Computational models of tax evasion with heterogeneous agents

The fundamental purpose of my dissertation is to further understand the roles played by taxpayer heterogeneity and social influence in tax-evasion dynamics, and to derive normative insights about optimal policy design considering both rational and non-rational taxpayers. How should fiscal policies be designed under the consideration of heterogeneous taxpayers? What is the role of social dynamics in public policy implementation? Which are the implications of rationality and (bounded) non-rationality of taxpayers in policy design? To answer these and similar questions, the first objective of the current thesis is to build on recent developments from behavioral economics, experimental economics and economic psychology to derive models with heterogeneous agents which have strong empirical and analytical micro-foundations. Mathematical models of individual and collective taxpayer behavior, however, are often too complex to be fully characterized analytically. To overcome this obstacle, we incorporate state-of-the-art computational models to simulate taxpayer behavior under different fiscal environments and to study the potential implications of diverse policy implementations. The second objective of my research concerns the optimal fiscal policy design under the consideration of taxpayer heterogeneity and social influence dynamics, with a normative focus. Acknowledging that taxpayer behavior evolves according to different dynamic processes, which depend on how taxpayers adapt their expectations about future enforcement regimes and penalty possibilities, we employ recent insights from game theory and network theory applied to social networks and information diffusion to capture the effects and repercussions of social influence dynamics. It is of utmost importance for policymakers to understand not only the individual decision-making processes of taxpayers, but also the emergence of aggregate behaviors as a consequence of idiosyncratic beliefs, social interactions, policy announcements, taxpayer expectations, diffusion of tax-related information, enforcement measures and potential policy implementations. The third objective of this work aims to design optimal fiscal policies in diverse economic scenarios where a social planner faces strategic tax evaders who can observe and react against the implemented fiscal policies. Harsher audit and penalty rates may affect taxpayer behavior. However, taxpayers' expectations, social influence and non-pecuniary factors also play prominent roles in both individual and collective taxpayer behaviors. The literature recognizes that taxpayer behavior is influenced not only by the probability of being audited and the possible incurred penalties, but also by tax rates, feelings of regret, uncertainty, risk aversion, psychic costs, peer effects, social interactions, social norms, the efficiency of government expenditures, the perceived power of the government, tax morale and the degree of trust that citizens place in the authority. Following, a central challenge is how to derive appropriate policy recommendations while considering the different ways in which taxpayers may behave, react, optimize and carry out their respective decision-making processes. The research presented in this work combines a set of interdisciplinary insights and findings from behavioral economics, experimental economics and economic psychology to derive models that formally characterize the prevalent theoretical and empirical results in tax-evasion literature. We also consider recent developments from network theory to model taxpayer social networks and to study peer-to-peer communication, the spread of information along the network, and the resulting social influence dynamics. Moreover, we aim to apply sophisticated game-theoretic techniques to identify the key individuals and links in a taxpayer network which ought to be targeted by a social planner or policymaker in order to minimize the aggregate level of tax evasion.