Working Paper: CEPR ID: DP6963
Authors: Wouter Den Haan; Pontus Rendahl
Abstract: We construct a method to solve models with heterogeneous agents and aggregate uncertainty that is simpler than existing algorithms; the aggregate law of motion is obtained neither by simulation nor by parameterization of the cross-sectional distribution, but by explicitly aggregating the individual policy rule. This establishes a link between the individual policy rule and the set of necessary aggregate state variables. In particular, the cross-sectional average of each basis function in the individual policy rule is a state variable. That is, if the individual capital stock, k, (or k²) enters the policy function then the mean of k (or the mean of k²) is a state variable. The laws of motions for these aggregate state variables are obtained by explicit aggregation of separate individual policy functions for the different elements.
Keywords: Numerical solutions; Projection methods
JEL Codes: C63; D52
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| individual policy rules (G52) | aggregate outcomes (E10) |
| individual behavior (D01) | aggregate laws of motion (C69) |
| cross-sectional averages of individual choices (D79) | aggregate capital stock (E22) |
| individual capital stocks (E22) | aggregate capital stock (E22) |
| first i moments of cross-sectional distribution (C46) | future aggregate states (C43) |
| cross-sectional distribution (D39) | accurate aggregate predictions (C59) |