In 2014, I wrote an article estimating the losses to Merlot farmers due to the movie Sideways at $400M. That article was written 10 years after Sideways was released, but that did not mean that it captured all of those losses, as the Sideways effect had not necessarily come to an end. I have heard many discussions lately, as to whether or not the Sideways-influenced consumer is done afflicting our Merlot farmers.
The short answer is, “Yes, pretty much.” A more correct, but slightly longer answer is, “Who really knows? But, yeah, it seems like it has.” What follows is the much longer answer, reflecting my examination of the issue.
Maybe the Sideways Effect was a Myth
I know, there’s a lot in the graphic, but don’t worry. We’ll go through it piece-by-piece. The numbers in the columns to the left-hand side are simple. They show vintage, nominal price, CPI adjustment factor and the price in constant, 2015 dollars. The 'SUMMARY OUTPUT' SECTION is the linear regression output. Ignore it, if it doesn't interest you More interesting is the graph:
The solid, blue line shows actual, average Merlot prices for the whole state, in 2015 dollars. As you can see, there was a distinct, long-run, downward trajectory between 1990 and 2007. This may have been caused by lower-end producers jumping on the Merlot bandwagon, both overplanting and growing lower-quality, lower-priced wine. Or not. It’s not too important to the subject at hand. Sideways came out in October of 2004, so it would have had a price effect starting in 2005.
The straight, dotted line is the linear trend line. As we can see, since Sideways, Merlot’s inflation-adjusted prices look to have dropped for a couple more years and then, despite the movie and the Great Recession, looks to have stabilized and might even be seeing a bit of a recovery. In fact, prices drifted further away from the trend line for only a couple years and have now converged with where we would expect prices to be in 2015, based on this overly simple analysis. The formula in the upper right hand corner corresponds to this trend line, where y equals price and x equals the vintage. The r-squared is the square root of the correlation. The closer to 1, the better.
The curved, dotted line indicates the fifth order, polynomial trend line. It has a much higher R-squared, but the formula is somewhat arbitrary and would not be predictive. It is “overfit.” The point is, though, as a descriptive formula it indicates that Merlot prices are right where they should be regardless of Sideways. Lower order polynomial regressions indicated that prices might actually be higher than one would expect.
So, are we done here, is the Sideways effect a myth, like the chupacabra or New York City sewer ‘gators? I’m sure you can guess the answer is “no.” Otherwise, this post would have had a different title and, also OF COURSE SIDEWAYS HURT MERLOT PRICES!
I decided to test the Sideways effect by adding in a dummy variable to represent the number of vintages since Sideways was released (e.g. 2015 = 1, 2005 = 11.) This addition increased the r-squared from .44 to .74. The Sideways variable’s coefficient was -57.76 and the vintage variable’s was -17.56. Both vintage, as in the linear regression, and the Sideways variable, were statistically significant. What this means is that, starting in 2005, there was a shift in the market. The coefficients indicate that the natural rate of inflation-adjusted, downward price pressure was $17.56 per year, accelerating to $75.23 (57.67+17.56) after the release of Sideways.
Note three things. First, this does not actually mean the price would drop that much, just that it is being suppressed that much, incrementally and annually, on a long-term basis. Second, the extra downward price pressure is coincident with Sideways, but this does not necessarily mean it is caused by the movie (correlation is not causality.) Finally, the model is still not really strong.
The next step was to build a more granular and predictive model. I wanted to keep the model relatively simple, so I started with previous year Merlot acreage and vintage only. This is a highly predictive variable (more so than previous year yield). This model had very high levels of statistical significance for its variables and boosted the r-squared to a whopping .95.
The next model worked the Sideways dummy variable back in. Statistical significance was still high and the r-squared inched closer to perfect (.97). Interestingly, the effect of the trend and Sideways shifted a good bit, with the overall trend being annual, marginal, downward price pressure of $31.54. For the Sideways effect it was $18.85.
OK, so now we know that there has been some downward pressure on prices since Sideways’ release. We also know it is significant, but weaker than the longer-running, general, negative trend for Merlot prices. Additionally, from our experiences in the real world, I think we can assume that it is more likely than not that this has actually been caused by the movie. But is this effect still a concern?
With a model this strong, we can try out a couple of iterations to gather more data. My dummy variable rated the strength of the Sideways in 2005 as an 11 and 2015 as a 1 (the effect strength decreases by 1 in each year). This was an arbitrary choice. I used trial and error to find a more powerful numbering system. The only one I found that was arguably stronger was to designate 2005 a 2 and 2015 a 12. Due to time constraints, I only looked at linear models, not logarithmic or exponential ones. Long story short, this yielded a strong indication that the Sideways effect will no longer be able to be distinguished after 2015 or 2016. I have run this test the past couple years and the evidence lined up with 2016 as the last year of an effect.
In reality, we might have been able to find other models - like one using a natural logarithm - as solutions and we can’t fully distinguish between all of the influences that are part of the unseen chaos of our world. This, however, is the strongest quantitative evidence I could hope for, short of surveys or studies of individual buyer behavior, that at this moment we are seeing the effect of Paul Giamatti wear off, presaging a return to a ‘normal’ Merlot market.
If that market is like the rest of the grape market, I would think that we are on the brink of real (inflation-adjusted) price hikes for Merlot for two reasons. First - and this is applicable only to premium grapes – coastal grapes are seeing prices rise and I see little reason now to assume that Merlot won’t follow. Second, we’ve been seeing acreage attrition for years now for Merlot, down 19% from the all-time high in 2004. I would guess there is a reasonable probability that demand has been or will soon start increasing, based on typical consumer behavior and the research detailed above.
My prediction, I’ll admit, is tentative and will vary by growing region. Furthermore, we already saw serious price-taking in 2011 and 2012 that may have captured future growth potential (about 14% per year in nominal terms.) Still, I expect that statewide average Merlot prices will see modest, inflation-adjusted rises in 2016 and 2017, as long as Cabernet Sauvignon prices don’t drop, which could weigh down Merlot prices. In any case, take these thoughts with a grain of salt. I modeled only for descriptive purposes, to answer my Sideways-related question. I have not built any predictive models. But, if you’d like me to, just pick up the phone!