The Estimation of Prewar Gnp Volatility, 1869-1938

Abstract

New evidence is provided to assess the recent controversy regarding the volatility of real economic activity before 1929 relative to the period since World War II. Some recent work claims that the longstanding stylized fact of greater prewar volatility is "spurious". In contrast, this paper reconfirms the greater amplitude of business fluctuations prior to the Great Depression. The basic technique is the regression method, which estimates equations for real GNP during 1909-38, with one or more explanatory variables for components of GNP, and then uses the estimated coefficients to "backcast" real GNP or the period 1869-1908. The paper contains an extensive examination of the sensitivity of these regression indexes to alternative dependent variables, sample periods, detrending methods, and the inclusion of alternative explanatory variables. Particular attention is paid to the conflicting evidence regarding the amplitude of cycles in construction activity between 1870 and 1890. The resulting prewar/postwar volatility ratios, for 1869-1928 as compared to 1950-1980, range from 1.43 to 2.16. The paper concludes by suggesting that this range of volatility ratios is more likely to understate than overstate the prewar/postwar volatility ratio.