Welfare-improving tax evasion

We study optimal income taxation in a two-group framework where the private cost of misreporting income is positively correlated with productivity. If high-wage types always reveal their income truthfully, then letting low-wage types cheat leads to Pareto-superior outcomes regardless of the audit costs (as compared to deterrence). With no cheating, redistribution takes place on first- or second-best frontiers and low-wage types always end up worse off than high-wage types. Letting low-wage types misreport reduces the need to recourse to second-best mechanisms. Additionally, it increases the reach of first-best redistribution to outcomes at which low-wage types are better off than high-wage types.