Reputation and Quality Eects on Wine Prices: A Comparison

Between En Primeur and Bottled Bordeaux Wine

Héla Hadj Ali, GREMAQ Toulouse

and Céline Nauges, LEERNA-INRA, Toulouse

23rd April 2003

Abstract

Introduction

The International Oce of Vine and Wine reports that the world total viticultural area reached

19.55 millions acres in 2000. Two-thirds of this surface is located in Europe (mainly in France,

Greece, Italy, Portugal and Spain) and represent 56% of the overall wine growing. France remains

a leading actor in the world wine market despite the sharp competition following the emergence

of new-world producers (South Africa, Chile, New Zealand among others) of high quality wine.

Even if the market share of France in the global wine production in volume has decreased from

38% in 1990 to 26% today,

1 the most prestigious Bordeaux wines,2 namely the ones with a long-

time established reputation like the Grands Crus Classés, remain the most famous and among

the most expensive wines worldwide. Could it be that reputation is better valued than quality

in the Bordeaux wine region? We propose to address this question by analyzing two distinct

Bordeaux wine markets: the `en primeur' market and the more traditional market for bottled

1
La revue du vin de France n 451, May 2001.
2
From a total value of sales of 3.1 billion euros in 2001, 1.2 billion come from wine sold on foreign markets
corresponding to a volume of 2.5 million of hectolitres (source: Conseil Interprofessionnel du Vin de Bordeaux).

1
wine.

The en primeur market, which is a kind of `future market' specic to the Bordeaux wine

region, occurs every year in spring. En primeur refers to the process of buying new wine several

months after the grape harvest while it is still in barrel. This market allows producers to get

liquidity before the wine is bottled,

3 but it also allows buyers to get rare wine at potential bargain

prices, hence attracting more and more nancial speculators (see Hadj Ali and Nauges, 2002).

The en primeur market in Bordeaux region may generate up to 600 million euros a year.
4

The opening of the en primeur market constitutes one of the most important events on the

Bordeaux market place. The new vintage is oered for the rst time and the prices, set by the

producers themselves, are revealed. Little is known about the way producers choose these prices.

If prices are expected to vary with objective characteristics such as appellation and ranking,
5

there is however a large uncertainty still attached to future quality as the wine is not yet matured.

Indeed, information regarding future quality of the wine mainly come from the observation of the

climatic conditions that prevailed during the growing year.

6 When the wine has been bottled

and released on the traditional (competitive) market, some additional information regarding

quality may be obtained through personal or experts' tasting. Is there any signicant eect of

the introduction of competition and the new information on quality?

Using panel data, we show that reputation and quality have very similar impacts on the en

primeur price and on the price of bottled wine. Reputation, that we choose to measure by the

3
Payment occurs at the time of the en primeur sales and bottled wine is delivered one to two years after.
4
Le Monde, 2001.
5
The appellation is a regional designation and the rank is a system of classication specic to the Bordeaux
region.
6
Tasting sessions involving wine experts may also occur before the opening of the en primeur market however
we do not observe such judgement or grade in our data set.

2
classication of wine (both in terms of ranking and appellation), is found highly correlated with

the en primeur price chosen by the producers but ranks and appellation are also highly priced

by the consumers in their purchase of bottled wine. Reputation is highly valued in particular for

the wines belonging to the most famous groups namely the Premiers Grands Crus Classés inside

both the 1855 and the Saint Emilion classications. Furthermore, the vintage year (a proxy for

climatic conditions of the growing year) is found to be a signicant determinant of both prices

(en primeur price and the price of bottled wine) while the current quality of each bottled wine

as measured by experts' ratings is not highly valued by consumers.

Results about en primeur price are new in this literature, while our results on bottled

wine conrm the ndings of Ginsburgh, Monzak and Monzak (1994), Combris, Lecocq and

Visser (1997) and Landon and Smith (1998), who also analyze price data from the Bordeaux

wine region. These three studies agree on the signicant impact of objective characteristics

(vintage, region etc.) and reputation (as measured by ranking) on wine price on the one hand,

and nd a moderate inuence of current quality (as measured by experts' grades or sensory

characteristics) on the other hand.

The paper is organized as follows. In Section 1, we dene precisely what we call reputation

and quality. In Section 2 we present the data and make some descriptive analysis of wine price

on the en primeur market and on the market for bottled wine. Econometric analysis addressing

the issue of reputation and quality impacts on wine price for the en primeur market and for the

market for bottled wine are detailed respectively in Sections 3 and 4. Section 5 summarizes the

main results and concludes.

3
1 Reputation and quality

The impact of the characteristics of a good on its price is commonly addressed through the

estimation of a hedonic function. If the good is traded on a competitive market, the estimated

coecients from the regression of price on characteristics can be interpreted as the valuation of

the good attributes by the consumers. This interpretation is obviously erroneous in the case of

en primeur wine as the price is set by the producers themselves. The estimated parameters will

instead measure the correlation between the price set by the château and the characteristics of

the wine.

In the Rosen (1974) framework, the characteristics used in the hedonic function should be the

ones for which consumers have perfect information. Objective characteristics such as appellation,

rank and vintage (or harvest) year are good candidates as they are easily accessible to the

consumers through the label.

