OECD Regional Development Papers No.

66

Migration and regional

OECD
innovation in Australia
https://dx.doi.org/10.1787/6d6ff472-en
 OECD Regional Development Papers

Migration and regional innovation in

Australia

This paper provides evidence on the impact of international migrants on

regional innovation. The study combines administrative individual-level
data covering all Australian residents with data on intellectual property
rights applications such as patents, trademarks, and design rights. The
analysis uses a standard shift-share instrument based on past migrant
settlements to identify the causal effects of migration on innovation. Its
four main findings are the following: First, on average, a one percentage
point increase in the regional employment share of higher-educated
migrants relative to total employment leads to a 4.8% rise in regional
patent applications in the medium run (five years). Second, while migrants
of all skill and education levels have a positive impact on patenting, those
in scientific occupations have the largest effect. Third, regions with lower
levels of patenting benefit relatively more from increases in migration
compared to those with higher patenting levels. Fourth, there is no effect
of migration on trademarks or design rights applications.

JEL codes: R10, O34, R23

Keywords: Australia, innovation, regions

Contact:
Cem Özgüzel- [email protected]
Jasper Hesse- [email protected]

PUBE
2

ABOUT THE OECD

The OECD is a multi-disciplinary inter-governmental organisation with member countries which engages
in its work an increasing number of non-members from all regions of the world. The Organisation’s core
mission today is to help governments work together towards a stronger, cleaner, fairer global economy.
Through its network of specialised committees and working groups, the OECD provides a setting where
governments compare policy experiences, seek answers to common problems, identify good practice, and
co-ordinate domestic and international policies. More information available: www.oecd.org.

ABOUT OECD REGIONAL DEVELOPMENT PAPERS

Papers from the Centre for Entrepreneurship, SMEs, Regions and Cities of the OECD cover a full range
of topics including regional statistics and analysis, urban governance and economics, rural governance
and economics, and multi-level governance. Depending on the programme of work, the papers can cover
specific topics such as regional innovation and networks, sustainable development, the determinants of
regional growth or fiscal consolidation at the subnational level. OECD Regional Development Papers are
published on http://www.oecd.org/cfe/regional-policy.
OECD Working Papers should not be reported as representing the official views of the OECD or of its
member countries. The opinions expressed and arguments employed are those of the author(s).
Working Papers describe preliminary results or research in progress by the author(s) and are published to
stimulate discussion on a broad range of issues on which the OECD works. Comments on Working Papers
are welcome, and may be sent to the Centre for Entrepreneurship, SMEs, Regions and Cities, OECD, 2
rue André-Pascal, 75775 Paris Cedex 16, France.
This paper is authorised for publication by Lamia Kamal-Chaoui, Director, Centre for Entrepreneurship,
SMEs, Regions and Cities, OECD.
This document, as well as any statistical data and map included herein, are without prejudice to the status
of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the
name of any territory, city or area.
The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities.
The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem
and Israeli settlements in the West Bank under the terms of international law.
Note by the Republic of Türkiye: The information in this document with reference to “Cyprus” relates to the
southern part of the Island. There is no single authority representing both Turkish and Greek Cypriot people
on the Island. Türkiye recognises the Turkish Republic of Northern Cyprus (TRNC). Until a lasting and
equitable solution is found within the context of the United Nations, Türkiye shall preserve its position
concerning the “Cyprus issue”.
Note by all the European Union Member States of the OECD and the European Union: The Republic of
Cyprus is recognised by all members of the United Nations with the exception of Türkiye. The information
in this document relates to the area under the effective control of the Government of the Republic of
Cyprus.

© OECD 2024
The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at
https://www.oecd.org/termsandconditions.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

 3

Executive summary

Migrants are an integral part of the Australian population. In 2021, almost three out of ten Australian
residents (29%) were born abroad, positioning Australia as the country with the third-highest migrant share
among OECD countries. On average, foreign-born residents in Australia are highly educated and well-
integrated into the labour market. Their level of education exceeds the average educational attainment of
the native population.
The arrival of higher-educated migrants into the labour supply of Australian regions increases
patenting activity. On average, a one percentage point increase in the regional employment share of
higher-educated migrants relative to total employment leads to a 4.8% rise in regional patent applications
in the medium run (five years). In contrast, the inflow of higher-educated migrants does not affect other
innovation measures, such as trademarks and design rights.
The positive influence on innovation is most significant for migrants in scientific occupations and
previously less innovative regions. Although the inflow of migrants across all education levels positively
impacts patent applications, those in scientific occupations have the most pronounced effect. Furthermore,
while migration positively affects all regions in Australia, the benefits are most pronounced in less
innovative regions.
The overall positive impact of migration on regional innovation reaffirms the significant
contribution of migrants to the Australian economy. The analysis suggests that the inflow of higher-
educated and higher-skilled migrants, facilitated by the selective migration policies, positively influenced
Australia’s innovation activities.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

4

Acknowledgements

This paper was prepared by the OECD Centre for Entrepreneurship, SMEs, Regions and Cities (CFE) led
by Lamia Kamal-Chaoui, Director. The work was conducted as part of the OECD Regional Development
Policy Committee’s program of work with financial support from the Centre for Population at the Australian
Treasury.
Cem Özgüzel, Economist at CFE, coordinated the preparation of the paper under the guidance of Ana
Moreno-Monroy, Head of the Statistics and Territorial Analysis Unit (CFE) and the supervision of Rüdiger
Ahrend, Head of the Economic Analysis, Data and Statistics Division at CFE. Gabriel Chaves Bosch,
Jasper Hesse, and Cem Özgüzel (all CFE) drafted the paper. The authors are thankful to Riccardo
Crescenzi (London School of Economics) and Giovanni Peri (University of California, Davis) for helpful
feedback. Thanks are also due to Nadim Ahmad, Lars Ludolph and Michelle Marshallian (all CFE), Ana
Damas de Matos (OECD Directorate for Employment, Labour and Social Affairs, ELS), and Ben Westmore
(OECD Economics Department, ECO) for their valuable comments. Eric Gonnard (CFE) provided
statistical support. Pilar Philip (CFE) prepared the paper for publication.
The paper also benefited from valuable comments by Patrick Fazzone IV, Rodrigo Rodrigues, Ian South,
and Kasey Stanfield (all Centre for Population), Joe Castellino, Kathleen Cross, Linda Velzeboer
(Department of Infrastructure, Transport, Regional Development, Communications and the Arts). Anna
Maria Mayda (Georgetown University), Ceren Özgen (University of Birmingham) and Riccardo Turati
(Autonomous University of Barcelona) provided guidance throughout the project as scientific advisors. The
report benefited from analytical input by Aneeq Sarwar (Economic Society Australia). The Secretariat
appreciates the feedback provided by the delegates during the 45 th session of the Working Party on
Territorial Indicators (WPTI) on 14 November 2023.
Michel Beine (University of Luxembourg) and Christopher Parsons (University of Western Australia) kindly
shared valuable data on the historical settlement of migrants across Australia. IP Australia provided data
on intellectual property rights applications, and the Australian Bureau of Statistics (ABS) provided access
to the Multi-Agency Data Integration Project (MADIP), which made the analysis possible.
This paper is the fourth output of a multi-annual collaboration between the Australian Centre for Population
and the OECD. The first working paper of the project provides a detailed descriptive overview of migrants
in Australia and the context of productivity and the labour market (OECD, 2023[1]). The second examines
the contribution of migrants to regional labour productivity differences in Australia (OECD, 2023[2]). The
third paper evaluates the impact of migration on regional labour markets (OECD, 2023[3]).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

 5

Table of contents

Executive summary 3
Acknowledgements 4
1 Introduction 7
2 Literature 9
3 Data and preliminary evidence 11
Data sources and sample construction 11

4 Empirical strategy 17
Empirical model 17
Endogeneity of the migrant share 18
Validity of the instrumental variable strategy 19

5 Results 20
Average effect of migration on innovation across regions 20
Uneven effects across workers and places 22

6 Concluding remarks 25
References 26
Annex A. Data sources 29
Annex B. Construction and validity of the instrumental variable 33
Annex C. Robustness checks 38
Annex D. Definitions of higher-skilled and higher-educated migrants 41
Annex E. Additional results 42

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

6

FIGURES
Figure 1. Patent applications and labour productivity 10
Figure 2. The presence of migrants varies across Australian labour markets 13
Figure 3. Geographical distribution of patenting activity across Australian regions 14
Figure 4. Correlations between changes in IP rights applications and higher-educated migrant shares 15

TABLES
Table 1. Descriptive statistics (unweighted) 16
Table 2. The average effect of migration on regional innovation 21
Table 3. Uneven effects of migrants with different skills and education backgrounds 22
Table 4. Uneven effects across regions with different characteristics 24
Table 5. Migrant decomposition in terms of country of origin 31
Table 6. Explanatory variables in 1981 34
Table 7. Controlling for past migration flows 36
Table 8. Impact of the instrumental variable on past IP rights applications 37
Table 9. Robustness of the main results 39
Table 10. Additional results for uneven effects of migrants with different education levels 42

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

 7

1 Introduction
With 29% of the population born abroad as of 2021 (OECD, 2022[4]), Australia has the third-highest migrant
share among OECD countries.1 Migrants can contribute to national economic growth through various
channels, for instance, by bringing new skills and ideas, as well as fostering innovation. Yet, evidence on
the potential contribution of migration to innovation remains scant.

This paper measures the impact of migration on regional innovation across Australian Statistical Area 4
(SA4) regions, focusing on higher-educated migrants who are most likely to innovate (Hunt, 2011[5]; Hunt
and Gauthier-Loiselle, 2010[6]; Bernstein et al., 2022[7]).23 It uses comprehensive administrative individual-
level data covering all Australian residents and detailed information on various measures of innovation to
assess the causal influence of higher-educated migrants on innovation across Australian regions from
2011 to 2018. The analysis employs a difference-in-difference design comparing innovative activities in
regions that received more higher-educated migrants with those that received less higher-educated
migrants before and after the arrival of the migrants. As migrants often select their location based on
economic opportunities, which can be linked to innovative activities, the study employs an instrumental
variable strategy to identify the causal effect of migration on regional innovation.
This paper uses patents, trademarks, and design rights applications as innovation indicators to provide
estimates that align with the current migration literature.4 Innovative activities – which are multifaceted and
intricate - tend to cluster geographically, resulting in significant disparities among regions within countries.
It is thus essential to employ diverse measures of intellectual property (IP) to capture various aspects of
the broader concept of innovation instead of relying on patents that, although widely used in the literature
(Ozgen, 2021[8]), may lead to skewed representations as they are concentrated in large cities. Unlike
patents, which typically capture innovations in STEM industries, trademarks typically capture innovations

1
The terms “migrants” and “foreign-born” are used interchangeably throughout this paper. Individuals born outside of
their country of residence are considered migrants. Unlike citizenship, this criterion does not change over time, it is
not subject to country differences in legislation, and it is thus adequate for international comparisons. In Australia,
migration is primarily measured as Net Oversea Migration (NOM), which refers to the net increase or decrease in the
Australian population resulting from immigration to and emigration from Australia, irrespective of the individual’s
country of birth or nationality.
2
The term “higher-educated” describe individuals who have obtained at least a college degree. Annex D provides
details on the definition.
3
The Australian Bureau of Statistics (ABS) has designed Statistical Areas 4 (SA4) regions by considering a range of
criteria that balance various factors. The two main criteria are population size and commuting patterns. As a result,
the 88 SA4 regions considered in this analysis cover the whole of Australia and represent all regional labour markets.
SA4 regions exhibit a functional characteristic in terms of capturing labour supply and demand. Throughout this paper,
the term “region” refers to SA4 regions unless indicated otherwise.
4
The analysis follows previous literature (Hunt and Gauthier-Loiselle, 2010[6]) and uses applications rather than
granted IP rights because the most immediate step after development of an innovation, typically, is to apply for its IP
right.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

8

in the retail or services sector, while design rights are more relevant in manufacturing.5 Jointly analysing
these measures provides a more comprehensive view of regional innovation activities in Australia.6
The paper makes four findings:

• Higher-educated migrants have a positive effect on patent applications across Australian

regions. Concretely, a one percentage point increase in the regional employment of higher-
educated migrants relative to total employment leads, on average, to a 4.8% rise in regional patent
applications in the medium run (five years).
• All migrant groups contribute to the positive effect on patent applications. Migrants of all
education levels have a positive impact on patent applications. While the effect is smaller for non-
higher-educated migrants, it is most pronounced for higher-educated migrants, particularly those
in scientific occupations.
• The positive impact of migrants on patenting activity is more significant in regions that
initially had lower levels of innovation. Although all regions benefit from the arrival of higher-
educated migrants, the increase in patenting activity is notably larger in regions with initially lower
levels of innovation. Moreover, migration benefits all regions irrespective of their regional income
level or population density.
• Migrants do not affect trademarks or design rights applications. 7 The analysis finds no
statistically significant regional effect of migration on trademarks or design rights applications.
The paper is organized as follows. The next section presents a brief review of the relevant migration
literature. Section 3 presents the data used in the analysis and provides descriptive statistics. Section 4
introduces the empirical strategy, Section 5 presents the results, and Section 6 concludes.

