11
The timing of this research and the early stage at which the study took place in the policy
cycle put more weight on the objective of ‘understanding the EU consumer’ than on the
‘maximisation of policy effectiveness’. This allowed the implementation of a gradual three-
stage research design investigating several policy areas:
Figure 2 Research design overview

Preparatory stage Online ‘Laboratory’ Offline ‘Field & Lab’

experiments or store experiments
 Review existing  Experiments covering  Experiments on
evidence and literature several policy areas trans fat only
 Develop hypothesis  8 countries  1 country (Germany)
 Consider relevant  N = 8076 interviews  N = 4841 observations
issues for the design of and observations  N = 1851 interviews
experiments

The preparatory stage comprised a review of existing evidence via desk research combined
with a series of meetings of key stakeholders at the European Commission to scope out the
hypotheses to be tested in the experiments and decided on overall survey design.
Online ‘Laboratory’ (‘lab’) experiments were conducted in eight member states (United
Kingdom, France, Germany, Italy, Spain, Finland, Poland, and Romania) in September 2013.
A questionnaire with online panellists included the simulation of various shopping and
consumption scenarios to collect relevant choice observations on the following policy areas:
trans fats, precautionary allergen warning, health warnings and calorific information on
alcoholic drinks. Further information without experiments was collected on the topics: origin
of primary ingredients and food waste.
Offline ‘Field & Lab’ or store experiments were conducted in super markets in one
member state (Germany) in July 2014. With the consent of a large retailer chain, customers
were observed when shopping for specific food categories (yoghurt, frozen pizza and biscuits)
and subsequently screened for immediate participation in a ‘lab’ experiment at a stand with
mock-up products of the category they just shopped. The focus of this experiment was on
testing the reaction to different levels of transparency of trans fat information.
All experiments are designed as randomised controlled trials. The difference between the ‘lab’
and the ‘field & lab’ design is a somewhat higher realism of the latter experiment due to the
realistic environment of the supermarkets and the sampling of shoppers who had just made
the same type of purchase decision in real life.

11

http://ec.europa.eu/dgs/health_consumer/information_sources/docs/30092013_jrc_scientific_policy_report_en.pdf

31
Impact of Food Information on Consumers’ Decision Making

As soon as it was decided to continue the next research step with the policy area on trans fat,
the initially intended ‘pure field’ experiment design had to be changed into the so-called ‘field
& lab’ design for several reasons:
 In a real field experiment all relevant products (real stakes) on the shelves had to be
manipulated with specific trans fat information. This would have required an arbitrary
or random allocation of high and low levels of trans fat to certain brands without
knowing whether this reflects reality or not.
 Since most shoppers were expected to be habitual buyers, the laboratory-like
experiments at the stand allowed us to switch their habitual shopping mood into a first
buyer mood (assuming that the usual product favourites are out of stock).
 The choice task between mock-up products allowed for a tighter control of influencing
variables and a more robust analysis by excluding brand, price, flavour, and package
format variations.

32
Impact of Food Information on Consumers’ Decision Making

3 EXPERIMENTAL ‘LAB’ DESIGN

3.1 Overview

The experimental design and the questionnaire were developed by TNS in co-operation with
DG SANCO. It consists of three main modules:
 Pre-experimental questions, to screen and segment the sample in order to route the
respondents to the appropriate experiment as well as questions aimed at collecting
information before the experiment and independently from this, to avoid the latter
influencing the replies (e.g. on habits, shopping goals and awareness of different fat
types and allergy information).
 The core part of the online study were the behavioural choice experiments designed
as randomised controlled trials to observe consumers’ preferences on food and
alcoholic beverages.
 Post-experimental questions captured further relevant issues including a knowledge
test on expiry dates, food waste habits, self-control, risk attitudes and socio-
demographics.
As the survey had to cover distinct sets of experiments for various topics, the core
questionnaire was split into two topics:
 One half of the respondents were presented with food shopping experiments to identify
drivers and barriers of a) healthy choices with selected TFA scenarios and b) cautious
choices with selected precautionary allergen warnings.
 The other respondents were submitted questions on decisions about their alcohol
consumption covering calorific and warning treatments for beer, wine and spirits.
The design of the experiments took into account typical shopping and consumption habits and
preferences, the specific context that consumers encounter when making decisions on what
to shop or what to drink in real life.
Some of these previous experiences and habits were relevant to direct respondents into the
appropriate experimental group in order to present an individually relevant decision scenario.
In addition, various treatments were developed to fit into the typical information and decision
process of a shopping scenario or a specific drinking occasion.

33
Impact of Food Information on Consumers’ Decision Making

Figure 3 Questionnaire flow for respondents

We took several decisions in the set up phase to finalise the experimental design. To maximise
the validity, realistic choice situations had to be replicated as closely as possible, while on the
other hand a reduction of real world complexity had to be achieved for analytical reasons but
also to avoid overburdening of respondents. The experimental design therefore focused on
key variables to reduce real world complexity and to avoid cognitive overburdening of
respondents as well as allow meaningful statistical analyses.
This chapter outlines the key challenges and issues which were considered for the final design
of the experimental core part of the questionnaire.

3.2 Respondent segmentation

Universe definition
The overall sample was screened for consumers who are responsible for making decisions on
everyday shopping for their household. Therefore, participants who did not at least share the
shopping responsibility with someone else in their household were screened out.
While the sample definition of household shoppers was appropriate for the shopping
experiments, it was not a logical pre-selection for alcohol consumers.
The reason why a shopper sample screening procedure was used as a preliminary base for all
experiments is that there was no reliable information available on the socio-demographic
structure of ‘alcohol consumers’ across all relevant countries.

34
Impact of Food Information on Consumers’ Decision Making

Therefore, the same more generic wording of the screener question was used as in the
Eurobarometer, which provided a frame (age and gender structure) which could be replicated
in the online survey to capture ‘the average shopper of a household’.
This sample base was further screened for consumers of specific types of alcoholic beverages.
This sample excludes consumers of alcoholic drinks who do not (or only rarely) shop for their
households.
Food vs. Alcohol experiments
Running all experiments for each respondent was not advised as the preceding experiments
will always have an effect on subsequent experiments (conditioning effect). Moreover, this
would have meant that the online questionnaire length would have exceeded 20 minutes, with
the potential consequence of “respondent fatigue”.
The first segmentation of respondents was to assign them to either the ‘food shopping’ or
‘alcohol drinking’ experiments.
The assignment of each respondent to one of these split groups was guided by the individual
relevance of the topics as captured by some questions at the beginning of the survey.
 Qualification for the food experiment was defined as when participants:
o Were either personally affected by food allergy / intolerance or when members
in their household were affected.
o Had shopped for crisps within the last 6 months.
 Qualification for the alcohol experiment was defined as when participants
o had consumed beer, wine, vodka or whiskey within the past twelve months.
Respondents who fulfilled both criteria were allocated to the split with the lowest number of
respondents in order to achieve an equal sample split between food and drinks experiments
within each country.
If someone did not qualify for either topic he or she was assigned to the food experiments,
which was then introduced as a hypothetical scenario to shop and select a product for a
friend12.
This design enabled presenting choice tasks to respondents that they are more or less familiar
with. For instance, the food split offered shopping decisions between different packs of crisps
and muesli to respondents with recent shopping experience of these categories, and the
alcohol split was also based on their own consumption.
Further splits and routings of respondents were applied within the experiments to develop a
framing of the choice tasks that took into account individual habits and experiences.
In the alcohol split, participants were routed to an experiment with an alcoholic drink, they
were familiar with: beer, wine, vodka or whiskey (spirits13). Respondents who said that they

12
The hypothetical scenario was applied to only 3% of the total sample, i.e. 233 respondents. The vast
majority qualified for both experiments (89%, 6337 respondents).
13
The goal was to cover the three most common alcohol categories across Europe: beer, wine and spirits. As
the latter category was estimated to be significantly smaller, vodka or whiskey was presented to the
respondents. In the analysis, the focus is on the overall category of spirits and not on the differences
between vodka and whiskey.

35
Impact of Food Information on Consumers’ Decision Making

drank alcohol from multiple categories were assigned to the type with the lowest number of
completed experiments to achieve an equal sample split across all three main alcohol types.
The following chapters outline the design rationale for each of the three experiments
conducted in this survey, i.e. experiments on
 TFA – trans fatty acids information in a food shopping scenario
 PAW - precautionary allergen warnings in a food shopping scenario
 Alcohol consumption scenario with calorific information or alcohol warning

3.3 Experiments on TFA information

The research objective was to investigate consumers’ understanding of information on TFA

itself as well as in the context of an overall diet. A core question was to explore whether
consumers were able to identify which combination of fat types (e.g., with versus without
TFA, TFA versus saturates) was the healthier option.
At present consumers in Europe are typically not informed about the existence of TFA in a
specific product, neither in the ingredients list nor on the nutrition fact label. For this reason,
an experiment based on the current information status did not make sense even not for the
control group.
However, current legislation requires that from the end of 2014 onwards, the nature of
hydrogenation of vegetable fat, i.e. whether it is fully or partially hydrogenated, needs to be
indicated on the ingredients list. This is intended to help consumers identify whether the
product may contain TFA or not. Only partially hydrogenated oil (PHO) can contain TFA and
the higher the ranking of this ingredient, the higher the amount of TFA could be.
The information on whether partially or fully hydrogenated oils are contained in the product
is the minimum level of information that will be shown in Europe from 2014 onwards and
therefore served as the control group in the experiment.
With this background, the specific challenge of this choice experiment will be something new
for all participants when shopping for food. Although it is likely that some consumers have
already heard of trans fats or partially hydrogenated fats, the current14 food information does
not allow comparisons at the point of sale.

3.3.1 Product category for TFA

The product category selected for this experiment was crisps. This choice was based on the
following criteria:
 Relevance to the policy area: Natural TFA can be found in fat-containing dairy products
like cheese, yoghurt, butter, as well as ruminant-based meats (beef, sheep, and goat).
Industrial TFA can be found in prepacked cakes, croissants, muffins, chips, crisps,
popcorn, chocolate products, fried potato products, margarines, pastries, etc. The
latter products can be found on the market with high differences in industrial TFA
contents, i.e., from being free of TFA to having 50% or more TFA in the total fat of the

14
The lab experiments were designed and delivered in 2013.

36
Impact of Food Information on Consumers’ Decision Making

product. These products were therefore more suitable for testing simple as well as
complex fat type combinations.
 Relevance to consumers: The selected category is popular across all relevant EU
member states, i.e. with only few non-buyers within the overall sample.
 Relevance of food information for consumers: Categories that attract more selective
buying behaviour (i.e. that are less prone to habitual buying or impulse buying) will
increase the ecological validity of an experiment that is based on evaluating differences
between food information. The most selective buying behaviour is typically observed
with baby food with a very high proportion of buyers reading food information in the
store. However this category would only be relevant to a very small share of
consumers. Chocolate bars represent the other extreme - a typical impulse product
that does not attract any reading of food information and was also therefore not
suitable.
Crisps are relevant to the policy area as they may contain industrial TFA. They are commonly
purchased by the majority of consumers across Europe. There was no data available to
evaluate how many consumers actually read food information when shopping for crisps,
although previous experience with observed shopping research allowed the exclusion of
certain categories known to attract low attention to food information, such as spreadable fat
and chocolate bars.

3.3.2 Framing of TFA experiments

For all participants directed to the food experiments split, the TFA scenario was the first
experiment. Hence, there was a two-step introduction:
 First step – a general introduction relevant to both topics, that explained the general
purpose and design of the choice tasks as well as the possibility to earn an additional
monetary bonus
 Second step – a more specific introduction to the TFA choice tasks with individual
reminders of relevant habits and attitudes that had been mentioned by the respondent
earlier in the survey (see figure below)

37
Impact of Food Information on Consumers’ Decision Making

Figure 4 Example of an individualised introduction for TFA experiments

Based on questions asked at the beginning of the questionnaire, the awareness and evaluation
of different types of fat was collected and brought forward again only to those who said they
were aware of it and had a healthy or an unhealthy perception of each one. If the type of fat
was never heard of or if the respondent was unsure whether it is healthy or unhealthy, there
was no specific comment at this stage.
The purpose of the framing with these reminders is to help respondents imagine a more
realistic and individual choice situation, which eventually should lead to more realistic
observations of choice behaviour in the experiment.

3.3.3 Treatment splits for TFA

The next dimension to be considered was the design of treatment stimuli and the number of
treatment splits to be tested. To examine the impact of possible treatments, the following
stimuli were developed to test their effectiveness on driving healthy choices:
 Additional TFA information on nutrition label
 Education intervention explaining all fat types
Additional TFA information on nutrition label
Since consumers will be confronted with new terminology (partially and fully hydrogenated
oil – PHO/FHO) in the ingredients list on many food products from 2014, it is one goal to find

38
Impact of Food Information on Consumers’ Decision Making

out whether this term is understood properly. If so, does it foster healthy choices although it
is not providing exact information on the presence and amount of TFA?
The alternative scenario is to provide concrete TFA information as it is common in other
countries outside the EU (see example below).

Figure 5 Example without TFA and example with TFA

The key question is, whether the additional and more concrete TFA information is actually
helping consumers to make healthier choices or whether it is misleading in certain scenarios.
Education intervention explaining all fat types
It was expected that a large share of respondents would have only little or no awareness of
specific fat types such as TFA and PHO. Thus an additional treatment was developed to
simulate an educational intervention for half of the participants. This was intended to help
understand whether an objective explanation of a complex topic to consumers would increase
the likelihood to make healthy choices.
While the treatment with TFA information was embedded into each choice task, the education
treatment was presented only once before the first task (see figure below).

39
Impact of Food Information on Consumers’ Decision Making

Figure 6 Education intervention about fat types

Four condition splits were developed to measure the impact of each of the two treatments
separately, as well as in combination. Split 4 serves as a control group without these
treatments although it will reflect only the reality from 2014 onwards.

40
Impact of Food Information on Consumers’ Decision Making

Table 7 TFA condition splits

N TFA information Education

Split 1 1 009 Yes Yes

Split 2 1 011 Yes No

Split 3 1 009 No Yes

Split 4 1 007 No No

This design allows a robust sample base for each split with approximately 1000 respondents
that are randomly assigned to each split.

3.3.4 Choice tasks and choice characteristics for TFA

All respondents received 2 subsequent choice tasks:

1. The first choice required a somewhat less complex knowledge regarding TFA, i.e. a
choice between a product with or without TFA/PHO. Everyone who is aware that TFA
is an unhealthy ingredient would be able to identify the healthier alternative (see
Figure 6).
2. The second choice required the knowledge that a product with only small amounts of
TFA/PHO is actually healthier than a product without TFA/PHO but substantially higher
amounts of saturated fat (see Figure 7).
In both choice tasks 1 and 2, the first choice offered a healthier product A (without TFA/PHO)
compared to product B (with 18g TFA).

41
Impact of Food Information on Consumers’ Decision Making

Figure 8 TFA choice Task 1 – example with TFA information and zoomed labels

As well the options of product A or B, participants could also choose the option ‘no difference
to me’. This additional option in principle could be selected by participants who either did not
notice the difference, did not understand the difference, or for whom the difference was not
relevant. Instead of offering such a detailed set of options, which would increase the likelihood
of a “non-choice”, instead respondents were asked after the experiment to explain the
rationale behind their choices.
To increase the realism of this exercise, the food information presented on the crisps pack
was hardly readable on a typical computer screen – except for the titles. The participant had
to actively select and zoom into the labels to read the fine print. This process was intended
to simulate the activity in an offline supermarket of taking packages from the shelf and
comparing food information.
Nobody was forced to read. The introduction on the previous screen suggested that consumers
might read before selection as in real life. However, unlike real life, all other variables were
controlled (no price, no brand, same packaging, etc.) and it can therefore be assumed that
the zooming behaviour measured in this lab experiment is certainly higher than label reading
in reality.
Whether respondents “zoomed” during the experiment or not was monitored in addition to
their selection. Overall participants could zoom on three areas: on ingredients, nutrition
information and/or origin of primary ingredient.

42
Impact of Food Information on Consumers’ Decision Making

The second choice offered a healthier product A with 2g TFA (lower ranked PHO) compared
to product B without TFA/PHO but 18g Saturates/FHO.

Figure 9 TFA choice Task 2 – example with TFA information and zoomed labels

After making these two choices, participants were asked to post-rationalise why they made
these choices, in order to assess whether they had understood the complexity of the tasks.
Finally, a small knowledge test was conducted as an additional measure of correct recall and
awareness of TFA information. Five statements about fat types (three true and two false
statements) were presented to all participants. Three statements had already been explained
in the education intervention and were presented to check whether they recalled the
information. Two statements (about which products can contain TFA) had not been explained
previously.

3.4 Experiments on PAW

Food allergy is an increasing public health problem linked to a variety of policy issues. In this
consumer study the aspect of precautionary allergen warnings (PAW) for the unintentional
presence of allergens in food was the key focus.
Food manufacturers can add an advisory ‘May contain’ labelling on a voluntary basis on their
products to pre-empt any liability claims for the unlikely event that an unintended cross-
contamination causes any health problems. These warnings can be found on a high proportion
of pre-packed food, even where such cross-contamination is unlikely.

43
Impact of Food Information on Consumers’ Decision Making

For some consumers with a severe food allergy it is sometimes not enough to check only the
list of ingredients. In particular, processed and pre-packed food can be subject to unintended
cross-contamination with allergens during the production and packaging process, for instance
if the allergen was an ingredient in another product that was processed before on the same
production line.
Since there is no clear rule about whether or when a PAW should be issued on a product or
not, the existence of a PAW actually does not indicate a more reliable choice compared with
a product without PAW.
Hence, this study set out to examine whether and how PAWs influence consumers’ choice.
Which PAW messages are perceived as most risky and are more likely to limit consumer
choice, perhaps unnecessarily? How do consumers react when offered one product with and
one without PAW? Do they prefer a cautious approach?
Of course, such a decision process is mainly relevant for consumers who are personally
affected by a food allergy, which represent only a small proportion of the population in Europe.
As this topic represents only one policy area among others in this research, the overall sample
of household shoppers was not boosted with allergy sufferers15.
Instead all respondents were asked about their awareness and experience with food allergies
and food intolerances (either because they were personally affected or through shopping for
an affected household member). All 4036 participants in the food split went through the PAW
experiment, as the framing of the shopping scenario had to be designed homogeneously as a
shopping task to be conducted ‘for a friend with a severe nuts allergy’.
This way, the analysis of this experiment would allow comparing the choice behaviour of
participants for whom the topic was relevant (n = 596; 15%) and those for whom it was not
relevant.
An additional aspect was considered in the design of the PAW experiment: information on the
country of origin of the primary ingredient of the product. Does the origin (local vs. imported
ingredient) affect how people assess the level of risk of a PAW? Qualitative research in UK
has indicated that PAWs on local products are often considered as safer than imported
products without a warning16 and this hypothesis was to be tested as part of the experiment.

3.4.1 Allergen and product category for PAW

There is a wide array of allergens that the experiment could have focused on. The choice of
the nut allergen was driven by the following considerations, and the assumption that a PAW
for nuts is likely to cause a higher level of consumer detriment than for other allergens17:
 Nuts are a more common allergy type and are typically related to more severe
consequences for an allergy sufferer.

15
A sample booster would have required identifying individuals with a specific food allergy such as nuts to
make it personally relevant and less hypothetical. This would have required a significantly larger sample size
and would have caused unintended bias in the results of the other food experiment.
16
Barnett J, Vasileiou K, Gowland MH, Raats MM, Lucas JS (2013): Beyond Labelling: What Strategies Do Nut
Allergic Individuals Employ to Make Food Choices? A Qualitative Study. PLoS ONE 8(1)
17
Ditto Barnett et. al. UK 2013

44
Impact of Food Information on Consumers’ Decision Making

 Due to the more severe allergic reactions (sometimes even fatal), the industry uses
PAWs more often in relation to nuts than other allergens for liability reasons.
 Unintended cross-contamination with nuts is more likely in a wide range of product
categories such as bakery, chocolate, cereals, desserts, processed or ready-made
meals. Allergic consumers consider these categories as particularly problematic as they
are strongly associated with nuts, and are more likely to require closer examination.
The choice of a specific allergen type was followed by a suitable product category. For this
experiment muesli was selected based on the following criteria:
 Relevance to the policy area: PAW for nuts are common among prepacked cakes,
chocolate, cereals, deserts, ready-made meals such as pizza.
 Relevance to consumers: The selected category should be popular across all relevant
EU member states, i.e. with only few non-buyers within the overall sample.
 Relevance of food information for consumers: Categories that attract more selective
buying behaviour (i.e. that are less prone to habitual buying) will increase the
ecological validity of an experiment that is based on evaluating differences between
food information.
 Suitability for field experiments: If the PAW experiment is to be repeated for
subsequent testing in real supermarkets, this requires a category with limited
brands/products. Moreover, the typical packaging for the products should allow for the
application of unobtrusive additional labelling.
Muesli is relevant to the policy area as it often contains nuts intentionally but may also contain
nuts as an unintended ingredient. Muesli is purchased by many consumers across Europe.
One can assume that allergy sufferers are less susceptible to impulse buying and more likely
to look for relevant food information than others when buying a product for the first time.
Finally, muesli was considered as a better fit for the shopping scenario as it is a healthier
category than chocolate.

3.4.2 Framing of PAW experiments

The introduction for the PAW experiments was presented to all 4036 participants of the food
split after completing the section on TFA. There was an homogeneous framing of the shopping
goal for everyone by asking them to select a muesli pack for a friend with a severe nuts
allergy. This standard opening was enhanced with individual reminders of habits when
shopping this category.

45
Impact of Food Information on Consumers’ Decision Making

Figure 10 Example for individualised introduction for PAW experiments

The purpose of the framing with these reminders is to help respondents more easily imagine
a more familiar choice situation, thereby allowing more realistic choice behaviour in the
experiment.

3.4.3 Treatment splits for PAW

The next dimension to be considered was the design of treatment stimuli (alternative PAW
wordings) and the number of treatment splits to be tested. A maximum of four sample splits
was recommended to allow a robust analysis when comparing alternative treatments across
all participants. Based on previous desk research the following messages were selected as
the most common wordings with a sufficient differentiation in style and detail:
 May contain nuts
 May contain traces of nuts
 Cannot guarantee nut free
 Made in a facility that also processes nuts

These advisory allergen warnings were displayed next to the ingredients list, where it is
usually placed.
Each respondent was assigned randomly to one of the four treatment splits.

46
Impact of Food Information on Consumers’ Decision Making

Table 11 PAW treatment splits

N N
PAW version
(all) (affected)

Split 1 1 011 149 May contain nuts

Split 2 1 010 143 May contain traces of nuts

Split 3 1 008 147 Cannot guarantee nut free

Split 4 1 007 157 Made in a facility that also processes nuts

There is no version that represents a standard or current status quo, which could serve as a
natural control group in this design. Therefore, the intention was to use the collected data to
suggest how the different treatments should be best compared.

3.4.4 Choice tasks and choice characteristics for PAW

Three choice tasks were presented in a randomised order (to avoid any order effect) to each
respondent:
 Choice 1 required a decision between one muesli pack with a PAW and local ingredients
and one muesli pack without a PAW and imported ingredients (see figure below).
 Choice 2 required a decision between two local mueslis, with only one of them
displaying a PAW.
 Choice 3 required a decision between two mueslis with imported primary ingredients
with only one of them displaying a PAW.

47
Impact of Food Information on Consumers’ Decision Making

Figure 12 PAW choice Task 1 – example with ‘May contain’ and zoomed labels

As well as the options of product A or B, participants could also choose the option ‘no
difference to me’. This additional option in principle could be selected by participants who
either did not notice the difference, did not understand the difference, or for whom the
difference was not relevant. Instead of offering such a detailed set of options, which would
increase the likelihood of a “non-choice”, instead respondents were asked after the
experiment to explain the rationale behind their choices.
To increase the realism of this exercise, the food information presented on the muesli pack
was hardly readable on a typical computer screen – except for the titles. The participant had
to actively select and zoom into the labels to read the fine print. This process was intended
to simulate the activity in an offline supermarket of taking packages from the shelf and
comparing food information.
Nobody was forced to read. The introduction on the previous screen suggested that consumers
might read before selection as in real life. However, unlike real life, all other variables were
controlled (no price, no brand, same packaging, etc.) and it can therefore be assumed that
the zooming behaviour measured in this lab experiment is certainly higher than label reading
in reality.
Whether respondents “zoomed” during the experiment or not was monitored in addition to
their selection. Overall participants could zoom on three areas: on allergen information,
nutrition information and/or origin of primary ingredient.
These choice scenarios were intended to explore whether the awareness of a specific PAW
creates more cautious consumer decisions by avoiding packages with a PAW and also the

48
Impact of Food Information on Consumers’ Decision Making

hypothesis of whether the origin of ingredients has an impact on the risk assessment of PAWs
can be checked.
After completing all three choices participants were asked to post-rationalise why they had
made these choices. The answers might help to better understand the experimental results.
Moreover, a direct risk assessment of six different allergen warnings was collected, which can
be compared with the experimental results.

3.5 Experiments on alcohol consumption

In contrast to the previously described food shopping experiments, the design of the alcohol
sample split focussed on the consumption of alcoholic drinks rather than purchase. It focused
on two core policy issues:
 Does calorific information for alcoholic beverages influence consumers’ intention to
drink less?
 Do health-related messages linked to alcohol consumption influence consumers’
intention to drink less?
While currently in the EU, typically there is no information to consumers on the calories
included in alcoholic beverages, health-related messages are increasingly used by
manufacturers on a voluntary basis. The aim of the study was to investigate whether the
awareness of this specific type of information on calorific content has any impact on the
consumer intention to drink less.
The specific challenge of this choice experiment was that the design of a consumption scenario
in an online survey is more hypothetical than a shopping scenario. Within the context of a
survey, it is not possible to simulate and observe the decision process beyond the stated
intention to drink less. Moreover, there are of course many influencing factors that are
relevant in a real ‘drink consumption’ situation that cannot be taken into account in a
hypothetical online scenario. Such factors would require a separate study which simulates the
more relevant facets of the decision process in this area.
Within a restricted questionnaire space and sample, the alcohol consumption experiment was
designed to focus on:
 Three different types of alcoholic beverages: beer, wine and spirits (see next Chapter
3.5.1)
 One specific drinking occasion: a party at a friend’s home (see Chapter 3.5.2)
 Two alternative treatments: education/information either on calories or on officially
recommended limits (see Chapter 3.5.3)
Therefore, the experiment covered only the consumers’ assessment phase of the given
treatment information and its immediate impact on the stated intention to drink less in the
specified situation and in the future.

3.5.1 Product categories for alcohol consumption

The product categories selected for this experiment were:

 Beer
 Wine

49
Impact of Food Information on Consumers’ Decision Making

 Spirits (Vodka/Whiskey)
This choice was based on the following criteria:
 Relevance to the policy area: The five main alcoholic beverage categories are beers,
wines, liquors, spirits and flavoured alcoholic drinks (alcopops). In particular the latter
three require further definition of sub-categories for a realistic framing, e.g. for
flavoured alcoholic drinks (shandy, wine cooler, prepacked cocktails, etc.) for spirits
(vodka, whiskey, gin, tequila, rum, etc.).
 Relevance to consumers: The selected (sub-) categories should be popular across all
relevant EU member states, i.e. with only few non-consumers within the overall
sample.
A maximum of three categories was set to allow a robust sample base and meaningful
statistical analysis. Alcopops and liquors were not covered as they were less frequently
consumed by a sample of household shoppers compared with the selected categories of beer,
wine and spirits. Spirits were specified as vodka or whiskey.
The alcohol experiment was only presented to participants who had drunk at least one type
of alcohol within the past 12 months and who were randomly assigned to this split if they also
qualified for the food questions. One question about general preferences when choosing an
alcoholic drink was asked before assigning a specific type of alcohol. The assignment to the
alcohol category (beer, wine, vodka, whiskey) was conducted at random among all types
consumed with the goal to achieve an equal sample size for the three main categories.

3.5.2 Framing of alcohol experiments

Before introducing the experiment, drinking habits and consumption patterns were collected,
i.e. types of occasions, frequency and amounts consumed.
Collecting data on realistic drinking volumes impose a real challenge to researchers as well
as to participants as there are several individual biases and external barriers expected to
influence the answer behaviour:
 The decision to drink a certain amount at a specific occasion is most often not a
conscious decision which is recalled properly the next day or whenever this question
is recalled. The reason for that is not only the influence of alcohol, but also the fact
that there is often no specific attention given to the volume of drinking.
 As a result, post-rationalisation or over-confidence about one’s level of restraint may
influence the answers given, which cannot be measured by comparison with a more
objective observation in a quasi-laboratory experiment.
 Social norms may withhold respondents from reporting real consumption, thereby
leading to systematic under-reporting of consumed volumes.
 Most multi-country studies on drinking patterns do not specify the serving size of a
drink as this often varies significantly across (and even within) countries. However,
one assumption was that the level of current drinking volume could be a key factor
which influences the willingness of drinking less in the future. Therefore this study
offered an average European serving size per category to the number of glasses when
collecting drinking volumes (see Table below).

50
Impact of Food Information on Consumers’ Decision Making

Table 13 Serving sizes used for collecting drinking volumes

Beer18 Wine Spirits

1 small glass or less 1 small glass or less 1 small shot or less
(up to 0.25l) (up to 0.1l = 100 ml) (up to 0.01l = 1cl)
2 small glasses 2 small glasses 2 small shots
(up to 0.5l) (up to 0.2l = 200 ml) (up to 0.02l = 2 cl)
3-4 small glasses 3-4 small glasses 3-4 small shots
(up to 1l) (up to 0.4l = 400 ml) (up to 0.04l = 4 cl)
5-6 small glasses 5-6 small glasses 5-6 small shots
(up to 1.5l) (up to 0.6l = 600 ml) (up to 0.06l = 6 cl)
7-8 small glasses 7-8 small glasses 7-8 small shots
(up to 2l) (up to 1l) (up to 0.1l = 10 cl)
More than 8 small glasses More than 8 small glasses More than 8 small shots
(2l) (1l) (0.1l = 10 cl)

The introduction to the decision scenarios was designed in two steps:

 First step – a general introduction valid for all alcohol types explained the general
purpose and design of the experiment as well as the possibility to earn an additional
bonus
 Second step – a more specific framing of a given drinking occasion (see Figure below)
Figure 14 Introduction for alcohol experiments – example for beer

18
In UK the serving size for beer was translated into pints.

51
Impact of Food Information on Consumers’ Decision Making

The framing of a given situation was necessary to control the impact of the variety of all
possible occasions and to be able to focus on the impact on treatments as well as selected
other factors that are less diverse.
The occasion of a party at a friends´ home was selected because it was expected to be a
common occasion, which excluded the relevance of costs and thereby should allow for a high
variance of answer behaviour.

3.5.3 Treatment splits for alcohol

To examine the impact of disclosure of calories and health warnings the following stimuli were
developed to test their effectiveness on the intention to drink less alcohol:
 Education on the amount of calories of a drink
 Education on the officially recommended limits of alcohol to drink per day
Both treatments were designed as an interactive education starting with a knowledge test
with a subsequent disclosure of the correct answer.

Calorific information treatment

Since consumers are typically not informed about the calories contained in alcoholic drinks
the knowledge test on calories was designed as a comparative test with food products that
usually display calorific information. The comparison with a reference product was intended
to make this task easier and more tangible for online respondents. A lower rate of ‘don’t know’
responses was expected as a result of this educational treatment which also included images.
The reference food type was selected on the basis of a high familiarity for consumers of the
respective alcohol type. Moreover, the reference food varied by setting alternate frames in
terms of more, the same or fewer calories. This allowed the testing of the impact of negative
vs. positive framing of the education treatment.
Beer drinkers were asked to compare 0.5 litres of beer with 200 grams of pizza and to decide
which of these contained more calories (see Figure below).

52
Impact of Food Information on Consumers’ Decision Making

Figure 15 Knowledge test on calories for beer

On the subsequent screen respondents were then told whether their answer was correct or
not. In this example pizza has more calories than beer (see Figure below).

53
Impact of Food Information on Consumers’ Decision Making

Figure 16 Education on calories for beer

A similar treatment was applied to

 Wine drinkers: 200 ml of wine has more calories (120-170 kcal) than 20 grams of
chocolate (100 kcal).
 Whiskey/Vodka drinkers: 2 cl of whiskey/vodka has about the same as 10 grams of
mixed nuts (50 kcal).

Limits treatment
There is an increasing variety of health warnings in existence for
alcohol, some of which are driven by mandatory regulation and
others by self-regulation by the industry in some EU member states.
The former tend to be more prescriptive and factual in describing
alcohol content and recommended safe consumption. The latter
voluntary information from the industry tends to comprise a vaguer,
more general advisory ‘enjoy in moderation’ message.
The health warning “Alcohol – Know your limits”19 was selected
because it is relevant to all consumers (in comparison to special
target populations such as pregnant women or car drivers).
It is suitable for all alcohol categories and works in all European
languages.
Moreover, it is concrete enough to conduct a knowledge test suitable for an education
intended to persuade people to drink less.

19
Published by the Public Health Agency in UK: http://www.knowyourlimits.info/ with a similar campaign by
the BZgA in Germany: http://www.kenn-dein-limit.info/

54
Impact of Food Information on Consumers’ Decision Making

Figure 17 Knowledge test on limits for wine

Figure 18 Education on limits for wine

Eight condition splits were developed to measure the impact of each of the two treatments
separately for the three alcohol categories.

55
Impact of Food Information on Consumers’ Decision Making

Table 19 Alcohol condition splits

N Category Education

Split 1 680 Beer Calories

Split 2 671 Beer Limits

Split 3 674 Wine Calories

Split 4 668 Wine Limits

Split 5/7 677 Spirits Calories

(Vodka/Whiskey)

Split 6/8 670 Spirits Limits

(Vodka/Whiskey)

This design allowed a robust sample base for each split with approximately 670 respondents
randomly assigned to each treatment version.

3.5.4 Choice tasks and choice characteristics for alcohol consumption

All respondents received two subsequent choice tasks after the education treatment:
 The first choice required a more short-term decision, whether one would consider a
revised intention to drink a certain amount at the specific occasion.
 The second choice required a more long-term decision, whether one would consider
drinking less, the same or more at any occasion in the future.

Both decision scenarios offered four answer options: drinking less, the same or more with the
additional option of not making a decision.
Both scenarios were framed by a comment to “Be quick and answer within 5 seconds”. The
idea behind this nudge was to receive answers closer to an intuitive reaction than a conscious
deliberate reaction. A quick and intuitively expressed intention is expected to reflect better
the well-known status quo bias of behaviour, while longer considered answers are more likely
to suffer from an over-confidence bias. According to behavioural theory the former should be
closer to an implemented behaviour than the latter.
The time that respondents actually stayed on this screen for reading and answering the choice
tasks was measured with the following results:
 The median duration across all respondents for the first choice was 15 seconds and for
the second choice 12 seconds.
 As it took approximately 10 seconds to read all of the information presented on the
screen in the first task and about 5 seconds for the second task, the majority of
respondents answered within the requested time frame.
 In the first task, 27% of the participants exceeded 20 seconds and in the second task
this was only 11%. However, in an online scenario it is difficult to assess whether these

56
Impact of Food Information on Consumers’ Decision Making

respondents needed more time to read or to decide on the appropriate answer or for
other reasons (e.g. unforeseen interruptions, slow internet connection).

Figure 20 Alcohol choice 1 – example with wine and calories treatment

57
Impact of Food Information on Consumers’ Decision Making

Figure 21 Alcohol choice 1 – example with wine and limits treatment

Figure 22 Alcohol choice 2 – example with wine

58
Impact of Food Information on Consumers’ Decision Making

Both choice tasks differed in the following aspects:

The first decision asked whether a previously stated intention to drink in a specific scenario
(party at a friend’s home) would change as a result of the information presented. This was
supported by an image of a bottle with information on calories or with a health warning on
limits. Moreover, they were reminded of the amount they had previously stated (prior to the
education treatment).
The second decision asked the respondent whether s/he would intend to change his/her
drinking behaviour at a similar occasion in the future. There was no accompanying image or
reminder of their previous statements. The respondent was required to make a decision based
only on what s/he knew by this stage.
These choice tasks allow an investigation of which of these treatments are effective in driving
a short-term attitude change towards a specific drinking occasion. The design also allows an
analysis of whether short term intention influences a more long-term attitude change.
After completing the two decision scenarios, participants faced questions on attitudes towards
drinking as well as information on alcoholic beverages.

59
Impact of Food Information on Consumers’ Decision Making

4 EXPERIMENTAL ‘LAB’ FINDINGS ON TRANS FATS

This section explains the most important findings of the experimental part on TFA to provide
answers for the ultimate research objectives:
b) What are the main individual biases and external barriers that prevent shoppers from
choosing foods with a healthier food composition? How does the disclosure of TFA
contents in the nutrition information and/or consumer education on TFA and fats
influence consumers' ability to identify healthier food choices?
To contextualise the findings of the research and to structure the analysis towards the
ultimate objective of understanding the drivers and barriers of the food choice decision
process we describe below a model of a simplified choice process in a shopping situation.
Such a process comprises three main stages:
 Accessing available information
 Assessing and analysing information
 Acting by taking a choice based on previous steps
While these steps may characterise a more rational and conscious decision making, the survey
design also allowed an exploration of whether there are less conscious elements that shape
the decision process. Thus, a more realistic assessment of the impact of potential treatments
should be possible to identify whether they would have the power to overcome “heuristics”
and habitual behaviours by nudging consumers towards more informed and healthy food
choices.
Of course, any online survey design is restricted in terms of capturing all of the relevant
parameters of a real world design. However, most of the parameters listed in the table below
were included in the study.
The treatment measures – as described earlier – are not easily assigned to one of the three
main stages of a decision process. In fact, each individual stimulus more or less follows the
theoretical order of accessing, analysing and acting. Any communication from education to
persuasion will be more geared towards influencing awareness and beliefs and attitudes for a
more sustainable and conscious behaviour change, which may not be seen immediately but
over time. Changes in the choice architecture are known to have a more direct impact at the
decision stage by overcoming status quo bias, i.e. more unconscious habits. Therefore, the
hypothesis is that steering methods can be expected to show a more immediate effect as
compared to pure information stimuli.

60
Impact of Food Information on Consumers’ Decision Making

Figure 23 Guideline for analysis of TFA results

The subsequent analysis and description of survey findings will start in Chapter 4.1 with a
description of disaggregated findings of the observed decision behaviour in the choice tasks
with suggestions for variables to be included in the statistical base model. The bivariate
examination of potentially influencing variables is more or less structured along the graphical
overview. This is supplemented by a descriptive overview of observed choice behaviour for
the tested treatments.
Chapter 4.2 combines the discussed variables by multivariate analysis to answer the key
questions about what are the statistically significant barriers and motivators for healthy
choices.

4.1 Observed choice behaviour

The first choice task required the identification of the healthier alternative between product A
which contained no PHO or TFA and product B which displayed 18g TFA or a high ranking of
PHO. About one third of all participants identified the healthier product A, while 58% stated
that there was no difference between the 2 products.
The second choice task presented product C with 2g TFA or low ranking PHO in comparison
to product D with no PHO or TFA, but 18g of saturated fats. This choice scenario achieved a

61
Impact of Food Information on Consumers’ Decision Making

significantly higher share of participants identifying the healthier product C (46%) while only
29% said there was no difference.
In both choices, half of the sample received additional information on the TFA amount within
the nutrition facts. The other half of the sample was only able to identify information on
TFA/PHO by zooming into the ingredients list. Similarly, only half of each of those subsamples
received education on TFA and fat, whereas the other half didn't.
Therefore, the observed choice behaviour displayed in the table below is a result of a mix of
treatments as described in Chapter 3.3.3. Separate results for the treatment splits are
presented further below in tables 33 and 34.

Table 24 Observed choice behaviour in TFA Tasks 1 and 2

Food split: TFA Choice Task 1 Choice Task 2

EU8 – Total 0 vs. high TFA/PHO 2g TFA/PHO vs. high SF
Base unweighted 4 036 4 036
Product A/C
31.1% 46.4%
healthier product
No difference
58.4% 29.3%
to me
Product B/D
10.5% 24.3%
less healthy product
Total 100% 100%

The majority of those who identified product A/C as the healthier alternative also recalled
correctly the respective choice rationale (72% in Task 1 and 84% in Task 2). Respondents
who were not able to decide between product A/C and B/D predominantly explained this by
not being able to spot any differences (66% in Task 1 and 53% in Task 2).
The latter result indicates that the simple provision of food information is often not enough to
make consumers aware of TFA/PHO content. This requires a more detailed analysis of barriers
and drivers of awareness, seen in the following chapter.

4.1.1 Potential awareness drivers

One key requirement for making healthier choices in such scenarios is a minimum level of
awareness and the correct evaluation of the various fat types. This data was collected before
the experimental part of the survey with the following results:
 Almost everyone had heard of saturated fat previously and around half correctly
classified it as something unhealthy.
 Compared to that, the general awareness of all of the other fat types is significantly
lower. Around 30% claim to have never heard of them. Amongst those aware of each
type of fat, only around half were able to judge whether it is something healthy or
unhealthy.
 Overall, trans fat seems to have a more unhealthy image than PHO or FHO.
 FHO seems to have a slightly more unhealthy image than PHO

62
Impact of Food Information on Consumers’ Decision Making

Figure 25 Awareness of fat types

Awareness of fat types

I have enough
I heard it, but I know roughly knowledge to
I have never I don´t know what it is/ what take the right
heard of it much about it it means decisions
%

TFA – 30 31 23 15
Trans fat

PHO – Partially 30 35 23 13
hydrogenated oil

FHO – Fully 29 34 22 14
hydrogenated oil

SF –
5 28 40 26
Saturated fat

Q12: Have you ever heard of these types of fat before? Which of the following applies to you regarding …?
Base: EU8 – all respondents (n = 8 076)

Figure 26 Evaluation of fat types

Evaluation of fat types

Very Quite I am Quite Very

healthy healthy not sure unhealthy unhealthy
%

TFA –
1 7 43 21 27
Trans fat
(n = 5 654)

PHO – Partially
1 14 49 26 10
hydrogenated oil
(n = 5 698)

FHO – Fully
hydrogenated oil 2 11 46 22 19
(n = 5 720)

SF –
Saturated fat 3 14 31 29 24
(n = 7 674)

Q13: Based on what you know, how healthy do you think these types of fat are?
Base: EU8 – respondents who heard of the term before (n = between 5 654 and 7 674)

63
Impact of Food Information on Consumers’ Decision Making

When looking at the choice behaviour of consumers with a correct knowledge of the relevant
fat types, (i.e. who knew already that TFA, PHO and Saturates are unhealthy prior to the
experiment), there is the suggestion of a possible interdependency in particular for the second
choice scenario (see figures below).
 Prior awareness of TFA being unhealthy seems to correlate with more healthy choices
in both tasks.
 The knowledge that PHO is unhealthy seems to correlate with healthier choices only in
the second task, but does not show significant differentiation in the first task.
 Saturated fats (SF) were a key differentiator in the second choice task. For this reason
it is not surprising that consumers who were previously aware that saturated fat is
unhealthy identified most often the healthier alternative.

Figure 27 Pre-experimental awareness of fat types vs. Task 1 behaviour

TFA Choice Task 1

Based on previous awareness of fat types

Product B Product A
(18g TFA/ high PHO) No difference (0g TFA/ no PHO)
less healthy to me healthier product

EU8 – Total food split (n = 4 036) 11% 58% 31%

TFA = unhealthy (n = 1 340) 9% 54% 37%

TFA = unknown/not sure (n = 2 436) 10% 61% 28%

TFA = healthy (n = 260) 19% 54% 27%

PHO = unhealthy (n = 997) 12% 54% 34%

PHO = unknown/not sure (n = 2 605) 9% 61% 30%

PHO = healthy (n = 434) 14% 52% 33%

Q271/272: Which product would you choose?

Base: EU8 – all respondents with food experiments and with the respective awareness of the specific fat type

64
Impact of Food Information on Consumers’ Decision Making

Figure 28 Pre-experimental awareness of fat types vs. Task 2 behaviour

TFA Choice Task 2

Based on previous awareness of fat types

Product D Product C
(0g TFA/PHO, 18g SF) No difference (2g TFA/PHO, 2g SF)
less healthy to me healthier product

24% 29% 46%

TFA = unhealthy (n = 1 340) 25% 22% 53%

TFA = unknown/not sure (n = 2 436) 24% 33% 43%

PHO = unhealthy (n = 997) 24% 22% 54%

PHO = unknown/not sure (n = 2 605) 24% 33% 44%

SF = unhealthy (n = 2 006) 21% 22% 57%

SF = unknown/not sure (n = 1 394) 23% 41% 36%

Q281/282: Which product would you choose?

Base: EU8 – all respondents with food experiments and with the respective awareness of the specific fat type

When looking at the previous awareness of fat types, the choice behaviour of the first task
displays a smaller variance compared to the second task.
One possible explanation is that consumers are already familiar with the term saturates on
food labels, while the other terms are new and less familiar in a shopping situation. Even if
someone is aware, they have not been able to compare this characteristic between products
in real life as they are not currently shown on food labels.
Another interesting observation was the zooming behaviour of participants. Participants were
instructed to select and read food information as in real life - therefore nobody was requested
to zoom into any label. However, the difference between the two products was barely readable
without zooming. As a result, a high zooming rate was observed:
 70% of all (food split) participants zoomed into at least one label in the first choice.
 72.5% zoomed into at least one label in the second choice.
The nutrition facts label was accessed most often. This provided key information for all splits
in Task 2, but only for selected splits in Task 1. The ingredients label would have provided all
respondents in all splits with the relevant information. However, very few looked more closely
at this. The origin label was seen by one in four respondents in Task 1, and by one in five
respondents in Task 2. There was no choice-relevant information included in this label.
Overall, such evidence indicates that the information on ingredients seems to be less relevant
to respondents – when comparing crisps – than nutrition facts or information on country of
origin.

65
Impact of Food Information on Consumers’ Decision Making

As zooming on the information was a necessary condition (albeit not sufficient in itself) for
the ability to select the healthier option, we need to control for such behaviour20:
 Respondents who zoomed on none of the labels displayed by far the lowest share of
healthy choices;
 The zooming of nutrition facts correlates most with healthy choices in both tasks;
 Zooming on the information on country of origin irrelevant to health seems to have a
negative influence in the first task and a positive in the second task.21
 Looking more closely at the essential ingredients information seems to have been
more helpful in the second task.

Table 29 Zooming behaviour22 in Task 1

Zooming Zooming
Food split: TFA EU8 – Total nutrition Zooming country of Nothing
Choice Task 1 food split facts ingredients origin zoomed
Base unweighted 4 036 2 568 620 1 034 1 210
Product A –
healthier product 31.1% 38.0% 33.0% 23.5% 18.3%
(0 TFA/ no PHO)
No difference
58.4% 53.5% 30.6% 65.0% 73.3%
to me
Product B 10.5% 8.5% 36.4% 11.5% 8.4%
(18g TFA/ high PHO)

Total 100% 100% 100% 100% 100%

20
In inferential analysis, “controlling for variables” implies studying the variability of one variable keeping the
other variables (the ones that we want to control for) constant.
21
The multivariate analysis will show in Chapter 4.2.2 that zooming on the country of origin label does have a
negative impact on healthy choices in both tasks. This misleading descriptive result can be due to the
correlation of “zooming country of origin” with the age of the respondents: we find that typically younger
consumers check more often for country labels than older consumers.
22
Each respondent was able to zoom into up to 3 different labels or none of them. Hence, the horizontal sum
of the total zooming behaviour is > 100% because those that zoomed in 2 or 3 segments were counted
multiple.

66
Impact of Food Information on Consumers’ Decision Making

Table 30 Zooming behaviour23 in Task 2

Zooming Zooming
Food split: TFA EU8 – Total nutrition Zooming country of Nothing
Choice Task 2 food split facts ingredients origin zoomed
Base unweighted 4 036 2 706 561 869 1 110
Product C –
healthier product 46.4% 56.4% 52.6% 53.1% 22.7%
(2g TFA/PHO, 2g SF)
No difference
29.3% 17.3% 20.4% 26.0% 57.9%
to me
Product D 24.3% 26.3% 27.1% 20.8% 19.4%
(0 TFA/PHO, 18g SF)

Total 100% 100% 100% 100% 100%

4.1.2 Potential beliefs & attitude drivers

Consumer motives and attitudes around food shopping were considered in this analysis, as it
was expected to make a difference on the ability to identify the healthier alternative. Some
questions on shopping engagement and goals were included, which can be looked at
separately or via a cluster segmentation.
The descriptive analysis based on individual statements revealed several motivational
barriers, i.e. fewer healthy choices by consumers who
 do not look for healthier alternatives (no health goal)
 do not take time for browsing and usually shop in a rush (not enough time)
 think price is more important than anything else and who cannot afford to buy what
they want (not enough money)
 do not read labels or pay attention to food information
Each of these individual statements seems to be logical. However, as every respondent has
an individual mix of motives, it can be more insightful looking instead at clusters of shopper
types with similar sets of motives.
Six distinct shopper segments were identified based on a distance analysis24 about the three
key resources that are obviously required for paying attention to food information in a
shopping situation:

23
Each respondent was able to zoom into up to 3 different labels or none of them. Hence, the horizontal sum
of the total zooming behaviour is > 100% because those that zoomed in 2 or 3 segments were counted
multiple.
24
A distance analysis works in the same way as a cluster analysis only with predetermined cluster definitions:
Each respondent was allocated to one of eight target clusters, with the best fit, e.g. enough money but no
time or interest, or high interest in quality/health but no time and money, etc. The original eight clusters
were reduced to six (by combining two clusters) in order to achieve robust sample base for each segment.

67
Impact of Food Information on Consumers’ Decision Making

 Time25,
 Money26,
 Interest in quality and health aspects27.

Figure 31 Shopper types

The Expert Shopper (21% of shoppers) has all the resources: time, money, interest in
health and quality when shopping food. This type is highly experienced and used to paying
attention to all types of food information except price. They are more informed about all fat
types, are least overwhelmed by choice and perform best in identifying the healthier

25
Based on Q9 statement 2 “I usually shop in a rush“ and statement 3 “I usually take time and browse while
shopping to see what is new“
26
Based on Q9 statement 5 “I can only afford to buy what I need rather than what I want“ and Q11 statement
2 “Price is more important than anything else“
27
Based on Q11 statement 1 “I’m never willing to compromise quality for a lower price“ and statement 5 “I
will always choose the healthier alternative when selecting a product“

68
Impact of Food Information on Consumers’ Decision Making

alternatives in the TFA experiments. Hence, this type suits well as a reference group in the
final analytical model.
The Ambitious Health Planner (17% of shoppers) wants to make healthy choices and
invests enough time to read food information but has the second lowest income level and
therefore cannot afford all of the things that s/he wants. This type plans its shopping most
often by making lists, has the highest rate of reading all kinds of food information and displays
the highest risk aversion in general. This segment is represented by a higher share of women,
of older people as well as having the highest average BMI28 amongst respondents. Although
the health planners were less familiar with the terminology of fat types, their performance in
the TFA experiment was second best in identifying healthy options.
The Rushed Quality Shopper (21% of shoppers) has sufficient money (income) and
interest in quality and healthy foods but not enough time for shopping. This type is well
educated, very loyal to brands and displays an above-average frequency of reading food labels
with focus on organic labels. The rushed quality shopper displays an average awareness and
behaviour in the TFA experiments.
The Self-Determined Shopper (14% of shoppers) represents the wealthiest segment
without any health goal. This cluster features rather younger respondents, often in single
households, the highest share of males and less food restrictions. The habit of accessing food
information is well below average. In the TFA experiment, this type was less often able to
identify the healthier alternative – in particular in the more complex Task 2.
The Bargain Hunter (14% of shoppers) has time and enjoys shopping for the best price
deals. This segment features the lowest income and education levels. Since the main focus is
on price, there is no attention to other food information or health issues. In the TFA
experiment this type was less often able to identify the healthy product in Task 1, but scored
averagely in Task 2.
The Frustrated Shopper (13% of shoppers) lacks all relevant resources: time, money
and interest in quality or healthy food choices and is least likely to enjoy the shopping process.
The focus of this segment is on price only, which comes together with the lowest readiness
to read any other food information. This group features the youngest respondents, with the
highest share of single households, with lower education and income levels. In the TFA
experiment, subjects in this group tended to select the less healthy product in both tasks.

28
BMI = Body Mass Index

69
Impact of Food Information on Consumers’ Decision Making

Table 32 Shopper type summary on TFA experiments

Performance in TFA Performance in TFA

SHOPPER TYPE choice Task 1 choice Task 2

Expert Shopper Healthiest (38%) Healthiest (52%)

Ambitious Health Shopper More healthy (34%) Average (46%)

Rushed Quality Shopper Average (29%) Average (47%)

Self-Determined Shopper Average (29%) Less healthy (42%)

Bargain Hunter Less healthy (28%) Average (48%)

Frustrated Shopper Least healthy (26%) Least healthy (40%)

All Participants (Food Split)  = 31%  = 46%

The descriptive analysis above suggests including the following motivational variables into the
final analytical model:
 Shopper types (covering health goals as well as the relevance of time and prices)
 (Dis-)Agreement with the statement “I read labels and pay close attention to product
information”

4.1.3 Experimental treatments

The tested treatments were also included in the final regression model. However, it is clearer
for non-statisticians to see how the choice behaviour differs when comparing different
treatment splits.
One key result in the table below is that the vast majority of consumers were not able to
make an informed decision in the first scenario without any additional treatment. As the
results of the control group reveal (split 4) about 64.4% (56.7%+7.7%) were either not able
to spot or to understand the difference between product A and B. Moreover, this choice
behaviour does not change at all through only an educational intervention (split 3). It should
be noted that in splits 3 and 4 only accessing and understanding the information in the
ingredients list would have enabled an informed choice. The low proportion of participants
accessing the ingredients list and the low knowledge on PHO/FHO discussed above may
explain that only very few participants may have succeeded with this task.
Only the additional information on concrete amounts of TFA on the nutrition fact label (as in
splits 1 and 2) seemed to provide significantly more clarity to consumers in choice Task 1.
However, the combination of both ‘education and TFA info’ (split 1) was associated with the
most healthy choices in the first scenario according to this bivariate analysis.

70
Impact of Food Information on Consumers’ Decision Making

Table 33 Treatments in Task 129

Split 1: Split 2: Split 3: Split 4:

Food split: TFA EU8 – Total with edu – no edu – with edu – no edu –
Choice Task 1 food split with TF info with TF info no TF info no TF info
Base unweighted 4 034 1 008 1 011 1 009 1 007
Product A –
healthier product 31.1% 61.6% 47.4% 7.7% 7.8%
(0 TFA/ no PHO)
*Not able to spot
38.3% 13.9% 23.5% 59.4% 56.7%
any differences
*Differences were
13.6% 11.2% 14.5% 13.3% 15.6%
not relevant to me
*Did not understand
6.3% 4.2% 5.5% 7.9% 7.7%
the difference
Product B 10.5% 9.1% 9.1% 11.8% 12.0%
(18g TFA/ high PHO)

Total 100% 100% 100% 100% 100%

In Task 2 the control group scenario (split 4) seemed to be easier compared to Task 1, as
49% of consumers were able to correctly identify the healthier product C without any
education or additional TFA information.
In this case, the most comprehensive treatment (split 1) was the least effective compared to
being the most effective in Task 1. The additional TFA information misled many consumers to
focus on TFA and neglect the high amount of saturates in product D. The highest share of
healthier choices can be observed in split 3, with the education as the sole treatment, which
seemed to improve, at least slightly, the choices made compared to the control group.

29
The table shows a combined bivariate analysis of the choice question and the post rationalisation for the
initial answer option “No difference to me”.

71
Impact of Food Information on Consumers’ Decision Making

Table 34 Treatments in Task 2

Split 1: Split 2: Split 3: Split 4:

Food split: TFA EU8 – Total with edu – no edu – with edu – no edu –
Choice Task 2 food split with TF info with TF info no TF info no TF info
Base unweighted 4 036 1 009 1 011 1 009 1 007
Product C –
healthier product 46.4% 41.0% 42.4% 53.1% 49.0%
(2g TFA/PHO, 2g SF)
*Not able to spot
15.4% 9.3% 14.1% 17.3% 21.1%
any differences
*Differences were
10.8% 11.4% 11.4% 9.2% 11.2%
not relevant to me
*Did not understand
3.1% 2.8% 3.3% 2.8% 3.4%
the difference
Product D 24.3% 35.6% 28.8% 17.7% 15.3%
(0 TFA/PHO, 18g SF)

Total 100% 100% 100% 100% 100%

In summary, the descriptive analysis of the choice behaviour between the different fat type
scenarios shows that the treatments impact on healthy choices in the following ways:
 TFA information on nutrition fact label
o Highly positive impact in scenario 1 (with or without TFA)
o Negative impact in scenario 2 (small TFA difference and high SF difference)
 Education stimulus
o Positive impact in scenario 1 only in combination with TFA information
o Positive impact in scenario 2 only without TFA information
This conflicting summary indicates that the analytical model outlined in the next chapter
should also investigate the interaction effects between these two treatments.

4.2 Findings of the multivariate analysis

In this section, we present the summary results of the multivariate analysis, conducted using
binary logistic regression to identify the major determinants of consumer choices in the
experiments. Good (or healthy) choices are defined as respondents choosing product A/C in
each of the tasks 1 and 2.
Firstly, we wanted to identify the drivers and barriers of making healthy choices. Secondly,
we wanted to test the potential effect of the treatments under evaluation in this study.
The presentation of the results begins in chapter 4.2.1 with the summary of results of what
we refer to as the ‘base model’, i.e. not controlling for the potential effect of treatments, but
focussing on the main individual biases and barriers.
This is followed by chapter 4.2.2, which adds the treatments to the base model to identify the
most effective measure in improving consumers´ choices.

72
Impact of Food Information on Consumers’ Decision Making

To make it easier to read, in the following analysis, exact technical values are replaced by
symbols showing the direction (positive or negative) and strength of influence, and the
significance levels30. The appendix provides a detailed overview of the results of the
multivariate analysis.

4.2.1 Main biases and barriers of healthy TFA choices

As already discussed, the selection of variables to be included within the base model was
based on conceptual considerations and in-depth analysis of the bivariate associations
between the variables considered, as well as for the dependent variable in the behavioural
experiments.
Since the two choice tasks involved different biases and barriers, there were three base
models: one for Task 1, one for Task 2 and a final one looking at what drove healthy choices
across both tasks.
All models were structured in the same way, covering the following four dimensions:
 Prior awareness of a specific subject and/ or accessing the necessary information
during the experiment:
o Zooming specific food information in the experiment
o Pre-experimental knowledge on all fat types (TFA, PHO, FHO, SF)
 Motives and attitudes (including shopper types31):
o Shopper types
o General interest in food information
 Habits:
o Reading habits when shopping for crisps
o Recent buying experience of crisps
o Whether following a weight loss programme / diet
 Socio-demographics:
o Age, gender, income, education, household composition, body mass index,
countries.
The socio-demographic variables were selected according to their expected and/or actual
impact on the quality of consumer decision-making.

30
Three stars (***) indicate a strongly significant impact with a probability p = 0.001, i.e. there is only one
chance in a thousand that the observed result could have happened by coincidence. Two stars (**) indicate
a moderately significant impact with p = 0.01 and one star (*) indicates a slightly significant impact with p
= 0.05.
31
Furthermore we have run all these models with the original variables that are the base for the typology
(shopping goal and engagement in Q9/11). On the whole, the multivariate analysis shows the same results.
Hence, the validity of the subsequent results can be additionally confirmed. To avoid multicolinearity the
regression models were run either with the typology or with the original variables of the typology.

73
Impact of Food Information on Consumers’ Decision Making

The three base models are all statistically significant and explain a reasonable portion of the
variation in the dependent variable.

Base model for the first choice task

The first TFA task required the identification and correct evaluation of a product with and
without TFA or PHO. All other food information was kept identical for the two products.
The most relevant factors driving choices in this task were awareness variables and, to a
lesser extent, motivational and habitual biases as well as socio-demographics.

Figure 35 Drivers of healthy choices in TFA Task 1 without treatment options32

Drivers of healthy choices in TFA task 1 14

R2: %
Base model without policy options

Reg.
Awareness Motives Ha- Socio-
coeff. & goals bits demographics
Zooming Shopper types
TFA is unhealthy

2,5
Reads ingredients
when shopping crisps

Higher education
Household with 2+ persons

Younger age (<34)

Ingredients

Frustrated Shopper
Bargain Hunter
Country of origin

Rushed Quality Shopper

Nutrition label

2,0

1,5

1,0

0,5

0,0 *** ** *** ** * * * * ** * *

Extract of results based on logistic regression analysis to identify statistically significant drivers of healthy fat choices
in task 1 (with and without TFA/PHO). The model explains 14% of the variance. Only variables with at least slight
significance (* = p < 0.05) are displayed. Base: EU8 (without missing variables) n = 3 945

Awareness
Accessing relevant food information by zooming on the nutrition label (strongly significant)
and zooming on the ingredients (moderately significant) was the main driver of good choices.
This is hardly surprising, for information to be used, it has to be accessed first.

32
The model explains about R2 = 14%. This represents a high proportion given the type of variable at stake.
(Categorial variables only take values 1 or 0. R2 and minimum least squares line fitting the observed points
is going to be far from both values: 0 and 1.)

74
Impact of Food Information on Consumers’ Decision Making

What is more surprising is the comparably stronger effect of zooming into the nutrition label
as opposed to the ingredients list. Only half of the respondents were provided with relevant
differences on the nutrition label, while all respondents were presented relevant differences
in the ingredients list. One would therefore expect that the ingredients list would have a higher
influence than the nutrition label.
However, the empirical evidence shows that the information accessed on the nutrition label
was much stronger than the information provided on the ingredients list. The terms PHO and
FHO which were used in the ingredients list are both less established in consumer minds than
other fat types (see chapter 4.1). This may explain why fewer consumers who zoomed on the
ingredients were able to identify the healthy option, compared with those who zoomed on the
nutrition facts.
Another strong barrier is the attention paid to ‘irrelevant’ food information in the experiment,
such as zooming on the country of origin. Consumers who were interested in this criteria did
not find any differentiating information when they zoomed on the fine print, and subsequently
were making less healthy choices.
Of the questions exploring existing knowledge of fat types, only one – namely “TFA is
unhealthy” – turned out to have a moderately positive impact on the choice. Those who knew
TFA is unhealthy were more likely to pick the healthier product.
Knowledge about PHO (correctly unhealthy, incorrectly healthy or no idea about PHO) had no
impact– despite also being a key criterion for an informed choice.

Motives & Goals

Of the six different shopper types, the expert shopper (who has sufficient interest, money
and time to look for healthy products when shopping for food) performed best in identifying
the healthier fat combination in both tasks. Therefore this type was used as a reference group
in the analytical model to identify the types with the higher motivational barriers. In the first
task there were three types who performed less well:
 The Bargain Hunter (no money, no interest but enough time),
 The Frustrated Shopper (no time, no money, no interest),
 The Rushed Quality Shopper (no time, but enough money and interest).
This result is not surprising - a stronger emphasis on time and cost (compared with quality or
health aspects) limits the ability to make healthier choices.