In the present paper, we call reputation the combination of appellation, ranking and name

of the château. We do not follow the approach used by Landon and Smith (1997, 1998) and

Oczkowski (2001), who measure reputation through lagged quality scores, because we are pri-

marily interested in measuring the impact of the long-time established fame of Bordeaux châteaux

on the price of wine.

Indeed, the rst system of ranking in the Bordeaux region dates back to 1855. It was

during the Exposition Universelle that Napoléon III, the Emperor of France at the time, invited

Bordeaux's wine brokers to rank the region's wines according to price. They eventually agreed

upon a ve-tier, Classed Growths or Crus Classés (CC), classication system ranging from

4
Premiers Crus (1CC, First Growth) to Cinquièmes Crus (5CC, Fifth Growth). The nal list

consisted of 61 leading châteaux. Nearly all the selections came from the Médoc region,
7 with the

exception of the most prominent Graves château, Haut-Brion (Pessac Léognan appellation), and

the highly rated, sweet wines of Sauternes and Barsac. Sweet wines are classied in Premiers Crus

Classés (1CC, First Growth) or Deuxièmes Crus Classés (2CC, Second Growth), the château

Yquem being the only one classied in Premier Cru Supérieur (1CS, First Superior Growth).

This Classication System of 1855, largely unchanged to this day, is more and more contested

because, as explained by Combris, Lecocq and Visser (1997), Many châteaux have changed

ownership, certain châteaux have doubled even tripled their cultivation surface, new vinication

methods have been introduced. The Crus Bourgeois (CB), which are the châteaux from the

Médoc region that were not originally ranked, were not classied until 1920. By 1932, there were

444 châteaux listed and numbers have uctuated ever since. A new classication of the Crus

Bourgeois (CB) of the Médoc and Haut Médoc is currently in process and should be displayed

around June 2003.

Saint Emilion wines (Saint Emilion Grand Cru appellation) were not formally classied until

1955 (subsequently revised in 1969, 1985/86 and 1996), with a distinction between Premiers

Grands Crus Classés A (C1A, First Top Growth A), Premiers Grands Crus Classés B (C1B,

First Top Growth B) and Grands Crus Classés (C, Top Growth). Similarly, some of the wines

of Graves (Pessac Léognan appellation) were not ocially classied (CC, Classed Growth) until

1953 with others following in 1959. The Pomerol wines always refused being classied. Thus

7
Haut Médoc, Margaux, Médoc, Moulis, Pauillac, Saint Estèphe, Saint Julien are the appellation groups found
in our sample.

5
the non-classied (NC) wines in the sample are the Pomerol wines and those wines that are not

classied while belonging to classied appellation groups.

Apart from reputation, quality is expected to be another determinant of the price of wine.

The debate on the price-quality relationship has always been controversial (see Lecocq and Visser,

2002 for a good survey and some illustrative examples). The quality of a wine depends on many

factors. We may distinguish two stages in the wine-making process. First stage would be the

growing of grapes until the harvest. It is well known that the quality of the grapes (aromatic

intensity, level of acidity and sugar, quantity of tannins, etc.), for a given grape variety, is highly

inuenced by the climatic conditions of the growing year, but also by decisions taken by the

producer such as the clearing of grapes, the date of harvest, the type of harvest (manual or

mechanical), the level of investment. The second stage would be the so-called making-process

of the wine, which takes place after the harvest. The cellar master (maître de chai) has to take

decisions (grape mixing, type of barrels, duration of fermentation and maturation processes,

temperature, etc.) during the whole making-process (which lasts often eighteen months in the

Bordeaux region) that will be determinant for the future quality of the wine. These decisions

are not observable neither by the econometrician, nor by the consumer in general. Only when

the wine is bottled some additional information on quality will be publicly known and easily

accessible to the consumer through wine guides or magazines. Note that the en primeur sales

take place before the second stage is completed so the only information available to the consumers

(and to the econometrician) at the opening of the en primeur market will be the expected overall

quality of the wine based on the information on the climatic conditions that prevailed before the

6
harvest.

Climatic conditions have been included in the hedonic price equation by Ashenfelter, Ashmore

and Lalonde (1995), Byron and Ashenfelter (1995) and Ginsburgh, Monzak and Monzak (1994)

and have been proved highly signicant. In the present study, we will use vintage dummies to

control for the climatic conditions of the growing year, both in the model of en primeur price

and in the model of the price for bottled wine. The information on the quality of each particular

wine, as provided by the experts (Robert Parker here) during tasting sessions, will be available

only when the wine is in bottle and thus will be included in the hedonic equation for bottled

wine only. We argue that it is better to use a single and publicly available measure of quality

such as experts' ratings

8 instead of sensory attributes (such as the wine's aroma, body, rmness

etc.) or chemical characteristics (such as the wine's sugar and acid level) as done in Combris,

Lecocq and Visser (1997, 2000) or Nerlove (1995) because these latter characteristics are not

easily observable by the consumer.

9

2 Descriptive analysis

The data record white and red wine produced by 185 châteaux between 1982 and 1998, except

for the 1984 vintage,

10 belonging to 15 dierent appellation groups and to 11 dierent ranks.

