GDP Estimates: Rationality Tests and Turning Point Performance

The reliability of BEA’s estimates, as measured by the magnitude and pattern of revisions, is highly important to economic policy-making and business decisions. We find evidence that the revisions are partially predicable using contemporaneously available information for the current quarterly estimates of GDP. Information about national income is found to significantly supplement the information found in the final current quarterly estimates of GDP in explaining the revisions to the latest-available estimates of GDP. However, there is little evidence of the predictability of revisions in GDI or national income. Finally, both the advance and final current quarterly estimates are found to do a reliable job of measuring GDP and GDI around cyclical peaks, but a less reliable job around cyclical troughs, the declines preceding the troughs are overstated and the upturns after the troughs are understated. An earlier and somewhat expanded version of this paper, “Revisions, Rationality, and Turning Points in GDP,” was presented at the session “Tracking the Turning Points in the Economy,” AEA meetings January 3–5 2003, Washington DC. It is available in the “working papers” section of BEA’s web site, www.bea.gov