Content uploaded by David A Jaeger
Author content
All content in this area was uploaded by David A Jaeger on Jan 15, 2015
Content may be subject to copyright.
136
American Economic Review: Papers & Proceedings 2011, 101:3, 136–141
http://www.aeaweb.org/articles.php?doi
=
10.1257/aer.101.3.136
Like other markets in which deviation from
Jevons’s “law of one price” is the norm rather
than the exception, the retail wine market in
the United States is characterized by enormous
price dispersions. For instance, in our data retail
prices for 2005 Chateau Latour range from
$695 in a Petaluma, California, wine store to
$2,000 in a wine store in Champaign, Illinois.
Similarly, at the lower end of the price distribu-
tion, the observed retail price of 2007 Yellowtail
Merlot ranges from $4.99 in Buffalo, New York,
to $9.99 in Jersey City, New Jersey. Price disper-
sion in the wine market can be caused by various
factors, such as differences in production and
distribution cost, differences in price elasticities
of demand, or different market regulations and
structures.
Since the ratication of the 21st Amendment
repealing Prohibition, the US wine market has
been primarily regulated at the state level, more
or less impairing or effectively abolishing com-
petition between wine retail outlets. In addition
to the federal wine tax, wine is levied by state-
specic wine and sales taxes. Eighteen states
maintain a monopoly over the wholesale and
retail sales of wine; others restrict the sales of
wine to certain outlets and/or certain times, or
do not allow the payment for wine purchases
TOPICS IN WINE ECONOMICS
†
Wine Retail Price Dispersion in the United States:
Searching for Expensive Wines?
By David A. Jaeger and Karl Storchmann*
†Discussants: Joyce Jacobsen, Wesleyan University;
Victor Ginsburgh, University of Brussels; Bronwyn
Hall, University of California-Berkeley; Charles Mason,
University of Wyoming.
* Jaeger: Center for Macroeconomic Research, University
of Cologne, Albertus-Magnus-Platz, 50923 Cologne,
Germany, and CUNY Graduate Center (e-mail: jaegerd@
uni-koeln.de); Storchmann: Economics Department, New
York University, 19 W. 4th St., 6FL, New York, NY 10012
(e-mail: karl.storchmann@nyu.edu). We are grateful to our
discussant, Joyce Jacobsen, and the participants of the Wine
Economics Session at the AEA Annual Conference for help-
ful comments. We also thank Etan Schwartz and Vadim
Zhitomirsky for excellent research assistance.
with credit cards. Many states prohibit direct
wine shipments from out-of-state producers and
retailers, while others even prohibit in-state pro-
ducers and retailers to ship wine to consumers.
Price differences between states or counties are
thus not surprising. In this paper we examine
whether state- or county-specic effects fully
explain the observed price dispersion or if price
variations remain, even after controlling for
location differences. If so, is the degree of price
dispersion identical across all price brackets, or
does the dispersion for expensive wines reect
greater returns to search?
A large body of information-theoretic litera-
ture suggests that markets, even for standard-
ized products, may exhibit considerable price
dispersion. Following George J. Stigler’s (1961)
paper, several authors model how equilibrium
price dispersion can arise as a result of heteroge-
nous information (e.g., Steven Salop and Joseph
E. Stiglitz 1977, 1982; Jennifer F. Reinganum
1979; Hal R. Varian 1980; Kenneth Burdett and
Kenneth L. Judd 1983; John A. Carlson and
R. Preston McAfee 1983; and Dale O. Stahl II
1989). In general, price dispersion can persist
in equilibrium if obtaining information is costly
(through, for example, search costs) and some
fraction of consumers chooses to be uninformed.
A variety of empirical studies have explicitly
examined the association between consumer
search and price dispersion for homogenous
goods. John W. Pratt, David A. Wise, and
Richard Zeckhauser (1979) examine price dis-
persions for 39 consumer goods in the Boston
area and report coefcients of variation (CV)
for the product prices between 4 and 71 percent.
They also nd that the price dispersion substan-
tially increases with the average price of the
good, suggesting that the search cost for expen-
sive items is higher. This may be explained by
the fact that expensive products are purchased
less frequently, reducing the incentive of a
VOL. 101 NO. 3 137
WINE RETAIL PRICE DISPERSION IN THE UNITED STATES
buyer to search. Bev Dahlby and Douglas S.
