Over the past decades awareness has grown that analyses based on the natural variation of choices and outcomes in a population are often a poor guide for optimal labor market policies, as they do not reveal the underlying parameters that are relevant for these policy decisions. Therefore, there has been an increased interest in research to estimate parameters using field or natural experiments. Such experiments give in a statistical sense more robust estimates, which apply to much more specific situations, though. Consequently, without coordination between policy and research, policy lacks precise and useful information to guide its policies. In this paper we explore what could happen if experimenting would become an integral part of labor market policies, providing policy makers with continuous information about parameters that are relevant for their decisions. We apply this to a simple case of income tax rates. In simulations we explore the effects of a policy in which tax rates are set which include some degree of randomization. This randomization provides the government with new information that allows for a continuous update of the tax rate. We discuss the costs and benefits of such an approach. We investigate the probability that this approach converges to a close to optimal tax rate and how long it will take before such optimum is reached. |
