Estimating linearized heterogeneous agent models using panel data

We develop a method to estimate heterogeneous agent models that uses not only time series of macroeconomic aggregates, but can also incorporate micro level data (repeated cross-section or panel). The micro data may be collected at lower frequency and time-aggregated. The method is based on the linearization approach of Reiter (2009), combined with optimal state aggregation as in Reiter (2010). The model may contain decision problems with both continuous and discrete choice. Linearity of the model solution allows fast computation of second moments and likelihood. We discuss various computational devices to maximize the speed of the estimation. (C) 2020 Published by Elsevier B.V.