West (1986) nd a CV of 18 percent for auto
insurance policies in Alberta. After ruling out
quality or cost differences, they conclude that
this price dispersion is almost exclusively due
to costly consumer search. Alan T. Sorenson
(2000) examines the retail prices of pharma-
cies in two geographically distinct markets and
nds a CV of 22 percent. While at most one-
third of the observed price dispersion is due to
pharmacy heterogeneity, most is due to costly
search. Sorenson also nds that frequently
purchased prescriptions exhibit lower price
variation.
Most relevant for our study is the hypothesis
that the Internet and the emergence of online
markets substantially lower search cost result-
ing in lower price dispersion (e.g., J. Yannis
Bakos 1997). Xing Pan, Brian T. Ratchford,
and Venkatesh Shankar (2002) analyze the price
dispersion of 581 goods in 8 product categories
in online markets. After controlling for sellers’
heterogeneity and especially service quality,
however, they nd online price dispersion to
be substantial and persistent. Erik K. Clemons,
Il-Horn Hann, and Lorin M. Hitt (2002) report
similar results for the market of airline tickets
sold by online travel agents. Kathy Baylis and
Jeffrey M. Perloff (2002) analyze Internet prices
of a specic type of digital camera and a atbed
scanner over a 14-week period, and also nd sig-
nicant price dispersion, which even increases
when controlling for service quality. In contrast
to Varian’s (1980) model of mixed strategies,
they nd a pure-strategies equilibrium, with
high-price rms and low-price rms remaining
xed in the overall ranking over time. They con-
clude that information costs (the time taken to
negotiate the website to discover stock and, to
some extent, price information) are an important
determinant of online price dispersion, and that
rms may discriminate among consumers based
on their knowledge, search costs, or patience.
Because a high degree of price dispersion
indicates large potential gains to search by
consumers, such dispersion may also suggest
that the market in question is inefcient with
regard to information. Empirical research has
shown that consumer search in most cases stops
before full information is obtained; sometimes
no search takes place at all (Ratchford 2009).
Given that search is costly, however, the opti-
mum search point is reached when marginal
search cost equals its marginal benet. Ratchford
and Narasimhan Srinivasan (1993), Edward J.
Fox and Stephen J. Hoch (2005), and Dinesh
K. Gauri, K. Sudhir, and Debabrata Talukdar
(2008) provide empirical evidence that is con-
sistent with this normative rule.
Given that the search cost is essentially xed
per wine and independent of its price (e.g.,
searching a website), it is possible that search
is more protable for expensive wines, result-
ing in smaller price dispersion with increasing
average prices. On the other hand, less expen-
sive wines face a stiffer competition from close
substitutes than expensive wines do. In contrast
to a $200 wine, when a consumer shops for a
$5 wine the brand and vintage are likely to be
of less importance. Monopoly pricing power
may therefore increase with price, potentially
leading to a price dispersion that increases with
a wine’s average price. Alternatively, learning
through experience may play a role and lead to
the same dispersion-price relationship. Low-
price wines sell at much higher quantities than
high-end wines. Information about quality and
prices of lower-tier wines may thus more easily
penetrate the market (for “learning-by-buying”
and “word-of-mouth,” see Ratchford 2009).
In this analysis, we draw on a large database
of wine retail prices to examine the relationship
between price level and price dispersion. We rst
examine the role of local characteristics such as
the number of retail wine establishments, per
capita income, and local demographics on wine
prices. We control for the regulatory environ-
ment by using state xed effects. After also
controlling for wine-vintage xed effects, we
then examine whether the residual variation in
prices is related to the wine’s average price. In
general, we nd a signicant and positive rela-
tionship between residual variation in prices and
(adjusted) price levels.
I. Data and Descriptive Statistics
We use wine retail prices from 2006 to 2008
provided by wine-searcher.com, an Internet
wine price search site on which wine retail out-
lets worldwide can post prices of their wines. For
sellers in the United States, wine-searcher.com
currently lists approximately 2.5 million prices
posted by about 6,300 wine stores. Since
many wines are available only in a few stores,
we restrict our analysis to 186 wine brands of
MAY 2011138 AEA PAPERS AND PROCEEDINGS
various vintages. For all but one of these wines
we observe well over 200 prices, and for many
we observe more than 1,000 prices.
