The Length of U.S. Business Expansions: When Did the Break in the Data Occur?

Abstract

It is widely accepted that the character of U.S. business cycles has changed in the post World War II era. In particular, as measured by the NBER business cycle dates, business expansions have grown from an average of 26.5 months in the prewar period to 51.5 months in the post-war period. Meanwhile, the length of contractions has fallen from 21.2 months prewar to 10.4 months in the post war era. While the changed character of the U.S. business cycle is generally accepted, the explanation for this change is not. World War II provides a convenient breakpoint in data set on business cycles, but in fact this choice is somewhat arbitrary. In this paper, we allow the data to choose the breakpoint. Following Diebold and Rudebusch (AER 1992) we conduct a nonparametric test to check for a break in the data. However, we vary the date at which we break the sample and compute a test statistic for each proposed break. The proposed date for which the probability of a break in the data set is maximized is chosen as the date for the which the break in the data occurred. For expansions, we find that March 1933 marks the breakpoint in the data. In April of 1933, the United States left the gold standard. Thus, the break in the series on business cycle expansions coincides with a major economic event that itself is theoretically consistent with this change in behavior. In particular, the timing of these two events is consistent with macroeconomic models in which there is a short-run nonneutrality of money and in which discretionary monetary policy has some scope for reducing business cycle fluctuations. We use NBER business cycle dates in conducting our analysis, but we check the robustness of our results by using an alternative set of business cycle dates.