The media has widely reported the new figures released last Friday by the Bureau of Economic Analysis (BEA) of the U.S. Department of Commerce. Real Gross Domestic Product (GDP) declined at an annual rate of 1.0 percent in the second quarter (April through June) of 2009. (Since this is an annualized figure, real GDP actually fell by only about 0.2 percent during the quarter, and even that percentage is not the raw change but seasonally adjusted.) Each new report typically revises the estimates from the previous few quarters, so now the BEA tells us that during the first quarter of 2009, real GDP fell by an annual rate of 6.4 percent rather than the 6.1 percent first reported, or the 5.7 percent in the agency's preliminary revision, or the 5.5 percent in its first final revision.

But the current release does more than revise the recent estimates. Using new input-output analysis, the BEA has just completed a comprehensive revision of all the numbers in the National Income and Product Accounts going way back to 1929. Consider the slight difference this can make in the story of the current recession, which the National Bureau of Economic Research (NBER) dates as beginning in December 2007. The old estimates reported that real GDP fell by 0.2 percent in the fourth quarter of 2007, whereas the new estimates report that it rose by 2.1 percent. For the first quarter of 2008, the old estimate is a 0.9 percent rise; the new estimate is 0.7 percent fall. Second quarter of 2008: old, 2.8 percent rise; new, 1.5 percent rise. Third quarter of 2008: old, 0.5 percent fall; new, 2.7 percent fall. Fourth quarter of 2008: old, 6.3 percent fall; new, 5.4 percent fall.

This of course dramatically underscores the inherent imprecision of all these numbers. Admittedly, imprecise numbers are better than no numbers at all, so long as one remains aware of their limitations. But these frequent changes are not only confusing but an incredible nuisance for both historians and economists. With three sets of series (or as we will see, actually more) that researchers might choose from, the BEA's futile quest for a false precision can wreak havoc with the efforts of those wishing to compare the work of different scholars.

Nor is this the first time that national income figures have undergone comprehensive revision. The BEA now does comprehensive revisions every five years. Just a quick perusal of selected back issues of the annual Economic Report of the President shows how this can affect the numbers. Richard K. Vedder and Lowell E. Gallaway caught probably the most egregious instance in their neglected Out of Work: Unemployment and Government in Twentieth-Century America (1993). The 1960 estimates of real Gross National Product (this was when the preferred aggregate was still GNP rather than GDP), using a 1929 base year, reported that output rose by 30 percent during World War II (1941-1944) and then fell by a modest 5 percent to 1948. In contrast, the 1990 estimates, using 1982 as a base year, had output rising by 50 percent during World War II and then falling by a massive 30 percent from 1944 to 1948.

In other words, the 1990 estimates discovered a massive post-World War II recession in the U.S. that absolutely nobody was aware of at the time and that failed to show up at all in the unemployment numbers. Yet most macro texts over the last decade have graphs that mindlessly depict this imaginary depression by statistical artifact. The BEA made another big change in 1996 when it switched from using an implicit Passche price index (with forward moving quantity weights) to using a chain-weighted index for deflating nominal GDP to real GDP. Under the latest revisions the severity of the post-World War II"depression" has been softened somewhat to a 13 percent fall in GDP that ends in 1947.

Obviously most revisions are nowhere near as significant. Indeed, given the dizzying frequency of its trivial adjustments, clearly the BEA is a government agency with far too much money at its disposal. All would be better served, in my opinion, if the imprecision of the National Income and Product Accounts was more openly acknowledged and the frequency of the BEA revisions significantly curtailed. Or better yet, abolish the BEA and turn the whole enterprise back over the private NBER, where it resided before being nationalized during the Great Depression.