UC Berkeley

We compare dynamic taxes and quotas in a stationary environment where a regulator and a non-strategic firm have asymmetric information. The regulator is able to learn about the unknown cost parameter either by using a tax or a quota that is slack with positive probability. With a tax, the information asymmetry is resolved in one period. Optimal learning using a quota is less transparent, though we show that this search problem has a simple solution. In particular, it is never optimal for the regulator to learn gradually. In the first period, he either ignores the possibility of learning, or he tries to improve his information. Regardless of the outcome in the first period, he never experiments in subsequent periods. We use this result to assess the informational advantage of taxes compared to quotas under asymmetric information.