Comment

Let me begin with a summary of this interesting paper before making some comments. Aruoba et al. use a two-step procedure to fit a factor model to aggregate data from the Group of 7 countries. In the first step, they take each of the seven countries in isolation and fit a one-factor model to growth rates or ratios of six series: quarterly GDP and disposable income, along with monthly employment, industrial production, retail sales, and initial claims for unemployment insurance. In the second step, they fit a one-factor model to the seven country factors constructed in the first phase. The country factors constructed in the first step have some reasonable and appealing properties. Almost all series load positively and significantly on the factor, with the exception of initial claims for unemployment insurance (top of table 2); the U.S. factor tends to fall sharply during NBER recessions (fig. 14); a fall in volatility (also known as the Great Moderation) is clear in most of the country factors (figs. 8–14). Things we learn from the country factors include the following: GDP generally has the largest effect on the country factor (panels A and B of table 3); factors generally are highly serially correlated (middle of table 2— somewhat surprising since most of the six observable series are entered in growth rates); factors are positively correlated across countries, with a correlation that changes over time (table 5); because of the ongoing crisis, this correlation is highest in the most recent decade (table 5). The global factor constructed in the second step also has some reasonable and appealing properties. All seven country factors load significantly on the global factor (table 6); the global factor falls sharply during worldwide recessions such as the 1970s and 2000s (fig. 16). Something we learn from the global factor is that some recessions occur when the global factor declines, some when thecountry-specific idiosyncratic component declines (discussion in Sec. V.B.2).