Tax reform from the gradient projection viewpoint

This paper examines the problems of gradual tax reform. Desirable small moves of the tax system are determined using the principles of the well-known gradient projection algorithm. Three classical models of public finance are considered. For each of them a continuous time dynamic process is exhibited along which the social welfare function increases until convergence to local second best extrema. Although the selection of trajectories according to the gradient projection method has some arbitrariness, it leads to intuitively and economically appealing rules such as a tax reform analog of the classical optimal taxation result on production efficiency.