Abstract
We present a refinement of the uniform truncation method of LeGrand and Ragot (2022) to solve heterogeneous-agent models with aggregate shocks. The method consists in providing a finite state-space representation of such economies by truncating idiosyncratic histories. The innovation compared to the uniform method is to allow for truncated histories of different lengths. This offers a finer representation when needed, while considerably reducing the model dimensionality. The method reproduces the steady-state distribution of any heterogeneous-agent model and solves for its dynamics in the presence of aggregate shocks. As with the uniform method, the refined method can be solved using perturbation methods and hence implemented with standard software, such as Dynare. We show that the refined truncation method provides accurate results that improve on those of the uniform method.