Estimation and identification in macroeconomic models with incomplete markets

This dissertation contributes to the econometric framework of macroeconomic models with heterogeneous agents. Heterogeneous agent models are used to study important questions related to the interaction of macroeconomics and inequality. However, where the economic literature is well-acquainted with econometric methods used for macroeconomic models with representative agents and the identification issues that one may encounter when trying to estimate these models, this is not the case for heterogeneous agent models. In heterogeneous agent models, as shown in the first part of this dissertation, new econometric challenges arise. For example, not all parameters are identifiable from aggregate variables, and to estimate them, additional information is needed, coming from cross-sectional or panel data. In response, this dissertation proposes a two-step identification and estimation procedure that can easily deal with these different data types and simplifies the identification analysis. The second and third part of this dissertation focus on a particular element in heterogeneous agent models, in particular, the earnings process that households in these models face. Because of the way heterogeneous agent macro models are solved, these earnings processes should have a discrete state space, yet most of the existing econometric literature focuses on continuous-support earnings processes. The second part of this dissertation proposes, identifies, and estimates a novel earnings process that features a rich notion of earnings risk heterogeneity, yet is discrete. In the last part, an improved method is proposed that can be used to discretize existing continuous-support stochastic processes.