Anchoring and Property Prices: The Influence of Echelle Des Crus

Anchoring and Property Prices: The Influence of Echelle Des Crus Ratings
on Land Sales in the Champagne Region of France
Olivier Gergaud
Kedge – Bordeaux Business School
Andrew J. Plantinga1
University of California, Santa Barbara
Aurélie Ringeval-Deluze
Université de Reims Champagne-Ardenne
Draft: 3-31-15
___________________
1
Corresponding author. Bren School of Environmental Science and Management, University of
California, Santa Barbara, 93106-5131. Email: [email protected]
Anchoring and Property Prices: The Influence of Echelle Des Crus Ratings on Land Sales
in the Champagne Region of France
Abstract: While a good deal of evidence for anchoring effects has been produced in
experimental settings, there have been relatively few studies testing for anchoring in actual
markets. We analyze vineyard sales in the Champagne region of France to determine whether
Echelle Des Crus (EDC) ratings are an anchor in the land market. The EDC is a set of numerical
scores for villages in the region. It was used as part of price-setting system for wine grapes that
began in 1919 and persisted until 1990. Although grape prices are now determined by the
market and the EDC no longer plays a direct role in determining them, we test whether the EDC
continues to be an anchor for participants in the land market. The econometric challenge is to
separately identify anchoring effects from the effects of relevant information the EDC may
convey about vineyard quality. We include in our hedonic regression observable characteristics
of vineyards (soils, slope, etc.) regarded as important determinants of quality and we instrument
for the EDC using the straight-line distance from each vineyard to the City of Reims, a major
center for champagne production. We draw on extensive historical evidence on the development
of Reims and the champagne industry to justify our instrument. Among other documentation, we
show that an important reason why champagne producers located in Reims in the 19th century
was to make use of a large network of limestone caves dating to Gallo-Roman period. We find
strong evidence for anchoring effects in the land market, which is further supported by analyses
of grape prices. The panel structure of our data allows us to examine whether the anchoring
effect is diminishing over time as market participants come to rely more on objective information
to determine prices. We find, instead, that the influence of the EDC appears to be increasing
over time, a result that could possibly be linked to climate change.
Anchoring and Property Prices: The Influence of Echelle Des Crus Ratings on Land Sales
in the Champagne Region of France
I.
Introduction
In a classic article, Tversky and Kahneman (1974) describe heuristic rules that people
frequently employ in making judgments under uncertainty. These rules simplify the complex
task of assessing probabilities and predicting values, but can often lead to systematic errors. One
such heuristic involves making adjustments to an initial value, or starting point, to arrive at a
final estimate. Laboratory experiments show that final estimates will often be systematically
related to initial values, even when these starting points are arbitrarily chosen (Northcraft and
Neale 1987, Ariely et al. 2003, Maniadis et al. 2014). This phenomenon is referred to as
anchoring. Anchoring has been recognized for years by economists conducting stated preference
surveys to elicit values for non-market goods (Brookshire et al. 1981, Boyle et al. 1985, Herriges
and Shogren 1996). A commonly-used format asks respondents whether or not they are willing
to pay X for a good. If the researcher’s choice of X has a systematic influence on the distribution
of responses, then respondents are said to be anchoring on X.
More recently, researchers have examined whether anchoring influences outcomes in
actual markets. Beggs and Graddy (2009) analyze fine art auctions and find that the previous
sale price for a painting has a strong influence on the current sale price. McAlvanah and Moul
(2013) show that Australian horseracing bookies fail to make sufficient adjustments to betting
odds after a late withdrawal of a horse, thus anchoring on the original odds. Several studies have
examined anchoring in the context of real estate transactions. Bokhari and Geltner (2011) and
Bucchianeri and Minson (2013) find that higher listing prices for single-family homes are
associated with higher final selling prices. Simonsohn and Loewenstein (2006) find that
migrants to a city rent more expensive apartments if their previous city had higher-priced
housing, and vice-versa. 1 A challenge in non-experimental studies is that the starting point is
outside of the researcher’s control. Correlation between the initial value and information that is
relevant to the estimation of the final value can confound measurement of the anchoring effect.
In studies that examine anchoring on past or list prices, hedonic predictions from different
periods are employed to identify effects of anchoring on current prices (Beggs and Graddy 2009,
Bokhari and Geltner 2011, Bucchianeri and Minson 2013).
Our study contributes further evidence on anchoring effects in actual markets. The panel
structure of our data allows us to examine whether the influence of the anchor changes over time,
a test that to our knowledge has not been considered in the previous literature. Were the
anchoring effect to diminish over time, for example, it would suggest that market participants
come to rely more on objective indicators of quality as they gain experience in the market.
We analyze vineyard sales in the Champagne region of France to determine whether
Echelle Des Crus (EDC) ratings influence land prices. The EDC (translated as “scale of
vineyards”) was part of a price-setting scheme for wine grapes that began in 1919. Under this
system, an appointed board of growers and champagne producers, the Comité Champagne,
would set the price for grapes during the harvest season. Grape growers would receive a
percentage of this price according to the EDC of the village in which their vineyard is located
(for example, after 1981 the scale of the EDC was 80 to 100 percent). This price-setting scheme
persisted until 1990, when it was abandoned in favor of a market system for establishing grape
prices. Currently, the EDC no longer plays a role in determining grape prices and, therefore,
should have no influence on vineyard prices, which reflect the discounted value of rents from
1
In the psychology literature, this behavior is referred to as a context effect rather than anchoring.
grape production. 2 However, given its importance throughout most of the 20th century, it is
possible that the EDC is an anchor for participants in the land market.
We make use of a large data set on vineyard sales for the period 2002-2012 to test
whether prices are anchored on the EDC. The empirical challenge we face is that the EDC is
likely to be correlated with quality attributes of vineyards that affect rents from grape production.
The main goal of the EDC was to establish a hierarchy of vineyards reflecting the quality of
grapes for making champagne. If we simply regress vineyard prices on the EDC and additional
vineyard attributes, unobservable vineyard characteristics that are correlated with the EDC rating
may bias our estimates of anchoring effects. To address endogeneity bias, we instrument for the
EDC using the distance from each vineyard to the city of Reims, the major champagne
production center. As we argue below, the cost of transporting grapes to Reims and Epernay was
an important consideration when the EDC ratings were first established in 1911 (Nollevalle
1961) but no longer has an appreciable effect on returns to grape production. We also draw on
extensive historical evidence to show that the location of Reims, and the decision by champagne
houses to locate there in the 18th century, was determined by factors other than proximity to high
quality vineyards. In particular, Reims had extensive underground caves, a remnant of limestone
quarries dating to the Gallo-Roman period, that could be converted to cellars for storing
champagne.
