Fiscal Multipliers in Globally Solved HANK Models with Aggregate Risk

Abstract

We show that a nonlinear perfect-foresight solution to standard HANK models, computed using standard fast methods, provides a good approximation to the full global solution with aggregate risk, except for a level shift driven by precautionary saving. This result holds in calibrations featuring extensive sign, size, and state dependence in the effects of fiscal policy. We build on, and extend, recent advances in solving heterogeneous-agent models in moving-average (MA) form using deep learning, and confirm the accuracy of this approach.

Draft available on request.

Citation

@techreport{DruedahlHuleuxRopke2026b,
author = {Jeppe Druedahl and Raphaël Huleux and Jacob Røpke},
year = {2026},
title = {Fiscal Multipliers in Globally Solved HANK Models with Aggregate Risk}}