A Taste for Equity: Wine Tasting Scores and Stock Prices
Robert Eyler
Sonoma State University
Rohnert Park, CA 94928
Copyright (c) 2001 Robert C. Eyler. All rights reserved.
Economists theorize that a firm's stock price is fundamentally a function of its expected cash flows (a forecast as to the dividend payout or appreciation in the stock price), the company-specific risks (default on debt, bad harvests, etc.), and market risks (inflation pressures, industry earnings problems, FED adjustments), regardless of other factors. This study focuses on a different type of stock "fundamental" that is intrinsically linked to others. What happens to the expected cash flows of the firm if tastes toward their products change? What happens to the company-specific risk if the quality of the firm's products is perceived as different than before new information is available? Unfortunately, it is very difficult to quantify quality measurements in such a way as to satisfy economists1.
The expected dividend payout and growth rate of the stock price are the first two fundamentals. The market and company risks are the other two fundamentals. Students of finance know the implications of changes in the fundamentals: if either expected dividend payouts or growth rates increase, the stock price should rise; if the risks rise, the stock price should fall, holding all else constant. For firms whose product depends greatly on qualitative measures of quality, profits to pay dividends or augment productivity are potentially tied directly to consumer perceptions. Also, company specific risk may be tied to these perceptions. Thus, we can use an indirect link to these fundamentals to explain movements of equity values after a qualitative shock in the wine industry.
A winery's financial performance is obviously a large part of equity price valuation and thus tasting scores should be a part of this valuation. However, can a link between the fundamentals and tasting scores be possible due to the qualitative nature and lack of continuous and consistent data for tasting scores? Financial analysts look at balance sheets, income statements, and raw financial numbers rather than publications like the Wine Spectator for their expectations on these stocks. Should they be looking at the Wine Spectator, however? This paper shows that winery tasting scores, the perceived quality of wines in the eyes, noses, and mouths of experts, have little influence on the firm's equity value while these scores have large consumer power. Observing this phenomenon is as easy as going into any winery tasting room, wine shop, or grocery store. The shelf space is dominated by advertisements concerning expert tasting scores in an attempt to influence wine sales 2 .
How much influence is there on wine sales due to new tasting accolades, such as wine expert scoring? It is difficult to make an industry-wide assessment, as every firm has different wines, different timing on score announcements, and different exogenous factors on firm revenues. The use of equity prices as an indicator of firm performance is more effective for our purposes, since they are also measurable in discrete time. The literature on equity prices is vast and the data are readily available. There has been an explosion of subtopics concerning the pricing of securities. However, there are few studies of how qualitative shocks may lead to changes in equity prices.
1The lack of studies on how quality
perception affects stock prices is not surprising. The literature
on hedonic preferences has not moved into applying expected
utility theory to equity valuations as of yet. This study is the
first to try and combine these worlds using an industry.
2 See the focus on tasting rooms in Wine Business Monthly,
April 2001.
Literature Review
There are two basic strains of literature on qualitative shocks and equity pricing topics. The first is a theoretical strain, coming from financial economics. The second is from the hedonic preference literature. The purpose of this paper is to shed light on a larger issue using the case of wineries to forward an evocative hypothesis: do product quality signals deliver information used by buyers and sellers of stocks? Since product quality is normally the result of a qualitative study or from information in summary form, we must analyze this phenomenon outside of an efficient market hypotheses (EMH) of theoretical finance. The hedonic preference literature gives some possibilities on topics concerning qualitative signals in equity markets. Using a mix of equity pricing formulae, intuition, hedonic preference theory, and econometrics, this study tests the hypothesis above.
The EMH literature entered its modern form via the rational expectations literature 3 (Muth, 1961; Lucas, 1976) in conjunction with Modigliani and Miller (1958) setting the foundation for university-level finance courses to date. In any of the above seminal works, the message is the same: in any market with many buyers and sellers, such as a securities market, perfect competition is seen as the reigning market structure, even in the EMH’s weakest form. Thus, equity prices carry some information about the stock indirectly in the EMH's weakest form; in its strongest form, a stock price finds a pure rational expectations equilibrium.
