An agent-based model of individual income tax compliance

Abstract

An agent-based model describing the dynamics of individual income tax compliance is presented using an iterated coordination game on a social network. The taxpayers' decisions whether to be honest or to underreport their income are influenced by rules governed by interactions with other taxpayers, prior audit experiences, and conformity to a social norm of honesty. Results show that increasing penalty rates and audit probabilities lead to a step-like transition in the increase of average compliance level. Moreover, even if a large fraction of dishonest taxpayers do not conform, high compliance levels are still possible.