Pushing the Frontier of Heterogeneous Agent Macro: Expectations Formation and Economic Crises

One of the key developments in macroeconomics research over the last three decades has been the incorporation of explicit heterogeneity into models of the macroeconomy. But one of the all-time biggest questions in macroeconomics has so far been out of reach for this approach: why do developed economies experience infrequent but large economic crises? This is due to a problematic feature of current heterogeneous agent (HA) models: the assumption of rational expectations means that agents forecast prices by forecasting distributions resulting in the entire cross-sectional distribution entering their decision problem. This extreme version of the curse of dimensionality hugely increases the models’ computational complexity.HA-FRONTIER aims to overcome this problem by developing alternative approaches to the rational expectations assumption. I will adapt ideas from the literature on reinforcement learning in psychology and computer science (temporal difference learning) to better model how agents form expectations about future prices. I will incorporate key empirical evidence on expectations formation, e.g. from the large literature on survey expectations. Departing from rational expectations thus “kills two birds with one stone”: it makes HA models with aggregate risk operational while, at the same time, making them more realistic. Finally, I will put this methodology to use to develop an empirically grounded macroeconomic theory of infrequent but large economic crises that incorporates asset-price booms driven by individuals reallocating their portfolios toward bubbly assets and crashes driven by the reverse movement.To disseminate the ideas developed by HA-FRONTIER, I will write a PhD-level textbook on heterogeneous-agent macroeconomics. Such a textbook is currently missing, thereby creating unnecessary entry barriers for junior researchers, practitioners, and policy makers.