In the next section, we provide details on the champagne industry and the history of the
EDC. The following sections provide justification for our instrument and describe the empirical
approach. In addition to the analysis of vineyard prices, we conduct similar tests of anchoring
2
There is an important caveat to this statement. As we discuss below, the EDC still determines whether champagne
can be labeled grand cru or premier cru, which likely adds a price premium to grapes. To ensure that we identify
effects of anchoring rather than labeling, we analyze sub-samples of vineyard sales within cru classes.
using village-level data on average grape prices for the period 1991-2012. Final sections present
estimation results and conclusions.
II.
History of Champagne and the Echelle Des Crus
Champagne is produced in a region of northeastern France (Figure 1). As with other
wines and food products, the French government requires that the grapes used to make
champagne be grown in a designated area, called the appellation d'origine contrôlée (AOC).
Sparkling wines are produced elsewhere in France and throughout the world, but these cannot be
legally labeled as champagne. Compared to other wine, champagne production is labor and
capital intensive. To give champagne its characteristic bubbles, still wine is fermented a second
time inside the bottle, which involves a lengthy process to remove residual yeast. Bottles are
placed upside down and rotated a quarter turn each day for a period weeks to coax the residues to
the opening. The end of the bottle is then frozen, and the bottle is opened briefly to remove the
yeast plug, refilled, and resealed. Two-thirds of all champagne, and almost all champagne for
export, are produced by négociants or champagne houses that purchase grapes from growers in
the surrounding area (Menival and Charters 2013). The champagne houses, which include
famous firms like Moët & Chandon, the makers of Dom Pérignon, are concentrated in the cities
of Reims and Epernay (Figure 1).
Romans are the first inhabitants of the Champagne region known to have planted
vineyards, although it is likely that the Gauls who preceded them did as well. 3 During the first
millennium, the main agricultural activity in the region was wool production. Grapes were
primarily grown by peasants to make small amounts of wine for local consumption. At one
3
The sources for the historical review are Kladstrup and Kladstrup (2005), Robinson (2006), and Brown (2003).
Peter (2003). The Rise of Western Christendom. Malden, MA, USA: Blackwell Publishing Ltd.
point, Emperor Domitian in Rome ordered that all vineyards in the region be converted to grains,
a decree that remained in place for 200 years. Wine making in the region did not begin to
flourish until the 12th century when monks planted vineyards to make wine for communion,
medicinal uses, and trade. This growth in wine production, however, was halted during the 14th
and 15th centuries as feudal lords fought for control of the region. Villages and farms were
abandoned and the region’s vineyards were destroyed. In 1668, wine-making resurged when the
Benedictine monk Dom Pérignon arrived at the abbey of Hautvillers. Dom Pérignon served as
the cellar master at Hautvillers for 47 years and during that time did much to improve vineyards
and wine quality in the region. The wines from Champagne were a favorite of Louis XIV and
would continue to be supplied to the royal court at Versailles up until the French Revolution in
1789.
Meanwhile, the city of Reims had emerged as an important center for the Roman
Catholic Church and the French monarchy. Clovis, the first king of the Franks, was baptized in
the basilica at Reims in 496. The adoption of Catholicism by Clovis helped to spread
Christianity across France and established the cathedral in Reims at the traditional site for the
coronation of French kings. Before the French Revolution, the Catholic Church played a central
role in the production and distribution of champagne. An outcome of the Revolution was the
nationalization of the Church’s property, including its extensive holdings of vineyards in the
Champagne region. When the champagne houses were established in the 18th century, many
would locate in Reims, the major city in the region. Besides being the major center for
commerce in the region, Reims offered another advantage to champagne houses. 4 Beneath the
city was a large network of caves dating back to the Gallo-Roman period. These caves were the
remains of limestone quarries that had been dug to provide stone for buildings in Reims,
4
http://www.maisons-champagne.com/en/orga_prof/presentation_umc_gb.htm
including its magnificent cathedral. The caves provided perfect conditions for storing
champagne and would be used as cellars by such famous houses as Charles Heidsieck, Pommery,
Ruinart, Taittinger and Veuve Clicquot Ponsardin. The other concentration of champagne
houses was found in Epernay. Like Reims, Epernay is an ancient city, but it emerged as the
trading post for champagne due to its location on the Marne River, which provided a direct
shipping route to Paris.
The champagne houses prospered during the 1800s, but the century would end with
violent protests by growers, ultimately resulting in a price-setting system for grapes. In 1890,
champagne producers could import up to 49% of their grapes from outside the region, and were
not prevented from using other fruits, such as apples. Combined with a string of bad harvests,
many growers in the region, most of whom farmed small plots, were facing bankruptcy.
Demonstrations by thousands of growers were held to protest the prices paid by the houses and
the policies of the national government. These demonstrations eventually turned violent,
resulting in the burning and looting of champagne houses and government buildings and
eventual military occupation.
In response to the “Champagne Riots,” the Champagne production area was formally
designated and the EDC was established. Under the first official version of the EDC, adopted in
1911, 145 villages were given a rating between 46 and 100 based on the prices received during
the previous 20 harvests (Table 1). Starting in 1919, the Comité Champagne, with
representatives from the champagne houses and the growers union, agreed to set the price
received by growers using an updated EDC. The board would determine the price for the top
villages, those with a 100% rating, and growers would receive a percentage of this price
according to the EDC of their village. The EDC was updated a number of times during the 20th
century to include new villages and to compress the scale. After 1981, for example, the EDC
ranged from 80 to 100. The EDC was used to set prices until 1990, when a free market system
for determining grape prices was adopted. The only remaining function of the EDC is to
determine the classification of wines as grand cru and premier cru. A champagne producer can
label their wine grand cru if the grapes come from a village with an EDC of 100. Premier cru
wines must be made with grapes from villages ranked 90-99. Champagne houses typically blend
wine with grapes sourced from many villages and even several vintages. Labeling their product
grand cru or premier cru is a way of signaling quality to consumers. Gergaud (1998) and later
Menival and Charters (2013) find evidence of higher prices for grapes from grand cru and
premier cru villages.
III.
An Instrument for the Echelle Des Crus
Our analysis of recent prices for vineyards and grapes will require an instrument for the
EDC. We argue in this section that the straight line distance from vineyards, or village centers,
to Reims satisfies the requirements of relevance and exogeneity. As noted in section II, grapes
are transported from vineyards to Reims where they are used to make champagne. The cost of
transporting grapes to Reims was an important consideration when the EDC was first established
in 1911 (La Champagne Viticole, 1961). At that time, grapes were transported by horse and cart
over dirt roads, increasing the potential for losses of grapes due to damage and spoilage.
Because the Comité Champagne specified an undelivered price (i.e., the price a grower would
receive at the vineyard), it is likely that the EDC would have been adjusted for distance to
Reims. Although we do not have direct evidence that this adjustment was made, we can test
whether the 1911 EDC and subsequent versions vary with distance to Reims.