5 Patents and design rights represent novel innovations introduced to the world, while trademarks register innovation
that is new to the local market or individual firms (Jensen and Webster, 2009[21]).
6
To ensure comparability with previous research (Crown, Faggian and Corcoran, 2020[18]; Blit, Skuterud and Zhang,
2020[23]; Hunt and Gauthier-Loiselle, 2010[6]) and due to lack of data at the regional level, this paper does not follow
the OSLO definition of innovation developed by the OECD and Eurostat (OECD/Eurostat, 2018[41]).
7
This finding is consistent with previous research finding migrant-owned firms are more innovative in terms of more
R&D spending and patenting activity but no are not more likely to apply for trademarks or copyrights in the US (Brown
et al., 2020[29]).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

 9

2 Literature
Migration can boost innovation through multiple channels. A recent study finds that migrants in the US
generate substantial positive externalities on innovation. While only 16% of all innovators in the US are
foreign-born, they contribute to 23% of the national innovation output (Bernstein et al., 2022[7]). Yet, beyond
their direct impact on innovation, migrants can also indirectly influence innovation by complementing the
efforts of individuals and firms in their host region through increased diversity and positive externalities
(Ozgen, 2021[8]; Perez-Silva, Partridge and Foster, 2019[9]). However, the impact of migration on innovation
might vary across countries as higher-skilled migrants settle unevenly (Kerr et al., 2016[10]).
Extensive evidence from OECD countries suggests that the impact of migration on regional innovation (like
the native population) depends on the characteristics of the migrant population, particularly education
composition. Recent studies find that an increase in the presence of higher-educated migrants has a strong
and positive effect on regional innovation in US counties (Hunt and Gauthier-Loiselle, 2010[6]), English
local authorities (Gagliardi, 2015[11]), French districts (Mayda, Orefice and Santoni, 2022[12]), and, more
generally, European countries (Bosetti, Cattaneo and Verdolini, 2015[13]). However, when assessing the
effect of an overall increase in the number of migrants, existing evidence is more nuanced, with studies
finding no effect in Italian provinces (Bratti and Conti, 2018[14]) or an effect limited to already innovative
municipalities across OECD countries (OECD, 2022[15]). Moreover, migration can also reduce patenting
activities if firms shift towards labour-intensive activities and reduce their investment in capital following
the arrival of lower-skilled migrants in the region (Imbert et al., 2022[16]).8 Increased innovation is also
closely linked to labour productivity, as Box 2.1 illustrates.
Less is known about the impact of migrants on innovation in Australia, as findings from other countries
may not hold for Australia given differences in industrial composition and national institutional settings, as
well as characteristics and presence of migrants. For instance, recent evidence shows that university-
educated migrant men from the same country have higher expected employment rates and weekly
earnings in the US than in Australia (Clarke, Ferrer and Skuterud, 2019[17]), which might incentivise most
innovative migrants to choose the US over Australia. Furthermore, it remains unclear to what extent the
Australian context can effectively leverage the expertise brought by migrants to support innovation.
Existing evidence from Australia shows that foreign-born graduates have a positive impact on patents but
not on trademarks or design rights (Crown, Faggian and Corcoran, 2020[18]). However, this effect might
not be translatable to other migrant groups that attained their education abroad. Hence, the overall impact
of arriving migrants with different education levels on innovative activities in Australia remains an open
question (Jensen, 2014[19]). Lastly, it is uncertain whether potential positive effects of migrants depend on
specific regional characteristics or apply to all regions.
This study looks to fill that gap and provide evidence at the regional level. It utilises individual-level data
covering all Australian residents and intellectual property rights information to provide a precise analysis
of the regional impact of migration and a nuanced understanding of the contribution of migrants based on

8
These findings are obtained in the context of Chinese prefectures and following an episode of large-scale internal
rural-to-urban migration. From a theoretical perspective, a similarly large-scale international migration inflow can also
have similar effects on the economy.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

10 

their educational levels or occupations. Additionally, it sheds light on uneven effects across regions, a
factor often overlooked in such analyses.

Box 2.1. Does innovation affect labour productivity?

Innovation affects productivity in various ways. On the one hand, it can increase the efficiency of
workers, also known as labour productivity (LP) (Sarwar, 2022[20]). On the other hand, innovation can
improve the output of capital per unit of labour, leading to an increase in total factor productivity (TFP).
Therefore, increasing innovation is key to boosting both LP and TFP.
In fact, innovation, proxied by patent applications, is positively associated with labour productivity. Panel
A of Figure 1 illustrates the positive relationship between labour productivity (vertical axis) and patent
applications (horizontal axis) in other OECD countries (blue markers) and Australia (red markers) over
47 years (1970-2017). The link between patent applications and labour productivity is stronger in
Australia than in the US, Canada, or Türkiye. Panel B of Figure 1 presents the correlation between
patent applications and productivity using an ordinary-least-square (OLS) regression, which controls for
time-invariant characteristics of each country using country fixed-effects. There is a clear positive
association between patent applications and productivity, although the strength of this relationship
varies across countries. In Australia, a 1% increase in patent applications is associated with a 7%
increase in labour productivity.

Figure 1. Patent applications and labour productivity

Scatterplot of labour productivity and patent applications (left panel) and impact of patent applications on labour
productivity (right panel) across selected OECD countries
Panel A: Patent applications have a strong positive relationship Panel B: An increase in patent applications leads to an increase in
with labour productivity labour productivity

Note: The figure presents a scatterplot of patent applications and labour productivity, both in logarithm (left panel) and the coefficients from
a regression of labour productivity on patent applications (right panel). Data include Australia, Canada, Denmark, Japan, South Korea, New
Zealand, Norway, Türkiye, and the US between 1970 and 2017 (47 years). The regression from the right panel controls for labour quality,
economic stability, and country-time fixed effects.
Source: Data are from Sarwar (2022[20]).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  11

3 Data and preliminary evidence

This section elaborates on the data sources and presents descriptive analysis and preliminary evidence
on migration and innovation in Australian regions. It starts by describing the data sources and construction
of the analysis sample. Next, it provides an overview of migration and innovation across Australian regions,
discussing their relationship in detail.

Data sources and sample construction

Data sources

This study uses rich individual-level administrative panel data provided by the Australian Bureau of
Statistics (ABS). The Multi-Agency Data Integration Project (MADIP) dataset compiles information from
various ministries related to health, education, government payments, income and taxation, employment,
population demographics, migration, as well as Census data. It contains 27.1 million individual records,
covering all Australian residents who interacted with the social security system, paid income tax, or
engaged with the health system at any point between 2006 and 2020.
The Intellectual Property Government Open Dataset (IPGOD) dataset, maintained by IP Australia, provides
information on intellectual property (IP) rights applications used to construct regional innovation measures.
This dataset offers detailed geolocated information on the number of patents, design rights, and trademark
applications. Each of these measures captures distinct dimensions of innovation. Patents and design rights
encapsulate innovations that are new to the world. Trademarks represent innovations that are not new to
the world but are new either to the firm or the regional market. These measures display a high correlation
with other firm-level indicators of innovation, such as survey-based measures of perceived innovation and
R&D expenditure (Jensen and Webster, 2009[21]). Annex A provides a comprehensive description of both
data sources.

Constructing the analysis sample

The primary analysis is conducted at the SA4 level, a classification developed by the ABS based on
regional labour demand and supply data to reflect local labour market areas.9 The individual-level
residence information is aggregated at the regional level and yearly frequency.
The analysis uses the MADIP dataset to generate socio-economic variables. It restricts the sample to
individuals aged 15-64 who were employed at some point between 2011 and 2018 and whose income was
above the minimum threshold for a tax declaration. Following the literature, the analysis excludes workers
in the public, health, agriculture, and mining sectors, as these sectors are generally less likely to employ

9
Australia is disaggregated into 89 SA4 regions with a population between 100 000 and 500 000. Following the OECD
territorial grid, the SA4 “Other territories” is excluded from the analysis, resulting in a total of 88 SA4 regions.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

 12 

migrants or apply for design rights, patents, or trademarks. The resulting sample contains nearly 26 million
individual-year observations over the 2011 to 2018 period.
Matching the sample with the 2016 Census allows to obtain education information. The primary variable
of interest is the inflow of higher-educated migrants, defined as the net change in higher-educated migrant
workers relative to the total number of workers in a region, defined based on individual-level residence
information.10 Higher-educated workers are workers with at least a college degree. 11 Additionally, the
analysis constructs relevant controls at the regional level, such as industry shares (defined as the share of
workers working for one-digit industry groups), the share of higher-educated native workers, and the
logarithm of the population count.
IPGOD microdata provides regional innovation measures, such as the number of annual applications for
patents, trademarks, and design rights. The dataset provides detailed information about each IP rights
application, including the involved parties, their location, and their roles. The analysis considers all patent
applications in Australia with at least one applicant located in Australia regardless of their country of birth
or nationality. In the years of the analysis, firms submitted about 95% of all patent applications. In cases
where applications involve applicants from different locations, the paper assigns a count of one to each
location. However, the analysis also conducts robustness tests using fractional counts across locations.12

Additional datasets

Measuring the causal impact of migration on innovation requires using instrumental variables based on
past settlement patterns of migrants, as explained in Section 4. Historical Census data from 1981 provide
reliable information on the migrant distribution across Australian regions. The historical census, adjusted
to 2016 borders by the ABS, provides information on the total employed population in 1981 disaggregated
by country of birth, industry of employment, and highest post-school qualification, where tertiary education
matches with the Census definition. Section 4 provides more details on the construction of the instrument.

10
Lacking information on migrants' workplace locations, the analysis assumes that migrants work in the same region
where they reside.
11
Annex C shows that using a higher-skilled definition based on occupations rather than education does not alter the
results, as the two definitions are highly correlated. The analysis considering higher-skilled migrants is not limited to
information based on the Census 2016. Hence, it also confirms that the results are not driven by the exclusion of
migrants without education information.
12
The fractional count allocates, for each party in an IP rights application, a share equal to one divided by the total
number of applicants for each applicant’s location. It gives a lower weight to innovation happening in subsidiary firms
if multi-party applications include parent companies which are typically located in denser metropolitan areas.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  13

Box 2. The geographical distribution of migrants

Australia is one of the largest migrant-receiving countries in the OECD. In 2021, Australia had the third-
highest share of migrants (29%) among OECD countries, after Luxembourg (49%) and Switzerland
(30%). This share is substantially higher than the migrant share in Canada (21%), Germany (16%), the
UK (14%), and the US (14%) (OECD, 2023[22]). Moreover, the share of migrants in Australia increased
by six percentage points from 23% in 2000. Over the same period, the migrant share across the total
OECD increased by just four percentage points from 10% to 14%.
The presence of migrants has a pronounced regional dimension in Australia. About 82% of all Australian
migrants concentrate in large and midsize metropolitan areas, such as Brisbane, Melbourne, Perth, and
Sydney, compared to 66% of natives. Consequently, only 18% of the migrant population lives in non-
metropolitan areas, compared to almost one-third (33%) of natives. As a result, migrants constitute a
high share of the population in large metropolitan areas (40%) such as Brisbane, Melbourne, Perth,
and Sydney. Similarly, in midsize metropolitan regions, the migrant share is around 29%. In non-
metropolitan areas, however, less than one-fourth of the population is born abroad, with some regions
in the southeast exhibiting values of less than 10% (OECD, 2023[1]).