Habits
The only habitual factor with a slightly significant impact is the common practice of reading
ingredients when shopping for crisps. Individual habits of paying attention to food information
was collected at the beginning of the questionnaire and used as a reminder in the introduction
of the experiment. This design helped to increase the proportion of respondents zooming on
the ingredients list, which again contributed to a more informed choice.

Socio-demographics
Younger respondents and those living in larger households were most likely to make healthier
choices in the first task:
 Shopping for at least one other person has a moderately significant positive impact
 Being younger (under 35) has a slightly significant positive impact

75
Impact of Food Information on Consumers’ Decision Making

 Higher education (those finishing full-time education aged 22 and over) has a slightly
significant negative impact
All other socio-demographic factors played no part in the choices. There was no country effect,
nor were there differences by gender, income or body mass index.

Base model for the second choice task

The second TFA task required the identification and correct evaluation of two types of food
information: the healthier product in this case contained a small amount of TFA/PHO and also
a small amount of saturated fat (SF/PHO), while the less healthy product included no TFA/PHO
but a very high amount of SF/PHO. All remaining food information was identical across the
two products.
The most relevant factors driving the choices in this task were again awareness factors but
very few additional variables, which means that the model only explains 18% of the variance
(although this is higher than the 14% explained by the base model for the first choice task).

76
Impact of Food Information on Consumers’ Decision Making

Figure 36 Drivers of healthy choices in TFA Task 2 without treatment options33

Drivers of healthy choices in TFA task 2 R2:

18
%

Base model without policy options

Reg.
Awareness Motives Habits Socio-
coeff. & goals demographics
Zooming
2,5
food information
General interest in
SF is unhealthy

Reads saturates
when shopping crisps

Finland

Younger age (<34)

Ingredients
Nutrition label

2,0

1,5

1,0

0,5

0,0 *** ** *** * * * *

Extract of results based on logistic regression analysis to identify statistically significant drivers of healthy fat
choices in task 2 (small TFA difference with high SF difference). The model explains 18% of the variance. Only
variables with at least slight significance (* = p < 0.05) are displayed.
Base: EU8 (without missing variables) n = 3 945

Awareness
Similarly to the first task, those who zoomed on food information performed better. The
attention given to the nutrition information was the strongest driver of choosing the healthier
product. Zooming on the ingredients list also had a moderately significant impact. Both labels
were relevant to making an informed decision.
In contrast to the observed behaviour in the first task, zooming on the country of origin label
did not have a negative impact on the quality of the choice. This may be the expected learning
effect from the earlier task. In the second task, fewer respondents zoomed on irrelevant
features.
Of the variables describing the pre-experimental knowledge of fat types, knowing that
“Saturated fat is unhealthy” had the most positive impact on the quality of the choice made.
Stated previous knowledge about TFA or PHO (whether correctly or not) had no impact.

33
The model explains about R2 = 18%. This represents a high proportion given the type of variable at stake.
(Categorial variables only take values 1 or 0. R2 and minimum least squares line fitting the observed points
is going to be far from both values: 0 and 1.)

77
Impact of Food Information on Consumers’ Decision Making

Motives & Goals

Shopping types do not show any noticeable effect in the second task. However, those who
said they had a general interest in food information were more likely to perform better.

Habits
The only habitual aspect – featuring a modest positive effect – was the practice of reading
the information on amount of saturates when shopping for crisps.

Socio-demographics
Healthier choices in the second task were observed more often among younger respondents
(below 35) and less often among Finnish respondents than in other countries34.
All of the other tested socio-demographic factors had no impact.

Base model for both tasks

As has been described, there were three common drivers of healthier choices in the tasks:
 Zooming on the nutrition label,
 Zooming on the ingredients,
 Age, with younger respondents more likely to choose the healthier option.
Having said this, all other determinants and barriers varied and a combined analytical model
was constructed to see if this was able to explain any further what was driving choice
behaviour in the tasks. Both choice tasks were included with the same weight, as it was not
possible to know which scenario is more common in reality.
This combined model required an additional variable representing the different choice
architecture of both tasks. As the second scenario achieved a higher share of healthy choices,
this task was used as the reference group35 in the combined base model with the key drivers
shown in the table below.

34
Poland was selected as the reference country in all TFA models since this country was the closest to
“average” choice behaviour in both TFA tasks.
35
If the first task were used as a reference group, the results would remain the same. The only difference
would be the presentation: instead of a negative direction for the first task, there would be a positive
direction for the second. The interpretation remains identical: Healthy choices are more likely in the second
scenario and less likely for the first.

78
Impact of Food Information on Consumers’ Decision Making

Figure 37 Drivers of healthy choices in both TFA tasks without treatment options

Drivers of healthy choices in TFA tasks 1+2 R2:

17
%

Base model without policy options

Reg. Con- Awareness Motives & goals Habits Socio-

coeff. text demogr.
Zooming Shopper types
2,5

TFA is unhealthy
Choice task 1

SF is unhealthy

Reads saturates
when shopping crisps

Reads ingredients
when shopping crisps
General interest in food info

Younger age (<34)

Self-Determined Shopper
Ingredients

Frustrated Shopper
Country of origin

Bargain Hunter

Rushed Quality Shopper

Nutrition label

2,0

1,5

1,0

0,5

0,0 *** *** *** *** *** n.s. ** * * * * * ** **

Extract of results based on logistic regression analysis to identify statistically significant drivers of healthy fat choices in both choice tasks.
The model explains 17% of the variance. Only variables with at least slight significance (* = p < 0.05) are displayed.
Base: EU8 (without missing variables) n = 3 945

The most relevant factors driving healthy choices in both choice scenarios were again primarily
awareness variables, followed by the choice architecture, the age group, habitual biases and
eventually to a lower extent some goals and motivations.

Context
The context and choice architecture in task one had in its essence a much simpler design than
the second task, if one assumes complete transparency and product literate consumers. There
was only one parameter to be identified and evaluated correctly. However, the majority of
consumers were not able to even spot this difference. The key barrier in the first choice was
that the tested type of food information (differing levels of TFA, PHO, FHO) is not yet seen in
reality on food products in Europe.
On the other hand, the second scenario presented a combination of more familiar food
information (differing amount of saturates) in addition to future/hypothetical food information
(differing levels of TFA, ranking of PHO). The fact that the primary piece of information was
more familiar to consumers certainly helped increase the number of healthy choices. This bias
in the first task could therefore be termed an ‘unfamiliarity’ bias.

79
Impact of Food Information on Consumers’ Decision Making

Awareness
Almost all awareness variables discussed previously in the description of the individual tasks
were very strong drivers in both scenarios. Paying attention to the relevant labels has a very
positive impact, while zooming on irrelevant information in this task was counterproductive
to making a healthier choice.
Having knowledge prior to the experiment that saturated fat is unhealthy remained a strong
driver overall in the combined model. However, previous knowledge of the unhealthiness of
TFA had no impact at all in this combined perspective. The positive influence in the first task
was flattened out by the negative influence (albeit not significant) of a superficial knowledge
of TFA in the second task.

Motives & Goals

All motivational drivers that were identified in the separate analyses were also visible in the
combined model with the same strength and direction, with the addition of one of the shopper
types – the self-determined shopper – as a negative influence on healthy choices.
Generally speaking, if a shopper’s focus is more on resources such as time or costs rather
than quality or health then the likelihood of making a healthy choice is slightly reduced – in
particular for the choices in the first scenario because they have less interest in reading food
information or trying to identify healthier alternates

Habits
Two consumer habits contribute positively to the ability to make healthy decisions. If someone
is used to looking for the ingredients list and checking the amount of saturates when shopping
for crisps, then this was also the behaviour that supported the correct identification of the
healthier alternatives in the experiment.

Socio-demographics
The only variable that remains a significant driver in the combined model is age. Younger
consumers (below 35) were significantly more able to make healthy choices compared to
older consumers.
One immediate theory for this is that the impact of age could be a result of the online design
of the experiment. The zooming functionality in an online survey is not very common yet. The
higher familiarity of younger participants with these digital features could be the reason
behind increased propensity to make healthy choices. However, there is no empirical evidence
supporting this hypothesis. An additional regression analysis conducted to understand the
drivers of zooming behaviour is described at the end of the subsequent Chapter 4.3.2.
All of the other variables identified in the separate models (household size, education and
country) were not relevant in the combined model.

4.2.2 Effectiveness of tested treatments on TFA

To answer the second key question in this research on the most effective interventions, the
base model was enhanced by adding the tested treatments.

80
Impact of Food Information on Consumers’ Decision Making

There are different ways of including these variables, either as condition splits based on the
treatment combinations described earlier:
 Split 1: with TFA information and with education
 Split 2: with TFA information but no education
 Split 3: no TFA information but with education
 Split 4: no TFA information and no education (control group)

Alternatively they can be presented as individual stimuli with the interaction effect observed
between both measures:
 With TFA information
 No TFA information (reference category)
 With education
 No education (reference category)
 Interaction effect of TFA info & education

Both options deliver identical results. The latter version was used in the base model. The
following tables present the same structure of variables as used in the base model described
earlier but with the treatments added to the analysis. Of course, the explanatory power of the
enhanced model is higher than the base models shown before by explaining 46%36 (up from
only 14% in the task 1 base model) of the choice behaviour in Task 1 and 20% (up from 18%
in the base model) in Task 2.

36
Binary logistical regression models are typically evaluated by Pseudo R2 to assess the quality of the model
by its explanatory power. The displayed R2 has been defined according to Nagelkerke.

81
Impact of Food Information on Consumers’ Decision Making

Figure 38 Drivers of healthy choices in TFA Task 1 with treatment options

Drivers of healthy choices in TFA task 1 R2: 46

%
Model with policy options
Reg.
Policy options Awareness Motives Habits Socio-
coeff. & goals demogr.
Zooming

TFA is unhealthy

Frustrated shopper
SF is unhealthy

General interest in
food information
Education

2,5
TFA info & education

when shopping crisps

when shopping crisps

Younger age (< 34)

Household with 2+ persons
Reads calories

Reads ingredients
Combination of

Lower income
TFA Info

Ingredients

Country of origin
Nutrition label

2,0

1,5

1,0

0,5

*** n.s. *** *** *** *** ***

0,0 * * * * * * * *
Extract of results based on logistic regression analysis to identify statistically significant drivers of healthy fat choices in task 1 (w ith and w ithout
TFA/PHO). The model explains 46% of the variance. Only variables w ith at least slight significance (* = p < 0.05) are displayed.
Base: EU8 (without missing variables) n = 3 945

The inclusion of the treatment options into the base model changes the impact of several
variables. A higher relevance and/or strength can be observed for:
 All awareness variables with relevant information i.e. zooming on the ingredients and
nutrition information as well as prior knowledge about TFA and saturated fat
 The general interest in reading food information
 The habit of checking calories when shopping for crisps
 Consumers with a very low income (less than 10k€)
This is clearly a result of the inclusion of the additional food information and education in the
model, which made the health disadvantage of TFA clearer to respondents who was presented
and saw the TFA amounts disclosed in the nutrition information and who are more interested
in this area. Respondents in the control group who were presented ‘no TFA info’ and ‘no
education’ are not included in this model, which focuses on the impact of new and additional
food information for consumers on the choice task.

Treatment options
The key driver of healthy choices in Task 1 is the presentation of a concrete amount of TFA
on the nutrition label. This is by far the strongest influencing factor if the decision is between
a product with TFA and a product without TFA.

82
Impact of Food Information on Consumers’ Decision Making

In contrast to this, the educational stimulus alone has no statistically significant influence.
The negative direction that was observed in the descriptive analysis is not significant when
looked at in this model.
The combination of an educational intervention and the TFA amount on the nutrition label
showed a strong momentum towards more healthy choices in Task 1.
In Task 2, the impact of the additional treatments is far less pronounced than in the first
scenario as can be seen in the table below.

Figure 39 Drivers of healthy choices in TFA Task 2 with treatment options

R2: 20
Drivers of healthy choices in TFA task 2 %

Model with policy options

Reg.
Policy options Awareness Motives Habits Socio-
coeff. & goals demogr.
Zooming
2,5
TFA is unhealthy

General interest in
food information
SF is unhealthy

Self-Determined Shopper
Education

TFA info & education

Reads saturates
when shopping crisps

when shopping crisps

TFA Info

Finland

Younger age (<34)

Reads ingredients
Combination of

Ingredients

Country of origin
Nutrition label

2,0

1,5

1,0

0,5

** n.s. n.s. *** ** * n.s. *** * * * n.s. * *

0,0
Extract of results based on logistic regression analysis to identify statistically significant drivers of healthy fat choices in task 2 (small TFA difference
with high SF difference). The model explains 20% of the variance. Only variables with at least slight significance (* = p < 0.05) are displayed.
Base: EU8 (without missing variables) n = 3 945

Regardless of whether the model includes the treatments or not, there are robust results for
most individual variables already included in the base model. Only two changes among the
previously explained variables can be observed: Participants who zoomed on the country of
origin label as well as the Self-determined Shopper type were not less likely to make healthy
choices if exposed to the additional TFA information.
Treatment options
In this second choice task, however, the additional TFA information presented a significant
barrier to healthy choices. This is no surprise as the negative impact was already observed in
the previous chapter. In contrast to the first scenario the product with TFA was healthier
overall because it included less saturated fat.

83
Impact of Food Information on Consumers’ Decision Making

The education stimulus presented alone seemed to work in a positive direction in the bivariate
analysis, but the statistical model shows that this effect is not significant.
Moreover, the combination of TFA information and education did not support consumers in
making healthy choices, although the negative tendency of this combination is not statistically
significant.
Overall, none of the tested treatments fostered healthy choices. The TFA information in
particular had a counterproductive impact, which while not on the scale as the positive impact
in Task 1 but strong enough for consumers to be distracted by the TFA content so that they
did not notice the unhealthy amounts of saturates. It should be noted, however, that the
second task was designed to test for overreaction when TFA information is provided, even
when the healthier choice contained some TFA and saturates compared to the unhealthy
choice with very high amounts of saturates but no TFA. In other words, the provision of TFA
information in the nutrition facts was not expected to improve participants' decision in this
task. At best, expert consumers could have understood that the high amounts of saturates
outweigh the small amounts of TFA.
To summarise all relevant drivers and barriers across both choice tasks with the treatments,
the following table presents the final overview.
Figure 40 Drivers of healthy choices in TFA tasks 1+2 with treatments

16
Drivers of healthy choices in TFA tasks 1+2
R2:
%

Model with policy options

Motives Socio-
Reg.
coeff.
Policy options Con-
text Awareness & goals Habits demogr.
Zooming Shopper
TFA is unhealthy
Choice task 1

SF is unhealthy
Education

TFA info & education

food information

when shopping crisps

when shopping crisps

Reads saturates

Reads ingredients
TFA Info

Younger age (<34)

2,5
General interest in
Combination of

Self-Determined Shopper
Frustrated Shopper
Ingredients

Country of origin
Nutrition label

2,0

1,5

1,0

0,5

n.s.
0,0 *** * *** *** *** *** *** * * * ** * * **
Extract of results based on logistic regression analysis to identify statistically significant drivers of healthy fat choices in both choice tasks.
The model explains 16,5% of the variance. Only variables with at least slight significance (* = p < 0.05) are displayed.
Base: EU8 (w ithout missing variables) n = 3 945

Treatment options

84
Impact of Food Information on Consumers’ Decision Making

When combining both tasks, the same treatments that were effective in task 1 are also
effective in making healthy choices overall. The negative overreaction in task 2 when
presenting TFA information on the nutrition label is not seen when looking at both choice
context scenarios. Presenting TFA information alone helps consumers to make better choices
as does to a lesser extent the combination of education and TFA transparency.
Education on fat types alone is not retained long enough to enable healthy choices – even
within the short, controlled timeframe of laboratory experiments. This might be due to the
fact that terms like PHO and FHO are not familiar to consumers yet and the terms are easily
confused.
Moreover, it is a very complex topic and requires time and expertise to understand the
differences between all fat types, how they relate to each other and which is less healthy in
relation to other types. It will be quite difficult to develop simple information and education
on ‘which fat combination is better or worse’ that is easily absorbed by the majority of
consumers.
In summary, these are the main drivers and barriers of healthy ‘fat’ choices:
The key drivers are
 Correct awareness of unhealthy fat types,
 Zooming on relevant food information and ignoring irrelevant labels,
 More familiar choice architecture as in Task 2.

Further less significant but positive drivers are

 General interest in food information,
 Habit of reading food information such as the amount of saturates and ingredients
when shopping crisps,
 Younger consumers under 35.

Further less significant barriers are

 Motivations of Frustrated Shopper and Self-determined Shopper types. Both types
have the highest risk-taking attitude and are not interested in healthy alternatives.
This summary accentuates the importance of the first and the last stage of the decision
process. Access to relevant food information is the main barrier that needs to be overcome.
The second hurdle is the context and framing at the decision stage. In order to better
understand how to reduce this initial barrier of consumers actively accessing the relevant food
information, the following explains the drivers and barriers of zooming behaviour in the
experiment. The conclusions based on these findings are outlined in Chapter 7.1.

Drivers and barriers of zooming behaviour in TFA choice tasks

In addition to the main research objectives, there are further insights that can help fine tune
our conclusions. This section explores what makes reading food information an issue which is
salient (or relevant) to consumers.

85
Impact of Food Information on Consumers’ Decision Making

While the question about whether consumers access food information in the experiment is
the key survey metric for making healthy choices, the study also indicates that not everyone
is interested in this kind of information although most say they regularly read food labels37.
Therefore, it can be valuable to look more closely at the factors that influenced the zooming
behaviour of consumers in the choice experiment. Three quarters (75%) of participants
zoomed on at least one relevant food label (nutrition facts and/or ingredients list) in the
experiment.
There are several hypotheses for how the zooming behaviour might be influenced e.g.
a) Did the educational intervention at the beginning of the experiment increase zooming
activity?
b) Were shoppers with less interest in health goals also less likely to zoom on food
information?
c) Did reading habits when shopping have an influence?
d) Were younger consumers more familiar with the technical zooming feature?

To identify the relevant drivers a logistical regression model was developed with the goal
variable “zoomed at least once on any label in either task” and a set of potentially influencing
variables covering similar dimensions as the base model but adjusted to the new analytical
focus:
• Education on fat types (no education as reference group)
• Context / choice task (as in base model)
• Awareness of fat types (as in base model)
• Relevance of motives and goals (shopper types as in base model)
• Habits (as in base model)
• Socio-demographics (countries, gender, age, education, income, household type,
BMI as in base model)
The result of this analysis (explaining about 8% of the variance with strong significances) is
displayed in the table below.

37
76% of all participants agree with the statement “I read labels and pay close attention to product
information”. It can be assumed that this reflects a high overconfidence bias rather than actual behaviour.
The observed zooming behaviour in this experiment certainly does not reflect the label reading behaviour in
the reality of a shopping trip with a wide range of products with brands, prices and much more food
information.

86
Impact of Food Information on Consumers’ Decision Making

Figure 41 Drivers of zooming any ingredients or nutrition facts in tasks 1+2

Drivers of zooming any ingredients or nutrition R2: %

8

facts in tasks 1+2

Model with policy options
Reg. Policy
Awareness Motives & goals Habits Socio-
coeff. option demogr.
Shopper types

Romania
TFA is healthy
SF is unhealthy

Reads ingredients
when shopping crisps

programme / diet

Women
Education

SF is healthy

when shopping crisps

Following a weight loss
General interest in food info
2,5

Reads saturates
Frustrated Shopper

Self-determined Shopper
Ambitious Health Planner

Bargain Hunter
Rushed Quality Shopper
2,0

1,5

1,0

0,5

0,0 *** *** ** *** *** *** *** *** ** ** *** ** ** *** **

Extract of results based on logistic regression analysis to identify statistically significant drivers of zooming behaviour in both choice tasks.
The model explains 8% of the variance. Only variables w ith at least slight significance (* = p < 0.05) are displayed.
Base: EU8 (without missing variables) n = 3 945

A wide variety of drivers can be identified with strong significance confirming the first three
hypotheses. Only the last hypothesis was not verified by the analysis. The age of participants
did not have an effect on whether the zooming functionality was used or not during the
experiment.

The drivers of zooming on food information were:

• The education treatment had a positive and strongly significant effect on
whether respondents zoomed on relevant food labels.
• Lack of motivation and goal to read food labels
Shopper types that lack either interest, money or time were less likely to access
food information. The expert shopper (characterised by having enough interest in
health, enough money and time when shopping) displayed the highest rate of
zooming in the experiment and was therefore used as a reference group. All other
shopper types were significantly less likely to zoom on the information.
• Habits of reading relevant food information
Consumers who usually read ingredients and check the amount of saturates were
more likely to do the same in the experiments. This was also the case for consumers
who are currently following a weight loss programme.

87
Impact of Food Information on Consumers’ Decision Making

• Specific images of fat types

Previous awareness of saturated fats (whether correct or not) had a strong positive
influence (with the exception of people who incorrectly thought that “TFA is
healthy” who were less likely to zoom on the information). Consumers who did not
know anything about types of fat prior to the experiment did not zoom at all in this
experiment.
• Women and Romanian respondents were more likely to zoom on the information.
As expected, the drivers of active attention (zooming) were a higher priority for people with
health goals when shopping, as well as those with a minimal level of existing knowledge of
fat types and the habit of reading relevant food labels.

4.3 Conclusions regarding TFA

For the summary of findings and conclusions of the behavioural study the graph below can
serve again as a conceptual guideline. It represents a simplified and idealised illustration of a
comprehensive decision process. It covers the typical stages of a shopper’s decision process
from accessing and assessing relevant food information to taking the final decision.
The identified individual biases and external barriers as well as the tested treatments are
marked with symbols to indicate their direction and effectiveness of impact on making
healthy choices.

88
Impact of Food Information on Consumers’ Decision Making

Figure 42 Summary of results on TFA analysis


The evidence from this study confirms that being able to identify the
healthier alternative is largely driven by initially accessing the relevant
information on the food label and the familiarity of the choice scenario.

These are the key findings of how individual biases and external barriers influence consumer
choices with varying combinations of fat types:
 The key driver of healthy choices is the zooming on the nutrition label for both
choice tasks. Among all three different food labels the nutrition label was clicked by
far the most often. This is surprising as only half of the participants received relevant
information on this label in particular for Task 1. Yet, the access to this label provided
more support for a correct identification than any of the other two food labels. This
indicates that the more structured design of the nutrition label is better suited to show
relevant food information (amounts in particular) compared with the ingredients list.
 Zooming on ingredients list also fostered healthy choices. This label contained all
of the relevant information for all participants in both choice tasks, but only a small

89
Impact of Food Information on Consumers’ Decision Making

proportion accessed this information. Therefore, its influence was less strong than the
nutrition label.
 Zooming on irrelevant labels such as the country of origin in these tasks presented
a strong barrier as they indicated other consumer preferences and distracted from the
relevant information about fat types.
 Familiarity with at least some of the choice architecture was the second most
important driver of healthy choices. The trade-off between zero and high amount of
TFA was a more difficult task to respondents, as they were not used to such a decision.
The additional differentiation of amounts of saturates in Task 2 was more familiar and
would lead to more healthy choices normally.
 Previous knowledge that saturates are unhealthy was also a key driver as it
directed consumers to focus on the relevant label in the tasks. Likewise pre-
experimental knowledge that TFA is unhealthy had a positive effect albeit smaller
and less significant. Incorrect perceptions of fat types and PHO or FHO had no impact.
 The ability to identify the healthier product was also driven by age. Younger
shoppers (below 35 years) were significantly more adept in both tasks.
 Individual interest in food information for shoppers with health objectives
contributed to more healthy choices.
 Individual habits such as checking saturates or ingredients when shopping for crisps
had a limited impact.

How can the conscious accessing of food information and the familiarity
and clarity of the choice architecture be improved in order to encourage
more healthy choices?

To identify the most effective treatment in driving healthy consumer choices, the analysis of
the experiments looked at the impact of each individual treatment (education on fat types or
concrete TFA information on the nutrition label) as well as at the combination of both.
Some of the key findings about the effectiveness of the tested treatments influencing the
consumer choices are as follows:


The application of concrete amounts of TFA on the nutrition label has a
strong and positive impact on healthy consumer choices with or without
an educational clarification treatment.

 The additional information about TFA is like a double-edged sword if the two observed
scenarios are looked at separately. It supports healthy choices in scenarios where only
the amount of TFA differs. At the same time it distracts consumers from healthy
choices in scenarios where there are other unhealthy fat types. However, the overall
effect of the treatment to display the amount of TFA on food has a positive impact.

90
Impact of Food Information on Consumers’ Decision Making

 A simple educational intervention alone is not effective at all in influencing healthy

consumer choices. The explanation and evaluation of different fat types only has an
impact if the terminology used (e.g. TFA) is also found on the nutrition label.
 If in reality the choice architecture of Task 1 (comparing only different amounts of
TFA) is the most prevalent, then the combination of education and information would
be recommended as the most effective treatment.
 If in reality the choice architecture of Task 2 (comparing TFA with other unhealthy fat
types) is more prevalent, then none of the tested treatments would be recommended
as they would have either a negative impact (TFA information only) or no impact
(combination of both treatments) on healthy choices.


Overall the most efficient of the tested treatments was the display of TFA
information on the nutrition label which provided more clarity in the final
decision stage for identifying the healthier alternative.

 The survey revealed a general lack of awareness of terms such as partially

hydrogenated oil (PHO) or fully hydrogenated oil (FHO). This explains why choices
between these terms (without TFA information) were mostly not possible because
people didn’t know the difference. Moreover, these terms were displayed in the
ingredients list and not the nutrition fact label which was much more likely to be
accessed (zoomed on) in the experiment.
 Whether consumers will become more familiar with the terms PHO and FHO once they
are implemented on food labels in Europe in 2014 could be subject to further research,
e.g. by field experiments.

91
Impact of Food Information on Consumers’ Decision Making

5 EXPERIMENTAL ‘LAB’ FINDINGS ON PRECAUTIONARY ALLERGEN

WARNINGS (PAW)

This section describes the key findings of the experimental part on the usage of PAW on food
labels in order to answer the following research objectives:
a) What do consumers cognitively understand, when they see a PAW? What is the level
of risk associated with different types of PAW wordings as well as with the absence of
any PAW?
b) Do consumers take a cautious approach? What leads consumers to restrict their choice
by choosing what appear to be safer products without a PAW?
c) What is the impact of the tested PAW labels on the choice process?
The inherent nature of PAWs required the identification of the relevant target group of
consumers that the subsequent analysis would be based on, as not all consumers are
personally affected by food allergies or experienced in shopping for someone with a food
allergy. Since this policy area represented only one topic amongst several within the study,
there was no booster of allergic consumers in the sampling design. Instead the individual
relevance and experience of allergen information for food was collected before the
introduction of the experiments. This was used as the base for a segmentation, allowing the
comparison of choice behaviour between a realistic scenario (affected consumers) and a
hypothetical scenario (unaffected consumers).
Overall 596 affected consumers were identified within the sample. This is a robust base for
covering personally affected consumers as well as consumers who shop for affected household
members.
Only 18% of these affected consumers said that they experienced severe allergic reactions if
food was contaminated by an allergen. Even so, this segment was quite homogeneous as
almost everyone was experienced in reading allergen information when shopping food. Hence,
the ‘affected’ sample is the base for the core analysis of results and wherever necessary there
is a comparison with the ‘unaffected’ sample base.

92
Impact of Food Information on Consumers’ Decision Making

Figure 43 Relevance of allergen/PAW information

Relevance of allergen/PAW information

15%
Personally affected or
Affected sample
9,5% Only household member affected (n = 596)
5,4%

Allergic Reading habit

reaction allergy labels
18% Severe
85,1%
47% Yes, regularly

46% Moderate

Not affected by any Yes,

food allergy / intolerance 45% occasionally
37%
or previously
Mild
8% No, never

Q14: Do you have any dietary restrictions? Q16: How would you rank the severity of the allergy/intolerance?
Q15: Does anyone else in your household (for whom you Q17: Do you ever look for allergy information, when shopping for food?
also shop at least occasionally) have any dietary restrictions? Base: EU8 – all respondents with food experiments that are affected
=> both with answer option “Food allergy or intolerance” either personally or by shopping for a household member (n = 596)
Base: EU8 – all respondents with food experiments
(n = 4 036)

To contextualise the findings of this and to structure the analysis in order to come to an
understanding of the factors that influence the choice process in a shopping scenario, we put
forward a map of a simplified decision process (as explained in Chapter 4) adapted to this
policy area.

93
Impact of Food Information on Consumers’ Decision Making

Figure 44 Guideline for analysis of results on PAW

The analysis starts with a description of the consumer’s risk assessment of various versions
of allergen information in Chapter 5.1. This section covers the cognitive perception of
consumers when shown a list of differently worded PAWs.
The following Chapter 5.2 covers the disaggregated findings of the observed decision
behaviour in the choice tasks with suggestions for variables to be included in the statistical
base model. This is supplemented by a descriptive overview of observed choice behaviour for
the tested PAW alternatives.
Chapter 5.3 combines the discussed variables in a multivariate analysis to identify the
statistically relevant drivers and barriers of allegedly safe choices.

5.1 Consumers’ cognitive risk assessment of PAWs

This section looks into the question about what consumers cognitively understand, when they
see a PAW on a product such as muesli. In particular, what is the level of risk they associate
with each individual PAW wording as well as the absence of any warning.
The statements to be evaluated were presented as a matrix question in a randomised order
to avoid ranking effects or any bias for the overall analysis.

94
Impact of Food Information on Consumers’ Decision Making

In addition the unambiguous option ‘contains nuts’ was included, which served as an indicator
of the difficulty of this task. If all respondents identified this option as definitely unsafe, then
the task could be deemed as easy. However this was not the case (see table below), although
participants affected by allergy issues were more likely to identify the ‘contains nuts’ label
correctly as definitely unsafe than not affected consumers.
In comparison to all other labels the ‘contains nuts’ information achieved the highest mean
risk score among both the affected sample (4.0) as well as the unaffected sample (3.7).
At the other end of the evaluation scale – if no information was given about an allergen – this
was regarded as less risky by the average consumer. The risk assessment of this option hardly
differed between both samples.
The four tested PAWs were all located between these two anchor points. The risk ranking was
largely identical for both sample segments although the risk levels assigned by the affected
consumers were generally higher.

Figure 45 Risk assessment of PAWs

The differences in the evaluation between the four PAW options were not strong enough to
justify a robust ranking as presented in the chart. However, there is a tendency indicating
that the more precise the wording of a PAW the lower the risk associated with it by consumers.
 ‘Made in a facility that also processes nuts’ achieved the relatively lowest risk score of
3.6
 ‘May contain traces of nuts’ ranked as second lowest level with a mean risk score of
3.7

95
Impact of Food Information on Consumers’ Decision Making

 ‘May contain nuts’ and ‘cannot guarantee nut free’ both have an identical mean score
of 3.8.
There is a slightly significant lower risk level for ‘Made in a facility’ compared with any of the
other three PAWs. ‘Made in a facility’ is also the only PAW whose assessment is not
significantly more risky than that that of no information at all. The labels ‘May contain’, ‘May
contain traces’ and ‘Cannot guarantee’ were all moderately significant with regards to a more
risky perception compared with ‘no information’.

Figure 46 Cognitive risk assessment of PAW wordings

Cognitive risk assessment of PAW wordings

Affected respondents – Significance levels

Higher risk Lower risk

4.0 3.9 3.8 3.7 3.6 3.5

May n.s. Made in a n.s.

contain facility …
Contains ** May contain No
nuts traces information
Cannot
guarantee *
**

Q40: Based on your understanding of the risk for someone with a nuts allergy, how safe would you consider muesli to be if it
had the following information on the pack? Scale: 5 = definitely unsafe, 4 = probably unsafe, 3 = not sure, 2 = probably safe,
1 = definitely safe. Base: EU8 – all respondents with food experiments that are affected by food allergy/intoleranc e (n = 596)

In summary, around two thirds of the affected consumers considered any of the displayed
PAWs as unsafe. Hence, one would assume that a similar share would also take a cautious
approach by avoiding products which display these types of PAW.

5.2 Observed choice behaviour

To understand how consumers react when given a choice between a product with and a
product without a PAW, an experimental choice task was designed to observe the choice
behaviour of participants.
Unlike the TFA experiment described in the previous chapter, the observation of the choice
behaviour was intended to capture not only the cognitive part of a decision but also intuitive,
habitual and less conscious factors that also influence this type of decision in reality.
This experimental part was presented to respondents before the cognitive risk assessment to
avoid any unwanted influence from this exercise. The framing of the experiment aimed to
provide a realistic scenario while also controlling for a manageable selection of variables to
isolate the impact of the most relevant factors.
As a result, there were three subsequent choice tasks presented in a randomised order, each
offering two different muesli products:

96
Impact of Food Information on Consumers’ Decision Making

 Task 1 offered a choice between a product with local ingredients and a PAW and a
product with imported ingredients without a PAW.
 Task 2 offered a choice between two products with local ingredients. Again, the only
difference was that one had a PAW and the other not.
 Task 3 offered a choice between two products with imported ingredients. The only
difference was that one had a PAW and the other not.

The rationale for including the origin of the ingredients was to test whether this has an impact
on the risk assessment of PAWs38. Moreover, it allowed comparing the relevance of a PAW on
the choice behaviour if it is the only differentiating variable (Tasks 2 and 3) vs. if there is a
second differentiating variable (Task 1).
When focussing on allergy affected consumers, the share of safer choices is almost equal
across all three scenarios with 40% on average. This is a significantly higher rate of safe
choices compared to the unaffected sample as one would expect.

38
Of course, this is just one possible influencing factor among many others. When comparing different
products in reality, there will be further differences in terms of price, brands, colours, packaging, etc. This
might be subject to further research, if required.

97
Impact of Food Information on Consumers’ Decision Making

Figure 47 PAW choice behaviour by affected and not affected respondents

PAW Choice Tasks

Affected vs. not affected respondents

With PAW No difference Without PAW

(risky choice) to me (secure choice)

Choice task 1
Allergy affected: 28% 33% 40%
Between local ingredient
with PAW and imported Not affected: 28% 44% 28%
without PAW

9% 51% 40%
Choice task 2
Between local ingredients 6% 64% 30%

7% 52% 41%
Choice task 3
Between imported ingredients 6% 63% 31%

Q341-365: Which product would you choose? Randomised order of tasks.

Base: EU8 – all respondents with food experiments that are affec ted by food allergy/intoleranc e (n = 596) or not affected (n=3 440)

The PAW labels restricted choices for a sizeable proportion of consumers regardless of the
origin of the ingredients. A striking distinction, however, is the higher proportion of less
safe choices and the lower proportion of those who thought there was no difference in Task
1. Clearly the preference for the local origin vs. imported ingredients outweighed or
distracted consumers from the PAW label, i.e. the local origin increased the proportion that
made a more risky choice.
Another indicator for the relevance of origin information in this context is the post-
rationalisation recorded after the experiment. Around a third of the affected participants
stated that they favoured products with local ingredients which obviously increased the more
risky choices seen in Task 1.
The proportion of respondents who said that they tried to avoid the product with allergen
information also matched the 40% observed in the choice tasks (see table below).

98
Impact of Food Information on Consumers’ Decision Making

Figure 48 Choice rationale for PAW experiments

Those participants who said that the allergen information was too vague actually had a higher
tendency to avoid the product with the PAW than to ignore it.
Overall a sizeable proportion of affected consumers was not able to spot any differences,
indicating that this task is not a typical habit even for all affected consumers39.

5.2.1 Potential awareness drivers

The key requirement to identify an appropriate muesli product for a friend with a nuts allergy
was, as in real life, to access the relevant food label section in the experiment. Participants
were instructed to select and read food information as in real life therefore nobody was
specifically asked to zoom on any label. However, the difference between the two products
was hardly legible without zooming. Hence, a high level of zooming was observed in this
experiment:
 83% of all affected participants zoomed on at least one label in each of the choices.
 79% of all unaffected participants zoomed on at least one label in each of the choices.
The nutrition facts label was accessed most often (by around two thirds of participants),
although it did not provide any differentiating or relevant information.

39
These are in particular elder consumers in Finland or Eastern European countries.

99
Impact of Food Information on Consumers’ Decision Making

The crucial PAW was accessible together with the ingredients list and was more often zoomed
on by affected participants (44-46% vs. 34-37%) in all tasks. The origin information was
accessed in both samples by around a third of participants in each choice task.
Overall, this indicates that the nutrition fact label dominated attention when comparing muesli
products, despite the goal to select something appropriate for an allergic friend, which should
have nudged participants into zooming on the ingredients lists.
The research design meant that it was necessary to zoom on the information in order to make
an informed choice. It is therefore unsurprising that zooming had a strong impact.

Table 49 Zooming vs. choice behaviour in Task 1

Food split: PAW EU8 – Zooming Zooming

Allergy affected Total nutrition Zooming country of Nothing
Choice Task 1 affected facts ingredients origin zoomed
Base unweighted 596 363 271 204 102
Product A 28% 33% 15% 60% 22%
(local with PAW)

No difference
33% 35% 9% 12% 54%
to me
Product B 40% 32% 77% 28% 24%
(foreign – no PAW)

Total 100% 100% 100% 100% 100%

Table 50 Zooming vs. choice behaviour in Task 2

Food split: PAW EU8 – Zooming Zooming

Allergy affected Total nutrition Zooming country of Nothing
Choice Task 2 affected facts ingredients origin zoomed
Base unweighted 596 394 260 231 100

Product C 40% 32% 82% 39% 22%

(local – no PAW)

No difference
51% 61% 8% 55% 54%
to me
Product D 9% 7% 10% 7% 24%
(local with PAW)

Total 100% 100% 100% 100% 100%

100
Impact of Food Information on Consumers’ Decision Making

Table 51 Zooming vs. choice behaviour in Task 3

Food split: PAW EU8 – Zooming Zooming

Allergy affected Total nutrition Zooming country of Nothing
Choice Task 3 affected facts ingredients origin zoomed
Base unweighted 596 385 269 223 102
Product E 7% 5% 9% 4% 13%
(foreign with PAW)

No difference
52% 62% 11% 58% 61%
to me
Product F 41% 33% 81% 39% 26%
(foreign – no PAW)

Total 100% 100% 100% 100% 100%

 The zooming on the ingredients correlates most highly with cautious choices in all
tasks.
 Respondents who zoomed on the irrelevant nutrition facts were mostly not able to
recognise any difference.
 The enlargement of the origin label only had a significant impact on more risky choices
in Task 1.
So far the descriptive analysis suggests the inclusion of the following awareness variables into
the final analytical model:
 Zooming behaviour in the experiment
o Ingredients label with(out) PAW as a driver of making less risky choices
o Nutrition label as a barrier to making less risky choices
o Country label as a barrier to making less risky choices and a driver of local
products despite the PAW

5.2.2 Potential beliefs and attitude drivers

Consumer motives and attitudes around food shopping as well as more general personality
traits should be considered as potentially influencing factors in whether one tends to avoid a
product with a PAW or ignore this information.
The first variable that was expected to be relevant was the degree of severity of their own
food allergy/intolerance and/ or the one of the affected member in the household. The
assumption was that the more severe the allergy, the more cautious the product choices.
Instead, the data based on 596 affected participants indicates no significant impact of the
degree of allergic reaction. In fact there is a slight tendency that participants with a mild
reaction are more likely to avoid products with a PAW, in particular when it is presented in
conjunction with imported ingredients as in Task 3 (see table below).

101
Impact of Food Information on Consumers’ Decision Making

Figure 52 Selection of product without PAW vs. severity of allergy reaction

Selection of product without PAW

Based on the degree of severity of the food allergy reaction

Secure choice …

43% in task 1
Mild 42% in task 2
(n=217) 48% in task 3

37%
Moderate 40%
(n= 274) 38%

39%
Severe 37%
(n=105) 38%

Q341-364 Which produc t would you choose? Selection of product without PAW
Base: EU8 – respondents in the food experiment affected by allergy with the respective severity of the allergy reaction

The analysis of shopping goals based on the individual statements indicated the following
tendencies:
Products without PAW were less likely to be selected by participants who (strongly) agreed
that they
 Never compromise quality for a lower price,
 Always try to choose the healthier alternative,
 Read labels and pay close attention to product information (only relevant in Task 3),
 Are loyal to the brands they buy (only relevant in Task 3).
Obviously some affected consumers motivated by health and quality goals when shopping do
not consider a product with a PAW as a risky choice. There are often other types of information
such as brands they look for and trust more.
However, since the sample base of affected consumers does not allow much further
segmentation, the significance of these variables will only be tested in the multivariate base
model. In this analysis, the segments will be aggregated to allow for a sufficient base in each
category. This is also the reason, why the shopper typology will not be applied to this model
since six segments will not provide a robust sample base among the affected participants.
Furthermore, there are two general personality traits, which might be influential in these
choice tasks: risk-aversion and self-control. Yet, the direction that they indicate in the
descriptive analysis does not confirm initial assumptions in particular for the variable of risk-
aversion. Participants were less likely to avoid choosing the product with a PAW who (strongly)
agreed that
 Security is more important to them than excitement and adventure;
 They prefer doing things that pay off right away than in the future.
A general propensity towards more security is not reflected by avoiding a product with a PAW.
The self-control variable indicates that consumers with a more impulsive (low self-control)
attitude are also less guided by a PAW when selecting products.

102
Impact of Food Information on Consumers’ Decision Making

To sum up this descriptive analysis the following motivational variables are included in the
final analytical model:
 The degree of severity of the allergic reaction:
o mild
o moderate
o severe
 (Dis-)agreement with the following statements on shopping goals
o I’m never willing to compromise quality for a lower price
o I read labels and pay close attention to product information
o I am extremely loyal to brands
o I always try to choose the healthier alternative when selecting a product
 (Dis-)agreement with the following statements on personality traits
o I much prefer doing things that pay off right away than in the future
o Security is more important to me than excitement and adventure

5.2.3 Potential habitual drivers

Analysis of questions in the pre-experimental part about individual habits when shopping for
muesli revealed the following impact on allergy affected consumers in the tasks:
 Consumers who did not buy muesli within the past 12 months were more likely to
avoid the product with a PAW.
 Those who look for allergen information when purchasing muesli for the first time were
also more likely to avoid products with a PAW.
 The habit of checking the ingredients list did not seem to have any influence on the
choice tasks, whereas tending to check the country of origin seemed to distract slightly
from the allergen information.
 When looking at general habits (not related to muesli shopping), the tendency to buy
the same foods all the time seems to reduce the relevance of a PAW.
 The frequency of reading allergy information when shopping for food did not seem to
have any impact on the choice behaviour.
Overall, there are no strong habitual drivers of choice identified in this descriptive analysis.
Nevertheless it is recommended to include all these variables in the analytical model to
confirm whether or not they have any impact.

5.2.4 Socio-demographic variables

The usual set of socio-demographical control variables were added to the analysis model,
such as country, gender, age, education and income level as well as household composition.
Body Mass Index was not included as it did not have any logical or statistical impact.

103
Impact of Food Information on Consumers’ Decision Making

5.2.5 Treatments

Four different wordings of a PAW were tested across the three choice tasks to see whether
they have a different effect. When looking at the choice behaviour of the affected consumers,
there were only slight variations visible that are not statistically significant (see table below).

Table 53 PAW choice behaviour vs. treatments

Food split: PAW May contain Cannot

Allergy affected EU8 – Total May contain traces of guarantee Made in a
Choice Task 1 affected nuts nuts nut free facility …
Base unweighted 596 149 143 147 157
Product A 28% 27% 26% 27% 31%
(local with PAW)

No difference
33% 31% 38% 31% 31%
to me
Product B 40% 42% 36% 42% 39%
(foreign – no PAW)

Total 100% 100% 100% 100% 100%

Food split: PAW May contain Cannot

Allergy affected EU8 – Total May contain traces of guarantee Made in a
Choice Task 2 affected nuts nuts nut free facility …
Base unweighted 596 149 143 147 157
Product C 40% 42% 38% 40% 41%
(local – no PAW)
No difference
51% 50% 55% 46% 51%
to me
Product D 9% 8% 7% 14% 8%
(local with PAW)

Total 100% 100% 100% 100% 100%

Food split: PAW May contain Cannot
Allergy affected EU8 – Total May contain traces of guarantee Made in a
Choice Task 2 affected nuts nuts nut free facility …
Base unweighted 596 149 143 147 157
Product E 7% 5% 6% 10% 6%
(foreign with PAW)

No difference
52% 52% 54% 53% 51%
to me
Product F 41% 43% 41% 38% 44%
(foreign – no PAW)

Total 100% 100% 100% 100% 100%

PAW AVOIDANCE
40.5% 42.3% 38.3% 40.0% 41.3%
SCORE (all tasks)

104
Impact of Food Information on Consumers’ Decision Making

The PAW avoidance score represents the average share of respondents choosing a product
without a PAW across all three choice tasks. However, it should be noted that the differences
in the overall score as well as within each task are too small to be statistically significant.
When looking at the larger sample of the unaffected consumers the PAW avoidance level is
generally higher (as outlined earlier) but the variations between the treatment splits are even
smaller.
The following table summarises the comparison of the observed choice behaviour with the
cognitive risk assessment (explained in Chapter 5.1) for each of the four tested PAW versions.

Figure 54 Cognitive risk assessment vs. observed avoidance behaviour

Cognitive risk assessment vs. observed avoidance

behaviour of allergy affected consumers
Unsafe / Unsure / Safe /
Avoidance Indifferent Ignorance
%

Cannot guarantee Cognifive: 67 14 19

nut free Observed: 40 43 17

68 9 24
May contain nuts 42 44 13

May contain traces 66 10 25

of nuts 38 49 13

Made in a facility that 61 15 24

also processes nuts 41 44 15

Cognitive risk assessment via Q40: Based on your understanding of the risk for someone with a nuts allergy, how safe would you
consider muesli to be if it had the following information on the pack?
Observed behaviour via framed choice tasks: average choice behaviour across all three tasks (choice of product with or without PAW)
Base: EU8 – all respondents with food experiments that are affected by food allergy/intolerance (n = 596)

There are three key insights to be concluded from this overview:

 Over 40% of affected consumers were not able to choose between a product with and
without a PAW, while very few consumers admitted that they were unsure in the
cognitive risk assessment.
 The cognitive ranking (with the lowest perceived risk level for ‘Made in a facility’) was
not repeated in the observed behaviour. Products with a ‘Made in a facility’ warning
were not chosen more often or avoided less often.
 These observations are typical when comparing the results of a joint evaluation mode
(risk assessment) and a single evaluation mode (observed behaviour with treatment
splits). According to behavioural theory, people tend to over-state the difference
between options when viewing them simultaneously (as in a matrix question)

105
Impact of Food Information on Consumers’ Decision Making

compared to when experiencing an option in isolation (as in treatment splits)40. Hence

it is not surprising that the cognitive risk assessment delivered a more differentiated
ranking of the PAW options than the observed behaviour.
Another question for the analysis is to consider whether or not everyone who said that they
considered a PAW to be unsafe actually took choices which reflected this and avoided the
products with a PAW in the experiment.
The table below shows the result of a cross-tabulation with the risk assessment as the base
for the sample segmentation. Among allergy affected consumers who consider a ‘Cannot
guarantee’ label as unsafe every other respondent also took a cautious approach and avoided
this product in Task 1. This is a substantially higher PAW avoidance level compared with
consumers who consider ‘Cannot guarantee’ as safe or who are unsure about its meaning.

Figure 55 Selection of product without PAW based on risk assessment of PAW labels

40
See Christopher K. Hsee and Jiao Zhang (2004): Distinction Bias: Misprediction and Mischoice Due to Joint
Evaluation http://faculty.chicagobooth.edu/christopher.hsee/vita/Papers/DistinctionBias.pdf

106
Impact of Food Information on Consumers’ Decision Making

This descriptive analysis indicates a high correlation between the risk assessment and the
choice behaviour for each PAW option. Therefore these variables were included in the
multivariate model although these assessments were collected after the experiments.
In summary, the descriptive analysis of the choice behaviour between different allergen and
origin information scenarios reveals only small and not significant variations between the four
tested treatments (the different PAW wordings).

5.3 Findings of multivariate analysis

In this section, we present the summary of results of the multivariate analysis conducted
using binary logistic regression to identify the major determinants which influenced the choice
of products without a PAW in the choice experiments. These apparently safe choices were
defined as respondents choosing product B in Task 1, C in Task 2 and F in Task 3 (each
without a PAW).
The first objective was to identify the drivers and barriers of cautious PAW avoiding choices.
The second objective paid particular attention to the impact of the four PAW labels under
evaluation in this study.
The presentation of results begins in Chapter 5.3.1 with the summary of results of the ‘base
model’, i.e. not controlling for the potential effects of the treatments, but focussing on the
main individual biases and barriers.
This is followed by Chapter 5.3.2 which adds the treatments to the base model to identify
whether and which of the PAW options are effectively influencing the choice behaviour.
The detailed overview of statistical results is provided in the appendix, while the key results
are presented in a more user-friendly chart format.

5.3.1 Main biases and barriers of PAW avoiding behaviour

The previous chapters have already discussed the potential variables included within the base
model in detail. The selection was based on conceptual considerations and in-depth analysis
of the bivariate associations between the variables as well as the dependent variable for the
behavioural experiments.
Since the analysis of the three choice tasks is based on an identical set of variables, there is
one base model for all three scenarios. The regression model for consumers affected by food
allergy or intolerance is structured along the following dimensions:
 Awareness and accessing behaviour:
o Zooming on specific food information in the experiment
 Beliefs and attitudes:
o Allergic reaction (mild, moderate, severe)
o Shopping goals (quality, health, brand, food information)
o Self-control and risk-aversion
o Risk assessment of allergen information (no information and PAW options)
 Habits:
o Recent buying experience of muesli

107
Impact of Food Information on Consumers’ Decision Making

o Reading habits when shopping for muesli (allergen information, country of

origin, ingredients list)
o Pre-shopping habits (making a list, searching for deals, always buying the same
foods)
o Frequency of reading allergy information when shopping for food
 Context: Choice tasks 1/2/3
 Socio-demographics:
o Age, gender, income, education, household composition, countries

The socio-demographic variables were selected according to their expected and/or actual
impact on choosing products without a PAW in the choice experiments.
The base model is statistically significant and explains a reasonable portion of the variation in
the dependent variable41.

Base model for all choices in Task 1, 2 and 3

All choice tasks required a decision between a product with a PAW and one without. It was
also possible to avoid any choice (“no difference to me”).
The most relevant factors driving a cautious choice (i.e. selecting the product without a PAW)
across all three tasks were mainly awareness variables and beliefs about risk levels associated
with certain PAW terms as well as some socio-demographics.

41
Binary logistical regression models are typically evaluated by Pseudo R2 to assess the quality of the model
by its explanatory power. This base model delivers an R2 of 70% defined according to Nagelkerke.

108
Impact of Food Information on Consumers’ Decision Making

Figure 56 Drivers of PAW avoiding choices without treatments

Drivers of cautious PAW avoiding choices R2:

Base model without policy options for all 3 choice tasks 70
%

Affected respondents
Reg. Awareness Beliefs & Socio-
coeff. Attitudes demogr.

4,0
Zoom ingredients

is considered unsafe

is considered unsafe
No PAW

“Cannot guarantee”
Zoom country of origin
Zoom nutrition label

Younger age (<34)

Country bias
3,0

***
ES
2,0
**
FR
RO
1,0 PL

*
IT
0,0 *** *** * *** ** *

Extract of results based on logistic regression analysis to identify statistically significant drivers of product choices w ithout PAWs
in all tasks (with and without PAW). The model explains 70% of the variance. Only variables with at least slight significance
(* = p < 0.05) are displayed. Reference category for country variables is Germany.
Base: EU8 only consumers affected by food allergy (without missing variables) n = 586

Awareness
By far the strongest influence on choosing products without a PAW was accessing the relevant
food information that displayed the PAW. Zooming on the ingredients list of both products
with and without the warning was a precondition for identifying the difference. If someone
did not pay attention to this area, then a cautious choice was less likely.
Zooming on the irrelevant nutrition label clearly distracted consumers from accessing the
more relevant area (ingredients and allergen information). The strength of this barrier on
making cautious choices might have been influenced to some extent by the previous
experiments on trans fats, where the nutrition facts were relevant for identifying differences
in food information. However, it is not an unrealistic scenario that shoppers get confused
looking for different food information on different types of product. The structured format of
the nutrition label seemed to attract far more attention than the ingredients list in both food
experiments.
The country of origin of the ingredients differed within and across the choice tasks. Consumers
who zoomed on this label were also less likely to make a cautious choice and instead showed
a preference for local ingredients. The attention given to the origin information was actually
only significant in the first choice task, where a local product with a PAW was offered with the
alternative of an imported product without a PAW. This result verifies the general assumption

109
Impact of Food Information on Consumers’ Decision Making

for products like muesli that local ingredients are trusted more (even if they display a PAW)
than imported ingredients.

Beliefs and attitudes

Among all beliefs and attitude variables, the only relevant factors that influenced a cautious
choice of allergy affected consumers was the individual risk assessment of:
 The lack of any advisory allergen information, i.e. if no information was considered
unsafe, then the product with a PAW was selected or it was stated that there was no
difference between these products.
 ‘Cannot guarantee’, i.e. if this wording was considered unsafe, then the product
without a PAW was preferred.
The risk assessments of all other PAW alternatives had no statistically significant effect.

Habits and Context

None of the tested habitual or contextual variables had any statistically relevant impact on
the choices made by allergen affected consumers. The relevance observed in the descriptive
analysis was neutralised in particular by the variables of the risk assessment.

Socio-demographics
Choosing to avoid the product with a PAW was also strongly influenced by cultural norms and
to a lesser extent by age.
 Younger consumers were more likely to make cautious choices in comparison with
middle-aged consumers.
 Consumers from Mediterranean countries as well as Eastern European countries were
less likely to consider the PAW in their product choice. The base for measuring the
country bias is Germany, which displayed an average choice behaviour across all tasks.
Thus, the countries with a higher likelihood of choosing the products without a PAW
were Finland and UK albeit not significantly higher than Germany.

5.3.2 Effectiveness of tested PAW options

To answer the final key question in this research on the most influential PAW alternative the
tested treatments were added to the base model:
 Split 1: May contain nuts
 Split 2: May contain traces of nuts
 Split 3: Cannot guarantee nut free
 Split 4: Made in a facility that also processes nuts

Since none of the treatment versions were suitable for a control group and none stood out
clearly enough in the descriptive analysis, the final model was selected after testing all
possible options:

110
Impact of Food Information on Consumers’ Decision Making

 When using split 1 ‘May contain’ as a control group, then ‘Cannot guarantee’ and ‘Made
in a facility’ becomes slightly significant in driving less cautious choices.
 When using split 2 ‘May contain traces’ as a control group, then no other PAW version
displays any significant impact.
 When using split 3 ‘Cannot guarantee’ or 4 ‘Made in a facility’ as a control group, then
‘May contain’ becomes slightly significant in driving more cautious choices.

As a result of running the regression model with these rotating control groups, the following
ranking of the PAW options can be established based on their impact on the selection of
products without a PAW in the choice experiments.

Figure 57 Impact of PAW wording on cautious choices

Impact of PAW wording on cautious choices

Affected respondents – Significance levels

Cautious choices Less cautious choices

Avoiding PAWs Not avoiding PAWs
B: +477 0 - 118 - 477

n.s. May contain n.s. Cannot

traces guarantee
May
contain Made in a
* *
facility …

Extract of results based on logistic regression analysis with policy options to identify statistically significant drivers of
cautious product choices in all three tasks. Base: EU8 – all respondents with food experiments that are affected by food
allergy/intolerance (n = 596)

Split 1 with the wording ‘May contain’ seems to be the most suitable as a control group as all
of the other PAW options have the tendency to create less cautious choices.

The following table presents the same structure of variables as used in the base model
described earlier but with the treatments added. The explanatory power of the enhanced
model does not increase, which indicates that the overall impact of the PAW wording is fairly
low compared to the other variables.

111
Impact of Food Information on Consumers’ Decision Making

Figure 58 Drivers of PAW avoiding choices with PAW options

Drivers of cautious PAW avoiding choices R2:

Base model with policy options for all 3 choice tasks 70
%

Affected respondents
Reg.
Policy/PAW options Awareness Beliefs & Socio-
coeff. Attitudes demographics
Zooming
(ref.cat.)

is considered unsafe

is considered unsafe
Cannot guarantee

Made in a facility

No PAW

“Cannot guarantee”
May contain traces

that also processes

Younger age (<34)

4,0
May contain

Country bias
Ingredients

Country of origin
Nutrition label
3,0

***
ES
2,0
**
FR
RO
1,0 PL

*
IT
n.s.
0,0 * * * *** *** * *** ** *

Extract of results based on logistic regression analysis to identify statistically significant drivers of product choices w ithout PAWs in all tasks
(with and without PAW). The model explains 70% of the variance. Only variables with at least slight significance (* = p < 0.05) are displayed.
Reference category for policy options is split 1 (may contain) and for country variables it is Germany.
Base: EU8 only consumers affected by food allergy (without missing variables) n = 586

The inclusion of the PAW options in the base model does not change the impact of any of the
previously described factors. Looking at the ingredient information with the PAW and
individual perceptions of the risk level of a PAW remained the key drivers of a cautious choice.

PAW options
The term ‘May contain’ is the only PAW wording which had any notable impact on people
making more cautious choices. All other wordings either made people less likely to choose
products without a PAW (‘Cannot guarantee’ and ‘Made in a facility…) or had negligible
influence on behaviour (‘May contain traces of’).
When comparing the choice behaviour with the cognitive risk assessment, then three out of
the four PAW options displayed the same risk ranking:
 ‘May contain’ is the only wording that contributed to more cautious choices and also
had the highest cognitive risk level.
 ‘May contain traces’ had a medium ranking in both evaluation formats without any
significant differentiation compared with the other options.

112
Impact of Food Information on Consumers’ Decision Making

 ‘Made in a facility that also processes’ contributed significantly less to cautious choices
in comparison to ‘May contain’ and also had the lowest cognitive risk level.
There was one evaluation that stood out in this comparison: ‘Cannot guarantee’ received the
highest cognitive risk score (together with ‘May contain’) in the post experimental evaluation,
where it was framed in a matrix question together with alternative allergen information.
Hence, one would have expected to see this PAW wording having a similar impact on choice
behaviour as ‘May contain’ whereas in fact it had a statistically relevant impact on driving less
cautious choices.
This surprising result is also confirmed by the fact that conversely ‘Cannot guarantee’ was the
only PAW where a low risk assessment increased the likelihood of cautious choices.
In order to find an explanation for this paradoxical evaluation, the previous model was
enhanced with the interaction effects between the cognitive risk assessment and the
treatment splits for each of the PAW options.
This analysis reveals that there was indeed a slightly significant interaction for the label
‘Cannot guarantee’ (see table below).

113
Impact of Food Information on Consumers’ Decision Making

Figure 59 Drivers of PAW avoiding choices with PAW options and interaction effects

Drivers of cautious PAW avoiding choices R2:

Base model with policy options for all 3 choice tasks 70
%

Affected respondents / incl. interaction effects

Reg.
Awareness
Beliefs & Socio-
Habits demogr.
coeff.
Policy/PAW options Attitudes
Zooming
Interaction effect
Cannot guarantee

is considered unsafe
‘Cannot guarantee’
of risk assessment for

Country bias
food information
Made in a facility

General interest in reading

No PAW
May contain traces

that also processes

4,0
Ingredients

Country of origin
Nutrition label
3,0

***
ES
2,0
**
FR
RO
1,0 PL

*
IT

0,0 n.s. * ** * *** *** * *** *

Extract of results based on logistic regression analysis to identify statistically significant drivers of product choices w ithout PAWs in all tasks
(with and without PAW). The model explains 70% of the variance. Only variables with at least slight significance (* = p < 0.05) are displayed.
Reference category for policy options is split 1 (may contain) and for country variables it is Germany.
Base: EU8 only consumers affected by food allergy (without missing variables) n = 586

For the interpretation of this result it is important to remember that the participants conducted
the cognitive risk assessment after the experiment. It is apparent that unlike the other
treatments, the priming with the treatment ‘Cannot guarantee’ had a significant and strong
influence on
a) Reducing the likelihood of cautious choices in a single evaluation scenario and at the
same time
b) Increasing the cognitive risk level associated with the label in a joint evaluation mode
with other labels.
This divergent impact is more understandable if one visualises the different framings of a PAW
that has a negative wording as compared to a positive wording:
 ‘Cannot guarantee’ vs. ‘no information’ are both negative / similar and therefore lead
to less cautious choices as compared to ‘May contain’ vs. ‘no information’.
 If ‘Cannot guarantee’ is evaluated simultaneously with four positive wordings, then the
reference framing was extended into the opposite direction (linguistically), which lead
to the more risky evaluation.

114
Impact of Food Information on Consumers’ Decision Making

The key factors which increase cautious (PAW avoiding) choices for allergy affected consumers
included:

 Zooming on relevant ingredients label including a PAW and ignoring other food
information,
 PAWs with the wording ‘May contain’,
 General interest in reading food information when shopping.

The elements which reduced the likelihood of making cautious choices were:
 Considering the lack of a PAW as unsafe with regards to any allergic reaction,
 PAWs with the wording ‘Cannot guarantee’ or ‘Made in a facility that also processes’,
 Preference for local ingredients, but only when considering a local product with a PAW
and a foreign product without a PAW.
 Cultural biases – in particular allergy affected consumers in Spain, France, Romania,
Poland and Italy were less likely to make PAW avoiding choices. They were less likely
to access the ingredients list and also assigned lower risk levels to the PAWs in the
cognitive assessment.

5.4 Conclusions regarding PAWs

Do consumers take a cautious approach? What are the main individual

biases and elements restricting consumer choice?

Around 40% of allergy affected consumers made a cautious choice in each of the three choice
tasks. To summarise the findings and conclusions of the experiment, the graph below can
serve again as a conceptual guideline. It represents a simplified and idealised illustration of a
comprehensive decision process. It covers the typical stages of a shoppers’ decision process
from accessing and assessing relevant food information to taking the final decision.

The identified individual elements as well as the tested PAW options are marked with symbols
to indicate their direction and effectiveness of impact.

115
Impact of Food Information on Consumers’ Decision Making

Figure 60 Summary of results on PAW analysis


Cautious consumer choice (i.e. avoiding products with a PAW) is largely
driven by a high issue salience (accessing relevant allergen information)
as well as by the belief that a product without any PAW is a safer choice.