We present regressions of each version of the EDC on straight-line distances to Reims
and Epernay, the squares of these variables, and three weather variables (Table 2). Ideally, our
distance measure would reflect travel times over the road network as it existed at the time. We
do not have this information and so use the straight-line distance, which seems like a better
alternative than using the current network distance given road improvements over the last
century. Also, we do not include village-level controls for characteristics of vineyards (e.g.,
soils, slope) because the extent and composition of vineyards changed greatly from 1911 to the
present and we only have information on current vineyards. We do include controls for weather
on the assumption that these variables have been relatively constant over time. Specifically, we
measure the average annual temperature, rainfall, and frost days recorded at the nearest weather
station to the village center over the period 1994-2011. 5 The results show that the 1911 EDC
declines in distance to Reims at a diminishing rate, and that a similar relationship holds for most
of the other versions of the EDC. Moreover, the coefficient on the distance to Reims is largest in
1911, and remains relatively large for the 1920 and 1944 EDCs, before stabilizing in versions
after 1945. The weather variables also have significant effects on most versions of the EDC,
which is consistent with the rating depending in part on grape quality.
Our argument for exogeneity has three parts. First, we establish that with modern
transportation infrastructure, the costs of transporting grapes within the region are now
negligible. Table 3 provides a breakdown of the costs of transporting a truckload of grapes from
three origins, the Marne, Aube, and Aisne 6 regions (see Figure 1). Even from the most distant
transportation costs are less than 0.5% of the value of the shipment. For the Marne and Aisne
5
Daily weather information from 33 stations spread across the Champagne region were used to compute these
variables.
6
Vineyards in Aisne are all located around the city of Château-Thierry, in the westernmost section of the Vallée de
la Marne region).
regions, the figure is 0.13% and 0.21%, respectively. This evidence suggests that differences in
the costs of transporting grapes to Reims will not be reflected in prices for vineyards and grapes.
Second, the quality of a site for growing wine grapes depends on factors determined on
geological time scales (e.g., type of soil, elevation, slope, and aspect) or by global processes such
as climate. In the Champagne region, there is little that growers can do to improve the quality of
their sites. Therefore, we can treat the spatial distribution of high quality vineyards as predetermined and exogenous.
The final part of the argument is that champagne houses located in Reims for reasons
other than proximity to high quality vineyards. The city was founded more than a millennium
before wine was successfully grown in the region, eliminating the possibility that the site for the
city was originally chosen because it was close to good vineyards. When private champagne
houses were established in the 18th century, Reims was the major center for commerce in the
region due in large part to the Church’s predominance in economic affairs at the time. The
Church’s presence in Reims traced back centuries to the adoption of Catholicism by Clovis and
continued through time as a long succession of French kings were crowned in the city’s
cathedral. Moreover, Reims had limestone caves that could be converted to cellars. These caves
were the remnants of quarries that had been dug to provide building materials for the city.
In sum, our argument for exogeneity is that errors in current vineyard and grape prices
should not be correlated with distances to Reims because transportation costs are a negligible
fraction of the value of shipments. The location of high quality vineyards is predetermined and
the location of champagne houses in Reims, as well as the location of the city itself, is the result
of historical events unrelated to vineyard quality. Finally, we measure the straight line distance
to Reims rather than the road network distance in case access to vineyards might have influenced
decisions about where to place roads.
IV.
Empirical Approach
A. Vineyard sales
Our first set of regressions makes use of data on vineyard sales. The dependent variable
is the log of the nominal sale price of parcel i in year t, which is denoted PVit . The independent
variables include the parcel area and other attributes including soil type, elevation, slope, aspect,
and weather variables. Such characteristics are thought to contribute to terroir, the special
characteristics of a place that impart unique qualities to wine. Their effects on vineyard and
wine prices have been investigated in earlier studies (Ashenfelter and Storchmann 2010,
Gergaud and Ginsburgh 2010, Cross et al. 2011). All of the parcel attributes are time-invariant
and collected in Xi . In the case of the weather variables, we expect land prices to depend on
long-term historical averages rather than annual values.
The variable of interest is the EDC, which is measured at the village level. We denote
this variable EDCv (i ) where v(i) is a mapping from parcel i to the village v in which it is located.
Our vineyard price data span the period 2002-2012. There were minor adjustments to the EDC
made between 2002 and 2003. We used the earlier version of the EDC for the 2002 observations
and the later version for observations after 2003, though this introduces little temporal variation
in EDCv (i ) .
We estimate two versions of the vineyard price model. The first version uses
observations for vineyards in premier cru villages. As discussed above, champagne can be
labelled as premier cru if the grapes come from villages with EDC ratings between 90 and 99. If
producers can sell premier cru wines for more, this should increase the prices of vineyards in
premier cru villages. Because the price premium is common to all vineyards in these villages,
we avoid the need to control for its influence by estimating the model with a sub-sample of
observations. We can still test for anchoring effects because the EDC varies across premier cru
villages. Our premier cru model is specified:
(1)
PVit = α + Xiβ + γ EDCv (i ) + δ t + ε it , 90 ≤ EDCv (i ) ≤ 99
where α is a constant term, β and γ are parameters, the δ t are a set of annual dummies, and ε it
is a random disturbance term. Our instrument for EDCv (i ) is the distance from each vineyard to
the center of Reims.
The second version of the model uses observations for vineyards in villages that have
neither a premier cru nor a grand cru designation. These villages, referred to as autre cru, have
an EDC rating between 80 and 89. Our autre cru model is specified:
(2)
PVit = α + Xiβ + γ EDCv (i ) + δ t + ε it , 80 ≤ EDCv (i ) ≤ 89
We do not estimate a grand cru version of the model because the EDC because there is no
variation in the EDC within these villages.
B. Grape prices
We provide further evidence on anchoring effects by regressing grape prices on the EDC
and other co-variates. The premier cru model is specified:
(3)
PGvt = α + Xvβ + Ζ vt θ + γ EDCv + δ t + ε vt , 90 ≤ EDCv ≤ 99
where PGvt is the average grape price in village v in year t and Xv are average time-invariant
characteristics of vineyards in village v. Annual grape yields, and thus annual prices, will be
influenced by the weather during the growing season. Therefore, we include in Z vt
contemporaneous measures of weather variables. The EDC variable is measured in the same
way as above, with the earlier version of the rating used for observations prior to 2003, and viceversa. The instruments for the EDC are the same distance measures that we used in the vineyard
model. We also estimate the grape price model for the autre cru villages:
(4)
PGvt = α + Xvβ + Ζ vt θ + γ EDCv + δ t + ε vt , 80 ≤ EDCv ≤ 89
As with the premier cru model, we estimate (4) using different samples that correspond to
different versions of the EDC.
C. Changes in the effects of the EDC over time
We test whether the EDC has had different marginal effects on vineyard and grape prices
over time. The grape price data span time periods when both the 1985 and the 2003 version of
the EDC were in effect, and so a natural division is to estimate separate models for the earlier
and later versions of the EDC. We have a shorter time series for the vineyard price data, but
many more cross-sectional observations, and so we estimate separate models for each year.