While it is believable that the forces governing securities markets are highly competitive, because product differentiation takes place for many goods as a means of non-price competition, it seems foolish to not look at certain industries as exceptions to the EMH. Using the wine industry, and the publicly-traded firms within, we have a great example of qualitative product characteristics influencing sales for and price changes by firms. Intuitively, this should lead to changes in stock fundamentals and prices. The temporal chain indicates that tasting scores should cause stock price changes. However, we must make a theoretical turn toward the hedonic preference literature, where issues of quality effects on price reside in economics.
Since the model below follows a hedonic framework, some specific works shape the model and analysis. Work of Grichiles (1961) and Gordon (1973) was refined by Rosen (1974) in the case of differentiated products 4 . Epple (1987) used these ideas to find estimaed supply and demand functions for differentiated products in terms of hedonic prices. These prices are simply demand functions that allow for quality characteristics to influence the "slope" of the demand curve. Theoretically, the Epple study posited price as a function of quantity demanded and a vector of quality attributes. Epple's analysis uses a functional form that delivers a stable equilibrium. Using winery stocks allows us to have an intuitive basis for linking product quality to firm performance. Cropper, et al. (1988), look at multiple functional forms for hedonic functions, expanding the Epple model. Few studies have shown how shocks to product quality, in a highly differentiated industry, change stock prices.
Do stock prices acts as hedonic prices in the case of the wine industry? If so, it may be possible to predict movements in these prices by simply learning wine tasting scores from experts. If not, the millions of dollars spent on advertising and tasting reviews, and the proliferation of these scores in such publications as the Wine Spectator, seem to be worth little to the publicly-traded winery if the firm's mission is stockholder wealth maximization.
3 See Sheffrin (1997) for a large
overview.
4 The article by Sherwin Rosen is, in the case of differentiated
markets and product quality, one of the key works and most often cited.
Theoretical Model
The model below looks at the stock price based on both quantitative and qualitative attributes of the firms. In certain cases, the amount of one firm's business involving qualitative attributes differs within an industry. For our purposes, the model reflects only the existence of the qualitative attributes. Suppose a publicly-traded firm, firm i, has common stock with price pi. Basic financial theory tells us, a la Gordon (1973), that the stock price is a function of the expected dividend payout (Di), the expected capital gains on the stock (KGi), and the required rate of return for the investor (Ri). We add the qualitative attributes of the firm's product as another possible indicator (Zi), indicating that a change in the perceived quality of the firm's product may have an effect on the quantitative variables listed above. This is the crux of this analysis: do qualitative signals change stock prices?
The price of the stock has the following generalized form:+ + - +
Pi = Pi(Di, KGi, Ri;
| Zi) = f(Zi) (Equation 1)
where is a set
of parameters governing the relationship between the exogenous variables
and the endogenous stock price. This is the classic model of equity prices
with the twist of hedonic preferences. The perceived quality of the underlying
good increases, the stock price should rise. If the quantitative factors
are in fact functionsof the qualitative attributes, the stock price moves
as a result of moving all the factors at once in an appropriate manner.
In this model, the following hypotheses can be tested using an econometric
analysis:
:
If the tasting score shock is positive, the augmentation in revenue
with no true change in the unit cost of the wine should increase profits
and thus increase expected dividends. This should increase the stock
price.
:
If the tasting score shock is positive, the expected capital gains increase
with the expectations that profits, because of higher revenues and no
changes in costs, rise causing new demand for the stock. This should
increase the stock price.
: If the tasting score shock is positive, the company-specific risk
in the stock should fall as consumer perception of quality is augmented.
The final market sees the product as relatively superior. This should
increase the stock price. Thus, the calculus of these changes allows
us to say that based on the theoretical model above.
>
0: As perceived product quality rises, the stock price rises.
