Learning Asymmetries in Real Business Cycles

When a boom ends, the downturn is generally sharp and short. When growth resumes, the boom is more gradual. Our explanation rests on learning about productivity. When agents believe productivity is high, they work, invest, and produce more. More production generates higher precision information. When the boom ends, precise estimates of the slowdown prompt decisive reactions: Investment and labor fall sharply. When growth resumes, low production yields noisy estimates of recovery. Noise impedes learning, slows recovery, and makes booms more gradual than downturns. A calibrated model generates growth rate asymmetry similar to macroeconomic aggregates. Fluctuations in agents’ forecast precision match observed countercyclical errors of forecasters. “There is, however, another characteristic of what we call the trade cycle that our explanation must cover; namely, the phenomenon of the crisis the fact that the substitution of a downward for an upward tendency often takes place suddenly and violently , whereas there is, as a rule, no such sharp turning point when an upward is substituted for a downward tendency.” J.M. Keynes (1936) ∗New York University, Stern School of Business, Department of Finance †Corresponding author: New York University, Stern School of Business, Department of Economics. [email protected], 44 West 4th St., New York, NY 10012, tel: 212 998-0527, fax: 212 995-4218. ‡First version: May 15, 2001. First submission: July 3, 2003. Many thanks to Tom Sargent, Tim Cogley, Paul Romer, Peter Henry, Bob Hall, Kenneth Judd, Pierre-Olivier Weill, Hanno Lustig, John Shea, Tim Johnson, Robert Rich, Robert King, and attendees of the Stanford Macro Lunch, the 2002 meetings of the Econometric Society in Los Angeles, the conference on learning and bounded rationality at the University of Illinois Urbana Champaign, and the economics seminar at the University of Leuven, Munich and Humbolt University for their helpful advice. Stijn Van Nieuwerburgh thanks SIEPR and the Flanders Fund for Scientific Research for financial support. All remaining errors are our own. JEL classification: E32, D83, E27.