V.
Data
The transactions data is from the Sociétés d’aménagement foncier et d’établissement
rural (SAFER). The data set includes the price, the date of the sale, and the size and location of
the parcel for 12,370 sales in the Champagne region over the period 2002 to 2012. We include
only sales for which the dominant use of the land is vineyards and exclude any sales that include
buildings. Each sale is matched to additional data sets to identify weather and geographical
features of the parcels, including altitude, slope, aspect, and soils. Using the location of each
sale and village boundaries, we identify the EDC for each parcel. Finally, we compute the
straight-line distance from each parcel to the center of Reims. Summary statistics for the data
used in the vineyard sale analysis are provided in Tables 4 and 5. The dominant soil type is
limestone and most vineyards have slopes between 2% and 20%. In addition, southern and
eastern orientations are most common. The average nominal price for vineyards increased
almost 2.5 times over the period of analysis.
A separate village-level data set includes average grape prices and EDC ratings, provided
by Comité Champagne (CIVC). At the village level, the price data are available from 1991 to
2012, except for 1992. For the region as a whole, average grape prices declined in the early
1990s during the global economic recession and then increased steadily after 1993 as a result of
growing demand for champagne (Table 5).
VI.
Estimation Results
A. Vineyard sales
Results for the vineyard price model are reported in Table 6. As a reference point, the
model 1 is estimated with least squares using all of the observations. The coefficient on the EDC
is approximately 0.02 and significantly different from zero at the 1% confidence level. The
results change considerably when we restrict the sample to parcels in autre cru or premier cru
villages and instrument for the EDC. With the autre cru sample, the OLS coefficient on the EDC
(model 2) is small and insignificant, however, the IV estimate (model 4) is much larger (0.07)
and significantly different from zero at the 1% level. The result indicates that a one-unit increase
in the EDC raises the price of vineyards by 7%, holding parcel size and other factors constant.
The results for the premier cru sample are similar. The IV estimate of the coefficient on the
EDC (model 7) is 0.09 and significantly different from zero, and approximately three times the
size of the OLS coefficient. For the autre cru and premier cru models, the C-statistic has a pvalue less than 0.000, confirming that the OLS estimates are affected by the presence of the
endogenous regressor.
A number of the other controls are found to have significant effects on vineyard prices.
Not surprisingly, prices are increasing in the size of the parcel and the year dummies (not
reported) show a strong upward trend in the nominal price of vineyards, consistent with Table 5.
Several of the soil variables have significant effects, and altitude affects autre cru parcels.
Differences in slope and aspect have limited influence. Finally, weather variables have
significant effects on autre cru parcels but not on premier cru parcels. These results are broadly
consistent with earlier studies by Gergaud and Ginsburgh (2010) and Cross et al. (2011) that find
limited effects of land characteristics on wine quality and vineyard prices.
B. Grape prices
The grape price results are presented in Table 7. When we use all of the data, the OLS
estimate of the coefficient on EDC (model 1) is 0.005. The results change considerably when we
restrict our sample to the autre cru villages. The IV estimate of coefficient on the EDC variable
(model 4) is 0.02 and significantly different from zero, compared to an OLS estimate (model 2)
of 0.003. A one unit increase in the EDC raises grape prices by 2%. As expected, the effect is
smaller than what we found for the vineyard prices, which measure the capitalized value of the
stream of profits from grape production. For the premier cru sample, the OLS and IV estimates
of the EDC coefficient (models 7 and 9) are both equal to 0.007 and significantly different from
zero. The Hansen/Sargan/C test indicates that endogenous regressor should not be treated as
exogenous. Many of the vineyard characteristics have significant effects on grape prices.
C. Changes in the effects of the EDC over time
Estimates of the annual coefficients on the EDC variable are presented in Table 8. For
autre cru vineyards, the coefficient is large and positive in 2003, but otherwise not significantly
different from zero for most years prior to 2007. After that, the coefficients show relatively little
variation, tending toward a value of about 0.09. Estimates for premier cru vineyards are more
stable. Except for 2002 and 2012, the estimates are all significantly different from zero.
Between 2003 and 2010, the coefficients are approximately equal to 0.07, but then rise in 2010
and 2011. If anything, these results suggest that the effect of the EDC is increasing over time.
The grape price models show a somewhat different pattern of change in the influence of
the EDC (Table 7). For the autre cru villages, the estimated coefficient on the EDC is 0.026 for
the period 1991-2002, declining to 0.015 for the period 2003-2012. For the premier cru villages,
the effect of the EDC increases 0.003 to 0.007.
VII.
Conclusions
While a good deal of evidence for anchoring effects has been produced in experimental
settings, there have been relatively few studies testing for anchoring in actual markets. One of
the reasons for this is likely the difficulty of separately identifying anchoring effects from the
effects of relevant information the anchor may convey. We estimate hedonic regressions that
include the anchor (the EDC) and a large number of observable attributes that are widely
regarded as important for determining the quality of grapes and vineyards. We instrument for the
anchor to avoid endogeneity bias that may arise from correlation between the anchor and
unobservable quality attributes. Overall, we find strong evidence for anchoring. In all of the IV
specifications we tested, the EDC is found to have a positive and significant effect on grape or
vineyard prices.
Our results also suggest that the effects of anchoring are large in magnitude. For the
vineyard price models, the estimated EDC coefficient is roughly 0.08 and the average EDC
rating is approximately 90. Given an average vineyard price (in logs) of 13.5 euros, this
indicates that the EDC accounts for about one-half of the log price. For grape prices, the
contribution of the EDC to log prices is even larger, about 88%. Another way to gauge the
magnitude of the anchoring effect is to consider the price dispersion due to the EDC. Within the
autre cru category, vineyards with a rating of 89 gain a price premium of about 70% compared to
a vineyard with a rating of 80. For the premier cru category, the price gain for a 10-point
increase in the rating is about 90%. These seem like large differences given that the EDC has no
actual effect on grape prices within either category.
The role of the EDC in determining prices for grapes ended in 1990 and, yet, our results
indicate that it continues to have a strong influence on the grape and vineyard markets. One
might expect that over time the effect of the EDC would diminish as market participants rely
more on observable information about the determinants of grape and vineyard values. We do not
find this to be the case in our application. With the exception of grapes prices from autre cru
villages, we find the effects of the EDC are getting stronger over time. One explanation is that
observable characteristics are becoming less reliable as signals of value, leading market
participants to depend more on the EDC for determining prices. Such a trend could be linked to
climate change that alters the effects of soils, aspect, elevation and other traditional determinants
of terroir on grape quality.