1
Overall,
our sample contains approximately 106,000
prices on red and white wines. In Table 1 we
report some basic descriptive statistics on price
levels and price dispersion. Most of the wines
in our sample are produced in the United States
and two-thirds of them are red. We observe sub-
stantial differences in price dispersion, mea-
sured by the coefcient of variation. Compared
to the results of other empirical analyses, the
overall price dispersion of 23.4 percent is rather
1
The data contain observations for sizes in addition to
the standard 750ml bottle. We have dropped all observations
for nonstandard sizes. In addition, for each wine we have
dropped the 5 percent lowest and 5 percent highest observed
prices, to be sure that we were not capturing (mislabeled)
case prices or other measurement issues. We have also
dropped any observations in which the description indicated
that the bottle was damaged or irregular in any way. For
wines with both vintage and nonvintage prices reported, we
dropped any nonvintage prices when these constituted less
than 25 percent of the total number of observations for that
wine. We also eliminated from the data rosé, sparkling, and
fortied wines.
high. It is higher for red than for white wines and
higher for French wines compared to domestic
wines and other imports (mainly from Australia
and Italy). Also, expensive wines exhibit higher
price dispersion than do wines in lower price
brackets, suggesting the dominance of the sub-
stitution effect or learning from buying over
search cost hypotheses.
II. Determinants of Wine Prices
To examine how local market characteristics
affect wine prices, we estimate the equation
(1) log(pivcsy) = β
0 + β
1 E
cy + β
2
I
c
+ β
3 W
c + β
4 A
c + β
5 O
ivy
+ β
6 N V
i + θ
y + δ
s
+ λ
iv + ε
ivcsy ,
where i indicates wine, v indicates vintage, c
indicates county, s indicates state, and y indi-
cates year of price posting. The variable E is the
number of retail wine establishments in county
Table 1—Descriptive Statistic s on Retail Wine Prices
Avg. Avg. coefcient Avg. N per Number of
Sample mean price of variation wine wines
Full sample
Red $80.25 0.2495 984.45 136
White $27.84 0.1925 746.32 50
United States
Red $45.61 0.2002 914.41 66
White $15.17 0.1814 746.84 31
France
Red $148.64 0.3492 1,173.49 47
White $130.53 0.3732 689.33 6
Other
Red $39.87 0.1874 799.13 23
White $10.66 0.1358 771.38 13
Average price < $15
Red $8.50 0.1668 661.92 36
White $8.71 0.1668 683.97 29
Average price < $50
Red $28.34 0.2576 842.47 32
White $22.83 0.1665 883.69 16
Average price ≥ $50
Red $142.65 0.2895 1,222.02 68
White $154.83 0.4099 668.4 5
Source: Authors’ calculations using data from wine-searcher.com. Observations are from 2006–
2008, measuring prices of nonvintage and vintage wines from the 1998–2007 vintages.
VOL. 101 NO. 3 139
WINE RETAIL PRICE DISPERSION IN THE UNITED STATES
c in year y divided by the county population in
2000, taken from the county business patterns
data of the United States Census Bureau. I is
per capita income in the county in 2000, W is
the white share of the population in the county
in 2000, A is the share of the county population
in 2000 that is 25 or over (the population most
likely to drink wine), O is how old the wine is in
year y (nonvintage wines are coded to zero), NV
is an indicator for nonvintage wines, θ is a year
xed effect, δ is a state xed effect (capturing
differences in state regulations), λ is a wine ×
vintage xed effect, and ε is the idiosyncratic
term. In some specications, we use only simple
wine xed effects without letting the coefcient
vary across vintages.
The results of estimating variants of equation
(1) are presented in Table 2. The rst three col-
umns contain results for red wine and the last
three contain results for white wine. In columns
1 and 4 we constrain the λs as well as the coef-
cient wine age and nonvintage to be equal to
zero. It is clear from both columns that prices
vary with local market conditions, even with our
sample drawn from sellers who list their prices
on the Internet. Local market conditions explain
only between 7 (red) and 13 (white) percent of
the variation in prices, however. This is not sur-
prising—in this regression we are treating all
wines the same, regardless of where or by whom
they were produced.
