Update and Summary: Economic Activity Measures

New aggregate indicators on the macroeconomy are out. How do they compare against a summary measure of the macro series the NBER BCDC focus on?

A week ago, Macroeconomic Advisers released their estimate of April GDP, while e-Forecasting released their estimate of May GDP. These two series are depicted in Figure 1.

Recall that the NBER BCDC examines four other variables to gauge economic activity: payroll employment, industrial production, real manufacturing and trade sales, and real personal income less transfers [1]. To see how GDP has moved differently from these other measures of economic activity, see this post. Rather than providing a welter of series, I’ve tried to summarize the movement in these four variables by generating an index that is the first principal component of the four underlying series (all logged). This is the blue line in Figure 1.

bcdcpc1.gif

Figure 1: First principal component of log payroll employment, industrial production, personal income less transfers, manufacturing and trade sales (blue); and log GDP (Macroeconomic Advisers (red), and e-Forecasting (green). Source: Macroeconomic Advisers, June 15 release; e-Forecasting June 17 release; FRED II; and author’s calcluations.

Loosely speaking, the first principal component is the linear combination of the underlying variables that has the maximum variance. In a four variable grouping, four principal components would explain all the variance of the grouping. The first principal component in this case accounts for about 96% of the total variance. (The loadings are about equal.)

What’s clear from this plot is that Macroeconomic Advisers’ measure of GDP peaked in January 2008, and e-Forecasting’s in March. (By the way, this jagged pattern in the Macroeconomic Advisers series is partly why the quarterly changes in the official BEA measure will not necessarily register a negative change q/q even if there is a 3 month changes in monthly GDP; see Jeff Frankel’s discussion.) The first principal component peaks earlier than both the GDP measures, in October of 2007, although there is a second peak that is close to the first, in January 2008. The drop from the peak is “only” 5.4% in log terms.

This summary measure ends in March because of the unavailability of manufacturing and trade sales for April. What is reported is retail and foods sales. While not conceptually similar to the former, the two series move together (in levels, at medium frequencies). The slope coefficient of a regression of log manufacturing and trade sales against log retail sales is 1.11, and an adjusted R2 of 0.98. Using this alternative measure, one obtains the alternate first principal component depicted in Figure 2.

bcdcpc2.gif

Figure 2: First principal component of log payroll employment, industrial production, personal income less transfers, retail sales ex food sales (blue); and log GDP (Macroeconomic Advisers (red), and e-Forecasting (green). Source: Macroeconomic Advisers, June 15 release; e-Forecasting June 17 release; FRED II; and author’s calcluations.

This alternative summary measure alos records a decline, in this case since the peak of November 2007: 3.7 percent.

All this is to say, GDP is an important indicator. But it’s not the only indicator of “economic activity”. Something to keep in mind when you read statements such as: [2]

The definition of a “recession” is very clear and straightforward: Two consecutive quarters of negative growth. We have not yet had one consecutive quarter of negative growth.

No, not quite so clear…especially after keeping in mind the possibility of revisions. [3]


Originally published at Econbrowser and reproduced here with the author’s permission.