Prices vary considerably across the wines in the data set. En primeur price and the price of

bottled wine are known for every wine that we dene as the production of one château and one

8
Oczkowski (1994, 2001), Gergaud (1998), Jones and Storchmann (2001), and Schamel and Anderson (2001)
also use grades as a measure of the quality of the wine.
9
To some extent, this has been conrmed by Combris, Lecocq and Visser (1997, 2000) who show that sensory
traits are statistically insignicant for Bordeaux and Burgundy wine.
10
This vintage has been removed from the catalogue by the broker.

7
 Table 1: En primeur prices for each rank (prices in 1990 euros equivalent)

1855 ranking Saint Emilion GC Graves

1CS∗ 1CC 2CC 3CC 4CC 5CC CB C1A C1B C CC NC
Mean 96 28 14 10 10 9 7 40 16 11 15 17
Std Dev 20 13 7 2 3 3 2 17 7 5 9 12
Nb obs. 7 127 194 83 103 138 158 28 92 153 166 205
∗
comprises Château Yquem only.

vintage. They correspond to the price of a 75cl bottle in euro equivalent (base 1990). The price

of the bottled wine is observed every quarter between June 1996 and July 2000, which amounts

to 16,586 observations in all. As our data were provided by one of the most famous broker house

in Bordeaux they can be considered as fairly representative of the market for Bordeaux wine on

the Bordeaux place.

11 For quality measures, we will refer in the present study to the overall

vintage judgement as given by Wine Spectator (see Appendix), the largest circulation U.S. wine

magazine, and the grade attributed to each wine by Robert Parker's famous guide.
12

2.1 The en primeur price

We report in Table 1 the average en primeur prices inside each ranking for the whole period

covering vintage years from 1982 to 1998 (1,454 observations in all). These statistics show that

en primeur prices match the hierarchy of ranking: inside each ranking, the greater the rank, the

higher the en primeur price. In the 1855 classication for example, prices of wine belonging to

the group of Premiers Crus Classés (1CC) lie largely above the price of those being classed from

Deuxièmes Crus Classés (2CC) to Cinquièmes Crus Classés (5CC).

Table 2, which gives some descriptive statistics on en primeur prices as observed for each

11
The broker's catalogue contains market values for wine computed from the prices proposed by all the wholesale
wine merchants on the Bordeaux place.
12
Both quality indices rise by unit intervals from a minimum of 50 to a maximum of 100.

8
 Table 2: En primeur prices for each vintage year (prices in 1990 euros equivalent)

82 83 85 86 87 88 89 90 91 92 93 94 95 96 97 98
Mean 15 17 19 15 17 13 16 14 14 9 9 10 13 17 20 17
Std Dev 16 15 15 14 13 7 10 11 15 5 5 7 8 12 15 13
Nb obs. 68 59 68 76 17 87 98 102 33 85 115 118 127 129 145 127

vintage year, shows the uctuation of en primeur prices between 1982 and 1998. Three sub-

periods may be distinguished: in the rst sub-period (82-91 vintage) the en primeur price does

not vary much from one vintage year to another. Between 1992 and 1994 the average en primeur

price decreased and increased after 1995 until 1998. Indeed, the wine of the 1991 to 1994 vintages

was not recognized by Wine Spectator as a `good' wine, as was also the case for the 1997 vintage.

However the latter was sold at the highest en primeur price for the period considered. So if

the trend of the en primeur prices matches the tasters judgement (price increases when overall

quality of the vintage is much more appreciated) at the beginning of the period, this seems no

longer true at the end as illustrated in Figure 1. In addition to the average en primeur price for

all wines and the grade attributed by Wine Spectator, we graph separately average en primeur

prices of the Premiers Grands Crus Classés from the Médoc region
13 and from the Saint Emilion

region (C1A). We set apart these two groups of châteaux as they are the most famous coming

from the Bordeaux wine region and they may be traded at prohibitive prices in auction markets

when getting old, as luxuries. Saint Emilion Premiers Grands Crus Classés A are sold at the

highest price on the en primeur market, followed by the Médoc Cinq Grands Crus (5GC). They

follow a similar trend all along the period except for the 1998 vintage (the châteaux from Saint

13
The Cinq Grands Crus (5GC) are the ve châteaux inside the 1CC group that produce red wine: château
Haut-Brion, château Late-Rothschild, château Latour, château Margaux and château Mouton-Rothschild.

9
 Figure 1: En primeur prices and Wine Spectator ratings for the 1982-1998 vintages

Emilion increased the en primeur price whereas the ones from Médoc lowered it). We clearly see

from the picture that a structural change occurred in the mid-nineties. This eect can also be

illustrated by computing the correlation coecients between the overall grade attributed to the

vintage by Wine Spectator and the average en primeur price (see Table 3).

