Abstract
In this paper, we analyze a discretized version of the dynamic programming algorithm for a parameterized family of infinite-horizon economic models, and derive error bounds for the approximate value and policy functions. If
h is the mesh size of the discretization, then the approximation error for the value function is bounded by
Mh
2, and the approximation error for the policy function is bounded by
Nh, where the constants
M and
N can be estimated from primitive data of the model.