Economic interpretations of the statistical discrepancy

Through many comprehensive revisions of the National Income and Product Accounts(NIPA) of the US, a significant discrepancy has persisted, namely that between the estimate of the headline total now called Gross Domestic Product (GDP) from the side of expenditures or from the side of imcome payments. This discrepancy is not trivial (now in the neighborhood of $-\$100^+$ bn.); it is not random; it is wrongly attributed exclusively to the income side estimates. There have been interesting proposals for systematically allocating it among NIPA entries according to some statistical rule. Students of the discrepancy, over the years, have noted systematic variation with respect to international trade, inventory investment, total output and other variables. In recent years, the income side total has given a different estimate of the historical rate of change of output per worker, obviously an extremely important statistic. In a fresh examination of the discrepancy through 1996 (quarterly), we find, in this paper, suggestive correlation with business earnings, itself a very important but difficult magnitude to measure. If the NIPA data are to be used in an important way for policy guidance, a more careful treatment based on economic and statistical analysis is called for.