UC Irvine

This dissertation comprises three chapters. Chapter 1 is based on my unpublished job market paper "Flavorants and Addiction: An Empirical Analysis of Tobacco Product Bans and Taxation." In this chapter, I investigate the potential impact of a proposed menthol cigarette ban on cigarette consumption and product substitution. The use of menthol cigarettes and other tobacco product flavorants has been a contentious issue, but empirical research on the effectiveness of flavorant bans is limited. To fill this gap, I use aggregate-level retail data and micro-level household data to estimate a random coefficient nested logit demand model that incorporates the effects of addiction and consumer heterogeneity. I pay particular attention to the Black American community and low-income households, who have the highest rates of menthol cigarette consumption. My findings show that in the absence of menthol cigarettes, overall cigarette usage reduces by 13%, and the Black smoking rate falls by 35%. In addition, e-cigarettes and cessation products experience a 4.9% and 1.7% increase in demand, respectively. These results are then contrasted with changes in consumption and consumer surplus stemming from a national cigarette tax and a ban on all flavored nicotine products.

Chapter 2 is based on unpublished work in collaboration with Jiawei Chen and Saad Andalib Syed Shah. In this chapter, we introduce a structural choice model that takes into account households' geographic and product substitution behavior to evaluate the impact of localized taxation policies. Using detailed retail and household data from Philadelphia's soda tax, we estimate the choice model to examine the relationship between households' demographic characteristics, proximity to the city border, and their tax avoidance behavior - such as switching from taxed to untaxed products or from Philadelphia to non-Philadelphia stores. Our findings show that travel time is a crucial factor for modeling households' heterogeneous responses, with an additional minute of travel time equivalent to adding 47₵ to the product price. By factoring in travel costs and the switch to less preferred products, we find that on average, Philadelphia households experience a loss in consumer surplus more than twice the amount of tax paid, with low-income households bearing the greatest burden.

Chapter 3 examines the importance of accounting for realistic substitution patterns in discrete choice models, particularly in the presence of large alternative sets. While multinomial probit models are the preferred method for modeling correlated unobservables across elements of the choice set, their estimation and identification can be challenging. As a result, researchers often resort to alternative methods or restrictive assumptions that disregard substitution patterns. However, ignoring these patterns can lead to biased estimation results. To address this issue, I propose a structured covariance matrix that models substitution patterns as a function of product similarity, allowing for feasible estimation in the presence of large alternative sets. In addition, I incorporate individual parameter heterogeneity and a two-stage consumer decision process, enabling dynamic and individual-level behavior. To estimate the model, I develop a Bayesian MCMC process that utilizes a Tailored Random Block Metropolis Hastings algorithm. Finally, I conduct a simulation experiment to demonstrate the superior performance of my proposed model compared to alternative estimation methods. My results suggest that restrictive substitution patterns can hinder proper estimation of parameter values, emphasizing the importance of considering realistic substitution patterns in choice models.