Tax Enforcement with Somewhat Honest Taxpayers

Standard compliance theory assumes that individuals evade to the extent it benefits them monetarily. However, a growing empirical literature suggests that many underlying cognitive considerations, including lying aversion, may play a non-trivial role when individuals make decisions. This study aims to analyze whether optimal audit rules that are the result of standard models survive a model with agents that are lying averse. I show the canonical cut-off audit rule where above a certain threshold no reported income is audited (introduced by Reinganum and Wilde (1986a) and further developed by Sanchez and Sobel (1993)) is not optimal in this setting. Moreover, a Bayesian incentive compatible audit probability does not need to be monotone in reported income, i.e. a higher report might be subject to a higher probability than a lower report.