Policy Uncertainty and Optimal Taxation

A kind of folk theorem in tax policy states that too much uncertainty about the impact of taxing (or subsidizing) a particular taxable attribute is cause for excluding that attribute from the tax base. I extend the optimal tax model to test this hypothesis. In my model, the government is not only unable to observe the underlying characteristics of any given taxpayer - as per the usual model - it is also uncertain regarding the population distribution of such characteristics. I study how optimal taxation is affected by a mean-preserving spread of the government’s beliefs with regard to potentially taxable attributes holding constant government beliefs with regard to already taxed attributes. I find that such additional uncertainty is as likely to increase as to decrease both the impetus for including potentially taxable attributes in the tax base and the intensity at which such attributes are optimally taxed if included. I highlight the decisive role of correlation across the components of the additional uncertainty regarding potentially taxable attributes. I connect this uncertainty covariance to the population covariance often found in optimal tax formulae following Diamond (1975)