The Statistical Discrepancy: A Source of Phony Wage Growth

A New York Times article today commented on the extraordinary jump in wages over the last two quarters. Before anyone breaks out the champagne, take a look at the statistical discrepancy in the GDP accounts.

This might be is a bit nerdy, but there is an important story here. In principle, it is possible to add up GDP on either the income side (e.g. wages, interest, profits) or the output side (e.g. consumption, investment, government) and get the same number. Of course, they never end up exactly the same � you don�t get perfect accounting in a $13 trillion economy.

Typically, the output side comes up slightly higher than the income side. (The conventional wisdom is that people might hide income in order to avoid taxes.) This gap between output side GDP and income side GDP is the statistical discrepancy.

The big story on the statistical discrepancy is that it fell by $150.8 billion over the last two quarters. The most obvious explanation is that the income side of GDP is currently overstated. My favorite explanation for overstated income is that capital gains income (e.g. exercises of stock options) is showing up as labor compensation.

Suppose this is right. The $150.8 billion shift in the statistical discrepancy accounts for more than half of the nominal wage growth over the last two quarters. If this change in the statistical discrepancy is subtracted from wage growth, then the extraordinary 7 percent annual rate of real wage growth over the last two quarters falls to a modest 1 percent. Of course, it is unlikely that the full shift in the statistical discrepancy is the result of misattributed labor income, but such an error can explain much of this gain.

Lest anyone think this is a backdoor effort to deny the Bush administration credit for some good economic news, let me point out that the magic of shifting statistical discrepancies worked the other way earlier in his administration. Between 2000 and 2003, the statistical discrepancy increased by $176 billion (almost 3 percent of labor compensation). This increase implies a potential accounting loss of labor income of this magnitude, which did not correspond to an actual decline in labor income.

The moral of this story is that labor compensation is poorly measured in the national income accounts. While the accounts give us a general picture of movements in wages and profits, the data must be assessed with caution. When the data in the national accounts diverge sharply from other data on compensation (e.g. the average hourly wage series and the employment cost index), it is best to look at it more carefully for errors before assuming that we are getting new information on the economy.

--Dean Baker

You may also like