Disecting the Cycle: A Methodological Investigation

Macroeconomics has a long tradition of inspecting and interpreting patterns in graphs of aggregate data. However, the move towards more precise quantification of macroeconomic phenomena has seen academics shift away from a study of turning points, which are a natural and obvious way of summarizing business cycles, towards measures of co-movement in detrended series. This shift arise from several developments, but an important one was the belief among academics that Burns and Mitchell's methods lacked the statistical basis and, hence, the precision required in modern macroeconomics. We adopt the older perspective that business cycles are to be defined in terms of the turning points in the level of economic activity. We show that such turning points can be associated with a well defined sequence of outcomes and can therefore be precisely analyzed. In turn this enables us to explore how various parametric models of aggregate output generate a cycle through the interaction of trend movements in activity with the volatility and serial correlation in growth rates. One of the strongest points in the rhetoric of modern business cycle theory is that trend and cycles should not be divorced. Consequently, any definition of the business cycle in terms of the co-movement of detrended data has to find the task of integration a difficult one. In contrast, we show that a return to the older tradition of studying the classical cycle in the level of economic activity produces a natural interpretation of the origin of the cycle in terms of the interaction of trend and the second moments of growth rates. This seems a critical advantage for the approach taken in this paper. An important issue that has also been debated in the literature is whether non-linear models are required to make a business cycle. Using the techniques developed in this paper we dissect the cycle of a number of countries and find little evidence that non-linearities, of the type investigated in the literature, are important in accounting for the broad features of the average cycle.