One problem is time lags, however. Since dividends are normally paid on a quarterly basis, if paid at all, there will be large lags in qualitative signals and their fruition in dividends. Also, investors may not react immediately to marginal quality augmentations or decreases. In an attempt to simplify the analysis, as there is no intuitive benefit to introducing large lag structures for events that are not continuous in time, we eliminate dividend payouts from the model. Our assumptions center on the expected capital gains and the required rate of return. Thus, the log-log model used below is as follows.
ln(Pit) = ß 0 + ß 1i ln(Pit-1) + … + ß 2i ln(Pit-10) + ß 3i ln(Zit) + e it (Equation 2)
Data and Empirical Results
The need to match data in a temporal sense is key. What we are looking for is a correlation between a change in the tasting score profile of the wine corporation and changes in its stock price. The hypothesis is that there is little connection between the two, even though marketing efforts, sales of full vintages, tasting room space allocation, and the large subscription base of publications such as the Wine Spectator, hinge on wine quality perception. If positive revenue gains take place due to this augmented quality perception, then we expect that investors may look at these changes as indicative of larger expected cash flows from the stock. However, even if there exists a statistical connection between tasting scores (the Wine Spectator acts as our source of perceived quality indicators as the best choice among few) and the stock price of the firm, it would be potentially anomalous due to the lack of fundamentals in that indicator choice.
Graphs of the stock prices for the four firms investigated are below 5 , where CHLN = Chalone Wine Group, MOND = Robert Mondavi Corp., WVVI = Willamette Valley Vineyards, and RVWD = Ravenswood Winery. Each wine corporation needs some basic explanation and background.
5 Initially, we must look at the stationarity of the stock price data series to be sure that there will be no volatility issues undermining the credibility of the stock price series from December 16, 1998 to March 16, 2001. An augmented Dickey-Fuller test shows that each data series is stationary in first differences. This inclusion of the lagged dependent variable eliminates any spurious correlation (Hoover, 1988).
- Chalone Wine Group (CHLN): CHLN is a Napa-based
company specializing in premium white and red varietals. CHLN
operates Chalone Winery, Acacia Vineyards, Jade Mountain,
Carmenet Vineyards, Echelon Vineyards (California); Sagelands
and Canoe Ridge Vineyards (Washington); also a joint venture in
France in Chateau Duhart-Milon. Earnings per share for the
quarter ended 12/31/2000 was $0.08/share. Revenues for year
2000 approximately 70 million dollars.
- Robert Mondavi Corporation (MOND): MOND sold 8 million cases (9 liter equivalents) of wine in year 2000. Their revenues were over 427 million dollars. MOND has many brands, including Luce, Caliterra, and Vichon as foreign holdings; Opus One is a joint venture with Rothschild of France; Mondavi and Woodbridge are California brands.
- Ravenswood Winery (RVWD): RVWD is a contrast to both CHLN and MOND. RVWD is based in Sonoma, California, and sell red wine varietals as its focus. It also produces some white wine varietals as a diversification device; red wine counted for 94% of RVWD sales in 2000. In year 2000, RVWD sold 412,860 cases of wine for gross revenues of 33.2 million dollars. RVWD was recently purchased by Constellation Brands.
- Willamette Valley Vineyards (WVVI): WVVI is located in Turner, Oregon. Their annual gross revenues for year 2000 were approximately 5.5 million dollars. They reported net losses in each quarter of 2000, thus their earnings per share is negative. WVVI concentrates on white varietals such as chardonnay, pinot gris, and riesling.
Figure 1
Stock Prices of CHLN, MOND, RVWD, WVVI ($/share)
12/16/1998 to 3/16/2001
Source: DJIA Interactive
Visually, CHLN and WVVI are trending down over the last two years while MOND and RVWD are trending up in price. By choosing these specific four firms, industry-wide problems causing changes in these equity values are not an issue as the correlations between these market-close prices are not strong. Table 1 below gives the correlations between these stock prices.
