Preferences for Wine and Food Consumption in China

Abstract

The overall purpose of this dissertation is to investigate the Chinese consumers' preferences for wine products, how these preferences drive their purchasing habits and behavior, and the food consumption patterns in Chinese households. The dissertation consists of three independent but related articles. The first article introduces an experimental auction method to elicit the WTP of the participants for wine products from four different countries, China, the US, France and Australia. Data is also collected through the experimental procedure on purchasing habits and behavior, as well as the socio-demographics. A linear model is developed based on imputed data sets to analyze how the WTP of the participants is determined by controlling factors and socio-demographics, such as age, monthly household income, whether there is information exposure, and so on. Socio-demographic summary and estimation results are both discussed and compared among the four different wine products. The second article focuses on the purchasing habits and behavior. First of all, the summary for this part of data is discussed to draw a sketch of the Chinese wine market in a microeconomic level. Second, logistic regression model is applied to analyze how their purchasing habits and behavior are determined by the same controlling factors and socio-demographics as in the first study, and to predict the probability of potential wine consumption occurrence and frequency in a short-term. Third, marketing suggestions for foreign wine producers are given based on the results from both the first and the second study. The third article investigates the food consumption in Chinese household based on the household survey which was conducted in 2007. An QUAIDS model was built to estimate both the expenditure and price elasticities for nine groups of food. QUAIDS model here is considered as a SUR (Seemingly Unrelated Regression)-Tobit model, so Bayesian estimation is applied, which can incorporate the Gibbs sampler to solve the problem of a large amount of zero observations.