Does Bad Data Explain the Unusual Recovery?

From last week’s Economist (subscription required):

DESPITE the welcome leap in American employment in March (see article), America’s job market has been surprisingly weak in the past couple of years—surprising, at least, to economists. Some have explained this by pointing to rapidly rising productivity figures. Perhaps firms have not needed more workers. But there is another explanation: America’s GDP figures, which have been strong, may be inaccurate, and may be exaggerating the extent of economic growth.

…Why might official statisticians be overstating America’s GDP—and productivity with it? Mr Hatzius [an economist at Goldman Sachs] suggests that they may be undercounting imports of intermediate inputs of goods and services produced abroad by American firms that have outsourced jobs to cheaper countries. Since GDP is calculated as domestic spending plus exports less imports (including imports of intermediate inputs), this would lead to an overstatement of GDP.

The picture below illustrates one piece of evidence to support this hypothesis.

Typically industrial production and GDP figures match each other pretty well. But since 2001 there has been a growing discrepancy.

The article does not address the obvious question, however: if there has been a sudden decrease in the quality of data on imports of intermediate goods and services (and I haven’t seen any direct evidence to support this theory, yet), then why did this just start happening 3 years ago? And why are the errors all in one direction, so that they understate US imports (rather than overstate imports or, as we would typically expect, have errors in both directions cancel each other out)? I don’t know the answers to these questions, though I think we can probably rule out one possibility right away: if it’s a tax avoidance story one would expect US firms to overstate imports, to make their US profits look lower so that they owe fewer US taxes.

Regardless, if it turns out that this theory is indeed true, at least this sure would make more sense out of some unusual features of this recovery.