References
Ariely, D., Loewenstein, G., and D. Prelec. 2003. Coherent Arbitrariness: Stable Demand Curves
without Stable Preferences.” Quarterly Journal of Economics 118 (1): 73–105.
Ashenfelter, O., and K. Storchmann. 2010. Using Hedonic Models of Solar Radiation and
Weather to Assess the Economic Effect of Climate Change: The Case of Mosel Valley
Vineyards. Review of Economics and Statistics 92(2): 333-349.
Beggs, A., and K. Graddy. 2009. Anchoring Effects: Evidence from Art Auctions. American
Economic Review 99(3): 1027-1039.
Bokhari, S., and D. Geltner. 2011. Loss Aversion and Anchoring in Commercial Real Estate
Pricing: Empirical Evidence and Price Index Implications. Real Estate Economics 39(4):
635-670.
Boyle, K.J., Bishop, R.C., and M.P. Welsh. 1985. Starting Point Bias in Contingent Valuation
Bidding Games. Land Economics 61(2): 188-194.
Brookshire, D.S., d’Arge, R.C., Schulze, W.D., and M.A. Thayer. 1981. Experiments in Valuing
Public Goods. In Advances in Applied Microeconomics, Vol. 1, ed. V. Kerry Smith.
Greenwich, CT: JAI Press.
Brown, P. 2003. The Rise of Western Christendom. Malden, MA, USA: Blackwell Publishing
Ltd.
Bucchianeri, G.W., and J.A. Minson. 2013. A homeowner’s dilemma: Anchoring in residential
real estate transactions. Journal of Economic Behavior and Organization 89: 76-92.
Cross, R., Plantinga, A.J., and R.N. Stavins. 2011. What is the Value of Terroir? American
Economic Review Papers and Proceedings, 101(3): 152-56.
Gergaud, O. 1998. Estimation d’une Fonction de Prix Hédonistiques pour le Vin de
Champagne, Economie et Prévision, 136: 93-105.
Gergaud, O., and V. Ginsburgh. 2010. Natural Endowments, Production Technologies and the
Quality of Wines in Bordeaux. Does Terroir Matter? The Economic Journal, 118 (June),
F142-F157.
Herriges, J., and J. Shogren. 1996. Starting point bias in dichotomous choice valuation with
follow-up questioning. Journal of Environmental Economics and Management 30:112131.
Kahneman, D. 1992. Reference Points, Anchors, Norms, and Mixed Feelings. Organizational
Behavior and Human Decision Processes 51: 296-312.
Kladstrup, D., and P. Kladstrup. 2005. Champagne: How the World’s Most Glamorous Wine
Triumphed Over War and Hard Times. HarperCollins Publishers.
Maniadis, Z., Tufano, F., and J.A. List. 2014. One Swallow Doesn’t Make a Summer: New
Evidence on Anchoring Effects. American Economic Review 104(1): 277–290.
McAlvanah, P., and C.C. Moul. 2013. Rhe house doesn’t always win: Evidence of anchoring
among Australian bookies. Journal of Economic Behavior and Organization 90: 87-99.
Menival, D., and S. Charters. 2013. The impact of geographic reputation on the value created in
Champagne. Australian Journal of Agricultural and Resource Economics 58: 171–184.
Nollevalle, J., 1961. 1911, L’agitation dans le vignoble champenois. La Champagne Viticole
(January): 6-30.
Northcraft, G., and M. Neale. 1987. Experts, amateurs, and real estate: an anchoring and
adjustment perspective on property pricing decision. Organizational Behavior and Human
Decision Processes 39: 84–97.
Robinson, J. 2006. The Oxford Companion to Wine. Oxford University Press, 3rd Edition.
Simonsohn, U., and G. Loewenstein. 2006. Mistake #37: The Effect of Previously Encountered
Prices on Current Housing Demand. Economic Journal 116 (508): 175–99.
Tversky, A., and D. Kahneman. 1974. Judgment Under Uncertainty: Heuristics and Biases.
Science 185(4157):1124-31.
Figure 1. The Champagne Region
Source: http://www.champagnesdevignerons.com/Découvrir-la-Champagne/Les-4-grandes-régions-deChampagne.html
Table 1. Historical Evolution of the Echelle des Crus System
Version
Number of Villages1
Mean Rating
Minimum
Rating
Maximum
Rating
Correlation with
1911 version
1911
1920
1944
1945
1971
1972
1980
1981
1985
2003
145
145
170
324
326
326
326
326
346
352
65
73
73
80
83
83
83
84
85
86
46
56
58
70
77
77
78
80
80
80
100
100
100
100
100
100
100
100
100
100
0.99
0.98
0.98
0.98
0.96
0.96
0.96
0.93
0.90
1
Some of the villages have separate ratings for red and white grapes
Table 2. The Effects of Straight-Line Distance and Weather Variables on the Echelle
des Crus Ratings
Variable
Distance to Reims
Distance to Reims squared
Distance to Epernay
Distance to Epernay squared
Average annual temperature
Average annual rainfall
Average annual frost days
Constant
Variable
Distance to Reims
Distance to Reims squared
Distance to Epernay
Distance to Epernay squared
Average annual temperature
Average annual rainfall
Average annual frost days
Constant
1911
1920
1944
1945
1971
Coef p -value
Coef p -value
Coef p -value
Coef p -value
Coef p -value
-1.187
0.021
-0.980
0.016
-0.890
0.001
-0.021
0.737
-0.095
0.059
0.025
0.018
0.020
0.014
0.015
0.000
-0.001
0.383
0.001
0.429
0.496
0.404
0.358
0.435
0.539
0.235
-0.384
0.000
-0.298
0.000
-0.039
0.012
-0.030
0.012
-0.034
0.007
0.003
0.000
0.002
0.026
0.535
0.933
0.125
0.980
3.720
0.404
-2.976
0.032
-2.031
0.105
-4.077
0.000
-3.368
0.000
-3.481
0.000
-0.569
0.012
-0.359
0.053
-3.460
0.709
-3.209
0.654
-0.527
0.937
-5.098
0.010
-4.582
0.011
151.908
0.048 149.134
0.013 105.922
0.043 143.162
0.000 129.132
0.000
1972
1980
1981
1985
2003
Coef p -value
Coef p -value
Coef p -value
Coef p -value
Coef p -value
-0.108
0.036
-0.095
0.054
-0.067
0.135
-0.098
0.032
-0.159
0.001
0.001
0.334
0.001
0.369
0.000
0.495
0.001
0.042
0.002
0.000
-0.290
0.000
-0.276
0.000
-0.248
0.000
-0.240
0.000
-0.129
0.005
0.002
0.040
0.002
0.031
0.002
0.016
0.001
0.312
-0.001
0.046
-2.397
0.064
-2.010
0.103
-1.263
0.264
-1.423
0.213
-2.319
0.051
-0.283
0.153
-0.289
0.140
-0.300
0.123
-0.328
0.087
-0.048
0.826
-5.265
0.006
-4.742
0.010
-3.732
0.031
-4.697
0.007
-6.548
0.000
133.244
0.000 127.772
0.000 117.169
0.000 121.807
0.000 130.334
0.000
Table 3. Costs of transporting grapes within the Champagne region (2014 estimates)
Origin
Cost component (in euros)
Marne
Aube
Aisne
Cost of gasoline per km (toll included)
0.501
0.501
0.501
Cost of gasoline per km (w/o toll)
0.433
0.433
0.433
Vehicle cost + cost structure (per day)
169.18
169.18
169.18
Labor cost (per hour)
24.42
24.42
24.42
Average distance to Reims
40.78
153.44
62.14
Total cost (Reims)
470.49
518.29
472.17
Average load (in liters – 9 marcs per load)
22,950
22,950
22,950
Total value of a load (2011)
120,717
112,455
114,062
Average grape price
5.26
4.9
4.97
(Transp. Cost to Reims / Total Value) × 100
0.13
0.46
0.21
Source: http://www.cnr.fr/fr/Indices-Statistiques/Citerne-liquide-alimentaire-40-T/Referentiel-prix-de-revient,
http://www.fierdetreroutier.com/zoom/vendanges.php.
