We show that the sequence-space Jacobians of stationary models have a special "quasi-Toeplitz" form. This result implies a simple test for existence and uniqueness of solutions, helps significantly reduce truncation error, and allows the solution of very large sequence-space systems. We apply these insights to a heterogeneous-agent New Keynesian model, showing how to identify thresholds for determinacy and also how to solve the model using fast iterations rather than direct operations on truncated matrices. We leverage these results to solve a 177-country version of the model with a realistic trade network almost instantly, in spite of its state space having dimension of almost 1 million.
I develop and estimate a quantitative Heterogeneous-Agent New Keynesian (HANK) model for macroeconomic forecasting and policy analysis. It is designed to compete with its Representative-Agent (RA) equivalents and with Vector Autoregressions (VARs) in out-of-sample forecasting of aggregates. I also assess the HANK model's strength in predicting the effects of targeted policies. I then analyze the drivers of the model's forecasting effectiveness and outline a framework for its use in policy analysis.
I develop methods for inference on historical decompositions in Bayesian Vector Autoregressions (BVARs) with external instruments and missing data to investigate the role of different shocks in the "Great Inflation," a peacetime inflation episode in the US from the 1960s to the 1980s. I find evidence that oil supply shocks were among its primary drivers. Moreover, oil supply shocks continued to play a meaningful role in the behavior of economic aggregates in the decades since. Monetary shocks also explain a considerable amount of variation in aggregates over the sample period.