These simple statistics give interesting insights: on the whole period the correlation coecient

between en primeur price and grade is moderate (0.21 when considering all wines). The results are

much dierent if correlation coecients are computed before and after 1994. When considering

the wine from the 1982 to 1993 vintages, en primeur price and grade are positively correlated, this

correlation being quite strong especially for the Premiers Grands Crus Classés : 0.84 and 0.86

respectively for the Cinq Grands Crus from Médoc and the Premiers Crus Classés A from Saint

Emilion. However, for the 1994 to 1998 vintages, the correlation coecient becomes negative,

10
 Table 3: Correlation between price and grade

Overall 5GC C1A

82-98 0.21 0.25 0.13
82-93 0.40 0.84 0.86
94-98 -0.41 -0.41 -0.30

showing clearly that the trend of en primeur prices no longer matches the experts' judgement.
14

2.2 The price of bottled wine

Prices of bottled wine from 1982 to 1998 vintages are observed between June 1996 and July

2000. During this period, the market for Bordeaux wine experienced feverish activity following

the entry on the market of wealthy buyers, namely multi-millionaires from emerging market

economies (Taïwan, Singapore, Brasil, Russia...). The `French paradox' assessing health benets

of wine consumption also contributed to the success of the wine market. This wealth eect

generated an extraordinary ination which is illustrated by our data (see Table 4). The average

price (all wines) per quarter almost doubled between the rst quarter (6-7/96) and the sixth

quarter (8-10/97). Because of large market uctuations we consider from now on a detrended

price in order to correctly assess the impact of the wine's characteristics on its price. A quarterly

price index is built from the regression of the log of price on quarter dummies.

Figures reported in Table 5 show that the price of bottled wine also matches the ranking.

We clearly see that the Premiers Crus Classés (1CC and C1A groups) are priced signicantly

above the others. Furthermore, it is shown from Table 6 that prices embody simultaneously an

`age' eect and a `vintage' eect. On the one hand, quality varies between vintages: `very good'

14
The correlation coecients in the column Overall of the table remain almost the same if we exclude the 5GC
and C1A wines.

11
 Table 4: Average price of bottled wine for each quarter (prices in 1990 euros equivalent)

6-7/96 8-10/96 11/96-1/97 2-4/97 5-7/97 8-10/97 11/97-1/98 2-4/98

Mean 28 33 36 43 51 54 54 52
Nb obs. 816 850 832 781 757 972 951 986

5-7/98 8-10/98 11/98-1/99 2-4/99 5-7/99 8-10/99 11/99-1/00 2-4/00 5-7/00

Mean 54 53 49 48 49 51 49 49 47
Nb obs 1,030 1,034 1,146 1,046 1,040 1,079 1,126 1,071 1,069

Table 5: Prices of bottled wine for each rank (detrended prices in 1990 euros equivalent)

1855 ranking Saint Emilion GC Graves

1CS∗ 1CC 2CC 3CC 4CC 5CC CB C1A C1B C CC NC
Mean 90 63 26 19 14 15 9 85 23 12 21 43
St Dev 21 57 20 12 7 11 5 50 13 8 20 87
Nb obs 115 1,632 2,405 965 897 1,506 1,563 380 1,176 1,760 1,709 2,478
∗
comprises Château Yquem only.

Table 6: Price of bottled wine for each vintage year (detrended prices in 1990 euros equivalent)

82 83 85 86 87 88 89 90 91 92 93 94 95 96 97 98
Mean 86 41 45 43 38 31 40 44 23 16 16 16 21 18 14 13
Nb obs. 697 381 686 826 60 1,034 1,341 1,385 286 951 1,825 2,013 2,158 1,557 1,023 363

12
ones (1982, 1990, 1995) as well as `bad' ones (1987, 1991 to 1994) are easily identiable. On the

other hand, the price gets higher when a bottle gets older: not only the wine gets better (up to

a certain limit) but it also becomes even more dicult to nd it on the market.

The `age' eect is more emphasized in Figure 2 where we graph the average price of bottled

wine at dierent ages (as measured in quarters).

15 We draw three separate curves: the average

price for wines belonging to the Cinq Grands Crus group (5GC), the Premiers Crus Classés A

group (C1A) and average price for all wines. We clearly see that the average price per bottle

increases with age. Prices of wines belonging to 5GC and C1A groups are pretty close to each

other but lie largely above average price of all wines. Note from the picture that the price starts

to increase after the age of ve years (20 quarters). The trend in prices of the top growth wines is

similar to the trend in prices of the whole average until the age of around 15 years (60 quarters).

Later, the price of the Premiers Grands Crus Classés (5GC and C1A) experiences a sharper

increase, probably due to the scarcity of these most famous bottles.

3 Econometric analysis of the en primeur price

The descriptive analysis shows that en primeur price varies with wine ranking and the overall

quality of the vintage as assessed by Wine Spectator. We propose to measure more precisely

the relative impact of reputation and quality by regressing en primeur price on wine's objective

characteristics (ranking, appellation), and its vintage year. The year dummies will act as a

proxy for climatic conditions of the growing year and thus for the overall quality of the vintage.

These are the only variables that will be used to measure the impact of quality in this model

15
The average eect of age is computed with wine from dierent vintage years and so of dierent quality.

13
 Figure 2: Price of bottled wine at dierent ages

as the grades that might be attributed to the wine during tasting sessions are not observed in

our data set. We distinguish the group of the Cinq Grands Crus (5GC), the ve red wines

of the Premiers Crus Classés (1CC) group, by including a zero-one dummy variable. We take

advantage of the panel form of the data by allowing for a random individual eect accounting

for an unobserved château eect. Note that this eect might include the unobservable decisions

taken by the producer during the growing process.