Within the restricted scope of variables that it was possible to test in this multi-topic online
survey, the following key findings were identified as significant biases and elements explaining
why consumers make a cautious choice:
 The key driver of a cautious choice was zooming on the ingredients list with or
without a PAW.
 Zooming on the nutrition label was a key barrier as it distracted attention from
the PAW label. As in the other food experiment the nutrition label was accessed
much more frequently than the ingredients list despite the task of shopping for
someone who is allergic to a specific ingredient.
 Zooming on the country of origin presented a barrier only when the choice was
between a local product with a PAW and an imported product without a PAW. In this

116
Impact of Food Information on Consumers’ Decision Making

context consumers with a strong preference for local products were less likely to
make a cautious choice.
 Consumers who believed that no information given about nuts on a product was
also unsafe were less likely to avoid choosing products with a PAW.
 Finally, a cultural/country bias was observed. Consumers in Mediterranean and
Eastern European countries were less likely to restrict their choices because of a
PAW.

What level of risk is associated by consumers with different types of PAW

wordings as well as with the absence of a PAW?

To identify the individual risk levels that consumers assigned to various versions of allergen
information a cognitive risk assessment exercise was conducted after the experiments. The
consumer understanding was collected by a matrix question presenting all statements
simultaneously with a randomised order and a 5-point verbal scale, which was transferred
into a mean risk score for the analysis.

The lack of any advisory allergen information was considered as less risky

 than most PAW options by the average affected consumer.

Only the wording ‘Made in a facility’ was rated at a similar lower risk level
as if no information is given.

Around two thirds of the allergy affected consumers considered any of the presented PAWs
as definitely or probably unsafe. The risk level was substantially higher for this affected
segment compared to consumers without personal experience of food allergies or
intolerances.
All PAW options were rated clearly as less risky than the label ‘contains nuts’ and also as more
risky than ‘no information’. The mean risk scores measured in this cognitive and joint
evaluation mode by affected consumers were42:
a) Contains nuts 4.0 ** (highest risk level)
b) May contain nuts 3.8 (not significant to c, d)
c) Cannot guarantee nut free 3.8 (not significant to b, d)
d) May contain traces of nuts 3.7 (not significant to b, c)
e) Made in a facility that also processes nuts 3.6 * (not significant to f)
f) No information given about nuts 3.5 ** (lowest risk level)

42
Based on the following scale: 5 = definitely unsafe, 4 = probably unsafe, 3 = not sure, 2 = probably safe, 1
= definitely safe

117
Impact of Food Information on Consumers’ Decision Making

It should be noted, that the slightly higher risk level of ‘Cannot guarantee’ was also driven by
the previous treatment in the experiments.

What is the impact of the tested PAW labels on making cautious choices?

To identify the most effective PAW in driving cautious choices (avoiding the product labelled
with a PAW), the analysis looked at the impact of each tested wording across all choice
tasks.

Overall, the impact of the PAW wording was only small.

 Only ‘May contain’ was effective in facilitating more cautious choices among
allergy affected consumers.
‘Cannot guarantee’ was more effective at driving less cautious choices.

As indicated before, the ranking of the PAW options based on the cognitive risk assessment
only partly matched the ranking based on the impact of these options in the observed
behaviour43:

Based on model with(out) interaction effects: Without With

a) May contain nuts +.477 * (control group)
b) May contain traces of nuts - .118 n. s. - .187 n. s.
c) Made in a facility that also processes nuts - .477 * - .692 *
d) Cannot guarantee nut free - .477 * -1.190 **

The key difference was the divergent positioning of the label ‘Cannot guarantee’, which
contributed to less cautious choices in the experiment with isolated treatments, but was rated
high risk in the cognitive assessment in a simultaneous evaluation framing.
When controlling for this interaction effect of the ‘Cannot guarantee’ treatment, then its
negative impact on cautious choices increases significantly. Hence, this label represents the
PAW option that is most efficient in driving less cautious choices, while the label ‘May contain’
is most efficient in driving more cautious choices.

43
Based on Beta values of the binary regression analysis which indicate the direction (+/-) and the strength of
the influencing variable on the goal variable (cautious choice)

118
Impact of Food Information on Consumers’ Decision Making

How can the choices be made easier for allergy affected consumers?

 Making allergen information clearer in terms of the risk level the product
represents will help to avoid misunderstandings by consumers.

The main policy goal is to identify whether allergy affected consumers have restricted choice
due to misunderstanding the risk represented by PAWs. Dealing with this issue requires an
understanding of the process of individual risk assessment and the level of trust in food
information which the study shows can vary depending on the framing that is presented to
consumers.
Cautious choices are more likely to be made by consumers who consider a product without a
PAW as safer compared to a product with a PAW. This is a significant contributor to a restricted
product choice.
If there was more clarity that there was no real difference in risk between a product with or
without a PAW, then fewer consumers would restrict their choices. Alternatively, if there were
common rules when to apply a PAW, in terms of the risk they represent together with a
standardised wording and format to be used by all manufacturers then this also may help
consumers make better informed choices.

119
Impact of Food Information on Consumers’ Decision Making

6 EXPERIMENTAL ‘LAB’ FINDINGS ON ALCOHOL CONSUMPTION

This section explains the key findings from the experiment on alcohol consumption to meet
the following research objectives:
a) What are the main individual biases and external barriers that prevent consumers´
attitude drinking less?
b) What are the most effective treatments – i.e. the most appropriate information or
health warning – to drive a positive attitude change?
To contextualise the findings of the research and to structure the analysis towards the
ultimate objective of understanding the drivers and barriers of an attitude change process,
we put forward a map of a simplified decision process (as explained in Chapter 4).

Figure 61 Guideline for analysis of results on alcohol consumption

120
Impact of Food Information on Consumers’ Decision Making

Unlike the previous food shopping experiments the focus of this research concentrated on
stated intentions (to drink less) rather than implemented behaviour44. As explained in chapter
3.5, a behavioural experiment that allowed observing actual drinking behaviour was not
possible within an online survey. Moreover, there was no zooming functionality implemented
to simulate the conscious awareness required for assessing new information. The insights
from the food experiments about the relevance of reading new information are likely to be
applicable also to the decision process of the alcohol consumption.
The main focus in this policy area was to understand the key drivers that influence the
intention to drink less, which required that the participants were exposed to specific
information (treatments).
The analysis presented here starts with a description of survey findings of the observed
intention change in the choice tasks with suggestions for variables to be included in the
statistical base model. The bivariate examination of potentially influencing variables is more
or less structured along the graphical overview. This is supplemented by a descriptive
overview of observed change of drinking intentions for the tested treatments.
Chapter 6.2 combines the discussed variables within a multivariate analysis to answer the
key questions about what are the statistically significant barriers and motivators for the
intention to drink less on a specific occasion as well as at any occasion in future.

6.1 Observed choice behaviour

The first choice task required the decision whether one would drink less, the same or more
alcohol at the specified occasion (at a friends´ party) in comparison to the volume that they
said they would normally drink before the treatment (with information on calories or limits).
The second choice task requested a decision with the same answer options45 but now geared
towards any typical drinking occasion in the future.

44
According to the more detailed stages of change model by DiClemente and Prochaska (1998) the analysis
above covers only the first two stages of pre-contemplation (i.e. not yet considering change) and
contemplation (considering change but not taking action). This excludes the stages of preparation (concrete
planning to change) and action (making changes in one’s behaviour).
45
The answer options are: 1: I would drink less 2: I would drink about the same 3: I would drink more and 8:
Don’t know. There were two differences between the tasks. In choice 1 the answer option 2 was
supplemented by the concrete volume stated before the treatment as a reminder. This was not required in
Task 2. The option “Don’t know” was not shown from the beginning in Task 1 (only if the respondent tried to
skip the decision). This explains the lower DK rate in Task 1.

121
Impact of Food Information on Consumers’ Decision Making

Table 62 Observed choice behaviour in alcohol tasks 1 and 2

Alcohol split: Choice Task 1 Choice Task 2

EU8 – Total specific occasion any occasion in future
Base unweighted 4 040 4 040

Less 17.4% 17.4%

Same 79.9% 76.3%

More 2.6% 2.9%

Don‘t know 0.1% 3.4%

Total 100% 100%

Both decision scenarios received the same share of participants with the intention to drink
less (17.4%). This observation suggests that those who answered with “less” in Task 1 are
identical with those who said “less” in Task 2. This assumption is true for two thirds of the
participants as can be seen in the table below.

Table 63 Intention to drink less on specific occasion vs. any occasion in future

Task 1:
intention to drink less on specific occasion
Alcohol split:
Task 1 vs. Task 2
EU8 –
EU8 – Total
Total Less Same More DK
Base unweighted 4 040 700 3231 106 3
Task 2: intention to

Less 17.4% 65.1% 7.5% 3.8% 33.8%

occasion in future
drink less at any

Same 76.3% 30.7% 87.4% 43.7% 32.4%

More 2.9% 1.1% 1.9% 44.1% -

Don‘t know 3.4% 3.1% 3.2% 8.4% 33.9%

Total 100% 100% 100% 100% 100%

Apparently the intention to drink less – based on the new information in the treatments – in
the specific scenario of Task 1 is already a good predictor of the general intention to drink
less in future. However, around a third of consumers who would drink less at the party would
not commit to drinking less at any occasion in future. Likewise, 65% of consumers who would
drink less in future did not indicate the same intention for the specified occasion.

122
Impact of Food Information on Consumers’ Decision Making

Both tasks were intended to measure attitude change in two different contexts or framings,
which obviously was understood by the participants. Therefore, it makes sense to look at the
potential drivers for each choice scenario in parallel.
Another differentiation of the choice scenario was applied by using three categories of
alcoholic drinks. The participants in the alcohol split were divided into three groups and
presented an experiment either with beer, wine or spirits (whiskey or vodka). The alcohol
type was randomly assigned among the respondents who drank this category within the last
12 months.
There is a slightly higher variance in the answer behaviour in Task 1 as compared with Task
2. Participants in the wine experiment seemed to be more likely to commit to drink less in
both scenarios, while those in the spirits experiment were somewhat less likely to reduce their
drinking volume on that specific occasion.

Table 64 Observed choice behaviour in Task 1 vs. alcohol category

Alcohol split: EU8 – Total

Choice Task 1 alcohol split Beer Wine Spirits
Base unweighted 4 040 1351 1342 1347

Less 17.4% 16.3% 20.2% 15.6%

Same 79.9% 81.3% 78.6% 79.9%

More 2.6% 2.3% 1.1% 4.4%

Don‘t know 0.1% 0.1% 0.1% 0.1%

Total 100% 100% 100% 100%

Table 65 Observed choice behaviour in Task 2 vs. alcohol category

Alcohol split: EU8 – Total

Choice Task 2 alcohol split Beer Wine Spirits
Base unweighted 4 040 1351 1342 1347

Less 17.4% 16.4% 19.6% 16.2%

Same 76.3% 78.3% 74.9% 75.8%

More 2.9% 2.4% 1.8% 4.4%

Don‘t know 3.4% 2.9% 3.7% 3.5%

Total 100% 100% 100% 100%

The rationale for this answer behaviour may become clearer through the analysis in the
following chapters.

123
Impact of Food Information on Consumers’ Decision Making

6.1.1 Potential beliefs and attitude drivers

Individual beliefs and attitudes about alcohol, choice preferences, the relevance of health
goals, as well as the drinking habits of the relevant peer group can all be assumed to be
potential drivers of alcohol consumption. Data on these issues were collected both before and
after the experiments.
The analysis in the table below indicates a strong relevance of social norms or peer group
behaviour for the intention to drink less. Consumers who strongly disagreed with the
statement that “alcohol is commonly drunk among the people they spend most of their time
with” were far more likely to drink less on the specific occasion or in general in future. This
refers to the hypothesis that most alcohol consumers are social drinkers, who orientate
themselves towards their personal network (peer group, e.g. friends, family, colleagues).
At the other end of the scale, consumers who spend their time with people who commonly
drink alcohol were far less likely to declare that they would reduce their drinking behaviour.

Figure 66 Intention to drink less vs. image and peer group behaviour

Intention to drink less alcohol in task 1 and 2

Based on alcohol image and peer group behaviour
 =17%
Q72.1 Peer group behaviour:
Drinking alcohol is common among the people I spend most of my time with I would drink less …

Strongly disagree 31% in task 2 (in future)

(n=490) 31% in task 1 (on occasion)

Disagree 19%
(n=1521) 19%

Agree 14%
(n=1673) 13%

Strongly agree 9%
(n=355) 12%

Q72.2 Alcohol image:

Overall, alcohol consumption has more advantages than disadvantages for our society I would drink less …

Strongly disagree 26% in task 2 (in future)

(n=1161) 24% in task 1 (on occasion)

Disagree 14%
(n=2060) 14%

Agree 13%
(n=671) 15%

Strongly agree 18%

(n=148) 18%

Base: EU8 – respondents in the alcohol experiment with the respective attitude in Q72

A similar correlation with a lower significance can be observed between the intention to drink
less and the overall image of alcohol. Again, respondents who saw more disadvantages for
society (and thereby had a more negative image) were more likely to say they would drink
less alcohol.

124
Impact of Food Information on Consumers’ Decision Making

Hence, both variables (peer group behaviour and alcohol image) will be included in the final
analytical model.
Another interesting set of variables are the different choice preferences when choosing an
alcoholic beverage. Whether the focus is on price, taste, alcoholic content or serving size, etc.
these individual preferences may also represent a barrier or a motivation to reduce alcohol
consumption.
The descriptive analysis in the table below demonstrates that there are three potential
motivators that are more likely to foster a healthy attitude change:
 Low alcoholic content
 Fewer calories
 Small serving size

Figure 67 Intention to drink less vs. choice preferences

Intention to drink less alcohol in task 1 and 2

Based on choice preferences for alcoholic beverages
I would drink less …
 =17%
Low alcoholic content 26% in task 2 (in future)
(n=630) 29% in task 1 (on occasion)

Fewer calories 25%

(n=304) 27%

Small serving size 25%

(n=351) 26%

Recommendation 18%
(n=646) 20%

Popular drink 19%

(n=991) 18%

Brand name 19%

(n=1725) 18%

High alcoholic content 17%

(n=257) 18%

Suitable for meal 17%

(n=2274) 17%

Low price/value 17%

(n=1043) 16%

Taste 17%
(n=2903) 16%

Large serving size 14%

(n=201) 15%

High price/value 16%

(n=259) 15%

Q45 When c hoosing alc oholic beverages, which aspects do you usually take into account?
Base: EU8 – respondents in the alc ohol experiment with the respective choice preferenc es

The final model will include these three variables as well as their complementary attributes
(high alcoholic content, large serving size) since they can logically be expected to also have
an influencing role in the experiment. All other preferences were skipped after confirming
their lack of relevance in the multivariate model.
Another motivator for drinking less alcohol can be the general relevance of health (vs. other
priorities). Indeed, shoppers who agreed that they always try to choose the healthier

125
Impact of Food Information on Consumers’ Decision Making

alternative when selecting a product were more likely to reduce their drinking volume (19%)
than those who disagreed (11/12%).
Although this variable originally related to goals when shopping for food it seems to also
influence attitude change about alcohol consumption. Hence, we also looked into the
motivation types and realised that these motivational mind-sets also help to explain attitude
change in this experiment. Three of the six types stand out in the table below.

Figure 68 Intention to drink less vs. motivation type

The Ambitious Health Planner displayed the highest share of reduced alcohol consumption
in both decision scenarios. When looking at the typical drinking patterns of this segment, this
is easily explained as their peers do not commonly drink either, their drinking volume is the
lowest of all types as well as their experience with consuming too much alcohol. This type
selects drinks most often by low alcoholic content, small serving size and fewer calories. On
the other hand the Health Planner has the lowest self-control across all types, which may
indicate that the goal to drink less in the future might be inflated by an overconfidence bias.
At the other end of the ranking the Frustrated and the Self-Determined Consumer were
the types with the lowest positive attitude change. Drinking alcohol was most common among
their social network for both types and also the experience of having had too much alcohol
was most often stated amongst these two segments.
The Frustrated Shopper has neither enough money, nor time for shopping nor interest in
health issues. This is the youngest consumer segment with an above average alcohol
consumption, low risk aversion and average self-control who only watch out for a low price
when selecting a drink.
The Self-Determined Shopper has enough monetary resources and lacks interest in health
issues. He (there are less women in this group) displayed the highest average drinking
volumes across all types, is most risk-taking with the highest self-control. His main selection
criterion for alcohol is taste.

126
Impact of Food Information on Consumers’ Decision Making

As already indicated in the profile of the segments above, there are two further attitude
dimensions, which may also directly influence the willingness to drink less alcohol: self-control
and risk attitude.
The answer pattern observed between self-control and the intention to drink less indicates
that respondents with a very low self-control were more likely to intend to drink less. However
it is likely to assume that these positive intentions are less likely to be implemented compared
with respondents with higher self-control. In other words: two thirds of consumers who intend
to drink less were also characterised by a low self-control.
Consumers who value security more than excitement and adventure were more aware of the
disadvantages of drinking alcohol. Therefore, makes sense that security orientation was also
an influence on the intention to reduce unhealthy alcohol consumption as indicated in the
table below.

Figure 69 Intention to drink less vs. self-control and risk attitude

Intention to drink less alcohol in task 1 and 2

Based on general self-control and attitude towards risk
 =17%
Q82.1 Self control:
I much prefer doing things that pay off right away than in the future I would drink less …
17%
(Strongly) Disagree in task 2 (in future)
(n=1430) 16% in task 1 (on occasion)

16%
Agree
(n=2045) 17%

23%
Strongly agree
(n=565) 25%

Q82.2 Risk attitude:

Security is more important to me than excitement and adventure I would drink less …

in task 2 (in future)

(Strongly) Disagree 13%
in task 1 (on occasion)
(n=707) 13%

Agree 16%
(n=2154) 15%

Strongly agree 23%

(n=1179) 24%

Base: EU8 – respondents in the alcohol experiment with the respective attitude in Q82

In summary the following attitudinal variables were identified for the final analytical model:
 Alcohol image and drinking habit of peer group (social norms)
o (Dis-)Agreement with the statement “Drinking alcohol is common among the
people I spend most of my time with”
o (Dis-)Agreement with the statement “Overall, alcohol consumption has more
advantages than disadvantages for our society”
 Choice preferences when choosing alcoholic beverages (only relevant criteria)
o Low and high alcoholic content
o Small and large serving size

127
Impact of Food Information on Consumers’ Decision Making

o Fewer calories
 Motivation types
o health goals and also money and time resources
 Interest in reading food information
o (Dis-)Agreement with the statement “I read labels and pay close attention to
product information”
 Risk aversion and self-control
o (Dis-)Agreement with the statement “I much prefer doing things that pay off
right away than in the future”
o (Dis-)Agreement with the statement “Security is more important to me than
excitement and adventure”

6.1.2 Potential habitual drivers

Questions in the pre-experimental survey asked about individual habits when drinking specific
types of alcohol. An initial descriptive analysis indicates that the following habitual barriers
seem to correlate with the intention of drinking less in both tasks:
 High frequency of drinking, i.e. more than once a week (13.6%/15.5%)
 High drinking volume on the most recent occasion, i.e. 7 or more glasses
(9.5%/11.6%)
 High drinking volume in experiment situation before treatment, i.e. 7 or more glasses
(9.2%/12.4%)
 Having drunk a wide variety of alcoholic drinks within the past 12 months, i.e. 4 or 5
categories (14.9%/14.6%)
On the other hand, a positive attitude change was more often observed amongst consumers
 Who follow a weight loss programme (23.8%/24.2%)
 Without any experience of consuming too much alcohol (21.7%/20.5%)
All these variables above form part of the analytical model.
In addition the analysis also includes familiarity with the experimental situation of ‘a party at
a friend’s home’ for drinking alcohol. Only around half of participants were familiar with
drinking the specific alcohol category at such an occasion. Although the bivariate analysis did
not indicate any correlation with the intention to drink less, it is a logical control variable.

6.1.3 Socio-demographic variables

The same set of socio-demographic variables used in the previous experiments has also been
included in this analysis.
The only difference is the change of the reference country. Spain represented the European
average in both choice tasks.

128
Impact of Food Information on Consumers’ Decision Making

6.1.4 Treatments

Since the attitude change towards drinking less alcohol had to be measured directly by stated
preference the effectiveness of treatments could only be compared directly. The table below
shows a slightly higher commitment to drink less after the treatment “Know your limits”
compared with the treatment with calorific information. This deviation between both
treatments is only slightly significant in Task 1.

Table 70 Choice behaviour in Task 1 vs. treatments

Alcohol split: EU8 – Total Calories Limits

Choice Task 1 alcohol split treatment treatment
Base unweighted 4 040 2031 2009

Less 17.4% 16.0% 18.8%

Same 79.9% 82.0% 77.8%

More 2.6% 1.9% 3.4%

Don‘t know 0.1% 0.1% -

Total 100% 100% 100%

Table 71 Choice behaviour in Task 2 vs. treatments

Alcohol split: EU8 – Total Calories Limits

Choice Task 2 alcohol split treatment treatment
Base unweighted 4 040 2031 2009

Less 17.4% 16.8% 18.1%

Same 76.3% 77.2% 75.5%

More 2.9% 2.6% 3.2%

Don‘t know 3.4% 3.4% 3.3%

Total 100% 100% 100%

The next step was to look at the more detailed treatment splits, i.e. whether the limits or
calorific treatment achieved different levels of attitude change in different alcohol categories.
The table below presents only very slight deviations for the calorific treatments. Participants
in the experiments with a treatment about calories for wine responded more often with a
positive attitude change compared with participants in experiments for beer and spirits.

Figure 72 Intention to drink less vs. alcohol category and treatments

129
Impact of Food Information on Consumers’ Decision Making

Intention to drink less alcohol in task 1 and 2

Based on treatment policy and alcohol category
I would drink less …
 =17%
Policy treatment ‘Limits’: in task 2 (in future)
Know your limits information/education in task 1 (on occasion)

Beer 19%
20%
(n=671)
Wine 18%
(n=668) 19%

Spirits 17%
(n=670) 18%

Policy treatment ‘Calories’:

Calorific information/education
Beer 14%
(n=680) 13%

Wine 21%
(n=674) 21%

Spirits 15%
(n=677) 14%

Base: EU8 – respondents in the alcohol experiment with the respec tive c ontext and treatment split

There are three hypotheses that could potentially bias and explain why the calories treatment
on wine had a higher impact:
A. Framing effects of the education treatment
B. Information was more surprising for wine drinkers
C. Wine drinkers are more susceptible to calories treatment

Hypothesis A: Framing effects of the education treatment

While the education (knowledge test) on ‘limits’ was identical with the middle option being
the correct answer across all types of drinks, the education on ‘calories’ varied as follows:
 Calories for beer were presented with a higher reference level, i.e. pizza had more.
 Calories for wine were presented with a lower reference level, i.e. chocolate had less.
 Calories for whiskey and vodka were presented with an equal reference level, i.e. nuts
had about the same.
Hence, the calorific education of wine had a more negative framing “Wine has more calories
than chocolate” than the other drinks. Negativity is a very common and strong bias observed
in many behavioural experiments. According to behavioural theory, people tend to weigh
negative information more heavily than positive information46. Such a negativity bias could
be one rationale for this observation.
Hypothesis B: Information was more surprising for wine drinkers

46
David E. Kanouse: Explaining negativity biases in evaluation an choice behaviour: theory and research, in
Advances in Consumer Research Volume 11, 1984

130
Impact of Food Information on Consumers’ Decision Making

Another potential explanation could be the effect of the information being unexpected, which
tends to make an education treatment more effective47. Participants who were told that their
initial choice was wrong might have paid more attention to the treatment information, which
was a required condition (although not sufficient on its own) for having any impact at all on
attitude change.
The analysis of the feedback results of the knowledge test in the table below shows a generally
higher DK rate for the ‘limits’ treatment than for the ‘calories’ treatment. Obviously the
respondents found it more difficult to answer the test on ‘know your limits’, which on average
had significantly fewer correct answers (18%) than the calories test (28%).

Figure 73 Knowledge test results vs. alcohol category and treatments

Knowledge test results

Based on treatment policy and alcohol category

Policy treatment ‘Limits’:

Know your limits information/education Correct Not correct Don’t know
Beer 16% 56% 28%
(n=671)
Wine 18% 59% 23%
(n=668)
Spirits 21% 52% 27%
(n=670)
Policy treatment ‘Calories’:
Calorific information/education Correct Not correct Don’t know
Beer 46% 40% 14%
(n=680)
Wine 15% 66% 18%
(n=674)
Spirits 22% 66% 12%
(n=677)
Base: EU8 – respondents in the alcohol experiment with the respec tive c ontext and treatment split

However, the only significant differentiation in terms of a more or less surprising feedback
can be observed with the calories treatment split for beer. Almost half chose the correct
answer “beer has fewer calories than pizza”. Hence, this group was less often surprised by
the feedback (education treatment), which may explain the slightly lower willingness to drink
less beer.
On the other hand, the proportion of wine drinkers that would have been surprised by the
fact that wine has more calories than chocolate (66%) is similar to that spirits drinkers, whose
attitude change was more comparable to beer than to wine.
This descriptive analysis as well as a test of interaction effects in the multivariate analysis
shows that there is no surprise effect with impact on attitude change.

47
Daniel Kahneman 2011: Thinking fast and slow; Charles Stangor and David McMillan, 1992: Memory for
expectancy-congruent and expectancy-incongruent information: a review of the social and social
developmental literatures

131
Impact of Food Information on Consumers’ Decision Making

Hypothesis C: Wine drinkers are more susceptible to calories treatment

If wine drinkers paid more attention to calories compared with drinkers of other alcoholic
beverages, this could explain the higher proportion who would drink less wine after the
education on calories.
However, neither the descriptive analysis nor the multivariate analysis confirmed a significant
interdependency. The choice preference for ‘fewer calories’ was almost identical as was
whether they followed a weight loss programme. There was a slightly higher share of women
among wine drinkers. Again, this interdependency does not have a statistically significant
effect on the ultimate goal to drink less. Likewise the higher share of men who drank spirits
had no significant effect on the intention to drink less48

Table 74 Susceptibility towards calories treatment vs. alcohol category

Beer Wine Spirits

Alcohol split: drinkers drinkers drinkers
Base unweighted 3433 3550 2127

Diet/ Weight loss

13.5% 13.8% 12.9%
programme (habit)

Fewer calories
7.6% 7.6% 6.2%
(choice preference)

Women (gender) 53.7% 57.6% 48.5%

Hence, the context-treatment variation in the experiment was neither due to the fact of
drinking only a specific type of alcohol (e.g. wine drinkers) nor the surprise effect of the
education treatment. The only explanation left is the negative framing of the calories
treatment for wine that may have impacted the positive attitude change.
In summary, the descriptive analysis about the attitude change to drink less alcohol indicated
only a slight impact of treatments:
- ‘Limits’ treatment had a slightly higher impact than calories in the first scenario across all
alcohol types.
- ‘Calories’ treatment on wine drinkers had the relative highest impact compared to all other
alcohol types and treatments due to its negative framing.

6.2 Findings of multivariate analysis

In this section, we present the summary results of the multivariate analysis conducted, using
binary logistic regression to identify the major determinants which influenced a positive

48
This was tested by adding the combinations of drinking wine and the female variable as well as drinking
spirits and the male variable to the base model of the multivariate analysis.

132
Impact of Food Information on Consumers’ Decision Making

attitude change in the alcohol experiment. Positive attitude change is defined as respondents
claiming that they would drink less in each decision scenario.
The first objective was to identify the drivers and barriers of healthy choices. The second
objective paid particular attention to the potential effects of the treatments under evaluation
in this study.
The presentation of the results begins in Chapter 6.2.1 with the summary of results of what
we refer to as the ‘base model’, i.e. not controlling for the potential effects of treatments, but
focussing on the main individual biases and barriers.
This is followed by Chapter 6.2.2 which adds the treatments to the base model to identify the
drivers and barriers that impact positive attitude change of consumers after each treatment.
As outlined for the previous experiments, the detailed overview of statistical results will be
provided in the appendix, while the key results are presented in a more user-friendly chart
format.

6.2.1 Main biases and barriers on alcohol consumption

The previous chapters have already discussed the potential variables included within the
multivariate base model in detail. The selection was based on conceptual considerations and
in-depth analysis of the bivariate associations between the variables considered as well as the
dependent variable for the experiments.
The analysis begins with the first decision scenario (attitude change in specific occasion),
which represents a more short-term or preliminary decision which logically has a strong
influence on the second decision for the more general and long-term scenario. Therefore, the
model for the second longer term decision will cover all variables as outlined for the first
decision plus the goal variable of the first decision (intention to drink less at the party).
All models are structured in the same way covering the following four dimensions:
 Context:
o Whether the decision was about beer or wine or spirits
 Motives and goals:
o Common peer group drinking
o Positive image of alcohol (has more advantages)
o Choice preferences for drinks (low or high alcoholic content, small or large
serving size, low calories)
o Consumer motivation types (based on interest in health and resources of
money and time)
o General interest in food information
o Risk attitude and self-control
 Habits
o Following a weight loss programme
o Number of different types of alcohol consumed within past 12 months

133
Impact of Food Information on Consumers’ Decision Making

o Familiarity of the specific occasion (a friend’s party) when drinking the

respective alcohol type
o Frequency of drinking the respective alcohol type
o Volume of drinking the respective alcohol type at last occasion in reality (vol0)
o Volume of drinking the respective alcohol type on a given occasion (friends´
party) without treatment (vol1)
 Socio-demographics:
o Age, gender, income, education, household composition, body mass index,
countries

Both base models are statistically significant and explain a reasonable portion of the variation
in the dependent variable. Task 1 explains 12%, Task 2 about 42% with strong significance
levels well below 0.001.

Base model for decision scenario 1

The first task required the decision whether to change the intended amount of alcohol they
would consume as the specific occasion of a friends´ party after an education treatment.
The most relevant factors driving a positive attitude change were motives and goals, followed
by habits and to a lower extent the context of the alcohol category as well as some socio-
demographics.

134
Impact of Food Information on Consumers’ Decision Making

Figure 75 Drivers of choice behaviour in alcohol Task 1 without treatments

Drivers of intention to reduce alcohol consumption 12

R2: %

in task 1 at the specific occasion (at a friends´ party)

Base model without policy options
Reg. Con- Motives & goals Habits Socio-
coeff. text demogr.

3,0

Poland/Finland
Frustrated Shopper
Ambitious Health Planner

Alcohol has more

Common peer

Women
alcoholic content

Weight loss programme

High drinking frequency

Wine

group drinking

advantages

Familiarity with
drinking occasion
Preference for low

2,5

2,0

1,5

1,0

0,5

0,0 * *** ** * *** * * ** * * *

Extract of results based on logistic regression analysis to identify statistically significant drivers of intention to reduce alcohol consumption in task 1.
Reference category for context (alcohol type) is “Spirits” which delivered the lowest rate for the intention to drink less.
The model explains 12% of the variance. Only variables with at least slight significance (* = p < 0.05) are displayed.
Base: EU8 (without missing variables) n = 3 967

Motives & goals

The strongest motive that influenced the intention to drink less was the preference of choosing
a drink with only low alcohol content. This attitude stands out from all other choice preferences
such as small serving size or low calories, which were not significant. Consumers who try to
avoid high alcoholic content were more prepared to further reduce their alcohol consumption
in general without controlling for any education treatment.
The other motivational driver is the typical mind-set of an Ambitious Health Planner. This type
is highly interested in making healthy choices, but s/he often lacks money and self-control to
implement the positive resolutions. In contrast to this segment, the Frustrated Consumer was
less likely to reduce alcohol drinking at the party.
The strongest barrier was peer group drinking behaviour. Since most drinking occasions are
social, it makes sense that the attitudes and behaviours of the individually relevant peer group
have such a strong impact.
If the consumer had a positive image about alcohol in general (advantages outweigh
disadvantages to society), this also presented a slight barrier to reducing alcohol
consumption.
The attitude towards risk as well as the variable on self-control did not have any impact on
the willingness to reduce alcohol in that specific occasion.

135
Impact of Food Information on Consumers’ Decision Making

Habits
There were three factors with a significant influence – one driver and two barriers.
Consumers who said that they currently follow a weight loss programme were more likely to
commit to drink less.
In contrast, the familiarity with the drinking occasion (friend’s party) as well as a high
frequency of drinking the respective type of alcohol represented a barrier to positive attitude
change.
No impact was seen for the reference measures such as the previously stated drinking volume
before the treatment or the habitual drinking volume. This is actually a surprise, since the
descriptive analysis indicated that consumers with a higher referential volume (7+ drinks)
were also less likely to reduce their high drinking volumes. Obviously, controlling for other
factors reduced the impact of high drinking volumes. Likewise the experience of having drunk
too much alcohol before was no predictor for the intention to drink less.

Context / Alcohol categories

Of all of the three alcohol categories, consumers in the spirits experiment revealed the lowest
rate of reducing their alcohol drinking. If spirits is used as a reference category, then wine is
the only tested alcohol type that represented a slightly significant driver of positive attitude
change. This means, that wine drinkers are somewhat more likely to intend to drink less on
that given occasion after an education treatment49.

Socio-demographics
The intention to drink less was also slightly influenced by additional cultural norms and a
gender effect:
 Women were more likely than men to change their mind and drink less in Task 1
 Finish and Polish consumers were less likely to reduce their drinking volume in Task 1
o When running the analysis per country, it becomes obvious that in Finland the
reference base for the pre-treatment drinking volume presents a significant
hurdle. The higher the initially stated drinking volume, the lower the motivation
to reduce this amount. This country specific barrier can be explained by the
high costs and regulations for alcohol in Scandinavian markets. On the one
hand this leads to a low drinking frequency but on the other hand to
uncompromisingly high volumes on drinking occasions such as at a party.
o The analysis for Poland did not reveal any new influencing variable, which
means that this specific cultural barrier was not captured by the survey.
Base model for decision scenario 2

49
This ultimate outcome does not change if the reference category changes. When using beer as a reference
point, then wine is again a slightly significant and positive driver. When using wine as a reference point,
then both beer and spirits both have a negative influence, but the negative impact of beer is stronger and
more significant than the impact of spirits.

136
Impact of Food Information on Consumers’ Decision Making

The second task required a decision as to whether one would drink less in future irrespective
of the occasion discussed before, but based on the same educational treatment as in Task 1.
The most relevant factors driving a positive and more long-term attitude change were
primarily based on motives and goals and a small country bias, which altogether explained
42% of the variance.

Figure 76 Drivers of choice behaviour in alcohol Task 2 without treatments

Drivers of intention to reduce alcohol consumption R2: 42

in task 2 at any occasion in the future %

Base model without policy options

Reg. Socio-
coeff. Motives & goals demogr.

3,0
Germany/France
High risk aversion
Alcohol has more
Common peer
Intention to drink less in task 1

group drinking

advantages

2,5

2,0

1,5

1,0

0,5

0,0 *** *** * * *

Extract of results based on logistic regression analysis to identify statistically significant drivers of intention to reduce alcohol
consumption in task 2. The model explains 42% of the variance. Only variables w ith at least slight significance (* = p < 0.05)
are displayed. Base: EU8 (w ithout missing variables) n = 3 967

Motives & goals

Not surprisingly, the overwhelming predictor of the long-term decision was the answer given
in the previous task. This outcome indicates that merely asking people about their drinking
behaviour and intentions may prompt a significant proportion to at least plan to reduce their
drinking. A possible explanation is that increasing self-awareness may be a motivating factor
in changing drinking patterns50. Whether these intentions are implemented into sustainable
behaviour remains to be subject to further field research.

50
Carlo DiClemente et.al. (1999): Motivation for change and alcoholism treatment

137
Impact of Food Information on Consumers’ Decision Making

Peer group drinking and positive image of alcohol affect both short term and long term
intentions
A new additional driver in Task 2 was the high level of risk aversion of consumers.

Socio-demographics
Germany and France were countries with a higher likelihood for a positive and long-term
attitude change. All other tested socio-demographic factors had no impact in this context.

6.2.2 Effectiveness of tested treatments on alcohol consumption

To measure the impact of the treatments, these were added to the base model. There are
again different ways of including these variables, either as treatment splits in combination
with the alcohol type as presented in the previous chapters:
 Split 1: Beer with calories
 Split 2: Beer with limits
 Split 3: Wine with calories
 Split 4: Wine with limits
 Split 5/7: Spirits with calories
 Split 6/8: Spirits with limits

Alternatively the treatments can be presented as aggregated stimuli:

 Calories treatment
 Limits treatment
Which are separate to the context stimuli:
 Beer
 Wine
 Spirits

Both options deliver almost identical results for the individual drivers and barriers as
documented in the chapter before. The combination with the 6 treatment splits delivers more
detailed insights on the effectiveness than the aggregated stimuli version. Overall the “Know
your limits” treatment was slightly more effective than the treatment with calories, but this
was also due to the differentiated framing of calories, which is not visible if added to the base
model in an aggregated manner. Therefore, the first version with the treatment splits was
used in the following analysis.
The explanatory power of the enhanced model is only slightly higher (13% instead of 12%)
in Task 1 and it does not increase at all in Task 2.

138
Impact of Food Information on Consumers’ Decision Making

Figure 77 Drivers of choice behaviour in alcohol Task 1 with treatments

Drivers of intention to reduce alcohol consumption R2: %

13

in task 1 at the specific occasion (at a friends´ party)

Base model with all treatment splits
Reg. Treatments Socio-
coeff. (policy & context) Motives & goals Habits demogr.

Poland/Finland
Wine with calories

Wine with limits

Spirits with limits

Beer with limits

Spirits with calories

Ambitious Health Planner

Alcohol has more

Common peer
alcoholic content

Weight loss programme

(negative framing)

(neutral framing)

(neutral framing)

(neutral framing)

(neutral framing)

group drinking

advantages

Familiarity with
drinking occasion
Preference for low
3,0

Frustrated Shopper
2,5

2,0

1,5

1,0

0,5

0,0 *** ** ** ** n.s. *** ** * *** * ** ** *

Extract of results based on logistic regression analysis to identify statistically significant drivers of intention to reduce alcohol consumption in task 1.
Reference category for treatment splits is “Beer with calories (positive framing)” which delivered the lowest rate for the intention to drink less.
The model explains 13% of the variance. Only variables with at least slight significance (* = p < 0.05) are displayed.
Base: EU8 (without missing variables) n = 3 967

The inclusion of the treatments into the base model changes the impact of only three of the
variables discussed earlier:
 A higher significance level can be observed for amongst those following a weight loss
programme, no doubt as a result of the calorific treatment.
 Gender and the high frequency of drinking no longer have any impact. The treatments
obviously reduced the female bias by making the health disadvantages of alcohol
somewhat clearer to both men and women with high and low frequencies of drinking.

Treatment options
The key driver of the intention to drink less in Task 1 was the negative framing of calorific
information for wine. This was the strongest influencing factor compared with all other
treatment splits.
The reference point for this analysis was the split with the lowest rate of positive attitude
change (beer with calories in a positive framing). Compared against this benchmark, four of
the five other treatment splits had a positive and significant impact on the goal variable. The
winning combination of ‘wine with a negative framing of calories’ was followed by the
treatments on ‘know your limits’, all showing a similar strength and significance. The

139
Impact of Food Information on Consumers’ Decision Making

treatment on spirits with a neutral framing of calories did not have any significant impact at
all.
In Task 2, the impact of the additional treatments is negligible as can be seen in the table
below.

Figure 78 Drivers of choice behaviour in alcohol Task 2 with treatments

Drivers of intention to reduce alcohol consumption R2: 42

in task 2 at any occasion in the future %

Base model with all treatment splits

Reg. Treatments Socio-
Motives & goals demogr.
coeff. (policy & context)

3,0

Germany/France
Spirits with calories

Wine with calories

Beer with limits

Spirits with limits

Wine with limits

Alcohol has more

Common peer
group drinking
(neutral framing)

(negative framing)

(neutral framing)

(neutral framing)

(neutral framing)

advantages
Intention to drink less in task 1

2,5

2,0

1,5

1,0

0,5

n.s.
*** n.s.
** n.s.
** n.s.
** n.s. *** *** ** *
0,0
Extract of results based on logistic regression analysis to identify statistically significant drivers of intention to reduce alcohol consumption in task 2.
Reference category for treatment splits is “Beer with calories (positive framing)” which delivered the lowest rate for the intention to drink less.
The model explains 42% of the variance. Only variables w ith at least slight significance (* = p < 0.05) are displayed.
Base: EU8 (without missing variables) n = 3 967

Regardless of whether the model includes the treatments or not, there were robust results
for most variables presented in the base model already. Only one change among the
previously explained variables was observed - high risk aversion no longer influenced the
intention to drink less in future if exposed to any treatment.

Treatments
In this second decision scenario all treatment options lost their significance in making any
positive impact on a more long-term decision to reduce alcohol.
The treatment of wine with calorific education seemed to be the winning combination in the
bivariate analysis also in Task 2, but the statistical model shows that this effect was not
significant.

140
Impact of Food Information on Consumers’ Decision Making

Overall, none of the tested treatments fostered a positive attitude change for any occasion in
the future.

In summary, these are the main drivers and barriers of the intention to reduce alcohol
consumption:

The key drivers are

 Preference for low alcoholic content
 Motivations of an Ambitious Health Planner, who is eager to make healthy choices,
who takes enough time to read food information. This risk-averse type prefers doing
things that pay off immediately rather than in the future.
 Following a weight loss programme

Significant barriers are

 Peer group drinking habits
 Positive alcohol image
 Familiarity with the drinking occasion
 Motivations of a Frustrated Consumer, who is not at all interested in health issues, who
does not invest any time in reading food information. This risk-taking type only
watches out for low prices when choosing alcoholic drinks.

Further cultural biases of lower significance are

 Consumers in Poland and Finland were least likely to commit drinking less in the
specific occasion (Task 1)
 Consumers in Germany and France were most likely to commit drinking less at any
occasion in future (Task 2)

This summary emphasises the importance of motivations, social norms and habits on attitude
change in relation to alcohol consumption. Obviously these informative or educational nudges
(as tested in this survey) mostly influenced those who were already following health goals.
Consumers who are not interested in health issues either in general or when consuming
alcohol will ignore health warnings as well as calorific information.
The conclusions based on these findings are outlined in the following chapter.

6.3 Conclusions regarding alcohol consumption

For the summary of findings and conclusions of the behavioural study the graph below can
serve again as a conceptual guideline. It represents a simplified and idealised illustration of a
comprehensive decision process.

141
Impact of Food Information on Consumers’ Decision Making

This scheme covers the typical stages of a consumers’ decision process from accessing and
assessing relevant food information, which then results into the intention to drink less on a
specific occasion or not.
This first decision scenario for a concrete drinking occasion was followed by a second decision
scenario for a more general and long-term attitude change, to drink less on any occasion in
future.
The identified individual biases and barriers as well as the tested treatments are marked with
symbols to indicate their direction and effectiveness of impact.

Figure 79 Summary of results on alcohol consumption analysis


The evidence from this study confirms that the motivation to drink less
strongly depends on the individual relevance of ‘healthy drinking’ as well
as the drinking habits and attitudes of peers.

142
Impact of Food Information on Consumers’ Decision Making

Within the restricted scope of variables that were possible to test in this multi-topic online
survey, the following key findings were identified as significant biases or barriers that explain
consumers´ intentions to drink less alcohol in the specified choice situation:
 The key motivation to reduce the drinking volume were healthy preferences and
habits, such as the preference of choosing drinks with lower alcoholic content and
following a weight loss programme.
 The key hurdle to intending to drink less was the social acceptance and relevance
of alcohol among peers. As most drinking occasions are social (i.e. with friends or
family), one’s own drinking intentions are highly correlated to the drinking habits of
the peer group.
 Consumers, who generally watch out for healthy alternatives and take time to read
any food information like the ‘Ambitious Health Planner’, were more likely to state
a positive attitude change, since their social milieu also drinks less alcohol.
 Frustrated Consumers, who are not at all interested in health issues and very
often have experience of consuming too much alcohol, were also less likely to reduce
their drinking intentions. This is in line with the peer group for this type which
typically was also more likely to drink alcohol.
 A positive image perception of alcohol represented a hurdle for some consumers
as well as the familiarity with the drinking occasion. Consumers with experience
of drinking the respective alcohol type in a similar situation were less likely to intend
to reduce their drinking.
 Consumers, who drank alcohol more frequently, were also less likely to reduce
their drinking volume. Overall, the drinking frequency had more impact than the
amount normally drunk. The psychological and social barriers were higher for those
who frequently drank than for those who drank only occasionally high volumes.
 In addition to these barriers described above a gender and cultural/country bias
was observed. While women were more likely to want a healthy lifestyle, consumers
in Finland and Poland faced additional local barriers.

143
Impact of Food Information on Consumers’ Decision Making

How does the additional information and education about the CALORIFIC
CONTENT of an alcoholic drink encourage consumers to drink less?

When looking at the sample split with the calorific treatment only, it is
possible to identify the drivers of the attitude change based on this new
type of food information, which nudged 16% of consumers to plan to
reduce their alcohol consumption on the specified occasion.


Education and information about CALORIES affected
consumers with distinct health goals – in particular
those who check for low alcoholic content and fewer
calories.

The calorific treatment was more effective:

 Among women
 Among consumers in France
 In relation to wine drinkers
 If presented with a negative framing, e.g. ‘Wine has more calories than chocolate’

The calories treatment was less effective among consumers:

 Who were not interested in any health issues
 Whose peers commonly drank alcohol
 Who saw more advantages than disadvantages of alcohol for society

How does the additional information and education about the

RECOMMENDED LIMITS of an alcoholic drink encourage consumers to drink
less?

144
Impact of Food Information on Consumers’ Decision Making

When looking at the sample split with the ‘Know your limits’ treatment
only, it is possible to identify the drivers of the attitude change based
on this type of food information, which nudged 19% of consumers
into reducing their planned alcohol consumption on the specified
occasion.


Education and information about ‘KNOW YOUR LIMITS’ influenced
consumers with a more general preference for low alcoholic content, and
seemed to reach a slightly wider audience than the calorific treatment.

The limits treatment was more effective among consumers with a lower level of full-time
education. This indicates that the ultimate message of ‘Know your limits’ is well understood
and motivating in particular for consumers with a presumable lower cognitive capacity.

At the same time the limits treatment was less effective among consumers:
 That were self-determined in their choices and not interested in health issues
 Whose peers commonly drank alcohol
 Who were familiar with the drinking scenario
 In UK and Germany51
 In households with children

Which conclusions can be drawn from the evidence collected on the

treatments effectiveness in motivating consumers to drink less alcohol at
the specific occasion?

Both treatments together had minimal impact on motivation as they explained only 1% of the
variance for claiming to drink less in Task 1 (in addition to the previously described factors).
Even despite the forced exposure in this experimental design, the immediate impact of these
information-based measures was still small.

51
The low impact in these two countries might be due to the fact that this campaign has already been
launched there and therefore may lack the novelty factor. However, this cannot be verified, since the
questionnaire did not include any awareness parameters of this campaign.

145
Impact of Food Information on Consumers’ Decision Making

Having said this, the alcohol warning ‘Know your limits’ was slightly more effective than the
calories treatment because addressing low alcohol content targeted more consumers intuitive
preferences than calories. In a pre-treatment question participants were asked which aspects
they usually take into account when choosing alcoholic beverages: Among thirteen choice
preferences presented …
 Only 8% mentioned ‘low calories’ (although 15% in the calories split were on a diet),
 ‘Low alcoholic content’ was mentioned by 14% of participants in the limits split.

It is clear that these pre-treatment preferences were likely to be influenced by the fact that
currently consumers are not usually presented with calorific information but with alcoholic
content on alcoholic beverages.
Whether the cognitive relevance of calorific information would increase if this information
were to become more widely available on alcoholic drinks might be indicated by looking at a
similar post-treatment question. Respondents were asked about the ideal information to be
provided on alcoholic drinks in the future. Among the given list with again thirteen different
answer options (topics)52:
 49% of participants with the calories treatment expected information on ‘calories’ for
alcoholic drinks and
 46% of participants with the limits treatment expected ‘daily unit guidelines/
recommended maximum limits’ to make more informed choices in the future.
The leap from 8% before to 49% after the calorific treatment indicates at least a significantly
higher ‘recall’ effectiveness than from 14% to 46% for the limits treatment. This does indeed
suggest its impact will increase once this type of information is available in reality.
It should be noted that the attention given to food information in a real shopping or
consumption scenario is always lower than the stated relevance in a survey. The 17% of
respondents, who claimed that they would reduce their alcohol consumption, were also
influenced by the fact that asking about the intention to do something generally delivers
answers that are more conscious, considered and rational. Stated preferences like these are
expected to be somewhat inflated by a typical overconfidence bias.
When looking at all tested treatments, the combination of a negative framing53 of the
education on calories for wine displayed the highest effectiveness with 21% of consumers
intending to drink less. On the other hand the least effective treatment was the positive
framing54 of the education on calories for beer which resulted in only 13% of consumers
planning to reduce their alcohol consumption.
This leads to a final important aspect when comparing the effectiveness between both
treatments: the consistency of the impact across different alcohol categories. The limits
treatment delivered consistent results across all categories (19% with +/-1%points) which
mirrors the homogeneous framing of the treatment.

52
The question as well as the answer options is not exactly comparable with the pre-treatment version, which
was designed not to influence the experiments.
53
Wine has more calories than chocolate.
54
Beer has less calories than pizza.

146
Impact of Food Information on Consumers’ Decision Making

In contrast to that the educational part of the calories treatment varied with different
framings, which were fully or partly responsible for the varying impact of this treatment (16%
with +5/-3%points). Whether a negatively framed education on calories would have had a
higher impact for beer and spirits as it had for wine requires further evidence.


While the ‘Know your limits’ treatment was slightly more effective than the
calorific treatment, the combination of a negative framing of calories for
wine achieved the highest impact. Further empirical evidence is required to
identify the optimal treatment and context combination.

Moving on from the analysis of the first choice task to the more general and long-term second
decision to drink less in the future:

What factors impact the more general and long-term motivation to drink
less alcohol?

Overall, the willingness to reduce alcohol consumption on any occasion reached the same
level as for the specific occasion (17%).
As expected the second decision was heavily influenced by the first decision. Two thirds of all
respondents who were motivated to drink less in the concrete scenario of a friend’s party also
intended to drink less in the future.
None of the treatment exerted any significant influence on this second decision. Instead the
analysis reveals that social norms and alcohol image impact the second decision again and in
addition to their impact in the first task. These two barriers are the recommended starting
points to be reduced for achieving a more effective motivation and behaviour change:
 Peer group drinking
 Positive image of alcohol (has more advantages than disadvantages)
In addition to these barriers one smaller motivational and two cultural drivers can be
observed:
 Consumers with a high security-orientation are more open to reduce their drinking.
 In Germany and France consumers were also more motivated by local drivers beyond
the overall reported drivers to drink less. While in France the additional motivation
originated from consumers who are still studying or who focus on low alcoholic content,
the positive attitude change in Germany was strongly driven by a preference for lower
calories and the opportunity to reduce high habitual drinking levels.


The general motivation to drink less in future is not driven by a one-time
exposure to new information on calories or recommended limits. Key
barriers to overcome are peer group drinking and positive alcohol image.

147
Impact of Food Information on Consumers’ Decision Making

148
Impact of Food Information on Consumers’ Decision Making

7 FINDINGS ON DATE MARKING AND FOOD WASTE

This section explains the findings of four questions placed in study addressing the subject of
food waste:

Do consumer misunderstandings of the ‘best before’ label contribute to

avoidable food waste in private households or do consumers throw food
away (consciously) for quality concerns?

7.1 Knowledge test and claimed behaviour

To answer this research question, first a knowledge test was conducted for the two most
commonly used expiry labels ‘best before’ and ‘use by’. Following this, participants were asked
whether they ever consumed outdated food and whether they threw away food before the
‘best before’ date.
The knowledge test started with the ‘best before’ label for all respondents in order to receive
an unbiased answer on the question “What information is shown on this label?” with several
answer options and the following results:

Table 80 Knowledge test results of best before by country

EXPIRY DATES
Knowledge test
EU8 –
Fin- Ger- Roma-
‘best before’
Total land UK Italy Spain many Poland France nia
Base unweighted 8 076 1 004 1027 1005 1002 1005 1003 1010 1023

Correct:
46.2% 69.3% 65.7% 54.2% 52.3% 44.1% 35.8% 35.3% 12.6%
highest quality

False:
36.8% 11.7% 20.8% 32.1% 31.8% 30.1% 52.4% 39.1% 76.8%
safe to eat/drink

False:
11.3% 17.9% 9.5% 8.6% 9.1% 15.0% 5.9% 18.1% 6.0%
Must be sold

None of these/
5.7% 1.1% 3.9% 5.1% 6.8% 10.9% 5.9% 7.6% 4.6%
Don‘t know

Total 100% 100% 100% 100% 100% 100% 100% 100% 100%

149
Impact of Food Information on Consumers’ Decision Making

Table 81 Knowledge test results of best before by sociodemographics

EXPIRY DATES
Knowledge test
EU8 –
Age Age Age Age Age
‘best before’
Total Male Female 18 - 24 25 - 34 35 - 44 45 - 54 55+
Base unweighted 8 076 3 280 4 796 942 1 423 1 454 1 368 2 889

Correct:
46.2% 46.7% 45.8% 41.4% 46.6% 49.2% 48.2% 45.1%
highest quality

False:
36.8% 36.6% 37.0% 45.4% 40.7% 35.5% 34.1% 34.0%
safe to eat/drink

False:
11.3% 11.8% 10.9% 9.2% 8.0% 9.0% 11.2% 14.7%
Must be sold

None of these/
5.7% 4.9% 6.3% 3.9% 4.6% 6.3% 6.5% 6.2%
Don‘t know

Total 100% 100% 100% 100% 100% 100% 100% 100%

The empirical evidence confirms previous assumptions that


 A majority of 54% of European household shoppers were not able to
identify the correct interpretation of ‘best before’ as a quality related
date.
 The most common misunderstanding (37%) confused the ‘best
before’ date with the safety related ‘use by’ date.

While 46% correctly interpreted the ‘best before’ date as the last day of highest quality, 11%
mistook ‘best before’ for ‘sold by’, and 6% had either no idea or another interpretation in
mind.
The crucial misunderstanding for 37% of participants was the belief that food is not safe
anymore past the ‘best before’ date. It was highest among the following groups:
- In Eastern European member states like Romania (77%) and Poland (52%)
- Among Ambitious Health Planners (47%), who are interested in health and quality
issues and take time reading food information.
- Among younger consumers aged under 24 (45%)

When looking at the results of the knowledge test for the ‘use by’ date a significantly higher
proportion of correct answers (72%) is observed. However, this is at least partly due to a

150
Impact of Food Information on Consumers’ Decision Making

learning effect from the previous evaluation55. Only 16% incorrectly associated the ‘use by’
label with the quality aspect.

Figure 82 Knowledge test on expiry dates

Overall, 33% consumers were able to identify the correct explanation for both labels. Four in
ten participants (39%) failed the knowledge test on ‘best before’ but succeeded in the second
‘use by’ test.

A more detailed cross-analysis of both knowledge tests reveals a slight tendency of

respondents to assign identical interpretations for both expiry dates (see table below). This
is true in particular for those who selected the ‘safety’ explanation.

55
Since the focus of this research was on food waste, the question order was fixed this way (also blocking
respondents from going back and changing previous answers) in order to achieve an unbiased answer
behaviour for the more crucial ‘best before’ test.

151
Impact of Food Information on Consumers’ Decision Making

Among the 37% who interpreted that ‘best before’ indicates the food is safe to eat until this
date, 83.5% assigned the same explanation to ‘use by’. Only 8.8% understood both labels
differently.

Table 83 Knowledge test results of best before vs. use by

EXPIRY DATES Knowledge test on ‚best before‘ date

Knowledge test
False:
EU8 – Correct: safe to False: None of these/
‘Use by’ date
Total highest quality eat/drink Must be sold Don‘t know
Base unweighted 8 076 3 732 2 976 908 460

False:
16.2% 22.0% 8.8% 17.0% 16.0%
highest quality

Correct:
72.1% 72.1% 83.5% 49.0% 43.4%
safe to eat/drink

False:
8.9% 5.4% 6.1% 32.6% 7.4%
Must be sold

None of these/
2.8% 0.5% 1.5% 1.3% 33.2%
Don‘t know

Total 100% 100% 100% 100% 100%

This answer pattern for both knowledge tests indicates that consumers tend to use the same
(safety) interpretation when confronted with either label version. One possible explanation is
that many consumers have a fixed perception of the meaning of an expiry date – no matter
with which wording it is presented.

In summary, the knowledge tests revealed a widespread misinterpretation of expiry dates by

consumers, which can be assumed to increase avoidable food waste.
This was further explored in subsequent questions which captured two complementing
behaviours:
 Consumption of outdated products kept in a fridge56:
15% said that they had never consumed or cooked food that had passed the ‘best
before date. 37% had done so, up to a maximum of two days beyond the date.

56
Q76: Thinking about products that you keep in a fridge, have you ever eaten or cooked a product when it
was past its 'best before' date? If yes, what is the longest period after the 'best before' date?

152
Impact of Food Information on Consumers’ Decision Making

 Throwing food away earlier than the best before date57:

32% said that they had thrown away a product before it had passed the ‘best before’
date.
Consuming outdated food showed a much higher differentiation than throwing food away
earlier than the best before date suggesting that both questions measure different aspects of
food waste behaviour, which can be explained by the segmentation explained below.

Figure 84 Food waste consumer behaviour

Food waste
Consumer behaviour

No, 1-2 days 3-4 days 5-6 days 7+ days Don’t

never old old old old know

Consumed 15% 37% 20% 10% 16%

3
%
outdated food
1-2 5-6
No, never days before 3-4 7+ DK
Food thrown
3
away before 60% 17% 6%
%
6% 8%
expiry
Base: EU8 – all respondents (n = 8076)

Consumed Date ignoramus: Conscious food waste

outdated food: avoider:

47% yes, 3+ days old  15% 32%

outdated outdated

53% no or max. 2 days old  Freshness maximiser: Date villain:

outdated 19% 34% outdated

Food waste before expiry: 34% yes   66% no

Base: EU8 – respondents without ‘Don’t know’ in both questions (76/77) on food waste behaviour (n = 7 319)

Again the cross-analysis of both questions reveals meaningful segments regarding the
relevance of the ‘best before’ date for consumer behaviour:
 The conscious food waste avoider (32%) has never thrown away food before the
‘best before’ date and has no problem eating significantly outdated food (3 days
beyond or older). These consumers are obviously more aware that the ‘best before’
date is not related to safety.

57
Q77: Have you ever thrown away a product that had not passed the best before date? If yes, what is the
earliest period before the 'best before' date when you have thrown something away?

153
Impact of Food Information on Consumers’ Decision Making

 The date villain (34%) has never thrown away food before the ‘best before’ date but
at the same time avoids consuming outdated food. For this segment the ‘best before’
date exudes a strong signal to discard food.
 The freshness maximiser (19%) has thrown food away even before its expiry date
and avoids consuming outdated food. This type of consumer obviously tries to
maximise the freshness of food by preferring to throw food away than eat something
that is not fresh anymore.
 The date ignoramus (15%) has thrown food away before its expiry date but has no
problem eating outdated food. Such inconsistent behaviour can only be explained by
the ignorance of expiry dates. Obviously there are other parameters (than ‘best before’
dates) relevant for the decision between edible or spoilable.

Among these four segments the most promising target groups for an education treatment
seem to be the date villain and the freshness maximiser, i.e. consumers who avoid eating
outdated food. As one would expect, there was a significant higher share among these two
behavioural segments of consumers who believed the ‘best before date’ represents a safety
limit (see table below).

Table 85 Food waste behaviour type vs. ‘best before’ awareness

FOOD WASTE Behaviour type

Knowledge test Conscious

EU8 – food waste Date Freshness Date
Best before
Total avoider villain maximiser ignoramus
Base unweighted 8 076 2 324 2 481 1 397 1 117

Is at its highest
46.2% 55.9% 43.4% 34.0% 49.6%
quality

Is safe to
36.8% 24.3% 42.5% 52.7% 31.5%
eat/drink

Must be sold 11.3% 13.9% 9.1% 9.2% 13.6%

None of these/
5.7% 5.9% 5.0% 4.1% 5.3%
Don‘t know

Total 100% 100% 100% 100% 100%

154
Impact of Food Information on Consumers’ Decision Making

So far, this descriptive analysis indicates that – all other things being equal -misunderstanding
of the ‘best before’ date is likely to foster throwing food away once the “best before” date has
passed over, but not before that date.58

7.2 Findings of multivariate analysis

For validating the hypothesis we conducted a multivariate analysis using binary logistic
regression to identify the major determinants59 of (claimed) consumer behaviour which
creates avoidable food waste. We present the summary of the results of two regression
models: one for the consumption of outdated food and one for the throwing away of food
which is not outdated.

Both models are structured in the same way, covering the following four dimensions:
 Awareness
o Incorrectly thinking that ‘best before’ is related to safety
o Incorrectly thinking that ‘use by’ is related to quality
 Motives and attitudes
o Shopper types
o Risk-aversion
o Self-control
 Habits
o Checking expiry dates when shopping for crisps or muesli
o Either “not consuming outdated food” or “throwing away not outdated food”
 Socio-demographics
o Age, gender, income, education, household composition, body mass index,
countries

All variables were selected according to their expected and/or actual impact on the (claimed)
food waste behaviour.
Both models are statistically significant and explain a reasonable proportion of the variation
in the dependent variable.

58
The share of misunderstanding the best before date as a safety limit is significantly higher among
consumers who would never consume outdated food, such as the date villain and the freshness maximiser
(as indicated in the table before).
59
As this topic was only one among many others in the survey, we certainly did not cover all relevant factors
that impact food waste behaviour. Moreover, there was no space for creating experiments on this topic,
which would have reduced an expected overconfidence bias.

155
Impact of Food Information on Consumers’ Decision Making

Base model for outdated food

The most relevant factors driving consumer waste of food past its ‘best before’ date were
awareness variables, motives and attitudes, socio-demographics and, to a lesser extent,
habitual bias.

Figure 86 Drivers of aversion to consume outdated food

Drivers of aversion to consume outdated food R2: 20

%
(More than 2 days past the ‚best before‘ date)

Reg. Aware-
coeff. Motives & goals Habits Socio-demographics
1,2 ness
Miscon- Shopper types Country bias
ceptions

Higher personal income

Men

Still studying or
higher education
when shopping crisps
Reading expiry dates

Reading expiry dates

Higher risk-aversion

when shopping muesli

1,0
Shopper

Ambitious Health Planner

Self-Determined
= safety limit

= quality limit

Frustrated Shopper

Bargain Hunter

Expert Shopper
Best before

Use by

RO
DE
FR
0,8

0,6 PL
ES

0,4

0,2

*** n.s. *** *** ** ** * * ** * *** * * *** ***

0,0
Extract of results based on logistic regression analysis to identify statistically significant drivers of NOT consuming food w ith a ‘best before’ date older
than 2 days. Reference category for countries is UK and for shopper types it is the rushed quality shopper. The model explains 20% of the variance.
Only variables with at least slight significance (* = p < 0.05) are displayed. Base: EU8 (without missing variables) n = 7 717

Awareness
Consumers, who believed that the ‘best before’ label represents a ‘safe to eat/drink’ limit were
significantly more likely to claim to not consume outdated food. The interpretation of the ‘use
by’ date had no impact at all.