Figure 2. The presence of migrants varies across Australian labour markets

Share of the foreign-born population across regions, 2016

Note: The figure presents the share of foreign-born among the working-age population (15-64 years) in Australia disaggregated by regions.
Data are for 2016.
Source: OECD calculations based on the Australian Census of Population and Housing 2016 accessed via ABS Census TableBuilder
(accessed May 2022).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

14 

Preliminary evidence: Innovation and migration at the regional level Patenting activities in Australia exhibit
a substantial geographical dimension. Figure 3 illustrates the number of patent applications per 10 000
workers across Australian regions in 2018. Patenting activity per workers varies widely, ranging from
around one in Murray to above 70 in districts of Sydney, Melbourne, and Perth. The major cities Adelaide,
Brisbane, Melbourne, Perth and Sydney are, on average, the most innovative, with almost 19 patents per
10 000 workers. This value is almost double the value of the remaining regions.

Figure 3. Geographical distribution of patenting activity across Australian regions

Number of patents per 10 000 workers in Australian regions, 2018

Note: The figure presents the cumulative number of patents per 10 000 workers in Australia disaggregated by regions. Data are for 2018.
Source: OECD calculations based on MADIP and IPGOD (accessed July 2023).

The regional presence of higher-educated migrants only shows a positive association with patents but not
with trade markets and design rights. Figure 4 provides scatterplots illustrating the correlations between
higher-educated migrant shares and IP rights applications across Australian regions. While correlations
provide insights into the relationship between migration and regional innovation, they do not reveal the true
impact of migration on regional innovation. As indicated by the red slope line, a positive relation exists
between the increases in the share of higher-educated migrants among the total workforce and increases
in patent applications per worker. At face value, the correlation would imply that a one percentage point
increase in the share of higher-educated migrants among the workforce correlates with an average 1.65%
increase in patent applications per worker. Conversely, the flat slope line suggests a lack of a clear
association for trademarks and design rights.
However, these figures do not necessarily indicate a causal relationship between migration and regional
innovation. As migrants may settle in already innovative urban areas, these associations may be
MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024
  15

correlational and not causal. The following section details the empirical strategy employed in this paper to
overcome this issue.

Figure 4. Correlations between changes in IP rights applications and higher-educated migrant

shares
Changes in IP rights applications and share of the higher-educated foreign-born population across regions, 2011-2016

Panel A: Patents Panel B: Trademarks

Panel C: Design Rights

Note: The figures present scatterplots showing the correlations of changes in the logarithm of IP rights applications per worker(vertical axis) and
changes in higher-educated migrant shares across Australian regions (horizontal axis) between 2011 and 2016. Panel A refers to patents, Panel
B to trademarks, and Panel C to design rights. Each circle represents a region, and its size corresponds to its population in 2011. Lines of best
fit are calculated from OLS regressions weighted by population in 2011. Estimated slope coefficients and robust standard errors are provided
on top.
Source: OECD calculations based on the MADIP and IPGOD 2022 data (accessed July 2023).

Summary statistics based on the sample

Table 1 shows the unweighted summary statistics of the main variables used in the analysis, including
changes from 2011 to 2016 across regions. On average, the regional share of migrants in the workforce
is 24%, and the regional share of workers with college degrees is 21% (comprising 8% migrants and 13%
natives). However, notable regional disparities exist. For example, the regional share of higher-educated
migrants in the workforce ranges from 1% to 30%. Similarly, substantial differences exist in IP rights
applications, with some regions showing minimal innovation while others have application rates per worker

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

16 

ten times higher than the average. Across regions, trademarks are the most common, followed by patents.
In contrast, design rights are less common, reflecting their specific nature, primarily used by the
manufacturing industry.
Panel B of Table 1 presents the changes in the shares of higher-educated migrants and natives from 2011
to 2016 across regions and the changes in IP rights applications per worker. The growth in the number of
higher-educated migrants is stronger than that of higher-educated natives both in levels and in relative
terms due to a substantial inflow of higher-educated migrants between 2011 and 2016 and a smaller initial
higher-educated migrant population. Moreover, there is a slight decrease in patent applications per worker,
while trademark applications per worker experience a substantial increase. Moreover, design rights
applications show a small increase.

Table 1. Descriptive statistics (unweighted)

Variable Mean Std. Dev. Min Max
Panel A: Levels, yearly variables (2011-2018)
Migrant share 0.239 0.134 0.065 0.607
Share of higher-educated migrants over total workers 0.076 0.070 0.009 0.318
Share of higher-educated natives over total workers 0.125 0.610 0.049 0.309
Number of patents applications per 10 000 workers 25.28 25.47 0 214.19
Number of trademarks applications per 10 000 workers 147.57 153.24 0 1322.20
number of design rights applications per 10 000 workers 6.94 10.07 0 98.37
Panel B: Relative changes, 2011-2016
Change in share of higher-educated migrants over total workers 0.045 0.043 0.003 0.174
Change in share of higher-educated natives over total workers 0.038 0.025 0.009 0.129
Change in number of patents applications per 10 000 workers -3.65 11.37 -40.42 62.44
Change in number of trademarks applications per 10 000 workers 42.76 54.44 -65.73 250.04
Change in number of design rights applications per 10 000 workers 5.39 10.61 -18.50 80.58

Note: The table presents descriptive statistics on regional measures of innovation and migration. Panel A contains yearly variables across 88
regions from 2011 to 2018, amounting to a total of 704 observations. In all variables, the denominator is the total (migrant and native) employed
population. The mean value is calculated across all 704 year-region observations and is not weighted by population size. Panel B contains
changes from 2011 to 2016, divided by the employed population in each region in 2011. Each statistic in Panel B is calculated in a cross-section
of 88 regions. Variables per 10 000 workers are per-worker variables multiplied b I bo h p , h co m “M ” p h m
value of the sample, wh h co m “M ” “M x” p h v of h g o w h h ow h gh ch g , p c v
Source: OECD calculations based on MADIP and IPGOD (accessed July 2023).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  17

4 Empirical strategy
This section presents the empirical strategy used to estimate the impact of migration on innovation in
Australian regions. First, it explains the employed empirical model. Second, it discusses the empirical
challenges in measuring the causal impact of migration on regional innovation and explains the methods
used to address them.

Empirical model

The analysis adopts a difference-in-difference approach to measure the impact of migration on regional
innovation, comparing regions that received more higher-educated migrants with those that received
less.13 The strategy employs a first-differences regression model which consists of measuring the changes
over time to eliminate the influence of certain time-invariant location-based (e.g., regional infrastructure or
population density) and group-based characteristics (e.g., age, sex). Accounting for such time-invariant
characteristics eliminates any factor specific to these places or groups that may affect the relationship
between migration and regional innovation.
Following the literature, the estimation uses five-year changes to capture medium-run effects (Hunt and
Gauthier-Loiselle, 2010[6]):

𝐻𝐸
𝛥 log 𝑌𝑟,𝑡+5 = 𝛽0 + 𝛽1 𝛥𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟,𝑡 + 𝛽2 𝑋𝑟,𝑡 + 𝛼𝑡 + 𝛥𝜀𝑟,𝑡 (1)

Where 𝛥 log 𝑌𝑟,𝑡+5 is the logarithm change in the number of applications in year 𝑡 + 5 and region 𝑟 per
employed population at baseline year 𝑡 for each period.14 Following the literature, in the baseline, the
dependent variable (patent, trademark, and design rights applications) measures the changes in the IP
rights applications not in the same year (i.e., time t) but after five years (i.e., time t+5) to allow time between
the migration-induced change in labour inputs and IP rights applications (Hunt and Gauthier-Loiselle,
2010[6]).15 Recent evidence for patents argues that the influence of newly arriving migrants reaches its
peak on patent applications four years after arrival (Blit, Skuterud and Zhang, 2020[23]). Robustness checks

13
In the migration literature, this is known as a spatial correlations approach (Dustmann, Schonberg and Stuhler,
2016[37]).
14
Following the literature, variables are divided by baseline employed population, preventing estimates from capturing
changes due to native mobility (Card and Peri, 2016[38]).
15
The results are robust to using contemporaneous rather than one-year leaded changes in IP rights applications. For
patents, estimated effects are smaller although still significant, consistent with the notion that innovation takes time to
materialise.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

18 

further include annual and triannual changes, as well as the effect on cumulative patents, i.e.,
log ∑𝑡+𝑑
𝑗=𝑡+1 𝑌𝑟,𝑗 over three and five years (d=3,5).
𝐻𝐸
𝛥𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟,𝑡 is the change in the headcount of higher-educated employed migrants in year t over the
employed population at baseline. 𝑋𝑟𝑡 contains baseline characteristics, including shares of workers in one-
digit industries, the change in higher-educated native workers, and the logarithm of the population; 𝛼𝑡 is a
vector of time fixed-effects that absorbs the effect of any shock that might affect all Australian regions each
year. 𝛽1 is the coefficient of interest indicating the effect of a change in migration on the outcome variables,
namely patent, trademark, and design rights applications. The analysis clusters the standard errors at the
regional level to account for possible within-region correlation of random disturbances.

Endogeneity of the migrant share

Assessing the causal impact of migration on regional innovation poses an empirical challenge because of
a possible “omitted variable bias”, whereby other regional factors that are not accounted for in the analysis
influence migrants’ location choices and regional innovation. Migrants generally prefer to live in cities
(OECD, 2022[15]), but innovation also clusters in cities to benefit from agglomeration economies (Carlino
and Kerr, 2015[24]). Moreover, measurement error in recording those migrants who innovate - resulting
from using changes in higher-educated or higher-skilled individuals as measures - could potentially weaken
the relationship between migration and innovation.
To address both challenges, the empirical strategy combines the difference-in-differences method with
instrumental variables. The analysis uses an instrument based on the past settlements of the migrant
population (also known as the shift-share instrument), which is the most extensively used instrumental
variable in the migration literature (Jaeger, Ruist and Stuhler, 2018[25]). The idea is that newly arrived
migrants tend to settle in places where they can find migrants from their own country of origin. This way,
the instrument predicts the inflow and settlement of migrants driven by networks rather than economic
factors that may be driving innovative activity (Bartik, 1991[26]). The instrument uses information on the
location of migrants who have arrived in earlier years to predict where the new migrants will settle, including
information on pre-existing migrant enclaves and the number of newly arrived migrants at the national level
by country of origin.
Building the shift-share instrument follows these steps (see Annex B for further details):
1. Split the migrant population of 1981 into 60 countries or regions of origin (See Table 5 (Annex B)
for a detailed list).
2. Calculate the regional distribution (settlement pattern) of each origin using the 1981 Census.
3. For each country and region of origin, predict the presence of higher-educated migrants in each
region and year using the total annual migrant population by origin between 2011 and 2018 (i.e.,
the shift) and the regional distribution of migrants by origin in 1981 (i.e., the share).
4. Sum up predicted settlements of higher-educated migrants across countries and regions of origin
to obtain the predicted total number of higher-educated migrants living in a given region and year.
5. Apply the same process to obtain the predicted number of natives to avoid the instrument capturing
the mobility response of natives due to the migrant inflow.
6. Use the predicted higher-educated migrant population to compute the predicted increase in higher-
educated migrants over the total predicted number of migrants and natives at baseline in each
region and year.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  19

Validity of the instrumental variable strategy

The main identification assumption is that the instrumental variable affects regional innovation only through
its impact on the increase of higher-educated migrants. This assumption is a combination of the
instrumental variable being associated with the migrant increase (instrument relevance) and not being
associated with other factors determining innovation (instrument exogeneity). In the context of shift-share
instruments, recent literature has shown that instrument exogeneity can be satisfied from either exogeneity
of the aggregate “shifts” (Borusyak, Hull and Jaravel, 2022[27]) or the baseline “shares” (Goldsmith-
Pinkham, Sorkin and Swift, 2020[28]). This study relies on identification based on the exogeneity of the
baseline shares, which means that the initial settlement of migrants across regions in 1981 is not correlated
with persistent omitted factors that could also determine regional innovation.16
Three tests provide evidence that the instrument used in the analysis satisfies the exogeneity condition.
First, initial shares of migrants in 1981 should be uncorrelated with regional characteristics in 1981
(Goldsmith-Pinkham, Sorkin and Swift, 2020[28]). Table 6 in Annex B shows that the shares of the top origin
nationalities that drive most of the variation during the 2011-2018 period, i.e., India, China, Philippines,
and Korea17, as well as the instrument built upon these shares, are not associated with a set of regional
characteristics reflecting regional labour markets characteristics and industry composition. Second, inflows
of higher-educated migrants prior to the study period should not influence current changes in innovation
(Jaeger, Ruist and Stuhler, 2018[25]). By including past inflows of higher-educated migrants in the main
estimation equation, in Annex B shows that the results are unchanged. Consequently, the main estimates
are driven by current migration inflows rather than by the long-term effects of past migration inflows. Finally,
the effect of higher-educated migrants on innovation could be explained by a spurious correlation between
predicted higher-educated migrants flows by the instrument and pre-existing trends in innovation. Annex
B shows that the instrument has no significant correlation with previous trends in IP applications.
These tests provide evidence in favour of the exogeneity of the baseline shares in 1981. In turn, the
instrument built upon this share is likely to be uncorrelated with other characteristics that could be driving
the impact on changes in regional innovation during the 2011-2018 period. These tests are a necessary
condition for the instrument to impact regional innovation only through its effect on actual inflows of higher-
educated migrants.