The model tting the logarithm of the en primeur price is estimated using Generalized Least

Squares. The overall regression is made on 1,454 wines produced by 163 dierent châteaux. Rank

dummies have to be interpreted with respect to the group Premiers Crus Classés including white

wine from 1CC and 1CS (we will call this group 1CCb from now on);
16 Barsac is chosen as the

base case of appellation dummies and 1982 is the base year.

16
The Premier Cru Supérieur (1CS) group includes the château Yquem only.

14
 Almost all vintage dummies are signicant at a 95% condence level. Note rst that estimated

coecients are found negative for those years in which wine was said (see Wine Spectator grades)

of `low' quality (1987, 1991 to 1994) and positive for `good' years, the only exception being the

1997 vintage (this vintage gets the highest coecient even if experts disliked it). Thus the

vintage dummies partly capture the impact of climatic conditions on the overall quality of the

vintage. Second, the coecients of vintage dummies corresponding to the years 1996 to 1998

are larger than the others in magnitude. As noted earlier, the post-1995 period corresponds to

an exceptional boom in the Bordeaux wine market. So it might be the case that, after 1995,

producers choose en primeur price considering not only the expected quality of the vintage but

also the increased number of (wealthy) buyers on the market.

Note the signicant eect of the rank on the en primeur price with respect to the group of

Premiers Crus Classés (1CCb) (most dummies are signicant at a 95% condence level), even

for the old 1855 classication which is today more and more contested: the lower the rank, the

lower the en primeur price. We nd in particular a large premium for those wines which belong

to the top groups: Cinq Grands Crus and the best-classied Saint Emilion wines (C1A group).

Any of the appellation dummies are signicant at the 95% condence level.

These results show that reputation (here appellation and above all ranking) has a higher

correlation than quality (as measured through vintage eects) with en primeur price. However,

quality in this model is measured only by vintage dummies and, as mentioned in the rst section,

quality might also be determined by some unobservable decisions taken by the producer. These

15
 Table 7: Econometric analysis of en primeur price

Coef. p-value Coef. p-value

constant 2.5942 0.0000

Vintage Appellation
Vint. 83 0.0377 0.1960 Bordeaux blanc 0.7063 0.0870
Vint. 85 0.1653 0.0000 Haut Médoc 0.1957 0.3840
Vint. 86 0.0214 0.4330 Margaux 0.0390 0.8460
Vint. 87 0.2914 0.0000 Médoc 0.2241 0.5720
Vint. 88 0.0225 0.3980 Moulis 0.1247 0.6650
Vint. 89 0.2050 0.0000 Pauillac 0.1662 0.4300
Vint. 90 0.0599 0.0200 Pessac Leognan 0.3319 0.1680
Vint. 91 0.2691 0.0000 Pomerol 0.3630 0.1240
Vint. 92 0.3695 0.0000 Sauternes 0.3351 0.0800
Vint. 93 0.2265 0.0000 Saint Emilion Grand Cru 0.1901 0.6450
Vint. 94 0.0943 0.0000 Saint Estèphe 0.1332 0.5320
Vint. 95 0.0962 0.0000 Saint Julien 0.2598 0.2000
Vint. 96 0.3662 0.0000
Vint. 97 0.5215 0.0000
Vint. 98 0.4196 0.0000

Rank
1855 classication
Cinq Grands Crus 0.6590 0.0090
2CC 0.3410 0.0600
3CC 0.5419 0.0200
4CC 0.6205 0.0060
5CC 0.7018 0.0020
CB 0.8823 0.0000
Saint Emilion
C1A 1.1435 0.0160
C1B 0.1806 0.6760
C 0.2898 0.4900
Graves
CC 0.5837 0.0220
Non-classied 0.4738 0.0290

Total number of observations 1,454

Total number of châteaux 163
R-squared 0.62

16
unobservable eects have been specied in our model as random château eects and estimations

show that they account for 81% of the total unexplained variance.

Another possible approach to weigh the relative impact of reputation and quality is to com-

pare the model tting the en primeur price using `quality' variables only (i.e. the model that

includes as explanatory variables the vintage dummies only) with the overall model including

also `reputation' variables (i.e. rank and appellation groups). The R-square of the latter reaches

0.62 whereas the former has an overall t equal to 0.12 only, which shows that the introduction of

`reputation' variables improves to a large extent the explanation of en primeur price variations.

