Wine ratings such as those by Wine Spectator and Wine Advocate are intended to describe the flavor of a wine. Rated in the blind by experts, they serve as a guide to buyers on the quality of the wine when it reaches maturity.
A natural question is, do ratings reflect price? It seems reasonable to assume that a high rating may drive up the price of a wine. Statistically we can only determine if a relationship exists, not whether ratings drive price.
That said, our analysis provides an ideal mechanism for measuring this relationship. We cannot simply compare auction prices to ratings, because the price changes with the age of the wine and market conditions. Instead, we normalize prices for both of those effects first in order to rephrase the question as, do auction price divergences from average correlate to wine ratings. The auction price data was also adjusted for other known effects such as bottle size and auction house.
Using the lifecycle and environment functions from the APC analysis as known, fixed inputs, we created a secondary model to test other effects. The model was created to predict auction prices as a function of unit size for the bottles, auction house, and rating. The database had enough Wine Spectator and Wine Advocate ratings to provide an effect statistical analysis. Wine Spectator ratings had a similar monotonic be-havior and even a similar scaling by rating, with the significant exception of the highest rated wines. For Lafite and Bordeaux excl. Lafite, 99-rated wines were dramatically more valuable than 98-rated wines at auction, but 100-rated wines exhibited auction values at the level of 98-rated wines or below. Although the confidence intervals on the estimates were large, the drop for 100-rated wines is significant. For Burgundy, both 98 and 99-rated wines were lower than 98.
All of the measured dynamics in this analysis are relative to auction price, not opinions on the taste of the wine, but the tests do show a strong relationship between price and rating for Lafite, Bordeaux, and Burgundy. Wine Advocate ratings show a fairly smooth, monotonic relationship to differences in log-price. For Bordeaux, the difference between an 85-rated wine and a 100-rated wine is roughly a factor of 100 in price throughout the life of the vintage.
Of course, none of this proves whether wine ratings are predictive of the wine’s auction price, or whether historic patterns in auction prices are predictive of the wine ratings for future vintages, but the correlations are very interesting.
Joseph L Breeden, PhD has been building forecasting models for over 20 years for such areas as currency futures, sporting events, agricultural commodities, and loans. The methods he pioneered in his book, Reinventing Retail Lending Analytics (2014) are considered are considered best practice in the industry and performed well through the US mortgage crisis and many other international economic crises over the last 20 years. His love of wine and data analysis led to his participation in auctionforecast.com.