In columns 2 and 5 we add xed effects for
each wine to the analysis, but constrain these to be
equal across vintages. The model now accounts
for 95 percent of the variation in log prices—
clearly the majority of variation in wine prices
comes from differences in origin and quality. The
coefcients on income, the share of whites in the
population, and age change somewhat, suggest-
ing that different wines are sold in different loca-
tions. In columns 3 and 6, we allow for a full set
of wine × vintage interactions. The results are
qualitatively similar to those in column 2.
III. The Relationship between Price and Variance
Our fundamental research question is whether
there is a relationship between residual price
variation and price level. In Table 3, we report
the slope coefcients from a regression of the
Table 2—Determinants of Wine Prices
Red White
Variable (1) (2) (3) (4) (5) (6)
Number of wine retailers per 0.0031 0.0002 0.0002 0.0023 −0.0002 −0.0001
2000 county population (0.0002) (0.0001) (<0.0001) (0.0004) (0.0001) (0.0001)
Log per capita income (2000)0.1568 −0.0300 −0.0083 0.1423 −0.0208 −0.0107
in county (0.0011) (0.0068) (0.0007) (0.0386) (0.0099) (0.0079)
White share of county population −0.0243 −0.0017 −0.0189 −0.2334 0.0214 −0.0066
(2000) (0.0073) (0.0085) (0.0060) (0.0515) (0.0129) (0.0101)
Share of county population that is 0.0247 0.1714 0.1987 1.4587 0.2767 0.2412
25 or older (2000) (0.0249) (0.0393) (0.0282) (0.2307) (0.0608) (0.0479)
Wine age (nonvintage = 0)0.0075 0.1411 0.0231 0.0533
(0.0008) (0.0034) (0.0023) (0.0042)
Nonvintage −0.0573 −0.0152
(0.0042) (0.0067)
Year xed effects X X X X X X
State xed effects X X X X X X
Wine xed effects X X
Wine × vintage xed effects X X
R20.07 0.95 0.98 0.13 0.94 0.97
Observations 82,698 23,919
Note: Dependent variable is log price. Estimated via OLS. Heteroskedasticity-consistent standard errors are in parentheses.
Source: Authors’ calculations using data from wine-searcher.com. Observations are from 2006–2008, measuring prices of non-
vintage and vintage wines from the 1998–2007 vintages.
MAY 2011140 AEA PAPERS AND PROCEEDINGS
average squared residual on the xed effect
for each wine × vintage combination, both
taken from columns 3 and 6 of Table 2 for red
and white wines, respectively. That is, we are
estimating
(2)
_
e iv
2
= ϕ + φ ˆ λ
iv + ξ
iv ,
where e are the residuals from the estimation of
equation (1). Here we nd that, overall, there is a
positive relationship between residual variation
in prices and their level. To put the magnitude of
the coefcient in context, the average value of
the dependent variable for red wines (that is, the
average mean squared residual) is 0.0287. Thus,
the estimated coefcient on the full sample of
1,117 wine × vintage combinations is about
one-tenth of this average. For white wines, the
average mean squared residual is 0.0209 and the
estimated coefcient for the full sample is about
three-tenths of this (0.0064). For both red and
white wines, we nd a stronger statistical rela-
tionship between dispersion and average price
for vintage wines than for nonvintage wines.
IV. Conclusion
In this paper we show that there is a fair
amount of price dispersion for red and white
wines in the United States, with an average per-
wine coefcient of variation of 23 percent. Some
of this is due to differential market conditions.
But our evidence suggests that dispersion also
depends (weakly) on price levels, after control-
ling for consumer, market, and state heterogene-
ity. These results are consistent with the theory
of “learning-by-buying” in which goods that are
purchased more often are predicted to have less
price heterogeneity. The results are less consis-
tent with a search costs story. To be consistent
with our results, search costs would have to
be higher for expensive wines relative to less-
expensive wines. This seems less plausible to
us because the search mechanisms are likely to
be the same for both inexpensive and expensive
wines. It may be, however, that buyers of more
expensive wines have a higher opportunity cost
of time and are less willing to spend time search-
ing for the lowest price.
REFERENCES
Bakos, J. Yannis. 1997. “Reducing Buyer Search
Costs: Implications for Electronic Market-
places.” Management Science, 43(12): 1676–
92.