Finally, to get a more precise idea of the impact of reputation and quality we compute

the marginal eect of each of the variables in the full model (see Table 8). As the dependent

variable has been transformed in logarithm, taking the exponential of each estimated coecient

gives the `multiplicative' marginal eect of the corresponding variable (relative to the base case

when looking at dummy variables). As an example the marginal eect of being classied in the

Cinquièmes Crus Classés (5CC) group is equal to 0.50; it means that, all other characteristics

being equal, a wine belonging to this group will be priced 0.50 times the price of a wine classied

in the Premiers Crus Classés (1CCb) group (the base case). Ranks exhibit marginal eects

ranging from 0.41 for the Crus Bourgeois (CB) to 3.14 for the Premiers Grands Crus Classés

A from the Saint Emilion region, which means that a wine classied C1A is priced around

eight times a wine ranked CB, all others things equal. The magnitude of reputation eects (as

measured by the marginal eects of the rank) is much larger than the magnitude of quality: the

marginal eects corresponding to vintage dummies vary from 0.69 (1992 vintage) to 1.69 (1997

17
Table 8: Marginal eects in the model for en primeur price

Quality eect Reputation eects

Vint. 83 1.038 1855 classication
Vint. 85 1.180 Cinq Grands Crus 1.933
Vint. 86 1.022 2CC 0.711
Vint. 87 0.747 3CC 0.582
Vint. 88 1.023 4CC 0.538
Vint. 89 1.227 5CC 0.496
Vint. 90 1.062 CB 0.414
Vint. 91 0.764 Saint Emilion
Vint. 92 0.691 C1A 3.138
Vint. 93 0.797 C1B 1.198
Vint. 94 0.910 C 0.748
Vint. 95 1.101 Graves
Vint. 96 1.442 CC 0.558
Vint. 97 1.685 Non-classied 0.623
Vint. 98 1.521
Appellation groups
Bordeaux blanc 0.493
Haut Médoc 0.822
Margaux 1.040
Médoc 0.799
Moulis 1.133
Pauillac 1.181
Pessac Léognan 1.394
Pomerol 1.438
Sauternes 1.398
Saint Emilion Grand Cru 0.827
Saint Estèphe 1.143
Saint Julien 1.297

18
vintage).

4 The price of bottled wine

On the competitive market for bottled wine, wines of dierent ages are traded between a high

number of sellers and buyers. As already spotted in the descriptive analysis, the impact of age

on price is quite complex and as this relationship could be non-linear we include as covariates in

the model the age of the wine both in level and in squares.

The model ts the logarithm of the detrended price of bottled wine, using Generalized Least

Squares on a panel data set of 185 châteaux. Unobserved heterogeneity is, as in the previous

model, accounted for by introducing a random château eect. Estimation results (see Table 9)

show that 70% of price variability is explained by the model. The overall quality of the vintage

as measured by Wine Spectator is highly valued by the consumers. Low quality wines such

as the ones from the 1987, 1991, 1992 and more particular the 1997 vintages exhibit large

negative estimated parameters. Indeed the vintage dummies capture almost exclusively the

vintage ratings attributed by Wine Spectator as the correlation coecient between estimated

parameters and grade is equal to 0.81. The quality of the wine in particular, as evaluated by

Robert Parker 's grade, is found a signicant but moderate determinant of price. The positive

coecient associated with the variable AGE conrms the intuition that a wine getting older will

be higher priced. Furthermore, the signicant and negative coecient associated with the age

squared conrms that the relationship between price and age is not linear. It is however dicult

to be more precise as age impacts prices simultaneously through a scarcity eect (mainly for

the good vintages) and a quality eect. It is important to know that this quality eect is not

19
 Table 9: Hedonic analysis of the price of bottled wine

Coef. p-value Coef. p-value

constant 1.0409 0.0000

Vintage dummies App. dummies

Vint. 83 0.5810 0.0000 Bordeaux Blanc 0.7745 0.0380
Vint. 85 0.4481 0.0000 Graves Blanc 0.1192 0.7540
Vint. 86 0.4770 0.0000 Haut Médoc 0.3903 0.0560
Vint. 87 0.9347 0.0000 Margaux 0.6848 0.0000
Vint. 88 0.5846 0.0000 Médoc 0.6832 0.0620
Vint. 89 0.4534 0.0000 Moulis 0.6362 0.0100
Vint. 90 0.4079 0.0000 Pauillac 0.8404 0.0000
Vint. 91 1.0737 0.0000 Pessac Léognan 0.7450 0.0010
Vint. 92 1.2072 0.0000 Pomerol 1.0415 0.0000
Vint. 93 0.9919 0.0000 Sauternes 0.0376 0.8250
Vint. 94 0.9533 0.0000 Saint Emilion Grand Cru 0.4946 0.0980
Vint. 95 0.8449 0.0000 Saint Estèphe 0.6928 0.0000
Vint. 96 0.9600 0.0000 Saint Julien 0.8149 0.0000
Vint. 97 1.0518 0.0000 Fronsac 0.1927 0.6050
Vint. 98 1.0119 0.0000

Rank dummies Parker's rating 0.0255 0.0000

1855 classication
Cinq Grands Crus 0.7804 0.0010
2CC 0.3735 0.0170 AGE 0.0052 0.0000
3CC 0.4925 0.0160 AGE×AGE 0.0001 0.0000
4CC 0.7048 0.0000
5CC 0.8245 0.0000
CB 1.0248 0.0000
Saint Emilion
C1A 1.1122 0.0030
C1B 0.0736 0.8150
C 0.6026 0.0450
Graves
CC 0.7373 0.0010
Non-classied 0.6233 0.0010

