International growth and volatility in historical perspective

Abstract

This paper studies the relationship between the volatility and growth of real GDP using a newly constructed panel data set from twelve countries over the 1870 to 1929 period. In addition, many other variables are examined that are related to economic growth. The goal has been to uncover robust empirical regularities on this issue for the period prior to the Great Depression – a period which has been relatively neglected in previous empirical work. The main finding is that there is a robust negative partial correlation between volatility and growth, after controlling for other factors. This result is consistent with recent empirical evidence on the post-World War II period.

Acknowledgements

Previous versions of this paper were presented at the Western Economic Association International Conference, the University of California, Davis, Bates College, and California State University, Sacramento, and I am grateful to participants for their comments and suggestions. I would especially like to thank David Aschauer, Kevin Hoover, David Lang, Peter Lindert, Margaret Maurer-Fazio, Stephen Perez, Alan Taylor, and two anonymous referees, although any errors and omissions are my own.

Notes

See Aghion and Howitt (1997), Chapter 8.

Acemoglu et al. (2003) and Fatas and Mihov (2003) also report a negative relationship between volatility and growth, but argue that the fundamental reason is due to deeper institutional causes such as poor economic policy. Both of these papers also rely on the Heston and Summers data set.

Exceptions are Zarnowitz and Moore (1986), and Altman (1995). Zarnowitz and Moore (1986) studied the cyclical behaviour during the past century in the USA, and noted that the standard deviation of growth tends to be higher during periods of slower growth. Altman (1995) examined the correlation between growth and volatility for thirteen countries over the 1870 to 1986 period. However, Altman only examined the raw correlations between volatility and growth, and not the multitude of variables examined in this paper.

The countries include Australia, Canada, Denmark, Finland, France, Germany, Italy, Japan, Spain, Sweden, the UK, and the USA.

This paper also uses data for Spain and the USA not included in Maddison, although the growth and volatility results are robust when Spain and/or the USA are excluded from the sample.

Due to space limitations, these additional results are not reported here. However, all of the results discussed are available from the author.

The total number of observations is usually 62 (out of a possible 72) because of missing data for Japan in the 1870s and 1880s, as well as missing data for some countries during the interwar period (1910s and 1920s), most particularly for France, Germany, and Italy.

A binary variable equal to one is included if a country was on the gold standard a majority of years during a given decade. This variable was constructed from the annual data kindly provided by Meissner (2002).