Learning (Not) to Evade Taxes

In this paper, lab experiments on tax compliance were theoretically investigated with dynamic and stochastic methods. It is well known from experimental games that learning allows a better understanding of participants’ behavior. However, it has not been explicitly applied so far in the theoretical analysis of tax compliance experiments. In this paper, it was shown that two decision-making processes may be distinguished: a discrete process in which all options are regarded and an all-or-nothing process in which either the respective tax is paid fully or not at all. The corresponding variant of the learning model was either a stochastic or a deterministic one, with the stochastic version as the more general model. In the additional empirical part of the paper, it was shown that tax payments decline in trend over the rounds of the considered experiment. This negative trend was interpreted as a learning effect, in accordance with the stochastic version of the theoretical model. However, the alternative interpretation that the observed behavior was driven by a tiring effect cannot be completely excluded.