Is Our Stimulus Working?

Paul Krugman is a very smart economist, far smarter than I am. So when I do not understand this post, I assume that I must be missing something.

Bruce Bartlett has written that the Obama administration underestimated how quickly the stimulus would affect the economy, reducing unemployment almost immediately. Krugman calls this "totally false":

Did Bartlett even look at the Bernstein-Romer paper? Here's the key graph:

The predicted impact from the stimulus is indicated by the difference between these two curves. We're now at the very beginning of 2009Q3; they predicted that the unemployment rate right now would be only a fraction of a percent lower now than it would otherwise be. The impact wasn't supposed to be really noticeable until late this year, and wasn't supposed to peak until late 2010.

The problem, in other words, is not that the stimulus is working more slowly than expected; it was never expected to do very much this soon. The problem, instead, is that the hole the stimulus needs to fill is much bigger than predicted. That -- coupled with the fact that yes, stimulus takes time to work -- is the reason for a second round, ASAP.

But when I look at the graph, it looks to me as if the stimulus was supposed to affect the unemployment rate immediately.  Specifically, it was supposed to dramatically lower the rate of increase in unemployment immediately.  By now, at the beginning of Q3, unemployment was supposed to start falling.  But unemployment has continued to rise apace.  It definitely isn't falling.

Also, I don't understand what he's trying to get at when he says that unemployment was only supposed to fall by a fraction of 1% by now.  In my experience, "a fraction of one percent" is usually used to refer to a small fraction of one percent, i.e. a trivial number.  As "fractions of 1%" approach 1%, it gets increasingly hard to distinguish them from 1%, because of measurement error inherent in collecting economic data.

But this fraction is certainly not trivial.  Eyeballing it, it seems to be in the neighborhood of .7%.  Since the maximimum reduction promised for the stimulus looks to be about 1.8% by Q3 2010, this means that we already should have seen more than a third of the employment benefit from the Obama plan.  By their own estimates, the stimulus seems to have failed badly.  I see little sign at this point that it's even moderating; rather, the seasonal adjustments seem to be improving because college graduates, who are being frozen out of the job market it record numbers, don't show up in job losses and unemployment claims.

It seems to me that a better defense against the Newsweek graph I linked is simply that models like this don't work very well, so expecting any sort of precise number from them is silly.  The problem is, those are the same models used to justify the stimulus in the first place. There's no question that in theory, ceteris paribus, a stimulus helps the economy return to full employment.  But as with tax cuts and other macroeconomic medicine, the size of the effect is an open question, one that has to be modeled against an extremely complicated and unpredictible economy.