Total number of obs. 16,586

Total number of châteaux 185
R-squared 0.70

20
Table 10: Marginal eects in the model for bottled wine

Reputation eects Quality eects

Appellation eects Vintage eect
Bordeaux Blanc 0.461 Vint. 83 0.559
Graves Blancs 1.127 Vint. 85 0.639
Haut Médoc 1.477 Vint. 86 0.621
Margaux 1.983 Vint. 87 0.393
Médoc 1.980 Vint. 88 0.557
Moulis 1.889 Vint. 89 0.635
Pauillac 2.317 Vint. 90 0.665
Pessac Léognan 2.106 Vint. 91 0.342
Pomerol 2.833 Vint. 92 0.299
Sauternes 1.038 Vint. 93 0.371
Saint Emilion Grand Cru 1.640 Vint. 94 0.385
Saint Estèphe 1.999 Vint. 95 0.430
Saint Julien 2.259 Vint. 96 0.383
Fronsac 0.825 Vint. 97 0.349
Vint. 98 0.364

Ranking eects Parker 's rating 1.026

1855 classication
Cinq Grands Crus 2.182 age 1.005
2CC 0.688
3CC 0.611
4CC 0.494
5CC 0.438
CB 0.359
Saint Emilion
C1A 3.041
C1B 0.929
C 0.547
Graves
CC 0.478
Non-classied 0.536

21
straightforward as the time before a wine reaches maturity can vary from a wine to another (it

depends among others on the degree of tannin).

17

Reputation inuences the price of bottled wine as was the case for the en primeur price.

However, unlike the en primeur model most of the appellation dummies are signicant relative

to the base case (Barsac). The highest coecient associated with the Pomerol appellation group

could be explained because of the presence of prestigious wines such as Petrus in this non-

classied appellation group.

We repeat the same exercises as before, tting rst the model of bottled wine price using

variables inuencing quality only (vintage year, Parker's grade, and age of the wine
18 ). The

overall R-square is found equal to 0.42 (0.12 in the en primeur price model), that we have to

compare to 0.70 which is the R-square obtained when `reputation' variables are incorporated.

Second, we measure that the unobservable château eects represent 71% (88% in the en primeur

price model) of total unexplained variance of the whole model. These two results might suggest

that there is more information on quality while the wine has been bottled and that quality has

a stronger impact on the price of bottled wine than on the en primeur price.

As in the previous section we compute the marginal eect of each variable (see Table 10).

This table shows that the marginal eects of the ranking are large, ranging from 0.36 for the Crus

Bourgeois (CB) group to 3.04 for wines from Saint Emilion belonging to the group of Premiers

Grands Crus Classés A (C1A). Note that the ordering as well as the magnitude of the marginal

eects of the ranks are very similar in both models. As for the en primeur price, there is a large

17
The tannin, which is a major component in the structure of red wine, depends on grape variety. A structure
involving more tannin will benet more from a long stay in the cellar.
18
The R-square of the sub-model is unchanged whether the age of wine is included or not.

22
 Table 11: Marginal eects of cross rates and ranks on the price of bottled wine

16,586 wines 6,383 wines 6,383 wines

Parker Parker Wine Sp.

1855 classication
Cinq Grands Crus (1CC red) 1.051 1.053 1.037
1CC white 1.035 1.035 1.024
2CC 1.030 1.030 1.023
3CC 1.020 1.028 1.022
4CC 1.022 1.024 1.018
5CC 1.024 1.027 1.019
CB 1.017 1.021 1.015
Saint Emilion
C1A 1.035 1.054 1.037
C1B 1.017 1.026 1.019
C 1.012 1.031 1.015
Graves
CC 1.024 1.026 1.020
Non-classied 1.029 1.030 1.020

price premium associated with the groups of the Premiers Grands Crus Classés (5GC and C1A).

For bottled wine, appellation eects are also quite important as shown by the magnitude of the

marginal eects. Relative to Barsac, these eects range from a minimum of 0.46 for Bordeaux

Blanc to a maximum of 2.83 for Pomerol. In contrast, age and Parker 's ratings have a much

smaller impact on price. An additional quarter in age increases the price by a factor of 1.01 or,

equivalently, an additional one year for a bottle drives the price up by a factor of 1.06. The

marginal impact of the quality as announced by Parker is the following: ve additional points

would push the price up by 14%.

Reputation and quality as judged by the experts might not be independent. This may

be checked by including the cross-product of Parker's rating and wine ranking as explanatory

variables in the model. To avoid multicollinearity, we only keep in the model the cross-products

23
and we remove the rank dummies and Parker 's rating as variables in level. Marginal eects

of variables crossing Parker's rating and ranks are reported in column 2 of Table 11. We also

propose to compare the ratings attributed by Robert Parker to the ones given by the Wine

Spectator magazine. As we do not observe Wine Spectator ratings for each of the wines, we

estimate the model and compute the marginal eects on the sub-sample of the 6,383 wines for

which we have both Parker and Wine Spectator grades. Marginal eects are found in columns 3

and 4 of the same table.

The marginal eect is larger for those wines belonging to top ranks (1CC and C1A) which

would mean that consumers value more a better judgement for top quality wines than for wines

belonging to lower ranks. From columns 3 and 4, we see that marginal eects are always larger

when Parker is the advisor, which may show the stronger inuence of Robert Parker on con-

sumers' choices.

