Monday, 15 August 2016

Quantifying value-for-money wines - part 4

This is the fourth of a four-part set of blog posts looking at how we might identify value-for-money wines. The topic is not as simple as we might like.
Quantifying value-for-money wines - part 1
   — issues with quantifying value for money
Quantifying value-for-money wines - part 2
   — empirically comparing wines within a specified wine group
Quantifying value-for-money wines - part 3
   — formulae for assessing individual wines against a baseline wine
Quantifying value-for-money wines - part 4
   — empirically comparing wines across wine groups

Empirically comparing wines across wine groups

In a previous post (Part 2) I considered empirical methods for trying to quantify value for money as related to wine, which is usually expressed as a quality to price ratio (QPR). The QPR method that I presented involved fitting the exponential (log-linear) model to the price and quality data for a set of comparable wines.

In this post I discuss a more sophisticated version of this approach. It is still based on the exponential model (and thus assumes a non-linear relationship between wine price and quality; see the post The relationship of wine quality to price), but it incorporates many more of the factors that might influence wine price. It estimates the QPR value for an "average" wine, which then allows the QPR value to be calculated for any other wines, of any type.

Hedonic Pricing

Hedonic Pricing is a method to estimate the "expected" price of the goods within a specified market, based on the principle that the goods are affected by certain factors that can raise or lower the "base" price. The factors can be intrinsic to the goods, or part of the external environment, or solely in the mind of the purchaser.

For wine, the obvious factors that could potentially affect the price include: grape type, vineyard location, winery reputation, vintage year, wine quality rating, cellaring potential of the wine, and current market demand. However, we can include as many factors as we like.

Based on this economic model, we can estimate the "expected" market price for wines by using the mathematical technique of multiple regression (called a hedonic regression). We collect a large dataset of wines, along with a record of all of their required characteristics, and we then perform the mathematical analysis. This yields a single equation that summarizes the model.

In practice, we use the model by entering into the pre-calculated equation the data for any specified wine, and this will produce the expected average price for that wine. If the wine costs less than this expected average then it is good value for money (a high QPR).


This is quite a popular approach to estimating expected prices, not just for wine but for many things, notably house prices. There is a list of many of the applications of Hedonic Pricing to wine in the post on The relationship of wine quality to price. In particular, the References listed below have explicitly applied hedonic regressions in order to identify value-for-money wines.

Unfortunately, I cannot draw a simple graph to illustrate Hedonic Pricing, the way I could in the previous Parts of this series of blog posts — the model forms a complex multi-dimensional pattern (one dimension for each factor included in the analysis). (The best way to illustrate the hedonic pricing model as applied to finding value-for-money wines would be to plot the wine value as predicted by the model against the actual wine price.) However, the principle is similar to that used in Part 2 of these blog posts. In particular, the model used is usually log-linear / exponential.

Importantly, the method is more comprehensive than that used in Part 2, because more influences on price are included, not just wine quality. Moreover, we do not need separate data analyses for different grape types, but instead they are all included in one giant analysis.

Also, the method is potentially easy to use in practice, because we have a single equation that we can use to estimate whether any given wine is good value for money. This is similar to the methods discussed in Part 3.

However, because everything is "built in", the model cannot be adjusted by the user for a new dataset. In particular, if vintage year is included as a factor in the model, then the equation needs to be recalculated after each new vintage.


Jon R. Miller, Robert W. Stone, Eric T. Stuen (2015) When is a wine a bargain? A comparison of popular and regression-based approaches. Journal of Wine Research 26:153-168.

Eddie Oczkowski (1994) A hedonic price function for Australian premium table wine. Australian Journal of Agricultural Economics 28:93-110.

Edward Oczkowski (2001) Hedonic wine price functions and measurement error. Economic Record 77:374-382.

Edward Oczkowski (2010) Hedonic wine price predictions and nonnormal errors. Agribusiness 26:519-535.

David A. Priilaid, Paul Van Rensburg (2006) Non-linearity in the hedonic pricing of South African red wines. International Journal of Wine Marketing 18:166-182.

David A. Priilaid, Paul Van Rensburg (2012) Nonlinear hedonic pricing: a confirmatory study of South African wines. International Journal of Wine Research 4:1-13.

Paul Van Rensburg, David A. Priilaid (2004) An econometric model for identifying value in South African red wine. International Journal of Wine Marketing 16:53-75.

Australian Wine Price Calculator

The first application of Hedonic Pricing to wine was by Eddie Oczkowski in 1994 (see the References). In 2002, he provided a web page for applying his pre-calculated Hedonic Pricing model for Australian wine, which is available at the Australian Wine Price Calculator. This appears to be the only available online calculator for wine that uses Hedonic Pricing.

This model includes five factors, as shown in the picture. It is currently based on a sample of 8,774 Australian premium wines (average price AU$34). The data cover the vintages 2004-2014 (as well as non-vintage wines), with cellaring potential until 2060. There are 82 grape varieties and blends included (with Shiraz as the most commonly occurring one in the dataset), along with 80 named grape-growing regions (McLaren Vale is the most common).

Where to from here?

The empirical QPR approach (see Part 2) should work best for identifying those wines with a good quality : price ratio, because it provides the best fit to the data for each wine category. However, this is rather impractical for assessing individual wines.

The wwpQPR approach (see Part 3) is much more practical for on-the-spot decisions when purchasing wine, because it uses a simple formula to evaluate any given wine. You do need to access the Wellesley Wine Press web page to do the calculations.

The Hedonic Pricing model (as discussed above) tries to combine the best aspects of these two approaches. It provides a good fit to the data across wine categories, as well as providing a simple formula to evaluate any given wine. However, there is currently only one web page available (for Australian wine only), which needs to be kept up to date after each vintage.

No comments:

Post a Comment