What are the limitations of ‘fair value’?

Liv-ex’s ‘fair value’ methodology is intended to be simple and transparent. It uses information from the secondary market to help identify a fair price for a wine, given its score. If more appropriate, its age is used instead. In order to use ‘fair value’ effectively, users must also understand its limitations.

The model does not account for differences in ‘vintage premiums’ over and above that already captured by differences in score. As in the previous example, a 96-point wine in an exceptional vintage may warrant a premium to the price of an identical 96-point wine in an average vintage. This means that the data in the regression models is somewhat biased.

An attempt can be made to quantify this bias by looking at the average premium or discount to the trend line across vintages. This is shown in the chart below using wines from Liv-ex Fine Wine 50 and Liv-ex Left Bank 200 indices. For example, wines from the exceptional 2009 and 2010 vintages trade at an average premium to their trend lines of 7.3% and 6.5%, respectively.

In theory, a more complicated regression model could be built to account for such bias. In practice, we prefer to keep the model as simple as possible at this stage.

Physical vintage premiums (Fine Wine 50 and Left Bank 200)

Our model also does not account for differences in age. Older wines are likely to trade at a premium to younger wines of the same score and vintage quality. Indeed, the chart above actually captures age premium as well as vintage premium.

Some of the bias due to differences in age is mitigated by limiting the regression models to the past 11 vintages. Also, age is simply substituted for score in regression models where age is more significant in explaining variations in price.

The regression models do not have to be based on either age or score. Both age and score could be incorporated into the regression model as independent variables. However, as before, we prefer to keep the model as simple as possible.

The sample size in the regression models is limited as only the past 11 vintages are considered. This is to reduce any bias due to differences in age. However, small sample sizes may magnify any errors in the models.

It is assumed that either score or age is the predictor variable as opposed to the other way around. However, a wine’s score may also be influenced by its price, unless it is tasted blind. In theory, individual critic scores may also be influenced by existing scores from other critics.

The regression model only considers scores from The Wine Advocate. It uses the most recent score for each wine from either Robert Parker or Neal Martin. The recent retirement of Parker presents two key issues:

  1. Scores from The Wine Advocate represent a mix of scores from Parker and Martin. They are generally from Parker up to the 2013 vintage, and Martin from 2014 onwards. However, Parker and Martin have some different preferences which may lead to errors in the models.
  1. It is generally accepted that Parker had far more influence over fine wine prices than any other critic. As such, it remains to be seen whether any other critic will match this level of influence. Alternatively, wine prices may be determined by the consensus opinion of a number of critics.

It is likely that future wine prices may be influenced more by consensus opinion rather than any individual critic. The ‘fair value’ model can be modified to reflect consensus opinion as the transition from Parker becomes clearer.

View next:

En Primeur has often been mispriced