16
The number of high-educated migrants increased as of 2005 due to the reforms in the migration policies (Nguyen
and Parsons, 2018[39]). While the dramatic increase also creates an exogenous “shift”, this study relies on the
exogeneity of the “shares”.
17
These are the four nationalities that contribute the most to the increase in migration during the 2011-2018 period.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

 20 

5 Results
This section presents the results in two steps. Firstly, it examines the impact of migration on regional
innovation. Secondly, it delves into the uneven effects due to different types of migrants and variations
across regions.

Average effect of migration on innovation across regions

The effect of migration on innovation may not materialise immediately, as innovation often takes time to
develop. To capture the varying horizons of innovation, the regressions consider 5-year, triannual, or
annual differences for each outcome. Table 2 provides the results from estimating the main empirical
model, using either patents, trademarks, or design rights as outcomes. While Panel A presents Ordinary
Least Squares (OLS) estimates, Panel B provides instrumental variable estimates using Two Stage Least
Squares (2SLS), which addresses potential endogeneity concerns.

Higher-educated migrants boost regional patenting activity with no impact on other

intellectual property rights

Migration is positively associated with patenting applications but not with trademarks or design rights. The
OLS and 2SLS estimates indicate a significant and positive impact of the inflow of higher-educated
migrants on regional patent applications per worker.18 On average, a one percentage point increase in the
regional employment of higher-educated migrants relative to total employment leads to a 4.8% rise in
regional patent applications in the medium run (five years). Similar effects are also observed in the shorter
run (one or three years). 19 In contrast, the effect of migration on regional trademarks and design rights
applications remains insignificant for both estimates and regardless of the timeframe. Moreover, increases
in the number of higher-educated native individuals do not appear to influence any type of IP rights.20
The arrival of higher-educated migrants between 2011 and 2016 lifted regional patent applications by
around 20%. Between 2011 and 2016, the average Australian region experienced a 4.5 percentage point
increase in employment due to higher-educated migration, as shown in Table 1. A back-of-the-envelope

18
The instrumental variable is a strong predictor of the main endogenous variable, conditional on the exogeneity of
the instrument (see Annex B). This is confirmed by a consistently high first-stage F-statistic exceeding 10. The small
and statistically insignificant differences between IV and OLS estimates suggest that the omitted variable bias and
measurement error are minimal.
19
The point estimates increase slightly when exploiting annual changes (from 4.8% to 6.6%). Yet, as the confidence
intervals at any conventional significance level overlap, the estimates are not statistically different.
20
The lack of significant effects for natives are possibly driven by two factors. Firstly, the proportion of higher-educated
natives is already substantial with limited variation across regions and time. Secondly, higher-educated migrants tend
to live in places where natives are also higher educated (OECD, 2022[15]). These factors make it challenging to
empirically measure the distinct effects of higher-educated migrants and natives on innovation simultaneously.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  21

calculation based on the five-year impact of migration on innovation (Column 1 in Panel B, Table 2)
suggests that higher-educated migration during this period contributed to an approximate increase in
patenting activity of 20%. This implies that if the baseline number of patent applications per 10 000 workers
is 33, the presence of migrants increased this value to 39.
The increase in patenting but not in trademarks or design rights aligns with previous findings. For Australia,
previous research finds that an increase in Temporary Graduate visa holders between 2007 and 2014 led
to more patent applications per worker, while trademarks or design rights were not significantly impacted
(Crown, Faggian and Corcoran, 2020[18]). Similarly, in the case of the US, previous research finds an
impact of migrants on innovation and patenting in high-tech sectors but no effect on trademarks (Brown
et al., 2020[29]). The inflow of higher-educated migrants between 2011 and 2016 in Australia predominantly
results from the arrival of migrants in STEM industries that are more likely to patent. This potentially
explains the significant regional effect on patterns and the absence of any effects on other innovation
measures.

The positive effect of migration on regional innovation holds against different robustness tests. Annex C
presents various robustness checks assessing the sensitivity of results and validating the empirical
strategy. Overall, the magnitude and statistical significance of the coefficients are consistent with the
baseline results.

Table 2. The average effect of migration on regional innovation

Estimated average effect of migration inflows on regional patents, trademarks, and design rights at the regional level

Patents Trademarks Design rights

5 years 3 years 1 year 5 years 3 years 1 year 5 years 3 years 1 year
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Panel A: OLS
∆ HE Migrants 3.572** 3.838*** 4.917* -1.066 -0.397 1.931 -0.766 -1.568 9.122
(1.242) (1.825) (2.523) (0.651) (0.849) (1.136) (3.743) (3.969) (5.344)
∆ HE Natives 0.701 -0.941 -3.529 -1.826 -1.454 -1.494 -2.374 12.036* -2.287
(2.375) (2.839) (3.134) (1.563) (1.912) (1.831) (7.178) (6.954) (8.510)
R2 0.211 0.082 0.052 0.054 0.077 0.189 0.066 0.204 0.142
Panel B: 2SLS
∆ HE Migrants 4.754** 6.351*** 6.564** -1.459 -1.919 -0.245 0.754 8.725 14.621*
(1.726) (2.832) (3.264) (1.225) (2.671) (2.403) (5.727) (7.235) (8.241)
∆ HE Natives 0.576 -1.542 -4.603 -1.785 -1.091 -0.075 -2.534 9.575 -5.873
(2.340) (2.909) (3.744) (1.540) (2.255) (2.487) (6.878) (6.471) (8.935)
N 88 176 616 88 176 616 88 176 616
F-stat 27.1 25.3 43.1 27.1 25.3 43.1 27.1 25.3 43.1

Note: The table presents results based on estimating Equation 1 using as outcomes either Patents (Columns 1-3), Trademarks (Columns 4-6)
and Design rights (Columns 7-9) applications. Panel A uses OLS as an estimator, while Panel B estimates IV results using a 2SLS estimator.
The dependent variable is expressed as log changes in IP applications per worker, using either 5, 3, or 1-year differences. The independent
variable is the increase in employment due to higher-educated migrants, where the measure of high education is tertiary education (at least with
a college degree). The columns present different time intervals for both the dependent and independent variables. Columns 1, 4 and 7 represent
five-year changes (2011-2016), Columns 2, 5 and 8 represent three-year changes (2011-2014 and 2014-2017), and Columns 3, 6, and 9
represent annual changes. No intervals overlap. All specifications control for baseline shares of higher-educated population, log of population
and industry shares. All specifications are weighted by the number of employed natives in the region at the baseline year. Time fixed-effects are
applied to account for time-varying shocks affecting the entire country, except for five-year changes due to collinearity. Standard errors, in
parentheses, are clustered at the regional level in all specifications. Statistical significance is denoted by ***, **, and * at the 1%, 5%, and 10%
levels, respectively. The analysis considers 88 regions, yielding 616 observations over five years.
Source: OECD calculations based on MADIP and IPGOD (accessed July 2023).
MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024
 22 

Uneven effects across workers and places

The impact of migration on innovation can differ based on the characteristics of migrants and the conditions
of the regional economy. Firstly, the analysis delves into the impact of distinct groups of migrants,
categorised by skill levels and educational attainments. Secondly, the analysis tests whether the effects
vary across different Australian regions, depending on their existing levels of innovation, population
density, and income.

Migrants of all education groups contribute to patenting across Australian regions

Estimating the effect of different migrant subgroups based on their education and occupation allows a
broader understanding of the impact of migration on innovation. Table 3 compares the baseline estimates
(Panel A) to alternative definitions of the migration inflow. Panel B focuses specifically on changes in
higher-skilled migrants, employing occupation-based skill assignments instead of education levels. Panel
C restricts the migrant inflow to scientists using a subset of occupations more likely to be involved in
patenting activities (for detailed information, see Annex D). Finally, Panel D includes all migrants,
irrespective of their educational background or occupation. Consistent with prior studies (Hunt and
Gauthier-Loiselle, 2010[6]), all specifications use the instrument based on 1981 settlement patterns.
In all panels, the findings consistently demonstrate positive effects on patent applications with similar
magnitudes across various specifications and timeframes. This suggests that the definition of higher-
educated workers does not substantially influence the results, and migrants of all skill and education groups
contribute to the patenting activity. Moreover, the effect is most pronounced when focusing only on
migrants employed as scientists. Similar to the baseline findings, there are no discernible effects on
trademarks or design rights. These patterns further affirm the reliability of the migration measure and the
empirical approach. Finally, Table 10 in Annex E shows that even when excluding scientists, the effect of
all migrants or higher-educated migrants on patenting activity is still significant, thus showing that migrant
scientists are not driving the results.

Table 3. Uneven effects of migrants with different skills and education backgrounds
Patents Trademarks Design rights
5 years 3 years 1 year 5 years 3 years 1 year 5 years 3 years 1 year
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Panel A: Baseline - Higher-educated (at least college)
∆ HE Migrants 4.754*** 6.351** 6.564** -1.459 -1.919 -0.245 0.754 8.725 14.621*
(1.726) (2.832) (3.264) (1.225) (2.671) (2.403) (5.727) (7.235) (8.241)
F-stat 27.1 25.3 43.1 27.1 25.3 43.1 27.1 25.3 43.1
Panel B: Higher-skilled occupations
∆ HS Migrants 4.916*** 6.407** 5.367* -1.532 -2.539 -0.251 1.912 6.994 12.402
(1.862) (2.733) 2.878 (1.216) (2.395) (2.161) (6.316) (7.058) (9.059)
F-stat 38.7 33.3 67.1 38.7 33.3 67.1 38.7 33.3 67.1
Panel C: Scientists
∆ Scientist 12.050*** 15.375** 17.197** -3.828 -5.862 -2.175 2.868 14.111 46.042**
Migrants
(4.239) (6.428) (7.640) (2.790) (5.446) (5.897) (14.882) (16.276) (21.865)
F-stat 31.9 26.6 37.8 31.9 26.6 37.8 31.9 26.6 37.8
Panel D: All migrants
∆ Migrants 1.550*** 2.135** 2.454** -0.407 -0.781 0.032 0.769 2.167 4.993
(0.597) (0.891) (1.148) (0.390) (0.829) (0.831) (2.029) (2.472) (3.342)
F-stat 33.1 30.4 77.1 33.1 30.4 77.1 33.1 30.4 77.1

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  23

Note: The table presents results based on estimating Equation 1 using as outcomes either Patents (Columns 1-3), Trademarks (Columns 4-6)
and Design rights (Columns 7-9) applications. Panel A uses the same definition as baseline, higher-educated individuals (with at least college
education). Panel B uses higher-skilled occupations, while Panel C uses a subset of applications that are more likely to submit patent
applications, as specified in Annex D. Panel D uses all migrants. The dependent variable is expressed as log changes in IP applications per
worker, using either 5, 3, or 1-year differences. The independent variable is the increase in employment by type of migrant. The columns present
different time intervals for both the dependent and independent variables. Columns 1, 4 and 7 represent five-year changes (2011-2016), Columns
2, 5 and 8 represent three-year changes (2011-2014 and 2014-2017), and Columns 3, 6, and 9 represent annual changes. No intervals overlap.
All specifications control for baseline shares of higher-educated population, log of population and industry shares. All specifications are weighted
by the number of employed natives in the region at the baseline year. Time fixed-effects are applied to account for time-varying shocks affecting
the entire country, except for five-year changes due to collinearity. Standard errors, in parentheses, are clustered at the regional level in all
specifications. Statistical significance is denoted by ***, **, and * at the 1%, 5%, and 10% levels, respectively.
Source: OECD calculations based on MADIP and IPGOD (accessed July 2023).