Table 4. Vineyard data – Summary Statistics
Mean
Standard
Minimum
Maximum
Deviation
Soil type:
Limestone
0.303
0.282
0
1
Alluvial Fans
0.141
0.136
0
1
Marl
0.141
0.225
0
1
Sedimentary Rocks
0.086
0.153
0
1
Silica Sand
0.109
0.164
0
1
Ferric Iron-bearing Sediments
0.102
0.128
0
1
161.226
33.729
98,143
346
0% ≤ Slope coeff. < 2%
0.005
0.013
0
0.5
2% ≤ Slope coeff. < 10%
0.306
0.227
0
1
10% ≤ Slope coeff. < 20%
0.443
0.189
0
1
20% ≤ Slope coeff. < 30%
0.12
0.116
0
1
Slope coeff. ≥ 30%
0.125
0.102
0
1
North
0.053
0.077
0
1
Northeast
0.119
0.135
0
1
East
0.144
0.123
0
1
Southeast
0.183
0.16
0
1
South
0.152
0.153
0
1
Southwest
0.101
0.117
0
1
West
0.051
0.08
0
1
Northwest
0.051
0.075
0
0.824
North
0.038
0.059
Note: Statistics based on 12,433 vineyard transactions
0
0.667
Altitude
Slope coefficients
Orientation
Table 5. Grape prices – Vineyard Prices Annual statistics
Grape Prices
Year
Mean
Std.
Vineyard Prices
Min
Max
N
Mean
Dev.
Std.
Min
Max
Dev.
1990
4.39
0.26
4.04
5.18
1991
4.01
0.24
3.72
4.57
1993
2.69
0.18
2.5
3.13
1994
2.8
0.19
2.59
3.24
1995
2.94
0.2
2.71
3.39
1996
3.18
0.21
2.93
3.66
1997
3.18
0.21
2.93
3.66
1998
3.31
0.22
3.05
3.81
1999
3.37
0.23
3.11
3.89
2000
3.54
0.21
3.32
4
2001
3.54
0.21
3.32
4
2002
3.64
0.24
3.4
4.22
1,411
380,189
252,067
3830
2,543,785
2003
3.77
0.24
3.38
4.38
1,127
507,847
263,346
2.37
6,700,000
2004
4.09
0.22
3.88
4.66
1,309
547,946
165,473
66,696
895,037
2005
4.12
0.23
3.9
4.77
1,208
613,125
174,710
22,301
989,999
2006
4.35
0.22
4.1
4.9
1,187
643,457
206,274
95,517
4,380,954
2007
4.79
0.23
4.5
5.33
1,234
713,308
195,382
125
1,809,211
2008
5.08
0.27
4.75
5.67
915
814,537
215,781
158,878
1,663,893
2009
4.88
0.25
4.56
5.43
830
830,598
185,125
184,615
2,840,001
2010
5.02
0.24
4.73
5.54
793
824,238
235,314
89.29
2,749,142
2011
5.19
0.26
4.89
5.77
1,107
883,383
270,917
43.29
4,438,983
1,309
986,143
438,899
1.49
1.12e+07
2012
Table 6. Log of Vineyard Price Regressions
(OLS and IV methods – Instrument for EDC is distance to Reims)
EDC ratings (85 and 03)
(1)
OLS
All
(2)
OLS
AC
0.019***
(20.90)
0.006
(0.97)
Distance to Reims
Size (in hectares)
-0.015*** 0.020***
(-79.66)
(3.44)
Limestone
0.016
-0.006
(0.42)
(-0.13)
Alluvial Fans
-0.030
-0.103**
(-0.71)
(-2.12)
Marl
-0.061
-0.018
(-1.52)
(-0.38)
Sedimentary Rocks
-0.036
0.031
(-0.74)
(0.70)
Silica Sand
-0.020
-0.034
(-0.49)
(-0.81)
Ferric Iron-bearing Sediments -0.034
-0.148***
(-0.80)
(-2.60)
Altitude (mean)
-0.001*** -0.001***
(-3.73)
(-4.28)
Northeast
0.026
0.108**
(0.92)
(2.20)
East
0.037
0.071
(1.39)
(1.28)
Southeast
0.007
-0.007
(0.24)
(-0.12)
South
0.022
0.054
(0.74)
(0.93)
Southwest
0.036
0.056
(1.20)
(0.95)
(3)
First stage
AC
-0.023***
(-32.87)
-0.006
(-0.29)
0.186
(0.69)
-0.057
(-0.21)
-1.045***
(-3.93)
-0.843***
(-3.17)
-0.788***
(-2.94)
-0.429
(-1.57)
-0.001
(-1.62)
0.776***
(8.69)
1.514***
(17.95)
1.536***
(19.80)
1.768***
(22.86)
1.416***
(16.70)
(4)
IV
AC
(5)
OLS
PC
(6)
First stage
PC
(7)
IV
PC
0.071*** 0.031***
(5.93)
(5.01)
0.094***
(7.07)
0.025*** 0.030***
(3.94)
(5.39)
-0.009
0.032
(-0.18)
(0.48)
-0.089*
0.055
(-1.76)
(0.73)
0.017
-0.120*
(0.33)
(-1.86)
0.073
-0.336*
(1.46)
(-1.95)
0.007
-0.020
(0.15)
(-0.25)
-0.122**
0.044
(-2.11)
(0.70)
-0.001**
-0.000
(-2.44)
(-0.06)
0.040
-0.059
(0.91)
(-1.36)
-0.038
-0.007
(-0.81)
(-0.18)
-0.108*
0.018
(-1.89)
(0.57)
-0.069
-0.009
(-1.36)
(-0.17)
-0.049
0.042
(-0.92)
(1.20)
0.106***
(30.27)
-0.172*** 0.042***
(-6.77)
(6.11)
-0.087
0.020
(-0.24)
(0.24)
-0.633*
0.047
(-1.65)
(0.53)
-1.816***
0.070
(-5.07)
(0.84)
-1.509*** -0.179
(-4.13)
(-1.04)
-2.112*** 0.183**
(-5.77)
(2.03)
-1.481*** 0.158**
(-4.12)
(1.99)
-0.002
-0.000
(-1.37)
(-0.69)
-0.223*
-0.048
(-1.94)
(-1.12)
-0.349*** -0.007
(-3.32)
(-0.20)
-0.648***
0.021
(-5.51)
(0.68)
-0.371*** -0.012
