Judging Bordeaux Vintages: Intuition and Super Crunching
March 26, 2008, by Karl Storchmann (Journal of Wine Economics)
While the first reviews of the 2007 Bordeaux vintage will flood the wine press shortly, we don’t even know whether 2005 was a mediocre, a good, or an outstanding vintage. While the Wine Spectator gave the 2005 vintage a disappointing 84-89 point rating, the influential wine critic Robert Parker deems it “a surprisingly good year for the finest terroirs.”
Why is this important? Because bottled 2007 Grands Crus do not exist yet and they will not be released before 2010. But the châteaux set prices at which one can buy futures in the year following the harvest, i.e., 2008, and – with some luck – one can realize a decent profit in two years. Buyers of Bordeaux wine futures, of course, hope that the market price in 2010 will be substantially higher than what they paid for their future. That is why we all want to know: how is the 2007 vintage? Is the Mouton-Rothschild worth the $650 per bottle future price of last year? Since private future buyers cannot taste the wine they want to invest in they rely on second hand reports. The top Bordeaux châteaux presented the 2006 vintage in the first week of April of 2007 to the public. However, only 100 wine journalists and about 5000 importers were invited. The wines were tasted fresh from the barrel – tart and astringent with only a slim resemblance to their future taste. Clear cut – only experienced experts are able to predict how a certain wine will develop and what it will be worth on the market place. And experts sell their knowledge. Each April, right after the first tastings, wine magazines are full with ranking, ratings and reports. From an economist’s point of view, the similarities to the stock market are astounding. Most of us trust experts and their assumingly superior knowledge. We buy expert-managed investment funds that promise super-natural payoffs, buy power tools that analyze the stock market and calculate future profits or invest our money into the hottest stock of the week. Needless to say that expert knowledge is not free.
Back in 1973, Princeton economics Professor Burton Malkiel wrote his million-copy bestseller A Random Walk Down Wall Street and introduced an academic approach to the stock market to the public. He shows that a portfolio chosen at random outperforms any expert-managed portfolio rendering expert-knowledge phony and worthless. Or as Malkiel put it, ”the sad truth is that there are only three kinds of financial prognosticators: those who don’t know, those who don’t know they don’t know and those who know they don’t know but who get paid big bucks to pretend they know.” Instead of paying an expert, you are better off picking your stock by pinning the Wall Street Journal at your wall and throwing darts at it. 
Almost two decades later, Orley Ashenfelter, another Princeton economics professor and longtime editor of the prestigious American Economic Review as well as founder and editor of the new Journal of Wine Economics, applied an academic approach to Bordeaux wine futures showing that wine experts’ knowledge is common sense at best or flawed at worst. Ashenfelter noticed that wine prices exhibit a substantial fluctuation from year to year. For instance, while a bottle of the 1991 Lafite-Rothschild was about $100 at Zachy’s auction, the 2003 vintage went for $700. Clearly, older does not always mean better or more expensive. But what makes the 2003 vintage superior? Ashenfelter did not rely on elusive expert opinion but ran a simple statistical model and found out that three factors are crucial for great wine: a warm growing season, the absence of rain during the harvest and a wet winter prior to the growing season. Of course, that wine quality depends on weather has been known for centuries and comes to no surprise for winemakers. He then came up with a mathematical equation that explains market prices for Bordeaux Grands Crus. In other words, if we know the weather of a certain vintage we can statistically predict the quality and eventually the market price of a Bordeaux wine without having drunk a single sip of it. An updated version of his paper was recently published as Working Paper No. 4 of the American Association of Wine Economists.
Ashenfelter started publishing his predictions in a newsletter called Liquid Assets. But his ideas reached a much larger audience in 1990, when his quantitative approach hit the press. Countless news channels, magazines, newspapers and many TV channels covered the Ashenfelter theory. Here is a video clip from ABC's "Good Morning America." On March 4, 1990, The New York Times praised the accuracy of the wine equation on its front page (“Wine Equation Puts Some Noses Out of Joint”).
Where Parker had rated the 1986 Bordeaux "very good and sometimes exceptional”, Ashenfelter disagreed. Moreover, he predicted the 1989 Bordeaux, barely three months in the cask and yet to be tasted by critics, would be “the wine of the century”. And, he said, 1990 was going to be even better. As it turned out a few years later Ashenfelter’s predictions were astonishingly accurate. The 1989s turned out to be a truly excellent vintage and the 1990s were even better.
What Ashenfelter’s equation also conveys is the fact that buying wine futures is hardly profitable. Most futures are more expensive than the market price a few years later.
As you already may have assumed, wine critics are not amused by this approach. After all, it threatens to put them out of work. Thus, Parker calls the Ashenfelter approach “ludicrous and absurd”, others call it “neanderthal and barbaric.” That was 15 years ago. Meanwhile, even Parker draws on the barbaric method and posts temperature and precipitation data on his website. In Ian Ayres’ new book Super Crunchers: Why Thinking-by-Numbers is the New Way to be Smart Orley Ashenfelter is one of the main protagonists. “Orley Ashenfelter really loves wine” Ayres begins his book that original title actually was “Super Crunchers: The End of Intuition.”