The estimates indicate that the impact of migrant scientists on regional innovation is less than half of what
was estimated for the US (Hunt and Gauthier-Loiselle, 2010[6]). The study for 1950-2000 shows that a one
percentage point increase in the share of scientist migrants led to a substantial 52.4% rise in patents over
a 10-year period.21 However, the analysis might underestimate the positive impact as migrants in Australia
could be patenting abroad, particularly in the US. This area remains a potential focus for future research.

The impact of migration on innovation depends on existing regional innovation activity

but not on density or income

This subsection explores uneven effects across regions by focusing on several regional characteristics in
the baseline year. The estimations in Table 4 expand the main estimation equation by interacting the
variable of interest (i.e., the regional change in higher-educated migrants) with an indicator equal to one if
the regions have below-the-median values for the considered regional characteristic in the initial year. The
interaction term is additionally instrumented by the analogous interaction on the instrument, following
previous literature (Özgüzel and Edo, 2023[30]).
The analysis focuses on three different regional characteristics to explore the uneven effects of migration
on regional innovation. First, density, defined as the number of workers over the regional built-up area, can
play a role in innovation as proximity among innovators can facilitate collaboration and foster innovation.
22
Secondly, income levels, measured using average regional labour income, may influence the degree of
innovation in different regions, as places with higher income levels may disproportionately attract migrants
who are better innovators. Lastly, regions already highly innovative may have innovation hubs that
potentially use the migrants’ skills more efficiently and, hence, generate larger effects. Alternatively, it is
also possible that such places benefit relatively less from new migrant inflows as increasing innovation
becomes more difficult at higher levels of innovation compared to less innovative places.

The impact of migrants on regional innovation is independent of regional income and density yet varies
depending on the pre-existing innovation levels. Table 4 outlines the results. The interaction parameter for
regional density and income (Panels A and B) lacks significance, suggesting no differences exist across
regions above and below the median value. In contrast, in Panel C, regions below the median in existing
innovation levels show significant interaction coefficients for patents (5- and 1-year changes) and design

21
The study finds similar estimates whether they look at 10-year or 50-year windows, which suggests that the
estimated effects are stable across time. This makes it possible to compare these magnitudes estimated for the US
with those estimated in this study for Australia.
22
Retrieved from the Global Human Settlement Layer (GHSL): Global Human Settlement - GHSL Homepage -
European Commission (europa.eu). The baseline year is 2020.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

24 

rights (5 and 1-year changes), indicating that the effect is stronger in less innovative regions. This contrasts
with the belief that innovation mainly occurs in already innovative places (Crescenzi, Dyevre and Neffke,
2022[31]). However, even though less innovative regions might catch up in patenting activity, the overall
innovation gap across regions does not necessarily decrease.

Table 4. Uneven effects across regions with different characteristics

Patents Trademarks Design rights
5 years 3 years 1 year 5 years 3 years 1 year 5 years 3 years 1 year
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Panel A: Density (built-up)
∆ HE Migrants 4.773** 6.215** 6.110* -1.318 -3.113 -0.157 2.466 9.098 13.857*
(1.875) (2.937) (3.431) (0.917) (2.061) (2.277) (4.657) (8.167) (8.359)
∆ HE Migrants x -0.061 0.615 4.501 -0.657 5.043 -0.744 -7.877 -1.697 6.09
Below Median
Density
(2.572) (3.582) (4.773) (2.608) (4.113) (4.174) (9.534) (10.293) (9.998)
Panel B: Income
∆ HE Migrants 4.335** 5.887** 5.784* -1.56 -1.648 -0.333 -0.502 7.987 12.387
(1.802) (2.842) (3.454) (1.214) (2.699) (2.371) (5.703) (6.681) (8.325)
∆ HE Migrants x 1.56 1.491 3.426 0.364 -0.845 0.387 4.631 2.229 8.922*
Below Median
Income
(1.423) (1.615) (2.442) (0.714) (1.144) (1.347) (3.91) (5.236) (4.779)
Panel C: Innovation
∆ HE Migrants 5.731*** 7.790** 8.000** -1.408 -1.649 -0.2 1.61 9.502 16.357**
(1.804) (3.364) (3.359) (1.379) (2.839) (2.638) (5.437) (7.237) (8.188)
∆ HE Migrants x 2.764* 2.821 4.932** 0.517 1.579 0.254 13.368*** 6.917 16.461***
Below Median
Innovation
(1.43) (1.923) (2.143) (1.486) (2.355) (2.631) (4.766) (5.554) (4.608)

Note: The table presents results based on estimating an extended version of Equation 1 using as outcomes either Patents (Columns 1-3),
Trademarks (Columns 4-6) and Design rights (Columns 7-9) applications. The specification includes an interaction term with an indicator of
whether each region has below the median innovation, density or income. Panel A provides the results for different levels of innovation. Panel
B provides the results for different levels of built-up density, and Panel C investigates the interaction with differing levels of income. The
independent variables of interest are the increase in employment due to higher-educated migrants, where the measure of high education is
tertiary educated (at least with a college degree), and its interaction with an indicator variable that takes value one if the region has below the
median innovation, density or income, depending on the panel. The columns present different time intervals for both the dependent and
independent variables. Columns 1, 4 and 7 represent five-year changes (2011-2016), Columns 2, 5 and 8 represent three-year changes (2011-
2014 and 2014-2017), and Columns 3, 6, and 9 represent annual changes. No intervals overlap. All specifications control for baseline shares of
higher-educated population, log of population and industry shares. All specifications are weighted by the number of employed natives in the
region at the baseline year. Time fixed-effects are applied to account for time-varying shocks affecting the entire country, except for five-year
changes due to collinearity. Standard errors, in parentheses, are clustered at the regional level in all specifications. Statistical significance is
denoted by ***, **, and * at the 1%, 5%, and 10% levels, respectively.
Source: OECD calculations based on MADIP and IPGOD (accessed July 2023).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  25

6 Concluding remarks
Migrants play a crucial role in regional development by contributing significantly to income, international
trade, and labour markets. This paper provides the first causal evidence of migrants' impact on regional
innovation in the country, using comprehensive administrative data covering all Australian residents and
IP rights applications. This study expands the understanding of migrants' influence on innovation, focusing
on the case of an OECD country where migrants are higher-educated. Additionally, it presents new insights
by examining how migrants affect regional innovation while considering regional characteristics.
The paper uncovers four main findings. First, it shows that a one percentage point increase in regional
employment due to the arrival of higher-educated migrants leads to a 4.8% increase in patent applications
per worker in the medium run (five years). Second, while migrants of all skill and education levels positively
impact patenting, the effect is most pronounced for migrants in scientific occupations. Third, regions with
lower pre-existing levels of patenting benefit relatively more from higher shares of migrants than those with
higher patenting levels. Other regional factors like income or population density do not influence the effect.
Finally, the analysis finds no effect on trademarks or design rights applications.
The analysis presented in this paper is an important step towards understanding the impact of migration
on innovation, but more research is needed for a deeper understanding of this relationship and underlying
mechanism. By contributing to innovation in patenting, migrants contribute to the development and growth
of businesses, impacting overall welfare. Investigating how innovation driven by migrants contributes to
productivity growth would be crucial for understanding their role in economic development. Moreover,
investigating the underlying mechanism of the impact of migrants on innovation would be an essential
complement to this research. Firm-level analysis using matched employer-employee data would allow to
unfold and understand how migrants contribute to the increased patenting activity. Further, it would be
valuable to investigate why the positive effects of migration on patents do not translate to trademarks or
design rights. This divergence is possibly due to a lack of migrant entrepreneurship in sectors that use
trademarks or design rights more intensively (Brown et al., 2020[29]). Lastly, examining the impact of
employer-sponsored visas on firm-level innovation would provide valuable insights for policymakers.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

26 

References

Bartik, T. (1991), “Who Benefits from State and Local Economic Development Policies”, W.E. [26]
Upjohn Institute.

Bernstein, S. et al. (2022), “The contribution of high-skilled immigrants to innovation in the [7]
United States”, Working Paper.

Blit, J., M. Skuterud and J. Zhang (2020), “Can skilled immigration raise innovation? Evidence [23]
from Canadian Cities”, Journal of Economic Geography, Vol. 20/4, pp. 879-901.

Borusyak, K., P. Hull and X. Jaravel (2022), “Quasi-Experimental Shift-Share Research [27]
Designs”, The Review of Economic Studies, Vol. 89, pp. 181-213,
https://doi.org/10.1093/restud/rdab030.

Bosetti, V., C. Cattaneo and E. Verdolini (2015), “Migration of skilled workers and innovation: [13]
A European Perspective”, Journal of International Economics, Vol. 96/2, pp. 311-322,
https://doi.org/10.1016/J.JINTECO.2015.04.002.

Bratti, M. and C. Conti (2018), “The effect of immigration on innovation in Italy”, Regional [14]
Studies, Vol. 52/7, pp. 934-947.

Butts, K. (2023), “Difference-in-Difference Estimation with Spatial Spillovers”, Working Paper. [33]

Card, D. and G. Peri (2016), “Immigration Economics by George J. Borjas: A Review Essay”, [38]
Journal of Economic Literature, Vol. 54/4, pp. 1333-49,
https://doi.org/10.1257/jel.20151248.

Carlino, G. and W. Kerr (2015), Agglomeration and Innovation, Elsevier B.V., [24]
https://doi.org/10.1016/B978-0-444-59517-1.00006-4.

Clarke, A., A. Ferrer and M. Skuterud (2019), “A Comparative Analysis of the Labor Market [17]
Performance of University-Educated Immigrants in Australia, Canada, and the United
States: Does Policy Matter?”, Journal of Labour Economics, Vol. 37/S2, pp. S443-S490.

Crescenzi, R., A. Dyevre and F. Neffke (2022), “Innovation Catalysts: How Multinationals [31]
Reshape the Global Geography of Innovation”, Economic Geography, Vol. 98/3, pp. 199-
227, https://doi.org/10.1080/00130095.2022.2026766.

Crown, D., A. Faggian and J. Corcoran (2020), “Foreign-Born graduates and innovation: [36]
Evidence from an Australian skilled visa program”, Research Policy, Vol. 49/9, pp. 1-17.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  27

Crown, D., A. Faggian and J. Corcoran (2020), “Foreign-Born graduates and innovation: [18]
Evidence from an Australian skilled visa program”, Research Policy, Vol. 49/9, p. 103945,
https://doi.org/10.1016/J.RESPOL.2020.103945.

Dotzel, K. and T. Wojan (2022), “An Occupational Approach to Analyzing Regional Invention”, [34]
National Center for Science and Engineering Statistics (NCSES),
https://ncses.nsf.gov/pubs/ncses22202/assets/ncses22202.pdf.

Dustmann, C., U. Schonberg and J. Stuhler (2016), “The Impact of Immigration: Why Do [37]
Studies Reach Such Different Results?”, Journal of Economic Perspectives, Vol. 30/4,
pp. 31-56.

Gagliardi, L. (2015), “Does skilled migration foster innovative performance? Evidence from [11]
British local areas”, Papers in Regional Science, Vol. 94/4, pp. 773-794.

Ganguli, I., S. Kahn and M. MacGarvie (eds.) (2020), Immigrant Entrepreneurs and Innovation [29]
in the US High-Tech Sector, University of Chicago Press.

Goldsmith-Pinkham, P., I. Sorkin and H. Swift (2020), “Bartik Instruments: What, When, Why, [28]
and How”, American Economic Review, Vol. 110/8, pp. 2586-2624,
https://doi.org/10.1257/aer.20181047.

Hunt, J. (2011), “Which Immigrants are Most Innovative and Entrepreneurial? Distinctions by [5]
Entry Visa”, Journal of Labor Economics, Vol. 29/3, pp. 417-457.

Hunt, J. and M. Gauthier-Loiselle (2010), “How much does immigration boost innovation”, [40]
American Economic Journal: Macroeconomics, Vol. 2/2, pp. 31-56,
https://doi.org/10.1257/mac.2.2.31.

