Are data unimportant in business cycle analysis?

Abstract

Recently, Michael Bergman and Lars Jonung (BJ) concluded that economic fluctuations in Sweden and in the U.S. over a hundred-year perspective have not been dampened.¹ In short, their method was “to measure the volatility as standard deviations in trend-adjusted data, and test for potential differences in variances across the periods 1873–1913 and 1948–1988.” (p. 24) In this comment I am not going to discuss their methodology per se, but instead the underlying production data for Sweden. Nobody would claim, I am sure, that even the most excellent econometric methodology can lead to robust and reliable results of an analysis when the data material is inappropriate or insufficient.