The empirical evidence shows that the misconception of the ‘best before’
date as a safety limit is one of the strongest factors which drive consumers
to throw away outdated food.

156
Impact of Food Information on Consumers’ Decision Making

Motives & goals

People who were more security conscious also had a strong impact on the decision to throw
away outdated food. This goes hand in hand with the previously stated misconception.
Consumers who said that they were generally cautious in their way of life were more likely to
watch out for any indicators (like any expiry date) that may affect their safety and well-being.
The impact of the shopper types indicated the relevance of quality and health in contrast to
time and money. All three types who lacked interest in quality and healthy products (bargain
hunter, frustrated shopper, self-determined shopper) were also less inclined to throw away
outdated food, whereas the other types (health planner, expert shopper) who highly value
quality and health aspects were more likely to avoid eating outdated food.

Habits
The habit of checking the expiry date – when shopping for products such as crisps or muesli
– suggests that this information is highly relevant to these consumers. Therefore it is not
surprising that these people are less likely to consume outdated food.
The tendency to throw away food which is not out of date does not have any influence on
whether people throw away food which actually is past its expiry date.

Socio-demographics
Men are less likely to consume outdated food than women.
Students, individuals with a higher level of education and/or with a higher personal income
are slightly more likely to consume outdated food.
However, the strongest impact on the decision to reject or consume outdated food was a
country or cultural bias60:
 Shoppers in Eastern European countries like Romania and Poland are most likely to
avoid outdated food;
 By comparison, the majority of shoppers in Germany, France and Spain have no
problem with the consumption of outdated food.

Base model for not-outdated food

The most relevant factors driving food waste before its ‘best before’ date were predominantly
socio-demographical and motivational variables and only to a lesser extent awareness of the
expiry date. No habitual drivers were identified among the tested variables.

60
The reference country for this analysis is the UK, which displays an average behaviour regarding food waste
among all countries researched.

157
Impact of Food Information on Consumers’ Decision Making

Figure 87 Drivers of throwing away not-outdated food

Drivers of throwing away not-outdated food R2: %

8

(Before the ‚best before‘ date)

Reg.
coeff. Awareness Motives & goals Socio-demographics
1,2
Misconceptions Shopper types

Lower personal income

Men
Low self-control

Higher risk-aversion

Younger age (<34)

Older age (55+)

Household without children

Self-Determined Shopper
= safety limit
Best before

= quality limit

Expert Shopper
Use by

1,0 Health Planner

Romania
0,8

0,6

0,4

0,2

* n.s. *** ** * * * *** *** *** ** * *

0,0
Extract of results based on logistic regression analysis to identify statistically significant drivers of throw ing food away before the ‘best before’ date.
Reference category for countries is UK and for shopper types it is the rushed quality shopper. The model explains 8% of the variance. Only variables
with at least slight significance (* = p < 0.05) are displayed. Base: EU8 (without missing variables) n = 7 717

Awareness
Consumers, who believed that the ‘best before’ label represented a ‘safe to eat/drink’ limit
were only slightly more likely to claim to throw away food not out of date. The interpretation
of the ‘use by’ date had no impact at all.
Motives & goals
Unlike the previous model, being more security conscious showed only a slight impact on the
decision. Instead, low self-control was more likely to drive food wastage. This indicates a
more impulsive behavioural pattern driven more by short-term emotions than by long-term
goals.
This finding was complemented by the motivation of the relevant shopper types. The more
strategic shoppers, who take time to carefully identify the best quality and healthy products,
were significantly less likely to throw away food still safe to eat. The Self-Determined Shopper
who is not at all interested in information on quality and health aspects was more likely to
discard unexpired food. The conclusion is that there must be further subjective reasons (not
captured in this study) in addition to the safety issue to explain this specific food waste
behaviour.
Socio-demographics
Wasting food before it reached the ‘best before’ date was most often seen among:

158
Impact of Food Information on Consumers’ Decision Making

 Consumers in Romania,
 Younger consumers;
 Women;
 Households without children.
It was least likely among consumers with a very low income.
In summary, the data indicates that different food waste scenarios are influenced by different
sets of variables. The misconception of the ‘best before’ date as a safety limit was only a
strong driver of wasting food past that date – but not before.
Further empirical evidence is required to
 Explore further reasons why consumers throw away food before and/or after the expiry
date;
 Better understand the strong country biases, which might be due to varying social
norms, differences in climate or storage equipment, cultural cooking or consumption
habits, specific food categories, etc.
 Test whether any information and/or education is effective enough to change at least
the intention to reduce food waste.
There are various potential treatments which may reduce the impact of the current
misconception of the ‘best before’ date – such as the following suggestions:

 Information based measure

o Different wordings of the term ‘best before’ in each language
 Educational intervention
o Education on the meaning of ‘best before’
o Raising awareness of the disadvantages of wasting food
 Choice architecture
o Offering only the date until when it is safe to eat (no quality related dates)
o Offering two dates: manufacturing date and last date when it is safe to consume
o Offering two dates: best quality and safe to consume

Behavioural theory would suggest that changes in the choice architecture will have the
highest impact, while the pure rewording of the ‘best before’ date is likely to be inefficient.
However, this has yet to be tested.

159
Impact of Food Information on Consumers’ Decision Making

8 DESIGN OF STORE EXPERIMENTS ON TRANS FAT

Following the analysis of the multi-topic multi-country study, the next step was further
investigation of consumer choices on trans fat issues in a more realistic offline shopping
environment.
The fieldwork was conducted in nine comparable supermarkets of a large retailer chain in
Germany in July 2014.

8.1 Updated research objectives for TFA experiments

The online survey provided a range of insights on how consumers react when confronted with
different TFA information in laboratory choice tasks (see Chapter 4).
The experiments based on the product category of crisps indicated that consumers tend to
react solely to TFA-related information on the nutrition facts label but not in the ingredients
list. While they performed well in avoiding TFA based on the nutrition information in a simple
task, consumers faced difficulties when needing to balance between the amounts of TFA and
saturates in a more complex choice situation. Only the combined treatment with an education
created an overall positive effect on making healthy choices.
The purpose of the follow-up research was to test whether these findings can be extrapolated
to other product categories with different choice architectures in the more natural offline
environment of a supermarket.
The key questions for the offline experiments were as follows:
a) Do consumers access food information while shopping in a bricks and mortar
supermarket?
b) How do consumers react, when presented with trans fat information? Will the
transparency of TFA amounts lead to more or less healthy product choices at the point
of sale?
c) Are shoppers able to identify a healthier product?

While the first research question was based on the observation of natural consumer behaviour
at the shelf when selecting products, the latter questions required the controlling of specific
context and treatment variables for a meaningful analysis:
 Two treatment scenarios vs. control group, i.e.
o Does the TFA amount on the nutrition panel decrease or increase good choices?
o Does the education treatment improve the impact of the display of the TFA
amount?
 Two task designs, i.e.
o Task 1: Does TFA transparency foster healthy choices when the product with
no/less TFA is healthier than the alternative?
o Task 2: Does TFA transparency hinder healthy choices when the product with more
TFA is healthier than the alternative?
 Three product categories (choice architecture), i.e.

160
Impact of Food Information on Consumers’ Decision Making

o Yoghurt/Dessert (Task 1: milk with some TFA and more saturates vs. soy without
TFA but more sugar / Task 2: slightly more TFA vs. significantly more sugar and
more saturates)
o Frozen pizza (Task 1: more TFA and saturates vs. a little more salt / Task 2: slightly
more TFA vs. more salt and saturates)
o Biscuits (Task 1: more TFA and saturates vs. more sugar / Task 2: slightly more
TFA vs. more sugar and saturates)

8.2 Overview of experimental design considerations

Since the experimental setting of the preceding multi-country study was driven by the more
economical online collection of large samples in many countries, the second research step
focused on purchase decisions in a more natural field environment of an ‘offline’ supermarket.
The reading behaviour of food information at a supermarket shelf can be expected to be very
different to the ‘zooming’ behaviour of consumers in an online choice task. As outlined in
Chapter 2.1.3 most choices in supermarkets are made by habitually choosing the same
products without reading food information at the point of sale.
Therefore, a pure field experiment by manipulating the TFA information on the real products
would have delivered a very small sample of shoppers who noticed the relevant information
in the shopping process. Moreover, the manipulation of real products would have required an
arbitrary allocation of real brands to a higher or lower trans fat amount, as the real TFA
content is unknown to us.
These were the main reasons why in this research step the experimental setting again was a
more controlled laboratory type by observing the decision making between mock-up products
at a stand located in a supermarket.61
The process of this in-store research was designed as follows:
1. Customers were observed while interacting with the product category (real stakes
without manipulation). The main purpose of the observation was to provide data on
the reading behaviour at the shelf and to identify shoppers for recruitment to the
experiment.
2. Shoppers who put at least one product in their basket were contacted after leaving the
category area and screened for eligibility to the experiments.
3. Eligible shoppers who were willing to participate were guided to the experimental stand
where another interviewer took over.
4. Initial questions covered shopping goals and the type of food information that was
looked for (if any reading was observed previously).
5. The interviewer introduced the experimental task and presented the mock-up products
and noted the observations of the participants’ reading and decision process.

61
A similar experimental setting at stands in supermarkets was conducted by Sheena Iyengar (2000): When choice
is demotivating http://www.columbia.edu/~ss957/articles/Choice_is_Demotivating.pdf

161
Impact of Food Information on Consumers’ Decision Making

6. Following the experiments the respondent received a detailed feedback on his/her

decisions, and the interviewer took notes on the reactions to this feedback.
7. All participants who completed the experiments received a shopping voucher as an
incentive.

Figure 88 Images of the data collection process in supermarkets

As in the online research, the offline experiments also excluded any differentiation by brand
or price. Participants were asked to assume that the presented mock-up products are new
offers from their favourite brands at the typical price that they normally pay.
In contrast to the online lab experiments, the offline ‘field & lab’ experiments in this section
were conducted
 in a real supermarket;
 with more realistic tangible products which contained more detailed food information;
 with shoppers who performed the same shopping decision in real life moments before
the experiment;
 by linking the observation data on the real reading behaviour at the shelves with the
questionnaire routing and the interview data.

162
Impact of Food Information on Consumers’ Decision Making

8.3 Product categories tested

The product categories selected for this experiment were yoghurt, frozen pizza and
biscuits. This choice was based on the following criteria:
 Relevance to the policy area:
Naturally occurring TFA can be found in fat-containing dairy products like cheese,
yoghurt, butter, as well as meat and meat products from ruminant animals, such as
cows, goat, and sheep.
Industrially produced TFA may be found in processed foods where PHO is added as
ingredient or where PHO is used during processing (e.g., frying) such as in prepacked
biscuits, cakes, croissants, crisps, popcorn, chocolate products, fried potato products,
margarines, pastries, etc.
 Relevance to consumers:
The selected categories had to be popular enough to achieve a reasonable sample size
within the time frame.
The selected range of product categories covered both trans fat types - ‘natural’ and
‘industrial’. Yoghurt contains only natural TFA, in biscuits TFA can occur naturally when butter
is used as ingredient or come from industrial origin if PHO is used as ingredient and frozen
pizza usually contains natural TFA from cheese and may in addition contain industrially
produced TFA in case PHO is used as ingredient.
The reason why yoghurt was chosen for testing is that it is a product with a healthy image,
with only small variations in the level of trans fat but higher levels saturates and potentially
sugar. Moreover, it was tested whether the display of trans fat information would deter
consumers to switch to alternative (soy) products instead.
Frozen pizza was selected to test higher variations of trans fat vs. salt and saturates. Biscuits
were selected to test higher variations of trans fat vs. sugar and saturates.
All categories are commonly purchased by consumers in German supermarkets. Only biscuits
had a slower turnover in the summer time, which required an extension of field work.
The mock-up products for these categories were developed by a professional packaging
supplier to represent a typical product in a neutral ‘black & white’ format.

163
Impact of Food Information on Consumers’ Decision Making

Figure 89 Product images

8.4 Treatment splits for TFA transparency

The next dimension to be considered was the design of treatment stimuli and the number of
treatment splits to be tested. To examine the impact of possible treatments the following
stimuli were developed to test their effectiveness on driving healthy choices:
 Control group: Nutrition facts panel containing no TFA information. The only indicator
for industrial TFA was the presence of PHO in the ingredients list.
 Treatment 2: Nutrition facts panel with additional information on TFA amounts.
 Treatment 3: Same as Treatment 2 plus an education intervention explaining about
an overall healthy diet and the fat types (presented as a nutrition guideline).

164
Impact of Food Information on Consumers’ Decision Making

Additional TFA information on nutrition label

Since consumers will be confronted with a new terminology (partially hydrogenated oil – PHO)
in ingredients list from 2014, it was another goal to find out whether the term is understood
properly. If so, does it foster healthy choices even if it does not provide exact information on
the amount of TFA?
The alternative scenario was to provide concrete TFA amounts as is common in other countries
outside the EU (see example below):

Figure 90 Example without TFA (control group) and with TFA (treatments 2/3)

The key question is whether the additional and more concrete TFA information actually helps
consumers to make healthier choices or whether it is in fact misleading and counter-
productively causes overreactions.

Education intervention explaining all fat types

It was expected that a large share of consumers would have only little or no awareness of
specific fat terms such as TFA and PHO. As a consequence an additional treatment was
developed62 to simulate an educational intervention for one third of the participants. This was
to help analyse whether an objective explanation of this rather complex topic to consumers
helps increase healthy choices.
While the treatments with TFA information were embedded into each choice task, the
education treatment (see figure below) was presented once before the first task, when asked
to identify the healthier product.

62
The content of this treatment was developed with the aim to balance the need for a short, targeted and easy-to-
understand leaflet to be quickly read and understood by participants with ensuring an existing evidence base and a
comprehensive education of a balanced diet.

165
Impact of Food Information on Consumers’ Decision Making

Figure 91 Education treatment for TFA field experiment

166
Impact of Food Information on Consumers’ Decision Making

The interviewers were requested to neither influence the reading time of respondents nor
discuss any questions they might have. After its initial presentation the respondents were
allowed to use and read the guideline again, if they wanted.

8.5 Selection of markets and respondent screening

The definition of three product categories and three different treatment scenarios required
the selection of nine separate but comparable stores.

Figure 92 Store segmentation

Hence, co-operation with a large retailer chain was required to select nine stores that shared
the following criteria:
 Organisational membership: all stores belonged to the same retailer brand “real,-“
 Outlet size: all stores were of a comparable size and turnover (+/- 10%)
 Equality of sales volume within the test category: same turnover (+/- 10%)
 Equality of competitive environment: same products on the shelves of the test
categories
 Geographical region and urbanity: similar customer structure as all stores are located
in the outskirts of larger cities in the North of Germany
For a reasonably robust analysis approximately 200 experimental interviews per store/split
were targeted and achieved. The overall gross number of observations at the shelves varied
between two and three times the net achieved sample depending on the recruitment rate in
each market.

167
Impact of Food Information on Consumers’ Decision Making

Table 93 Sample sizes per split

# Category TFA amounts Education Observations Interviews

Split 1 Yoghurt No No 383 200

Split 2 Yoghurt Yes No 501 202

Split 3 Yoghurt Yes Yes 574 201

Split 4 Frozen Pizza No No 598 211

Split 5 Frozen Pizza Yes No 372 201

Split 6 Frozen Pizza Yes Yes 615 203

Split 7 Biscuits No No 473 202

Split 8 Biscuits Yes No 663 239

Split 9 Biscuits Yes Yes 662 192

Total sample 4 841 1 851

While the gross sample of 4841 shopper observations allows us to further understand typical
consumer reading behaviour at the shelves, the core analysis of the experimental design is
based on the net sample of 1851 respondents.
Between the observation and the experiments a thorough screening process was conducted
to make sure that the net sample included the envisaged target group features. Below is an
overview of the screening process from gross to net sample:

Table 94 Overview of screen out process

Sample Definition

4 841 All observed shoppers who interact with the relevant test category

-236 Screen out of non-buyers (Q10)

-2 247 Not willing to participate in survey (Q15)

-8 Screen out of <16 year old shoppers (Q18)

-31 Screen out of only low fat shoppers (Q19)

-12 Screen out of previous participants (Q20)

-456 Not willing to participate in full interview (Q21)

= 1 851 Respondents with complete experiments and interview data

168
Impact of Food Information on Consumers’ Decision Making

The target sample included only shoppers aged 16 and older who selected at least one non-
diet product of the observed category. The exclusion of “diet” shoppers was due to the fact
that the experiments involved products with high variation of fat levels, which would have
been too hypothetical for these shoppers.

8.6 Framing of TFA experiments

The experiments were embedded within an interview following the screening process. The
interviewer at the stand asked about the previous observation as to whether food information
was read at the shelf or not.
All questions asked before the experiment were carefully selected and worded in order to
achieve unbiased choices in the experiment. The introduction to the choice tasks (see figure
below) presented reminders of reading and selection habits. Moreover, it explained how to
overcome the lack of brand and price information, which normally represents a key choice
feature for consumers. Each respondent had to imagine that the presented mock-up products
came from their individually favoured brand at the usual price.

Figure 95 Introduction to TFA experiments

The purpose of the framing with these reminders was to help respondents imagine a more
realistic and individual choice situation, to allow more realistic observations of choice
behaviour in the experiment.

8.7 Choice tasks and choice characteristics

The experimental design reflected two subsequent choice sets in order to answer the two
different research objectives:
a) How consumers react, when presented with trans fat information?
b) Are shoppers able to identify a healthier product?

The first objective required the observation of initial – often less conscious – choices that are
as unbiased as possible also in terms of health goals. Only if shoppers care about selecting a
healthier product is it likely that they will search for relevant food information. Hence, this set

169
Impact of Food Information on Consumers’ Decision Making

of choice tasks was positioned at the beginning of the experiment without any prior education
treatment or consideration of the healthiness of individual nutrition elements. It should be
kept in mind that the experimental task with its controlled variation of specific features was
more likely to automatically draw the attention of attentive consumers towards a healthy
choice. However, the framing of the choice tasks nudged more the habitual behaviour which
can be expected to achieve more realistic choices.
The questions in the initial choice set were:
 “Which [product] would you choose – A or B or both or none of them?”
 “Which [product] would you choose – X or Y or both or none of them?”
Following the initial choices, the respondent was asked to reflect on the healthiness of the
key nutrition elements based on his/her current knowledge. According to the previous
research step, the correct perception of healthiness of saturates and trans fats was a
significant prerequisite of the ability to identify a healthy product.
Before introducing the final choice set, one third of the respondents received the education
treatment in the format of a nutrition guideline.
Subsequently, the participants were asked to make ‘good choices’:
 “Which [product] is healthier – A or B or are they both equally healthy or equally
unhealthy?”
 “Which [product] is healthier – X or Y or are they both equally healthy or equally
unhealthy?”
In contrast to the initial choices this set of questions reflected more cognitive decisions in
particular for consumers who less used to looking out for healthy choices. The preceding
assessment of the healthiness of nutrition elements provided a balanced framing for all
respondents. Even those, who already made their initial choices consciously towards a health
goal, had the opportunity to double check the food labels again after this deliberative exercise.

170
Impact of Food Information on Consumers’ Decision Making

Figure 96 Overview of TFA experiments flow for respondents

With this experimental design, the impact of the education treatment can only be measured
in the latter choice set and not for the initial choices. This limitation is acceptable, as the key
purpose of the education is to help consumers to identify healthy products. To measure the
impact of education treatments on more habitual initial choices would require a more long-
term field experiment.
The following guidelines were established to develop the choice architecture details:
 Presentation of pairwise choices with one product clearly superior than the other from
a nutritional perspective.
 Each choice set (initial and good choices) consisted of two tasks to cover two different
choice settings:
o Task 1 presented as healthier choice product (B) without or with significantly less
TFA than the alternative  noticing and understanding the TFA content was critical
for a correct decision and therefore TFA transparency was expected to increase
healthy choices in this setting.
o Task 2 presented as healthier choice product (X) with slightly more TFA than the
alternative
 for a correct choice, TFA information needed to be ignored or participants
needed to understand that the difference in TFA contents did not outweigh the

171
Impact of Food Information on Consumers’ Decision Making

high amounts of other unhealthy nutrients . TFA transparency was expected to

decrease the number of healthy choices in this setting.
 Products did not have more variations than absolutely necessary
 Products contained roughly realistic amounts and combinations of nutrition elements

Of course, in reality people also choose between products that are equally healthy or
unhealthy options. However, such choices do not call for interventions of this type
(transparency measures only make sense if there is an opportunity for consumers to identify
a healthier alternative). The descriptions of Task 1 and 2 covered the key situations, where
consumers may make bad choices because of misunderstanding or ignorance.
When creating quasi field experiments there is a temptation to include as many variations as
possible to simulate the complexity of a real life decision. On the other hand, the observation
of choice behaviour also requires an objective reasoning and interpretation. The more
variations the less we can determine what really triggered the choice.
In contrast to the previous online experiments, there was now one additional variable in each
product (either sugar or salt), which competed with the variation of saturates and trans fats.
The purpose of this additional dimension was to compare the relative impact of these nutrition
elements on the choices.

Figure 97 Choice architecture of TFA experiments

Product Task rule Superiority TFA (g) SF (g) Fat (g) Sugar (g) Salt (g) other

Dessert A more TFA less healthy 0,2 1,7 2,5 7,3 0,1 Cow milk
TASK 1

Dessert B less TFA healthier 0 0,4 2,5 9,0 0,1 Soy milk

Pizza A more TFA less healthy 1,3 3,9 9,4 2,7 1,2 -

Pizza B less TFA healthier 0,2 3,0 9,4 2,7 1,6 -

Biscuit A more TFA less healthy 3,2 7,1 13,3 28,2 0,2 -

Biscuit B less TFA healthier 0,2 5,0 13,3 40,9 0,2 -

Yoghurt X more TFA healthier 0,4 5,6 10,0 6,6 0,1 Cream

Yoghurt Y less TFA less healthy 0,2 6,9 10,0 17,9 0,1 Cream
TASK 2

Pizza X more TFA healthier 1,4 4,4 11,2 2,6 0,8 -

Pizza Y less TFA less healthy 1,2 5,6 11,2 2,6 1,8 -

Biscuit X more TFA healthier 2,5 10,1 25,0 18,2 0,2 -

Biscuit Y less TFA less healthy 2,3 12,6 25,0 26,9 0,2 -

In Task 1 product A in all categories was designed to be less healthy than product B because
of more trans fat and slightly more saturated fat, and slightly lower sugar or salt content than
product B. However, without knowing that TFA is a very unhealthy ingredient product A
appeared to be the healthier option. This is a scenario, where – in an ideal world with rational
consumers - more transparency about TFA should increase better choices.

172
Impact of Food Information on Consumers’ Decision Making

The lowest transparency level was simulated in the control group without any information on
TFA amounts and without any education treatment. The only way to identify significant levels
of TFA was the presence of partially hydrogenated oil in the ingredients list.
The fact that ‘natural’ trans fats are generally not listed in the ingredients list required an
additional distinction in the yoghurt category in Task 1. The ingredients differed by cow milk
vs. soy milk63. This design also allowed testing whether consumers confronted with low TFA
amounts on a yoghurt carton would actually switch to soy alternatives without TFA.
In Task 2 product Y in all categories was designed to be less healthy than product X despite
a slightly lower TFA content. The significantly higher amount of sugar or salt in combination
with somewhat more saturates in product Y outweighed the small variations in TFA levels.
The purpose of this scenario was to test whether consumers overreact when they are
confronted with the presence of TFA and forget about other unhealthy ingredients.
After each of the four choices, the respondents were asked why they chose this product and
what type of food information they were looking for when making their choice.
Eventually, the experimental part of the interview was finalised by giving the respondents
feedback on the healthiness of the four different products through a verbal and visual
explanation (see examples below).

Figure 98 Feedback with choice characteristics of yoghurt category in Task 1

63
Since the natural TFA content of cow milk usually does not vary very much for making differentiated choices, soy
milk would be an alternative to avoid any TFA at all.

173
Impact of Food Information on Consumers’ Decision Making

Figure 99 Feedback with choice characteristics of pizza category in Task 2

The interview at the stand was concluded with some questions on how consumers reacted to
the disclosure of TFA information and which treatments they would prefer.

174
Impact of Food Information on Consumers’ Decision Making

9 FINDINGS ON STORE EXPERIMENTS REGARDING TRANS FATS

This section explains the most important findings of the follow-up TFA experiments conducted
on the premises of selected supermarkets in Germany.
Again we present the data in a structured way based on a simplified map of a shopper’s
decision process, which includes the potential drivers and barriers as identified in the previous
research plus some new variables.

Figure 100 Guideline for analysis of results on store experiments

The ultimate objective is to identify the drivers and barriers of making healthy food choices
and whether more transparency on TFA amounts has any significant impact.
The analysis starts with a description of the observation of consumers´ shopping and reading
behaviour at the shelf in Chapter 9.1. This section focusses on the observed reading behaviour
when selecting yoghurt, pizza and biscuits in a supermarket. In addition, this section describes
the subjective consumer preferences, collected after the experiments.
Chapter 9.2 covers the disaggregated findings of the observed decision behaviour in the initial
choice tasks with suggestions for variables to be included in the statistical base model. The

175
Impact of Food Information on Consumers’ Decision Making

disaggregated findings of the more conscious identification of healthy products follows in

Chapter 9.3.
Chapter 9.4 combines the discussed variables through a multivariate analysis to answer the
key questions about what are the statistically relevant drivers and barriers of choosing or
identifying the healthy products and whether the transparency of TFA amounts impacts these
choices in a positive or a negative way.
Chapter 9.5 summarises the key insights and discusses the main conclusions.

9.1 Observed choice behaviour at the shelf and stated preferences

This section briefly describes the extent to which consumers read food information when
selecting products at the shelf.
The vast majority were so-called ‘grab & go buyers’, who quickly or habitually made a
purchase decision without reading anything. Less than a quarter of all observed shoppers
(who selected at least one product within the monitored categories) actually read the food
information. This is in line with the empirical evidence of multiple previous shopper surveys
conducted by TNS in European supermarkets.
The likelihood to read food information varied between product categories:
 Products with a healthier image such as yoghurt were more likely to be read.
 Products from a freezer compartment such as pizza attracted fewer readers.

Table 101 Shopping and reading behaviour at the shelf

Shelf behaviour of observed All categories Yoghurt Frozen pizza Biscuits

shoppers: Base unweighted 4841 1458 1585 1798

- Reading buyers 23% 34% 10% 24%

- Grab & go buyers 73% 62% 87% 68%

All buyers 95% 97% 96% 93%

Non-buyers 5% 3% 4% 7%

Total shoppers with at least

100% 100% 100% 100%
one product interaction

There was no significant gender difference in reading behaviour. Older shoppers

(approximately 55 years or older) were somewhat more likely to read food information at the
shelf.
Reading food information clearly takes more time than just “grab & go” shopping. The average
duration of the shopping/selection process at the shelf takes twice the time (63 seconds) for
readers than for non-readers (36 seconds).

176
Impact of Food Information on Consumers’ Decision Making

Table 102 Average duration of shopping processes at the shelf

Average duration of observed

shopping processes (seconds): All categories Yoghurt Frozen pizza Biscuits
Base unweighted 4841 1458 1585 1798

- Reading buyers 63 sec. 58 sec. 57 sec. 69 sec.

- Grab & go buyers 36 sec. 41 sec. 32 sec. 38 sec.

All buyers 43 sec. 47 sec. 35 sec. 46 sec.

Non-buyers 52 sec. 40 sec. 45 sec. 60 sec.

Total shoppers 43 sec. 47 sec. 35 sec. 47 sec.

The readers of biscuit food information took longer on average (69 seconds) than readers of
other products. The fact that “grab & go” shoppers of yoghurt took more time than in the
other categories can be explained by the wider range of products in this category.
The typical package formats of the product categories also determined the ease of access for
food information.
 Almost all reading yoghurt shoppers looked at the information printed on the top, while
only a minority also looked at other areas, where the ingredients and nutrition panels
are normally located.
 Frozen pizza is typically stored in a way where the upper side serves as a ‘front panel’
which attracts the most attention. Detailed ingredients and nutrition facts are mostly
printed on the back panel which was used only by about one third of the readers.
 Biscuits come in differently shaped packaging and attracted more careful readers who
searched information on several sides. Overall, there was only one shopper in the
biscuits category who actually used a smartphone to scan a bar or QR code for more
information.

177
Impact of Food Information on Consumers’ Decision Making

Participants in the experiment were asked about the type of information they normally look
for. The three most relevant factors were flavour, price and brand. These were particularly
important for habitual shoppers who did not read information at the shelf. These consumers
were also less interested in nutrition elements such as sugar or other fat types.

Figure 103 Type of food information read at the shelf

Type of food information read at the shelf

Recall by habitual vs. reading shoppers
Information searched by
71% Grab & Go shoppers
Flavour/Toppings 61% Reading shoppers
28%
Price 26%

25%
Brand 22%

11%
Ingredients 13%

12%
Size/Volume/Weight 12%

6%
Other fat types 10%

5%
Best before 12%

5%
Sugar 10%

5%
Nutrition facts 8%

5%
Calories 7%

4%
Saturates 5%

Q25: So today, when selecting your [category] you were reading something. What specific type of information did you ac tually
look up? Base: respondents, who read at the shelf (n=454) – open answers with prec odes
Q26: Today when you selected [category] you were not reading any spec ific product information. What kind of information would
you normally be reading, if your favourite brand or product type would not have been available?
Base: all respondents, who did not read at the shelf (n=1397) – open answers with precodes

The product category also influences the type of information looked for. Details on saturates
or other fat types were relevant in particular for yoghurt shoppers (18%). Information on
sugar was checked by one in ten of those choosing yoghurt or biscuits. Pizza shoppers had
almost no interest in any nutrition elements. Almost nobody searched for salt or sodium
information.

178
Impact of Food Information on Consumers’ Decision Making

This search pattern for food information correlated with specific shopping goals. For the
average consumer health aspects were less important than other factors such as taste,
convenience and price. A notable difference between ‘grab & go’ and reading shoppers was in
the relevance of
 Convenience, which was more important to ‘grab & go’ shoppers,
 Health, which was more important to reading shoppers.

Figure 104 Relevance of shopping goals

Perceived relevance of aspects on shopping decisions

By habitual vs. reading shoppers
Very strong
impact (5)

Grab & Go 75% 19% 5%

Taste Reading
76% 19% 3%

Grab & Go 25% 24% 24% 10% 17%

Convenience
Reading 18% 23% 22% 15% 22%

Grab & Go 19% 19% 34% 14% 14%

Price
Reading 18% 23% 33% 13% 13%

No impact at all (1)

Grab & Go 17% 18% 22% 17% 26%
Health
Reading 18% 23% 27% 14% 18%

Q27: How much of an impact do the following aspects usualy have on your decision to buy [c ategory]?
Please use a scale where 1 means ‘no impac t at all’ and 5 means ‘a very strong impact’.
Base: all respondents – yoghurt split (n=603), pizza split (n=615), biscuit split (n=633)

This suggests that consumers who care about the healthiness of their food are also more
likely to spend time reading food information when shopping. Indeed, 18% of shoppers who
stated a (very) strong impact of health aspects on their decisions were also most likely to
consider reading food information as important for making right choices.

179
Impact of Food Information on Consumers’ Decision Making

Another perspective on shopping goals is the comparison of shoppers of different product

categories. As one may have expected, yoghurt shoppers regarded health aspects as more
important than pizza or biscuit shoppers. This reinforces the previous assumption that yoghurt
is more often subject to a conscious decision towards a healthy diet than the other two
categories.

Figure 105 Relevance of health goal by product categories

Perceived relevance of aspects on shopping decisions

By product categories

Very strong No
impact (5) impact (1)

Yoghurt shoppers 24 29 25 12 10

Health Pizza shoppers 9 13 22 21 35

Biscuit shoppers 18 16 24 15 28

Q27: How much of an impact do the following aspects usualy have on your decision to buy [category]?
Please use a scale where 1 means ‘no impact at all’ and 5 means ‘a very strong impact’.
Base: all respondents – yoghurt split (n=603), pizza split (n=615), biscuit split (n=633)

The individual relevance of health aspects on shopping decisions will be included into the
overall analysis model as well as the habits of reading specific food information at the shelves.
Another set of potentially relevant factors is knowledge about the (un)healthiness of
nutritional elements. Whether consumers know that trans fat is very unhealthy or whether
they think it is healthy or whether they have no idea about it – this could be a driver or barrier
to making good choices.
The overall evaluation of selected nutrition elements (see figure below) indicates that the
term ‘trans fat’ (translated into ‘Transfette’) is still widely unknown to German shoppers.
While the vast majority correctly identified sugar as unhealthy and protein as healthy, there
were more substantial knowledge gaps when evaluating saturated fats, salt and partially
hydrogenated oil. However it is still worth noting that even so, two thirds recognised partially
hydrogenated oil to be unhealthy.

180
Impact of Food Information on Consumers’ Decision Making

Figure 106 Healthiness assessment of nutrition elements

Assessment of healthiness of nutrition elements

Very Quite Don’t Quite Very

healthy healthy know unhealthy unhealthy

%
Sugar 1 9 1 56 33

Partially 1 12 20 41 25
hydrogenated oil

Salt 2 29 4 54 12

Saturated fats 6 23 16 36 20

Trans fats 13 66 13 18

Protein 24 63 6 8

Q41: Please have a look at this question and indicate for each element whether you consider it to be very or quite healthy or
unhealthy. If you have no idea then tick “don’t know”.
Base: all respondents (n = 1851)

Before discussing the observed choice behaviour in the experiments, here we briefly describe
the participants’ reactions to the disclosure of TFA information as well as their preferred
interventions.
As indicated before, the majority of participants had never heard of TFA before the
experiments. While only around one in ten was aware of the existence of TFA in the product
category, 15% of the overall sample were surprised by this fact.
There was little difference by product type. One might have expected more people in the
‘healthier’ yoghurt category to be unaware of the presence of TFA. However, the empirical
evidence indicates that many consumers who had heard of TFA before did not really associate
it with any of the tested categories.
Likewise the comparison between the treatment groups does not show any significant impact.

181
Impact of Food Information on Consumers’ Decision Making

Figure 107 Surprise effect of TFA disclosure

Consumer reactions after feedback/education

Surprise effect by products and treatment splits

Heard of TFA Heard before

Never heard but not in this that TFA can
of TFA before category be in this cat.

Yoghurt shoppers
(n= 603) 74 17 9

Pizza shoppers
(n=615) 77 14 9

Biscuit shoppers
73 13 14
(n=633)

With TFA info and

education (n=596) 72 15 12

With TFA info only

(n=642) 79 12 10

Control group
73 17 10
(n=613)

Q54: As products sold in Germany are currently not labelled that way, we would like to understand how familiar the isue of trans
fatty acids is to you. Which of the following statements apply to you?
Base: all respondents (n=1851)

One in four participants felt confused after the experiments and the feedback. The confusion
was slightly higher among pizza shoppers and it was lowest among those who received the
education treatment.

182
Impact of Food Information on Consumers’ Decision Making

Figure 108 Confusion effect of TFA disclosure

Consumer reactions after feedback/education

Confusion effect by products and treatment splits

Agree Neither nor Disagree

strongly Agree or DK Disagree strongly
Product segmentation:

Yoghurt shoppers
(n= 603) 6 17 19 31 27

Pizza shoppers
(n=615)
12 19 17 26 27

Biscuit shoppers
4 16 13 39 28
(n=633)

Treatment segmentation:

With TFA info and

education (n=596) 3 16 23 37 21

With TFA info only

(n=642) 9 16 15 28 33

Control group
10 19 12 32 27
(n=613)

Q55_1: I am going to read out some statements that may or may not reflect how you see the issue on trans fat. Please tell me
how much you agree or disagree with each statement using a five-point-scale.
I am confused now – all this is too difficult for me.
Base: all respondents (n=1851)

Many ignore health aspects when making food choices. Almost 40% agreed that they didn’t
care about the presence of trans fats and made their choice on other factors. Pizza
shoppers were slightly more likely to agree with, whereas biscuit shoppers displayed the
highest share of disagreement.

183
Impact of Food Information on Consumers’ Decision Making

Figure 109 Strategy of ignorance on TFA disclosure

Consumer reactions after feedback/education

Strategy of ignorance by products and treatment splits

Agree Neither nor Disagree

strongly Agree or DK Disagree strongly
Product segmentation:

Yoghurt shoppers
(n= 603) 14 26 18 26 16

Pizza shoppers
(n=615)
15 34 16 26 10

Biscuit shoppers
9 20 16 40 15
(n=633)

Treatment segmentation:

With TFA info and

education (n=596) 11 28 18 21 12

With TFA info only

(n=642) 12 24 20 26 18

Control group
15 27 12 36 11
(n=613)

Q55_2: I am going to read out some statements that may or may not reflect how you see the issue on trans fat. Please tell me
how much you agree or disagree with each statement using a five-point-scale.
I don’t care about this – I make my choices on other aspects.
Base: all respondents (n=1851)

The intention to avoid products with high TFA levels in future was fairly high amongst biscuit
and pizza shoppers. By comparison, the proportion of yoghurt shoppers who intended to
reduce their yoghurt consumption in the future because of trans fat was significantly lower at
17%. These stated preferences are most likely inflated by an overconfidence bias, as there
were several consumers who agreed with both statements about ‘avoiding’ and ‘ignoring’ TFA.

184
Impact of Food Information on Consumers’ Decision Making

Figure 110 Strategy of avoidance on TFA disclosure

Consumer reactions after feedback/education

Strategy of avoidance by products and treatment splits

Agree Neither nor Disagree

strongly Agree or DK Disagree strongly
Product segmentation:

Yoghurt shoppers
(n= 603) 5 12 16 29 38

Pizza shoppers
(n=615)
22 36 13 18 11

Biscuit shoppers
22 42 17 11 8
(n=633)

Treatment segmentation:

With TFA info and

education (n=596) 12 30 15 25 19

With TFA info only

(n=642) 12 27 17 17 27

Control group
26 34 14 15 10
(n=613)

Q55_3: I am going to read out some statements that may or may not reflect how you see the issue on trans fat. Please tell me
how much you agree or disagree with each statement using a five-point-scale.
- If yoghurt split: I will reduce my yoghurt consumption in the future because of trans fat.
- If pizza/biscuit: I will try to avoid any [pizza/biscuits] with high trans fat levels in the future.
Base: all respondents (n=1851)

An interesting analysis is the comparison of treatment splits. Participants in the control group
(i.e. without any TFA information and education) were significantly more likely to say they
would avoid TFA than participants who had more transparency in the experiments.
The higher agreement level among the control group is again an indicator for an
overstatement bias. While people in the control group assumed they would act rationally in
the future, participants with the experience of transparency in the experiments were more
aware of how difficult it is to make a good decision.
When presented with a list of possible interventions, there were two clear favourites. A
strong majority wanted more education and for more transparency in the general labelling of
TFA amounts in nutrition facts.

185
Impact of Food Information on Consumers’ Decision Making

Figure 111 Preferred interventions by consumers

Preferred policy actions by consumers

By treatment splits

There should be more 73%

education for consumers
75%
in how to make better
food choices. 79%

Trans fat amounts should 63%

always be stated in 60%
the nutrition facts.
62%
Preferences by
TFA info + education
The industry should not 21%
With TFA info only
be allowed to sell any products 15%
Control group
with high levels of trans fat. 26%

The industry should not 18%

be allowed to sell any products 11%
with trans fat. 30%

8%
Trans fat amounts should
9%
only be stated if high.
7%

Q56: What should be done in your opinion, to help making good choic es easier for consumers like you?
Base: all respondents (n=1851) – multiple answers possible

For the remaining interventions there was one notable difference: Control group participants
were most likely to support the most extreme intervention to ban trans fats from all products
or to ban at least products with high TFA levels.
As mentioned before, the control group had less information, which normally makes it more
difficult to make a healthy choice for a rational decision maker. This may explain why these
participants were more likely to claim they would avoid trans fats in future (as in figure 106),
and without information, the easiest way to achieve this would be if the products would not
contain any (high levels of) trans fats.
Whether the top two preferred interventions are also more effective in making better food
choices, will be analysed in the following chapters focussing on the experimental choice
behaviour.

186
Impact of Food Information on Consumers’ Decision Making

9.2 Observed initial choice behaviour – disaggregated findings

The first choice set required ‘initial choices’ based on individual goals and habits. The choice
set consisted of two subsequent tasks each offering two different products. The question was
worded as follows: “Which product would you choose – A or B (X or Y), both or neither of
them?”
Task 1 presented as less healthy choice product A (with significantly more TFA) and as
healthier choice product B (with clearly less TFA). Overall, only 17% of all participants chose
the healthier product B in this task, while half of the participants selected the less healthy
product A.
Task 2 presented as the healthier choice product X (with slightly more TFA, but less saturates,
salt and sugar) and as less healthy choice product Y (with slightly less TFA, but more
saturates, salt and sugar). This choice scenario achieved a significantly higher share of
participants choosing the healthier product X (39%), whereas only 9% selected the less
healthy product Y. Around half avoided choosing by selecting both or none in this task.
It is worth stressing that the experimental choice context was overly simplified. In real life,
the choice context would be significantly more complex, as all elements in the nutritional
declaration are expected to differ among any two products. It might be expected that
consumers would be less likely to make the healthier choice in more complex choice contexts
although this assumption is beyond the scope of this particular piece of research.
Table 112 Initial choice behaviour in Tasks 1 and 2

INITIAL CHOICES Task 1 (A/B) Task 2 (X/Y)

No/less TFA is healthier With/more TFA is healthier
Base unweighted 1851 1851

Product B/X 17% 39%

healthier product

Product A/Y 49% 9%

less healthy product

Both 25% 35%

None 10% 17%

Total 100% 100%

In both choices, two thirds of the sample received information about TFA amounts if they read
the nutrition labels. The remaining third had to identify the healthier product by looking for
partially hydrogenated fats in the ingredients list and understand its meaning. Therefore, the
observed choice behaviour displayed in the table above is a result of a mix of treatments and
product categories.

187
Impact of Food Information on Consumers’ Decision Making

Across all three product categories, the level of initial healthy product choices was significantly
higher in the second task compared to the first task. This may suggest that differences in the
level of trans fats (or the existence of PHO) are less relevant for healthy product choices than
more familiar elements like saturates, sugar or salt.
The experiments with biscuits displayed significantly more healthy product choices in both
tasks, whereas the yoghurt/dessert experiments had the fewest healthy product decisions.

Table 113 Initial choice behaviour in Task 1 by product category

INITIAL CHOICE All Frozen

Task 1 categories Yoghurt pizza Biscuits
Base unweighted 1851 603 615 633

Product B –
healthier product 17% 8% 16% 25%
(no/less TFA)

Product A 49% 80% 26% 42%

(with/more TFA)

Both 25% 5% 51% 18%

None 10% 7% 7% 15%

Total 100% 100% 100% 100%

Table 114 Initial choice behaviour in Task 2 by product category

INITIAL CHOICE All Frozen

Task 2 categories Yoghurt pizza Biscuits
Base unweighted 1851 603 615 633

Product X –
healthier product 39% 28% 36% 52%
(with/more TFA)

Product Y 9% 6% 7% 14%
(no/less TFA)

Both 35% 37% 51% 18%

None 17% 29% 6% 17%

Total 100% 100% 100% 100%

188
Impact of Food Information on Consumers’ Decision Making

9.2.1 Rationale given for initial choices

Approximately half of those who selected the healthier product B also gave a correct
explanation of their choice, compared with only 25% who mentioned TFA or PHO in their
reasoning. The lower amount of saturated fat was mentioned slightly more often than the
level of TFA or PHO.

Figure 115 Rationale for initial choice Task 1 if product A or B was chosen

Rationale for initial choice task 1

if A or B was chosen
correct recall
Recall of rationale if … chosen for product B
30% Healthier product B 50% (net sum)
Saturates were lower 2% Less healthy product A

18%
TFA was lower 1%
25% (net sum)
11%
No/Less PHO 1%

17%
Healthier 3%

6%
Sugar was lower 19%

2%
Salt was lower 13%
75% (net sum)

correct recall
I don’t like soy 44% for product A

Ingredients were 21%

better 34%

44%
Other / Don’t know 28%

Q33: Why did you choose this produc t?

Base: respondents choosing product A (n=907) or product B (n=308) in initial choice task 1

In comparison, the rationale for the choice of the less healthy product A was based on a higher
proportion of concrete and correct recall. Whereas the majority of the yoghurt shoppers simply
disliked any soy dessert, pizza shoppers identified the difference in salt and biscuit shoppers
the difference in sugar.

189
Impact of Food Information on Consumers’ Decision Making

In Task 2 the higher share of healthy choices also revealed a high level of correct reasoning.
70% selected product X because it had less sugar, saturates or salt.
The expected overreaction of picking product Y because it contained less TFA/PHO was only
observed in a tiny proportion of those who selected this product type.

Figure 116 Rationale for initial choice Task 2 if product X or Y was chosen

Rationale for initial choice task 2

if X or Y was chosen
correct recall
Recall of rationale if … chosen for product X

37% Healthier product X 70% (net sum)

Sugar was lower 4% Less healthy product Y

26%
Saturates were lower 7%
70% (net sum)

Salt was lower 22%

1%

7%
Healthier 6%

4%
TFA was lower 4%
11% (net sum)
4% (net sum)
2% correct recall
No/Less PHO 2% for product Y

Ingredients were 9%
better 11%

29%
Other / Don’t know 80%

Q39: Why did you choose this produc t?

Base: respondents choosing product X (n=716) or product Y (n=169) in initial choice task 2

The most common explanation that was given for choosing neither or both products in both
tasks is that there was no difference between the products.

9.2.2 Potential access and awareness drivers

Accessing and understanding the relevant food information is a key requirement for making
informed choices. Therefore we monitored in the experiments which information was accessed
for each choice. The descriptive overview below indicates that the accessing behaviour is a
strong driver (in line with the previous online study).
 Respondents who looked at irrelevant or no specific information were least likely to
choose the healthier product.
 Reading relevant food information like nutrition facts, fat (or sugar or salt in Task 2)
seemed to drive initial healthy product choices. Only reading the ingredients list
appeared to be less helpful.

190
Impact of Food Information on Consumers’ Decision Making

Table 117 Initial choice behaviour in Task 1 by accessed information

INITIAL CHOICE
Ingre- Nutrition Nothing/
Task 1
Total dients list facts Fat Sugar Salt irrelevant
Base unweighted 1851 572 409 209 229 148 1139

Product B –
healthier product 17% 20% 29% 61% 15% 14% 10%
(no/less TFA)

Product A 49% 34% 48% 23% 73% 75% 51%

(with/more TFA)

Both 25% 35% 16% 9% 5% 7% 26%

None 10% 11% 7% 6% 7% 3% 12%

Total 100% 100% 100% 100% 100% 100% 100%

Table 118 Initial choice behaviour in Task 2 by accessed information

INITIAL CHOICE
Ingre- Nutrition Nothing/
Task 2
Total dients list facts Fat Sugar Salt irrelevant
Base unweighted 1851 566 491 286 306 172 1090

Product X –
healthier product 39% 33% 66% 66% 84% 90% 25%
(with/more TFA)

Product Y 9% 8% 8% 8% 5% 3% 10%
(no/less TFA)

Both 35% 43% 18% 7% 4% 6% 43%

None 17% 16% 8% 19% 8% 1% 23%

Total 100% 100% 100% 100% 100% 100% 100%

In addition to the information accessed, the regression model will also include the awareness
of the (un)healthiness of nutrition elements, which was collected after the initial choice tasks.
Although the descriptive analysis did not reveal any significant differences, these variables
were relevant drivers in the online survey.

9.2.3 Potential goals and habitual drivers

Individual shopping goals were collected at the beginning of the interview. For the multivariate
analysis we included only a selection of relevant goals with a significant impact in the bivariate
analysis. These were:
 Health orientation (Q27_3)
 Convenience orientation (Q27_4)

191
Impact of Food Information on Consumers’ Decision Making

 Novelty – I like looking for new and exciting products (Q28_1)

 Variety - I want a lot of variety such as sizes, flavours, forms, etc. (Q28_2)
 Rush - I want to get in and out of the section as fast as I can (Q28_4)
 Food information reading for making right choices (Q28_5)
Further questions and observations at the shelf covered potential habitual factors that may
have influenced the shoppers decisions. The following aspects were thus included in the final
model:
 Duration of the observed shopping process (Q4-9)
 Shelf behaviour (Q10): reading vs. grab & go
 Reading fat, sugar, nutrition facts, ingredients at the shelf (Q25/26)
 Influenced by shopping company (Q11/12)
 General frequency of shopping the category (Q16)
 Planned to buy category before entering the store (Q22)
 Bought specific product type or brand before (Q24)

Moreover, the usual set of socio-demographics was added as control variables to the analytical
model, such as gender, age, household size, working status, education level and income.

9.2.4 Treatments

The analysis of the initial choices focuses on comparing the control group (not given any TFA
information) with the combined treatments 2 and 3 (with TFA amounts displayed on the
nutrition table). As the additional education intervention (treatment 3) was only given after
the initial choices, this differentiation does not make sense to include in this chapter.
When comparing the descriptive results for Task 1 in the following tables, there is no
significant impact on the initial choice of the healthier product B by showing the
amount of TFA. However, the multivariate analysis will show in Chapter 9.4.1 that in task 1
the display of TFA amounts does have a significant and positive impact on initial healthy
choices. This is an indicator that the results of the descriptive analysis are changed by
controlling for other variables.

192
Impact of Food Information on Consumers’ Decision Making

Table 119 Initial choice behaviour in Task 1 by treatments

INITIAL CHOICE Control group: Treatment 2/3:

Task 1 Total no TF info with TF info
Base unweighted 1851 613 1238

Product B –
healthier product 17% 17% 16%
(no/less TFA)

Product A 49% 53% 47%

(with/more TFA)

Both 25% 20% 27%

None 10% 9% 10%

Total 100% 100% 100%

Table 120 Initial choice behaviour in Task 1 by treatments and products

INITIAL CHOICE Control group: Treatment 2/3:

= B (Task 1) Total no TF info with TF info

All categories 17% 17% 16%

Yoghurt/Dessert 8% 6% 9%

Frozen Pizza 16% 18% 15%

Biscuits 25% 27% 24%

The yoghurt vs. soy dessert choice showed a slight tendency towards the healthier soy dessert
if TFA amounts are displayed, although not a statistically significant one. Likewise the
differences between the initial choices within the categories of frozen pizza or biscuits was
again not significant.
As a reminder: The scenario of Task 1 was designed with the expectation of a positive impact
of TFA transparency as the product with less TFA was healthier (the level of TFA weighed
more than the smaller deviations in sugar or salt).

193
Impact of Food Information on Consumers’ Decision Making

When comparing the descriptive results for Task 2 in the following tables, there was a slightly
lower proportion of consumers who chose the healthier product B when TFA amounts were
presented – in particular within the pizza category.64
As a reminder: The scenario of Task 2 was designed with the expectation of a negative impact
of TFA transparency as the product with more TFA is healthier (the level of TFA was
outweighed by higher levels of saturates and sugar or salt).

Table 121 Initial choice behaviour in Task 2 by treatments

INITIAL CHOICE
Control group: Treatment 2/3:
Task 2
Total no TF info with TF info
Base unweighted 1851 613 1238

Product X –
healthier product 39% 45% 36%
(with/more TFA)

Product Y 9% 8% 10%
(no/less TFA)

Both 35% 31% 37%

None 17% 16% 18%

Total 100% 100% 100%

Table 122 Initial choice behaviour in Task 2 by treatments and products

INITIAL CHOICE Control group: Treatment 2/3:

= X (Task 2) Total no TF info with TF info

All categories 39% 45% 36%

Yoghurt/Dessert 28% 27% 28%

Frozen Pizza 36% 52% 27%

Biscuits 52% 55% 50%

However, only the multivariate analysis can establish whether the treatment had a direct and
significant impact in both choices or whether these bivariate results were influenced by other
factors.

64
If controlling for other relevant variables in the multivariate analysis (see Chapter 9.4.1) there is no
statistically significant impact on healthy choices by displaying the TFA amount in task 2.

194
Impact of Food Information on Consumers’ Decision Making

9.3 Observed ability to identify healthy products – disaggregated

analysis

The second choice set required participants to identify the healthier products by using the
same product sets as in the initial choices. The question wording was as follows: “Which
product is healthier – A or B (X or Y) or are they both equally healthy or equally unhealthy?”
It was also noted where the participant was not able to give an answer.
The Tasks 1 and 2 were identical to the initial choices. The only difference was the more
cognitive decision-making. While the initial choices reflected more habitual choice behaviour
without stipulating a health goal for everyone, the second choice set was solely to find out
whether consumers were able to identify a healthier product or not.
Overall, the healthy product was identified more frequently in this ‘knowledge test’ than it
was observed in the initial choices, but the pattern of a higher proportion of “correct” choices
in the Task 2 than Task 1 was the same.
 In Task 1 40% of all participants identified the healthy product B (with less TFA), while
one third of the participants selected the less healthy product A.
 In Task 2 54% correctly selected the healthier product X (with more TFA, but less
saturates, salt or sugar), whereas only 7% selected the less healthy product Y.

Table 123 Identification of healthy products in Task 1 and 2

CORRECT choice Task 1 (A/B) Task 2 (X/Y)

of healthier product No/less TFA is healthier With/more TFA is healthier
Base unweighted 1851 1851

Product B/X
40% 54%
healthier product

Product A/Y
32% 7%
less healthy product

Equally healthy 6% 8%

Equally unhealthy 16% 25%

Don’t know 6% 6%

Total 100% 100%

In both choices, one third of the sample received the education intervention and the TFA
amounts on the products, one third received only the information about TFA amounts (without
education). The remaining third had to identify the healthier product by looking for partially
hydrogenated fats in the ingredients list and understanding its meaning. Therefore, the
observed choice behaviour displayed in the table above is a result of a mix of treatments and
product categories.

195
Impact of Food Information on Consumers’ Decision Making

Although the average level of correct choices was higher in Task 2, this pattern varied when
comparing the choices between the three product categories. Yoghurt shoppers identified the
healthy product more often in Task 1 than in Task 2. Biscuits and pizza shoppers were more
likely to choose the healthier product in Task 2 than in Task 1.
Another interesting observation was that Task 1 led to more decisions favouring the less
healthy product, whereas Task 2 led to more consumers saying that both products are equally
unhealthy.

Table 124 Identification of healthy product in Task 1 by product category

CORRECT Choice
All Frozen
Task 1
categories Yoghurt pizza Biscuits
Base unweighted 1851 603 615 633

Product B –
healthier product 40% 54% 34% 32%
(no/less TFA)

Product A
32% 27% 25% 43%
(with/more TFA)

Equally healthy 6% 9% 7% 4%

Equally unhealthy 16% 4% 29% 14%

Don’t know 6% 6% 5% 7%

Total 100% 100% 100% 100%

Table 125 Identification of healthy product in Task 2 by product category

CORRECT Choice
All Frozen
Task 2
categories Yoghurt pizza Biscuits
Base unweighted 1851 603 615 633

Product X –
healthier product 54% 38% 54% 69%
(with/more TFA)

Product Y
7% 6% 6% 8%
(no/less TFA)

Equally healthy 8% 16% 7% 3%

Equally unhealthy 25% 32% 30% 14%

Don’t know 6% 8% 4% 6%

Total 100% 100% 100% 100%

196
Impact of Food Information on Consumers’ Decision Making

9.3.1 Rationale given for initial choices

Approximately two thirds of those who selected the healthier product B also gave a correct
rationale for their choice, compared with 36% who mentioned TFA or PHO in their reasoning.
The lower amount of saturated fat was significantly more likely to be mentioned than the
deviation in levels of TFA or PHO.

Figure 126 Rationale for healthy product identification in Task 1

Rationale for healthy product identification

in task 1 (A vs. B)
correct recall
for product B
53% 65% (net sum)
Saturates were lower 6% Recall of rationale if … chosen

32% product B (healthier)

TFA was lower 2% product A
36% (net sum)
6%
No/Less PHO 1%

5%
Sugar was lower 49%
70% (net sum)
1% correct recall
Salt was lower 21% for product A

Ingredients were 13%

8%
better
31%
Other / Don’t know 29%

Q52: Why did you think this product is healthier?

Base: respondents choosing product A (n=589) or product B (n=743) in healthy choice task 1

The reasoning for choosing the less healthy product A was based on a slightly higher
proportion of correct recall. Pizza shoppers identified the small deviations in salt. Biscuit and
yoghurt shoppers registered the deviations in sugar.
The higher proportion of healthy choices in Task 2 also came with a notably higher proportion
in correct reasoning. Nine out of ten (92%) selected product X because it had less sugar,
saturates or salt.
The expected reaction of picking product Y because it contained less TFA/PHO was observed
for around a third of the 7% who selected this product type.

197
Impact of Food Information on Consumers’ Decision Making

Figure 127 Rationale for healthy product identification in Task 2

Rationale for healthy product identification

in task 2 (X vs. Y)
correct recall
for product X

54% 92% (net sum)

Sugar was lower 9% Recall of rationale if … chosen

48% product X (healthier)

Saturates were lower 16% product Y

Salt was lower 28%

2%

7%
TFA was lower 30%
30% (net sum)
2%
No/Less PHO correct recall
5% for product Y

Ingredients were 4%
better 9%

11%
Other / Don’t know 55%

Q47: Why do you think this produc t is healthier?

Base: respondents choosing produc t X (n=1000) or produc t Y (n=122) in healthy choic e task 2

9.3.2 Potential access and awareness drivers

The monitoring of the information accessed by participants was adapted slightly for the second
choice set:
 The newly presented education leaflet was added, i.e. whether it was read again while
making a choice.
 The differentiation between ingredients list and nutrition facts was repeated as in the
previous choice set.
 Not accessing any information or accessing information that was not relevant for
evaluating the healthiness of the products was captured as well.
The descriptive overview below indicates only a few significant differences, i.e. not accessing
any relevant information reduces the likelihood of making the healthy choice. Nutrition facts
were again a strong potential driver of good choices, in particular for Task 2.
As only very few participants voluntarily used the education leaflet during their choices, the
differences for these were not significant.

198
Impact of Food Information on Consumers’ Decision Making

Table 128 Healthy product identification in Task 1 by accessed information

CORRECT Choice
All Education Ingredient Nutrition None/
Task 1
categories leaflet list facts Other
Base unweighted 1851 6665 454 1185 592

Product B –
healthier product 40% 52% 38% 44% 32%
(no/less TFA)

Product A 32% 30% 29% 39% 21%

(with/more TFA)

Equally healthy 6% 5% 10% 4% 9%

Equally unhealthy 16% 12% 17% 11% 26%

Don’t know 6% 2% 6% 2% 12%

Total 100% 100% 100% 100% 100%

Table 129 Healthy product identification in Task 2 by accessed information

CORRECT Choice
All Education Ingredient Nutrition None/
Task 2
categories leaflet list facts Other
Base unweighted 1851 6866 464 1266 555

Product X –
healthier product 54% 57% 47% 73% 19%
(with/more TFA)

Product Y 7% 18% 7% 7% 6%
(no/less TFA)

Equally healthy 8% 0% 12% 3% 15%

Equally unhealthy 25% 24% 30% 15% 46%

Don’t know 6% 1% 4% 2% 15%

Total 100% 100% 100% 100% 100%

65
This sample base is very low, therefore any differences are not statistically significant.
66
This sample base is very low, therefore any differences are not statistically significant.

199
Impact of Food Information on Consumers’ Decision Making

Further awareness variables are also included in the analytical model as one would expect
also in this experiment that previous knowledge about the healthiness of certain elements
could drive or hinder making the correct choice. However, the descriptive analysis did not
reveal any significant differences except that those who did perceive sugar as healthy had a
reduced likelihood of making healthier choices in Task 2.

9.3.3 Potential goals and habitual drivers

For contextual and comparison reasons, the analytical model consists of the same set of
control variables representing selected shopping goals, habits and socio-demographics, which
are
 Health orientation (Q27_3)
 Convenience orientation (Q27_4)
 Novelty – I like looking for new and exciting products (Q28_1)
 Variety - I want a lot of variety such as sizes, flavours, forms, etc. (Q28_2)
 Rush - I want to get in and out of the section as fast as I can (Q28_4)
 Food information reading for making right choices (Q28_5)
 Duration of the observed shopping process (Q4-9)
 Shelf behaviour (Q10): reading vs. grab & go
 Reading fat, sugar, nutrition facts, ingredients at the shelf (Q25/26)
 Influenced by shopping company (Q11/12)
 General frequency of shopping the category (Q16)
 Planned to buy category before entering the store (Q22)
 Bought specific product type or brand before (Q24)
 Socio-demographics: gender, age, household size, working status, education level and
income

9.3.4 Treatments

The analysis of the identification of the healthy product allows a more differentiated approach
than the initial choices because the education intervention served as an additional treatment.
The expectation was that participants who received the education would be able to make
better choices with fewer knowledge gaps and overreactions or misunderstandings.
Since the duration of the education intervention varied by participant (they were able to read
the leaflet for as long or as short a time as they wanted to), this treatment can also be
analysed by two additional splits:
 A short education (< 45 seconds)
 A long education (45 seconds or more)
The length of time that a participant took when reading the nutrition guideline can be an
indicator of interest and commitment to this topic. Therefore, one would expect that the longer
education should lead to more good choices.
When comparing the descriptive results for Task 1 in the following tables, there seems to be
a strong and positive impact of the education treatment as well as a positive impact of showing
TFA amounts without education. This was expected in this scenario.

200
Impact of Food Information on Consumers’ Decision Making

Table 130 Healthy product identification in Task 1 by treatments

Treat 2: Treat 3: Treat 3a: Treat 3b:

CORRECT Choice
Control only TF info, with with short with long
Task 1
Total group no edu. education education education
Base unweighted 1851 613 642 596 264 207

Product B –
healthier product 40% 34% 38% 49% 42% 54%
(no/less TFA)

Product A 32% 40% 36% 19% 17% 20%

(with/more TFA)

Equally healthy 6% 8% 7% 4% 5% 4%

Equally unhealthy 16% 13% 12% 23% 28% 20%

Don’t know 6% 6% 7% 5% 7% 3%

Total 100% 100% 100% 100% 100% 100%

Table 131 Healthy product identification in Task 1 by treatments and products

Treat 2: Treat 3: Treat 3a: Treat 3b:

CORRECT Choice Control only TF info, with with short with long
= B Task 1 Total group no edu. education education education

All categories 40% 34% 38% 49% 42% 54%

Yoghurt/Dessert 54% 47% 53% 62% 57% 66%

Frozen pizza 34% 31% 30% 42% 39% 47%

Biscuits 32% 24% 32% 42% 25% 49%

When comparing the descriptive results for Task 2 in the following tables, one can see a
negative impact for consumers who had only a short education intervention.
The expectation was that treatment 2 (showing TFA info without any education) could lead to
overreactions to TFA and incorrect choices as observed in the online survey. This did not
happen in the store experiments across all product categories. Moreover, the education
treatment did not have the expected positive impact overall.

