Probabilistic Lump-Sum Taxation

In this paper, we describe an elementary, fully implementable, and revenue neutral randomizing mechanism that leads to a Pareto improvement over allocations induced by labor income tax functions. We illustrate, by providing a theoretical example, that our mechanism can be applied to optimal tax functions identified by Saez and Diamond. Furthermore, we provide an explicit numerical example, which confirms that our procedure can lead to a Pareto improvement over allocations induced by optimal tax functions of Mirrlees. Moreover, we show that our randomizing procedure is not only implementable, as it requires less than the standard informational demands, but can also lead to sizable welfare gains amounting to about 10% of the underlying distortion. Finally, we show that, unlike most of the traditional approaches involving randomness, the mechanism we describe can preserve horizontal equity.