Hunt, J. and M. Gauthier-Loiselle (2010), “How Much Does Immigration Boost Innovation?”, [6]
American Economic Journal: Macroeconomics, Vol. 2/2, pp. 31-56.

Imbert, C. et al. (2022), “Migrants and Firms: Evidence from China”, American Economic [16]
Review, Vol. 112/6, pp. 1885-1914.

Jaeger, D., J. Ruist and J. Stuhler (2018), Shift-share instruments and the impact of [25]
immigration, National Bureau of Economic Research.

Jensen, P. (2014), “Understanding the impact of migration on innovation”, The Australian [19]
Economic Review, Vol. 47/2, pp. 240-250.

Jensen, P. and E. Webster (2009), “Another look at the relationship between innivation [21]
proxies”, Australian Economic Papers, Vol. 48, pp. 252-269.

Kerr, S. et al. (2016), “Global Talent Flows”, Journal of Economic Perspectives, Vol. 30/4, [10]
pp. 83-106.

Mayda, A., G. Orefice and G. Santoni (2022), “Skilled Immigration, Task Allocation and the [12]
Innovation of Firms”, IZA Discussion Paper Series, Vol. 15693.

Nguyen, T. and C. Parsons (2018), “The Labour Market Impact of a High Skilled Migration [39]
Wave:”, Working Paper.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

28 

OECD (2023), International Migration Outlook 2023, OECD Publishing, Paris, [22]
https://doi.org/10.1787/b0f40584-en.

OECD (2023), “Migration and regional productivity: Evidence from individual wages in [2]
Australia”, OECD Regional Development Papers, No. 60, OECD Publishing, Paris,
https://doi.org/10.1787/7bc64c78-en.

OECD (2023), “Regional productivity, local labour markets, and migration in Australia”, OECD [1]
Regional Development Papers, No. 39, OECD Publishing, Paris,
https://doi.org/10.1787/3cc8f669-en.

OECD (2023), “The impact of migration on regional labour markets in Australia”, OECD [3]
Regional Development Papers, No. 64, OECD Publishing, Paris,
https://doi.org/10.1787/d72110b5-en.

OECD (2022), International Migration Outlook 2022, OECD Publishing, Paris, [4]
https://doi.org/10.1787/30fe16d2-en.

OECD (2022), The Contribution of Migration to Regional Development, OECD Regional [15]
Development Studies, OECD Publishing, Paris, https://doi.org/10.1787/57046df4-en.

OECD (2022), “Unlocking Rural Innovation”, OECD Rural Studies, Vol. OECD Publishing, [35]
Paris, https://doi.org/10.1787/9044a961-en.

OECD/Eurostat (2018), Oslo Manual 2018: Guidelines for Collecting, Reporting and Using [41]
Data on Innovation, 4th Edition, The Measurement of Scientific, Technological and
Innovation Activities, OECD Publishing, Paris/Eurostat, Luxembourg,
https://doi.org/10.1787/9789264304604-en.

Ozgen, C. (2021), “The economics of diversity: Innovation, productivity, and the labour [8]
market”, Journal of Economic Surveys, Vol. 35/4, pp. 1168-1216.

Özgüzel, C. and A. Edo (2023), “Immigration and employment dynamics in European regions”, [30]
OECD Regional Development Papers, No. 52, OECD Publishing, Paris,
https://doi.org/10.1787/a20155a2-en.

Perez-Silva, R., M. Partridge and W. Foster (2019), “Are foreign-born researchers more [9]
innovative? Self-selection and the production of knowledge among PhD recipients in the
USA”, Journal of Geographical Systems, Vol. 21/4, pp. 557-594.

Proctor, M. (2023), https://www.matthewproctor.com/australian_postcodes. [32]

Sarwar, A. (2022), “Australia’s productivity challenge is an innovation challenge”, Australian [20]

Industry Group (AI group).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  29

Annex A. Data sources

Multi-Agency Data Integration Project (MADIP)

The Multi-Agency Data Integration Project (MADIP) dataset by the Australian Bureau of Statistics (ABS) is
an individual-level panel dataset that provides longitudinal information for more than 27 million individual
records between 2011 and 2020. MADIP combines administrative information from different departments,
such as the Australian Taxation Office (ATO), the Department of Education, the Department of Health and
Aged Care, the Department of Social Services, Services Australia, and the Department of Home
Affairs.The availability of the dataset is subject to the agreement of the data custodians of the individual
agencies and depends on the individual research question. In addition to administrative data, the MADIP
includes one of the quinquennial Australian Census of Housing and Population. Besides the MADIP core
data, this analysis relies on tax data by ATO, migration data by the Department of Home Affairs, and the
Census 2018. The following subsections describe the individual components of the dataset.

MADIP core data

The MADIP core dataset is at the centre of every analysis using MADIP data. It contains demographic
information like date of birth, gender, and date of death as well as location information on all residents in
Australia. Moreover, the dataset includes a spine ID integral to merging the individual datasets from
different agencies. By default, the dataset covers every Australian resident recorded in either Social
Security and Related Information, Personal Income Tax data, or Medicare Benefits Schedule data between
2006 and 2020, resulting in a total of 27.1 million individual records. However, not every recorded person
is listed in every individual dataset. For instance, income tax data is not available if the person has never
reported taxes (e.g., children).
The geographical information is available at different granularity levels, including SA4, SA3, and SA2.
Given the overwhelming coverage of the Australian population, the data is expected to be representative
at every geographical level. Location information is distinguished by residential and mail address. For
migrants, the business address is also reported. In the analysis, the individual location information is based
on the residential address or mail address, depending on data availability.
MADIP is expected to cover the vast majority of Australian citizens and residents due to the combination
of medicare, social benefits, and income tax records. According to the ABS, the following groups are
potentially underrepresented: i) recently arrived migrants without Medicare, ii) non-earning partners and
family members of working visa holders, iii) non-earning foreign students, iv) military personnel, v)
prisoners, vi) recently born individuals, not yet included in the Medicare Benefits Schedule.

Census of Population and Housing 2016

The Australian Census of Population and Housing is conducted every five years and includes, among
others, information on educational attainment, employment and work, family, and personal characteristics.
This paper uses data from the Census wave of 2016, the latest available for research, linked to the MADIP
universe. Due to Australian data confidentiality rules, only one Census wave at a time can be used in the
MADIP environment. The Census data refers to the data collected on the 9th of August, 2016.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

30 

In Australia, participating in the census is mandatory for Australian residents, with very few exceptions.
The ABS linked 20.7 million records of the Census 2016 to the MADIP data, which corresponds to 88% of
all collected Census records in 2016. According to the ABS, the following groups are not within the scope
of the Census: i) Australians overseas, ii) residents for less than six months, iii) visitors, iv) diplomatic
personnel and their families. The paper retrieves information on age, occupation, industry, and country of
birth from the Census.

Australian Taxation Office

The Australian Taxation Office provides administrative information on all employed individuals in Australia
based on official tax returns. The dataset covers around 16.7 million individual records, including everyone
with a tax return in Australia in at least one year from 2010/2011 – 2017/2018. The Australian financial tax
year ranges from July until June of the following year. However, in order to combine the data with other
datasets, the tax records are assumed to follow the calendar year (January-December rather than July-
June).23 Data spans from wages, total income, and insurance payments to job sector information. The
variables of interest to this analysis are age, individual wage/salary, the main salary or wage occupation
code, and industry. Employed individuals with an income below the threshold imposed by the ATO and,
hence, without a tax record are not considered in the data. This also includes most migrants on a working
holiday maker (WHM) visa.

Department of Home Affairs

The Department of Home Affairs provides administrative data on the native and migrant populations. The
data includes every individual (native- or foreign-born) who crossed the border of Australia between 1990
and 2020. The dataset is used to retrieve information on the country of birth, date (month and year) of
birth, and gender. Visa information is not available for all migrants. Moreover, due to changes in the visa
status after arriving in Australia, the visa information might not be reliable for all migrants.

IP Government Open Data

The IP Government Open Data (IPGOD) is a data set provided by IP Australia, containing information on
over 100 years of information on IP rights applications. The data are updated yearly, and this project uses
the most recent available data, IPGOD 2022. The data set allows researchers to investigate the state of
IP rights since applications were first filed to IP Australia. It includes rich microdata on applicants 24, key
dates and events, the classification of IP rights and the history of IP transfers and exchanges.
The data cover four types of IP rights administered by IP Australia: trademarks, patents, design rights and
plant breeder’s rights. The latter are not included in this paper due to their limited scope.
Importantly, the data include information on applicant’s location, namely postcode. These are used to
assign applications to different regions, using an updated cross-walk of postcodes to SA4 regions (Proctor,
2023[32]).

23
This means that the tax return for the financial year 2011/12 is treated as the tax return for the year 2012. The age
retrieved from the ATO, is adjusted accordingly.
24
For some applicants, the Australian Business number (ABN) is provided.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  31

Historic Census

As discussed in Section 5, the identification strategy of the paper requires the use of a historical instrument
based on the settlement patterns of migrants in the past. The information on the past settlement patterns
is obtained from census data from 1981, 1986, 1991, and 2001. The historical data has been adjusted to
2016 borders by the ABS and provides data on the total employed population of the respective year
disaggregated by country of birth, the industry of employment, and highest post-school qualification. The
country of birth consists of 60 national groupings. Table 5 lists the national groupings and their share of
the total migrant population in 1981.

Table 5. Migrant decomposition in terms of country of origin

Share of employed migrants among the employed migrant population, 1981
Country groupings Share of total foreign-born
employed population
Albania, Bulgaria & Romania 0.32%
Argentina & Uruguay 0.51%
Austria 0.87%
Bangladesh 0.03%
Belgium 0.15%
Brazil 0.05%
Cambodia, Laos & Myanmar 0.42%
Canada 0.46%
Chile 0.40%
China 0.99%
Colombia, Ecuador & Peru 0.13%
Cyprus 0.82%
Denmark, Finland, Norway & Sweden 0.87%
Egypt 1.20%
Fiji 0.27%
Former Czechoslovakia 0.63%
Former USSR 1.67%
Former Yugoslavia 5.63%
France 0.42%
Germany 4.52%
Greece 6.00%
Hong Kong & Macau 0.41%
Hungary 1.11%
India 1.44%
Indonesia & Timor-Leste 0.54%
Iran 0.11%
Iraq 0.11%
Ireland 1.62%
Israel 0.21%
Italy 10.97%
Japan 0.27%
Kenya, Malawi, Zimbabwe, Tanzania, Uganda & Zambia 0.27%
Korea 0.13%
Lebanon 1.48%
Malaysia & Brunei 0.76%
Malta 2.32%

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

32 

Mauritius 0.33%
Mexico 0.01%
Netherlands 3.92%
New Zealand 4.63%
Other Africa 0.37%
Other Middle East 0.05%
Pakistan 0.08%
Papua New Guinea 0.25%
Philippines 0.51%
Poland 2.18%
Portugal 0.40%
Singapore 0.30%
South Africa & Namibia 0.73%
Spain 0.54%
Sri Lanka 0.56%
Switzerland 0.26%
Syria 0.11%
Taiwan 0.03%
Thailand 0.09%
Türkiye 0.66%
United Kingdom 33.58%
United States of America 0.96%
Vietnam 0.82%
All other countries 0.54%

Note: The table presents the share of the foreign-born employed population in 1981. The countries of origin are aggregated into 60 national
groupings. Grouping was conducted by the ABS and refers to the international borders of 1981.
Source: Australian Census of Population 1981 from the Australian Bureau of Statistics (ABS).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  33

Annex B. Construction and validity of the

instrumental variable

This section details the construction of the instrumental variable and then discusses the validity tests to
provide evidence on the main identifying assumption, i.e., that the instrumental variable affects regional
innovation only through its impact on the increase of higher-educated migrants. The analysis draws from
data from the 2001 and 2006 Australian Population Census.

