Adam Klug, Ben-Gurion University
Eugene White, Rutgers University

This paper uses a unique historical survey to study business expectations in the U.S. during the Great Depression. The survey consists of forecasts of railway traffic made by traffic managers and covers 29 commodity groups and 91% of all railway traffic. We show that an appropriately weighted aggregate of these commodity groups is highly correlated with industrial production. These data show that on aggregate, businessmenŐs forecast errors rose from 10% to 30% at the height of the Depression and that errors of similar magnitude were made during the recovery. An optimal ARMA model, estimated for the period 1919-1929, forecasts better out of sample than the railroad shippers forecasts, showing that they did not use past information efficiently. At the disaggregate level, we find that over half the forecasts were biased, although the bias is too small to support the view that the poor forecasting performance was the outcome of strategic behavior. We also find that these data reject the hypothesis that individual behavior in the Depression was characterized by rational expectations.