THE STASH DECEMBER 20, 2009
Last week I noted the evolving thinking on unemployment and the recovery--in particular, the growing number of analysts who think the job numbers might increase pretty quickly over the next several months. The most compelling argument I'd read came courtesy of a Deutsche Bank report arguing that companies overshot with their layoffs during the recession, so they'd have to hire more than the increase in GDP would normally justify during the recovery.
Alas, a few days later, the ultra-sharp economics team at Goldman Sachs came out with a report (not online) explicitly rebutting this view.
Deutsche Bank's optimism derived from an empirical regularity known as "Okun's law," which suggests that, for every one-point change in GDP (relative to potential), the unemployment rate changes by half a percentage point. The DB economists noted that unemployment had risen significantly faster than would normally be associated with the drop in GDP that we observed during the recession, so it should fall significantly faster as GDP increases.
The Goldman economists made three arguments in response:
1.) If you look at the Okun's Law relationship over the last 25 years, rather than all the way back to World War II (as most analyses do), the last two years (i.e., the recession) look pretty similar to the previous 23. So it's possible that the world simply changed a bit in the 1980s.
2.) If you look at the Okun relationship in terms of payrolls (i.e., number of people employed) rather than the unemployment rate (which can fluctuate for arbitrary reasons), the relationship look like it's been pretty stable during the recession--i.e., that there was no overshooting on the way down.
3.) Most compellingly to my eyes, to the extent the unemployment rate did overshoot a bit during the recession relative to the post-war experience, it actually overshot on the upside a bit for a couple years beforehand, so we may just be seeing an evening out of sorts. As the Goldman team puts it:
If the “overshooting” hypothesis has merit, then Okun’s Law should have predicted a smaller increase in unemployment during the Great Recession than actually occurred. At first blush, the residuals from our equation – the differences between the actual change in unemployment and its fitted value – lean slightly in this direction. ... More importantly, the [differences] for the latter part of the preceding expansion (2004-2007) lean systematically the other way. In other words, this year’s modestly larger-than-predicted run-up in the unemployment rate appears to correct for a lower-than-predicted unemployment rate during the latter half of the 2001-2007 expansion, leaving Okun’s Law intact over the period as a whole [emphasis added].
I haven't entirely sorted it out in my head, but I have a vague feeling that points 3 and points 1 and 2 contradict each other slightly. That is, either the relationship was stable for each year during the 2004-2009 period (as points 1 and 2 imply), or we overshot on the upside in 2004-2007, then on the downside in 2007-2009, leaving the relationship intact for the overall period (point 3). But I'm not sure both those things can be true. On the other hand, I'm not sure they can't be true either. Might depend on what you're measuring.
Whatever the case, the upshot is that Goldman is sticking with its prediction that this recovery, like its predecessors in 2001 and 1991, will tilt in the jobless direction. And they've been right about a lot of things so far this recession.