A Markov-switching dynamic factor framework for dating global economic cycles

Abstract

An important issue in identifying global recessions is the limited availability of output data at the quarterly and monthly frequencies over longer time horizons. A related issue is the heterogeneity in evidence about specific recessionary episodes. We utilize the context that commodity prices are determined in the global markets, and four base metals have flexible nominal prices at the monthly frequency from the 1960s. We use the base metal prices to supplement the information about the global economy in the GDP data of 32 countries and the World Industrial Production index. We estimate the quarterly episodes of global recessions from the 1960s using extended Markov-switching dynamic factor models with multiple indicators. We further adapt the quarterly models to a mixed-frequency Markov-switching dynamic factor model to estimate the monthly episodes. Our estimates show eight episodes of global recessions at the quarterly frequency. Monthly estimates also capture the eight quarterly episodes of global recessions. The results are robust to several model and data sensitivity analyses. Regressions using 32 countries show reductions in GDP growth for all countries during the global recession episodes. Further analysis shows that the four global recessions that are common with other studies are deeper and more widespread recessions than the other four downturns. The analysis highlights heterogeneity in the size and the spread of global recessions while providing empirical evidence in favor of four specific recessions with mixed support in the past literature.