We analysed a sample of 93 investment grade Bordeaux Châteaux in order to better understand – and if possible, to predict - the dynamics undelaying the price variation in the short term after their release in the market.
We considered vintages between 2007 and 2016*.
The following figures show the distribution of price variation between release price and market price at 3 years after release date.
The release price is the first price that appears on the market in a given vintage, usually between April and June during the En Primeur campaign), when the wine is still in barrel (equivalent to futures in the financial market). The market price in the present analysis is recorded after 3 years from release, when the wine has already become physical (i.e. is in bottle).
For the 2015 vintage, as an example, we measured the difference between the average market price of our sample in 2019 (average of 3 months, May to July) and the average release price in the springtime 2016.
* The data acquired for the 2016 vintage are the last available ones, meaning that the average of May-July is not available at this time. Final data for this vintage may differ slightly.
When we look at the shape of the distribution of price variation of our Châteaux sample for various vintages, we notice that they tend to resemble to a normal distribution, skewed towards the right. This longer right tail indicates that a small number of Châteaux increase in price more than proportionally when compared to those that decrease in price. In this sample analysis, price variation ranges roughly between -60% and +200%.
There are basically 3 different identifiable trends in our analysis: vintages (2010, 2011) in which a higher number of Châteaux show a marked decrease in price compared to their release price ; vintages (2007, 2008, 2009, 2013, 2014, 2015) in which a higher number of Châteaux show an increase in price and vintages (2012, 2016) in which the gains and losses are slightly more equally distributed.
When we look at how much each vintage has increased or decreased on average, we can observe wide variations.
The horizontal axis shows the vintage being analysed, and the vertical axis shows the percentage of variation from its release price after 3 years.
The averaged data shown here are consistent with their distribution (ch.1), as mentioned above, but what type of information can we derive from these tendencies?
To try to answer to that question, we analysed different independent variables that might influence the increase or decrease in price in the short term.
We noticed that the price of wines of recent past vintages tends to be more volatile during the En Primeur campaign of a newly released vintage.
We hypothesize that the relative quality of recent vintages might have an influence on the price of wines that have recently been offered on the market, either as futures or as physical wines.
Although the variation in price cannot be attributed solely to one or two variables, we can clearly see that there is a correlation between it and the difference in quality – measured in 100 points scores (Global Wine Score in our case) - of recent vintages.
For example, in the first column on the left (2014) we observe that the average difference of score among recent vintages (multiple recent vintages are taken into account by our algorithm) is +4.63, and this is correlated positively with the increase in price (37.3%) after 3 years from release, i.e. in the springtime of 2018. In the case of 2010 vintage, the average difference is negative (-4.34), which correlates with a decrease in price (13.96%), and so on.
Another variable that we wanted consider is the so-called Fair Value (as defined by Liv-ex), which is the best fit between a wine price and its scores by critics.
By observing the price history of individual wines, we can assume that in the long term, generally speaking, the market price of a wine will tend to converge to its Fair Value.
As we can see from the graph above, for the vintage 2008 the average distance of our sample from their Fair Value is -41.7%, which correlated well with the average increase in price of +34.8%. On the other hand, the 2010 vintage was on average released at higher price than Fair Value, and that coincided with their decrease in price down the road.
Again, as already mentioned, this trend is generally observed but there are exceptions, as in the 2012 and 2015 vintages in which other factors might have played a significant role in the market (i.e. China anticorruption measures, financial crisis, US tariffs, etc.).
After having analysed the historical data and performances and certain variables that can influence the price variation of our Bordeaux sample, we went to perform a series of predictions by creating algorithms that take into account the most significant variables discussed earlier.
The graph above shows the result of our predictions for the average price variation after 3 years from the release price, as described earlier. The blue line indicates the predicted average variations and the orange bars the observed average variations.
As shown in the analysis, our model can indeed predict in most cases the sign of the variation (increase or decrease), whereas the extent of the variation is not always accurate. This of course reflects the fact the in real life the price variations are influenced by many variables, internal and external to the specific market segment analysed.
Nevertheless, this method can provide a good estimate of the expected trends, particularly in conjunction with Saturnalia early score prediction model.
By knowing the relative value of a new vintage is therefore possible to estimate the knock-on effects that this may have on existing recent vintages with a 6-month advantage compared to the standard En Primeur campaign.
Our predictive algorithms do not just apply to aggregated data, as shown here, but also to individual Châteaux wines, as shown in a separate analysis.
To follow Saturnalia news and updates, and access to exclusive reports, please subscribe to our newsletter or send us an email by filling the form down below.