Essays in Behavioral Economics and Taxation

Abstract

This dissertation studies behavioral economics and the design of public policy. Chapters 1 and 2 investigate tax salience—the notion that the prominence of a tax shapes the extent to which taxpayers account for it when making purchasing decisions. In many contexts, policymakers can control a tax’s salience and Chapter 1 investigates how such control should be exercised to best promote consumer welfare. Chapter 2 turns to distributional concerns relating to tax salience. For some taxes, salience effects may vary by income; in work co-authored with Tatiana Homonoff, I investigate how policymakers can take advantage of this fact to make the distribution of such taxes less regressive. Drawing on state and time variation in cigarette tax rates, we empirically investigate whether salience effects vary by income and find evidence consistent with this theory.

Chapter 3 turns away from tax salience to broader issues concerning behavioral economics and public policy. In numerous settings, behavior varies according to seemingly arbitrary features of the decision-making environment, such as which option is the default, the order in which options are presented, or which option characteristics are salient. Optimal policy design requires accounting for the preferences of decision-makers whose choices are sensitive to such factors, but traditional revealed preference analysis breaks down in that setting. In work co-authored with Daniel Reck, I consider binary choice problems in which preference-irrelevant “frames” affect the behavior of decision-makers and develop an empirical framework for identifying decision-makers’ ordinal preferences given limited data. We show that preference identification hinges upon understanding the empirical relationship between decision-makers’ preferences and their consistency. By recasting the behavioral preference recovery problem in these terms, familiar insights from the program evaluation literature can be fruitfully adapted to this new setting. We illustrate our proposed techniques with data from a range of recent empirical studies.