5 Conclusion

This paper addresses the issue of measuring the relative impact of reputation and quality on the

price of wine, where the reputation of a wine is dened through its appellation and its rank.

The analysis is made on a panel data set of châteaux from the Bordeaux wine region, covering

sixteen vintages and including time-series observations of prices representative of the Bordeaux

region. The analysis of reputation is highly relevant as the old system of classication dates

back to the nineteenth century. The well-established reputation of the châteaux is found to be

determinant in the setting of en primeur price by the producers but also to be highly valued by

the consumers when they purchase bottled wine. Reputation is highly priced, in particular when

24
the wine belongs to the most famous groups namely the Premiers Grands Crus Classés inside

both the 1855 and the Saint Emilion classications.

The overall quality of the vintage, as measured in the present study by Wine Spectator 's

grade, is signicantly valued by the consumers on the market for bottled wine, but this impact

is much smaller than the impact of reputation. On the en primeur market it seems that during

the last four years the way the producers choose the en primeur prices have changed. Producers'

behaviour might have been inuenced by the boom experienced by the wine market during the

second part of the decade.

The impact of Robert Parker 's ratings is found to have a small even if signicant impact on

the price of bottled wine.

We also show that reputation and quality eects are not independent and that a better grade

is more highly valued by the consumers when they consider top quality wines. Furthermore, our

results emphasize the dominant inuence of Robert Parker on consumers' choices.

In addition to pioneering results on en primeur price, this analysis conrms previous ndings

that reputation is a major determinant of the price of bottled wine and that present quality as

measured by experts' ratings has a much smaller impact. Thus despite the criticism that the

ranking system should be updated, we show that consumers still highly value the rank of the

wine in the Bordeaux region.

25
References
Ashenfelter, O., D. Ashmore, and R. Lalonde (1995): Bordeaux Wine Vintage Quality

and the Weather, Chance, 8, 714.

Byron, P. R., and O. Ashenfelter (1995): Predicting the Quality of an Unborn Grange,

Economic Record, 71, 4053.

Combris, P., S. Lecocq, and M. Visser (1997): Estimation of a Hedonic Price Equation

for Bordeaux Wine: Does Quality Matter?, Economic Journal, (107), 309402.

(2000): Estimation of a Hedonic Price Equation for Burgundy Wine, Applied Eco-

nomics, (32), 961967.

Gergaud, O. (1998): Estimation d'une Fonction de Prix Hédonistiques Pour le Vin de Cham-

pagne, Economie et Prévision, 136, 93105.

Ginsburgh, V., M. Monzak, and A. Monzak (1994): Red Wines of Medoc: What is Wine

Tasting Worth, Verona: Vineyard Data Quantication Society.

HadjAli, H., and C. Nauges (2002): Vente en Primeur et Investissement : une Etude sur

les Grands Crus de Bordeaux, LEERNA-INRA Working Paper 02.16.091, in revision for

Economie et Prévision.

Jones, G., and K. Storchmann (2001): Wine Market Prices and Investment under Un-

certainty: an Econometric Model for Bordeaux Crus Classés, Agricultural Economics, 26,

115133.

Landon, S., and C. Smith (1997): The Use of Quality and Reputation Indicators by Con-

sumers: The Case of Bordeaux Wine, Journal of Consumer Policy, 20, 289323.

(1998): Quality Expectations, Reputation and Price, Southern Economic Journal, 64,

628647.

Lecocq, S., and M. Visser (2003): What Determines Wine Prices: Objective vs. Sensory

Characteristics, Prepared for The Economics of Wine, edited by O. Ashenfelter and V. Gins-

burgh.

Nerlove, M. (1995): Hedonic Price Functions and the Measurement of Preferences: The Case

of Swedish Wine Consumers, European Economic Review, (39), 16971716.

Oczkowski, E. (1994): A Hedonic Price Function for Australian Premium Table Wine, Aus-

tralian Journal of Agricultural Economics, 38, 93110.

(2001): Hedonic Wine Price Functions and Measurement Error, The Economic Record,

77, 374382.

26
Rosen, S. (1974): Hedonic Prices and Implicit Markets: Product Dierenciation in Pure Com-

petition, Journal of Political Economy, 82, 3455.

Schamel, G., and K. Anderson (2001): Wine Quality and Varietal, Regional and Winery

Reputations: Hedonic Prices for Australia and New Zealand, CIES Discussion Paper 0103,

Adelaide University.

27
Appendix: Overall Wine Spectator ratings for vintage years
Wine Spectator magazine makes a four-tier classication: Classic, Outstanding, Good to very

good and Average ; the Classic gathering the best wines.

Vintage year Ratings

1982 95 (Classic)
1983 86 (Good to very good)
1985 93 (Outstanding)
1986 95 (Classic)
1987 76 (Average)
1988 93 (Outstanding)
1989 98 (Classic)
1990 97 (Classic)
1991 72 (Average)
1992 72 (Average)
1993 82 (Good to very good)
1994 85 (Good to very good)
1995 95 (Classic)
1996 85 (Good to very good)
1997 81 (Good to very good)
1998 90 (Outstanding)

28