Construction of the instrumental variable

The construction of the instrument follows several steps. First, migrants are grouped into 60 origin groups,
as reported in Table 5. Second, the settlement patterns of the pre-existing migrant population by country
of origin across Australian regions is calculated using data from the 1981 Census. 25 The shares are
calculated as follows:
1981
1981
𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑛,𝑟
𝑆ℎ𝑎𝑟𝑒𝑛,𝑟 = 1981
∑ 𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑛,𝑟
1981
The numerator 𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑛,𝑟 is the number of employed migrants in 1981 in each of the 60 national
1981
groupings n in region r. The denominator ∑ 𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑛,𝑟 is the total migrant population from national
grouping n in 1981 across Australia. These shares are used to estimate the total number of higher-
educated migrants in a given region r in year t, 𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟𝑡 . This is done by predicting the total number of
higher-educated migrants from origin country n in year t across Australia, 𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑛𝑡 , using the shares of
individuals by nationality in 1981:
60
̂ 𝐻𝐸 = ∑ 𝑆ℎ𝑎𝑟𝑒 1981 ∗ 𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠 𝐻𝐸
𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟𝑡 𝑛,𝑟 𝑛𝑡
𝑛=1
By assigning migrants from each country across different regions, an estimate of the expected number of
migrants in each region based on settlement patterns from 1981 is obtained. Baseline shares of total
employed people rather than higher-educated employed people are used in order to emphasize the role
of non-economic factors and ethnic networks in determining location choices.
Similar to the migrant population, the settlement decision of natives may not be random as natives might
also be attracted to places that are more dynamic and innovative. Furthermore, natives potentially react to
the arrival of migrants by moving out of regions where migrants disproportionately locate (OECD, 2023[3]).
Therefore, native population numbers, used in the denominator of the higher-educated migrant share, may

25
This is the earliest year in which local data can be matched to the borders used in the analysis. Additionally, a forty
years lag is useful to claim that shares are unrelated to contemporary economic factors.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

34 

also suffer from endogeneity problems. To address this concern, the current regional native population
𝑁𝑎𝑡𝑖𝑣𝑒𝑠𝑟𝑡 is also predicted based on the settlement patterns in 1981:

̂ 𝑁𝑎𝑡𝑖𝑣𝑒𝑠1981
𝑟
𝑁𝑎𝑡𝑖𝑣𝑒𝑠 𝑟,𝑡 = 1981
∗ 𝑁𝑎𝑡𝑖𝑣𝑒𝑠𝑟𝑡
∑ 𝑁𝑎𝑡𝑖𝑣𝑒𝑠𝑟

Finally, the predicted numbers of migrants and natives are used to predict the inflow of migrants, which is
used to instrument variables for the migrant inflow relative to the total population:
̂ 𝐻𝐸 ̂ 𝐻𝐸
̂ 𝐻𝐸 = 𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟,𝑡 − 𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟,𝑡−1
𝛥𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟,𝑡 ̂ 𝑟,𝑡−1 + 𝑁𝑎𝑡𝑖𝑣𝑒𝑠
̂𝑟,𝑡−1
𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠

Validity of the instrumental variable

Validity test 1: Past settlement patterns are not correlated with past regional
characteristics

To provide evidence on the exogeneity of the shares, previous research suggests investigating whether
initial shares of migrants in 1981 are correlated with regional characteristics, which can, in turn, be
correlated with current innovation (Goldsmith-Pinkham, Sorkin and Swift, 2020[28]). In this case, the initial
shares of the top-origin nationalities that drive most of the variation during the 2011-2018 period, i.e., India,
China, Philippines, and Korea26, should not be associated with regional characteristics in 1981.
Most of the regional variables are not associated with origin country shares in 1981. Table 6 provides the
results of regressions of top origin-specific shares on a set of regional characteristics reflecting regional
labour market characteristics and industry composition in 1981. These include shares of higher-educated
workers, the distribution of workers across sectors, and the logarithm of wages and employment. Columns
1 to 4 show that out of 20 coefficients, only 2 are statistically significant. Additionally, Column 5 shows that
when these four top nationalities are grouped together, none of the regional characteristics is correlated
with the aggregate origin shares. In consequence, the instrument is not correlated with regional
characteristics in 1981. Columns 6 to 8 assess the association of regional variables in 1981 with the
predicted increase in the number of higher-educated migrants, i.e., the instrument. Neither the share of
higher-educated individuals nor the sectoral shares or wage or employment levels are correlated with the
instrument. Taken together, these results provide further evidence that the instrument is affecting current
innovation only through its effect on migration flows.

Table 6. Explanatory variables in 1981

India China Philippines Korea Top 4 ∆ Predicted HE Migrant
(1) (2) (3) (4) (5) (6) (7) (8)
Share higher- 0.038 -0.116* -0.127** -0.120 -0.205 0.057 0.060 -0.102
educated
(0.067) (0.063) (0.056) (0.144) (0.135) (0.062) (0.075) (0.158)
Share primary 0.023 0.010 0.027 0.047 0.059 -0.015 -0.072
sector
(0.028) (0.026) (0.023) (0.060) (0.056) (0.047) (0.056)

26
These are the four nationalities that contribute the most to the increase in migration during the 2011-2018 period.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  35

Share secondary 0.042 -0.021 0.028 0.043 0.050 0.032 0.040

sector
(0.035) (0.033) (0.029) (0.075) (0.071) (0.060) (0.060)
Wages (logarithm) 0.100 -0.084 -0.066 -0.015 -0.050 0.020
(0.064) (0.061) (0.054) (0.139) (0.130) (0.087)
Employment -0.025 0.053 0.048 0.009 0.076 0.189
(logarithm)
(0.037) (0.035) (0.032) (0.081) (0.076) (0.135)
R2 0.511 0.686 0.721 0.463 0.753 0.628 0.633 0.666

Note: The table presents correlations of origin shares in 1981 for the main origin countries (India, China, Philippines, Korea, and their sum) and
the instrument, with respect to regional characteristics in 1981. Each column represents a dependent variable. The independent variables are
regional characteristics, which include the share of higher-educated individuals, the share of workers in the primary or secondary sector, with
the tertiary sector omitted to avoid multicollinearity and the logarithms of wage and employment levels in 1981. All specification control for
shares of higher-educated population, log of population and industry shares in 2011. All specifications are weighted by the number of employed
natives in the region at the baseline year. Robust standard errors are provided in parentheses. Statistical significance is denoted by ***, **, and
* at the 1%, 5%, and 10% levels, respectively.
Source: OECD calculations based on MADIP and Australian Census of Population 1981 (accessed July 2023).

Validity test 2: The effect is not driven by adjustment to previous inflows of

higher-educated migrants

The arrival of migrants may continue affecting regional outcomes beyond their immediate effects in the
short run. If inflows of higher-educated migrants prior to the study period have long-lasting effects, these
can be conflated with the effects of contemporaneous increases in higher-educated migrants. To overcome
this problem, accounting for past migration flows - the so-called lags - allows to separately identify the
effect of past migration episodes (Jaeger, Ruist and Stuhler, 2018[25]).
To investigate the presence of such a bias, the following model is estimated, which adds the lagged
increase in higher-educated migrants in Equation 1:
𝐻𝐸 𝐻𝐸
𝛥 𝑙𝑜𝑔 𝑌𝑟,𝑡+5 = 𝛽0 + 𝛽1 𝛥𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟𝑡 + 𝛽2 𝛥𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟𝑡−5 + 𝛽3 𝑋𝑟𝑡 + 𝛼𝑡 + 𝛥𝜀𝑟𝑡 (2)
𝐻𝐸
The equation is estimated for the period 2011 to 2016 using a 5-year change. 𝛥𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟𝑡−1 represents
the increase in higher-educated migrant population from 2001 to 2006, which corresponds to the previous
period.
Accounting for the lagged migration inflows does not change the results. Table 7 provides the estimates
of current and past increases in higher-educated migration on different measures of innovation. The
estimates for the increase in higher-educated migration are similar to the baseline specification, and the
difference is not statistically significant. In contrast, the effect of past migration inflows on current innovation
is not statistically different from zero. Taken together, this points shows that the main estimates are driven
by current migration inflows rather than long-term effects of past migration inflows. In consequence, the
main estimates are not biased by adjustment to previous changes in higher-educated migrant population.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

36 

Table 7. Controlling for past migration flows

Patents Trademarks Design
Rights
(1) (2) (3)
∆ HE Migrants 4.077* -1.825 -0.177
(2.144) (1.533) (6.460)
∆ HE Migrants, past 0.476 1.581 2.132
(2.591) (1.713) (7.023)

Note: The table presents results based on estimating Equation 2 using as outcomes either Patents (Column 1), Trademarks (Column 2) and
Design rights (Column 3) applications. The dependent variables are expressed as log changes in IP applications per worker, using 5-year
differences. The independent variable is the increase in employment due to higher-educated migrants, where the measure of high education is
tertiary education (at least with a college degree), which is instrumented. Additionally, the increase in higher-educated migrants from 2001 to
2006 is included as an additional explanatory variable. All specification control for baseline shares of higher-educated population, log of
population and industry shares. All specifications are weighted by the number of employed natives in the region at the baseline year. Robust
standard errors are provided in parentheses. Statistical significance is denoted by ***, **, and * at the 1%, 5%, and 10% levels, respectively.
Source: OECD calculations based on MADIP, IPGOD and the Australian Census of Population of 1981, 2001 and 2006 (accessed July 2023).

Validity test 3: Pre-existing growth in innovation is not associated with the current
migrant flows

Another potential concern is the presence of trends on pre-existing levels of innovation. The effect of
higher-educated migrants on innovation could be driven by regions experiencing higher growth in
innovation even before the arrival of migrants. In consequence, if the instrument is correlated with these
pre-existing levels, then the exogeneity condition is violated. To test whether the instrument is related to
previous trends in innovation, the following reduced-form regression is estimated:
̂ 𝐻𝐸 + 𝛽 𝑋 + 𝛼 + 𝛥𝜀 (3)
𝛥 𝑙𝑜𝑔 𝑌𝑟,𝑡−1 = 𝛽0 + 𝛽1 𝛥𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠𝑟,𝑡 2 𝑟𝑡 𝑡 𝑟𝑡

̂ 𝑆 , which
Where the explanatory variable is the predicted change in higher-educated migration, 𝛥𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑠 𝑟,𝑡
is the instrumental variable, and the outcome represents the change in innovation in the 2001-2006 period.
The instrument is not correlated with previous changes in IP rights applications. Table 8 shows that the
instrument is not associated with previous applications of patents, trademarks, or design rights. None of
the coefficients is significant, indicating that the instrument is not associated with trends in the outcomes.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  37

Table 8. Impact of the instrumental variable on past IP rights applications

Patents, past Trademarks, past Design rights, past
5 years 3 years 1 year 5 years 3 years 1 year 5 years 3 years 1 year
(1) (2) (3) (4) (5) (6) (7) (8) (9)
∆ Predicted HE 0.354 2.384 3.386 -0.045 0.148 -0.919 0.230 -5.612 -7.164
Migrant
(1.651) (2.041) (2.886) (1.044) (1.694) (1.734) (2.920) (5.528) (6.846)
N 88 176 616 88 176 616 88 25.357 616

Note: The table presents results based on estimating Equation 3 using as outcomes either Patents (Column 1), Trademarks (Column 2) and
Design rights (Column 3) applications. The dependent variables are expressed as log changes in IP applications per worker, using 5-year
differences, but in the period 2001-2006. The independent variable is the predicted increase in employment due to higher-educated migrants,
as calculated by the instrumental variable, where the measure of high education is tertiary education (at least with a college degree). The
columns present different time intervals for both the dependent and independent variables. Columns 1, 4 and 7 represent five-year changes
(2011-2016), Columns 2, 5 and 8 represent three-year changes (2011-2014 and 2014-2017), and Columns 3, 6, and 9 represent annual changes.
No intervals overlap. All specification control for baseline shares of higher-educated population, log of population and industry shares. All
specifications are weighted by the number of employed natives in the region at the baseline year. Standard errors, in parentheses, are clustered
at the regional level in all specifications. Statistical significance is denoted by ***, **, and * at the 1%, 5%, and 10% levels, respectively.
Source: OECD calculations based on MADIP and IPGOD (accessed July 2023).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

38 

Annex C. Robustness checks

Conducting various robustness checks allows to assess the sensitivity of the results and validates the
empirical strategy. This section provides the robustness checks, where each panel includes a different
econometric specification. All regressions are estimated using 2SLS. For comparison, Panel A provides
the main results from the 2SLS estimates in Panel B of Table 2. Overall, the magnitude and statistical
significance of the coefficients are consistent with the baseline results, proving the robustness of the
results.