201
Impact of Food Information on Consumers’ Decision Making

Table 132 Healthy product identification in Task 2 by treatments

Treat 2: Treat 3: Treat 3a: Treat 3b:

CORRECT Choice
Control only TF info, with with short with long
Task 2
Total group no edu. education education education
Base unweighted 1851 613 642 596 264 207

Product X –
healthier product 54% 56% 57% 49% 38% 58%
(with/more TFA)

Product Y 7% 6% 6% 8% 8% 9%
(no/less TFA)

Equally healthy 8% 11% 9% 5% 7% 4%

Equally unhealthy 25% 23% 19% 33% 42% 27%

Don’t know 6% 4% 10% 4% 6% 2%

Total 100% 100% 100% 100% 100% 100%

Table 133 Healthy product identification in Task 2 by treatments and products

Treat 2: Treat 3: Treat 3a: Treat 3b:

CORRECT Choice Control only TF info, with with short with long
= X Task 2 Total group no edu. education education education

All categories 54% 56% 57% 49% 38% 58%

Yoghurt/Dessert 38% 40% 31% 43% 36% 49%

Frozen pizza 54% 64% 57% 42% 34% 51%

Biscuits 69% 62% 78% 64% 47% 71%

When looking at the choice behaviour for each product category, the expected effects were
only seen among yoghurt shoppers but not the other categories.
However, only multivariate analysis can establish whether these patterns were influenced by
other factors or whether they represent the final empirical evidence.

9.4 Findings of the multivariate analysis

In this section we present the summary of the multivariate analysis conducted using binary
logistic regression to identify the major factors which influenced the selection of the healthier
product in the choice experiments. ‘Healthy product choices’ are defined as “respondents
choosing product B” in Task 1 (product X in Task 2).

202
Impact of Food Information on Consumers’ Decision Making

The previous chapters already discussed the potential variables to be included within the
model in detail. The selection was based on conceptual considerations and in-depth analysis
of the bivariate associations between the dependent and independent variables.
There is one model for each task, i.e. four models overall. We tested binomial and multinomial
models with and without interaction effects. Since the more complex models did not deliver
additional insights, we present below the binominal version without interaction effects. Each
model was run with and without treatments to examine any differences. As this comparison
did not deliver any relevant changes, the models presented in the following chapters are all
including treatments.

9.4.1 Drivers of initial healthy product choices in Task 1

The research question for this scenario was as follows: Does TFA transparency improve
participants' ability to identify the healthy choices when the product with no or less TFA is
healthier than the alternative?
Figure 134 Drivers of initial choice of healthy product in Task 1

Drivers of initial choice of healthy product B R2:

34
%

in task 1 Model with policy option

Reg.
coeff. Policy Context Access Goal Habits
3,0 Ref. cat. biscuits
Fat
nutrition table

Yoghurt/Dessert

Reading ingredients at
TFA amount on

Frozen Pizza

Irrelevant labels
or nothing

Strong convenience
Sugar

orientation

the shelf

Influenced by shopping
company at the shelf

2,5

2,0

1,5

1,0

0,5

0,0 * *** *** *** *** *** ** * *

Extract of results based on binary logistic regression analysis to identify statistically significant drivers of initial healthy product choices in task 1
(Q29-31). The model explains 34% of the variance. Only variables with at least slight significance (* = p < 0.05) are displayed.
Base: Store experiments in German supermarkets (n = 1851)

The analysis of the store experiments across all product categories shows a slightly significant
positive impact on making healthier initial choices with more TFA transparency.
Other positive drivers were
 Accessing information about saturates when making the decision

203
Impact of Food Information on Consumers’ Decision Making

 Habitual reading of ingredients at the shelf (as a “first buyer”)

 Being influenced while shopping by family members or friends when selecting products
at the shelf
Typical barriers that prevented participants from selecting the healthier product were
 Accessing information about sugar (which was slightly higher in the healthier
product)67
 Not accessing any information or accessing information that was irrelevant to evaluate
the healthiness of the products
 A strong convenience orientation when making shopping decisions
Moreover, the product category influenced whether consumers selected the healthy product
or not. The decision between two biscuit products was more likely to deliver a healthy choice
in Task 1 than pizza or desserts.

9.4.2 Drivers of initial healthy product choices in Task 2

The research question for this scenario was as follows: Does TFA transparency reduce healthy
choices when the product with more TFA is healthier than the alternative?

67
The access of information on amounts of salt did not show any statistically significant impact on making
healthy choices in this task1.

204
Impact of Food Information on Consumers’ Decision Making

Figure 135 Drivers of initial choice of healthy product in Task 2

Drivers of initial choice of healthy product X R2: 50

in task 2 Model with policy option

%

Reg.
coeff. Policy Context Access & Awareness Goals Socio-demo.
3,0 Ref. cat. biscuits

Nutrition table

Fat

Ingredients list
nutrition table
TFA amount on

Weak health goal orientation

Frozen Pizza

Salt

No awareness about

Variety seeker
Irrelevant labels
Yoghurt

or nothing
Sugar

unhealthiness of sugar

Lower education
Single households
2,5

2,0

1,5

1,0

0,5

n.s.
0,0 *** *** *** *** *** ** *** *** * * * ** *
Extract of results based on binary logistic regression analysis to identify statistically significant drivers of initial healthy product choices in task 2
(Q35-37). The model explains 50% of the variance. Only variables with at least slight significance (* = p < 0.05) are displayed.
Base: Store experiments in German supermarkets (n = 1851)

The analysis of the store experiments across all product categories shows no significant impact
on making healthier initial choices with more TFA transparency. Overall, there was no
“overreaction” observed to the TFA level.
Positive drivers of initial healthy product choices in Task 2 were
 Accessing the relevant elements in the nutrition table, i.e. the amounts of salt, sugar
and saturates
 The shopping objective of looking at a wide variety of product types also helped in
selecting the healthier product.
Barriers or negative drivers in this scenario were
 Not accessing any information or accessing the ingredients list or food information that
was irrelevant to the healthy choice
 Not being aware of the negative health aspects of sugar
 Not generally considering any health aspects when making food choices
 Living in a single household or a lower education level also hindered a healthy choice.
Again, the context of the product category influenced this choice in a similar way as in the
first choice scenario. The healthier product was selected more often when comparing biscuits
rather than yoghurt or pizza.

205
Impact of Food Information on Consumers’ Decision Making

9.4.3 Drivers of the ability to identify healthy products in Task 1

The research question for this chapter was as follows: Are consumers able to identify the
healthier product with no or less TFA if there is more transparency?
Participants were presented again with the same product sets as in the initial choices, after
now having evaluated the healthiness of individual nutrition elements and after one third of
the sample had received an additional education treatment. The purpose in this chapter is to
look at the consumers’ ability to identify the healthier product.

Figure 136 Drivers of the ability to identify healthy products in Task 1

Drivers of correctly identifying the healthier 18

R2: %

product B in task 1 Model with policy option

Reg. Con- Access & Awareness
coeff. Policy options Goal Socio.
text
3,0 Ref. cat.
biscuits
Yoghurt/Dessert

Nutrition table

Ingredients list
TFA amount on nutrition
table & short education

TFA amount on nutrition

table & long education

convenience orientation
Either weak or strong

Men
Only TFA amount on nutrition
table & no education

No awareness about

Irrelevant labels
or nothing
unhealthiness of PHO

2,5

2,0

1,5

1,0

0,5

0,0 * * *** *** *** * * *** ** *

Extract of results based on binary logistic regression analysis to identify statistically significant drivers of the ability to identify the healthy
product in task 1 (Q44-46). The model explains 18% of the variance. Only variables with at least slight significance (* = p < 0.05) are
displayed. Base: Store experiments in German supermarkets (n = 1851)

The analysis of the store experiments across all product categories shows a significant positive
impact of all treatments in comparison to the control group. The gradual variation of the
differentiation shows clearly that the more transparency the higher the impact.
A further driver of correct choices was accessing the nutrition table when making a decision.
Key barriers were
 Not accessing any information or looking at irrelevant information
 The use of the ingredients list hindered healthy choices, which is in line with the other
barrier of people not knowing that PHO is unhealthy
 Having either a weak or a strong convenience orientation also hindered good choices.

206
Impact of Food Information on Consumers’ Decision Making

 Men had more problems identifying the healthier product in Task 1.

Within the context of yoghurt products, it was easier for consumers to identify the healthier
product as compared to biscuits.

9.4.4 Drivers of the ability to identify healthy products in Task 2

The research question for this chapter was as follows: Are consumers able to identify the
healthier product if it has slightly more TFA but an overall healthier combination of saturates,
sugar and salt? Overreactions were probable if only TFA amounts are shown without education
treatment, however they were not observed (see figure 133 above).

Figure 137 Drivers of the ability to identify healthy products in Task 2

Drivers of correctly identifying the healthier R2: 46

%
product X in task 2 Model with policy option
Reg.
coeff. Policy options Context Access & Awareness Socio.
3,0 Ref. cat. biscuits
Nutrition table

Ingredients list

No awareness about
table & long education
TFA amount on nutrition
TFA amount on nutrition
table & short education

Frozen Pizza

unhealthiness of sugar
Irrelevant labels
Yoghurt

or nothing

Men
Only TFA amount on nutrition
table & no education

2,5

2,0

1,5

1,0

0,5
n.s.
n.s.
0,0 * *** *** *** *** ** *** * *
Extract of results based on binary logistic regression analysis to identify statistically significant drivers of the ability to identify the healthy product in
task 2 (Q49-51). The model explains 46% of the variance. Only variables with at least slight significance (* = p < 0.05) are displayed.
Base: Store experiments in German supermarkets (n = 1851)

The analysis of the store experiments across all product categories shows no overreaction to
the display of TFA amounts. Moreover, the education treatment did not improve consumer
choices as expected. In fact, if the guideline was only scanned briefly by participants, then
the impact was negative.
The only positive driver of good choices in this experiment was looking at the nutrition table.
Key barriers of good choices were

207
Impact of Food Information on Consumers’ Decision Making

 Not accessing any information or accessing the ingredients list or food information that
was not relevant for making healthy choices
 Not being aware of the unhealthiness of sugar
 Men were less able to make good choices also in this task.
The context of product categories shows the same pattern as in the initial choice set. Biscuits
decisions delivered better results than yoghurt or pizza decisions.

9.5 Conclusions regarding TFA

Will the display of TFA amounts on the nutrition label lead to more healthy
choices or will people misunderstand and over-react?

The treatment “TFA amounts on nutrition table” has a small positive effect in scenario 1
(product with less TFA is healthier) in the initial choice as well as a somewhat bigger positive
effect in the more conscious choice.
In scenario 2 (product with more TFA is healthier) there was no overreaction observed as
seen in the online survey. A possible explanation is the more complex choice architecture as
well as the offline field environment in this experiment. Making complex choices in a busy
supermarket, people are less likely to be influenced by information that they are unfamiliar
with (such as TFA).
Therefore, the empirical evidence of the store experiments are more suitable to predict real
consumer behaviour in the initial implementation phase. However, the more familiar
consumers become with the term TFA or PHO (through communication campaigns), the more
likely future consumer reactions might involve overreactions to the presence of TFA as
observed in the online survey.

The empirical evidence from the store experiments reveals only slightly


significantly more healthy choices and no overreactions in 2 out of 3
products when TFA is disclosed in the nutrition information. Overreactions
may increase if consumers become more familiar with the terms TFA or
PHO.

Does the education intervention (presented in addition to the TFA amounts

on the nutrition table) help consumers to identify the healthier product?

The education treatment was effective in scenario 1 but not in scenario 2, where it was even
counterproductive if the participant only superficially scanned the leaflet. This contrasting
effect of the education highlights the impact of the different choice architectures in Task 1
and 2.
Task 2 was the easier and more familiar task for consumers who did not see or who ignored
the TFA information. Those who received the education leaflet but did not read it thoroughly,
were actually making worse choices. Hence, the expected overreaction actually was more

208
Impact of Food Information on Consumers’ Decision Making

likely where consumers were alerted by the leaflet and subsequently looked for TFA
information (which they did not do in their initial choices).
In contrast, Task 1 required participants only to focus on TFA and ignore (or correctly
evaluate) all other elements. In such a scenario it is not surprising that the more transparency
and the longer the exposure to the education treatment, the better the consumer choices.
This task required people to evaluate an unfamiliar term (TFA/PHO), which is why fewer
consumers made healthier choices overall. However, the education message that TFA is very
unhealthy was understood more quickly and had a more effective impact on this choice
scenario.


The education treatment generated an ambivalent impact on consumer
choices in the store experiments: either significantly positive or negative
– depending on the choice scenario and the duration time of reading the
nutrition guideline.

Which conclusions can be drawn from the evidence collected in the online
study and the store experiments?

The empirical evidence in the store experiments supports the results of the online study,
though the experiments are not exactly comparable.
Making healthy food choices is predominantly driven by accessing the relevant food
information. In a supermarket less than a quarter of shoppers actually take time and read
selected information on the package. Of these, only a fraction of consumers read relevant
food information that informs about the healthiness of the product.
However, we also know, that consumers often shop under time pressure and that a more
intense reading of food information can also happen at home. Communication campaigns,
discussions among friends and family members can change awareness and shopping
behaviour over time.
When asked directly, consumers said they would prefer to receive more education on this
topic and more transparency, i.e. the TFA amounts should be stated on the nutrition tables.
These two measures, however, are only effectively driving good choices, if the products with
less TFA are overall the healthier option.
Consumers are regularly overconfident that with more transparency they will be able to make
better choices, but the task to determine which combination of nutritional elements is the
healthier option is often more difficult than expected. Therefore, many consumers simplify
their food decisions by using heuristics, such as the availability bias68, habits, etc.
The initial choice for Task 2 revealed that most consumers simply did not recognise the less
familiar TFA/PHO information. Only after the educational nudge, this information was noticed
with the outcome of an overreaction among those who only read the leaflet superficially. This
scenario can be compared with Task 2 in the online survey where a similar overreaction was

68
i.e. relying on immediate examples that come to mind

209
Impact of Food Information on Consumers’ Decision Making

measured in a more controlled environment. Here, the treatment “TFA info without education”
was less liked to be ignored leading to consumer overreaction.
Overall, the objective of making healthier food choices easier to consumers is as important
as it is difficult to implement. It requires a high level of expertise, time and commitment to
correctly evaluate the presented information. The additional information about trans fats only
adds another variable to an already complex decision.
This report focuses only on consumer reactions to TFA transparency.

 The provision of TFA information and education were not consistently

effective in driving good choices.

210
Impact of Food Information on Consumers’ Decision Making

APPENDIX A – TECHNICAL DETAILS OF THE SURVEYS

211
APPENDIX A – TECHNICAL DETAILS OF THE SURVEYS

Online survey – technical summary

The online survey was carried out in eight EU Member States. The study was conducted
by means of CAWI (Computer Assisted Web Interviews). All participants were invited by
email to participate in the online survey. Invitations were sent out at the beginning of
fieldwork, with further emails sent out during the course of the fieldwork period.
Respondents who did not respond to these emails were re-invited by email.

The fieldwork was conducted in June 2012.

A target sample size of approximately 1000 respondents was set for each country. The
table below shows the achieved sample size.

Sample size of online survey

COMPLETED
COUNTRY SURVEYS

France 1 007

Germany 1 005

Italy 1 005

Spain 1 002

UK 1 027

Finland 1 004

Poland 1 003

Romania 1 023

Total 8 076

Once fieldwork was completed, a data file for each country was generated following a
specific data map.

To produce tables and other outputs based on the data set, the data was weighted
according to target figures for gender and age distribution in each country. We applied
rim weighting, using an iterative procedure to achieve an even distribution of results
across the entire dataset while balancing the gender and age figures to pre-determined
totals. It simultaneously weights the specified characteristics and disturbs each variable
as little as possible.

The country selection for the online study on food labels reflected a maximum population
coverage criteria as well as, based on our pre-analysis of available market indicators,
maximum heterogeneity coverage principle.
 Country selection criteria

Family of Population Food quality Label

Country Nations size label checking comprehension

France Central 48.754.823 Average Average

Germany Central 66.451.766 Less often Average

Italy Southern 48.517.748 More often Low

Spain Southern 36.793.761 Average Low

UK Anglo-Saxon 47.508.811 Average High

Finland Northern 4.154.280 More often High

Poland Eastern 30.455.706 Less often Very low

Romania Eastern 17.268.511 More often Low

The country sample included the five EU countries with the highest population share plus
three further countries. All eight countries cover approximately 300 million consumers
aged 18 and older, representing 76% of the EU universe.

The selection takes into account not only the “families of nations” typology which has
guided the bulk of sophisticated comparative public policy studies for the last few
decades, but also reflects our pre-analyses of market specific country profiles based on
the data from recent surveys TNS has conducted for the European Commission1.

In addition, the country selection also took into account that certain countries with
specific legislations would not be suitable for testing all requested dimensions in the
experiment. For instance, Austria, Sweden and Denmark have either banned trans fats or
restricted the amount to below 2% and therefore were excluded from our recommended
selection.

Finland was included as the only remaining country from the Northern family and
because it showed in previous research a very high awareness of quality labels and also
above average literacy levels when reading food labels.

Romania was included as it represents together with Italy the two countries with the
highest discrepancy between checking food labels (more often) and label literacy (low
comprehension), which may indicate that these countries should have a higher need for
improving food information.

The decision for an online methodology will have a significant effect on the universe and
the interpretation of results. The online universe is more homogenous in terms of higher
education and higher income level than the general population. On the one hand, there
will be a much higher incidence of shopping pre-packed food in modern grocery shops
also in Eastern European markets. On the other hand, the survey does not cover the
lowest social classes which are likely to need more assistance for making the right
choices when buying food.

1
Based on Eurobarometer data (July 2012 and March 2010)
As the focus of this research was about consumer behaviour and decisions made at the
point of sale, then it was agreed that the target should be the person (solely or jointly)
responsible for everyday shopping decisions. This target has a different profile to the
general population as outlined in the following table.

Universe stratification for EU272

EU27 EU27
population shoppers

Gender
1 Male 48% 40%

2 Female 52% 60%

Total 100% 100%

Age
1 18-24 12% 12%

2 25-34 17% 17%

3 35-44 19% 18%

4 45-54 18% 17%

4 55+ 34% 36%

Total 100% 100%

2
Source: based on 2,000 F2F interviews per country within Eurobarometer 73.2+3 in 2010 – shoppers
are defined as persons living in single households or saying that it is more them who shop for their
household or saying that they have a balanced responsibility together with their partner to decide on
everyday shopping for their household.
 Universe stratification of shoppers in the surveyed countries

When targeting the average EU shopper instead of consumer, the sample contains a
somewhat higher bias towards women and slightly older citizens. However, this
stratification allows a more robust analysis when looking at potential gender and age
effects and their implications for food product consumption in multi-person households in
all selected countries.

The process was the following:

 Target the average shopper of a household

 Ensure robust sample size for gender and age comparisons within each country

 Use the shopper profiles per country for weighting the overall sample

The online sample excluded anyone below 18 years - since they are not allowed to
purchase alcohol in most countries; buying beer or wine in retail stores is typically
allowed by the age of 16 or 18 (only in Sweden the minimum age is 20).
Offline store survey – technical summary

The objectives for the offline store experiments were:

 Do consumers access the relevant food information in a shopping environment? If

yes, who does, which parts, how long? Do they understand it?

 Understanding how consumers react, when confronted with TFA/PHO in various

treatment scenarios (with or without TFA on nutrition panel? and does the
education treatment improve the effect of concrete TFA amounts) and product
categories (assumption that products with a healthy image like yoghurt will
surprise more than less healthy products like biscuits)

 Are shoppers able to identify a healthier product? What are the barriers of making
good choices?

Fieldwork

Fieldwork took place in Germany in May and June 2014 with the achieved sample sizes
shown in the table below.

Sample size of offline survey in supermarkets

# Category Treatment Observations Interviews

Store 1: Isernhagen-
Control
Altwarmbüchen - Opelstraße Yoghurt 383 200
group
3-5

Store 2: Garbsen - Havelser

Yoghurt TFA info only 501 202
Str. 2-8

Store 3: Hemmingen - TFA info &

Yoghurt 574 201
Alfred Bentz Str. 1 education

Store 4: Hamburg-
Control
Oststeinbek - Willinghusener Frozen Pizza 598 211
group
Weg 1

Store 5: Hamburg-Lurup -
Frozen Pizza TFA info only 372 201
Grandkuhlenweg 11

Store 6: Hamburg-Farmsen TFA info &

Frozen Pizza 615 203
- Berner Heerweg 173-175 education

Store 7: Bremen-Vahr - Control

Biscuits 473 202
Vahrer Str. 197 group

Store 8: Ritterhude-Ihlpohl
Biscuits TFA info only 663 239
- Rosenhügel 5

Store 9: Bremen-
TFA info &
Habenhausen - Steinsetzer Biscuits 662 192
education
Str. 15
Design

The data collection process starts with an observation of shoppers at the shelf, who were
then screened and recruited when they leave the product category.

The interview with the experiments was conducted within the supermarket at a separate
stand with mock-up products (to control for brands, price and product variations), which
allowed a more robust analysis of the drivers of observed choices.

For further details of the data collection flow, please see the CAPI Questionnaire.

The design allowed the analysis of initial choices to simulate typical choice behaviour as
in reality (no education, no focus on health issues in preliminary interview). Although the
variations between the products more or less forced the attentive participant to reflect
health issues, the framing of the experiment tried to minimise this effect by reminding
participants of individual shopping goals and habits.

Following the observation of the initial choice behaviour and before presenting the
education treatment – participants were asked to give a cognitive health assessment of
the key elements in the nutrition panel. This order allowed the collection of an unbiased
awareness of the image of fat types, etc.

The education treatment was offered to every third participant (randomly allocated within
each product category).

The subsequent choice set allowed for the analysis choice behaviour aimed at a common
health goal (where appropriate).
Characteristics and choice architecture for the two tasks:

Task 1 = simplified choice with only one variation (with and without TFA/PHO), which
was comparable across all product categories

 Allowed analysis to understand whether consumers ignore or misunderstand or don’t

understand the meaning of TFA within a simplified context

Task 2 = more realistic choice with two or three variations (Saturates, sugar and/or salt)
and not always a superior choice

 Allowed analysis to understand how consumers react when the choices are more
difficult as in reality and whether they tend to overreact by choosing products without
TFA that are not healthier than products with TFA

Detailed characteristics and their rationale are outlined on the following pages for each
task and category.
Suggestion for nutrition facts in choice tasks for YOGHURT
Task1: set of yoghurts with a superior choice: Product A without TFA and B with TFA
Nutrition facts are selected more on the healthy side of an ordinary full fat yogurth with 3.5% fat. Higher fat levels are lik ely to reduce any potential surprise effect
(10% fat is lik ely to be considered less healthy and is less often part of a regular diet). However, if it is unrealistic to have 0,4% TFA within a 3,5% total fat yoghurt, then
we should use a slightly higher total fat level.
TFA level for product B is slightly higher than the identified amounts in the NEVO database. However, we assume it is still a realistic level, which would allow us to
compare the impact of all product categories (yoghurt, pizza, biscuits) at least for task 1 (only possible if same level used for all products in one task ).
Energy Energy Protein Carbohydra Sugar Fat total Saturates Trans fat Salt
Yoghurt set for task 1
(kJ) (kcal) total (g) tes total (g) (g) (g) (g) (g) (g)
Product A: (healthier) 278 67 4,0 3,3 3,3 3,5 2,5 0,0 0,1

Product B: (less healthy) 278 67 4,0 3,3 3,3 3,5 2,5 0,4 0,1

Task2: yoghurt without superior choice (both with equally less healthy nutrition facts): X with TFA, Y with more Sugar
Nutrition facts are selected more on the less healthy side of a yogurth with 7.6% fat (still avoiding the extreme). Product X is similar to the Greek yoghurt in the NEVO
database and Product Y is similar to the yoghurt cream with fruit in the database. All elements (other than TFA, sugar and carbohydrates) were consolidated to average
amounts between both yoghurt types.
Energy Energy Protein Carbohydra Sugar Fat total Saturates Trans fat Salt
Yoghurt set for task 2
(kJ) (kcal) total (g) tes total (g) (g) (g) (g) (g) (g)
Product X: (less healthy) 555 133 3,7 3,6 3,6 7,6 4,9 0,2 0,1

Product Y: (less healthy) 555 133 3,7 14,9 14,9 7,6 4,9 0,0 0,1

NEVO DATABASE - http://nevo-online.rivm.nl/ProductenZoeken.aspx

Energy Energy Protein Carbohydra Sugar Fat total Saturates Trans fat Natrium
Yoghurts - full fat only
(kJ) (kcal) total (g) tes total (g) (g) (g) (g) (g) (g)
Yoghurt Greek full fat
519 125 4,7 3,6 3,6 10,0 6,5 0,2 0,1
(base for product x)
Yoghurt cream with fruit
590 141 2,7 15,1 14,9 7,6 4,9 0,1 0,0
(base for product y)
Yoghurt Turkish 10% fat 485 117 3,5 3,0 3,1 10,1 6,0 0,2 0,0

Yoghurt full fat stracciatella 557 133 3,7 17,3 14,6 5,2 3,3 0,1 0,0

Yoghurt full fat w fruit Activia 398 95 3,5 13,1 12,7 3,1 1,9 0,1 0,1

Yoghurt Bulgarian whole milk 366 87 5,1 6,0 6,0 4,5 2,9 0,1 0,0

Yoghurt Turkish 4% fat 278 67 3,9 3,3 3,3 4,2 2,5 0,1 0,1

Yoghurt full fat 237 57 3,8 3,3 3,3 2,9 1,9 0,1 0,1

Yoghurt full fat with fruit 408 97 4,0 14,0 13,5 2,5 1,6 0,0 0,0

MIN 237 57 2,7 3,0 3,1 2,5 1,6 0,0 0,0

AVERAGE 426 102 3,9 8,7 8,3 5,6 3,5 0,1 0,0

MAX 590 141 5,1 17,3 14,9 10,1 6,5 0,2 0,1

NHS guidelines
Sugar Fat total Saturates Trans fat
General per 100 g Salt
(g) (g) (g) (g)
LOW (if … or lower) 5,0 3,0 1,5 0,2 0,3

HIGH (if more than …) 22,5 17,5 5,0 0,5 1,5

Suggestion for nutrition facts in choice tasks for PIZZA
Task1: set of pizzas with a superior choice: Product A without TFA and B with TFA
Nutrition facts reflect the pizza average in the NEVO database.

TFA level for product B is slightly higher than the identified amounts in the NEVO database. However, we assume it is still a realistic level, which would allow us to
compare the impact of all product categories (yoghurt, pizza, biscuits) at least for task 1 (only possible if same level used for all products in one task ).
Energy Energy Protein Carbohydra Sugar Fat total Saturates Trans fat Salt
Pizza set for task 1
(kJ) (kcal) total (g) tes total (g) (g) (g) (g) (g) (g)
Product A: (healthier) 986 236 9,0 27,5 2,7 9,4 3,0 0,0 0,6

Product B: (less healthy) 986 236 9,0 27,5 2,7 9,4 3,0 0,4 0,6

Task2: pizza set with superior choice: healthier pizza X with some TFA and less salt, Y without TFA and more salt
Nutrition facts are similar to the database average as in task 1. Product X is similar to the cheese and vegetable pizza in the NEVO database and Product Y is
similar to the Turk ish pizza in the database. All elements (other than TFA, natrium) were consolidated to average amounts between both pizza types. The salt level
was adjusted based on Jan´s feedback . Please advise, whether the variations of TFA and salt establish an objectively superior choice for product X or what would be
the maximum variation which is still realistic?
Energy Energy Protein Carbohydra Sugar Fat total Saturates Trans fat Salz
Pizza set for task 2
(kJ) (kcal) total (g) tes total (g) (g) (g) (g) (g) (g)
Product X: (healthier) 929 221 8,5 27,4 2,6 8,2 2,4 0,2 0,2

Product Y: (less healthy) 929 221 8,5 27,4 2,6 8,2 2,4 0,0 1,8

NEVO DATABASE - http://nevo-online.rivm.nl/ProductenZoeken.aspx

Energy Energy Protein Carbohydra Sugar Fat total Saturates Trans fat Natrium
Pizza with …
(kJ) (kcal) total (g) tes total (g) (g) (g) (g) (g) (g)
cheese and vegetables
1041 249 9,0 24,5 1,9 12,3 4,4 0,2 0,5
(base for product x)
Turkish with extras
862 204 7,6 33,7 3,4 3,7 1,2 0,1 0,7
(base for product y)
salami 1087 259 9,9 27,4 2,1 11,8 3,9 0,1 0,7

meat 980 234 9,1 26,9 2,3 9,5 3,3 0,1 0,6

fish 1014 242 9,6 25,3 2,1 10,9 3,0 0,1 0,5

ham and pineapple 929 221 8,5 27,4 4,4 8,2 2,4 0,1 0,5

MIN 862 204 7,6 24,5 1,9 3,7 1,2 0,1 0,5
AVERAGE 986 235 9,0 27,5 2,7 9,4 3,0 0,1 0,6
MAX 1.087 259 9,9 33,7 4,4 12,3 4,4 0,2 0,7

NHS guidelines
Sugar Fat total Saturates Trans fat
General per 100 g Salt
(g) (g) (g) (g)
LOW (if … or lower) 5,0 3,0 1,5 0,2 0,3
HIGH (if more than …) 22,5 17,5 5,0 0,5 1,5
Suggestion for nutrition facts in choice tasks for BISCUITS
Task1: set of biscuits with a superior choice: Product A without TFA and B with TFA
Nutrition facts reflect the biscuits average in the NEVO database.

TFA level for product B represents a slightly higer than average amounts in the NEVO database.
Energy Energy Protein Carbohydra Sugar Fat total Saturates Trans fat Salt
Biscuits set for task 1
(kJ) (kcal) total (g) tes total (g) (g) (g) (g) (g) (g)
Product A: (healthier) 2.050 490 6,4 65,7 27,7 21,6 10,7 0,0 0,3

Product B: (less healthy) 2.050 490 6,4 65,7 27,7 21,6 10,7 0,4 0,3

Task2: biscuits without superior choice: X with more TFA and SF but less sugar, Y with more Sugar but less SF and no TFA
Nutrition facts are selected more on the less healthy side with above average calories and fat levels. Product X is similar to the Dutch shortbread spritsstuk k en in
the NEVO database and Product Y is similar to the biscuit chocolate coated chocoprins in the database. All elements (other than TFA, sugar and saturates) were
consolidated to average amounts between both biscuit types. Please advise, whether the variations of TFA, Saturates and Sugar establish an objectively superior
choice or not?
Energy Energy Protein Carbohydra Sugar Fat total Saturates Trans fat Salt
Biscuits set for task 2
(kJ) (kcal) total (g) tes total (g) (g) (g) (g) (g) (g)
Product X: (less healthy) 2.139 511 5,1 65,7 24,4 24,9 16,2 0,6 0,2

Product Y: (less healthy) 2.139 511 5,1 65,7 40,9 24,9 7,5 0,0 0,2

NEVO DATABASE - http://nevo-online.rivm.nl/ProductenZoeken.aspx

Energy Energy Protein Carbohydra Sugar Fat total Saturates Trans fat Natrium
Selected biscuits
(kJ) (kcal) total (g) tes total (g) (g) (g) (g) (g) (g)
Biscuit Dutch shortbread
2201 526 5,0 62,4 24,4 28,2 16,2 0,6 0,2
spritsstukken (base for product x)
Biscuit chocolate coated Chocoprins
2077 495 5,2 69,0 40,9 21,6 11,9 0,1 0,2
(base for product y)
Biscuit chocolate 1978 472 7,1 63,5 30,2 20,0 10,7 1,3 0,2

Biscuit Dutch shortbread w chocolate 2254 539 4,9 61,4 29,4 30,1 13,0 0,6 0,2

Waffle penny 2261 542 5,9 56,1 32,4 32,5 22,2 0,2 0,2

Biscuit filled Prince 1971 469 5,0 73,0 36,0 17,0 5,5 0,2 0,2

Rice cakes puffed w chocolate 2099 502 6,8 60,3 25,3 24,9 14,7 0,1 0,1

Biscuit digestive w chocolate 2049 489 6,5 64,3 28,1 22,2 7,4 0,1 0,4

Biscuit brown/ wholemeal 1919 457 7,6 70,0 21,9 15,0 6,9 0,1 0,4

Biscuit sweet 1869 443 7,6 77,2 18,6 11,0 5,1 0,1 0,4

MIN 1.869 443 4,9 56,1 18,6 11,0 5,1 0,1 0,1
AVERAGE 2.050 489 6,4 65,7 27,7 21,6 10,7 0,3 0,3
MAX 2.261 542 7,6 77,2 36,0 32,5 22,2 1,3 0,4

NHS guidelines
Sugar Fat total Saturates Trans fat
General per 100 g Salt
(g) (g) (g) (g)
LOW (if … or lower) 5,0 3,0 1,5 0,2 0,3
HIGH (if more than …) 22,5 17,5 5,0 0,5 1,5
EDUCATION TREATMENT FOR OFFLINE EXPERIMENT

How do you know if a food fits well in a balanced and healthy diet?

For a more balanced and healthy diet you should cut down on:
- Fat – especially saturated and trans fat,
- Salt and
- Added sugars.

How do I know that a food is high or low in fat, sugar or salt?

For example, watch out for healthy and unhealthy fat types:
There are different kinds of fat included in most types of food. Many people know that
- saturated fat is unhealthy while
- unsaturated fat is generally healthy.
It is also recommended that people avoid trans fat, which is even more unhealthy than
saturated fat.
- On the one hand, trans fats can occur naturally in certain foods from
ruminant animals such as in milk and dairy products (e.g. yoghurt, cheese)
 where levels are low and not of concern.
- On the other hand, trans fats can be produced industrially and added to
processed foods, in particular in the form of partially hydrogenated
(vegetable) oils and fats
 where levels can be worryingly high.
Today most food producers have reformulated their products and removed partially
hydrogenated oils from their products or greatly reduced them. Nevertheless, few
products with high trans fat levels are still present on the market and it is important
to identify them and use products from alternative producers.
 You can either identify trans fat by spotting partially hydrogenated oils and fats
in the ingredients list or by recognising products with more than 0.8g trans fats per
100g.
Important: Do not lose sight of the other nutrients when avoiding partially
hydrogenated oils or trans fats: prefer products and brands where all sugar, salt,
saturated and trans fats are low or at least not high
Impact of Food Information on Consumers’ Decision Making

APPENDIX B – QUESTIONNAIRES

212
CAWI
Questionnaire

Name of survey
EC Food Information - MCP Survey 2013

Client name
DG SANCO

Author(s)
Elke Himmelsbach
Anna Rysina

This questionnaire was written according to TNS quality procedures

checked by Quality Department and Translations

Version from August 05th 2013

67.02.131647

© TNS 16-07-2013 Version number: 2

Index
SCREENER & INTRO
Q1 | country | Country
T1 | intro0 | Intro to screener
Q2 | age1 | Age open
Q3 | age2 | Age group
Q4 | age3 | Age combined
Q5 | sex | Gender
Q6 | hhsize | Household size
Q7 | resp | Shopping responsibility
T2 | outro1 | Screen out
T3 | intro1 | Intro to study
End SCREENER & INTRO
SHOPPING HABITS - GENERAL
Q8 | freq-shop | Frequency of shopping
Q9 | engage | Shopping engagement
Q10 | plan | Pre-shopping habits
Q11 | goals | Shopping goals
End SHOPPING HABITS - GENERAL
AWARENESS OF TF/PHO/FHO
Q12 | tf-awa | Awareness of fat types
Q13 | tf-eva | Evaluation of fat types
End AWARENESS OF TF/PHO/FHO
AWARENESS OF ALLERGY INFORMATION
Q14 | diet1 | Dietary restrictions - personal
Q15 | diet2 | Dietary restrictions - household
Q16 | allergy1 | Allergic reaction
Q17 | allergy2 | Reading allergy labels
Q18 | seg-allergen | Segmentation on allergen awareness
End AWARENESS OF ALLERGY INFORMATION
CATEGORY SPECIFIC HABITS
Q19 | cat-food | Category buyers: Food
Category Repeater Questions
Q20 | type1 | Selectors / Habitual Buyers
Q21 | type2 | Shopper type
Q22 | labels | Relevant food information per category
End Category Repeater Questions
Q23 | cat-alc | Category non-drinker: Consumption of alcoholic beverages
Q24 | seg-catuser | Segmentation on category usage
Q25 | seg-exp | Respondent split for food or drinks experiment
End CATEGORY SPECIFIC HABITS
FOOD EXPERIMENTS
T4 | exp-intro-food1 | Experiments Intro Food 1
CHOICE TASK: TRANS FATS
Q26 | split-tf | Treatment splits: Trans fats
T5 | exp-intro-tf | Experiments Intro Trans Fats
T6 | edu-tf1 | Education on Trans Fats
Q271 | exp-tf1-with | Trans fats choice 1 - with TFA
Q272 | exp-tf1-without | Trans fats choice 1 - without TFA
Q273 | dum_zoom_tf1 | Zooming areas TF1
Q281 | exp-tf2- with | Trans fats choice 2 - with TFA
Q282 | exp-tf2- without | Trans fats choice 2 - without TFA
Q283 | dum-zoom-tf2 | Zooming areas TF2
Q29 | tf-health-cho | Always healthier choice for Trans fat
Q30 | rationale-tf2 | Rationale for second choices on trans fats
Q31 | rationale-tf1 | Rationale for first choice on trans fats
Q32 | tf-truth | Trans fats myths and truths
 End CHOICE TASK: TRANS FATS
CHOICE TASK: ALLERGENS & PRIMARY INGREDIENT
Q33 | split-allergy | Treatment splits: Allergy
T7 | intro-allergy | Experiment intro allergen labels
ROTATE ORDER OF QUESTIONS IN THIS BLOCK
Q341 | exp-all1-split1 | Allergen choice 1 - May contain
Q342 | exp-all1-split2 | Allergen choice 1 - May contain traces
Q343 | exp-all1-split3 | Allergen choice 1 - Cannot guarantee
Q344 | exp-all1-split4 | Allergen choice 1 - Manufactured in a facility that also processes
Q345 | dum-zoom-allergy1 | Zooming areas Allergy task 1
Q351 | exp-all2-split1 | Allergen choice 2 - May contain
Q352 | exp-all2-split2 | Allergen choice 2 - May contain traces
Q353 | exp-all2-split3 | Allergen choice 2 - Cannot guarantee
Q354 | exp-all2-split4 | Allergen choice 2 - Manufactured in a facility that also processes
Q355 | dum-zoom-allergy2 | Zooming areas Allergy task 2
Q361 | exp-all3-split1 | Allergen choice 3 - May contain
Q362 | exp-all3-split2 | Allergen choice 3 - May contain traces
Q363 | exp-all3-split3 | Allergen choice 3 - Cannot guarantee
Q364 | exp-all3-split4 | Allergen choice 3 - Manufactured in a facility that also processes
Q365 | dum-zoom-allergy3 | Zooming areas Allergy task 3
Q37 | all-sec-cho | Always secure choice for Allergens
Q38 | num-sec-cho | Number of secure choice for Allergens
End ROTATE ORDER OF QUESTIONS IN THIS BLOCK
Q39 | rationale-all-pi | Rationale for choices on allergen and primary ingredient
Q40 | risk-allergy | Risk assessment of precautionary allergen label options
Q41 | recall-pi1 | Recall of primary ingredient in Muesli
Q42 | pi-pizza | Primary ingredient in pizza
Q43 | pi-yogurt | Primary ingredient in yogurt
End CHOICE TASK: ALLERGENS & PRIMARY INGREDIENT
Q44 | food-info| Expected food information
End FOOD EXPERIMENTS
DRINKS EXPERIMENTS
Q45 | alc-pref | Choice preferences for alcohol
Q46 | seg-drinks | Respondent segmentation - treatment splits on alcohol
PREVIOUS ALCOHOL CONSUMPTION
Q471 | alc-occa-beer | Drinking occasions - Beer
Q472 | alc-occa-wine | Drinking occasions - Wine
Q473 | alc-occa-spirits | Drinking occasions - Spirits
Q481 | alc-freq-beer | Frequency of drinking alcohol - Beer
Q482 | alc-freq-wine | Frequency of drinking alcohol - Wine
Q483 | alc-freq-spirits | Frequency of drinking alcohol - Spirits
Q491 | alc-occa-last-beer | Drinking occasions - last time - Beer
Q492 | alc-occa-last-wine | Drinking occasions - last time - Wine
Q493 | alc-occa-last-spirits | Drinking occasions - last time - Spirits
Q50 | beer-vol0 | Alcohol consumption per occasion - beer
Q51 | wine-vol0 | Alcohol consumption per occasion - wine
Q52 | spirits-vol0 | Alcohol consumption at last occasion - vodka/whiskey
End PREVIOUS ALCOHOL CONSUMPTION
T8 | exp-intro-drinks | Experiments Intro Drinks
Q53 | beer-vol1 | Beer consumption ex ante
Q54 | wine-vol1 | Wine consumption ex ante
Q55 | spirits-vol1 | Spirits consumption ex ante
TASK: BEER - CALORIES
Q56 | beer-cal-test | Knowledge test on calories for beer
T9 | beer-cal-edu | Education - Calories in beer vs pizza
Q57 | beer-vol2-cal | Consumption change - Calories - Beer
End TASK: BEER - CALORIES
TASK: BEER - LIMITS
 Q58 | beer-lim-test | Knowledge test on limits for beer
T10 | beer-lim-edu | Education - Limits with beer
Q59 | beer-vol2-lim | Consumption change - Limits - Beer
End TASK: BEER - LIMITS
TASK: WINE - CALORIES
Q60 | wine-cal-test | Knowledge test on calories for wine
T11 | wine-cal-edu | Education - Calories in wine vs. chocolate
Q61 | wine-vol2-cal | Consumption change - Calories - Wine
End TASK: WINE - CALORIES
TASK: WINE - LIMITS
Q62 | wine-lim-test | Knowledge test on limits for wine
T12 | wine-lim-edu | Education - Limits with wine
Q63 | wine-vol2-lim | Consumption change - Limits - Wine
End TASK: WINE - LIMITS
TASK: SPIRITS - CALORIES
Q64 | spirits-cal-test | Knowledge test on calories for spirits
T13 | spirits-cal-edu | Education - Calories in spirits vs. nuts
Q65 | spirits-vol2-cal | Consumption change - Calories - Spirits
End TASK: SPIRITS - CALORIES
TASK: SPIRITS - LIMITS
Q66 | spirits-lim-test | Knowledge test on limits for spirits
T14 | spirits-lim-edu | Education - Limits with spirits
Q67 | spirits-vol2-lim | Consumption change - Limits - Spirits
End TASK: SPIRITS - LIMITS
Q68 | beer-vol3 | Future intention to drinking volume of beer
Q69 | wine-vol3 | Future intention to drinking volume of wine
Q70 | spirits-vol3 | Future intention to drinking volume of spirits
Q71 | alc-info | Expected food information on alcoholic drinks
Q72 | alc-att | Attitudes about alcohol consumption
Q73 | too-much | Experience with too much alcohol
End DRINKS EXPERIMENTS
FOOD WASTE
Q74 | bestbefore | Knowledge test: Best Before
Q75 | useby | Knowledge test: Use By
Q76 | waste1 | Consumed outdated food
Q77 | waste2 | Food thrown away before expiry
Q78 | waste-att1 | Food waste attitudes 1
Q79 | waste-habit | Order in fridge
Q80 | waste-vol0 | Food waste volume
T15 | waste-fb1 | Food Waste knowledge test - Feedback1
T16 | waste-edu1 | Education on Food Waste 1
T17 | waste-fb2 | Food Waste knowledge test - Feedback2
T18 | waste-edu2 | Education on Food Waste 2
Q81 | waste-att2 | Food waste attitude change
End FOOD WASTE
DEMOGRAPHICS
Q82 | att-control-risk | Self control & Risk attitude
Q83 | living-status | Living Status
Q84 | work-status | Working status
Q85 | education | Education - terminal age
Q86 | income | Individual Income
Q87 | height-uk | Height UK in feet and inches
Q88 | height-dum-uk | Height UK in cm
Q89 | height-non-UK | Height non-UK in cm
Q90 | height-all-dum | Height all in cm Dummy
Q91 | weight-uk | Weight UK in stones and pounds
Q92 | weight-dum-uk | Weight UK in kg
Q93 | weight-non-UK | Weight non-UK in kg
 Q94 | weight-all-dum | Weight all in kg Dummy
End DEMOGRAPHICS
Q95 | qual1 | Quality-Check: Straightliners
Q96 | qual2 | Quality-Check: Speeder
T19 | outro0 | Outro
B1 | SCREENER & INTRO Begin block

Q1 | country | Country Single coded

Dummy

1 UK
2 France
3 Germany
4 Italy
5 Spain
6 Finland
7 Poland
8 Romania

T1 | intro0 | Intro to screener Text

Welcome
... to our survey!

Thank you so much for participating in this survey!

We treat your personal data strictly confidentially. It will under no circumstances be given to any third
parties including our clients. Studies are always analysed anonymously (e.g. "75 per cent of the people
say ...")

It can take up to 14 days after the survey has closed for the remuneration to appear in your account.

Q2 | age1 | Age open Numeric

Min 18 | Max 99
Before we begin, please answer the following questions:
What was your age on your last birthday?

Scripter notes: - IF AGE younger than 18, THEN => SCREEN OUT
- IF respondent skips this question, THEN write ‘Please review your responses on this page. One or more
questions require further input.’ at top of the screen AND after the question text ‘Please provide an
answer’ AND add ‘No answer' option with code 9.
- IF 'No answer' is given, THEN ask Q3 with age categories.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 1

Q3 | age2 | Age group Single coded

Which of the following age groups do you fall into?

1 Less than 18 years old

 GO TO T2outro1
2 Between 18 and 24 years old
3 Between 25 and 34 years old
4 Between 35 and 44 years old
5 Between 45 and 54 years old
6 55 years old or more
9 No answer
 GO TO T2outro1

Scripter notes: - ADD FILTER: This question is only asked if Q2 is not answered.
- IF respondent skips this question, THEN write ‘Please review your responses on this page. One or more
questions require further input.’ at top of the screen AND after the question text ‘Please provide an
answer’ AND add ‘No answer’ option with code 9.

Q4 | age3 | Age combined Single coded

Dummy
[DP: sum of answers in Q1 and Q2]

1 Between 18 and 24 years old

2 Between 25 and 34 years old
3 Between 35 and 44 years old
4 Between 45 and 54 years old
5 55 years old or more

Researcher notes: Translator: no translation required

Scripter notes: CHECK QUOTAS on age stratification per country

Q5 | sex | Gender Single coded

Are you...?

1 Male
2 Female
9 No answer
 GO TO T2outro1

Scripter notes: - IF respondent skips this question, THEN write ‘Please review your responses on this
page. One or more questions require further input.’ at top of the screen AND after the question text
‘Please provide an answer’ AND add ‘No answer’ option with code 9.
- CHECK QUOTAS on gender stratification per country

EC Food Information - MCP Survey 2013 | version 2 | © TNS 2

Q6 | hhsize | Household size Single coded

Including yourself, how many people are there in your household?

1 1
2 2
3 3
4 4
5 5 or more
9 No answer
 GO TO T2outro1

Scripter notes: IF respondent skips this question, THEN write 'Please review your responses on this page.
One or more questions require further input.' at the top of the screen AND after the question text 'Please
provide an answer' AND add 'No answer' option with code 9.

ASK ONLY IF Q6 | hhsize=2,3,4,5

Q7 | resp | Shopping responsibility Single coded

Thinking of you and other members of your household, who is more likely to make the decisions on
everyday shopping?

1 More you
2 Balanced between you and someone else
3 More your partner or someone else in your household
 GO TO T2outro1
9 No answer
 GO TO T2outro1

Scripter notes: IF respondent skips this question, THEN write 'Please review your responses on this page.
One or more questions require further input.' at the top of the screen AND after the question text 'Please
provide an answer' AND add 'No answer' option with code 9.

T2 | outro1 | Screen out Text

We are very sorry, but we have already received enough answers from people corresponding to your
profile for this study.
Thank you for taking the time to answer!

Scripter notes: Display this text for screen-out if screening requirements are not fulfilled or quotas are
full.

T3 | intro1 | Intro to study Text

Welcome and thank you for volunteering to take part in our survey. The survey is about your habits and
preferences when you go shopping for food. It starts with some questions about your shopping
behaviour, when purchasing groceries.

This survey also includes a brief choice experiment where it is possible for you to earn an additional
bonus on top of your usual incentive for completing the questionnaire.

Your answers are, of course, completely confidential and will only be analysed and shown in an
aggregated format.

Please try to complete the survey in one sitting. It should take no longer than 20 minutes.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 3

B1 | SCREENER & INTRO End block

B2 | SHOPPING HABITS - GENERAL Begin block

Q8 | freq-shop | Frequency of shopping Single coded

How often do you shop for groceries?

5 More than twice a week

4 Twice a week
3 Once a week
2 Every other week
1 Once a month or less often
8 Don't know

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

Q9 | engage | Shopping engagement Matrix

Which of the following best describes how you feel about shopping groceries?

Rotated
Strongly Strongly
agree disagree
I enjoy doing the     
shopping
I usually shop in a rush     
I usually take time and     
browse while shopping
to see what is new
When shopping, I am     
often overwhelmed by
the choice available
I can only afford to buy     
what I need rather than
what I want

Scripter notes: IF respondent forgets some answers in a matrix question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer in line [insert and highlight statement]’ .

Please program a slider scale, but please avoid the necessity of a flash player for this question

Researcher notes: Translator note: Please translate also statements 2 and 3, since we don’t
know if they can be skipped.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 4

Q10 | plan | Pre-shopping habits Matrix

When you are shopping or planning to shop for groceries.

Before you enter a store - how often do you ...

Rotated
Always Most of the Some of the Rarely Never Don't know
time time
Make a list of products      
you want to buy
Search for special      
deals, best prices or
coupons
Already know which      
product and brand you
will buy because you
buy the same foods
every time

Scripter notes: IF respondent forgets some answers in a matrix question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer in line [insert and highlight statement]’ AND add ‘Don’t know’
option with code 8.

Q11 | goals | Shopping goals Matrix

Thinking about how you shop for food, how well does each of the statements describe you and your
shopping behaviours and beliefs?

Rotated
Strongly agree Agree Disagree Strongly Don't know
disagree
I'm never willing to     
compromise quality for
a lower price
Price is more important     
than anything else
I read labels and pay     
close attention to
product information
I am extremely loyal to     
the brands I buy
I always try to choose     
the healthier
alternative when
selecting a product

Scripter notes: IF respondent forgets some answers in a matrix question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer in line [insert and highlight statement]’ AND add ‘Don’t know’
option with code 8.

B2 | SHOPPING HABITS - GENERAL End block

B3 | AWARENESS OF TF/PHO/FHO Begin block

EC Food Information - MCP Survey 2013 | version 2 | © TNS 5

Q12 | tf-awa | Awareness of fat types Matrix

Have you ever heard of these types of fat before?

Which of the following applies to you regarding ... ?

Rotated
I have never heard I have heard of it, I know roughly I have enough
of it but I do not know what it is / what it knowledge to take
much about it means the right decisions
for myself
Trans fat    
Partially hydrogenated    
oil
Fully hydrogenated oil    
Saturated fat    

Scripter notes: IF respondent forgets some answers in a matrix question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer in line [insert and highlight statement]'.

ASK ONLY IF Q12 | tf1 ST=1 & SC=2,3,4 or Q12 | tf1 ST=2 & SC=2,3,4 or Q12 | tf1 ST=3 & SC=2,3,4
or Q12 | tf1 ST=4 & SC=2,3,4

Q13 | tf-eva | Evaluation of fat types Matrix

Based on what you know, how healthy do you think these types of fat are?

Rotated
Very healthy Quite healthy Not sure Quite Very unhealthy
unhealthy

Trans fat     
Partially hydrogenated     
oil
Fully hydrogenated oil     
Saturated fat     

Scripter notes: ASK only for statements (fat types) if in question before codes 2-4!
IF respondent forgets some answers in a matrix question, THEN write ‘Please review your responses on
this page. One or more questions require further input.’ at top of the screen AND after the question text
‘Please provide an answer in line [insert and highlight statement]'.

B3 | AWARENESS OF TF/PHO/FHO End block

B4 | AWARENESS OF ALLERGY INFORMATION Begin block

EC Food Information - MCP Survey 2013 | version 2 | © TNS 6

Q14 | diet1 | Dietary restrictions - personal Multi coded

Do you have any dietary restrictions?

Tick all that apply

Rotated

1 Food allergy or intolerance

2 Diet / Weight loss
3 Diabetic
4 Vegetarian
5 Vegan
6 Other *Position fixed
7 No food restrictions *Exclusive *Position fixed
8 Don't know *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

Q15 | diet2 | Dietary restrictions - household Multi coded

Does anyone else in your household (for whom you also shop at least occasionally) have any dietary
restrictions?

Tick all that apply

Rotated

1 Food allergy or intolerance

2 Diet / Weight loss
3 Diabetic
4 Vegetarian
5 Vegan
6 Other *Position fixed
7 No food restrictions *Exclusive *Position fixed
8 Don't know *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

ASK ONLY IF Q14 | diet1=1 or Q15 | diet2=1

Q16 | allergy1 | Allergic reaction Single coded

How would you rank the severity of the allergy/intolerance?

If there is more than one food allergy or intolerance, please answer for the more severe type.

1 Mild
2 Moderate
3 Severe
8 Don't know

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 7

 ASK ONLY IF Q14 | diet1=1 or Q15 | diet2=1

Q17 | allergy2 | Reading allergy labels Single coded

Do you ever look for allergy information, when shopping for food?

4 Yes, regularly
3 Yes, but only occasionally, e.g. when shopping for/with someone else
2 Yes, I have done previously
1 No, never
8 Don't know

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

Q18 | seg-allergen | Segmentation on allergen awareness Single coded

Dummy
[DP: target group consolidation based on previous questions]

1 Personally affected with label experience [DP: if in Q14 code 1 AND Q17 code 2-4]
2 Personally affected without label experience [DP: if in Q14 code 1 AND Q17 code 1 or 8]
3 Only household member affected with label experience [DP: if in Q15 code 1 AND in Q14 NOT
code 1 AND in Q17 code 2-4]
4 Only household member affected without label experience [DP: if in Q15 code 1 AND in Q14 NOT
code 1 AND in Q17 code 1 or 8]
5 Not affected [DP: if in Q14 NOT code 1 AND in Q15 NOT code 1]

Researcher notes: Translator: no translation required

B4 | AWARENESS OF ALLERGY INFORMATION End block

B5 | CATEGORY SPECIFIC HABITS Begin block

Q19 | cat-food | Category buyers: Food Matrix

When was the last time you bought any of the following products for yourself or your household, friends
or colleagues in a self-service shop like a supermarket, discount store, grocery store, local corner shop,
health food shop or in an online shop?

Rotated
Last week Last 4 Last 6 Last 12 Less often Never
weeks/ last months months bought
month
Crisps      
Muesli      

Scripter notes: Scripter notes: - IF respondent forgets some answers in a matrix question, THEN write
‘Please review your responses on this page. One or more questions require further input.’ at top of the
screen AND after the question text ‘Please provide an answer in line [insert and highlight statement]’ .

B6 | Category Repeater Questions Begin block

Scripter notes: Repeat block for each category bought within last 6 months, i.e. IF in Q19 codes 4-6

EC Food Information - MCP Survey 2013 | version 2 | © TNS 8

Q20 | type1 | Selectors / Habitual Buyers Single coded

Which statement best describes when you usually decide on the brand of [DP: category] to buy?

1 I have no particular brand in mind and make up my mind in the store
2 I have more than one specific brand in mind and make up my mind in the store
3 I usually start off wanting a certain brand but sometimes end up buying something different
4 I always know which brand I want before reaching the store

Scripter notes: - Please insert category name in question text and underline it
- IF respondent skips this question, THEN write ‘Please review your responses on this page. One or more
questions require further input.’ at top of the screen AND after the question text ‘Please provide an
answer’.

Q21 | type2 | Shopper type Single coded

Dummy

1 Selecting buyer [DP: if in Q20 codes 1-2]

2 Habitual buyer [DP: if in Q20 codes 3-4]
3 Non-Buyer [DP: if in Q19 code 1 ]

Researcher notes: Translator: no translation required

Q22 | labels | Relevant food information per category Multi coded

What kind of information do you normally watch out for when selecting a product or brand of [DP:
category] for the first time?

Tick all that apply

Rotated

1 Allergen information
2 Amount of carbohydrates
3 Amount of sugar
4 Amount of salt
5 Amount of saturated fat
6 Calories
7 Country of origin
8 Expiry date
9 Health claims
10 Ingredients
11 Organically produced
12 Price
13 Other *Position fixed
14 None of these *Exclusive *Position fixed
15 Don't know *Exclusive *Position fixed

Scripter notes: - Please insert category name in question text and underline it
- IF respondent skips this question, THEN write ‘Please review your responses on this page. One or more
questions require further input.’ at top of the screen AND after the question text ‘Please provide an
answer’ AND add ‘Don’t know’ option with code 15.

Client notes: The purpose of this question is to capture the previous behaviour when selecting a product
for the first time (to exclude habitual buying for repeat purchases). There will be another question after
the experiments to capture the expectation which information should be given on the product packages.

B6 | Category Repeater Questions End block

EC Food Information - MCP Survey 2013 | version 2 | © TNS 9

Q23 | cat-alc | Category non-drinker: Consumption of alcoholic Matrix
beverages

During the past 12 months, did you drink any beer, wine, vodka and/or whiskey?

Rotated
Yes No
Beer  
Wine  
Vodka  
Whiskey  

Scripter notes: - IF respondent forgets some answers in a matrix question, THEN write ‘Please review
your responses on this page. One or more questions require further input.’ at top of the screen AND after
the question text ‘Please provide an answer in line [insert and highlight statement]’.

Q24 | seg-catuser | Segmentation on category usage Multi coded

Min 6 | Max 6 | Dummy
[DP: target group consolidation based on previous questions]

1 Crisps shopper [DP: if in Q19 ST1 codes 3-6]

2 Crisps non-shopper [DP: if in Q19 ST1 code 1-2]
3 Muesli shopper [DP: if in Q19 ST2 codes 3-6]
4 Muesli non-shopper [DP: if in Q19 ST2 codes 1-2]
5 Beer drinker [DP: if in Q23 ST1 code 1]
6 Beer non-drinker [DP: if in Q23 ST1 code 2]
7 Wine drinker [DP: if in Q23 ST2 code 1]
8 Wine non-drinker [DP: if in Q23 ST2 code 2]
9 Vodka drinker [DP: if in Q23 ST3 code 1]
10 Vodka non-drinker [DP: if in Q23 ST3 code 2]
11 Whiskey drinker [DP: if in Q23 ST4 code 1]
12 Whiskey non-drinker [DP: if in Q23 ST4 code 2]

Researcher notes: Translator: no translation required

Q25 | seg-exp | Respondent split for food or drinks experiment Single coded
Dummy
[DP: assign respondents to the split where at least one category is used, if respondent uses products in
both categories, then choose the split with the lowest number of respondents to achieve an equal sample
participation between food and drinks experiments]

1 FOOD: crisps, high allergen awareness (DP: if in Q23 code 2 for all alcohol types OR in Q19 ST1
codes 4-6 OR high allergen awareness Q18 codes 1-4)
2 ALCOHOL: beer, wine, vodka, whiskey (DP: if in Q23 code 1 for at least one alcohol type)

Researcher notes: Translator: no translation required

B5 | CATEGORY SPECIFIC HABITS End block

EC Food Information - MCP Survey 2013 | version 2 | © TNS 10

 ASK ONLY IF Q25 | seg-exp=1

B7 | FOOD EXPERIMENTS Begin block

T4 | exp-intro-food1 | Experiments Intro Food 1 Text

In the following part of the survey we will take you through a series of shopping scenarios. There, you
will be asked to choose between different products.

By completing this choice experiment carefully and honestly it is possible to earn an additional bonus on
top of your usual incentive for completing the questionnaire.

When shopping in reality there are many features, which can be considered when buying food. In the
following scenarios we would like you to focus on a selection of the most common features except price
and brand.

We would like to know more about your preferences when comparing different offers for certain types of
food. On each of the following pages we will show you two choices. When comparing the two, some food
information may differ, whereas others may be identical.

B8 | CHOICE TASK: TRANS FATS Begin block

Q26 | split-tf | Treatment splits: Trans fats Single coded

Dummy
[DP: Equal sample size across splits - assign at random to split with fewest respondents]

1 Split 1: education - with TF info

2 Split 2: no education - with TF info
3 Split 3: education - no TF info
4 Split 4: no education - no TF info

Researcher notes: Translator: no translation required

EC Food Information - MCP Survey 2013 | version 2 | © TNS 11

T5 | exp-intro-tf | Experiments Intro Trans Fats Text
Not back
Please make the choices as you would in real life.
For example, you said earlier that you ... [DP: add if Q11 ST codes 3-4 or Q13 NOT code 3 or Q14 code 2
or Q22 code 5]

- watch out for the healthier alternative when selecting a product [DP: add if in Q11 ST5 codes 3-4]
- watch out for information on the amount of saturates [DP: add if in Q22 code 5]
- are following a diet or a weight loss programme [DP: add if in Q14 code 2]
- consider trans fat to be [DP: if in Q26 code 1 or 2 and if Q13 ST1 not code 3, then answer from Q13
ST1: very healthy / quite healthy / quite unhealthy / very unhealthy]
- consider partially hydrogenated oil to be [DP: if Q13 ST2 not code 3, then insert answer from Q13 ST2
very healthy / quite healthy / quite unhealthy / very unhealthy]
- consider fully hydrogenated oil to be [DP: if Q13 ST3 not code 3, then insert answer from Q13 ST3 very
healthy / quite healthy / quite unhealthy / very unhealthy]
- consider saturates to be [DP: if Q13 ST4 not code 3, then insert answer from Q13 ST4 very healthy /
quite healthy / quite unhealthy / very unhealthy]

[DP: add text if (Q19 ST1 code 1) respondent never bought product before:]
On the following screens we will ask you to shop crisps. You mentioned before, that you never bought
this type of product. Therefore, please assume you are choosing the crisps for a friend.

Please consider this also in your subsequent choices.

Scripter notes: Show this screen only if any of the conditions listed apply.

ASK ONLY IF Q26 | split-tf=1,3

T6 | edu-tf1 | Education on Trans Fats Text

Show picture(s): Show picture for education on trans fat

Before choosing, please read the following information carefully:

There are different kinds of fat included in most types of food. Many people know that saturated fat is
unhealthy while unsaturated fat is generally healthy. On a food label, saturated fat is labelled as
'saturates'.
It is also recommended that people avoid trans fat, which is even more unhealthy than saturated fat.
A food may contain hydrogenated oil. This oil is either hydrogenated fully or partially. Fully hydrogenated
oil contains only saturated fat and does not contain trans fat, while partially hydrogenated oil contains
both saturated fat and trans fat in different quantities.
[DP: Insert education image below this text]

Researcher notes: Translator notes: Please check/ advice whether the sentence "On a food
label, saturated fat is labelled as 'saturates'." is really needed in your language. If not, skip it.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 12

 ASK ONLY IF Q26 | split-tf=1,2

Q271 | exp-tf1-with | Trans fats choice 1 - with TFA Single coded

Show picture(s): insert 2 product images

Not back
Which product would you choose A or B?

