The Desirability of Cheating Incentives and Randomnness in the Optimal Income Tax

Abstract

A common feature of income tax plans is the opportunity to evade taxes illegally. To make this unattractive, governments audit a random sample and punish agents who cheat, usually with fines. The decision to cheat, from the agent's viewpoint, is analogous to the purchase of a risky asset. The availability of risky financial assets will generally affect the relative attractiveness between labor and leisure. It follows that the creation of incentives to cheat will alter the amount of labor agents wish to supply for any given wage and tax rate. It is well known that an income tax distorts the marginal incentives and lowers the supply of labor below that which agents would choose if, instead, a lump-sum tax which yielded the same revenue were levied. This paper will explore the possibility that incentives to cheat might offset this undesirable substitution effect and improve resource allocation sufficiently to compensate agents for their usual aversion to risk. As in Mirrlees's (1971) important article on the optimal redistributive income tax, an atemporal, artisan economy (no firms, no capital, constant marginal product, labor-leisure trade-off) is analyzed. The government employs an income tax to raise an exogenous amount of revenue. For expositional simplicity, only linear tax schedules are examined, but the restriction is not crucial. The real-world consideration that much effort is wasted by devising and thwarting evasion schemes is assumed away by postulating that the acts of evasion and detection are, in themselves, costless. Lump-sum taxation is ruled inadmissable, so that the problem is inherently in the nature of the second best. A simple model of tax evasion formulated by Allingham and Sandmo (1972), Srinivasan (1973), and others is extended to allow for variable labor supply. It is observed that, starting from a position in which no one