Table 1
Correlation Matrix for CHLN, MOND, RVWD, and WVVI Stock Prices
12/16/1998 to 3/16/2001
| CHLN | MOND | RVWD | WVVI | |
|---|---|---|---|---|
| CHLN | 1.000 | |||
| MOND | -0.064 | 1.000 | ||
| RVWD | -0.416 | 0.562 | 1.000 | |
| WVVI | 0.420 | -0.248 | -0.356 | 1.000 |
With a lack of statistical correlation between these prices, we assume that the volatility in each stock is more company-specific than market-specific. Looking at the tasting score data themselves, we need to consider the nature and timing of this data to analyze it correctly.
The Wine Spectator is a highly-respected industry source with a large subscription. When wineries receive positive accolades from the Wine Spectator, as defined by this study a tasting score equal or greater to 90 points out of 100, advertising and public relations efforts at small, privately-held wineries become focused on that number. The tasting facility of the winery, if one exists, displays this information prominently for any customer to see. In certain cases, the entire supply of a vintage sells within days of the accolade's announcement. There exists much anecdotal evidence to believe that tasting scores influence firm performance, especially at a small firm. The tasting data also have another peculiar aspect: the data series are not continuous. Making the data continuous forces assumptions upon us about the way historical data influences our decision mechanism today. We assume that investors, if tracking the tasting scores of the wine corporation, adjust their viewpoint on new news and then averages what happened previously. This is a slight deviation from a rational expectations approach, but allows inference from a discrete data series with no major theoretical issues. The graphs for each of the sample wineries above is below, where ln(Zi) lies between 4.25 and 4.605.
Figure 2
ln(Zi) for CHLN, MOND, RVWD, WVVI
12/16/98 to 3/16/01
Source: Wine Spectator
Looking at Figure 2, and comparing it to stock price movements shown in Figure 1, we see that CHLN and RVWD are trending down, while MOND and WVVI have little trend. The more important aspect of this data is not in the trend but in the volatility. CHLN has little volatility, which means that the Wine Spectator views CHLN wines at consistently below 90. The remaining firms have tasting profiles over the two years that show some volatility and also some stability. This stability, either high or low, may lead to expectations based on long-standing quality perception. Thus, the wine market may be affected greatly by a shock (random effect if seen as anomalous) or affected very little because of historically good or poor tasting scores.
The statistical analysis below attempts to find connections between these tasting scores and the price of the stocks for the four wine corporations chosen. The model is an autoregressive model with 10-period lags, looking for connections where the tasting score changes. With 118 time periods and over 100 degrees of freedom, there are plenty of observations to show a connection, especially with the current volatility of the stock market.
Table 2
Regression Results using Equation 2 above
|
|
R2 |
DW Stat |
|---|---|---|
| CHLN |
0.78 |
1.85 |
| MOND |
0.88 |
2.06 |
| RVWD |
0.96 |
1.86 |
| WVVI |
0.65 |
2.04 |
Table 2
Regression Results using Equation 2 above (continued)
| CHLN | MOND | RVWD | WVVI | |||||
|---|---|---|---|---|---|---|---|---|
| Coeff | t-stat | Coeff | t-stat | Coeff | t-stat | Coeff | t-stat | |
| Constant | -1.86 | -2.45 | 1.06 | 0.86 | 1.38 | 1.32 | -2.99 | -2.53 |
| Winery-1 | 0.47 | 2.15 | 0.81 | 7.91 | 0.84 | 8.01 | 0.37 | 3.69 |
| Winery-2 | -0.31 | -1.08 | 0.15 | 1.43 | 0.16 | 1.45 | 0.29 | 2.99 |
| SCORE | 0.25 | 0.83 | 0.13 | 0.61 | 0.24 | 0.81 | -0.34 | -0.52 |
| SCORE-1 | -0.20 | -0.63 | 0.04 | 0.15 | -0.43 | -1.10 | -0.30 | -0.34 |
| SCORE-2 | 0.24 | 0.75 | 0.04 | 0.15 | -0.46 | -1.28 | 0.39 | 0.45 |
| SCORE-3 | -0.47 | -1.49 | -0.18 | -0.68 | 0.23 | 0.64 | -0.10 | -0.11 |
| SCORE-4 | 0.40 | 1.24 | -0.37 | -1.35 | 0.46 | 1.26 | 1.22 | 1.41 |
| SCORE-5 | -0.32 | -0.98 | 0.60 | 2.20 | -0.09 | -0.25 | -0.71 | -0.82 |