1 Product A
2 No difference to me
3 Product B

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only separate 3 zooming areas per screen. For the experiment with crisps it will be:
-Ingredients
-Nutrition information
-Origin information „made in“
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

ASK ONLY IF Q26 | split-tf=3,4

Q272 | exp-tf1-without | Trans fats choice 1 - without TFA Single coded

Show picture(s): insert 2 product images

Not back
Which product would you choose A or B?

1 Product A
2 No difference to me
3 Product B

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with crisps it will be:
-Ingredients
-Nutrition information
-Origin information „made in“
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q271

EC Food Information - MCP Survey 2013 | version 2 | © TNS 13

Q273 | dum-zoom-tf1 | Zooming areas TF1 Multi coded
Min 1 | Max 3 | Dummy

Tracking what area was zoomed in Q271 or Q272

1 Ingredients
2 Country of origin
3 Nutrition facts
4 None *Exclusive *Position fixed

Researcher notes: Translator: no translation required

ASK ONLY IF Q26 | split-tf=1,2

Q281 | exp-tf2- with | Trans fats choice 2 - with TFA Single coded

Show picture(s): insert 2 product images

Not back

Which product would you choose C or D?

1 Product C
2 No difference to me
3 Product D

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only separate 3 zooming areas per screen. For the experiment with crisps it will be:
-Ingredients
-Nutrition information
-Origin information „made in“
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q271. Just please replace A or B
by C or D.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 14

 ASK ONLY IF Q26 | split-tf=3,4

Q282 | exp-tf2- without | Trans fats choice 2 - without TFA Single coded

Show picture(s): insert 2 product images

Not back

Which product would you choose C or D?

1 Product C
2 No difference to me
3 Product D

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only separate 3 zooming areas per screen. For the experiment with crisps it will be:
-Ingredients
-Nutrition information
-Origin information „made in“
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q281.

Q283 | dum-zoom-tf2 | Zooming areas TF2 Multi coded

Min 1 | Max 3 | Dummy

Tracking what area was zoomed in Q281 or Q282

1 Ingredients
2 Country of origin
3 Nutrition facts
4 None *Exclusive *Position fixed

Researcher notes: Translator: no translation required

Q29 | tf-health-cho | Always healthier choice for Trans fat Single coded
Dummy

1 Not always healthier choice [DP: if Q271 | exp-tf1 OR Q281| exp-tf2 are not code 1 OR Q272 |
exp-tf1 OR Q282| exp-tf2 are not code 1]
2 Always healthier choice [DP: if Q271 | exp-tf1 code 1 AND Q281 | exp-tf2 code 1 AND Q272 |
exp-tf1 code 1 AND Q282 | exp-tf2 code 1]

Researcher notes: Translator: no translation required

EC Food Information - MCP Survey 2013 | version 2 | © TNS 15

Q30 | rationale-tf2 | Rationale for second choices on trans fats Multi coded
Not back
What was guiding your choice in the previous question between products C and D?

Tick all that apply [DP: show only if in Q281 OR Q282 code 1 or 3]

Rotated

1 Lowest amount of saturated fat [DP: show only if in Q281 OR Q282 code 1 or 3]
2 Lowest amount of trans fat [DP: show only in relevant splits and if in Q281 OR Q282 code 1 or 3]
3 Without partially hydrogenated oil [DP: show only if in Q281 OR Q282 code 1 or 3]
4 Without fully hydrogenated oil [DP: show only if in Q281 OR Q282 code 1 or 3]
5 The ranking of ingredients [DP: show only if in Q281 OR Q282 code 1 or 3]
6 I was not able to spot any differences [DP: show only if in Q281 OR Q282 code 2] *Exclusive
*Position fixed
7 The differences I saw were not relevant for me [DP: show only if in Q281 OR Q282 code 2]
*Exclusive *Position fixed
8 I did not understand the difference [DP: show only if in Q281 OR Q282 code 2] *Exclusive
*Position fixed
9 Don't know *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 9.

Q31 | rationale-tf1 | Rationale for first choice on trans fats Multi coded
Not back
And what was guiding your choice in the illustrated question before: between product A and B?

Tick all that apply [DP: show only if in Q271 OR Q272 code 1 or 3]

Rotated

1 Lowest amount of saturated fat [DP: show only if in Q271 OR Q272 code 1 or 3]
2 Lowest amount of trans fat [DP: show only in relevant splits and if in Q271 OR Q272 code 1 or 3]
3 Without partially hydrogenated oil [DP: show only if Q271 OR Q272 code 1 or 3]
4 Without fully hydrogenated oil [DP: show only if Q271 OR Q272 code 1 or 3]
5 The ranking of ingredients [DP: show only if Q271 OR Q272 code 1 or 3]
6 I was not able to spot any differences [DP: show only if Q271 OR Q272 code 2] *Exclusive
*Position fixed
7 The differences I saw were not relevant for me [DP: show only if Q271 OR Q272 code 2]
*Exclusive *Position fixed
8 I did not understand the difference [DP: show only if in Q271 OR Q272 code 2] *Exclusive
*Position fixed
9 Don't know *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 9.

Translator note: Please just copy and paste the answers. Same answers like in Q30.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 16

Q32 | tf-truth | Trans fats myths and truths Matrix
Not back
The following statements on certain types of fat can be true or false. For each one, please tell us what
you think about it.

Rotated
True False Don't know
Saturated fat is less   
damaging to health
than trans fat if you
compare equal
amounts
Fully hydrogenated oil   
contains trans fat
Partially hydrogenated   
oil may contain trans
fat
Bakery foods like cakes   
or cookies never
contain to trans fat
Trans fat can be found   
in milk products

Scripter notes: IF respondent forgets some answers in a matrix question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer in line [insert and highlight statement]’ AND add ‘Don’t know’
option with code 8.

Researcher notes: Translator notes: Since we are still waiting for feedback please translate
also first statement.

B8 | CHOICE TASK: TRANS FATS End block

B9 | CHOICE TASK: ALLERGENS & PRIMARY INGREDIENT Begin block

Q33 | split-allergy | Treatment splits: Allergy Single coded

Dummy
[DP: Equal sample size across splits - assign at random to split with fewest respondents]

1 Split 1: may contain

2 Split 2: may contain traces
3 Split 3: cannot guarantee nut free
4 Split 4: made in a facility that also processes

Researcher notes: Translator: no translation required

EC Food Information - MCP Survey 2013 | version 2 | © TNS 17

T7 | intro-allergy | Experiment intro allergen labels Text
Not back
Imagine a friend is visiting you for a few days. Your guest suffers from a severe food allergy and should
not consume any NUTS. He mentioned that he loves MUESLI for breakfast.
Therefore, when we ask you on the following screens to select a muesli for this visit, you should select
between the options as you normally do - and in addition, you should keep in mind your friends' food
allergy.
Please make the choices as you would in real life.

[DP: add if Q17 code 2-4 or Q22 codes 1, 7 or 10]

For example, you said earlier that you …
- read allergen information [DP: add if Q17 codes 3-4]
… regularly [DP: add if Q17 code 4]
… only occasionally, e.g. when shopping for/with someone else [DP: add if Q17 code 3]
- have read allergen information previously [DP: add if Q17 code 2]
- when selecting muesli for the first time – you watch out for [DP: add if Q22 codes 1, 7 or 10]
… allergen information [DP: add if in Q22 code 1 for Muesli]
… country of origin [DP: add if in Q22 code 7 for Muesli]
… ingredients [DP: add if in Q22 code 10 for Muesli]
Please consider this also in your subsequent choices. [DP: add if Q17 code 2-4 or Q22 codes 1, 7 or 10]

B10 | ROTATE ORDER OF QUESTIONS IN THIS BLOCK Begin block

ASK ONLY IF Q33 | split-allergy=1

Q341 | exp-all1-split1 | Allergen choice 1 - May contain Single coded

Show picture(s): insert 2 product images

Not back
Which product would you choose A or B?

1 Product A
2 No difference to me
3 Product B

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q271.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 18

 ASK ONLY IF Q33 | split-allergy=2

Q342 | exp-all1-split2 | Allergen choice 1 - May contain traces Single coded

Show picture(s): insert 2 product images

Not back
Which product would you choose A or B?

1 Product A
2 No difference to me
3 Product B

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.
Translator note: Please just copy and paste. Same text like in Q271.

ASK ONLY IF Q33 | split-allergy=3

Q343 | exp-all1-split3 | Allergen choice 1 - Cannot guarantee nut Single coded

free

Show picture(s): insert 2 product images

Not back
Which product would you choose A or B?

1 Product A
2 No difference to me
3 Product B

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q271.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 19

 ASK ONLY IF Q33 | split-allergy=4

Q344 | exp-all1-split4 | Allergen choice 1 - Made in a facility that Single coded

also processes nuts

Show picture(s): insert 2 product images

Not back
Which product would you choose A or B?

1 Product A
2 No difference to me
3 Product B

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q271.

Q345 | dum-zoom-allergy1 | Zooming areas Allergy task 1 Multi coded

Min 1 | Max 3 | Dummy

Tracking what area was zoomed in Q341 - Q344

1 Ingredients
2 Country of origin
3 Nutrition facts
4 None *Exclusive *Position fixed

Researcher notes: Translator: no translation required

ASK ONLY IF Q33 | split-allergy=1

Q351 | exp-all2-split1 | Allergen choice 2 - May contain Single coded

Show picture(s): insert 2 product images

Not back
Which product would you choose C or D?

1 Product C
2 No difference to me
3 Product D

EC Food Information - MCP Survey 2013 | version 2 | © TNS 20

 Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.
Translator note: Please just copy and paste. Same text like in Q281.

ASK ONLY IF Q33 | split-allergy=2

Q352 | exp-all2-split2 | Allergen choice 2 - May contain traces Single coded

Show picture(s): insert 2 product images

Not back
Which product would you choose C or D?

1 Product C
2 No difference to me
3 Product D

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q281.

ASK ONLY IF Q33 | split-allergy=3

Q353 | exp-all2-split3 | Allergen choice 2 - Cannot guarantee nut Single coded

free

Show picture(s): insert 2 product images

Not back
Which product would you choose C or D?

1 Product C
2 No difference to me
3 Product D

EC Food Information - MCP Survey 2013 | version 2 | © TNS 21

 Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q281.

ASK ONLY IF Q33 | split-allergy=4

Q354 | exp-all2-split4 | Allergen choice 2 - Made in a facility that Single coded

also processes nuts

Show picture(s): insert 2 product images

Not back

Which product would you choose C or D?

1 Product C
2 No difference to me
3 Product D

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

C Translator note: Please just copy and paste. Same text like in Q281.

Q355 | dum-zoom-allergy2 | Zooming areas Allergy task 2 Multi coded

Min 1 | Max 3 | Dummy

Tracking what area was zoomed in Q351 - Q354

1 Ingredients
2 Country of origin
3 Nutrition facts
4 None *Exclusive *Position fixed

Researcher notes: Translator: no translation required

ASK ONLY IF Q33 | split-allergy=1

Q361 | exp-all3-split1 | Allergen choice 3 - May contain Single coded

Show picture(s): insert 2 product images

Not back
Which product would you choose E or F?

EC Food Information - MCP Survey 2013 | version 2 | © TNS 22

1 Product E
2 No difference to me
3 Product F

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q281. Please replace C or D by E
or F

ASK ONLY IF Q33 | split-allergy=2

Q362 | exp-all3-split2 | Allergen choice 3 - May contain traces Single coded

Show picture(s): insert 2 product images

Not back
Which product would you choose E or F?

1 Product E
2 No difference to me
3 Product F

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q361.

ASK ONLY IF Q33 | split-allergy=3

Q363 | exp-all3-split3 | Allergen choice 3 - Cannot guarantee nut Single coded

free

Show picture(s): insert 2 product images

Not back
Which product would you choose E or F?

1 Product E
2 No difference to me
3 Product F

EC Food Information - MCP Survey 2013 | version 2 | © TNS 23

 Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q361.

ASK ONLY IF Q33 | split-allergy=4

Q364 | exp-all3-split4 | Allergen choice 3 - Made in a facility that Single coded

also processes nuts

Show picture(s): insert 2 product images

Not back
Which product would you choose E or F?

1 Product E
2 No difference to me
3 Product F

Scripter notes: When the respondent sees the first screen, s/he cannot read the details of the food labels
but the headers must be readable, so that there will be a conscious choice, whether and which labels are
zoomed in, which we also need to track for our analysis.
There will be only 3 separate zooming areas per screen. For the experiment with muesli it will be:
-Ingredients with or without allergy information
-Nutrition facts
-Country of origin of primary ingredients
If a respondent zooms into e.g. „Ingredients“, then both ingredient labels will be shown at the same
time.

Translator note: Please just copy and paste. Same text like in Q361.

Q365 | dum-zoom-allergy3 | Zooming areas Allergy task 3 Multi coded

Min 1 | Max 3 | Dummy

Tracking what area was zoomed in Q361- Q364

1 Ingredients
2 Country of origin
3 Nutrition facts
4 None *Exclusive *Position fixed

Researcher notes: Translator: no translation required

EC Food Information - MCP Survey 2013 | version 2 | © TNS 24

Q37 | all-sec-cho | Always secure choice for Allergens Single coded
Dummy | Not back

1 Not always secure choice [DP: if Q341-Q344 NOT code 3 OR Q351-Q354 NOT code 1 OR Q361 -
Q364 NOT code 3]
2 Always secure choice [DP: Q341-Q344 code 3 AND Q351-Q354 code 1 AND Q361-Q364 code 3]

Researcher notes: Translator: no translation required

Q38 | num-sec-cho | Number of secure choice for Allergens Numeric

Max 3 | Dummy

Count the number of secure choices: 1 point for every time a respondent has code 3 in Q341-Q344, code
1 in Q351-Q354 or code 3 in Q361-Q364

Researcher notes: Translator: no translation required

B10 | ROTATE ORDER OF QUESTIONS IN THIS BLOCK End block

Q39 | rationale-all-pi | Rationale for choices on allergen and Multi coded

primary ingredient
Not back
What guided your choice in the previous questions?

Tick all that apply.

Rotated

1 Product without allergen information

2 Allergen information was too vague
3 Local ingredients
4 Imported ingredients
5 I was not able to spot any differences (at least sometimes) *Position fixed
6 The differences I saw were not relevant for me (at least sometimes) *Position fixed
8 Don't know *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 25

Q40 | risk-allergy | Risk assessment of precautionary allergen Matrix
label options

Based on your understanding of the risk for someone with a nuts allergy, how safe would you consider
muesli to be if it had the following information on the pack?

Random
Definitely safe Probably safe Not sure Probably not Definitely not
safe safe
May contain traces of     
nuts
May contain nuts     
Cannot guarantee nut     
free
Made in a facility that     
also processes nuts
Contains nuts     
No information given     
about nuts

Scripter notes: - IF respondent forgets some answers in a matrix question, THEN write ‘Please review
your responses on this page. One or more questions require further input.’ at top of the screen AND after
the question text ‘Please provide an answer in line [insert and highlight statement]’.

Q41 | recall-pi1 | Recall of primary ingredient in Muesli Single coded

What was the primary ingredient in the different muesli choices?

Rotated

1 Whole grain oats

2 Nuts
3 Raisins
4 Whole grain wheat
5 Mixed dried fruits
6 Sugar
8 Don't remember *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t remember’ option with code 8.

Q42 | pi-pizza | Primary ingredient in pizza Single coded

If we now look at other products, what do you consider the primary ingredient to be in this pizza?

Rotated

1 Flour
2 Tomato Sauce
3 Mozzarella cheese
4 Salami
8 Don't know *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 26

Q43 | pi-yoghurt | Primary ingredient in yoghurt Single coded

What do you consider to be the primary ingredient in this yoghurt?

Rotated

1 Milk
2 Sugar
3 Berries
8 Don't know *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

B9 | CHOICE TASK: ALLERGENS & PRIMARY INGREDIENT End block

Q44 | food-info | Expected food information Multi coded

Not back
The previous shopping scenarios provided you with less information than normal. On the other hand,
some of the information presented is not yet provided on food sold in your country. Therefore, we are
highly interested in your expectations on the ideal set of information for your future shopping.

Based on what you know by now, what information should be provided to you for making more informed
choices when shopping food?

Tick all that apply

Rotated

1 Allergen information based on actual ingredients

2 Precautionary allergen information
3 Amount of carbohydrates
4 Amount of sugar
5 Amount of salt
6 Amount of saturated fat
7 Amount of trans fat
8 Calories
9 Country of origin
10 Country of origin for primary ingredient
11 Expiry date
12 Health claims
13 Ingredients
14 Organically produced
15 Price
16 Other *Position fixed
17 None of these *Exclusive *Position fixed
18 Don't know *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 18.

B7 | FOOD EXPERIMENTS End block

EC Food Information - MCP Survey 2013 | version 2 | © TNS 27

 ASK ONLY IF Q25 | seg-exp=2

B11 | DRINKS EXPERIMENTS Begin block

Q45 | alc-pref | Choice preferences for alcohol Multi coded

When choosing alcoholic beverages, which aspects do you usually take into account?

Tick all that apply

Rotated

1 Taste
2 Brand name
3 Low price/ value
4 High price/ value
5 Low alcoholic content
6 High alcoholic content
7 Small serving size
8 Large serving size
9 Fewer calories
10 Suitability for meal or occasion
11 Recommendation
12 Popular drink at this occasion
13 Don't know *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 13.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 28

Q46 | seg-drinks | Respondent segmentation - treatment splits on Single coded
alcohol
Dummy
[DP: selection rules for subsequent questions covering certain alcohol types based on Q23
a) if respondent drinks only one type, then select only this type
b) if respondent drinks more types, then select the type [beer, wine, spirits] with the lowest number of
respondents
c) within the alcohol type select split between calories and limits equally (lowest number of respondents)
d) within spirit type select the type [vodka or whiskey] with the lowest number of respondents
The goal is to achieve that one third of respondents gets questions about beer, one third about wine and
another third about spirits - provided that enough respondents within a country consumed these alcohol
types within past 12 months. Thereby the spirit third is divided into vodka and whiskey.
Hence the sample sizes should be like this:
Split 1: 1/6 of respondents
Split 2: 1/6 of respondents
Split 3: 1/6 of respondents
Split 4: 1/6 of respondents
Split 5: 1/12 of respondents
Split 6: 1/12 of respondents
Split 7: 1/12 of respondents
Split 8: 1/12 of respondents
1/6 equates ca. 667 respondents and 1/12 equates ca. 333 respondents]

1 Split 1: Beer - calories

2 Split 2: Beer - limits
3 Split 3: Wine - calories
4 Split 4: Wine - limits
5 Split 5: Vodka - calories
6 Split 6: Vodka - limits
7 Split 7: Whiskey - calories
8 Split 8: Whiskey - limits

Researcher notes: Translator: no translation required

B12 | PREVIOUS ALCOHOL CONSUMPTION Begin block

EC Food Information - MCP Survey 2013 | version 2 | © TNS 29

 ASK ONLY IF Q46 | seg-drinks=1,2

Q471 | alc-occa-beer | Drinking occasions - Beer Multi coded

Not back
In which of the following situations have you drunk beer in the past 12 months?

Tick all that apply

Rotated

1 At home
2 At a friends´ home
3 At workplace, in the office
4 In a restaurant or pub with a meal
5 In a pub, bar, night club without a meal
6 In a public place e.g. concert, festivals
7 Other situation(s)/occasion(s) *Position fixed
8 Don't remember *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add "Don't remember" option with code 8.

ASK ONLY IF Q46 | seg-drinks=3,4

Q472 | alc-occa-wine | Drinking occasions - Wine Multi coded

Not back
In which of the following situations have you drunk wine in the past 12 months?

Tick all that apply

Rotated

1 At home
2 At a friends´ home
3 At workplace, in the office
4 In a restaurant or pub with a meal
5 In a pub, bar, night club without a meal
6 In a public place e.g. concert, festivals
7 Other situation(s)/occasion(s) *Position fixed
8 Don't remember *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add "Don't remember" option with code 8.

Translator note: Please just copy and paste. Same text like in Q471. But wine instead of beer.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 30

 ASK ONLY IF Q46 | seg-drinks=5,6,7,8

Q473 | alc-occa-spirits | Drinking occasions - Spirits Multi coded

Not back
In which of the following situations have you drunk [DP: if split 5 or 6 insert vodka / if split 7 or 8 insert
whiskey] in the past 12 months?

Tick all that apply

Rotated

1 At home
2 At a friends´ home
3 At workplace, in the office
4 In a restaurant or pub with a meal
5 In a pub, bar, night club without a meal
6 In a public place e.g. concert, festivals
7 Other situation(s)/occasion(s) *Position fixed
8 Don't remember *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add "Don't remember" option with code 8.

Translator note: Please just copy paste. Same text like in Q471. But vodka/ whiskey instead of
beer.

ASK ONLY IF Q46 | seg-drinks=1,2

Q481 | alc-freq-beer | Frequency of drinking alcohol - Beer Single coded

How often in the past 12 months have you drunk beer?

Please consider all occasions from drinking at home or when going out.

6 4 times a week or more often

5 2 to 3 times a week
4 Once a week
3 2 to 3 times a month
2 Once a month
1 Less often
8 Don't remember

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add "Don't remember" option with code 8.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 31

 ASK ONLY IF Q46 | seg-drinks=3,4

Q482 | alc-freq-wine | Frequency of drinking alcohol - Wine Single coded

How often in the past 12 months have you drunk wine?

Please consider all occasions from drinking at home or when going out.

6 4 times a week or more often

5 2 to 3 times a week
4 Once a week
3 2 to 3 times a month
2 Once a month
1 Less often
8 Don't remember

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add "Don't remember" option with code 8.

Translator note: Please just copy and paste. Same text like in Q481. But wine instead of beer.

ASK ONLY IF Q46 | seg-drinks=5,6,7,8

Q483 | alc-freq-spirits | Frequency of drinking alcohol - Spirits Single coded

How often in the past 12 months have you drunk [DP: if split 5 or 6 insert vodka / if split 7 or 8 insert
whiskey]?

Please consider all occasions from drinking at home or when going out.

6 4 times a week or more often

5 2 to 3 times a week
4 Once a week
3 2 to 3 times a month
2 Once a month
1 Less often
8 Don't remember

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add "Don't remember" option with code 8.

Translator note: Please just copy and paste. Same text like in Q481. But vodka/ whiskey
instead of beer.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 32

 ASK ONLY IF Q46 | seg-drinks=1,2

Q491 | alc-occa-last-beer | Drinking occasions - last time - Beer Single coded

When was the last time you drank beer?

Rotated

1 At home
2 At a friends´ home
3 At workplace, in the office
4 In a restaurant or pub with a meal
5 In a pub, bar, night club without a meal
6 In a public place e.g. concert, festivals
7 Other situation(s)/occasion(s) *Position fixed
8 Don't remember *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add "Don't remember" option with code 8.

Translator note: Please copy and paste the answers. Same answers like in Q471.

ASK ONLY IF Q46 | seg-drinks=3,4

Q492 | alc-occa-last-wine | Drinking occasions - last time - Wine Single coded

When was the last time you drank wine?

Rotated

1 At home
2 At a friends´ home
3 At workplace, in the office
4 In a restaurant or pub with a meal
5 In a pub, bar, night club without a meal
6 In a public place e.g. concert, festivals
7 Other situation(s)/occasion(s) *Position fixed
8 Don't remember *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add "Don't remember" option with code 8.

Translator note: Please just copy and paste. Same text like in Q491. But wine instead of beer.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 33

 ASK ONLY IF Q46 | seg-drinks=5,6,7,8

Q493 | alc-occa-last-spirits | Drinking occasions - last time - Single coded

Spirits

When was the last time you drank [DP: if split 5 or 6 insert vodka / if split 7 or 8 insert whiskey]?

Rotated

1 At home
2 At a friends´ home
3 At workplace, in the office
4 In a restaurant or pub with a meal
5 In a pub, bar, night club without a meal
6 In a public place e.g. concert, festivals
7 Other situation(s)/occasion(s) *Position fixed
8 Don't remember *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add "Don't remember" option with code 8.

Translator note: Please just copy and paste. Same text like in Q491. But vodka/whiskey
instead of beer.

ASK ONLY IF Q46 | seg-drinks=1,2

Q50 | beer-vol0 | Alcohol consumption per occasion - beer Single coded

In Europe, the average serving size of a small glass of beer is a quarter of a litre (0.25l) [DP: insert for
UK: " i.e. just under half a pint"]. With this in mind, roughly how much did you drink the last time you
drank beer?

If you had mixed drinks (e.g. shandy), please try to estimate only the volume of beer.

1 1 small glass or less (up to 0.25l)

2 2 small glasses (up to 0.5l)
3 3-4 small glasses (up to 1l)
4 5-6 small glasses (up to 1.5l)
5 7-8 small glasses (up to 2l)
6 More than 8 small glasses (2l)
8 Don't know

Scripter notes: Scripter notes: - IF respondent skips this question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer’ AND add ‘Don’t know’ option with code 8.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 34

 ASK ONLY IF Q46 | seg-drinks=3,4

Q51 | wine-vol0 | Alcohol consumption per occasion - wine Single coded

In Europe, the average serving size of a small glass of wine is about 0.1 litre (or 100 millilitres). With this
in mind, roughly how much did you drink the last time you drank wine?

If you had mixed drinks (e.g. spritzer), please try to estimate only the volume of wine.

1 1 small glass or less (up to 0.1l = 100ml)

2 2 small glasses (up to 0.2l = 200ml)
3 3-4 small glasses (up to 0.4l = 400ml)
4 5-6 small glasses (up to 0.6l = 600ml)
5 7-10 small glasses (up to 1l)
6 More than 10 small glasses (1l)
8 Don't know

Scripter notes: Scripter notes: - IF respondent skips this question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer’ AND add ‘Don’t know’ option with code 8.

ASK ONLY IF Q46 | seg-drinks=5,6,7,8

Q52 | spirits-vol0 | Alcohol consumption at last occasion - Single coded

vodka/whiskey

In Europe, the average serving size of a small shot of [DP: insert vodka if split 5 or 6 and whiskey if split
7 or 8] is about 0.01 litre (or 1cl). With this in mind, roughly how much did you drink the last time you
drank [DP: insert vodka if split 5 or 6 and whiskey if split 7 or 8]?

If you had mixed drinks (e.g. with soda water or as an ingredient of a cocktail), please try to estimate
only the volume of [DP: insert vodka if split 5 or 6 and whiskey if split 7 or 8].

1 1 small shot or less (up to 0.01l = 1cl)

2 2 small shots (up to 0.02l = 2cl)
3 3-4 small shots (up to 0.04l = 4cl)
4 5-6 small shots (up to 0.06l = 6cl)
5 7-10 small shots (up to 0.1l = 10cl)
6 More than 10 small shots (0.1l = 10cl)
8 Don't know

Scripter notes: Scripter notes: - IF respondent skips this question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer’ AND add ‘Don’t know’ option with code 8.

B12 | PREVIOUS ALCOHOL CONSUMPTION End block

T8 | exp-intro-drinks | Experiments Intro Drinks Text

Not back
In the following part of the survey we will take you through a short series of decision scenarios. There,
you will be asked about your preferences when drinking a certain type of alcoholic beverage.

By completing this choice experiment carefully and honestly it is possible to earn an additional bonus on
top of your usual incentive for completing the questionnaire.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 35

 ASK ONLY IF Q46 | seg-drinks=1,2

Q53 | beer-vol1 | Beer consumption ex ante Single coded

Imagine you are invited to a party at a friends' home with enough food and drink for everyone. There are
non-alcoholic beverages as well as beer available for you to help yourself. Assume that you do not need
to drive a car afterwards or have to work the next day.

How much beer would you drink at this occasion?

Please select one answer with the most likely volume based on your previous habits for this or a similar
occasion. The answer options below are again based on the average serving size of a small glass of beer
in Europe.

1 1 small glass or less (up to 0.25l)

2 2 small glasses (up to 0.5l)
3 3-4 small glasses (up to 1l)
4 5-6 small glasses (up to 1.5l)
5 7-8 small glasses (up to 2l)
6 More than 8 small glasses (2l)
8 Don't know

Scripter notes: Scripter notes: - IF respondent skips this question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer’ AND add ‘Don’t know’ option with code 8.

Translator note: Please just copy and paste the answers. Same answers like in Q50.

ASK ONLY IF Q46 | seg-drinks=3,4

Q54 | wine-vol1 | Wine consumption ex ante Single coded

Imagine you are invited to a party at a friends' home with enough food and drink for everyone. There are
non-alcoholic beverages as well as wine available for you to help yourself.
Assume that you do not need to drive a car afterwards or have to work the next day.

How much wine would you drink at this occasion?

Please select one answer with the most likely volume based on your previous habits for this or a similar
occasion. The answer options below are again based on the average serving size of a small glass of wine
in Europe.

1 1 small glass or less (up to 0.1l = 100ml)

2 2 small glasses (up to 0.2l = 200ml)
3 3-4 small glasses (up to 0.4l = 400ml)
4 5-6 small glasses (up to 0.6l = 600ml)
5 7-10 small glasses (up to 1l)
6 More than 10 small glasses (1l)
8 Don't know

Scripter notes: Scripter notes: - IF respondent skips this question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer’ AND add ‘Don’t know’ option with code 8.

Translator note: Please just copy and paste. Same question like in Q53 and same answers like
in Q51. But wine instead of beer.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 36

 ASK ONLY IF Q46 | seg-drinks=5,6,7,8

Q55 | spirits-vol1 | Spirits consumption ex ante Single coded

Imagine you are invited to a party at a friends' home with enough food and drink for everyone. There are
non-alcoholic beverages as well as [DP: insert vodka if split 5, 6 and whiskey if split 7, 8] available for
you to help yourself. Assume that you do not need to drive a car afterwards or have to work the next
day.
How much [DP: insert vodka if split 5, 6 and whiskey if split 7, 8] would you drink at this occasion?

Please select one answer with the most likely volume based on your previous habits for this or a similar
occasion. The answer options below are again based on the average serving size of shot of [DP: insert
vodka if split 5, 6 and whiskey if split 7, 8] in Europe.

1 1 small shot or less (up to 0.01l = 1cl)

2 2 small shots (up to 0.02l = 2cl)
3 3-4 small shots (up to 0.04l = 4cl)
4 5-6 small shots (up to 0.06l = 6cl)
5 7-10 small shots (up to 0.1l = 10cl)
6 More than 10 small shots (0.1l = 10cl)
8 Don't know

Scripter notes: Scripter notes: - IF respondent skips this question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer’ AND add ‘Don’t know’ option with code 8.

Translator note: Please just copy and paste. Same question like in Q53 and same answers like
in Q51. But vodka/ whiskey instead of beer.

ASK ONLY IF Q46 | seg-drinks=1

B13 | TASK: BEER - CALORIES Begin block

Q56 | beer-cal-test | Knowledge test on calories for beer Single coded

Show picture(s): 2 images (beer and pizza)

Which contains more calories - the pizza half or half a litre of beer?

1 Beer has more

2 About the same
3 Pizza has more
8 Don't know

EC Food Information - MCP Survey 2013 | version 2 | © TNS 37

T9 | beer-cal-edu | Education - Calories in beer vs pizza Text
Not back
You said before
[DP: if in Q56 code 8]: you don't know whether pizza or beer has more calories.
[DP: if in Q56 code 3]: you think that pizza has more calories. This answer was correct. Well done!
[DP: if in Q56 code 2]: you think that beer and pizza has the same amount of calories. This answer was
not correct.
[DP: if in Q56 code 1]: you think that beer has more calories. This answer was not correct.
In fact, on average, half a litre [DP insert for UK: i.e. 0.9 pints] of BEER contains between 200 - 300
calories whereas 200 g of PIZZA contains at least 400 calories.

Q57 | beer-vol2-cal | Consumption change - Calories - Beer Single coded

Show picture(s): Beer bottle with calories

Knowing the amount of calories in a beer would you intend to drink less, the same or more at a friend’s
party than you said earlier?

-> Be quick and answer within 5 seconds!

1 I would drink less

2 I would drink about the same, i.e. [DP: insert answer from Q53]
3 I would drink more
8 Don't know

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

B13 | TASK: BEER - CALORIES End block

ASK ONLY IF Q46 | seg-drinks=2

B14 | TASK: BEER - LIMITS Begin block

Q58 | beer-lim-test | Knowledge test on limits for beer Single coded

Thinking about the officially recommended maximum amount of beer a person should drink on a day,
which of the following answers is correct?

Please note that 1 unit = 10 ml of pure alcohol. For example, one litre of beer with 5% alcohol contains 5
units of pure alcohol.

1 Men: 2-3 units / Women: 1-2 units

2 Men: 3-4 units / Women: 2-3 units
3 Men: 4-5 units / Women: 3-4 units
8 Don't know

EC Food Information - MCP Survey 2013 | version 2 | © TNS 38

T10 | beer-lim-edu | Education - Limits with beer Text
Not back
You said before
[DP: if in Q58 code 8]: you don't know the correct answer.

[DP: if in Q58 code 3]: you think that men should not exceed 4-5 units / women 3-4 units of pure
alcohol. This answer was not correct.

[DP: if in Q58 code 2]: you think that men should not exceed 3-4 units / women 2-3 units of pure
alcohol. This answer was correct. Well done!

[DP: if in Q58 code 1]: you think that men should not exceed 2-3 units / women 1-2 units of pure
alcohol. This answer was not correct.

In fact, on average, men should not exceed 3-4 units / women 2-3 units of pure alcohol per day in order
to sustain healthiness.

Q59 | beer-vol2-lim | Consumption change - Limits - Beer Single coded

Show picture(s): Beer bottle with health warning

Knowing the recommended limits of drinking alcohol would you intend to drink less, the same or more at
a friend’s party than you said earlier?

-> Be quick and answer within 5 seconds!

1 I would drink less

2 I would drink about the same, i.e. [DP: insert answer from Q53]
3 I would drink more
8 Don't know

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

B14 | TASK: BEER - LIMITS End block

ASK ONLY IF Q46 | seg-drinks=3

B15 | TASK: WINE - CALORIES Begin block

Q60 | wine-cal-test | Knowledge test on calories for wine Single coded

Show picture(s): calories wine and chocolate

Which contains more calories –a glass of wine or 20g of chocolate?

1 Wine has more

2 About the same
3 Chocolate has more
8 Don't know

EC Food Information - MCP Survey 2013 | version 2 | © TNS 39

T11 | wine-cal-edu | Education - Calories in wine vs. chocolate Text
Not back
You said before
[DP: if in Q60 code 8]: you don't know whether wine or chocolate has more calories.
[DP: if in Q60 code 3]: you think that chocolate has more calories. This answer was not correct.
[DP: if in Q60 code 2]: you think that wine and chocolate has the same amount of calories. This answer
was not correct.
[DP: if in Q60 code 1]: you think that wine has more calories. This answer was correct. Well done!
In fact, on average, 200 ml of WINE contains between 120-170 calories whereas 20 g of CHOCOLATE
contains about 100 calories.

Q61 | wine-vol2-cal | Consumption change - Calories - Wine Single coded

Show picture(s): Wine bottle with calories

Knowing the amount of calories in wine would you intend to drink less, the same or more at a friend’s
party than you said earlier?

-> Be quick and answer within 5 seconds!

1 I would drink less

2 I would drink about the same, i.e. [DP: insert answer from Q54]
3 I would drink more
8 Don't know

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

Translator note: Please just copy and paste. Same question like and answers like in Q57. But
wine instead of beer

B15 | TASK: WINE - CALORIES End block

ASK ONLY IF Q46 | seg-drinks=4

B16 | TASK: WINE - LIMITS Begin block

Q62 | wine-lim-test | Knowledge test on limits for wine Single coded

Thinking about the officially recommended maximum amount of wine a person should drink on a day,
which of the following answers is correct?

Please note that 1 unit = 10 ml of pure alcohol. For example, one litre of wine with 12% alcohol contains
12 units of pure alcohol.

1 Men: 2-3 units / Women: 1-2 units

2 Men: 3-4 units / Women: 2-3 units
3 Men: 4-5 units / Women: 3-4 units
8 Don't know
Translator note: Please just copy and paste. Same question and answers like in Q58. But wine
instead of beer

EC Food Information - MCP Survey 2013 | version 2 | © TNS 40

T12 | wine-lim-edu | Education - Limits with wine Text
Not back
You said before
[DP: if in Q62 code 8]: you don't know the correct answer.

[DP: if in Q62 code 3]: you think that men should not exceed 4-5 units / women 3-4 units of pure
alcohol. This answer was not correct.

[DP: if in Q62 code 2]: you think that men should not exceed 3-4 units / women 2-3 units of pure
alcohol. This answer was correct. Well done!

[DP: if in Q62 code 1]: you think that men should not exceed 2-3 units / women 1-2 units of pure
alcohol. This answer was not correct.

In fact, on average, men should not exceed 3-4 units / women 2-3 units of pure alcohol per day in order
to sustain healthiness.

Translator note: Please just copy and paste. Same text like in T10.

Q63 | wine-vol2-lim | Consumption change - Limits - Wine Single coded

Show picture(s): Wine bottle with health warning

Knowing the recommended limits of drinking alcohol would you intend to drink less, the same or more at
a friend’s party than you said earlier?

-> Be quick and answer within 5 seconds.

1 I would drink less

2 I would drink about the same, i.e. [DP: insert answer from Q54]
3 I would drink more
8 Don't know

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 8.

Translator note: Please just copy paste. Same question and answers like in Q59.

B16 | TASK: WINE - LIMITS End block

ASK ONLY IF Q46 | seg-drinks=5,7

B17 | TASK: SPIRITS - CALORIES Begin block

EC Food Information - MCP Survey 2013 | version 2 | © TNS 41

Q64 | spirits-cal-test | Knowledge test on calories for spirits Single coded

Show picture(s): vodka/ whiskey and nuts

If split 5 insert the image for vodka

If split 7 insert the image for whiskey

Which contains more calories - a handful of mixed nuts or a shot of [DP: if split 5 insert vodka and if split
7 insert whiskey]?
1 [DP: if split 5 insert Vodka/ if split 7 insert Whiskey] has more
2 About the same
3 Nuts have more
8 Don't know

Scripter notes: Please show for split 5 the picture of a shot of vodka and for split 7 a shot of whiskey

T13 | spirits-cal-edu | Education - Calories in spirits vs. nuts Text

Not back
You said before
[DP: if in Q64 code 8]: you don't know whether [if split 5: vodka/ if split 7: whiskey] or nuts has more
calories.

[DP: if in Q64 code 3]: you think that nuts has more calories. This answer was not correct.

[DP: if in Q64 code 2]: you think that [if split 5: if vodka/ if split 7: whiskey] and nuts has the same
amount of calories. This answer was correct. Well done!

[DP: if in Q64 code 1]: you think that [if split 5: vodka/ if split 7: whiskey] has more calories. This
answer was not correct.

In fact, on average, a 2cl shot of [if split 5: VODKA/ if split 7: WHISKEY] contains between 40-50 calories
whereas 10g of NUTS contains between 50-60 calories.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 42

Q65 | spirits-vol2-cal | Consumption change - Calories - Spirits Single coded

Show picture(s): vodka or whiskey with calories information

If split 5 insert the image for vodka

If split 7 insert the image for whiskey

Knowing the amount of calories in [DP: insert vodka if split 5/ whiskey if split 7 according to Q46] would
you intend to drink less, the same or more at a friend’s party than you said earlier?

-> Be quick and answer within 5 seconds!

1 I would drink less

2 I would drink about the same, i.e. [DP: insert answer from Q55]
3 I would drink more
8 Don't know

Scripter notes: if split 5 show bottle of vodka with calories

if split 7 show bottle of whiskey with calories

IF respondent skips this question, THEN write ‘Please review your responses on this page. One or more
questions require further input.’ at top of the screen AND after the question text ‘Please provide an
answer’ AND add ‘Don’t know’ option with code 8.

Translator note: Please just copy and paste. Same question like and answers like in Q57. But
vodka/whiskey instead of beer

B17 | TASK: SPIRITS - CALORIES End block

ASK ONLY IF Q46 | seg-drinks=6,8

B18 | TASK: SPIRITS - LIMITS Begin block

Q66 | spirits-lim-test | Knowledge test on limits for spirits Single coded

Thinking about the officially recommended maximum amount of [DP: if split 6 insert vodka/ if split 8
insert whiskey according to Q46] a person should drink on a day, which of the following answers is
correct?

Please note that 1 unit = 10 ml of pure alcohol. For example, one litre of [DP: insert vodka if split 6/
whiskey if split 8 according to Q46] with 40% alcohol contains 40 units of pure alcohol.

1 Men: 2-3 units / Women: 1-2 units

2 Men: 3-4 units / Women: 2-3 units
3 Men: 4-5 units / Women: 3-4 units
8 Don't know
Translator note: Please just copy and paste. Same question and answers like in Q58. But
vodka/whiskey instead of beer

EC Food Information - MCP Survey 2013 | version 2 | © TNS 43

T14 | spirits-lim-edu | Education - Limits with spirits Text
Not back
You said before
[DP: if in Q66 code 8]: you don't know the correct answer.

[DP: if in Q66 code 3]: you think that men should not exceed 4-5 units / women 3-4 units of pure
alcohol. This answer was not correct.

[DP: if in Q66 code 2]: you think that men should not exceed 3-4 units / women 2-3 units of pure
alcohol. This answer was correct. Well done!

[DP: if in Q66 code 1]: you think that men should not exceed 2-3 units / women 1-2 units of pure
alcohol. This answer was not correct.

In fact, on average, men should not exceed 3-4 units / women 2-3 units of pure alcohol per day in order
to sustain healthiness.

Translator note: Please just copy and paste. Same text like in T10.

Q67 | spirits-vol2-lim | Consumption change - Limits - Spirits Single coded

Show picture(s): vodka or whiskey with health warning

If split 7 insert the image for vodka

If split 8 insert the image for whiskey

Knowing the recommended limits of drinking alcohol would you intend to drink less, the same or more at
a friend’s party than you said earlier?

-> Be quick and answer within 5 seconds!

1 I would drink less

2 I would drink about the same, i.e. [DP: insert answer from Q55]
3 I would drink more
8 Don't know

Scripter notes: if split 6 show bottle of vodka with health warning

if split 8 show bottle of whiskey with health warning

IF respondent skips this question, THEN write ‘Please review your responses on this page. One or more
questions require further input.’ at top of the screen AND after the question text ‘Please provide an
answer’ AND add ‘Don’t know’ option with code 8.

Translator note: Please just copy and paste. Same question and answers like in Q59.

B18 | TASK: SPIRITS - LIMITS End block

EC Food Information - MCP Survey 2013 | version 2 | © TNS 44

 ASK ONLY IF Q46 | seg-drinks=1,2

Q68 | beer-vol3 | Future intention to drinking volume of beer Single coded

Based on what you know now do you think you will drink less, the same amount or more beer in future?

-> Be quick and answer within 5 seconds!

1 I would drink less

2 I would drink about the same
3 I would drink more
8 Don't know
Translator note: Please just copy and paste the answers. Same answers like in Q57.

ASK ONLY IF Q46 | seg-drinks=3,4

Q69 | wine-vol3 | Future intention to drinking volume of wine Single coded

Based on what you know now do you think you will drink less, the same amount or more wine in future?

-> Be quick and answer within 5 seconds!

1 I would drink less

2 I would drink about the same
3 I would drink more
8 Don't know
Translator note: Please just copy and paste. Same question and answers like in Q68. But wine
instead of beer.

ASK ONLY IF Q46 | seg-drinks=5,6,7,8

Q70 | spirits-vol3 | Future intention to drinking volume of spirits Single coded

Based on what you know now do you think you will drink less, the same amount or more [DP: if split 5,6
insert vodka/if split 7,8 insert whiskey] in future?

-> Be quick and answer within 5 seconds!

1 I would drink less

2 I would drink about the same
3 I would drink more
8 Don't know

Translator note: Please just copy and paste. Same question and answers like in Q68. But
vodka/ whiskey instead of beer.

Q701 | dum-alc-future | Future intention to drink volume Single coded

Consolidation of what was selected in Q68 – Q70

1 I would drink less [DP: if Q68 OR Q69 OR Q70 code 1]

2 I would drink about the same [DP: if Q68 OR Q69 OR Q70 code 2]
3 I would drink more [DP: if Q68 OR Q69 OR Q70 code 3]
8 Don’t know [DP: if Q68 OR Q69 OR Q70 code 8]

Researcher notes: Translator: no translation required

EC Food Information - MCP Survey 2013 | version 2 | © TNS 45

Q71 | alc-info | Expected food information on alcoholic drinks Multi coded

The previous questions provided you only with less information that normal. On the other hand, some of
the information presented is not yet provided on alcoholic beverages sold in your country.
Therefore, we are highly interested in your expectations on the ideal set of information for your future
consumption of alcohol.

Based on what you know by now, what information should be provided to you for making more informed
choices when shopping alcoholic beverages?

Tick all that apply

Rotated

1 Allergen information based on actual ingredients

2 Precautionary allergen information
3 Calories
4 Country of origin
5 Daily unit guidelines (recommended maximum limits)
6 Expiry date
7 Ingredients
8 Organically produced
9 Price
10 Alcohol level
11 Other *Position fixed
12 None of these *Exclusive *Position fixed
13 Don't know *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don’t know’ option with code 13.

Q72 | alc-att | Attitudes about alcohol consumption Matrix

To what extent do you agree or disagree with each of the following statements?

Rotated
Strongly agree Agree Disagree Strongly Don't know
disagree
Drinking alcohol is     
common among the
people I spend most of
my time with
Overall, alcohol     
consumption has more
advantages than
disadvantages for our
society

Scripter notes: IF respondent forgets some answers in a matrix question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer in line [insert and highlight statement]’ AND add ‘Don’t know’
option with code 8.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 46

Q73 | too-much | Experience with too much alcohol Single coded

Have you ever drunk too much alcohol e.g. you felt unwell afterwards?

1 Yes
2 No
9 Prefer not to say

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Prefer not to say’ option with code 9.

B11 | DRINKS EXPERIMENTS End block

B19 | FOOD WASTE Begin block

Q74 | bestbefore | Knowledge test: Best Before Single coded

You may have noticed that the products shown in some of the previous questions also displayed a label
with a date. What information is shown on this label?
[DP: insert best before label]

1 The last day on which the food is at its highest quality

2 The last day on which the food is safe to eat
3 The day the food must be sold by
4 None of these
8 Don't know

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’

Q75 | useby | Knowledge test: Use By Single coded

Not back
What information is shown on this label?
[DP: insert use by label]

1 The last day on which the food is at its highest quality

2 The last day on which the food is safe to eat
3 The day the food must be sold by
4 None of these
8 Don't know

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’

EC Food Information - MCP Survey 2013 | version 2 | © TNS 47

Q76 | waste1 | Consumed outdated food Single coded

Thinking about products that you keep in a fridge, have you ever eaten or cooked a product when it was
past its 'best before' date?
If yes, what is the longest period after the 'best before' date?

1 No, never
2 Yes, with a date that was 1-2 days old
3 Yes, with a date that was 3-4 days old
4 Yes, with a date that was 5-7 days old
5 Yes, with a date that was more than a week old
8 Don't know

Q77 | waste2 | Food thrown away before expiry Single coded

Have you ever thrown away a product that had not passed the best before date?
If yes, what is the earliest period before the 'best before' date when you have thrown something away?

1 No, never
2 Yes, 1-2 days before the date
3 Yes, 3-4 days before the date
4 Yes, 5-7 days before the date
5 Yes, more than a week before the date
8 Don't know

Client notes: The purpose of this question and the one before is to anchor the respondents imagination to
his/her individually experienced 'extreme' situations at both ends of possible actions from throwing
healthy food away to eating old and unhealthy food. Only if we ask both questions upfront, we may
minimise the desirability bias when the respondent answers the subsequent attitude questions. If we skip
this question, we expect to receive more desirable answers that are less reflecting real beliefs and
attitudes.

Q78 | waste-att1 | Food waste attitudes 1 Matrix

To what extent do you agree or disagree with each of the following statements?

Strongly Agree Neither Disagree Strongly Don't know

agree agree nor disagree
disagree
I always use more      
ways than the date
given on the label to
check if food is safe to
eat.
I prefer to throw food      
away than eat it when
it is past its best
It will be difficult for      
me to reduce the
amount of food I throw
away

Scripter notes: Scripter notes: - IF respondent forgets some answers in a matrix question, THEN write
‘Please review your responses on this page. One or more questions require further input.’ at top of the
screen AND after the question text ‘Please provide an answer in line [insert and highlight statement]’
AND add ‘Don’t know’ option with code 8.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 48

Q79 | waste-habit | Order in fridge Single coded

Do you arrange the food in your fridge by expiry date?

1 Yes
2 No
8 Don't know

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Don't know' option with code 8.

Q80 | waste-vol0 | Food waste volume Single coded

How would you describe the amount of food that you regularly throw away in your household - would you
say this is ...

5 A very high amount

4 A rather high amount
3 An average amount
2 A rather small amount
1 Hardly anything
8 Don't know

T15 | waste-fb1 | Food Waste knowledge test - Feedback1 Text

Not back
You said before
[DP: if in Q74 code 8]: you don't know what the label stating "best before" means.
[DP: if in Q74 code 4]: that none of the given answers explain the label "best before".
[DP: if in Q74 code 3]: you understand that the "best before" label indicates the day the food must be
sold by. This answer was not correct.
[DP: if in Q74 code 2]: you understand that the "best before" label indicates the last day on which the
food is safe to eat. This answer was not correct.
[DP: if in Q74 code 1]: you understand that the "best before" label indicates the last day on which the
food is at its highest quality. This answer was correct. Well done!

EC Food Information - MCP Survey 2013 | version 2 | © TNS 49

T16 | waste-edu1 | Education on Food Waste 1 Text

"Best before" indicates the date until when the food retains its expected quality.

Food is still safe to consume after the indicated 'best before' day on the condition that storage
instructions are respected and packaging is not damaged, but it might begin to lose its flavour and
texture.

'Best before' dates appear on a wide range of frozen, dried (pasta, rice), tinned and other foods
(vegetable oil, chocolate, etc.).

Check if the packaging is intact, and if the food looks, smells and tastes good before throwing away food
past its 'best before' date.

Once a food with a 'best before' date on it has been opened, follow any instructions such as 'eat within
three days of opening', when applicable.

Scripter notes: insert html version

T17 | waste-fb2 | Food Waste knowledge test - Feedback2 Text

Not back
You said before
[DP: if in Q75 code 8]: you don't know what the label stating "use by" means.
[DP: if in Q75 code 4]: that none of the given answers explain the label "use by".
[DP: if in Q75 code 3]: you understand that the "use by" label indicates the day the food must be sold
by. This answer was not correct.
[DP: if in Q75 code 2]: you understand that the "use by" label indicates the last day on which the food is
safe to eat. This answer was not correct.
[DP: if in Q75 code 1]: you understand that the "use by" label indicates the last day on which the food is
at its highest quality. This answer was correct. Well done!

EC Food Information - MCP Survey 2013 | version 2 | © TNS 50

T18 | waste-edu2 | Education on Food Waste 2 Text

'Use By' indicates the date until when the food can be eaten safely.

Don’t use any food after expiration of the 'use by' date.

'Use by' dates appear on highly perishable food, from the microbiological point of view, such as fresh fish,
fresh minced meat, etc.

Follow the storage instructions, such as 'keep in a refrigerator' or 'keep at 2-4° C'; if not the food will
spoil quicker and you may risk food poisoning.

By freezing the food at home soon after purchase, you can extend its life beyond the 'use by' date, if it is
frozen properly. But make sure you follow any instructions on the pack, such as 'freeze up to the use by
date', 'cook from frozen' or 'defrost thoroughly before use and use within 24 hours'.

Once a food with a 'use by' date on it has been opened, follow any instructions for storage and use, such
as 'eat within three days of opening', bearing in mind that food should be consumed before the expiration
of the 'use by' date.

Scripter notes: insert html version

Q81 | waste-att2 | Food waste attitude change Single coded

Based on what you know now - how do you think about your future way of dealing with food waste?
Would you intend to throw away the same or more or less food?

Be quick and answer within 5 seconds.

1 I would throw away less food

2 I would throw away the same amount of food
3 I would throw away more food
8 Don't know

Scripter notes: Scripter notes: - IF respondent skips this question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer’ AND add ‘Don’t know’ option with code 8.

B19 | FOOD WASTE End block

B20 | DEMOGRAPHICS Begin block

EC Food Information - MCP Survey 2013 | version 2 | © TNS 51

Q82 | att-control-risk | Self control & Risk attitude Matrix

How well does each of the following statements decribe you?

Strongly agree Agree Disagree Strongly Don't know

disagree
I much prefer doing     
things that pay off
right away than in the
future
Security is more     
important to me than
excitement and
adventure

Scripter notes: IF respondent forgets some answers in a matrix question, THEN write ‘Please review your
responses on this page. One or more questions require further input.’ at top of the screen AND after the
question text ‘Please provide an answer in line [insert and highlight statement]’ AND add ‘Don’t know’
option with code 8.

ASK ONLY IF Q6 | hhsize=2,3,4,5

Q83 | living-status | Living Status Multi coded

Who do you currently live with?

Rotated

1 Spouse/Partner
2 Children
3 Parents
4 Other family members
5 Friends
6 None of these *Exclusive *Position fixed
9 Prefer not to say *Exclusive *Position fixed

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Prefer not to say’ option with code 9.

Q84 | work-status | Working status Single coded

Which of the following best describes your current working status?

1 Working full-time (over 30 hours per week)

2 Working part-time
3 Temporarily unemployed/looking for work
4 Retired
5 Not working for other reasons (looking after family, ill etc.)
6 At school/college/university
9 Prefer not to say

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Prefer not to say’ option with code 9.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 52

 ASK ONLY IF not Q84 | work_status=6

Q85 | education | Education - terminal age Numeric

Min 7 | Max 99
What age were you when you finished full time education?

Scripter notes: IF respondent skips this question, THEN write ‘Please review your responses on this page.
One or more questions require further input.’ at top of the screen AND after the question text ‘Please
provide an answer’ AND add ‘Prefer not to say’ option with code 9.

[DP: compare answers in Q2 (age1) with answers in Q85 (education)] education age should be lower or
equal to current age

IF the terminal age is higher than the actual age, THEN write error message: 'Please ensure that the age
you have specified is not higher than your actual age'.

Q86 | income | Individual Income Single coded

What is your total annual personal income before tax?

1 [DP: insert for FR, DE, IT, ES,FI: 0 - 9 999 EUR/

for UK: 0 - 8 999 GBP /
for PL: 0 - 24 999 zł/
for RO: 0 - 14 999 lei]
2 [DP: insert for FR, DE, IT, ES,FI: 10 - 19 999 EUR/
for UK: 9 000 - 17 999 GBP /
for PL: 25 000 - 49 999 zł/
for RO: 15 000 - 29 999 lei]
3 [DP: insert for FR, DE, IT, ES,FI: 20 000 - 29 999 EUR/
for UK: 18 000 - 26 999 GBP /
for PL: 50 000 - 74 999 zł/
for RO: 30 000 - 44 999 lei]
4 [DP: insert for FR, DE, IT, ES,FI: 30 000 - 39 999 EUR/
for UK: 27 000 - 35 999 GBP /
for PL: 75 000 - 99 999 zł/
for RO: 45 000 - 59 999 lei]
5 [DP: insert for FR, DE, IT, ES,FI: 40 000 - 49 999 EUR/
for UK: 36 000 - 44 999 GBP /
for PL: 100 000 - 124 999 zł/
for RO: 60 000 - 74 999 lei]
6 [DP: insert for FR, DE, IT, ES,FI: 50 000 - 59 999 EUR/
for UK: 45 000 - 53 999 GBP /
for PL: 125 000 - 149 999 zł/
for RO: 75 000 - 89 999 lei]
7 [DP: insert for FR, DE, IT, ES,FI: 60 000 EUR/
for UK: 54 000 GBP /
for PL: 150 000 zł/
for RO: 90 000 lei]
and more
9 Prefer not to say

Scripter notes: Show 'Prefer not to say' from the beginning.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 53

 ASK ONLY IF Q1 | country=1

Q87 | height-uk | Height UK in feet and inches Numeric

Max 11
Now, here is the last but one question.
How tall are you without shoes?

Please enter your height in feet and inches. If you are for example exactly 6 feet tall, please enter 00 at
the field for inches.

Researcher notes: Translator: no translation required

Scripter notes: Please ADD "Prefer not to say" option code 9 and "Don't know" option code 8

and two numeric fields and the example:

_____ feet _______inches

e.g. 5 feet 09 inches

for feet: Min. 3 Max. 7

for inches: Min. 00 Max. 11

IF respondent gives an answer that is beyond limits. Please show an error message:
"Please type a digit in the range between 3 and 7 in the numeric box for feet and a digit in the range
between 00 and 11 in the numeric box for inches."

ASK ONLY IF Q1 | country=1

Q88 | height-dum-uk | Height UK in cm Numeric

Min 100 | Max 220 | Dummy

[DP: conversion from feet to inches and to cm]

Researcher notes: Translator: no translation required

Scripter notes: Filter only if Q1=1 and if in Q87 NOT code 8 or code 9

please convert feet to inches with this formula:

(feet * 12) + inches

and then from inches to cm with this formula

inches * 2.54

===> ((feet * 12) + inches) * 2.54

EC Food Information - MCP Survey 2013 | version 2 | © TNS 54

 ASK ONLY IF not Q1 | country=1

Q89 | height-non-UK | Height non-UK in cm Numeric

Min 100 | Max 220
Now, here is the last but one question.
How tall are you without shoes?

Please enter your height in centimetres.

Scripter notes: Please ADD "Prefer not to say" option code 9 and "Don't know" option code 8

and "cm" after the numeric field and example

____cm
e.g. 173cm

IF respondent gives an answer that is beyond limits. Please show an error message:
"Please type a digit in the range between 100 and 220 in the numeric box for cm."

Q90 | height-all-dum | Height all in cm Dummy Numeric

Min 100 | Max 200 | Dummy

Researcher notes: Translator: no translation required

Scripter notes: Consolidated answers from Q88 and Q89, who didn't answer DK or Prefer not to say.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 55

 ASK ONLY IF Q1 | country=1

Q91 | weight-uk | Weight UK in stones and pounds Numeric

Max 35
And finally, how much do you weigh without clothes?

Please enter your weight in stones and pounds. If you weigh for example exactly 8 stones, please enter
08 at the filed for stones and 00 at the field for pounds.

Researcher notes: Translator: no translation required

Scripter notes: Please ADD "Prefer not to say" option code 9 and "Don't know" option code 8

and two numeric fields with an example

____ stones ____pounds

e.g. 10 stones 05 pounds

for stones: Min: 06 Max: 35

for pounds: Min: 00 Max 13

IF respondent gives an answer that is beyond limits. Please show an error message:
"Please type a digit in the range between 06 and 35 in the numeric box for stones and a digit in the
range between 00 and 13 in the numeric box for pounds."

ASK ONLY IF Q1 | country=1

Q92 | weight-dum-uk | Weight UK in kg Numeric

Min 40 | Max 220 | Dummy

[DP: conversion from stones to pounds/ to kg]

Researcher notes: Translator: no translation required

Scripter notes: Filter only if Q1 code 1 AND Q91 NOT code 8 or code 9

Please convert stones to pounds with this formula

(stones * 14) + pounds

and then to kg with this formula

pounds / 2.2

===> ((stones * 14) + pounds) / 2.2

EC Food Information - MCP Survey 2013 | version 2 | © TNS 56

 ASK ONLY IF not Q1 | country=1

Q93 | weight-non-UK | Weight non-UK in kg Numeric

Min 40 | Max 220
And finally, how much do you weigh without clothes?

Please enter your weight in kilograms.

Scripter notes: Please ADD "Prefer not to say" option code 9 and "Don't know" option code 8

add "kg" after the numeric field and the example

____kg
e.g. 78 kg

IF respondent gives an answer that is beyond limits. Please show an error message:
"Please type a digit in the range between 40 and 220 in the numeric box for kg."

Q94 | weight-all-dum | Weight all in kg Dummy Numeric

Min 40 | Max 220 | Dummy

Researcher notes: Translator: no translation required

Scripter notes: Consolidated answers from Q92 and Q93, who didn't answer DK or Prefer not to say

B20 | DEMOGRAPHICS End block

Q95 | qual1 | Quality-Check: Straightliners Single coded

Dummy
Code "Yes", if in all of the following matrix questions the same code for all statements within the question
is used Q9 engage, Q11 goals, Q78 waste-att1

1 No Straightliner => stay in final sample

2 Yes Straightliner => Redirect to Invalidates due to quality parameters

Researcher notes: Translator: no translation required

Q96 | qual2 | Quality-Check: Speeder Single coded

Dummy
Code "Yes", if total answering time below 8 minutes

1 No Speeder => stay in final sample

2 Yes Speeder => Redirect to Invalidates due to quality parameters

Researcher notes: Translator: no translation required

EC Food Information - MCP Survey 2013 | version 2 | © TNS 57

T19 | outro0 | Outro Text

Thank you
... for your completed participation in this survey.

You've earned it!

The bonus points for this project will be paid within the 14 days after the study has closed.

As mentioned earlier, you have a chance to win an extra bonus. With only a little bit of luck you will be
amoung the winners who will soon receive a note with the extra amount.