(-3.02)
(-0.23)
0.163
0.001
(1.08)
(0.04)
West
Northwest
2% ≤ Slope coeff. < 10%
10% ≤ Slope coeff. < 20%
20% ≤ Slope coeff. < 30%
Slope coeff. ≥ 30%
Av. temperature (current)
Av. rainfall (current)
Av. frost days (current)
Year dummies
Constant
Observations
R-squared
Adj.R-squared
Endogeneity test
C-Statistic (p-value)
0.034
(1.14)
0.069**
(2.45)
-0.071**
(-2.34)
0.000
(0.00)
-0.021
(-0.62)
-0.054
(-1.30)
-0.077***
(-2.61)
0.003
(0.98)
-0.027
(-0.75)
Yes
0.054
(0.92)
0.108*
(1.90)
-0.037
(-0.45)
0.006
(0.07)
-0.033
(-0.41)
-0.064
(-0.75)
-0.147**
(-2.39)
-0.002
(-0.43)
-0.164**
(-2.50)
Yes
1.008***
(8.81)
0.792***
(7.27)
0.268
(0.53)
0.542
(1.07)
0.388
(0.77)
0.264
(0.51)
-1.894***
(-13.03)
-0.062***
(-3.37)
-5.088***
(-28.16)
Yes
-0.019
(-0.35)
0.045
(0.86)
-0.054
(-0.59)
-0.034
(-0.37)
-0.067
(-0.73)
-0.088
(-0.92)
0.088**
(2.03)
0.013**
(2.18)
0.336***
(3.73)
Yes
-0.027
(-0.75)
0.034
(1.11)
-0.086
(-1.62)
0.036
(0.60)
-0.013
(-0.15)
0.055
(0.63)
-0.204
(-1.10)
-0.004
(-0.15)
-0.201
(-1.22)
Yes
0.448***
(2.79)
0.479***
(3.01)
-0.582
(-1.20)
-1.126**
(-2.32)
-1.371***
(-2.79)
-1.051*
(-1.76)
3.964***
(11.07)
0.258***
(6.58)
3.691***
(6.80)
Yes
-0.048
(-1.32)
0.027
(0.84)
0.002
(0.04)
0.153**
(2.04)
0.118
(1.32)
0.111
(1.22)
-0.288
(-1.61)
-0.002
(-0.11)
-0.161
(-0.92)
Yes
12.205*** 14.598*** 114.352*** 5.361*** 12.812*** 36.133*** 7.777**
(31.97)
(12.26)
(59.95)
(3.71)
(4.63)
(6.66)
(2.31)
10,853
0.37
0.36
5,602
0.18
0.17
5,603
0.65
0.65
5,602
0.15
0.15
21.618
0.000
2,944
0.16
0.15
Robust t-statistics in parentheses ; *** p<0.01, ** p<0.05, * p<0.1
2,946
0.59
0.59
2,944
0.13
0.12
35.656
0.000
Table 7. Grape Price Regressions
(OLS and IV methods – Instrument for EDC is distance to Reims)
EDC ratings
(1)
OLS
All
(2)
OLS
AC
0.005***
(34.86)
0.003***
(8.48)
Distance to Reims
Limestone
Alluvial Fans
Marl
Sedimentary Rocks
Silica Sand
Ferric Iron-bearing Sediments
Altitude (mean)
Northeast
East
Southeast
South
Southwest
West
Northwest
2% ≤ Slope coeff. < 10%
10% ≤ Slope coeff. < 20%
20% ≤ Slope coeff. < 30%
Slope coeff. > 30%
Av. annual temp.
0.048***
(14.26)
0.025***
(5.64)
-0.027***
(-9.24)
-0.034***
(-11.07)
-0.036***
(-11.57)
-0.016***
(-3.93)
-0.000***
(-10.39)
0.025***
(5.80)
0.015***
(4.38)
0.028***
(9.01)
0.019***
(5.94)
0.018***
(4.85)
0.019***
(5.29)
0.015***
(2.72)
0.024***
(3.42)
0.016**
(2.46)
0.006
(0.86)
0.020***
(3.05)
-0.026***
(-8.23)
0.061***
(18.18)
0.034***
(7.17)
-0.021***
(-7.45)
-0.029***
(-9.31)
-0.033***
(-9.92)
-0.023***
(-5.56)
-0.000***
(-13.10)
0.027***
(6.59)
0.017***
(5.05)
0.029***
(9.22)
0.023***
(7.17)
0.022***
(6.00)
0.024***
(7.00)
0.017***
(3.27)
0.029***
(4.37)
0.019***
(2.98)
0.009
(1.36)
0.020***
(3.16)
-0.041***
(-11.86)
(3)
First stage
AC
-0.023***
(-27.09)
-0.908***
(-4.31)
-0.446*
(-1.87)
-1.329***
(-8.24)
-0.786***
(-4.64)
-0.408**
(-2.20)
-0.858***
(-3.58)
0.005***
(5.75)
1.034***
(4.47)
1.228***
(6.12)
0.475***
(2.64)
1.256***
(6.99)
0.935***
(4.53)
-0.769***
(-3.60)
1.166***
(4.42)
-1.393***
(-4.82)
0.084
(0.34)
-0.412
(-1.58)
-0.416*
(-1.81)
-0.873***
(-4.39)
(4)
IV
AC
(5)
IV – AC
19912002
(6)
IV - AC
20032012
(7)
OLS
PC
0.020***
(22.86)
0.026***
(15.44)
0.015***
(18.40)
0.007***
(12.22)
0.082***
(15.48)
0.059***
(8.52)
0.013***
(3.20)
-0.002
(-0.43)
-0.009**
(-2.09)
0.005
(0.94)
0.000
(0.82)
0.004
(0.56)
-0.012**
(-2.10)
0.015***
(2.88)
-0.002
(-0.32)
-0.002
(-0.39)
0.029***
(5.03)
-0.011
(-1.33)
0.051***
(5.82)
0.015*
(1.92)
0.017**
(2.06)
0.035***
(4.93)
0.024***
(3.59)
0.125***
(12.81)
0.082***
(6.69)
0.016**
(2.12)
-0.000
(-0.02)
-0.019**
(-2.41)
0.014
(1.36)
-0.000**
(-2.47)
-0.023*
(-1.91)
-0.035***
(-3.01)
0.018*
(1.87)
-0.008
(-0.82)
-0.015
(-1.31)
0.032***
(2.94)
-0.033**
(-2.18)
0.070***
(4.97)
0.026**
(2.26)
0.042***
(3.28)