| SCORE-6 | 0.40 | 1.24 | -0.44 | -1.58 | -0.36 | -0.98 | -0.63 | -0.72 |
| SCORE-7 | -0.32 | -1.00 | 0.26 | 0.94 | 0.08 | 0.21 | 0.07 | 0.08 |
| SCORE-8 | 0.43 | 1.75 | 0.16 | 0.59 | -0.26 | -0.71 | -0.08 | -0.09 |
| SCORE-9 | 0.56 | 5.34 | -0.24 | -0.91 | 0.54 | 1.38 | 0.94 | 1.07 |
| SCORE-10 | 0.14 | 1.38 | -0.19 | -0.92 | -0.26 | -0.85 | 0.23 | 0.34 |
Some standout statistics are the R2 and the coefficients on the tasting score variable (SCORE) in each case. The regression model uses the natural logarithm of the closing stock price for each week from December 16, 1998, to March 14, 2001. By lagging this variable on the right-hand side by two periods, any serial correlation present in the closing stock prices should be eliminated 6 . The Durbin-Watson statistic for each equation provides the information that the serial correlation is statistically eliminated. The R2 statistic is not consistently high by the standards of most time series studies. In the case of RVWD, the explanation of the variation of the stock price by the lagged stock prices and the tasting score is extremely strong, but in the other three cases the explanation is weak. The use of the lagged dependent variables should eliminate these problems, but did not. This does tell a story, however, that is very credible. If the firm's wine market has little to do with the tasting score received by experts, then the tasting score should have very little explanatory power as to stock price variation. For RVWD and MOND, perceived quality is shown to be more explanatory, as a result leaves with the idea that these two corporations may spend more resources on sources such as the Wine Spectator to market their products. The main feature of the R2 statistic is its inability to indicate the extent to which the independent variables' variation explains the dependent variable’s variation.
By using natural logarithms, we derive elasticities from the regression output directly. Each of the coefficients above is essentially tasting score multipliers: when the tasting scores change by a certain percentage, the stock price is affected in a percentage sense by the tasting score elasticity multiplied by the percentage change in the tasting score. Thus, having tasting scores be continuously over 90 points may not have much of an effect on the stock price as investors see no change in the perceived quality of the wine. Volatility upward in tasting score, however, may have profound effects as it suddenly may change the perception of the entire market at one time. Since the multipliers are either small or not statistically different from zero, we can conclude that tasting scores have little effect on the stock price of these publicly-held firms, regardless of size and market. From the results one can say that CHLN, MOND, RVWD, and WVVI individually.
6The DW stat, when statistically equal to 2 shows the presence of no serial correlation. The range of the DW stat is 0 (positive serial correlation) to 4 (negative serial correlation). As a time series property, the elimination of serial correlation is to assure the strength of the regression coefficient and eliminate bias.
Conclusions
Hedonic indicators are strange tools for financial advisors to use when rating firms. The ephemeral nature of perceived quality lends itself to scrutiny as an indicator of any economic phenomena. Tasting scores for wineries are a great area to investigate the perceived quality of goods and also potential reactions to them. This study looked at the effects of tasting scores on wineries that are publicly-held for effects on the stock price. Assuming that the timing of the tasting score shock was easily found in the data and that the stock was somewhat immune to macroeconomic phenomena, we looked at the relationship between the tasting score lagged over ten weeks from the initial shock and its effects on the stock price. We found mixed results in the four firms investigated, and found little in the way of true patterns to be discerned. Three conclusions flow from the parsimony in the results.
