Pricing Models for German Wine: Hedonic Regression vs. Machine Learning

Abstract

This article examines whether there are different hedonic price models for different German wines by grape variety, and identifies influential factors that focus on weather variables and direct and indirect quality measures for wine prices. A log linear regression model is first applied only for Riesling, and then machine learning is used to find hedonic price models for Riesling, Silvaner, Pinot Blanc, and Pinot Noir. Machine learning exhibits slightly greater explanatory power, suggests adding additional variables, and allows for a more detailed interpretation of results. Gault&Millau points are shown to have a significant positive impact on German wine prices. The log linear approach suggests a huge effect of different quality categories on the wine prices for Riesling with the highest price premiums for Auslese and “Beerenauslese/Trockenbeerenauslese/Eiswein (Batbaice),” while the machine learning model shows, that additionally the alcohol level has a positive effect on wines in the quality categories “QbA,” “Kabinett,” and “Spätlese,” and a mostly negative one in the categories “Auslese” and “Batbaice.” Weather variables exert different affects per grape variety, but all grape varieties have problems coping with rising maximum temperatures in the winter and with rising minimum and maximum temperatures in the harvest season. (JEL Classifications: C45, L11, Q11)

On Fine Wine Pricing across Different Trading Venues

“…Third, we employ a Bayesian approach to overcome the problem of sparse and non-synchronous data, which is typical for the fine wine market. Although hedonic regressions (Oczkowski, 2016;Cardebat et al, 2017;Faye and Le Fur, 2019;Niklas and Rinke, 2020) or repeat-sales regressions (Burton and Jacobsen, 2001;Masset and Weisskopf, 2018) are commonly applied to investigate the price behavior of fine wines, we believe that the Bayesian approach corresponds better to our dataset.…”

Expensive and Cheap Wine Words Revisited

“…Table 1 reproduces all the 2 The fact that some studies are still using NB classifiers to study wine reflects the groundbreaking nature of Krumme's work, but her work can also be seen as groundbreaking in the more general sense that she used methods popular among data scientists to study wine. More recent work in the same vein has used support vector machines (Croijmans et al, 2020;Dong et al, 2020), neural networks (Niklas and Rinke, 2020), penalized logistic regression (Hovy, Melumad, and Inman, 2021), principal component analysis (Barth et al, 2021), and other popular data-science methods to predict wine prices, ratings, varietals, and other aspects. 3 Although the ratio of P(w i | c H ) to P(w i | c L ) with each of those probabilities defined as in Equation ( 1) is not necessarily the statistic Krumme used to identify expensive and cheap wine words, that statistic is consistent with the above-given quotes from Krumme (2011) about trying to identify words that are preferentially used or more likely to be used to describe high-priced wines.…”

Could Collected Chemical Parameters Be Utilized to Build Soft Sensors Capable of Predicting the Provenance, Vintages, and Price Points of New Zealand Pinot Noir Wines Simultaneously?

“…Soft sensors are built using machine learning methods, alongside easy-to-measure variables to predict highly related difficult-tomeasure variables [8]. For instance, a soft sensor employing the machine learning method artificial neural network (ANN) was used to predict Pinot noir wines' retail price based on viticultural data with an R2 value 0.80 [9]. The soft sensor utilising machine learning method least-square support vector machine (LS-SVM) successfully predicted 1-year-aged, 3-year-aged and 5-year-aged rice wines based on alcohol content, titratable acidity (TA) and pH, with R2 values of 0.91, 0.82 and 0.96, respectively [10].…”