The Congressional Budget Office says the $862 billion stimulus package already has “increased the number of people employed by between 1.2 million and 2.8 million,” among other things.
The CBO’s director, Douglas Elmendorf, explained this back in March, but it’s worth revisiting now. When the CBO says the stimulus did this or that, it presents as “evidence” the very same economic models that were used to predict the effects of the stimulus before it was even passed. Put another way, here is all the CBO is saying:
In other words, the models haven’t changed, so we will assume that the stimulus worked as intended.
The CBO is not measuring outputs. The feds’ accountability website for the stimulus, Recovery.gov, attempts to do that and finds fewer than 700,000 jobs created, as reported by fund recipients themselves (for almost $400 billion spent so far). Remember, too, that some of these numbers are considered sketchy, to say the least.
As the authors of the latest CBO report point out, measuring outputs is difficult to do because, well, the stimulus might be stimulating people to create jobs without their realizing it was the stimulus that stimulated them. The reported figure of fewer than 700,000, they observe, could be significantly lower than or significantly higher than what’s happened in the real world.
So, they try to estimate outputs (jobs created) by measuring inputs (money spent). If the models had it right before, then they probably still have it right. If they didn’t, then as Elmendorf acknowledged in March, the CBO wouldn’t know it.
It comes down, then, to whether you trust the models. It would be interesting to know how similar the models that CBO uses are to the ones that White House economics adviser Christina Romer used in creating this now-infamous graph in support of the stimulus:
If you want to hear Elmendorf explain this himself, click here and fast-forward to the 38:24 mark to hear the relevant exchange (it lasts about 2.5 minutes). But, being the swell guy that I am, I’ve listened to it for you and transcribed it below (with emphases added):
QUESTIONER: I’ll be brief. You mentioned that with the ARRA [American Recovery and Reinvestment Act, or stimulus bill] as it was moving through Congress, you provided projections based on a set of multipliers and econometric models — and then later, when evaluating the effects, you also used multipliers and econometric models. And that gives the impression of assuming what one’s trying to prove in terms of measuring the effects. And so as a two-part question, first, is that treatment required by the rules Congress sets for CBO? And second, how would that be different if you compared the initial projections, both baseline and with the stimulus bill, to the actual experience?
ELMENDORF: So our method of analysis of that is not required by the Congress. We try to be very clear in our reports on ARRA that we don’t think one can learn much by watching the evolution of particular components of GDP over the last few quarters about the effects of the stimulus. We think the best evidence about the effects of past policies comes from more detailed studies done often several years later about the behavior of particular households that got tax rebates sooner or later or what have you. So we don’t think you can learn much from that, and therefore we fall back on repeating the sort of analysis we did before. And we’ve tried to be very explicit about that, that is essentially repeating the same exercise we did rather than as an independent check on it. The part that’s a check is that we watch how the money’s flowed out of the government budget; we can update that. We’re reading new evidence, and if we thought we saw evidence that substantially shifted the body of work in this area, then we would shift our views. We haven’t seen that at this point.
QUESTIONER: If the stimulus bill did not do what it was originally forecast to do, that would not have been detected, by the subsequent analysis? Is that correct?
ELMENDORF: That’s right. That’s right. And, in terms of what we would have found otherwise, I mean, I don’t remember each of our forecasts. Certainly, by last March, our economic forecast took on board a very large decline in employment, a run-up in the unemployment rate, weak GDP growth with a recovery in the second half of last year. Our January forecast, last January’s forecast, did not have that. We marked down the forecast considerably from January to March. Our first estimates of the effects of the stimulus package I think were coming out n between those benchmarks. So it’s hard for me to go back and disentangle those pieces entirely.
What we’re left with in the way of evidence, then, is more like faith-based accountability. Do you believe?
On a related topic, some Harvard professors — doing the kind of after-the-fact, real-world analysis Elmendorf mentioned favorably above — found that powerful members of Congress may bring home the bacon to their districts, but it doesn’t lead to a boom among private business and can even lead to lower private-sector employment.