Baylis, Kathy, and Jeffrey M. Perloff. 2002. “Price
Dispersion on the Internet: Good Firms and
Bad Firms.” Review of Industrial Organiza-
tion, 21(3): 305–24.
Burdett, Kenneth, and Kenneth L. Judd. 1983.
“Equilibrium Price Dispersion.” Economet-
rica, 51(4): 955–69.
Carlson, John A., and R. Preston McAfee. 1983.
“Discrete Equilibrium Price Dispersion.” Jour-
nal of Political Economy, 91(3): 480–93.
Clemons, Eric K., Il-Horn Hann, and Lorin M.
Hitt. 2002. “Price Dispersion and Differentia-
tion in Online Travel: An Empirical Investiga-
tion.” Management Science, 48(4): 534–49.
Dahlby, Bev, and Douglas S. West. 1986. “Price
Dispersion in an Automobile Insurance Mar-
ket.” Journal of Political Economy, 94(2):
418–38.
Fox, Edward J., and Stephen J. Hoch. 2005.
“Cherry-Picking.” Journal of Marketing,
69(1): 46–62.
Gauri, Dinesh K., K. Sudhir, and Debabrata Taluk-
dar. 2008. “The Temporal and Spatial Dimen-
sions of Price Search: Insights from Matching
Household Survey and Purchase Data.” Jour-
nal of Marketing Research, 45(2): 226–40.
Pan, Xing, Brian T. Ratchford, and Venkatesh
Shankar. 2002. “Can Price Dispersion in
Online Markets Be Explained by Differences
Table 3—Relationship between Residual Price
Dispersion and Average Price
Red White
Sample Coefcient NCoefcient N
All 0.0024 1,117 0.0064 398
(0.0010) (0.0020)
Vintage 0.0025 1,056 0.0062 352
(0.0010) (0.0020)
Nonvintage 0.0291 61 0.0171 46
(0.0151) (0.0133)
Notes: Each coefcient comes from a separate regression.
We regress the average squared residuals for each wine ×
vintage combination on the xed effects for those combi-
nations. These are taken from columns 2 and 6 of Table 2.
Heteroskedasticity-consistent standard errors in parentheses.
Source: Authors’ calculations using data from wine-
searcher.com. Observations are from 2006–2008, measuring
prices of nonvintage and vintage wines from the 1998–2007
vintages.
VOL. 101 NO. 3 141
WINE RETAIL PRICE DISPERSION IN THE UNITED STATES
in E-Tailer Service Quality?” Journal of the
Academy of Marketing Science, 30(4): 433–45.
Pratt, John W., David A. Wise, and Richard Zeck-
hauser. 1979. “Price Differences in Almost
Competitive Markets.” Quarterly Journal of
Economics, 93(2): 189–211.
Ratchford, Brian T. 2009. “Consumer Search and
Pricing.” In Handbook of Pricing Research
in Marketing, ed. Vithala R. Rao, 91–107.
Northampton, MA: Edward Elgar Publishing.
Ratchford, Brian T., and Narasimhan Srinivasan.
1993. “An Empirical Investigation of Returns
to Search.” Marketing Science, 12(1): 73–87.
Reinganum, Jennifer F. 1979. “A Simple Model
of Equilibrium Price Dispersion.” Journal of
Poli ti ca l Ec on om y, 87(4): 851–58.
Salop, Steven, and Joseph E. Stiglitz. 1977. “Bar-
gains and Ripoffs: A Model of Monopolistically
Competitive Price Dispersion.” Review of Eco-
nomic Studies, 44(3): 493–510.
Salop, Steven, and Jospeh E. Stiglitz. 1982. “The
Theory of Sales: A Simple Model of Equilib-
rium Price Dispersion with Identical Agents.”
American Economic Review, 72(5): 1121–30.
Sorensen, Alan T. 2000. “Equilibrium Price Dis-
persion in Retail Markets for Prescription
Drugs.” Journal of Political Economy, 108(4):
833–50.
Stahl, Dale O., II. 1989. “Oligopolistic Pricing
with Sequential Consumer Search.” American
Economic Review, 79(4): 700–712.
Stigler, George J. 1961. “The Economics of Infor-
mation.” Journal of Political Economy, 69(3):
213–25.
Varian, Hal R. 1980. “A Model of Sales.” Ameri-
can Economic Review, 70(4): 651–59.
