Learning leverage shocks and the Great Recession

This paper develops a simple business-cycle model in which financial shocks have large macroeconomic effects when private agents are gradually learning their economic environment. When agents update their beliefs about the unobserved process driving financial shocks to the leverage ratio, the responses of output and other aggregates under adaptive learning are significantly larger than under rational expectations. In our benchmark case calibrated using US data on leverage, debt-to-GDP and land value-to-GDP ratios for 1996Q1-2008Q4, learning amplifies leverage shocks by a factor of about three, relative to rational expectations. When fed with the actual leverage innovations, the learning model predicts the correct magnitude for the Great Recession, while its rational expectations counterpart predicts a counter-factual expansion. In addition, we show that procyclical leverage reinforces the impact of learning and, accordingly, that macro-prudential policies enforcing countercyclical leverage dampen the effects of leverage shocks. Finally, we illustrate how learning with a misspecified model that ignores real/financial linkages also contributes to magnify financial shocks.