EC Food Information - MCP Survey 2013 | version 2 | © TNS 58

Impact of Food Information on Consumers’ Decision Making

APPENDIX C – STATISTICAL MODELS

213
APPENDIX C - STATISTICAL MODELS

Model 1 – Online experiments on TFA

BINOMINAL LOGISTIC REGRESSION - TFA CHOICE 1 - WITHOUT POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
TFA choice 1 = healthy product A B error Sig. Exp(B) low er bound upper bound
Intercept -1,971 ,299 ,000 ***
Healthy image TFA -,236 ,168 ,159 ,790 ,569 1,097
Unhealthy im age TFA ,283 ,097 ,003 1,327 1,098 1,604 **
Healthy image PHO ,180 ,158 ,255 1,197 ,878 1,633
Unhealthy image PHO -,081 ,148 ,585 ,922 ,691 1,232
Aw areness (rc.:
Healthy image FHO -,057 ,164 ,726 ,944 ,685 1,301
no aw aress of
Unhealthy image FHO -,070 ,144 ,629 ,933 ,704 1,236
fat types/ no
Healthy image saturates ,095 ,120 ,428 1,100 ,869 1,393
zoom ing)
Unhealthy image saturates ,176 ,097 ,070 1,192 ,986 1,442
Zoom ingredients task 1 ,268 ,102 ,009 1,307 1,070 1,596 **
Zoom country of origin task -1,138 ,095 ,000 ,320 ,266 ,386 ***
Zoom nutrition facts task 1 1,225 ,086 ,000 3,404 2,875 4,030 ***
Diet / Weight loss ,084 ,111 ,451 1,087 ,875 1,351
Crisps buyer ,053 ,146 ,716 1,054 ,792 1,404
Habits Look at saturates of crisps ,182 ,102 ,076 1,199 ,981 1,465
Look at calories of crisps ,138 ,092 ,133 1,148 ,959 1,373
Look at ingredients of ,208 ,082 ,011 1,231 1,048 1,445 *
i
Healthy planner -,143 ,118 ,223 ,866 ,688 1,091
Rushed quality shopper -,279 ,113 ,014 ,757 ,606 ,945 *
Motives and
Self-determined shopper -,209 ,131 ,111 ,811 ,627 1,049
attitudes (rc.:
Bargain hunter -,326 ,133 ,014 ,722 ,557 ,936 *
expert shopper)
Frustrated shopper -,282 ,140 ,043 ,754 ,573 ,992 *
Look at food info label ,093 ,055 ,095 1,097 ,984 1,223
Age (rc.: 35-54 Age (18 - 34 years) ,222 ,100 ,026 1,249 1,028 1,518
years) Age (55 and more years) ,045 ,097 ,641 1,046 ,865 1,266
Gender (rc.:
Man -,086 ,080 ,281 ,917 ,784 1,073
w om an)
Incom e (rc.: Income (0 - 9 999) ,163 ,093 ,079 1,177 ,982 1,411
10 000 - 29 999) Income (30 000 and more) -,126 ,097 ,192 ,881 ,729 1,065
Education (rc.: Still study -,247 ,160 ,122 ,781 ,570 1,069
finished Education terminal age max. 17 -,121 ,122 ,322 ,886 ,697 1,126
education w ith
Education term inal age 22+ -,185 ,088 ,035 ,831 ,699 ,987 *
17-22 years)
Living status Hh 2+ persons w ith children ,262 ,118 ,027 1,299 1,031 1,637 *
(rc.: singles) Hh 2+ persons w ithout ,299 ,111 ,007 1,348 1,084 1,675 **
BMI (rc.: BMI
Overw eight (BMI 25+) -,011 ,080 ,894 ,989 ,846 1,157
under 25)
United Kingdom -,289 ,162 ,074 ,749 ,545 1,029
France -,110 ,155 ,478 ,896 ,661 1,214
Germany -,107 ,158 ,495 ,898 ,660 1,223
Countries (rc.:
Italy ,058 ,148 ,697 1,059 ,793 1,415
Poland)
Spain -,217 ,148 ,143 ,805 ,602 1,076
Finland ,065 ,164 ,691 1,067 ,774 1,471
Romania ,020 ,152 ,895 1,020 ,758 1,373

N = 3.945 - Nagelkerke's R² = 0,144

Model 2 – Online experiments on TFA

BINOMINAL LOGISTIC REGRESSION - TFA CHOICE 1 - WITH POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
TFA choice 1 = healthy product A B error Sig. Exp(B) low er bound upper bound
Intercept -4,519 ,383 ,000 ***
Policy: Education on TFA -,175 ,174 ,315 ,839 ,597 1,181
Policy options
Policy: Label info TFA 2,558 ,146 ,000 12,911 9,707 17,172 ***
(rc.: control
Policy: Com bination of ,840 ,201 ,000 2,317 1,561 3,438
group)
education and inform ation ***
Healthy image TFA -,262 ,197 ,183 ,769 ,523 1,132
Unhealthy im age TFA ,456 ,117 ,000 1,577 1,254 1,983 ***
Healthy image PHO ,279 ,186 ,133 1,322 ,918 1,904
Unhealthy image PHO -,208 ,177 ,240 ,812 ,573 1,150
Aw areness (rc.:
Healthy image FHO -,077 ,191 ,687 ,926 ,637 1,347
no aw aress of
Unhealthy image FHO ,026 ,173 ,882 1,026 ,731 1,440
fat types/ no
Healthy image saturates ,221 ,143 ,124 1,247 ,941 1,651
zoom ing)
Unhealthy im age saturates ,233 ,114 ,041 1,263 1,009 1,580 *
Zoom ingredients task 1 ,844 ,126 ,000 2,325 1,815 2,979 ***
Zoom country of origin task -,661 ,115 ,000 ,516 ,412 ,646 ***
Zoom nutrition facts task 1 1,389 ,101 ,000 4,011 3,290 4,890 ***
Diet / Weight loss ,255 ,134 ,058 1,290 ,991 1,679
Crisps buyer ,049 ,173 ,777 1,050 ,748 1,474
Habits Look at saturates of crisps ,207 ,123 ,092 1,230 ,967 1,564
Look at calories of crisps ,251 ,110 ,023 1,285 1,035 1,596 *
Look at ingredients of ,221 ,099 ,025 1,247 1,028 1,513 *
Healthy planner ,018 ,142 ,900 1,018 ,771 1,345
Rushed quality shopper -,238 ,135 ,079 ,788 ,605 1,028
Motives and
Self-determined shopper -,129 ,157 ,412 ,879 ,645 1,196
attitudes (rc.:
Bargain hunter -,232 ,158 ,142 ,793 ,582 1,081
expert shopper)
Frustrated shopper -,347 ,166 ,036 ,707 ,511 ,978 *
Look at food info label ,152 ,065 ,020 1,164 1,024 1,322 *
Age (rc.: 35-54 Age (18 - 34 years) ,283 ,119 ,017 1,327 1,051 1,675 *
years) Age (55 and more years) -,091 ,115 ,427 ,913 ,729 1,143
Gender (rc.:
Man -,039 ,095 ,683 ,962 ,798 1,159
w om an)
Incom e (rc.: Incom e (0 - 9 999) ,246 ,110 ,026 1,279 1,030 1,587 *
10 000 - 29 999) Income (30 000 and more) -,034 ,116 ,771 ,967 ,771 1,213
Education (rc.: Still study -,373 ,190 ,050 ,689 ,475 1,000
finished
Education terminal age max. 17 -,098 ,145 ,497 ,906 ,682 1,204
education w ith
Education terminal age 22+ -,168 ,105 ,111 ,845 ,688 1,039
17-22 years)
Living status Hh 2+ persons w ith children ,303 ,140 ,030 1,354 1,029 1,782 *
(rc.: singles) Hh 2+ persons w ithout ,319 ,131 ,015 1,375 1,064 1,778 *
BMI (rc.: BMI
Overw eight (BMI 25+) -,031 ,095 ,743 ,969 ,805 1,168
under 25)
United Kingdom -,220 ,192 ,252 ,802 ,550 1,169
France -,131 ,184 ,475 ,877 ,612 1,257
Germany -,149 ,189 ,429 ,862 ,595 1,247
Countries (rc.:
Italy ,087 ,178 ,625 1,091 ,770 1,545
Poland)
Spain -,232 ,176 ,187 ,793 ,561 1,120
Finland -,023 ,196 ,908 ,978 ,666 1,435
Romania -,051 ,180 ,776 ,950 ,667 1,353

N = 3.945 - Nagelkerke's R² = 0,463

Model 3 – Online experiments on TFA

BINOMINAL LOGISTIC REGRESSION - TFA CHOICE 2 - WITHOUT POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
TFA choice 2 = healthy product C B error Sig. Exp(B) low er bound upper bound
Intercept - 1,7 9 2 ,280 ,0 0 0 ***
Healthy image TFA ,010 ,156 ,950 1,010 ,744 1,370
Unhealthy image TFA -,020 ,093 ,827 ,980 ,816 1,176
Healthy image PHO -,068 ,152 ,652 ,934 ,693 1,258
Unhealthy image PHO -,032 ,141 ,821 ,968 ,734 1,278
Aw areness (rc.:
Healthy image FHO ,016 ,156 ,917 1,016 ,749 1,380
no aw aress of
Unhealthy image FHO ,081 ,138 ,558 1,084 ,828 1,420
fat types/ no
Healthy image saturates -,183 ,115 ,111 ,833 ,665 1,043
zoom ing)
Unhealthy im age saturates ,710 ,091 ,000 2,035 1,703 2,431 ***
Zoom ingredients task 2 ,328 ,103 ,002 1,389 1,134 1,701 **
Zoom country of origin task -,100 ,088 ,260 ,905 ,761 1,077
Zoom nutrition facts task 2 1,237 ,081 ,000 3,445 2,939 4,038 ***
Diet / Weight loss ,128 ,107 ,234 1,136 ,921 1,401
Crisps buyer ,199 ,138 ,147 1,221 ,932 1,599
Habits Look at saturates of crisps ,198 ,100 ,049 1,219 1,001 1,484 *
Look at calories of crisps ,075 ,089 ,395 1,078 ,906 1,283
Look at ingredients of crisps ,113 ,078 ,149 1,119 ,961 1,304
Healthy planner -,041 ,114 ,720 ,960 ,768 1,200
Rushed quality shopper -,055 ,108 ,609 ,946 ,766 1,170
Motives and
Self-determined shopper -,237 ,125 ,059 ,789 ,617 1,009
attitudes (rc.:
Bargain hunter -,042 ,125 ,740 ,959 ,750 1,226
expert shopper)
Frustrated shopper -,227 ,132 ,086 ,797 ,616 1,032
Look at food info label ,106 ,052 ,041 1,111 1,004 1,230 *
Age (rc.: 35-54 Age (18 - 34 years) ,237 ,095 ,012 1,268 1,053 1,526 *
years) Age (55 and more years) -,135 ,091 ,139 ,874 ,730 1,045
Gender (rc.:
Man -,050 ,075 ,508 ,951 ,821 1,103
w om an)
Incom e (rc.: Income (0 - 9 999) -,022 ,088 ,806 ,979 ,823 1,164
10 000 - 29 999) Income (30 000 and more) ,048 ,091 ,600 1,049 ,878 1,253
Education (rc.: Still study ,072 ,153 ,637 1,075 ,797 1,450
finished
Education terminal age max. 17 -,083 ,115 ,473 ,921 ,734 1,154
education w ith
Education terminal age 22+ -,059 ,083 ,476 ,942 ,801 1,109
17-22 years)
Living status Hh 2+ persons w ith children -,067 ,108 ,535 ,935 ,756 1,156
(rc.: singles) Hh 2+ persons w ithout children -,105 ,102 ,303 ,900 ,737 1,099
BMI (rc.: BMI
Overw eight (BMI 25+) -,024 ,075 ,748 ,976 ,842 1,132
under 25)
United Kingdom ,199 ,154 ,196 1,220 ,903 1,650
France ,017 ,146 ,910 1,017 ,763 1,355
Germany ,146 ,149 ,325 1,158 ,865 1,549
Countries (rc.:
Italy -,147 ,144 ,307 ,864 ,652 1,144
Poland)
Spain ,109 ,141 ,438 1,116 ,846 1,471
Finland -,361 ,159 ,023 ,697 ,511 ,951 *
Romania -,021 ,145 ,884 ,979 ,737 1,301

N = 3.945 - Nagelkerke's R² = 0,184

Model 4 – Online experiments on TFA

BINOMINAL LOGISTIC REGRESSION - TFA CHOICE 2 - WITH POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
TFA choice 2 = healthy product C B error Sig. Exp(B) low er bound upper bound
Intercept -1,622 ,289 ,000 ***
Policy options Policy: Education on TFA ,161 ,098 ,101 1,175 ,969 1,425
(rc.: control Policy: Label info TFA -,295 ,100 ,003 ,745 ,613 ,906 **
group) Policy: Combination of -,263 ,139 ,059 ,769 ,585 1,010
Healthy image TFA -,002 ,157 ,991 ,998 ,734 1,357
Unhealthy image TFA -,031 ,094 ,742 ,970 ,807 1,165
Healthy image PHO -,069 ,153 ,652 ,933 ,692 1,259
Unhealthy image PHO -,017 ,143 ,903 ,983 ,743 1,299
Aw areness (rc.:
Healthy image FHO ,008 ,157 ,959 1,008 ,741 1,372
no aw aress of
Unhealthy image FHO ,065 ,139 ,640 1,067 ,813 1,400
fat types/ no
Healthy image saturates -,198 ,115 ,086 ,820 ,654 1,028
zoom ing)
Unhealthy im age saturates ,715 ,091 ,000 2,045 1,709 2,446 ***
Zoom ingredients task 2 ,293 ,105 ,005 1,340 1,092 1,645 **
Zoom country of origin task -,190 ,091 ,037 ,827 ,692 ,988 *
Zoom nutrition facts task 2 1,263 ,082 ,000 3,536 3,012 4,150 ***
Diet / Weight loss ,115 ,108 ,283 1,122 ,909 1,386
Crisps buyer ,202 ,138 ,143 1,224 ,934 1,604
Habits Look at saturates of crisps ,213 ,101 ,036 1,237 1,015 1,509 *
Look at calories of crisps ,073 ,089 ,416 1,075 ,903 1,281
Look at ingredients of crisps ,116 ,079 ,140 1,123 ,963 1,310
Healthy planner -,058 ,114 ,611 ,943 ,754 1,181
Rushed quality shopper -,068 ,109 ,534 ,935 ,755 1,157
Motives and
Self-determ ined shopper -,258 ,126 ,041 ,772 ,603 ,989 *
attitudes (rc.:
Bargain hunter -,073 ,126 ,565 ,930 ,726 1,191
expert shopper)
Frustrated shopper -,236 ,132 ,075 ,790 ,610 1,024
Look at food info label ,103 ,052 ,048 1,108 1,001 1,227 *
Age (rc.: 35-54 Age (18 - 34 years) ,246 ,095 ,010 1,280 1,062 1,542 *
years) Age (55 and more years) -,114 ,092 ,214 ,892 ,745 1,068
Gender (rc.:
Man -,056 ,076 ,458 ,945 ,815 1,097
w om an)
Incom e (rc.: Income (0 - 9 999) -,029 ,089 ,741 ,971 ,816 1,156
10 000 - 29 999) Income (30 000 and more) ,030 ,091 ,744 1,030 ,861 1,232
Education (rc.: Still study ,078 ,154 ,614 1,081 ,800 1,460
finished
Education terminal age max. 17 -,091 ,116 ,432 ,913 ,727 1,146
education w ith
Education terminal age 22+ -,075 ,084 ,373 ,928 ,788 1,094
17-22 years)
Living status Hh 2+ persons w ith children -,064 ,109 ,557 ,938 ,758 1,161
(rc.: singles) Hh 2+ persons w ithout children -,095 ,103 ,355 ,910 ,744 1,112
BMI (rc.: BMI
Overw eight (BMI 25+) -,023 ,076 ,759 ,977 ,842 1,134
under 25)
United Kingdom ,184 ,155 ,235 1,202 ,887 1,629
France ,009 ,147 ,952 1,009 ,756 1,347
Germany ,141 ,149 ,345 1,151 ,859 1,543
Countries (rc.:
Italy -,159 ,144 ,271 ,853 ,643 1,132
Poland)
Spain ,099 ,142 ,485 1,104 ,836 1,458
Finland -,359 ,159 ,024 ,699 ,511 ,955 *
Romania -,023 ,146 ,874 ,977 ,734 1,300

N = 3.945 - Nagelkerke's R² = 0,196

Model 5 – Online experiments on TFA

BINOMINAL LOGISTIC REGRESSION - TFA CHOICE 1+2 - WITHOUT POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
TFA choice 1+2=healthy product A/C B error Sig. Exp(B) low er bound upper bound
Intercept -1,529 ,203 ,000 ***
Choice
architecture (rc.: TFA choice task 1 -,651 ,050 ,000 ,522 ,473 ,575 ***
task 2)
Healthy image TFA -,106 ,112 ,341 ,899 ,722 1,119
Unhealthy image TFA ,120 ,066 ,069 1,127 ,991 1,283
Healthy image PHO ,054 ,109 ,620 1,055 ,853 1,306
Unhealthy image PHO -,057 ,101 ,570 ,944 ,776 1,150
Aw areness (rc.:
Healthy image FHO -,023 ,112 ,840 ,978 ,785 1,217
no aw aress of
Unhealthy image FHO ,003 ,098 ,974 1,003 ,828 1,215
fat types/ no
Healthy image saturates -,050 ,083 ,549 ,951 ,809 1,119
zoom ing)
Unhealthy im age saturates ,452 ,065 ,000 1,571 1,382 1,787 ***
Zoom ingredients ,308 ,071 ,000 1,361 1,184 1,564 ***
Zoom country of origin -,595 ,063 ,000 ,552 ,488 ,624 ***
Zoom nutrition facts 1,206 ,059 ,000 3,341 2,977 3,749 ***
Diet / Weight loss ,105 ,076 ,167 1,111 ,957 1,289
Crisps buyer ,136 ,099 ,169 1,146 ,943 1,392
Habits Look at saturates of crisps ,182 ,071 ,010 1,199 1,044 1,377 *
Look at calories of crisps ,105 ,063 ,095 1,111 ,982 1,256
Look at ingredients of ,153 ,056 ,006 1,166 1,045 1,300 **
Healthy planner -,084 ,081 ,301 ,919 ,784 1,078
Rushed quality shopper -,159 ,077 ,040 ,853 ,733 ,993 *
Motives and
Self-determ ined shopper -,228 ,090 ,011 ,796 ,668 ,949 *
attitudes (rc.:
Bargain hunter -,176 ,089 ,049 ,839 ,704 ,999 *
expert shopper)
Frustrated shopper -,246 ,095 ,009 ,782 ,649 ,941 **
Look at food info label ,096 ,037 ,010 1,101 1,023 1,184 *
Age (rc.: 35-54 Age (18 - 34 years) ,220 ,067 ,001 1,247 1,092 1,423 **
years) Age (55 and more years) -,052 ,066 ,430 ,949 ,834 1,080
Gender (rc.:
Man -,065 ,054 ,231 ,937 ,843 1,042
w om an)
Incom e (rc.: Income (0 - 9 999) ,056 ,063 ,378 1,057 ,934 1,196
10 000 - 29 999) Income (30 000 and more) -,033 ,065 ,609 ,967 ,851 1,099
Education (rc.: Still study -,059 ,108 ,586 ,943 ,762 1,166
finished
Education terminal age max. 17 -,088 ,083 ,287 ,916 ,779 1,077
education w ith
Education terminal age 22+ -,112 ,060 ,061 ,894 ,795 1,005
17-22 years)
Living status Hh 2+ persons w ith children ,073 ,078 ,349 1,076 ,923 1,254
(rc.: singles) Hh 2+ persons w ithout children ,074 ,074 ,311 1,077 ,933 1,244
BMI (rc.: BMI
Overw eight (BMI 25+) -,019 ,054 ,720 ,981 ,882 1,091
under 25)
United Kingdom -,040 ,110 ,717 ,961 ,775 1,191
France -,044 ,105 ,677 ,957 ,779 1,176
Germany ,018 ,107 ,869 1,018 ,825 1,255
Countries (rc.:
Italy -,047 ,102 ,647 ,954 ,781 1,166
Poland)
Spain -,044 ,101 ,660 ,957 ,785 1,165
Finland -,170 ,113 ,132 ,843 ,676 1,053
Romania -,006 ,104 ,954 ,994 ,811 1,219

N = 3.945 - Nagelkerke's R² = 0,165

Model 6 – Online experiments on TFA

BINOMINAL LOGISTIC REGRESSION - TFA CHOICE 1+2 - WITH POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
TFA choice 1+2=healthy product A/C B error Sig. Exp(B) low er bound upper bound
Intercept -2,155 ,214 ,000 ***
Choice
architecture (rc.: TFA choice task 1 -,695 ,051 ,000 ,499 ,452 ,552
task 2) ***
Policy: Education on TFA ,036 ,075 ,627 1,037 ,896 1,200
Policy options
Policy: Label info TFA ,785 ,074 ,000 2,191 1,896 2,532 ***
(rc.: control
Policy: Combination of ,229 ,102 ,024 1,258 1,030 1,535
group)
education and information *
Healthy image TFA -,086 ,114 ,450 ,917 ,734 1,147
Unhealthy image TFA ,135 ,067 ,046 1,144 1,003 1,306 *
Healthy image PHO ,068 ,111 ,538 1,070 ,862 1,330
Unhealthy image PHO -,083 ,103 ,419 ,921 ,753 1,125
Aw areness (rc.:
Healthy image FHO -,023 ,114 ,837 ,977 ,782 1,221
no aw aress of
Unhealthy image FHO ,038 ,100 ,702 1,039 ,854 1,263
fat types/ no
Healthy image saturates -,030 ,085 ,721 ,970 ,822 1,146
zoom ing)
Unhealthy im age saturates ,473 ,067 ,000 1,605 1,408 1,830 ***
Zoom ingredients ,424 ,073 ,000 1,529 1,326 1,763 ***
Zoom country of origin -,382 ,065 ,000 ,683 ,601 ,775 ***
Zoom nutrition facts 1,199 ,060 ,000 3,316 2,948 3,730 ***
Diet / Weight loss ,141 ,078 ,070 1,151 ,988 1,341
Crisps buyer ,138 ,102 ,175 1,148 ,940 1,401
Habits Look at saturates of crisps ,168 ,072 ,020 1,183 1,027 1,362 *
Look at calories of crisps ,122 ,064 ,057 1,130 ,996 1,281
Look at ingredients of ,146 ,057 ,011 1,157 1,035 1,294 *
Healthy planner -,038 ,083 ,648 ,963 ,818 1,133
Rushed quality shopper -,130 ,079 ,099 ,878 ,752 1,025
Motives and
Self-determ ined shopper -,195 ,092 ,033 ,822 ,687 ,984 *
attitudes (rc.:
Bargain hunter -,127 ,091 ,166 ,881 ,736 1,054
expert shopper)
Frustrated shopper -,248 ,097 ,011 ,781 ,645 ,944 *
Look at food info label ,108 ,038 ,005 1,114 1,034 1,200 **
Age (rc.: 35-54 Age (18 - 34 years) ,219 ,069 ,001 1,245 1,087 1,424 **
years) Age (55 and more years) -,099 ,067 ,140 ,906 ,794 1,033
Gender (rc.:
Man -,051 ,055 ,353 ,950 ,852 1,059
w om an)
Incom e (rc.: Income (0 - 9 999) ,066 ,065 ,306 1,068 ,941 1,212
10 000 - 29 999) Income (30 000 and more) ,007 ,067 ,912 1,007 ,884 1,148
Education (rc.: Still study -,070 ,111 ,526 ,932 ,750 1,158
finished
Education terminal age max. 17 -,072 ,084 ,392 ,930 ,788 1,098
education w ith
Education terminal age 22+ -,092 ,061 ,133 ,912 ,810 1,028
17-22 years)
Living status Hh 2+ persons w ith children ,064 ,080 ,421 1,066 ,912 1,247
(rc.: singles) Hh 2+ persons w ithout children ,059 ,075 ,436 1,060 ,915 1,229
BMI (rc.: BMI
Overw eight (BMI 25+) -,027 ,055 ,625 ,973 ,873 1,085
under 25)
United Kingdom ,004 ,112 ,973 1,004 ,806 1,250
France -,034 ,108 ,752 ,967 ,783 1,193
Germany ,026 ,109 ,813 1,026 ,828 1,271
Countries (rc.:
Italy -,039 ,105 ,711 ,962 ,784 1,181
Poland)
Spain -,021 ,103 ,837 ,979 ,801 1,197
Finland -,201 ,116 ,082 ,818 ,652 1,026
Romania -,014 ,106 ,893 ,986 ,801 1,214

N = 3.945 - Nagelkerke's R² = 0,165

Model 7 – Online experiments on TFA

BINOMINAL LOGISTIC REGRESSION - ZOOMING BEHAVIOUR IN TFA CHOICES 1+2
Zooming any relevant labels in both Std. 95% conf. interval for Exp(B)
TFA choices B error Sig. Exp(B) low er bound upper bound
Intercept ,489 ,201 ,015 *
Choice
architecture (rc.: TFA choice task 1 -,132 ,051 ,009 ,876 ,793 ,968
task 2) **
Policy (rc.: no
Policy: Education on TFA ,190 ,051 ,000 1,209 1,094 1,337 ***
education)
Healthy im age TFA -,396 ,110 ,000 ,673 ,543 ,835 ***
Unhealthy image TFA ,031 ,071 ,663 1,032 ,897 1,186
Healthy image PHO ,044 ,110 ,690 1,045 ,842 1,298
Aw areness (rc.: Unhealthy image PHO ,013 ,108 ,906 1,013 ,819 1,253
no aw aress) Healthy image FHO -,163 ,112 ,145 ,850 ,682 1,058
Unhealthy image FHO ,125 ,105 ,235 1,133 ,922 1,393
Healthy im age saturates ,273 ,082 ,001 1,315 1,120 1,543 **
Unhealthy im age saturates ,498 ,066 ,000 1,646 1,446 1,875 ***
Diet / Weight loss ,248 ,083 ,003 1,281 1,088 1,509 **
Crisps buyer ,150 ,095 ,116 1,161 ,964 1,399
Habits Look at saturates of crisps ,215 ,080 ,007 1,240 1,061 1,449 **
Look at calories of crisps -,030 ,068 ,664 ,971 ,850 1,109
Look at ingredients of ,375 ,060 ,000 1,456 1,295 1,637 ***
Healthy planner -,512 ,088 ,000 ,599 ,504 ,712 ***
Rushed quality shopper -,355 ,085 ,000 ,701 ,594 ,828 ***
Motives and
Self-determ ined shopper -,422 ,095 ,000 ,656 ,545 ,790 ***
attitudes (rc.:
Bargain hunter -,323 ,096 ,001 ,724 ,599 ,875 **
expert shopper)
Frustrated shopper -,487 ,098 ,000 ,615 ,507 ,745 ***
Look at food info label ,103 ,037 ,006 1,109 1,031 1,193 **
Age (rc.: 35-54 Age (18 - 34 years) ,099 ,070 ,160 1,104 ,962 1,266
years) Age (55 and more years) -,058 ,067 ,380 ,943 ,828 1,075
Gender (rc.:
Man -,324 ,055 ,000 ,723 ,650 ,806 ***
w om an)
Income (0 - 9 999) -,021 ,066 ,749 ,979 ,861 1,114
Incom e (rc.: 10
000 - 29 999) Income (30 000 and more) -,022 ,067 ,745 ,979 ,858 1,116
Education (rc.: Still study ,205 ,120 ,088 1,227 ,970 1,553
finished
Education terminal age max. 17 -,053 ,082 ,522 ,949 ,808 1,114
education w ith
Education terminal age 22+ ,029 ,062 ,642 1,029 ,912 1,161
17-22 years)
Living status Hh 2+ persons w ith children -,097 ,079 ,221 ,908 ,778 1,060
(rc.: singles) Hh 2+ persons w ithout children -,082 ,074 ,268 ,921 ,796 1,065
BMI (rc.: BMI
Overw eight (BMI 25+) -,019 ,055 ,738 ,982 ,880 1,094
under 25)
United Kingdom -,123 ,112 ,272 ,884 ,709 1,102
France -,047 ,107 ,659 ,954 ,774 1,176
Germany ,037 ,109 ,735 1,038 ,838 1,285
Countries (rc.:
Italy -,058 ,107 ,585 ,943 ,765 1,163
Poland)
Spain -,145 ,104 ,165 ,865 ,705 1,061
Finland ,032 ,115 ,781 1,032 ,825 1,293
Rom ania ,331 ,112 ,003 1,392 1,119 1,733 **

N = 3.945 - Nagelkerke's R² = 0,077

Model 8 – Online experiments on PAW

BINOMINAL LOGISTIC REGRESSION - PAW CHOICE 1 WITHOUT POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
Allergy choice 1 = PAW avoiding B error Sig. Exp(B) low er bound upper bound
Intercept -2,588 1,269 ,041
Zoom ingredients task 1 4,236 ,349 ,000 69,097 34,844 137,022 ***
Aw areness Zoom country of origin task -,869 ,334 ,009 ,419 ,218 ,807 **
Zoom nutrition facts task 1 -,877 ,322 ,006 ,416 ,221 ,782 **
Muesli bought last 6 months ,032 ,452 ,943 1,033 ,426 2,507
Look at allergen info of muesli ,408 ,399 ,306 1,504 ,688 3,285
Look at country of origin of -,162 ,423 ,701 ,850 ,371 1,946
Look at ingredients of muesli ,017 ,428 ,967 1,018 ,440 2,353
Habits
Make a list before shopping -,032 ,138 ,816 ,969 ,740 1,268
Search for special deals -,017 ,150 ,908 ,983 ,732 1,319
Buy alw ays same products -,269 ,202 ,183 ,764 ,515 1,135
Frequent allergy info reading ,401 ,356 ,259 1,493 ,744 2,998
Allergy/intolerance mild ,097 ,325 ,765 1,102 ,583 2,085
Allergy/intolerance severe ,556 ,405 ,170 1,744 ,789 3,856
Strong relevance of label ,160 ,341 ,639 1,173 ,601 2,289
Choose healthier alternative ,157 ,363 ,665 1,170 ,575 2,382
Don't compromise quality for ,151 ,324 ,641 1,163 ,616 2,197
Goals & Brand loyal -,182 ,323 ,573 ,834 ,442 1,570
attitudes (rc.: Low self-control ,149 ,313 ,635 1,160 ,628 2,143
m oderate/ High risk aversion -,482 ,327 ,140 ,617 ,325 1,172
m edium ) May contain traces - not safe ,454 ,633 ,474 1,574 ,455 5,444
May contain - not safe -,130 ,622 ,835 ,878 ,259 2,972
Cannot guarantee - not safe 1,145 ,415 ,006 3,144 1,393 7,094 **
Made in a facility - not safe ,156 ,433 ,718 1,169 ,500 2,733
Contains nuts - not safe -,326 ,567 ,565 ,722 ,238 2,191
No inform ation given - not -,771 ,318 ,015 ,463 ,248 ,862 *
Age (rc.: 35-54 Age (18 - 34 years) ,600 ,389 ,123 1,822 ,850 3,906
years) Age (55 and more years) ,687 ,391 ,079 1,987 ,923 4,278
Gender (rc.:
Man ,102 ,314 ,744 1,108 ,599 2,049
w om an)
Incom e (rc.: 10 Incom e (0 - 9 999) ,719 ,360 ,045 2,053 1,015 4,155 *
000 - 29 999) Income (30 000 and more) -,599 ,387 ,122 ,549 ,257 1,173
Education (rc.: Still study -,832 ,558 ,136 ,435 ,146 1,299
finished
Education terminal age max. 17 -,227 ,531 ,668 ,797 ,281 2,254
education w ith
Education terminal age 22+ ,578 ,339 ,088 1,782 ,917 3,462
17-22 years)
Living status Hh 2+ persons w ith children ,820 ,485 ,091 2,271 ,878 5,874
(rc.: singles) Hh 2+ persons w ithout children ,491 ,442 ,267 1,634 ,687 3,888
United Kingdom 1,031 ,682 ,130 2,805 ,737 10,678
France -1,324 ,790 ,094 ,266 ,057 1,253
Italy -1,144 ,617 ,064 ,319 ,095 1,068
Countries (rc.:
Spain -2,579 ,784 ,001 ,076 ,016 ,353 **
Germ any)
Finland ,413 ,538 ,442 1,512 ,527 4,341
Poland -,707 ,595 ,235 ,493 ,153 1,583
Romania -1,156 ,631 ,067 ,315 ,091 1,085

N = 586 - Nagelkerke R² = 0,698

Model 9 – Online experiments on PAW

BINOMINAL LOGISTIC REGRESSION - PAW CHOICE 1 WITH POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
Allergy choice 1 = PAW avoiding B error Sig. Exp(B) low er bound upper bound
Intercept -2,317 1,306 ,076
Policy options Split 2: May contain traces -,054 ,435 ,901 ,947 ,404 2,223
(rc.: split 1: m ay Split 3: Cannot guarantee -,356 ,409 ,384 ,700 ,314 1,561
contain) Split 4: Made in a facility -,667 ,411 ,105 ,513 ,229 1,149
Zoom ingredients task 1 4,337 ,362 ,000 76,475 37,592 155,578 ***
Aw areness Zoom country of origin task -,919 ,340 ,007 ,399 ,205 ,777 **
Zoom nutrition facts task 1 -,890 ,323 ,006 ,411 ,218 ,774 **
Muesli bought last 6 months ,011 ,457 ,981 1,011 ,413 2,477
Look at allergen info of muesli ,437 ,400 ,274 1,548 ,707 3,389
Look at country of origin of -,172 ,425 ,685 ,842 ,366 1,935
Look at ingredients of muesli ,026 ,431 ,952 1,026 ,441 2,386
Habits
Make a list before shopping -,026 ,139 ,849 ,974 ,742 1,278
Search for special deals -,026 ,151 ,861 ,974 ,725 1,309
Buy alw ays same products -,275 ,202 ,175 ,760 ,511 1,130
Frequent allergy info reading ,406 ,358 ,257 1,500 ,744 3,028
Allergy/intolerance mild ,162 ,329 ,622 1,176 ,617 2,243
Allergy/intolerance severe ,578 ,406 ,155 1,782 ,804 3,948
Strong relevance of label ,206 ,347 ,552 1,229 ,623 2,424
Choose healthier alternative ,151 ,370 ,683 1,163 ,563 2,400
Don't compromise quality for ,135 ,327 ,678 1,145 ,604 2,172
Goals & Brand loyal -,213 ,328 ,517 ,808 ,425 1,538
attitudes (rc.: Low self-control ,157 ,315 ,617 1,170 ,632 2,169
m oderate/ High risk aversion -,466 ,328 ,155 ,627 ,330 1,193
m edium ) May contain traces - not safe ,522 ,634 ,411 1,685 ,486 5,839
May contain - not safe -,155 ,625 ,804 ,856 ,252 2,912
Cannot guarantee - not safe 1,115 ,417 ,008 3,048 1,346 6,907 **
Made in a facility - not safe ,220 ,433 ,611 1,247 ,534 2,912
Contains nuts - not safe -,330 ,578 ,569 ,719 ,231 2,235
No inform ation given - not -,785 ,322 ,015 ,456 ,243 ,857 *
Age (rc.: 35-54 Age (18 - 34 years) ,599 ,392 ,127 1,821 ,844 3,926
years) Age (55 and more years) ,665 ,392 ,090 1,945 ,901 4,197
Gender (rc.:
Man ,136 ,317 ,668 1,145 ,615 2,133
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,688 ,362 ,058 1,989 ,978 4,044
000 - 29 999) Income (30 000 and more) -,566 ,388 ,144 ,568 ,265 1,214
Education (rc.: Still study -,859 ,567 ,130 ,424 ,140 1,286
finished
Education terminal age max. 17 -,222 ,538 ,681 ,801 ,279 2,300
education w ith
Education terminal age 22+ ,583 ,339 ,086 1,791 ,921 3,483
17-22 years)
Living status Hh 2+ persons w ith children ,833 ,488 ,088 2,300 ,884 5,985
(rc.: singles) Hh 2+ persons w ithout children ,499 ,444 ,262 1,647 ,689 3,936
United Kingdom ,984 ,688 ,152 2,676 ,695 10,301
France -1,343 ,787 ,088 ,261 ,056 1,221
Italy -1,246 ,622 ,045 ,288 ,085 ,974 *
Countries (rc.:
Spain -2,637 ,786 ,001 ,072 ,015 ,334 **
Germ any)
Finland ,353 ,544 ,517 1,423 ,490 4,134
Poland -,713 ,598 ,234 ,490 ,152 1,584
Romania -1,227 ,634 ,053 ,293 ,085 1,017

N = 586 - Nagelkerke R² = 0,702

Model 10 – Online experiments on PAW

BINOMINAL LOGISTIC REGRESSION - PAW CHOICE 2 WITHOUT POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
Allergy choice 2 = PAW avoiding B error Sig. Exp(B) low er bound upper bound
Intercept ,002 1,339 ,999
Zoom ingredients task 2 4,654 ,385 ,000 104,993 49,409 223,110 ***
Aw areness Zoom country of origin task -,590 ,366 ,107 ,554 ,271 1,135
Zoom nutrition facts task 2 -1,499 ,381 ,000 ,223 ,106 ,471 ***
Muesli bought last 6 months -,430 ,478 ,368 ,650 ,255 1,660
Look at allergen info of muesli ,189 ,423 ,655 1,208 ,527 2,766
Look at country of origin of ,023 ,445 ,960 1,023 ,427 2,449
Look at ingredients of muesli ,248 ,448 ,580 1,282 ,532 3,086
Habits
Make a list before shopping -,066 ,141 ,641 ,936 ,710 1,235
Search for special deals -,127 ,159 ,424 ,880 ,644 1,203
Buy alw ays same products -,181 ,215 ,399 ,834 ,548 1,271
Frequent allergy info -,841 ,406 ,038 ,431 ,195 ,956 *
Allergy/intolerance mild -,125 ,342 ,716 ,883 ,452 1,726
Allergy/intolerance severe ,371 ,431 ,390 1,449 ,622 3,375
Strong relevance of label ,747 ,365 ,041 2,111 1,031 4,320 *
Choose healthier alternative -,669 ,391 ,087 ,512 ,238 1,101
Don't compromise quality for ,127 ,351 ,719 1,135 ,570 2,259
Goals & Brand loyal ,111 ,335 ,739 1,118 ,580 2,156
attitudes (rc.: Low self-control -,138 ,332 ,678 ,871 ,455 1,670
m oderate/ High risk aversion -,459 ,357 ,199 ,632 ,314 1,273
m edium ) May contain traces - not safe ,490 ,727 ,500 1,632 ,393 6,781
May contain - not safe 1,125 ,687 ,102 3,080 ,801 11,842
Cannot guarantee - not safe ,449 ,472 ,341 1,567 ,622 3,949
Made in a facility - not safe -,586 ,497 ,239 ,557 ,210 1,475
Contains nuts - not safe -,764 ,630 ,226 ,466 ,135 1,603
No inform ation given - not -,692 ,327 ,035 ,501 ,264 ,951 *
Age (rc.: 35-54 Age (18 - 34 years) ,537 ,420 ,201 1,711 ,752 3,897
years) Age (55 and more years) ,029 ,414 ,944 1,029 ,457 2,318
Gender (rc.: -,357 ,335 ,287 ,700 ,363 1,350
Man
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,056 ,389 ,885 1,058 ,493 2,269
000 - 29 999) Income (30 000 and more) -,026 ,408 ,950 ,975 ,438 2,169
Education (rc.: Still study -,705 ,612 ,250 ,494 ,149 1,642
finished
Education terminal age max. 17 -,361 ,526 ,492 ,697 ,248 1,954
education w ith
Education terminal age 22+ ,053 ,355 ,881 1,054 ,526 2,112
17-22 years)
Living status Hh 2+ persons w ith children 1,019 ,535 ,057 2,771 ,970 7,915
(rc.: singles) Hh 2+ persons w ithout ,980 ,480 ,041 2,663 1,039 6,829 *
United Kingdom ,448 ,679 ,510 1,565 ,413 5,928
France -1,704 ,922 ,065 ,182 ,030 1,108
Italy -,371 ,627 ,555 ,690 ,202 2,361
Countries (rc.:
Spain -2,824 ,854 ,001 ,059 ,011 ,317 **
Germ any)
Finland ,042 ,577 ,942 1,043 ,337 3,230
Poland -1,119 ,644 ,082 ,327 ,092 1,155
Romania -1,204 ,691 ,082 ,300 ,077 1,163

N = 586 - Nagelkerke R² = 0,733

Model 11 – Online experiments on PAW

BINOMINAL LOGISTIC REGRESSION - PAW CHOICE 2 WITH POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
Allergy choice 2 = PAW avoiding B error Sig. Exp(B) low er bound upper bound
Intercept ,133 1,372 ,923
Policy options Split 2: May contain traces -,034 ,440 ,939 ,967 ,408 2,290
(rc.: split 1: m ay Split 3: Cannot guarantee -,398 ,441 ,367 ,672 ,283 1,595
contain) Split 4: Made in a facility -,191 ,430 ,657 ,826 ,356 1,918
Zoom ingredients task 2 4,697 ,390 ,000 109,651 51,014 235,684 ***
Aw areness Zoom country of origin task 2 -,577 ,368 ,117 ,562 ,273 1,155
Zoom nutrition facts task 2 -1,507 ,382 ,000 ,222 ,105 ,469 ***
Muesli bought last 6 months -,445 ,481 ,355 ,641 ,250 1,644
Look at allergen info of muesli ,209 ,424 ,622 1,232 ,537 2,828
Look at country of origin of ,013 ,446 ,976 1,013 ,422 2,431
Look at ingredients of muesli ,236 ,449 ,598 1,266 ,526 3,051
Habits
Make a list before shopping -,065 ,143 ,650 ,937 ,708 1,240
Search for special deals -,139 ,160 ,385 ,870 ,636 1,190
Buy alw ays same products -,185 ,216 ,392 ,831 ,544 1,269
Frequent allergy info -,826 ,407 ,043 ,438 ,197 ,973 *
Allergy/intolerance mild -,078 ,346 ,821 ,925 ,470 1,820
Allergy/intolerance severe ,397 ,432 ,359 1,487 ,637 3,469
Strong relevance of label ,773 ,369 ,036 2,167 1,051 4,466 *
Choose healthier alternative -,715 ,394 ,069 ,489 ,226 1,058
Don't compromise quality for ,168 ,355 ,635 1,183 ,590 2,375
Goals & Brand loyal ,098 ,342 ,774 1,103 ,564 2,157
attitudes (rc.: Low self-control -,118 ,334 ,725 ,889 ,462 1,710
m oderate/ High risk aversion -,436 ,359 ,224 ,646 ,320 1,307
m edium ) May contain traces - not safe ,519 ,728 ,476 1,681 ,403 7,007
May contain - not safe 1,127 ,696 ,106 3,085 ,788 12,076
Cannot guarantee - not safe ,457 ,475 ,335 1,579 ,623 4,003
Made in a facility - not safe -,548 ,497 ,271 ,578 ,218 1,533
Contains nuts - not safe -,826 ,642 ,199 ,438 ,124 1,542
No inform ation given - not -,688 ,329 ,037 ,503 ,264 ,958 *
Age (rc.: 35-54 Age (18 - 34 years) ,536 ,419 ,201 1,709 ,751 3,887
years) Age (55 and more years) ,013 ,416 ,976 1,013 ,448 2,288
Gender (rc.:
Man -,337 ,338 ,319 ,714 ,368 1,385
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,073 ,392 ,853 1,075 ,498 2,321
000 - 29 999) Income (30 000 and more) -,004 ,414 ,993 ,996 ,443 2,241
Education (rc.: Still study -,726 ,621 ,242 ,484 ,143 1,634
finished
Education terminal age max. 17 -,349 ,528 ,509 ,706 ,251 1,985
education w ith
Education terminal age 22+ ,052 ,354 ,882 1,054 ,526 2,110
17-22 years)
Living status Hh 2+ persons w ith children 1,041 ,540 ,054 2,832 ,983 8,162
(rc.: singles) Hh 2+ persons w ithout ,980 ,483 ,043 2,664 1,033 6,872 *
United Kingdom ,398 ,686 ,561 1,489 ,388 5,708
France -1,710 ,925 ,065 ,181 ,029 1,109
Italy -,443 ,638 ,487 ,642 ,184 2,242
Countries (rc.:
Spain -2,889 ,870 ,001 ,056 ,010 ,306 **
Germ any)
Finland ,025 ,582 ,966 1,025 ,328 3,209
Poland -1,102 ,645 ,087 ,332 ,094 1,175
Romania -1,244 ,693 ,073 ,288 ,074 1,121

N = 586 - Nagelkerke R² = 0,734

Model 12 – Online experiments on PAW

BINOMINAL LOGISTIC REGRESSION - PAW CHOICE 3 WITHOUT POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
Allergy choice 3 = PAW avoiding B error Sig. Exp(B) low er bound upper bound
Intercept -,933 1,296 ,472
Zoom ingredients task 3 4,586 ,387 ,000 98,091 45,985 209,238 ***
Aw areness Zoom country of origin task 3 ,082 ,377 ,828 1,086 ,518 2,274
Zoom nutrition facts task 3 -2,221 ,420 ,000 ,109 ,048 ,247 ***
Muesli bought last 6 months ,579 ,459 ,208 1,784 ,725 4,389
Look at allergen info of muesli ,780 ,420 ,063 2,182 ,959 4,969
Look at country of origin of ,126 ,423 ,766 1,134 ,495 2,596
Look at ingredients of muesli -,487 ,430 ,257 ,615 ,265 1,427
Habits
Make a list before shopping ,087 ,145 ,546 1,091 ,822 1,450
Search for special deals ,118 ,160 ,459 1,125 ,823 1,539
Buy alw ays same products -,156 ,219 ,479 ,856 ,557 1,316
Frequent allergy info reading -,058 ,380 ,878 ,943 ,447 1,988
Allergy/intolerance mild ,578 ,346 ,095 1,782 ,905 3,511
Allergy/intolerance severe ,326 ,419 ,436 1,386 ,610 3,149
Strong relevance of label ,286 ,362 ,430 1,331 ,655 2,702
Choose healthier alternative -,084 ,383 ,826 ,919 ,434 1,949
Don't compromise quality for ,158 ,352 ,653 1,172 ,588 2,334
Goals & Brand loyal -,796 ,341 ,020 ,451 ,231 ,880 *
attitudes (rc.: Low self-control -,579 ,324 ,074 ,560 ,297 1,059
m oderate/ High risk aversion -,074 ,354 ,835 ,929 ,464 1,860
m edium ) May contain traces - not safe ,508 ,696 ,465 1,662 ,425 6,505
May contain - not safe -,852 ,695 ,220 ,427 ,109 1,664
Cannot guarantee - not safe ,611 ,406 ,132 1,842 ,832 4,080
Made in a facility - not safe ,483 ,469 ,303 1,621 ,647 4,063
Contains nuts - not safe ,764 ,539 ,157 2,146 ,745 6,178
No inform ation given - not -,861 ,346 ,013 ,423 ,215 ,832 *
Age (rc.: 35-54 Age (18 - 34 years) ,300 ,398 ,450 1,350 ,619 2,943
years) Age (55 and more years) ,133 ,412 ,746 1,143 ,510 2,560
Gender (rc.: -,056 ,331 ,867 ,946 ,495 1,809
Man
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,223 ,387 ,564 1,250 ,585 2,669
000 - 29 999) Income (30 000 and more) -,387 ,391 ,321 ,679 ,316 1,460
Education (rc.: Still study -,094 ,633 ,882 ,910 ,263 3,148
finished
Education terminal age max. 17 -,385 ,521 ,460 ,681 ,245 1,890
education w ith
Education terminal age 22+ ,144 ,341 ,674 1,155 ,591 2,254
17-22 years)
Living status Hh 2+ persons w ith children -,406 ,501 ,417 ,666 ,249 1,778
(rc.: singles) Hh 2+ persons w ithout children -,290 ,463 ,532 ,749 ,302 1,856
United Kingdom ,155 ,675 ,819 1,167 ,311 4,385
France -1,532 ,848 ,071 ,216 ,041 1,138
Italy -1,085 ,629 ,085 ,338 ,098 1,160
Countries (rc.:
Spain -3,049 ,802 ,000 ,047 ,010 ,228 ***
Germ any)
Finland -,360 ,596 ,546 ,698 ,217 2,243
Poland -1,533 ,658 ,020 ,216 ,059 ,784 *
Romania -1,106 ,681 ,104 ,331 ,087 1,256

N = 586 - Nagelkerke R² = 0,734

Model 13 – Online experiments on PAW

BINOMINAL LOGISTIC REGRESSION - PAW CHOICE 3 WITH POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
Allergy choice 3 = PAW avoiding B error Sig. Exp(B) low er bound upper bound
Intercept -,520 1,327 ,695
Policy options Split 2: May contain traces -,393 ,454 ,387 ,675 ,277 1,643
(rc.: split 1: m ay Split 3: Cannot guarantee -,804 ,440 ,068 ,447 ,189 1,060
contain) Split 4: Made in a facility -,783 ,439 ,074 ,457 ,193 1,079
Zoom ingredients task 3 4,712 ,402 ,000 111,229 50,597 244,518 ***
Aw areness Zoom country of origin task 3 ,089 ,384 ,817 1,093 ,515 2,321
Zoom nutrition facts task 3 -2,281 ,428 ,000 ,102 ,044 ,236 ***
Muesli bought last 6 months ,573 ,460 ,213 1,774 ,720 4,369
Look at allergen info of muesli ,827 ,427 ,053 2,287 ,991 5,279
Look at country of origin of ,076 ,433 ,861 1,079 ,462 2,520
Look at ingredients of muesli -,481 ,433 ,266 ,618 ,265 1,444
Habits
Make a list before shopping ,069 ,147 ,638 1,072 ,803 1,429
Search for special deals ,130 ,162 ,424 1,138 ,829 1,563
Buy alw ays same products -,135 ,221 ,541 ,874 ,567 1,347
Frequent allergy info reading -,010 ,386 ,979 ,990 ,465 2,109
Allergy/intolerance m ild ,707 ,356 ,047 2,027 1,009 4,074 *
Allergy/intolerance severe ,392 ,426 ,356 1,481 ,643 3,410
Strong relevance of label ,329 ,366 ,368 1,390 ,678 2,849
Choose healthier alternative -,139 ,389 ,720 ,870 ,406 1,863
Don't compromise quality for ,179 ,354 ,614 1,195 ,597 2,393
Goals & Brand loyal -,830 ,349 ,017 ,436 ,220 ,864 *
attitudes (rc.: Low self-control -,617 ,327 ,059 ,539 ,284 1,024
m oderate/ High risk aversion -,049 ,355 ,889 ,952 ,475 1,908
m edium ) May contain traces - not safe ,504 ,711 ,478 1,656 ,411 6,671
May contain - not safe -,935 ,716 ,192 ,393 ,096 1,597
Cannot guarantee - not safe ,573 ,408 ,161 1,773 ,797 3,946
Made in a facility - not safe ,572 ,475 ,228 1,773 ,698 4,499
Contains nuts - not safe ,839 ,558 ,133 2,314 ,775 6,916
No inform ation given - not -,839 ,352 ,017 ,432 ,217 ,861 *
Age (rc.: 35-54 Age (18 - 34 years) ,309 ,400 ,440 1,362 ,622 2,982
years) Age (55 and more years) ,104 ,411 ,800 1,110 ,496 2,486
Gender (rc.:
Man -,015 ,333 ,964 ,985 ,513 1,890
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,162 ,396 ,683 1,175 ,541 2,555
000 - 29 999) Income (30 000 and more) -,374 ,396 ,345 ,688 ,316 1,496
Education (rc.: Still study -,113 ,646 ,861 ,893 ,252 3,170
finished
Education terminal age max. 17 -,346 ,526 ,511 ,708 ,253 1,982
education w ith
Education terminal age 22+ ,119 ,344 ,728 1,127 ,574 2,211
17-22 years)
Living status Hh 2+ persons w ith children -,417 ,505 ,409 ,659 ,245 1,772
(rc.: singles) Hh 2+ persons w ithout children -,309 ,466 ,508 ,735 ,295 1,831
United Kingdom ,115 ,680 ,865 1,122 ,296 4,259
France -1,600 ,852 ,060 ,202 ,038 1,073
Italy -1,273 ,636 ,045 ,280 ,080 ,974 *
Countries (rc.:
Spain -3,177 ,826 ,000 ,042 ,008 ,210 ***
Germ any)
Finland -,396 ,602 ,511 ,673 ,207 2,189
Poland -1,508 ,668 ,024 ,221 ,060 ,819 *
Romania -1,152 ,685 ,093 ,316 ,083 1,210

N = 586 - Nagelkerke R² = 0,738

Model 14 – Online experiments on PAW

BINOMINAL LOGISTIC REGRESSION - PAW CHOICES 1-3 WITHOUT POLICY

OPTIONS
Std. 95% conf. interval for Exp(B)
Allergy choices 1-3 = PAW avoiding B error Sig. Exp(B) low er bound upper bound
Intercept -1,096 ,728 ,132
Context/ Choice Allergy choice task 1 -,310 ,192 ,106 ,733 ,503 1,069
architecture (rc.:
Allergy choice task 2 -,035 ,192 ,857 ,966 ,663 1,408
task 3)
Zoom ingredients 4,139 ,189 ,000 62,770 43,371 90,847 ***
Aw areness Zoom country of origin -,466 ,193 ,016 ,627 ,430 ,915 *
Zoom nutrition facts -1,292 ,199 ,000 ,275 ,186 ,405 ***
Muesli bought last 6 months ,026 ,257 ,921 1,026 ,620 1,698
Look at allergen info of muesli ,417 ,227 ,067 1,518 ,972 2,371
Look at country of origin of ,001 ,237 ,995 1,001 ,630 1,592
Look at ingredients of muesli -,041 ,241 ,866 ,960 ,599 1,540
Habits
Make a list before shopping -,006 ,079 ,936 ,994 ,851 1,160
Search for special deals ,006 ,086 ,948 1,006 ,849 1,191
Buy alw ays same products -,193 ,117 ,098 ,825 ,656 1,036
Frequent allergy info reading -,108 ,208 ,603 ,897 ,596 1,350
Allergy/intolerance mild ,169 ,187 ,365 1,184 ,821 1,708
Allergy/intolerance severe ,361 ,230 ,117 1,435 ,914 2,254
Strong relevance of label ,352 ,197 ,075 1,422 ,966 2,093
Choose healthier alternative -,156 ,209 ,456 ,856 ,568 1,289
Don't compromise quality for ,127 ,189 ,502 1,135 ,784 1,643
Goals & Brand loyal -,260 ,184 ,158 ,771 ,538 1,106
attitudes (rc.: Low self-control -,178 ,179 ,320 ,837 ,590 1,189
m oderate/ High risk aversion -,310 ,190 ,103 ,734 ,506 1,065
m edium ) May contain traces - not safe ,427 ,379 ,260 1,533 ,729 3,222
May contain - not safe -,005 ,368 ,989 ,995 ,484 2,046
Cannot guarantee - not safe ,666 ,239 ,005 1,946 1,220 3,106 **
Made in a facility - not safe ,083 ,252 ,743 1,086 ,662 1,781
Contains nuts - not safe -,079 ,319 ,806 ,924 ,494 1,728
No inform ation given - not -,718 ,180 ,000 ,488 ,343 ,694 ***
Age (rc.: 35-54 Age (18 - 34 years) ,468 ,223 ,035 1,597 1,033 2,471 *
years) Age (55 and more years) ,266 ,222 ,232 1,305 ,844 2,017
Gender (rc.:
Man -,067 ,181 ,713 ,936 ,656 1,334
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,337 ,208 ,106 1,401 ,931 2,108
000 - 29 999) Income (30 000 and more) -,280 ,218 ,198 ,756 ,493 1,157
Education (rc.: Still study -,525 ,332 ,114 ,592 ,309 1,134
finished
Education terminal age max. 17 -,324 ,290 ,265 ,723 ,409 1,278
education w ith
Education terminal age 22+ ,257 ,191 ,178 1,294 ,889 1,881
17-22 years)
Living status Hh 2+ persons w ith children ,395 ,279 ,157 1,484 ,859 2,565
(rc.: singles) Hh 2+ persons w ithout children ,335 ,256 ,191 1,398 ,846 2,310
United Kingdom ,516 ,379 ,174 1,675 ,797 3,522
France -1,486 ,476 ,002 ,226 ,089 ,576 **
Italy -,772 ,348 ,027 ,462 ,234 ,914 *
Countries (rc.:
Germ any)
Spain -2,545 ,445 ,000 ,078 ,033 ,188 ***
Finland ,055 ,317 ,863 1,056 ,567 1,967
Poland -1,049 ,352 ,003 ,350 ,176 ,698 **
Rom ania -1,065 ,369 ,004 ,345 ,167 ,711 **
N = 586 - Nagelkerke R² = 0,698
Model 15 – Online experiments on PAW

BINOMINAL LOGISTIC REGRESSION - PAW CHOICES 1-3 WITH POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
Allergy choices 1-3 = PAW avoiding B error Sig. Exp(B) low er bound upper bound
Intercept -,851 ,747 ,255
Context/ Choice Allergy choice task 1 -,308 ,192 ,109 ,735 ,504 1,072
architecture (rc.:
Allergy choice task 2 -,038 ,193 ,842 ,962 ,660 1,404
task 3)
Policy options Split 2: May contain traces -,118 ,245 ,629 ,888 ,550 1,435
(rc.: split 1: m ay Split 3: Cannot guarantee -,477 ,237 ,045 ,621 ,390 ,988 *
contain) Split 4: Made in a facility -,477 ,236 ,044 ,621 ,391 ,986 *
Zoom ingredients 4,200 ,193 ,000 66,705 45,733 97,294 ***
Aw areness Zoom country of origin -,465 ,194 ,016 ,628 ,429 ,918 *
Zoom nutrition facts -1,301 ,199 ,000 ,272 ,184 ,402 ***
Muesli bought last 6 months ,007 ,259 ,978 1,007 ,607 1,672
Look at allergen info of muesli ,447 ,228 ,050 1,564 1,000 2,447
Look at country of origin of -,016 ,238 ,948 ,985 ,618 1,569
Look at ingredients of muesli -,036 ,242 ,880 ,964 ,600 1,548
Habits
Make a list before shopping -,005 ,080 ,947 ,995 ,851 1,163
Search for special deals -,001 ,087 ,991 ,999 ,843 1,184
Buy alw ays same products -,190 ,117 ,104 ,827 ,658 1,040
Frequent allergy info reading -,098 ,209 ,638 ,906 ,601 1,365
Allergy/intolerance mild ,240 ,190 ,206 1,271 ,877 1,843
Allergy/intolerance severe ,404 ,231 ,081 1,498 ,952 2,358
Strong relevance of label ,386 ,200 ,053 1,471 ,995 2,177
Choose healthier alternative -,197 ,212 ,352 ,821 ,542 1,244
Don't compromise quality for ,132 ,189 ,484 1,142 ,788 1,655
Goals & Brand loyal -,277 ,187 ,139 ,758 ,525 1,094
attitudes (rc.: Low self-control -,180 ,180 ,316 ,835 ,588 1,188
m oderate/ High risk aversion -,282 ,190 ,139 ,754 ,519 1,096
m edium ) May contain traces - not safe ,470 ,382 ,219 1,600 ,757 3,382
May contain - not safe -,035 ,373 ,925 ,965 ,465 2,004
Cannot guarantee - not safe ,642 ,240 ,007 1,900 1,188 3,038 **
Made in a facility - not safe ,139 ,253 ,584 1,149 ,699 1,887
Contains nuts - not safe -,096 ,325 ,767 ,908 ,480 1,717
No inform ation given - not -,714 ,182 ,000 ,490 ,343 ,699 ***
Age (rc.: 35-54 Age (18 - 34 years) ,461 ,223 ,039 1,586 1,024 2,456 *
years) Age (55 and more years) ,253 ,222 ,255 1,288 ,833 1,990
Gender (rc.:
Man -,033 ,183 ,855 ,967 ,676 1,384
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,319 ,210 ,130 1,376 ,911 2,078
000 - 29 999) Income (30 000 and more) -,260 ,219 ,236 ,771 ,502 1,185
Education (rc.: Still study -,537 ,337 ,111 ,585 ,302 1,132
finished
Education terminal age max. 17 -,317 ,292 ,278 ,728 ,411 1,292
education w ith
Education terminal age 22+ ,257 ,191 ,180 1,292 ,888 1,880
17-22 years)
Living status Hh 2+ persons w ith children ,406 ,280 ,147 1,501 ,867 2,597
(rc.: singles) Hh 2+ persons w ithout children ,329 ,257 ,201 1,389 ,839 2,299
United Kingdom ,483 ,382 ,206 1,621 ,767 3,428
France -1,474 ,473 ,002 ,229 ,091 ,579 **
Italy -,869 ,351 ,013 ,419 ,211 ,835 *
Countries (rc.:
Spain -2,608 ,452 ,000 ,074 ,030 ,179 ***
Germ any)
Finland ,016 ,320 ,960 1,016 ,542 1,904
Poland -1,035 ,353 ,003 ,355 ,178 ,709 **
Rom ania -1,114 ,371 ,003 ,328 ,159 ,678 **
N = 586 -Nagelkerke R² = 0,700

Model 16 – Online experiments on PAW

BINOMINAL LOGISTIC REGRESSION - PAW CHOICES 1-3 WITH POLICY OPTIONS
AND INTERACTION VARIABLES
Std. 95% conf. interval for Exp(B)
Allergy choices 1-3 = PAW avoiding B error Sig. Exp(B) low er bound upper bound
Intercept -,506 ,768 ,510
Context/ Choice Allergy choice task 1 -,312 ,193 ,106 ,732 ,501 1,069
architecture (rc.: Allergy choice task 2 -,048 ,193 ,804 ,953 ,653 1,392
task 3)
Split 2: May contain traces -,187 ,354 ,597 ,829 ,414 1,661
Split 3: Cannot guarantee -1,190 ,374 ,001 ,304 ,146 ,633 **
Policy options
Split 4: Made in a facility -,692 ,347 ,046 ,501 ,254 ,988 *
(rc.: split 1: m ay
Interaction: split 3 & 1,077 ,431 ,012 2,936 1,261 6,836 *
contain)
Interaction: split 2 & traces ,069 ,427 ,872 1,071 ,464 2,474
Interaction: split 4 & made in ,348 ,416 ,403 1,416 ,626 3,203
Zoom ingredients 4,238 ,194 ,000 69,258 47,316 101,375 ***
Aw areness Zoom country of origin -,479 ,195 ,014 ,620 ,423 ,908 *
Zoom nutrition facts -1,303 ,200 ,000 ,272 ,184 ,402 ***
Muesli bought last 6 months -,016 ,261 ,952 ,984 ,590 1,643
Look at allergen info of muesli ,440 ,229 ,054 1,553 ,993 2,431
Look at country of origin of -,042 ,240 ,860 ,959 ,599 1,534
Look at ingredients of muesli ,016 ,244 ,949 1,016 ,629 1,640
Habits
Make a list before shopping -,018 ,080 ,820 ,982 ,839 1,149
Search for special deals -,005 ,087 ,950 ,995 ,839 1,179
Buy alw ays same products -,178 ,118 ,132 ,837 ,664 1,055
Frequent allergy info reading -,160 ,213 ,452 ,852 ,561 1,293
Allergy/intolerance mild ,238 ,191 ,211 1,269 ,873 1,845
Allergy/intolerance severe ,404 ,233 ,084 1,498 ,948 2,366
Strong relevance of label ,393 ,200 ,049 1,482 1,001 2,194 *
Choose healthier alternative -,248 ,214 ,246 ,780 ,513 1,186
Don't compromise quality for ,130 ,190 ,495 1,138 ,784 1,652
Goals & Brand loyal -,260 ,189 ,168 ,771 ,533 1,116
attitudes (rc.: Low self-control -,182 ,181 ,313 ,833 ,585 1,187
m oderate/ High risk aversion -,246 ,193 ,200 ,782 ,536 1,140
m edium ) May contain traces - not safe ,371 ,407 ,362 1,449 ,653 3,218
May contain - not safe -,091 ,380 ,812 ,913 ,433 1,925
Cannot guarantee - not safe ,394 ,261 ,132 1,483 ,888 2,475
Made in a facility - not safe ,062 ,272 ,820 1,064 ,624 1,814
Contains nuts - not safe -,057 ,331 ,864 ,945 ,494 1,806
No inform ation given - not -,656 ,185 ,000 ,519 ,361 ,746 ***
Age (rc.: 35-54 Age (18 - 34 years) ,428 ,225 ,057 1,534 ,987 2,385
years) Age (55 and more years) ,234 ,224 ,296 1,264 ,815 1,961
Gender (rc.:
Man -,064 ,184 ,729 ,938 ,654 1,345
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,322 ,213 ,131 1,379 ,909 2,093
000 - 29 999) Income (30 000 and more) -,264 ,222 ,235 ,768 ,497 1,188
Education (rc.: Still study -,530 ,337 ,116 ,589 ,304 1,140
finished
Education terminal age max. 17 -,366 ,296 ,216 ,693 ,388 1,239
education w ith
Education terminal age 22+ ,228 ,193 ,238 1,256 ,860 1,834
17-22 years)
Living status Hh 2+ persons w ith children ,406 ,284 ,153 1,500 ,860 2,618
(rc.: singles) Hh 2+ persons w ithout children ,336 ,261 ,199 1,399 ,839 2,333
United Kingdom ,595 ,387 ,124 1,813 ,850 3,868
France -1,508 ,472 ,001 ,221 ,088 ,558 **
Italy -,855 ,352 ,015 ,425 ,213 ,847 *
Countries (rc.:
Spain -2,659 ,453 ,000 ,070 ,029 ,170 ***
Germ any)
Finland ,008 ,323 ,980 1,008 ,536 1,897
Poland -1,015 ,355 ,004 ,362 ,181 ,727 **
Rom ania -1,179 ,374 ,002 ,307 ,148 ,640 **
N = 586 Nagelkerke R² = 0,703