Sometimes a few data and simple algebra attain more than elusive wine expert philosophy. And most importantly, it is free. Monthly and daily weather data for 
I guess I am really out of it. I didn't realize there was such a thing as a wine economist. But obviously you guys have something to contribute, and it appears to be sound thought and common sense. For me this thread is one of the most interesting and the discussion could go on forever. I would certainly stake my purchase on weather before relying on a "critic."
Two points come to mind. The first is that while average temp. during growing season provides some correlation it can be refined to see how the heat came during the season. This was the case for UC Davis's heat summation numbers. A heat spell during the middle of the growing season can damage quality and can hide a cold harvest. Also Day/night temps have a lot to do with grape composition as well and are not addressed by an average temp.
The second point is a bit tougher to deal with. That is regarding the data sets. I do not know the source of these data you summarize, but you generally find the weather stations at airports in Europe and there can be a lot of contamination from development. An American example is the contaminated data in the Napa Valley from two weather stations - one at the Napa Airport, the other at the State Hospital. There has been a huge amount of development around those stations in the last two decades with industrial parks and shopping centers whose asphalt and structures capture and release heat into the evenings. This distorts the data and makes it hard to correlate historical vintages with the present. While there has been a degree of warming in Bordeaux like everywhere else, from you summary I would not be surprised to see the same issue there that we have here. Generally the better data comes from weather stations in the vineyard itself.
Reply to this
Morton, it appears that there are enough economists interested in wine and enough wine people interested in economics that it can fill a journal. So, we started the "American Association of Wine Economics" and its journal the "Journal of Wine Economics" in 2006. Let's see where this will go.
Your points are very good. Certainly, monthly or even daily (or hourly) temperature can differentiate better and can gauge the impact of extreme cold/heat (that would not be possible when you use growing season averages). But for our limited purpose, namely assessing Bordeaux vintages, growing season averages will do it. There are statistical test procedures that confirm this.
As for regional/local weather. You are right, vineyard weather data are better. Taking weather only from one station (Bordeaux Merignac) assumes that local weather varies proportional with the weather at the station. As long as that is the case, there is no problem. Normally, you can assume when its a hot day at Merignac, it will also be a hot day 20 miles away. -- For the Bordeaux region, Michael Visser and Sebastien Lecoq (both INRA Paris) tested empirically if there is any explanatory gain from referring to weather data from several stations compared to taking only data from one station. They show that there is no difference. It is published in the Journal of Wine Economics, Vol. 1 (2006), No. 2, 114-124. But, I am sure, these result don't necessarily apply for all wine regions.
Reply to this
I agree with the proportional relationship of weather measurement. I have actually used it to fill in data of my own weather stations, knowing that a particular station lags an average degree behind another station. The problem comes as I said with comparing vintage historically with the present. If you look, for instance, at the historical temperature record around Napa Airport where there has been massive development you see significantly warmer night temperatures and an alarming upward pattern over the last two decades. Twenty miles away at the Angwin airport, where there has been no development and we do not see the same pattern. If I rely on the Napa Airport data to investigate the 1998 vintage in that region I would assume it to be much warmer than any in the 70's or 80's. Maybe the vintage of the century or maybe too hot for quality winemaking! The uncontaminated Angwin data presents a much different and much more accurate picture.
While I don't know whether the Bordeaux Merignac data is similarly contaminated, if you compare what actually happened on the ground in Bordeaux in 1982 (hot Summer and Harvest) with what happened in 2002, the graph doesn't represent the degree of difference that existed on the ground. The contamination of data caused by development is extensive in the U.S. (Check out http://www.surfacestations.org/odd_sites.htm) I would not be surprised if it existed elsewhere.
Reply to this
Very good point. Of course, if the difference in temperature between the reference weather station and the vineyard is not constant but grows or falls over time, the estimates become biased. That's clearly a problem. According to the Visser & Lecocq paper, for the Bordeaux region this doesn't seem to be the case (their data range from 1993-2002). But I can see that this may not be true elsewhere. Would be nice to check this empirically for different wine regions. Then you can model the difference between vineyard weather and weather station data as a function of various variables.
Reply to this
Thanks for the interesting article and commentary. I foresee a problem developing with this type of analysis. I'm assuming Professor Ashenfelter is using prices as an indicator of quality, rather than more subjective critical ratings. We have a trend towards higher average temperatures and also a trend towards greater spending on "high end" wines in most developed (and some developing) economies. Assuming no slippage in the relative prestige/image of Bordeaux Crus, it would seem these two trends together would guarantee a positive correlation between temperature and the price of top Bordeaux, without necessarily implying linkage. What do you think?
Reply to this
What Ashenfelter did was a cross sectional analysis. He compared wines from different vintages at one point in time. Clearly, wine prices will change over time due to factors other than the wine's age (income, wealth , etc.). I think it is reasonable to assume that these changes will affect all wines equally. That is, why should a Margaux benefit from income increases more than a Lafite. That is, in order to quantify the impact of weather on wine prices a cross sectional analysis is just fine. Actually, it is better than running a time series model. You are right, a times series model might confound income and temperature effects.
Reply to this