Heuristic Driven Agents in Tax Evasion: An Agent-Based Approach

The Allingham and Sandmo model is an adaptation of the standard expected utility maximization framework where the taxpayer is defined as a representative agent who is coping with a risky choice. The main limit of this model regards the assumption of perfect rationality from the agent's side and the impossibility to study at the macro level a situation where many heterogeneous agents interact together. The aim of this work is to try to overcome, at least partially, some of the neoclassical standard approaches in this field. More precisely, we present a very simplified, agent-based, fiscal system with heterogeneous tax payers, interacting within a public good game framework. Heterogeneity has been introduced in our model by designing the agents like simple heuristics. The environment has been designed by a voluntary supply public good context reinforced through tax audits and fines. Looking for more realism, we also allowed agents to mutate their heuristics and we introduced two cross sectional types of agents: the "employees" and the "self-employees" allowing our agents to switch from one type to the other and vice-versa. Finally, the system is dynamic over time, since new agents can join over time to mimic the idea of having a growing population over time. We obtained is a complex adaptive system (CAS) with heterogeneous agents, dynamically evolving, able to describe the adaptation of agent's behaviour either concerning evading decision and the preferred kind of heuristic.