Collart Dinarte, Alba Jeanette (2013-12). Econometric Methods to Analyze Consumer Behavior Using Hypothetical and Non-hypothetical Approaches. Doctoral Dissertation. Thesis uri icon

abstract

  • This dissertation examines consumer behavior using hypothetical and non-hypothetical approaches. Across the study, emphasis is made on brand awareness and Willingness-to-Pay (WTP) measures of consumer preferences and the measurement of unobserved individual heterogeneity in the econometric analysis of consumer valuations. The methodologies used to elicit valuations and gather consumer preferences are hypothetical and non-hypothetical. The statistical tools used to analyze these data include econometric models for categorical and limited dependent variables, linear and non-linear random parameters models, and Latent Class Analysis (LCA). The first essay evaluates the effectiveness of a point-of-purchase advertising program conducted for two local horticultural brands. Results based on electronic surveys gathered before and after the program was launched suggest that the campaign size was not sufficient to significantly increase brand awareness and overall demand, yet it increased WTP by 5.5% for those consumers aware of one of the brands. A major factor found to influence preferences was purchase frequency, which suggests that other advertising methods aimed to increase buying frequency might affect demand more effectively. The second essay involved the econometric analysis of data collected using experimental auctions, which are often multidimensional. Panel data models commonly used consider bid-censoring and random effects that capture heterogeneity in the intercepts, but overlook heterogeneity in the coefficients. This essay compares different models, and provides evidence that a Random Parameters Tobit model extends the measurement of heterogeneity, accounts for bid-censoring, and provides the most efficient and consistent estimates. When the model is applied to data collected in a non-hypothetical Vickrey auction to elicit WTP for government (Food Safety Modernization Act, FSMA) and industry-issued (Global GAP) food safety standards in specialty melons, findings indicate that valuations are censored and heterogeneous. Finally, heterogeneity in valuations is assumed to occur discretely. Using a LCA approach, an examination was done to segment consumers based on their unobserved motivation to participate in experimental auctions. Moreover, Random Effects Tobit models are estimated to investigate differences in WTP among latent classes. The three latent classes found were characterized as: "Fee-Chasers", "Certification Conscious", and "Taste Conscious". Results reveal that the classes differed significantly in terms of their WTP estimates.
  • This dissertation examines consumer behavior using hypothetical and non-hypothetical approaches. Across the study, emphasis is made on brand awareness and Willingness-to-Pay (WTP) measures of consumer preferences and the measurement of unobserved individual heterogeneity in the econometric analysis of consumer valuations. The methodologies used to elicit valuations and gather consumer preferences are hypothetical and non-hypothetical. The statistical tools used to analyze these data include econometric models for categorical and limited dependent variables, linear and non-linear random parameters models, and Latent Class Analysis (LCA).

    The first essay evaluates the effectiveness of a point-of-purchase advertising program conducted for two local horticultural brands. Results based on electronic surveys gathered before and after the program was launched suggest that the campaign size was not sufficient to significantly increase brand awareness and overall demand, yet it increased WTP by 5.5% for those consumers aware of one of the brands. A major factor found to influence preferences was purchase frequency, which suggests that other advertising methods aimed to increase buying frequency might affect demand more effectively.

    The second essay involved the econometric analysis of data collected using experimental auctions, which are often multidimensional. Panel data models commonly used consider bid-censoring and random effects that capture heterogeneity in the intercepts, but overlook heterogeneity in the coefficients. This essay compares different models, and provides evidence that a Random Parameters Tobit model extends the measurement of heterogeneity, accounts for bid-censoring, and provides the most efficient and consistent estimates. When the model is applied to data collected in a non-hypothetical Vickrey auction to elicit WTP for government (Food Safety Modernization Act, FSMA) and industry-issued (Global GAP) food safety standards in specialty melons, findings indicate that valuations are censored and heterogeneous.

    Finally, heterogeneity in valuations is assumed to occur discretely. Using a LCA approach, an examination was done to segment consumers based on their unobserved motivation to participate in experimental auctions. Moreover, Random Effects Tobit models are estimated to investigate differences in WTP among latent classes. The three latent classes found were characterized as: "Fee-Chasers", "Certification Conscious", and "Taste Conscious". Results reveal that the classes differed significantly in terms of their WTP estimates.

publication date

  • December 2013