0.035***
(3.57)
0.055***
(4.25)
0.042***
(9.47)
0.025***
(4.48)
-0.002
(-0.46)
-0.013***
(-3.20)
-0.010**
(-2.22)
-0.006
(-1.06)
0.000*
(1.72)
0.019***
(3.59)
0.007
(1.41)
0.017***
(4.34)
0.005
(1.30)
0.009**
(2.21)
0.026***
(5.74)
0.011*
(1.91)
0.028***
(3.73)
0.006
(0.84)
-0.003
(-0.38)
0.028***
(4.39)
0.002
(0.30)
0.084***
(6.06)
0.098***
(7.33)
0.037**
(2.04)
-0.015
(-0.82)
-0.002
(-0.07)
0.053**
(2.11)
0.000***
(6.79)
0.072***
(5.42)
0.050***
(4.41)
0.020**
(2.00)
0.061***
(4.52)
-0.004
(-0.27)
-0.129***
(-7.50)
0.213***
(10.08)
-0.164***
(-5.60)
-0.183***
(-5.81)
-0.282***
(-9.41)
-0.086***
(-2.68)
0.020**
(2.50)
(8)
First stage
PC
-0.087***
(-4.64)
2.048
(0.90)
8.570***
(3.45)
7.683***
(2.71)
4.444*
(1.95)
-18.547***
(-9.83)
11.695***
(3.00)
-0.035***
(-6.04)
3.759**
(2.04)
0.260
(0.24)
-1.852**
(-2.32)
2.871**
(2.07)
-1.598
(-0.86)
9.069***
(4.70)
-5.473***
(-2.71)
-21.216***
(-8.65)
-27.967***
(-9.43)
-32.601***
(-9.38)
-28.391***
(-12.02)
7.507***
(7.52)
(9)
IV
PC
(10)
IV - PC
19912002
(11)
IV - PC
20032012
0.007***
(9.55)
0.003***
(3.13)
0.007***
(8.12)
0.095***
(6.93)
0.111***
(8.50)
-0.009
(-0.52)
-0.081***
(-5.09)
0.060**
(2.48)
-0.037
(-1.51)
0.001***
(14.65)
0.115***
(10.45)
0.043***
(4.33)
0.028***
(2.92)
0.115***
(9.41)
-0.013
(-1.13)
-0.239***
(-14.89)
0.335***
(14.93)
-0.090***
(-3.56)
-0.091***
(-3.46)
-0.284***
(-11.49)
0.023
(0.72)
0.000
(0.01)
0.184***
(10.08)
0.230***
(12.35)
0.154***
(6.67)
0.106***
(4.84)
-0.096***
(-3.89)
0.226***
(4.34)
0.000
(1.22)
0.183***
(6.26)
0.004
(0.21)
-0.017
(-1.58)
0.061***
(2.59)
-0.027
(-0.38)
-0.013
(-0.27)
-0.069
(-0.99)
-0.009
(-0.15)
-0.043
(-0.56)
-0.266***
(-9.41)
0.093
(1.31)
0.043*
(1.85)
0.114***
(8.87)
0.123***
(9.44)
0.033**
(2.01)
-0.027
(-1.55)
0.057**
(2.49)
0.031
(1.41)
0.001***
(12.75)
0.092***
(7.60)
0.031***
(3.16)
0.020*
(1.86)
0.070***
(5.97)
0.017
(1.39)
-0.247***
(-13.97)
0.332***
(13.61)
-0.102***
(-2.92)
-0.111***
(-3.14)
-0.258***
(-7.53)
0.014
(0.34)
0.003
(0.41)
Av. annual rainfall
Av. annual frost days
Year dummies
Constant
Observations
R-squared
Adj.R-squared
Endogeneity test
C-Statistic (p-value)
-0.001*
(-1.69)
-0.030***
(-8.13)
Yes
-0.001**
(-2.51)
-0.042***
(-11.03)
Yes
-0.196***
(-10.85)
-3.007***
(-12.62)
Yes
0.003***
(6.54)
0.071***
(7.89)
Yes
0.005***
(7.17)
0.121***
(6.74)
Yes
0.002***
(4.75)
0.034***
(4.87)
Yes
-0.001
(-0.68)
-0.054***
(-4.95)
Yes
0.695***
(4.26)
7.137***
(6.36)
Yes
-0.004***
(-7.06)
-0.127***
(-10.66)
Yes
0.006**
(2.12)
0.042*
(1.72)
Yes
-0.002***
(-4.24)
-0.131***
(-11.11)
Yes
0.941***
(20.97)
1.678***
(26.30)
102.441***
(42.08)
-0.493***
(-3.32)
-1.814***
(-6.21)
0.237*
(1.86)
0.580***
(4.81)
14.175
(1.03)
1.193***
(10.12)
0.266
(0.73)
1.130***
(10.03)
3,898
0.99
0.99
3,196
0.99
0.99
3,985
0.70
0.70
622
0.75
0.73
522
1.00
1.00
29.732
(0.000)
225
0.99
0.99
0.371
(0.5423)
297
0.99
0.99
2.068
(0.1504)
3,178
1,611
1,567
522
0.97
0.82
0.97
1.00
0.97
0.81
0.97
1.00
633.593
392.395
183.585
(0.000)
(0.000)
(0.000)
Robust t-statistics in parentheses ; *** p<0.01, ** p<0.05, * p<0.1
Table 8. Log of Vineyard Price Regressions (year-by-year)
(IV method – Instrument for EDC is distance to Reims)
EDC 1985-2003 Ratings (coefficients and t-statistics)
Autre cru
Premier cru
EDC 1985-2003 Ratings
Year
Coeff.
t-stat
Coeff.
t-stat
2002
-0.002
(-0.04)
0.019
(0.72)
2003
0.372**
(2.09)
0.060***
(3.60)
2004
0.012
(0.62)
0.084**
(2.09)
2005
-0.058
(-1.47)
0.053***
(3.88)
2006
0.001
(0.04)
0.057***
(3.53)
2007
0.069*
(1.69)
0.096***
(5.67)
2008
0.090***
(4.45)
0.071***
(4.67)
2009
0.087***
(4.06)
0.074***
(3.81)
2010
0.097***
(4.30)
0.176***
(4.40)
2011
0.059
(0.89)
0.151***
(3.24)
2012
0.081***
(4.39)
0.150
(1.54)
Coefficients and t-statistics obtained from year-by-year regressions.