Model 17 – Online experiments on Alcohol

BINOMINAL LOGISTIC REGRESSION - ALCOHOL CHOICE 1 WITHOUT POLICY
OPTIONS
Intention to reduce alcohol at specific Std. 95% conf. interval for Exp(B)
occasion B error Sig. Exp(B) low er bound upper bound
Intercept -,871 ,411 ,034 *
Context (ref. Experiments about beer -,051 ,136 ,709 ,951 ,728 1,241
cat.: spirits) Experim ents about w ine ,275 ,130 ,035 1,317 1,020 1,700 *
Health planner ,373 ,133 ,005 1,451 1,118 1,884 **
Rushed quality shopper -,229 ,136 ,093 ,796 ,609 1,039
Self determined shopper -,313 ,177 ,077 ,731 ,517 1,034
Bargain hunter ,157 ,152 ,300 1,170 ,869 1,576
Frustrated shopper -,395 ,178 ,027 ,674 ,475 ,955 *
Look at food info label ,069 ,064 ,276 1,072 ,946 1,214
Prefer low alcoholic content ,527 ,117 ,000 1,693 1,345 2,131 ***
Prefer high alcoholic content ,256 ,191 ,180 1,292 ,888 1,880
Attitude
Prefer small serving size ,120 ,146 ,411 1,128 ,847 1,501
Prefer large serving size -,089 ,222 ,688 ,915 ,592 1,413
Prefer few er calories ,258 ,154 ,094 1,294 ,957 1,750
Com m on peer group -,297 ,063 ,000 ,743 ,656 ,842 ***
Alcohol has m ore
-,135 ,062 ,029 ,874 ,774 ,986 *
advantages
Low self control ,149 ,096 ,119 1,160 ,962 1,399
Risk aversion ,087 ,130 ,502 1,091 ,846 1,408
Diet / Weight loss ,312 ,122 ,011 1,366 1,075 1,735 *
Number of alcoholic bever. ,042 ,061 ,490 1,043 ,926 1,174
Fam iliarity w ith drinking
-,243 ,092 ,008 ,784 ,655 ,939
occasion (friends' party) **
Habits Too much alcohol -,005 ,101 ,961 ,995 ,817 1,213
Frequency of drinking -,141 ,068 ,039 ,869 ,760 ,993 *
Alcohol volume last time -,063 ,103 ,541 ,939 ,768 1,148
Alcohol volume friends' party ,069 ,091 ,448 1,071 ,897 1,280
Age (rc.: 35-54 Age (18 - 34 years) ,107 ,133 ,422 1,113 ,857 1,445
years) Age (55 and more years) -,011 ,112 ,921 ,989 ,794 1,231
Gender (rc.:
Man -,209 ,103 ,043 ,811 ,663 ,993 *
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,112 ,109 ,307 1,118 ,902 1,385
000 - 29 999) Income (30 000 and more) ,045 ,121 ,710 1,046 ,825 1,325
Education (rc.: Still study ,139 ,210 ,507 1,149 ,762 1,733
finished
Education terminal age max. 17 ,231 ,140 ,099 1,260 ,957 1,657
education w ith
Education terminal age 22+ ,165 ,105 ,118 1,179 ,959 1,449
17-22 years)
Living status Hh 2+ persons w ith child. -,238 ,134 ,074 ,788 ,606 1,024
(rc.: singles) Hh 2+ persons w ithout child. -,112 ,119 ,349 ,894 ,708 1,130
BMI (rc.: BMI
Overw eight (BMI 25+) ,015 ,096 ,879 1,015 ,840 1,225
under 25)
United Kingdom -,352 ,196 ,073 ,703 ,478 1,034
France ,146 ,175 ,404 1,157 ,821 1,631
Germany -,355 ,192 ,064 ,701 ,481 1,021
Countries (rc.:
Italy ,179 ,171 ,295 1,196 ,855 1,673
Spain)
Finland -,491 ,208 ,018 ,612 ,407 ,921 *
Poland -,478 ,190 ,012 ,620 ,427 ,900 *
Romania ,236 ,172 ,170 1,267 ,904 1,775

N = 3.967 - Nagelkerke R² = 0,117

Model 18 – Online experiments on Alcohol

BINOMINAL LOGISTIC REGRESSION - ALCOHOL CHOICE 1 WITH POLICY OPTIONS

Intention to reduce alcohol at specific Std. 95% conf. interval for Exp(B)
occasion B error Sig. Exp(B) low er bound upper bound
Intercept -1,111 ,399 ,005 **
Split 2: Beer & Lim its ,481 ,163 ,003 1,617 1,176 2,224 **
Policy options Split 3: Wine & Calories ,611 ,160 ,000 1,842 1,345 2,521 ***
(ref. cat.: Split 1: Split 4: Wine & Lim its ,507 ,163 ,002 1,661 1,206 2,288 **
Beer & Calories) Split 5: Spirits & Calories ,106 ,188 ,572 1,112 ,770 1,606
Split 6: Spirits & Lim its ,494 ,186 ,008 1,639 1,139 2,359 **
Aw areness Correct answ er know ledge test -,174 ,116 ,134 ,840 ,669 1,055
Health planner ,370 ,134 ,006 1,448 1,115 1,882 **
Rushed quality shopper -,242 ,137 ,076 ,785 ,600 1,026
Self determined shopper -,317 ,177 ,074 ,728 ,515 1,031
Bargain hunter ,153 ,152 ,316 1,165 ,864 1,570
Frustrated shopper -,386 ,178 ,030 ,680 ,479 ,964 *
Look at food info label ,072 ,064 ,256 1,075 ,949 1,218
Prefer low alcoholic content ,542 ,118 ,000 1,719 1,364 2,166 ***
Prefer high alcoholic content ,260 ,192 ,175 1,298 ,891 1,890
Attitude
Prefer small serving size ,141 ,147 ,336 1,151 ,864 1,534
Prefer large serving size -,097 ,223 ,662 ,907 ,586 1,405
Prefer few er calories ,255 ,155 ,099 1,291 ,953 1,748
Com m on peer group -,301 ,064 ,000 ,740 ,653 ,838 ***
Alcohol has m ore
-,139 ,062 ,025 ,870 ,771 ,983
advantages *
Low self control ,158 ,096 ,099 1,171 ,971 1,413
Risk aversion ,088 ,131 ,500 1,092 ,845 1,411
Diet / Weight loss ,328 ,123 ,007 1,389 1,092 1,766 **
Number of alcoholic bever. ,030 ,061 ,620 1,031 ,914 1,163
Fam iliarity w ith drinking
-,244 ,092 ,008 ,783 ,654 ,939
occasion (friends' party) **
Habits Too much alcohol -,011 ,101 ,913 ,989 ,811 1,206
Frequency of drinking alcohol -,134 ,068 ,050 ,875 ,765 1,000
Alcohol volume last time -,059 ,103 ,564 ,942 ,770 1,153
Alcohol volume friends' party ,063 ,091 ,486 1,065 ,892 1,273
Age (rc.: 35-54 Age (18 - 34 years) ,105 ,134 ,434 1,110 ,854 1,443
years) Age (55 and more years) -,031 ,112 ,781 ,969 ,778 1,208
Gender (rc.:
Man -,199 ,104 ,055 ,820 ,669 1,005
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,115 ,110 ,295 1,122 ,905 1,391
000 - 29 999) Income (30 000 and more) ,046 ,121 ,704 1,047 ,826 1,328
Education (rc.: Still study ,156 ,211 ,460 1,169 ,773 1,766
finished
Education terminal age max. 17 ,225 ,141 ,109 1,253 ,951 1,650
education w ith
Education terminal age 22+ ,175 ,106 ,097 1,191 ,969 1,465
17-22 years)
Living status Hh 2+ persons w ith child. -,247 ,134 ,065 ,781 ,601 1,016
(rc.: singles) Hh 2+ persons w ithout child. -,106 ,120 ,374 ,899 ,711 1,137
BMI (rc.: BMI
Overw eight (BMI 25+) ,015 ,096 ,875 1,015 ,840 1,226
under 25)
United Kingdom -,329 ,197 ,096 ,720 ,489 1,060
France ,152 ,176 ,387 1,164 ,825 1,642
Germany -,362 ,192 ,060 ,696 ,478 1,015
Countries (rc.:
Italy ,169 ,172 ,325 1,184 ,846 1,658
Spain)
Finland -,483 ,209 ,021 ,617 ,409 ,929 *
Poland -,484 ,191 ,011 ,616 ,424 ,895 *
Romania ,220 ,173 ,203 1,246 ,888 1,749

N = 3.967 - Nagelkerke R² = 0,125

Model 19 – Online experiments on Alcohol

BINOMINAL LOGISTIC REGRESSION - ALCOHOL CHOICE 2 WITHOUT POLICY
OPTIONS
Intention to reduce alcohol at any Std. 95% conf. interval for Exp(B)
occasion in future B error Sig. Exp(B) low er bound upper bound
Intercept - 2 ,3 5 5 ,496 ,0 0 0 ***
Context (ref. Experiments about beer -,218 ,160 ,172 ,804 ,588 1,099
cat.: spirits) Experiments about w ine -,074 ,155 ,633 ,929 ,685 1,258
Intention to drink less after
Intention 3 ,10 3 ,111 ,0 0 0 22,261 17,892 27,698 ***
education
Health planner -,103 ,164 ,532 ,902 ,654 1,245
Rushed quality shopper -,274 ,163 ,093 ,760 ,553 1,046
Self determined shopper -,136 ,197 ,490 ,873 ,594 1,284
Bargain hunter -,054 ,182 ,768 ,948 ,664 1,353
Frustrated shopper -,280 ,204 ,170 ,756 ,507 1,128
Look at food info label ,134 ,076 ,078 1,144 ,985 1,328
Prefer low alcoholic content ,111 ,147 ,450 1,118 ,837 1,492
Prefer high alcoholic content ,032 ,233 ,892 1,032 ,654 1,630
Attitude
Prefer small serving size ,295 ,178 ,097 1,344 ,948 1,904
Prefer large serving size -,140 ,272 ,608 ,870 ,510 1,482
Prefer few er calories ,131 ,192 ,493 1,140 ,783 1,660

Com m on peer group - ,2 9 0 ,075 ,0 0 0 ,748 ,646 ,867 ***

Alcohol has m ore - ,19 1 ,074 ,0 10 ,827 ,716 ,955
advantages *
Low self control -,077 ,113 ,495 ,926 ,742 1,155

Risk aversion ,3 14 ,156 ,0 4 4 1,369 1,008 1,859 *

Diet / Weight loss ,207 ,149 ,165 1,230 ,918 1,647
Number of alcoholic bever. -,049 ,071 ,490 ,952 ,828 1,095
Familiarity w ith drinking
-,061 ,110 ,580 ,941 ,759 1,167
occasion (friends' party)
Habits Too much alcohol ,087 ,122 ,475 1,091 ,859 1,385
Frequency of drinking alcohol ,067 ,080 ,404 1,069 ,914 1,251
Alcohol volume last time ,042 ,116 ,719 1,043 ,830 1,310
Alcohol volume friends' party ,131 ,106 ,218 1,139 ,926 1,402

Age (rc.: 35-54 Age (18 - 34 years) ,157 ,159 ,323 1,170 ,857 1,598
years) Age (55 and more years) ,002 ,135 ,988 1,002 ,768 1,307
Gender (rc.:
Man -,156 ,123 ,206 ,856 ,672 1,089
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,060 ,132 ,650 1,062 ,820 1,374
000 - 29 999) Income (30 000 and more) -,262 ,146 ,072 ,769 ,578 1,023
Education (rc.: Still study ,084 ,249 ,738 1,087 ,667 1,772
finished -,172 ,166 ,302 ,842 ,608 1,166
Education terminal age max. 17
education w ith
Education terminal age 22+ -,110 ,125 ,381 ,896 ,701 1,145
17-22 years)
Living status Hh 2+ persons w ith child. -,046 ,159 ,774 ,955 ,700 1,304
(rc.: singles) Hh 2+ persons w ithout child. ,014 ,141 ,922 1,014 ,768 1,338
BMI (rc.: BMI
Overw eight (BMI 25+) ,173 ,116 ,135 1,189 ,947 1,491
under 25)
United Kingdom ,136 ,233 ,560 1,146 ,725 1,810

France ,5 11 ,215 ,0 17 1,668 1,095 2,540 *

Germ any ,5 5 4 ,222 ,0 13 1,741 1,126 2,692 *
Countries (rc.:
Italy ,220 ,215 ,307 1,246 ,817 1,900
Spain)
Finland -,039 ,249 ,875 ,962 ,590 1,566
Poland ,009 ,226 ,968 1,009 ,649 1,570
Romania -,119 ,219 ,589 ,888 ,578 1,365

N = 3.967 - Nagelkerke R² = 0,415

Model 20 – Online experiments on Alcohol

BINOMINAL LOGISTIC REGRESSION - ALCOHOL CHOICE 2 WITH POLICY OPTIONS
Intention to reduce alcohol at any Std. 95% conf. interval for Exp(B)
occasion in future B error Sig. Exp(B) low er bound upper bound
Intercept -2,740 ,480 ,000 ***
Split 2: Beer & Limits ,268 ,192 ,162 1,307 ,898 1,904
Policy options Split 3: Wine & Calories ,370 ,189 ,051 1,448 ,999 2,099
(rc.: Split 1: Beer Split 4: Wine & Limits ,230 ,192 ,232 1,258 ,864 1,833
& Calories) Split 5: Spirits & Calories ,372 ,214 ,083 1,451 ,953 2,208
Split 6: Spirits & Limits ,366 ,218 ,093 1,442 ,941 2,208
Aw areness Correct answ er know ledge test ,112 ,133 ,400 1,119 ,861 1,452
Intention to drink less after
Intention 3,099 ,112 ,000 22,186 17,815 27,629 ***
education
Health planner -,108 ,165 ,512 ,898 ,650 1,240
Rushed quality shopper -,279 ,163 ,086 ,756 ,549 1,041
Self determined shopper -,132 ,197 ,502 ,876 ,596 1,289
Bargain hunter -,051 ,182 ,780 ,950 ,665 1,358
Frustrated shopper -,286 ,205 ,162 ,751 ,503 1,122
Look at food info label ,135 ,076 ,077 1,145 ,986 1,329
Prefer low alcoholic content ,118 ,148 ,422 1,126 ,843 1,504
Attitude Prefer high alcoholic content ,034 ,234 ,883 1,035 ,655 1,636
Prefer small serving size ,299 ,178 ,094 1,348 ,950 1,913
Prefer large serving size -,155 ,273 ,571 ,857 ,502 1,463
Prefer few er calories ,117 ,192 ,542 1,124 ,771 1,639
Com m on peer group -,291 ,075 ,000 ,748 ,645 ,867 ***
Alcohol has m ore -,191 ,074 ,009 ,826 ,715 ,954
advantages **
Low self control -,073 ,113 ,521 ,930 ,745 1,160
Risk aversion ,307 ,157 ,050 1,359 1,000 1,847
Diet / Weight loss ,203 ,149 ,174 1,225 ,914 1,642
Number of alcoholic bever. -,048 ,072 ,505 ,953 ,828 1,098
Familiarity w ith drinking
-,062 ,110 ,571 ,940 ,758 1,166
occasion (friends' party)
Habits Too much alcohol ,082 ,122 ,499 1,086 ,855 1,379
Frequency of drinking alcohol ,066 ,080 ,415 1,068 ,912 1,250
Alcohol volume last time ,045 ,117 ,702 1,046 ,832 1,314
Alcohol volume friends' party ,128 ,106 ,228 1,136 ,923 1,399
Age (rc.: 35-54 Age (18 - 34 years) ,160 ,159 ,315 1,173 ,859 1,603
years) Age (55 and more years) ,004 ,136 ,977 1,004 ,769 1,310
Gender (rc.:
Man -,153 ,123 ,216 ,859 ,674 1,093
w om an)
Incom e (rc.: 10 Income (0 - 9 999) ,062 ,132 ,638 1,064 ,822 1,377
000 - 29 999) Income (30 000 and more) -,264 ,146 ,069 ,768 ,577 1,021
Education (rc.: Still study ,080 ,250 ,749 1,083 ,664 1,767
finished
Education terminal age max. 17 -,169 ,166 ,309 ,844 ,609 1,170
education w ith
Education terminal age 22+ -,103 ,125 ,410 ,902 ,706 1,153
17-22 years)
Living status Hh 2+ persons w ith child. -,043 ,159 ,789 ,958 ,702 1,308
(rc.: singles) Hh 2+ persons w ithout child. ,015 ,142 ,914 1,015 ,769 1,340
BMI (rc.: BMI
Overw eight (BMI 25+) ,170 ,116 ,142 1,186 ,945 1,488
under 25)
United Kingdom ,122 ,234 ,602 1,130 ,714 1,788
France ,511 ,215 ,017 1,668 1,095 2,540 *
Germ any ,554 ,222 ,013 1,740 1,125 2,691 *
Countries (rc.:
Italy ,223 ,215 ,301 1,250 ,819 1,906
Spain)
Finland -,047 ,249 ,851 ,954 ,586 1,555
Poland ,011 ,226 ,960 1,011 ,650 1,574
Romania -,118 ,220 ,590 ,888 ,578 1,366

N = 3.967 - Nagelkerke R² = 0,416

Model 21 – Online experiments on Food Waste

BINOMINAL LOGISTIC REGRESSION - AVOIDING OUTDATED FOOD
Exp(B Exp(B)
Not eating outdated food B Std. error Sig. ) bound bound
Intercept -,260 ,141 ,064
Misconceptions
best before = safety lim it ,602 ,057 ,000 1,825 1,633 2,040 ***
(rc.: no
use by = quality limit -,079 ,068 ,247 ,924 ,808 1,056
m isconceptions
no food w aste ,002 ,053 ,968 1,002 ,904 1,111
)
Reading expiry dates w hen ,167 ,060 ,005 1,182 1,051 1,330
Habits (rc.: not shopping crisps **
reading) Reading expiry dates w hen ,137 ,063 ,030 1,147 1,013 1,298
shopping m uesli *
expert_shopper ,169 ,077 ,029 1,184 1,018 1,378 *
Motives (rc.: health_planner ,186 ,084 ,026 1,205 1,022 1,421 *
rushed quality self_determ ined_shopper -,337 ,087 ,000 ,714 ,602 ,847 ***
shopper) bargain_hunter -,269 ,087 ,002 ,764 ,645 ,906 **
frustrated_shopper -,292 ,089 ,001 ,747 ,628 ,889 **
Low self-control ,093 ,054 ,083 1,098 ,988 1,220
Attitudes
High risk-aversion ,270 ,068 ,000 1,310 1,146 1,497 ***
Age (rc.: 35-54 Age (18 - 34 years) -,136 ,071 ,054 ,873 ,759 1,003
years) Age (55 and more years) -,023 ,063 ,712 ,977 ,863 1,106
Gender Man ,236 ,053 ,000 1,267 1,142 1,405 ***
Incom e (rc.: 10 Income (0 - 9 999) ,022 ,065 ,733 1,022 ,900 1,161
000 - 29 999) Income (30 000 and more) -,143 ,064 ,026 ,867 ,765 ,983 *
Education (rc.: Still study -,248 ,116 ,033 ,781 ,622 ,980
*
finished Education terminal age max. 17 ,068 ,078 ,381 1,070 ,919 1,246
education w ith
17-22 years) Education terminal age 22+ -,151 ,060 ,012 ,860 ,765 ,967
*
Living status Hh 2+ persons w ith child. ,025 ,074 ,733 1,026 ,886 1,187
(rc.: singles) Hh 2+ persons w ithout child. ,009 ,068 ,900 1,009 ,883 1,153
BMI (rc.: BMI Overw eight (BMI 25+) -,100 ,053 ,057 ,904 ,815 1,003
under 25)
France -,883 ,100 ,000 ,413 ,340 ,503 ***
Germ any -,990 ,101 ,000 ,372 ,305 ,453 ***
Italy ,069 ,101 ,495 1,071 ,879 1,304
Countries (rc.:
Spain -,564 ,098 ,000 ,569 ,469 ,689 ***
UK)
Finland ,119 ,101 ,237 1,126 ,925 1,372
Poland ,630 ,109 ,000 1,877 1,517 2,323 ***
Rom ania 1,011 ,127 ,000 2,748 2,142 3,525 ***
2
N = 7717 - Nagelkerke R = 0,201
Model 22 – Online experiments on Food Waste

BINOMINAL LOGISTIC REGRESSION - THROWING AWAY NOT-OUTDATED FOOD

Exp(B Exp(B)
Throwing away not-outdated food B Std. error Sig. ) bound bound
Konstanter Term -,858 ,139 ,000 ***
Misconceptions
best before = safety lim it ,135 ,056 ,015 1,145 1,027 1,277 *
(rc.: no
use by = quality limit ,001 ,068 ,991 1,001 ,876 1,144
m isconceptions
no food w aste -,002 ,053 ,972 ,998 ,900 1,107
)
Habits (rc.: not Reading expiry dates w hen ,038 ,058 ,510 1,039 ,927 1,165
reading) Reading expiry dates w hen ,014 ,061 ,823 1,014 ,899 1,143
expert_shopper -,199 ,076 ,008 ,819 ,706 ,950 **
Motives (rc.: health_planner -,205 ,081 ,011 ,815 ,696 ,955 *
rushed quality self_determ ined_shopper ,167 ,085 ,048 1,182 1,001 1,395 *
shopper) bargain_hunter -,074 ,086 ,386 ,928 ,785 1,098
frustrated_shopper -,013 ,087 ,884 ,987 ,832 1,172
Low self-control ,223 ,053 ,000 1,250 1,127 1,387 ***
Attitudes
High risk-aversion ,152 ,068 ,024 1,164 1,020 1,329 *
Age (rc.: 35-54 Age (18 - 34 years) ,467 ,067 ,000 1,595 1,399 1,819 ***
years) Age (55 and m ore years) -,428 ,062 ,000 ,652 ,577 ,736 ***
Gender Man -,170 ,052 ,001 ,844 ,763 ,934 **
Incom e (rc.: 10 Incom e (0 - 9 999) -,145 ,063 ,021 ,865 ,765 ,979 *
000 - 29 999) Income (30 000 and more) ,095 ,063 ,132 1,100 ,972 1,245
Education (rc.: Still study -,095 ,109 ,387 ,910 ,734 1,127
finished Education terminal age max. 17 -,050 ,080 ,534 ,951 ,813 1,114
education w ith
17-22 years) Education terminal age 22+ ,062 ,057 ,280 1,064 ,951 1,191
Living status Hh 2+ persons w ith child. ,055 ,074 ,460 1,056 ,913 1,222
(rc.: singles) Hh 2+ persons w ithout ,148 ,068 ,030 1,160 1,014 1,326 *
BMI (rc.: BMI Overw eight (BMI 25+) ,036 ,052 ,490 1,036 ,936 1,147
under 25)
France -,126 ,103 ,218 ,881 ,721 1,078
Germany ,050 ,101 ,622 1,051 ,862 1,283
Italy -,052 ,104 ,618 ,949 ,774 1,164
Countries (rc.:
Spain -,167 ,102 ,102 ,846 ,692 1,034
UK)
Finland ,179 ,103 ,084 1,196 ,976 1,464
Poland ,106 ,107 ,325 1,111 ,901 1,371
Rom ania ,890 ,116 ,000 2,435 1,940 3,055 ***
2
N = 7717 - Nagelkerke R = 0,084
Model 23 – Offline store experiments on TFA

BINOMINAL LOGISTIC REGRESSION - INITIAL CHOICE 1 - WITH POLICY OPTIONS

Std. 95% conf. interval for Exp(B)
Initial choice 1 = healthier Product B B error Sig. Exp(B) lower bound upper bound
Intercept -,829 ,473 ,080
Context (rc: pizza -,841 ,210 ,000 ,431 ,286 ,651 ***
biscuits) yoghurt -,936 ,234 ,000 ,392 ,248 ,621 ***
Policy (rc: control
group - no TFA policy1 (w ith TFA info) ,407 ,175 ,020 1,503 1,067 2,117 *
info)
Saturates = DK/healthy -,117 ,159 ,464 ,890 ,651 1,216
PHO = DK/healthy -,265 ,178 ,137 ,767 ,541 1,088
Aw areness Q41
TFA = DK/healthy -,236 ,163 ,148 ,790 ,574 1,087
(rc: unhealthy)
Sugar = DK/healthy ,014 ,251 ,954 1,015 ,620 1,659
Salt = DK/healthy ,101 ,169 ,552 1,106 ,794 1,541
Fat 2,383 ,208 ,000 10,841 7,207 16,308 ***
Salt -,528 ,316 ,095 ,590 ,318 1,095
Access Q32 (rc: Sugar -1,803 ,268 ,000 ,165 ,098 ,278 ***
no access) Nutrition fact panel ,315 ,188 ,093 1,371 ,948 1,981
Ingredients list ,168 ,172 ,331 1,182 ,844 1,657
Irrelevant labels or nothing -,887 ,184 ,000 ,412 ,287 ,591 ***
Observation short (< 15 seconds) ,292 ,181 ,106 1,339 ,940 1,907
duration Q4-9
(rc: medium) long (60 seconds plus) -,014 ,199 ,942 ,986 ,667 1,456
Shelf behaviour reading at shelf before
-,312 ,202 ,123 ,732 ,492 1,089
Q10 (rc: grab&go) selection to buy
Fat ,071 ,283 ,802 1,073 ,617 1,868
Reading habits
Nutrition fact panel ,039 ,314 ,901 1,040 ,562 1,923
at shelf Q25/26
Sugar ,164 ,313 ,600 1,179 ,638 2,178
(rc: no reading)
Ingredients list ,541 ,210 ,010 1,718 1,139 2,591 *
Influenced by w as influenced ,532 ,231 ,021 1,702 1,083 2,675 *
shopping company
Q11/12 (rc: alone) not influenced ,140 ,203 ,489 1,151 ,774 1,711
Shopping fre- less often ,299 ,212 ,159 1,348 ,890 2,042
quency Q16 (rc:
every 2-3 w eeks) more often ,114 ,203 ,572 1,121 ,754 1,668
Planned to shop
before Q22 (rc: not planned before -,120 ,165 ,465 ,887 ,642 1,225
yes)
Bought before
no, bought first time ,016 ,193 ,933 1,016 ,696 1,484
Q24 (rc: yes)
I look for food info … -,143 ,289 ,621 ,867 ,492 1,527
Main shopping I w ant to get in and out ... as
-,135 ,280 ,629 ,874 ,505 1,512
goal Q28 (rc: not fast as I can
chosen) I w ant a lot of variety … -,203 ,216 ,348 ,816 ,534 1,247
I like looking for new and … -,134 ,201 ,505 ,875 ,590 1,296
Relevance of health impact = w eak -,295 ,201 ,142 ,744 ,502 1,104
health Q27 (rc:
medium) health impact = strong ,223 ,200 ,265 1,250 ,844 1,851
Relevance of convenience = w eak -,519 ,202 ,010 ,595 ,400 ,884 *
convenience
Q27 (rc: medium) convenience = strong -,539 ,183 ,003 ,583 ,408 ,834 **
Gender man -,215 ,174 ,218 ,807 ,573 1,135
Age (rc: 35-54 young (16-34 years) -,139 ,192 ,470 ,871 ,598 1,268
years) old (55+ years) -,059 ,211 ,779 ,943 ,624 1,424
hhsize (rc: 2 single ,002 ,221 ,993 1,002 ,649 1,546
persons) 3+ persons -,071 ,184 ,698 ,931 ,650 1,335
Occupation (rc: full-time (AC) ,144 ,210 ,494 1,155 ,765 1,744
part-time) w ork_other (EFGHJL) -,107 ,224 ,634 ,899 ,579 1,394
Education (rc: low _edu (AB) ,203 ,205 ,322 1,225 ,820 1,832
medium) high_edu (EFH) ,051 ,177 ,774 1,052 ,744 1,487
low _inc (up to 2000 €) ,218 ,215 ,310 1,244 ,816 1,897
Incom e (rc:
high_inc (3000 € plus) ,089 ,224 ,691 1,093 ,704 1,696
medium)
na_inc (no answ er) ,392 ,246 ,111 1,480 ,914 2,398
2
N = 1851 - Nagelkerke R = 0,339

Model 24 – Offline store experiments on TFA

BINOMINAL LOGISTIC REGRESSION - INITIAL CHOICE 1 - WITHOUT POLICY
Std. 95% conf. interval for Exp(B)
Initial choice 1 = healthier Product B B error Sig. Exp(B) lower bound upper bound
Intercept -,509 ,452 ,260
Context (rc: pizza -,811 ,210 ,000 ,444 ,294 ,671 ***
biscuits) yoghurt -,976 ,233 ,000 ,377 ,239 ,595 ***
Saturates = DK/healthy -,123 ,159 ,442 ,885 ,647 1,209
PHO = DK/healthy -,250 ,178 ,160 ,778 ,549 1,104
Aw areness Q41
TFA = DK/healthy -,235 ,163 ,148 ,791 ,575 1,087
(rc: unhealthy)
Sugar = DK/healthy -,020 ,250 ,937 ,980 ,601 1,600
Salt = DK/healthy ,120 ,169 ,476 1,128 ,810 1,569
Fat 2,362 ,208 ,000 10,608 7,056 15,946 ***
Salt -,598 ,315 ,058 ,550 ,296 1,020
Access Q32 (rc: Sugar -1,783 ,267 ,000 ,168 ,100 ,284 ***
no access) Nutrition fact panel ,250 ,186 ,179 1,284 ,892 1,847
Ingredients list ,125 ,171 ,467 1,133 ,810 1,585
Irrelevant labels or nothing -,853 ,184 ,000 ,426 ,297 ,611 ***
Observation short (< 15 seconds) ,253 ,179 ,157 1,288 ,907 1,829
duration Q4-9 long (60 seconds plus) ,012 ,199 ,951 1,012 ,685 1,495
Shelf behaviour reading at shelf before
-,301 ,202 ,136 ,740 ,498 1,099
Q10 (rc: grab&go) selection to buy
Fat ,101 ,282 ,720 1,106 ,637 1,921
Reading habits
Nutrition fact panel -,008 ,314 ,979 ,992 ,536 1,834
at shelf Q25/26
Sugar ,217 ,312 ,486 1,243 ,675 2,288
(rc: no reading)
Ingredients list ,555 ,209 ,008 1,743 1,156 2,626 **
Influenced by w as influenced ,510 ,230 ,027 1,665 1,060 2,615 *
shopping company
Q11/12 (rc: alone) not influenced ,128 ,203 ,529 1,136 ,764 1,690
Shopping fre- less often ,279 ,212 ,187 1,322 ,873 2,001
quency Q16 (rc:
every 2-3 w eeks) more often ,140 ,202 ,490 1,150 ,774 1,708
Planned to shop
not planned before -,103 ,164 ,533 ,903 ,654 1,245
before Q22 (rc:
Bought before
no, bought first time ,042 ,193 ,829 1,043 ,715 1,521
Q24 (rc: yes)
I look for food info … -,161 ,288 ,576 ,851 ,483 1,498
Main shopping
I w ant to get in and out ... as
goal Q28 (rc: not -,166 ,279 ,552 ,847 ,490 1,463
fast as I can
chosen)
I w ant a lot of variety … -,247 ,216 ,253 ,781 ,512 1,192
I like looking for new and … -,139 ,200 ,488 ,870 ,588 1,289
Relevance of health impact = w eak -,310 ,201 ,124 ,734 ,495 1,088
health Q27 (rc: health impact = strong ,216 ,200 ,279 1,241 ,839 1,837
Relevance of convenience = w eak -,540 ,202 ,007 ,583 ,392 ,866 **
convenience
Q27 (rc: medium) convenience = strong -,528 ,182 ,004 ,590 ,413 ,843 **
Gender man -,185 ,173 ,285 ,831 ,592 1,167
Age (rc: 35-54 young (16-34 years) -,124 ,191 ,516 ,883 ,607 1,285
years) old (55+ years) -,045 ,210 ,830 ,956 ,633 1,443
hhsize (rc: 2 single -,005 ,221 ,983 ,995 ,646 1,535
persons) 3+ persons -,084 ,183 ,646 ,919 ,642 1,316
Occupation (rc: full-time (AC) ,119 ,209 ,571 1,126 ,747 1,697
part-time) w ork_other (EFGHJL) -,144 ,223 ,519 ,866 ,559 1,342
Education (rc: low _edu (AB) ,171 ,205 ,403 1,186 ,795 1,772
medium) high_edu (EFH) ,070 ,176 ,689 1,073 ,760 1,515
low _inc (up to 2000 €) ,203 ,215 ,345 1,225 ,804 1,867
Incom e (rc:
high_inc (3000 € plus) ,113 ,223 ,614 1,119 ,723 1,734
medium)
na_inc (no answ er) ,373 ,246 ,130 1,452 ,896 2,353
2
N = 1851 - Nagelkerke R = 0,335
Model 25 – Offline store experiments on TFA
BINOMINAL LOGISTIC REGRESSION - INITIAL CHOICE 2 - WITH POLICY
Std. 95% conf. interval for Exp(B)
Initial choice 2 = healthier Product X B error Sig. Exp(B) lower bound upper bound
Intercept ,406 ,407 ,319
Context (rc: pizza -,990 ,191 ,000 ,372 ,256 ,540 ***
biscuits) yoghurt -,917 ,179 ,000 ,400 ,281 ,568 ***
Policy (rc: control
group - no TFA policy1 (w ith TFA info) ,058 ,145 ,691 1,059 ,797 1,408
info)
Saturates = DK/healthy -,248 ,134 ,064 ,780 ,600 1,015
PHO = DK/healthy ,081 ,146 ,578 1,084 ,815 1,442
Aw areness Q41
TFA = DK/healthy ,207 ,144 ,151 1,231 ,927 1,633
(rc: unhealthy)
Sugar = DK/healthy -,505 ,227 ,026 ,604 ,387 ,941 *
Salt = DK/healthy -,099 ,143 ,490 ,906 ,684 1,200
Fat ,505 ,183 ,006 1,657 1,157 2,372 **
Salt 3,088 ,306 ,000 21,931 12,035 39,963 ***
Access Q32 (rc: Sugar 2,179 ,201 ,000 8,838 5,965 13,097 ***
no access) Nutrition fact panel ,926 ,158 ,000 2,525 1,851 3,444 ***
Ingredients list -,934 ,158 ,000 ,393 ,288 ,536 ***
Irrelevant labels or nothing -,845 ,154 ,000 ,430 ,318 ,581 ***
Observation short (< 15 seconds) ,136 ,159 ,392 1,146 ,839 1,564
duration Q4-9 long (60 seconds plus) -,062 ,164 ,705 ,940 ,682 1,295
Shelf behaviour reading at shelf before
-,151 ,159 ,344 ,860 ,630 1,175
Q10 (rc: grab&go) selection to buy
Fat -,036 ,238 ,880 ,965 ,605 1,539
Reading habits
Nutrition fact panel ,219 ,279 ,433 1,245 ,720 2,153
at shelf Q25/26
Sugar -,131 ,286 ,648 ,878 ,501 1,537
(rc: no reading)
Ingredients list -,009 ,208 ,967 ,991 ,660 1,490
Influenced by w as influenced ,008 ,205 ,969 1,008 ,674 1,508
shopping company
Q11/12 (rc: alone) not influenced ,140 ,173 ,418 1,150 ,820 1,614
Shopping fre- less often -,021 ,195 ,913 ,979 ,668 1,435
quency Q16 (rc:
every 2-3 w eeks) more often ,007 ,174 ,969 1,007 ,716 1,416
Planned to shop
not planned before -,159 ,144 ,270 ,853 ,643 1,131
before Q22 (rc:
Bought before
no, bought first time -,123 ,165 ,456 ,884 ,640 1,221
Q24 (rc: yes)

I look for food info … ,348 ,255 ,172 1,416 ,859 2,335
Main shopping
I w ant to get in and out ... as
goal Q28 (rc: not -,195 ,248 ,432 ,823 ,506 1,337
fast as I can
chosen)
I w ant a lot of variety … ,399 ,180 ,027 1,491 1,047 2,123 *
I like looking for new and … ,136 ,176 ,440 1,146 ,811 1,619
Relevance of health im pact = w eak -,376 ,169 ,026 ,686 ,493 ,955 *
health Q27 (rc: health impact = strong -,247 ,172 ,151 ,781 ,558 1,094
Relevance of convenience = w eak ,236 ,176 ,181 1,266 ,896 1,787
convenience
Q27 (rc: medium) convenience = strong ,084 ,164 ,607 1,088 ,789 1,499
Gender man -,266 ,149 ,073 ,766 ,572 1,025
Age (rc: 35-54 young (16-34 years) ,213 ,163 ,192 1,237 ,899 1,703
years) old (55+ years) -,114 ,182 ,532 ,893 ,625 1,275
hhsize (rc: 2 single -,580 ,201 ,004 ,560 ,378 ,830 **
persons) 3+ persons -,222 ,155 ,153 ,801 ,591 1,086
Occupation (rc: full-time (AC) ,029 ,178 ,869 1,030 ,727 1,458
part-time) w ork_other (EFGHJL) -,037 ,188 ,842 ,963 ,666 1,393
Education (rc: low _edu (AB) -,378 ,180 ,036 ,685 ,481 ,976 *
medium) high_edu (EFH) -,069 ,149 ,643 ,933 ,698 1,249
low _inc (up to 2000 €) -,202 ,185 ,275 ,817 ,569 1,174
Incom e (rc:
high_inc (3000 € plus) -,092 ,182 ,614 ,912 ,638 1,304
medium)
na_inc (no answ er) -,166 ,214 ,438 ,847 ,558 1,288
2
N = 1851 - Nagelkerke R = 0,502
Model 26 – Offline store experiments on TFA
BINOMINAL LOGISTIC REGRESSION - INITIAL CHOICE 2 - WITHOUT POLICY
Std. 95% conf. interval for Exp(B)
Initial choice 2 = healthier Product X B error Sig. Exp(B) lower bound upper bound
Intercept ,455 ,389 ,242
Context (rc: pizza -,988 ,191 ,000 ,372 ,256 ,541 ***
biscuits) yoghurt -,925 ,178 ,000 ,397 ,280 ,563 ***
Saturates = DK/healthy -,248 ,134 ,064 ,780 ,600 1,015
PHO = DK/healthy ,082 ,146 ,575 1,085 ,816 1,443
Aw areness Q41
TFA = DK/healthy ,207 ,144 ,151 1,230 ,927 1,633
(rc: unhealthy)
Sugar = DK/healthy -,510 ,227 ,024 ,601 ,385 ,937 *
Salt = DK/healthy -,098 ,143 ,495 ,907 ,685 1,201
Fat ,499 ,183 ,006 1,647 1,152 2,357 **
Salt 3,081 ,306 ,000 21,784 11,964 39,665 ***
Access Q32 (rc: Sugar 2,177 ,201 ,000 8,820 5,953 13,068 ***
no access) Nutrition fact panel ,920 ,157 ,000 2,508 1,842 3,416 ***
Ingredients list -,942 ,157 ,000 ,390 ,287 ,530 ***
Irrelevant labels or nothing -,841 ,154 ,000 ,431 ,319 ,583 ***
Observation short (< 15 seconds) ,131 ,158 ,408 1,140 ,836 1,555
duration Q4-9 long (60 seconds plus) -,059 ,163 ,720 ,943 ,685 1,299
Shelf behaviour reading at shelf before
-,148 ,159 ,351 ,862 ,631 1,177
Q10 (rc: grab&go) selection to buy
Fat -,030 ,238 ,899 ,970 ,609 1,547
Reading habits
Nutrition fact panel ,215 ,280 ,441 1,240 ,717 2,145
at shelf Q25/26
Sugar -,124 ,285 ,664 ,883 ,505 1,546
(rc: no reading)
Ingredients list -,005 ,208 ,981 ,995 ,663 1,495
Influenced by w as influenced ,007 ,205 ,974 1,007 ,673 1,506
shopping company
Q11/12 (rc: alone) not influenced ,141 ,173 ,415 1,151 ,821 1,616
Shopping fre- less often -,025 ,195 ,897 ,975 ,666 1,428
quency Q16 (rc:
every 2-3 w eeks) more often ,010 ,174 ,954 1,010 ,719 1,420
Planned to shop
not planned before -,158 ,144 ,272 ,854 ,644 1,132
before Q22 (rc:
Bought before
no, bought first time -,118 ,164 ,472 ,888 ,644 1,226
Q24 (rc: yes)
I look for food info … ,347 ,255 ,173 1,415 ,858 2,333
Main shopping
I w ant to get in and out ... as
goal Q28 (rc: not -,197 ,248 ,426 ,821 ,505 1,334
fast as I can
chosen)
I w ant a lot of variety … ,395 ,180 ,028 1,484 1,043 2,112 *
I like looking for new and … ,135 ,176 ,444 1,144 ,810 1,617
Relevance of health im pact = w eak -,378 ,169 ,025 ,685 ,492 ,953 *
health Q27 (rc: health impact = strong -,248 ,172 ,149 ,780 ,557 1,093
Relevance of convenience = w eak ,233 ,176 ,185 1,262 ,894 1,782
convenience
Q27 (rc: medium) convenience = strong ,084 ,164 ,606 1,088 ,790 1,500
Gender man -,263 ,149 ,076 ,769 ,574 1,028
Age (rc: 35-54 young (16-34 years) ,214 ,163 ,189 1,239 ,900 1,705
years) old (55+ years) -,113 ,182 ,536 ,893 ,625 1,276
hhsize (rc: 2 single -,581 ,201 ,004 ,559 ,378 ,829 **
persons) 3+ persons -,223 ,155 ,150 ,800 ,590 1,084
Occupation (rc: full-time (AC) ,025 ,177 ,889 1,025 ,724 1,450
part-time) w ork_other (EFGHJL) -,042 ,188 ,825 ,959 ,664 1,386
Education (rc: low _edu (AB) -,381 ,180 ,034 ,683 ,480 ,972 *
medium) high_edu (EFH) -,067 ,148 ,652 ,935 ,699 1,251
low _inc (up to 2000 €) -,204 ,185 ,270 ,816 ,568 1,171
Incom e (rc:
high_inc (3000 € plus) -,091 ,182 ,619 ,913 ,639 1,305
medium)
na_inc (no answ er) -,169 ,213 ,427 ,844 ,556 1,282
2
N = 1851 - Nagelkerke R = 0,501
Model 27 – Offline store experiments on TFA
BINOMINAL LOGISTIC REGRESSION WITHOUT INTERACTIONS
- HEALTHY CHOICE 1 - WITH POLICIES
Std. 95% conf. interval for Exp(B)
Healthy choice 1 = Product B B error Sig. Exp(B) lower bound upper bound
Intercept -,771 ,346 ,026
Context (rc: pizza ,175 ,145 ,227 1,192 ,897 1,584
biscuits) yoghurt 1,505 ,163 ,000 4,502 3,272 6,195 ***
TFA info w ithout education ,275 ,132 ,037 1,316 1,017 1,703 *
Policy (rc: control TFA info w ith short
,421 ,173 ,015 1,523 1,085 2,138 *
group - no TFA education (< 45 seconds)
info) TFA info w ith long
,837 ,162 ,000 2,310 1,682 3,173 ***
education (45+ seconds)
Saturates = DK/healthy -,204 ,110 ,063 ,815 ,658 1,011
PHO = DK/healthy -,276 ,120 ,022 ,759 ,599 ,961 *
Aw areness Q41
TFA = DK/healthy -,189 ,116 ,104 ,828 ,659 1,039
(rc: unhealthy)
Sugar = DK/healthy -,066 ,177 ,707 ,936 ,662 1,323
Salt = DK/healthy ,053 ,119 ,657 1,054 ,835 1,331
Education ,194 ,292 ,506 1,214 ,685 2,153
Access Q53 (rc: Ingredients list -,331 ,131 ,012 ,718 ,555 ,929 *
no access) Nutrition fact panel ,558 ,159 ,000 1,747 1,280 2,386 ***
Irrelevant areas or none -,632 ,165 ,000 ,532 ,385 ,734 ***
Observation short (< 15 seconds) -,064 ,132 ,627 ,938 ,724 1,215
duration Q4-9 long (60 seconds plus) -,165 ,135 ,221 ,848 ,651 1,104
Shelf behaviour
reading at shelf before
Q10 (rc: grab & ,165 ,130 ,202 1,180 ,915 1,520
selection to buy
go)
Fat ,160 ,197 ,416 1,174 ,798 1,727
Reading habits
Nutrition fact panel -,237 ,246 ,334 ,789 ,488 1,277
at shelf Q25/26
Sugar -,190 ,236 ,420 ,827 ,521 1,312
(rc: no reading)
Ingredients list ,140 ,166 ,398 1,150 ,831 1,592
Influenced by w as influenced ,089 ,166 ,592 1,093 ,790 1,513
shopping company
Q11/12 (rc: alone) not influenced -,044 ,142 ,755 ,957 ,724 1,264
Shopping fre- less often ,127 ,154 ,408 1,136 ,840 1,536
quency Q16 (rc:
every 2-3 w eeks) more often -,149 ,141 ,290 ,862 ,654 1,135
Planned to shop
before Q22 (rc: not planned before ,125 ,116 ,284 1,133 ,902 1,423
yes)
Bought before
no, bought first time ,063 ,132 ,631 1,066 ,823 1,380
Q24 (rc: yes)
I look for food info … ,054 ,212 ,797 1,056 ,697 1,599
Main shopping I w ant to get in and out ... as
goal Q28 (rc: not ,025 ,194 ,898 1,025 ,702 1,498
fast as I can
chosen) I w ant a lot of variety … ,022 ,146 ,882 1,022 ,767 1,361
I like looking for new and … -,146 ,141 ,301 ,865 ,656 1,139
Relevance of health impact = w eak -,206 ,136 ,128 ,813 ,623 1,061
health Q27 (rc: health impact = strong ,010 ,140 ,946 1,010 ,767 1,330
Relevance of convenience = w eak -,447 ,146 ,002 ,639 ,480 ,852 **
convenience
Q27 (rc: medium) convenience = strong -,319 ,132 ,016 ,727 ,561 ,942 *
Gender m an -,243 ,119 ,041 ,784 ,621 ,990 *
Age (rc: 35-54 young (16-34 years) ,225 ,133 ,091 1,252 ,965 1,625
years) old (55+ years) ,020 ,148 ,892 1,020 ,763 1,365
hhsize (rc: 2 single ,202 ,156 ,195 1,224 ,901 1,662
persons) 3+ persons -,057 ,127 ,656 ,945 ,736 1,213
Occupation (rc: full-time (AC) -,057 ,146 ,699 ,945 ,710 1,258
part-time) w ork_other (EFGHJL) ,140 ,157 ,372 1,150 ,846 1,564
Education (rc: low _edu (AB) ,015 ,145 ,917 1,015 ,764 1,348
medium) high_edu (EFH) ,083 ,121 ,495 1,086 ,857 1,377
low _inc (up to 2000 €) -,120 ,148 ,416 ,887 ,664 1,184
Incom e (rc:
high_inc (3000 € plus) ,196 ,147 ,182 1,217 ,912 1,622
medium)
na_inc (no answ er) ,022 ,178 ,902 1,022 ,722 1,448
2
N = 1851 - Nagelkerke R = 0,183
Model 28 – Offline store experiments on TFA
BINOMINAL LOGISTIC REGRESSION WITHOUT INTERACTIONS
- HEALTHY CHOICE 1 - WITHOUT POLICIES
Std. 95% conf. interval for Exp(B)
Healthy choice 1 = Product B B error Sig. Exp(B) lower bound upper bound
Intercept -,360 ,326 ,269
Context (rc: pizza ,183 ,143 ,201 1,201 ,907 1,589
biscuits) yoghurt 1,448 ,161 ,000 4,255 3,105 5,830 ***
Saturates = DK/healthy -,205 ,109 ,060 ,815 ,658 1,009
PHO = DK/healthy -,256 ,119 ,032 ,774 ,613 ,978 *
Aw areness Q41
TFA = DK/healthy -,202 ,115 ,079 ,817 ,652 1,024
(rc: unhealthy)
Sugar = DK/healthy -,109 ,175 ,535 ,897 ,637 1,264
Salt = DK/healthy ,074 ,118 ,530 1,077 ,855 1,357
Education ,580 ,280 ,038 1,786 1,032 3,090 *
Access Q53 (rc: Ingredients list -,393 ,128 ,002 ,675 ,525 ,868 **
no access) Nutrition fact panel ,503 ,156 ,001 1,654 1,219 2,244 **
Irrelevant areas or none -,655 ,162 ,000 ,520 ,378 ,714 ***
Observation short (< 15 seconds) -,130 ,131 ,320 ,878 ,680 1,134
duration Q4-9 long (60 seconds plus) -,108 ,133 ,414 ,897 ,692 1,164
Shelf behaviour
reading at shelf before
Q10 (rc: grab & ,192 ,128 ,131 1,212 ,944 1,556
selection to buy
go)
Fat ,217 ,194 ,264 1,242 ,849 1,817
Reading habits
Nutrition fact panel -,269 ,241 ,264 ,764 ,476 1,225
at shelf Q25/26
Sugar -,124 ,231 ,593 ,884 ,562 1,390
(rc: no reading)
Ingredients list ,185 ,164 ,260 1,203 ,872 1,658
Influenced by w as influenced ,091 ,164 ,578 1,095 ,795 1,510
shopping company
Q11/12 (rc: alone) not influenced -,059 ,141 ,673 ,942 ,715 1,241
Shopping fre- less often ,108 ,153 ,480 1,114 ,826 1,502
quency Q16 (rc:
every 2-3 w eeks) more often -,126 ,140 ,368 ,882 ,671 1,160
Planned to shop
before Q22 (rc: not planned before ,145 ,116 ,210 1,156 ,921 1,450
yes)
Bought before
no, bought first time ,094 ,131 ,472 1,099 ,850 1,420
Q24 (rc: yes)
I look for food info … ,050 ,208 ,809 1,052 ,699 1,582
Main shopping
I w ant to get in and out ... as
goal Q28 (rc: not ,016 ,191 ,935 1,016 ,698 1,477
fast as I can
chosen)
I w ant a lot of variety … -,028 ,145 ,847 ,973 ,732 1,291
I like looking for new and … -,159 ,140 ,254 ,853 ,649 1,121
Relevance of health impact = w eak -,247 ,135 ,066 ,781 ,600 1,017
health Q27 (rc: health impact = strong -,006 ,139 ,968 ,995 ,757 1,306
Relevance of convenience = w eak -,478 ,145 ,001 ,620 ,467 ,824 **
convenience
Q27 (rc: medium) convenience = strong -,337 ,131 ,010 ,714 ,553 ,923 *
Gender man -,229 ,118 ,052 ,795 ,631 1,002
Age (rc: 35-54 young (16-34 years) ,235 ,132 ,076 1,264 ,976 1,638
years) old (55+ years) ,033 ,147 ,824 1,033 ,774 1,379
hhsize (rc: 2 single ,199 ,155 ,199 1,220 ,901 1,653
persons) 3+ persons -,051 ,126 ,688 ,951 ,743 1,216
Occupation (rc: full-time (AC) -,102 ,144 ,480 ,903 ,681 1,198
part-time) w ork_other (EFGHJL) ,099 ,155 ,521 1,104 ,815 1,496
Education (rc: low _edu (AB) ,009 ,143 ,948 1,009 ,762 1,337
medium) high_edu (EFH) ,094 ,120 ,435 1,098 ,868 1,389
low _inc (up to 2000 €) -,136 ,146 ,351 ,872 ,655 1,162
Incom e (rc:
high_inc (3000 € plus) ,228 ,146 ,118 1,256 ,944 1,670
medium)
na_inc (no answ er) -,006 ,175 ,975 ,994 ,706 1,401
2
N = 1851 - Nagelkerke R = 0,165
Model 29 – Offline store experiments on TFA

BINOMINAL LOGISTIC REGRESSION WITHOUT INTERACTIONS

- HEALTHY CHOICE 2 - WITH POLICIES
Std. 95% conf. interval for Exp(B)
a
Healthy choice 2 = Product X B error Sig. Exp(B) lower bound upper bound
Intercept -,355 ,404 ,379
Context (rc: pizza -,647 ,171 ,000 ,523 ,375 ,731 ***
biscuits) yoghurt -1,123 ,177 ,000 ,325 ,230 ,460 ***
TFA info w ithout education ,086 ,150 ,567 1,090 ,811 1,464
Policy (rc: control TFA info w ith short
-,408 ,198 ,039 ,665 ,451 ,980 *
group - no TFA education (< 45 seconds)
info) TFA info w ith long education
,170 ,188 ,367 1,185 ,819 1,715
(45+ seconds)
Saturates = DK/healthy ,034 ,127 ,788 1,035 ,807 1,328
PHO = DK/healthy ,139 ,139 ,316 1,149 ,875 1,509
Aw areness Q41
TFA = DK/healthy ,012 ,136 ,927 1,012 ,776 1,321
(rc: unhealthy)
Sugar = DK/healthy -,410 ,200 ,041 ,664 ,448 ,984 *
Salt = DK/healthy -,151 ,137 ,272 ,860 ,657 1,126
Education ,021 ,333 ,949 1,021 ,532 1,961
Access Q48 (rc: Ingredients list -,417 ,140 ,003 ,659 ,501 ,868 **
no access) Nutrition fact panel 2,334 ,180 ,000 10,315 7,246 14,683 ***
Irrelevant areas or none -,879 ,176 ,000 ,415 ,294 ,586 ***
Observation short (< 15 seconds) ,034 ,151 ,820 1,035 ,770 1,392
duration Q4-9 long (60 seconds plus) ,043 ,158 ,786 1,044 ,766 1,423
Shelf behaviour
reading at shelf before
Q10 (rc: grab & -,130 ,151 ,390 ,878 ,652 1,181
selection to buy
go)
Fat -,143 ,230 ,533 ,866 ,552 1,360
Reading habits
Nutrition fact panel -,130 ,273 ,634 ,878 ,514 1,499
at shelf Q25/26
Sugar ,308 ,275 ,263 1,361 ,793 2,334
(rc: no reading)
Ingredients list -,002 ,196 ,990 ,998 ,679 1,466
Influenced by w as influenced -,063 ,193 ,744 ,939 ,643 1,371
shopping company
Q11/12 (rc: alone) not influenced ,271 ,167 ,105 1,311 ,945 1,818
Shopping fre- less often -,167 ,177 ,345 ,846 ,598 1,197
quency Q16 (rc:
every 2-3 w eeks) more often ,047 ,165 ,776 1,048 ,758 1,448
Planned to shop
before Q22 (rc: not planned before -,022 ,135 ,870 ,978 ,750 1,275
yes)
Bought before
no, bought first time ,188 ,156 ,229 1,207 ,888 1,640
Q24 (rc: yes)
I look for food info … -,086 ,243 ,725 ,918 ,570 1,478
Main shopping
I w ant to get in and out ... as
goal Q28 (rc: not -,378 ,220 ,085 ,685 ,446 1,054
fast as I can
chosen)
I w ant a lot of variety … -,149 ,172 ,387 ,862 ,615 1,207
I like looking for new and … -,236 ,166 ,153 ,790 ,571 1,092
Relevance of health impact = w eak -,297 ,161 ,065 ,743 ,542 1,018
health Q27 (rc: health impact = strong -,157 ,167 ,347 ,855 ,617 1,185
Relevance of convenience = w eak ,122 ,170 ,472 1,130 ,810 1,576
convenience
Q27 (rc: medium) convenience = strong -,073 ,155 ,635 ,929 ,686 1,258
Gender m an -,271 ,138 ,049 ,762 ,582 ,999 *
Age (rc: 35-54 young (16-34 years) ,240 ,153 ,118 1,271 ,941 1,717
years) old (55+ years) -,019 ,170 ,909 ,981 ,703 1,368
hhsize (rc: 2 single -,275 ,182 ,130 ,760 ,532 1,084
persons) 3+ persons -,071 ,147 ,628 ,931 ,698 1,242
Occupation (rc: full-time (AC) ,244 ,171 ,154 1,276 ,912 1,784
part-time) w ork_other (EFGHJL) -,148 ,179 ,408 ,862 ,607 1,225
Education (rc: low _edu (AB) ,224 ,168 ,182 1,251 ,900 1,740
medium) high_edu (EFH) ,250 ,143 ,080 1,284 ,971 1,698
low _inc (up to 2000 €) ,027 ,170 ,876 1,027 ,735 1,435
Incom e (rc:
high_inc (3000 € plus) ,246 ,175 ,160 1,279 ,907 1,803
medium)
na_inc (no answ er) -,285 ,203 ,161 ,752 ,505 1,120
2
N = 1851 - Nagelkerke R = 0,464

Model 30 – Offline store experiments on TFA

BINOMINAL LOGISTIC REGRESSION WITHOUT INTERACTIONS
- HEALTHY CHOICE 2 - WITHOUT POLICIES
Std. 95% conf. interval for Exp(B)
a
Healthy choice 2 = Product X B error Sig. Exp(B) lower bound upper bound
Intercept -,383 ,387 ,321
Context (rc: pizza -,696 ,169 ,000 ,498 ,358 ,694 ***
biscuits) yoghurt -1,142 ,176 ,000 ,319 ,226 ,451 ***
Saturates = DK/healthy ,037 ,127 ,769 1,038 ,810 1,331
PHO = DK/healthy ,127 ,139 ,359 1,136 ,865 1,491
Aw areness Q41
TFA = DK/healthy ,022 ,135 ,872 1,022 ,784 1,332
(rc: unhealthy)
Sugar = DK/healthy -,417 ,201 ,037 ,659 ,445 ,976 *
Salt = DK/healthy -,162 ,137 ,237 ,851 ,651 1,112
Education ,038 ,318 ,906 1,038 ,557 1,936
Access Q48 (rc: Ingredients list -,408 ,138 ,003 ,665 ,508 ,870 **
no access) Nutrition fact panel 2,360 ,179 ,000 10,594 7,461 15,042 ***
Irrelevant areas or none -,872 ,175 ,000 ,418 ,297 ,590 ***
Observation short (< 15 seconds) ,033 ,150 ,827 1,033 ,770 1,387
duration Q4-9 long (60 seconds plus) ,067 ,157 ,669 1,069 ,786 1,455
Shelf behaviour
reading at shelf before
Q10 (rc: grab & -,153 ,150 ,308 ,858 ,639 1,152
selection to buy
go)
Fat -,152 ,228 ,505 ,859 ,549 1,343
Reading habits
Nutrition fact panel -,163 ,271 ,547 ,850 ,500 1,444
at shelf Q25/26
Sugar ,317 ,274 ,248 1,373 ,802 2,349
(rc: no reading)
Ingredients list ,003 ,195 ,989 1,003 ,684 1,470
Influenced by w as influenced -,088 ,192 ,649 ,916 ,628 1,336
shopping company
Q11/12 (rc: alone) not influenced ,289 ,166 ,082 1,335 ,964 1,850

Shopping fre- less often -,147 ,177 ,404 ,863 ,610 1,220
quency Q16 (rc:
every 2-3 w eeks) more often ,049 ,164 ,763 1,051 ,762 1,448

Planned to shop
before Q22 (rc: not planned before -,030 ,135 ,823 ,970 ,745 1,263
yes)
Bought before
no, bought first time ,198 ,156 ,204 1,219 ,898 1,655
Q24 (rc: yes)
I look for food info … -,053 ,242 ,828 ,949 ,591 1,524
Main shopping
I w ant to get in and out ... as
goal Q28 (rc: not -,371 ,218 ,089 ,690 ,450 1,059
fast as I can
chosen)
I w ant a lot of variety … -,134 ,171 ,431 ,874 ,626 1,222
I like looking for new and … -,215 ,165 ,193 ,807 ,584 1,114
Relevance of health impact = w eak -,299 ,160 ,062 ,742 ,542 1,015
health Q27 (rc:
medium) health impact = strong -,166 ,167 ,320 ,847 ,611 1,174
Relevance of convenience = w eak ,134 ,169 ,428 1,143 ,821 1,592
convenience
Q27 (rc: medium) convenience = strong -,051 ,154 ,739 ,950 ,703 1,284
Gender man -,261 ,137 ,057 ,770 ,588 1,008
Age (rc: 35-54 young (16-34 years) ,244 ,153 ,110 1,277 ,946 1,723
years) old (55+ years) ,006 ,169 ,974 1,006 ,721 1,402
hhsize (rc: 2 single -,295 ,181 ,102 ,745 ,523 1,061
persons) 3+ persons -,081 ,146 ,577 ,922 ,692 1,227
Occupation (rc: full-time (AC) ,262 ,170 ,123 1,299 ,931 1,812
part-time) w ork_other (EFGHJL) -,153 ,178 ,390 ,858 ,605 1,217
Education (rc: low _edu (AB) ,234 ,167 ,162 1,263 ,910 1,754
medium) high_edu (EFH) ,221 ,142 ,118 1,248 ,945 1,647
low _inc (up to 2000 €) ,025 ,170 ,883 1,025 ,735 1,430
Incom e (rc:
high_inc (3000 € plus) ,232 ,174 ,183 1,261 ,896 1,774
medium)
na_inc (no answ er) -,273 ,202 ,177 ,761 ,512 1,131
2
N = 1851 - Nagelkerke R = 0,460