Solving Heterogeneous Agent Models with Non-convex Optimization Problems: Linearization and Beyond %

This paper presents methods for heterogeneous agent models where agents solve non-convex optimization problems. It shows how to apply the linearization approach of Reiter (2009) to non-convex models, and develops a theory of state and value function reduction to handle models with very large state spaces. It shows the potential problems of the linearization approach and ways to diagnose them. To overcome these problems, global nonlinear solution algorithms are presented, based on temporary equilibrium concepts. The methods are applied to models with heterogeneous households and indivisible labor, as well as to a model of heterogeneous firms with lumpy investment. \end{abstract}