Does GDI Point to a Stronger Recovery?
Gross Domestic Product (GDP) and Gross Domestic Income (GDI) both measure the same economic variable - the aggregate production of goods and services within the US in a year. GDP computes it as the sum of all expenditures (consumption, investment, government spending, and net exports), while GDI computes it as the sum of all incomes (employee compensation, profits, interest, rent, income from unincorporated businesses, indirect taxes minus subsidies, depreciation). These two measures may, sometimes, diverge because of measurement errors. In the current recovery, in particular, GDI has been growing faster than GDP. Between the second quarter of 2009 and the second quarter of 2013, GDI grew at a 2.65 annual rate, while GDP grew only at a 2.23 annual rate, quite a large difference for a four-year-long period. What is the actual rate at which the economy is growing?
Both GDP and GDI have measurement strengths and weaknesses, but if we compare the source data used to compute the two measures, we are led to put more trust in GDP than GDI as an indicator of aggregate output. The source data used to compute GDP is generally better because it is mainly based on business surveys collected for statistical purposes, and it uses a consistent set of concepts and definitions. In contrast, the source data used to compute GDI is produced for a variety of other purposes, since it is mainly based on financial statements and information from tax and regulatory agencies, and it uses heterogeneous concepts and definitions. GDP source data is also timelier - a much larger fraction of source data is available for the early GDP estimates than for the early GDI estimates, so judgment and trend adjustments play a much smaller role in the early GDP estimates (See Landefeld 2010 for a thorough discussion of the topic in this paragraph).
Some evidence, however, favors GDI over GDP as a measure of aggregate output. Aruoba, Diebold, Nalewaik, Schorfheide, and Song (2013) estimate the unknown path of aggregate output solely based on the known paths of GDP and GDI, and find that GDI is, overall, a more accurate measure than GDP. For instance, if we use their estimation method (and the second model in their paper), we find that aggregate output grew at a 2.51 annual rate between the second quarter of 2009 and the second quarter of 2013, closer to the GDI growth rate than to the GDP growth rate.
On balance, the evidence suggests that the growth rate of aggregate output in the recovery has been in between the growth rates of GDP and GDI, up to 0.25 percentage points faster than indicated by GDP. The implications of this upward revision, however, are rather limited.
The overall picture of the recovery is not much changed. Even if we focus on the estimate of aggregate output in the second chart above - which is obtained with a method that favors GDI over GDP - the recovery still lacks an initial strong rebound after the Great Recession, and its pace continues to be moderate. The current level of output is still well below the forecasts that were made back in 2007, before the beginning of the crisis (see Behind the Slowdown of Potential GDP, Jacobson and Occhino, February 2013).
Higher estimated output growth leads to only slightly higher statistical estimates of trend output growth. For instance, if we compute trend GDP using a band-pass filter that eliminates all cycles shorter than 30 years from GDP data, we find that trend output is currently growing at a 2.21 annual rate. If we apply the same method to the estimates of aggregate output obtained using the estimation method of Aruoba, Diebold, Nalewaik, Schorfheide, and Song, we find that trend output is currently growing at a 2.28 percent annual rate, just a few basis points higher.
Faster estimated growth is only slightly more consistent with the improvement that the labor market has experienced in the recovery. While typically output grows fast when the unemployment rate declines, in this recovery it has grown slowly, even though the unemployment rate has declined steadily. Higher estimated output growth, then, is slightly more in line with past business cycles. However, the size of the revision is small relative to the overall decline of the unemployment rate. An upward revision to output of 0.25 percentage points over a four-year period - for a cumulative 1 percentage point - roughly corresponds to a decline in the unemployment rate of only 0.25 to 0.50 percentage points, based on common estimates of Okun's Law. This is rather small compared with the overall decline of the unemployment rate over the same period - almost 2 percentage points.
- B. Aruoba, F. X. Diebold, J. Nalewaik, F. Schorfheide and D. Song, "Improving GDP Measurement: A Measurement-Error Perspective", unpublished manuscript, April 2013.
- J. Steven Landefeld, Spring 2010, comment to Jeremy J. Nalewaik, "The Income- and Expenditure-Side Estimates of U.S. Output Growth", Brookings Papers on Economic Activity, pages 112-123.