Overborrowing Crises and The Role of Expectations

MARTIN GUZMAN

Providence

Seminario; Macroeconomics Seminar of Brown University; 2013

Brown University

I study the role of expectations and learning about productivity trends in the occurrence of overborrowing crises. I consider two different types of learning in a quantitative small open economy model of default. The first is a Bayesian process in which agents use the Kalman filter to learn about the nature of output shocks. The second is a non-Bayesian stochastic-gain learning process characterized by a gain learning coefficient that is related to forecast errors. The model is calibrated for two emerging economies - Argentina and Mexico - and one developed economy - Canada. In order to match the observed frequency of crises in emerging countries, the variance of expected permanent income must be large. In Argentina and Mexico, the two learning models I consider significantly raise the variance of expected permanent income in comparison to full information rational expectations. As a result, the models with learning, especially with stochastic-gain learning, improve the match between theoretical and actual frequencies of crises in emerging economies.