Inferring Individual Preferences and Latent Behavioral Factors with Incomplete Information

This dissertation extends prior research on inferring individual preferences from the following two aspects: one is to examine important latent behavioral factors affecting consumers\u27 consumption decisions; the other is to overcome the challenges arising from incomplete information. Regarding latent behavioral factors, this dissertation considers the following two aspects: (1) three types of intragroup dynamics behavior, and (2) variety-seeking behavior. Regarding incomplete information, this dissertation focuses on two types of incomplete information: individual\u27s behavior and identity, and order of consumption. Specifically, Chapter 2 presents a method to infer heterogeneous individual preferences and three components of intragroup dynamics using just aggregate and de-identified data. Chapter 3 emphasizes the effect of consumption outcomes on an individual\u27s propensity for variety-seeking when the order of consumption is unobserved. To overcome the challenges arising from incomplete information, this dissertation develops advanced individual-level Bayesian models and uses two-step iterative algorithms to estimate the proposed models in an MCMC framework. In-depth simulation studies show that the parameters are well recovered, suggesting that the proposed models are identified. Furthermore, this dissertation shows that ignoring latent behavioral factors may lead to biased estimation of individual preferences, which could result in many consequences. This dissertation applies the proposed methods to two empirical settings: an individual-level TV viewing and targeted TV advertising setting using Nielsen People Meter (NPM) data, and an online video game environment. In the TV viewing setting, it is shown that the proposed method could significantly improve the efficiency of TV ad targeting through counterfactual analysis. In the video-game environment, results show that although there is extensive heterogeneity, on average, positive consumption outcomes lead to inertial preferences, while negative consumption outcomes lead to variety-seeking. In sum, this dissertation shows the importance to incorporate important latent behavioral factors in inferring heterogeneous individual preferences especially when data are incomplete, and proposes innovative methods to overcome the challenges emerging from incomplete information