First, it is highly unlikely that the equity markets view wine tasting scores as indicative of changes in expected cash flows from these firms, or in changes in the firm-specific risk. In most cases, potentially because the equity markets are fully aware that the tasting scores can change very quickly in different directions and also may have little impact on the typical consumer's consumption pattern, and thus on the expected cash flows from the stock or the company-specific risk. Second, firms with the size and scope of Chalone Wine Group and Mondavi have little effect to be expected by a change in the perceived quality of one product. This is the point of taking advantage of economies of scale and scope. Diversifying wine brands, having international brands in particular, should be viewed as an insulation device from volatile financial and goods markets. It also should be viewed, in a market such as wine, as having the inability to take full advantage of positive, minor shocks to perceived quality or to any one product in particular. The other firms, much smaller in both scope and size, show little effect of tasting scores changing. This leads us to the third idea that quality is difficult to generalize. One taste may not easily transfer to another person, group of consumers, or the market as a whole. It is true that in individual cases, tasting score shocks can sell entire lots of wine but the effect of one vintage selling out is difficult to translate across vintages, brands, or varietals.
It goes without saying that the subscription base of a publication like the Wine Spectator makes it a powerful publication regardless of actual effect on the stock price. Should these publications have that much power? Should the tasting score be an indicator of anything from a profitability or risk standpoint. If we think deeply about the subjective basis of these scores, they should not translate into business analysis directly. However, if sales do rise as a result of strong tasting scores, it seems foolish not to watch them. For our purposes, it is best to believe that tasting scores have little purpose other than stimulate good conversation at your next dinner party. However, the landscape of the financial may change if a true link is found between a subjective measure such as tasting score and equity prices. Further research should be done to either solidify or dispel the myth and mystique held by these wine publications.
References:
Cochrane, John H. (2001) Asset Pricing. Princeton University Press, Princeton, NJ.
Cropper, Maureen and Deck, Leland B. and McConnell, Kenneth. (1988) "On the Choice of Functional Form for Hedonic Price Functions", Review of Economics and Statistics, vol. 70, no. 4, pp. 668-75.
Dow Jones Interactive. (2001) "Historic Stock Prices", www.djinteractive.com.
DRI/McGraw Hill Database (2001) McGraw-Hill Publishers, Boston, MA.
Epple, Dennis. (1987) "Hedonic Prices and Implicit Markets: Estimating Demand and Supply Functions for Differentiated Products", Journal of Political Economy, v. 95, n. 1, p. 59-80.
Golan, A. and Shalit, H. (1993) "Wine Quality Differentials in Hedonic Grape Pricing",Journal of Agricultural Economics, vol. 44, no. 2, pp. 311-321.
Greene, William. (2000) Econometric Analysis, 4th Edition, Prentice-Hall, New York.
Grichiles, Zvi. (1961) "Hedonic Price Indexes with Automotive Examples" The Price Statistics of the Federal Government, New York.
Laube, James. (1995) California Wine, Wine Spectator Press, New York.
Lucas, Robert E. "Econometric Policy Evaluation: A Critique", The Phillips Curve and Labor Markets, ed. Karl Brunner and Allan Metzler. Carnigie-Rochester Series in Public Policy, Amsterdam, North Holland.
Modigliani, Franco and Miller, Merton. (1958) "The Cost of Capital, Corporation Finance and the Theory of Investment", vol. 48, no. 3, pp. 261-297
Muth, John F. (1961) "Rational Expectations and the Theory of Price Movements", Econmetrica, vol. 29, pp. 315-335.
Oster, Sharon M. (1999) Modern Competitive Analysis, 3rd Edition, Oxford University Press, New York, pp. 128-140.
Porter, Michael. (1985) Competitive Advantage, Free Press, New York.
Rosen, Sherwin. (1974) "Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition", Journal of Political Economy, v82, n.1, pp. 34-55.
Sheffrin, Steven M. (1997) Rational Expectations, 2nd Edition, Cambridge University Press, Boston, MA.
Wine Business Monthly, Jan-May 1998 - 2001, Sonoma, CA.
Wine Spectator. (2001) "Wine Spectator Tasting Scores", www.winespectator.com.
