Computational accuracy and distributional analysis in models with incomplete markets and aggregate uncertainty

This paper shows that grid-based numerical solutions to models with incomplete markets and aggregate uncertainty are sensitive to the number and placement of grid points in the aggregate asset holdings direction. Higher moments of the cross-sectional distribution of asset holdings can be particularly affected, which is important for welfare analysis. The paper recommends using grids that have few grid points for aggregate assets overall but are denser around the mean of the ergodic distribution of individual asset holdings. The fact that the accuracy of the approximation to decision functions can be much improved this way is captured by Den Haan’s (2010) dynamic Euler equation test but not by the traditionally used R2 statistic.