GDI is a better indicator of the business cycle than GDP

24 Aug 2011 by Jim Fickett.

GDP, based on expenditure data, and GDI, based on income data, both attempt to estimate the same abstract quantity – the value added by all economic activity in the US. But the difference between the two is cyclical, suggesting one picks up on the business cycle better than the other. Checking the relationship with various indicators shows that GDI correlates better with the business cycle.

In an earlier post ("It's tough to make predictions, especially about the future") I mentioned the work of Jeremy Nalewaik, a Fed economist who proposes that there is a separately detectable “stall” period that precedes some recessions. His work places more emphasis on Gross Domestic Income than Gross Domestic Product, and I wanted to see what his justification was for this preference, before delving into the details of his most recent prediction algorithm.

In Apr 2010 he published a paper entitled, The Income- and Expenditure-Side Estimates of U.S. Output Growth. The paper is approachable, interesting in the details, and quite convincing. For those that don't have time, what follows is a concise summary of one line of reasoning from the paper that I found particularly persuasive.

GDI and GDP are both attempts to measure the same thing – the total value added by all economic activity in the US. GDI looks at income data, and GDP looks at expenditure data. Although both estimates should, in theory, lead to the same number, the data used is quite different, and the difference in estimates, termed the “statistical discrepancy”, is substantial.

Most people assume the statistical discrepancy is just random noise. But Nalewaik (and others) note that in fact the statistical discrepancy is correlated with the business cycle.

The correlation is far from perfect, and does break down at times, but it is clear that the statistical discrepancy is not just random noise. Note that the correlation seem to be better in recent years. Nalewaik suggests that this is due to the service sector of the economy growing in relative importance.

the correlation between ∆GDP(E) [quarterly change in GDP] and ∆GDP(I) [quarterly change in GDI] dropped sometime around the mid-1980s, and the divergences between the estimates also became highly cyclical around that time. …

While PCE for services has always held a relatively large share of GDP(E), averaging 30 percent from 1947 to 1984, its share shot up to an average of 43 percent from 1985 to 2009, and the share reached 48 percent in 2009. As the share of services PCE has increased, the measurement problems in GDP(E) may have become more severe, and more plainly visible. In addition, booms and busts in financial services may have accounted for a much larger share of the variability of the business cycle since the mid-1980s, with the junk bond boom and bust (as well as the savings and loan boom and bust) from the mid to late 1980s, the day-trading boom in the mid to late 1990s and subsequent stock market crash from 2000 to 2002, and the mortgage securitization boom and bust from 2002 to 2008. GDP(E) may have missed much of this variation.

If the difference between GDP and GDI is cyclical, then that means one of the measures is more sensitive to the cycle than the other. Which one? The answer is GDI, as seen by checking the correlation of each with other indicators that come from different data sources. Regressions of GDI growth and GDP growth on several indicators, over the period 1984Q3-2006Q4, shows a tighter correlation of GDI with each cyclical indicator:

Indicator Adj. R2 GDI Adj. R2 GDP
YOY change in unemployment 0.25 0.10
ISM PMI 0.33 0.19
ISM services 0.29 0.18
NBER recessions 0.29 0.24

That is quite strong evidence.

As an aside, when this paper came out, in Apr 2010, Nalewaik analyzed various sources of error in the data, and made the claim that the drop in GDP during the great recession was being underestimated. With the recent annual revision, we now know that this was correct.

This is not the first time I've run across an analysis suggesting GDI might be a better business cycle indicator. So why does everyone concentrate on GDP? For the very simple reason that it comes out earlier. Nalewaik suggests that at the very least, one should pay attention to both:

These results strongly suggest that economists and statisticians interested in business cycle fluctuations in U.S. output should pay attention to the income-side estimates, and consider using some sort of weighted average of the income- and expenditure-side estimates in their analyses.

I am convinced, and will start reporting on both when the BEA makes its regular estimates.