Considering an alternative measure of cumulative innovation does not alter the results. The first robustness
test employs cumulative innovation activity, acknowledging that the impact of migrants on innovation might
not materialise immediately. While some previous research suggests a time frame of half a year to two
years for innovation to develop (Hunt and Gauthier-Loiselle, 2010[6]), recent studies indicate that migrants
may reach their peak in innovative activity four years after their arrival (Blit, Skuterud and Zhang, 2020[23]).
Since the specific data required to calculate this time frame for Australia is unavailable, the analysis
implements a robustness check using cumulative innovation after the increase in migration.27 Panel B of
Table 9 uses cumulative IP rights applications during the three to five years after the increase in migration
as the dependent variable. The coefficients remain robust, indicating that the influence of migration on
innovation does not dissipate immediately after migrants’ arrival.

Considering contemporaneous changes in the migrant and patenting activity allows to uncover whether
the arrival of higher-educated migrants has an immediate effect on innovation. To demonstrate leading
outcomes by one year does not drive the results, Panel C of Table 9 presents results using
contemporaneous changes in IP rights applications. The fact that one-year changes do not exhibit a
significant impact on innovation further validates the chosen strategy, as any existing impact within a
maximum of one year should be relatively moderate, if present at all. Moreover, the coefficients are smaller
compared to the baseline strategy, which is consistent with the notion that the impact of migration on
innovation takes time to materialise.

To ensure that outliers do not drive the effects, the sample is winsorised by dropping regions in the top 5%
and bottom 5% in terms of patenting activity. Panel D of Table 9 presents these results. The estimated
effect supports the baseline estimates, as it is even stronger, indicating that outliers drive the significance
and magnitude of the results.

Dropping specific regions from the sample does not affect the findings. Panels E and F of Table 9 drop
either the Outback or the largest two metropolitan areas (Sydney and Melbourne), respectively. Panel E
shows that excluding the Outback from the sample leaves results unchanged, thus showing that sparsely
populated regions do not drive the results. Panel F excludes regions which are part of the two largest
metropolitan (Sydney and Melbourne), home to 40% of the population and a large number of migrant
populations. Yet, the results remain roughly unchanged, although the levels of significance become lower
due to larger standard errors.

27
Cumulative innovation is defined as the total number of applications three or five years after the migration inflow.
For instance, for the five-year period 2011-2016, the outcome is cumulative IP rights applications from 2016 to 2021.

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  39

Aggregating regions composing the capital regions of the Australian States and Territories yield
comparable but slightly less significant estimates. Native workers may either move to other regions as a
reaction to migration, creating a problem of spatial interdependence. Such mobility might lead to spill-overs
across different regions, potentially biasing the results in Australian metropolitan areas that are
disaggregated into multiple regions. To test whether spatial interdependence drives the results, the
analysis uses the OECD’s TL3 classification. This classification merges regions of metropolitan areas in
the capital regions of Australian States and Territories into a single geographical unit, reducing the number
of regions from 88 to 50. The remaining regions remain unchanged. Panel G of Table 9 shows that the
results are robust to this choice, although the point estimate from the regression using one-year differences
becomes noisier. The point estimates are generally slightly smaller but always within the standard errors
of the baseline estimates. The positive bias of coefficients in regressions using variation across regions is
likely due to increases in innovation in metropolitan regions being driven by increases in higher-educated
migrants in neighbouring regions (Butts, 2023[33]).

Alternative definitions of innovation confirm that the total count definition does not drive the effect. A
definition used in the literature is the fractional count of patents. Panel H of Table 9 investigates the
sensitivity to using fractional rather than the total count of IP rights assigned to each region. The
magnitudes remain similar to the total count definition, thus showing that the potential under-representation
of less dense regions due to the fractional count does not affect the results.

Excluding the increase in higher-educated natives from the regression does change the estimated
coefficients of the impact of higher-educated migrants. A correlation between increases in higher-educated
migrants and higher-educated natives can potentially lead to biased results due to high multicollinearity.
However, the correlation between these two variables is 0.73 for one-year changes, 0.71 for three-year
changes and 0.68 for five-year changes. Although high, these variables are not perfectly correlated, and
Panel I of Table 3 shows that dropping the increase in higher-educated natives from the estimation does
not significantly alter the results.

Table 9. Robustness of the main results

Patents Trademarks Design rights
5 years 3 years 1 year 5 years 3 years 1 year 5 years 3 years 1 year
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Panel A: Baseline
∆ HE Migrants 4.754*** 6.351** 6.564** -1.459 -1.919 -0.245 0.754 8.725 14.621*
(1.726) (2.832) (3.264) (1.225) (2.671) (2.403) (5.727) (7.235) (8.241)
F-stat 27.1 25.3 43.1 27.1 25.3 43.1 27.1 25.3 43.1
Panel B: Cumulative innovation
∆ HE Migrants 3.063** 4.284*** -1.066 -1.057 4.202* 5.592
(1.137) (1.949) (0.651) (1.166) (2.420) (4.562)
F-stat 29.9 18.3 29.9 18.3 29.9 18.3
Panel C: Contemporaneous changes
∆ HE Migrants 3.794* 5.412*** 2.934 -0.157 -1.946 0.434 1.915 -6.413 -0.425
(2.045) (1.988) (2.469) (1.130) (1.393) (1.515) (4.982) (6.914) (7.531)
F-stat 27.1 25.3 43.1 27.1 25.3 43.1 27.1 25.3 43.1
Panel D: Dropping top and bottom five percentiles of innovation
∆ HE Migrants 6.243*** 8.751*** 9.025*** -1.951 -4.790** -1.62 2.348 13.486 10.817
(1.597) (2.778) (3.219) (1.207) (1.976) (2.281) (6.585) (8.919) (9.412)
F-stat 19.1 18.6 34 19.1 18.6 34 19.1 18.6 34
Panel E: Omitting Outback
∆ HE Migrants 4.582** 6.885** 6.872** -1.597 0.343 0.104 2.509 10.293 15.207

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

40 

(1.789) (2.956) (3.354) (1.395) (3.358) (2.920) (6.267) (7.909) (8.817)

F-stat 23.1 21.8 37.4 23.1 21.8 37.4 23.1 21.8 37.4
Panel F: Omitting Sydney and Melbourne
∆ HE Migrants 5.026** 8.111** 7.374* -1.597 0.343 0.104 -4.215 1.668 3.423
(2.178) (3.653) (4.182) (1.395) (3.358) (2.920) (3.371) (7.013) (7.637)
F-stat 26.7 21.8 28.5 26.7 21.8 28.5 26.7 21.8 28.5
Panel G: TL3
∆ HE Migrants 3.890** 5.735** 5.587 -1.482 0.199 -0.347 -4.215 1.668 3.423
(1.657) (2.374) (4.446) (1.300) (3.104) (3.388) (3.371) (7.013) (7.637)
F-stat 111.3 121.6 75.7 111.3 121.6 75.7 111.3 121.6 75.7
Panel H: Fractional count
∆ HE Migrants 4.229** 7.701*** 7.367** -0.818 -2.082 0.416 3.365 6.645 15.834*
(1.667) (2.940) (3.318) (1.150) (2.326) (2.230) (6.074) (6.920) (8.750)
F-stat 27.1 25.3 43.1 27.1 25.3 43.1 27.1 25.3 43.1
Panel I: Excluding ∆ HE Natives
∆ HE Migrants 4.788** 6.329** 5.480* -1.550 -1.940 -0.259 0.631 8.895 13.115*
(1.676) (2.827) (2.835) (1.256) (2.655) (1.946) (5.752) (7.365) (7.639)
F-stat 26.8 27.1 48.4 26.8 27.1 48.4 26.8 27.1 48.4

Note: The table presents results based on estimating Equation 1 using as outcomes either Patents (Columns 1-3), Trademarks (Columns 4-6)
and Design rights (Columns 7-9) applications. Panel A provides baseline results, while Panels B to I perform different robustness checks, as
explained in the text. The independent variable is the increase in employment due to higher-educated migrants, where the measure of high
education is tertiary education (at least with a college degree). The columns present different time intervals for both the dependent and
independent variables. Columns 1, 4 and 7 represent five-year changes (2011-2016), Columns 2, 5 and 8 represent three-year changes (2011-
2014 and 2014-2017), and Columns 3, 6, and 9 represent annual changes. No intervals overlap. All specification control for baseline shares of
higher-educated population, log of population and industry shares. All specifications are weighted by the number of employed natives in the
region at the baseline year. Time fixed-effects are applied to account for time-varying shocks affecting the entire country, except for five-year
changes due to collinearity. Standard errors, in parentheses, are clustered at the regional level in all specifications. Statistical significance is
denoted by ***, **, and * at the 1%, 5%, and 10% levels, respectively.
Source: OECD calculations based on MADIP and IPGOD (accessed July 2023).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

  41

Annex D. Definitions of higher-skilled and

higher-educated migrants

Different definitions of skills and education are used: higher-educated (at least a college degree), higher-
skilled occupations and scientists. To assign whether an individual has tertiary education or not, data from
the 2016 Census is used, using the Level of Highest Educational Attainment (HEAP) variable. Individuals
with Postgraduate Degree Level, Graduate Diploma and Graduate Certificate Level, and Bachelors Degree
Level, which represent categories 1 to 3 in the data, are the ones classified as tertiary educated.
The classification of individuals to higher-skilled occupations follows previous research and uses ANZSCO
occupation classification28. Categories 1 (Managers) and 2 (Professionals) of the ANZSCO one-digit
classification are used to identify higher-skilled individuals. Similarly, to identify scientists, occupations that
are more likely to innovate are used (Dotzel and Wojan, 2022[34]), following previous literature (OECD,
2022[35]). The following two-digit occupations are used: 13 (Specialist Managers), 23 (Design, Engineering,
Science and Transport Professionals), 26 (ICT Professionals), and 31 (Engineering, ICT and Science
Technicians).

28
More information on the Occupation and Skill Classification are provided by the ABS (1220.0 - ANZSCO -- Australian
and New Zealand Standard Classification of Occupations, 2013, Version 1.3 (abs.gov.au)).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024

42 

Annex E. Additional results

Table 10. Additional results for uneven effects of migrants with different education levels
Patents Trademarks Design rights
5 years 3 years 1 year 5 years 3 years 1 year 5 years 3 years 1 year
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Panel A: All migrants except scientists
∆ Migrants, no scientist 1.776** 2.518** 2.653* -0.438 -0.848 0.090 0.997 2.653 5.191
(0.713) (1.109) (1.653) (0.459) (1.019) (0.973) (2.271) (2.990) (3.890)
F-stat 30.5 28.8 79.4 30.5 28.8 79.4 30.5 28.8 79.4
Panel B: Tertiary-educated migrants, except scientists
∆ HE Migrants, no 7.707*** 10.349** 8.100* -2.381 -2.882 0.119 1.937 17.830 17.899
scientists
(2.843) (1.408) (4.493) (2.191) (4.712) (4.712) (9.575) (12.869) (11.881)
F-stat 21.4 21.4 49.2 21.4 21.4 49.2 21.4 21.4 49.2

Note: The table presents results based on estimating Equation 1 using as outcomes either Patents (Columns 1-3), Trademarks (Columns 4-6)
and Design rights (Columns 7-9) applications. Panel A uses all migrants but omits scientists. Panel B uses higher-educated migrants but omits
scientists. The dependent variable is expressed as log changes in IP applications per worker, using either 5, 3, or 1-year differences. The
independent variable is the increase in employment by each type of migrant. The columns present different time intervals for both the dependent
and independent variables. Columns 1, 4 and 7 represent five-year changes (2011-2016), Columns 2, 5 and 8 represent three-year changes
(2011-2014 and 2014-2017), and Columns 3, 6, and 9 represent annual changes. No intervals overlap. All specification control for baseline
shares of higher-educated population, log of population and industry shares. All specifications are weighted by the number of employed natives
in the region at the baseline year. Time fixed-effects are applied to account for time-varying shocks affecting the entire country, except for five-
year changes due to collinearity. Standard errors, in parentheses, are clustered at the regional level in all specifications. Statistical significance
is denoted by ***, **, and * at the 1%, 5%, and 10% levels, respectively.
Source: OECD calculations based on MADIP and IPGOD (accessed July 2023).

MIGRATION AND REGIONAL INNOVATION IN AUSTRALIA © OECD 2024