Yesterday the CBO released, as mandated by the American Recovery and Reinvestment Act (ARRA, the stimulus bill), an analysis commenting on the reports filed by recipients of stimulus funds. I have argued in the past that interpretation of such analyses is difficult, and that the Obama administration’s public interpretations have not demonstrated a corresponding level of exactitude. Now I see that the same left-of-center economists who dismissed my previous argument are presenting yesterday’s CBO report as confirmation of their beliefs that the stimulus saved or created about 1 million jobs and that the administration is free to advertise “created or saved” numbers as if they could point to each specific job.
Nevertheless, I take yesterday’s CBO report as affirmation both that 1) the number of jobs created by the stimulus cannot be determined without making judgment calls about the underlying economic model and that 2) the administration is on shaky ground when it claims to have evidence of hundreds of thousands of jobs created or saved.
In the second paragraph of yesterday’s report, the CBO agrees with me on the second count.
Recipients report that about 640,000 jobs were created or retained with ARRA funding through September 2009. Such reports, however, do not provide a comprehensive estimate of the law’s impact on employment in the United States. That impact may be higher or lower than the reported number for several reasons (in addition to any issues about the quality of the data in the reports). First, it is impossible to determine how many of the reported jobs would have existed in the absence of the stimulus package. [Emphasis added.]
Case closed: the nonpartisan CBO’s conclusions support my belief that the administration crosses the line between methodological soundness and propaganda whenever it states, in the context of touting reports filed by stimulus recipients, that it has created or saved large numbers of jobs.
As for the more general claim — 1) — that aside from the recipient reports, historical and model-based approaches lead us to conclude that the stimulus is successfully saving jobs: that is true. In this most recent report, the CBO tries to use both those approaches. But the CBO’s methods include some sweeping assumptions (which are typical in standard Keynesian models) about the economy, and anyone trumpeting the CBO’s findings should include qualifiers about those assumptions.
Here’s how the CBO comes up with the employment numbers. Instead of trying to count the reported jobs and extrapolate from those numbers, they look at the total amounts spent by the government so far, and then estimate the effects of those expenditures using evidence from similar policies in the past and sophisticated macroeconomic models.
One key assumption in the stimulus’s formulation is that these government expenditures have a multiplier effect — that they increase consumption and output both directly and indirectly. That means that each dollar the government spends or gives back as a tax cut not only raises the consumption of the worker employed using that dollar or the recipient of the tax rebate, but also raises the consumption of others as the recipients spend their extra money.
In the CBO’s report, there is no evidence from current macroeconomic conditions for what that assumed multiplier might be. Here is the relevant passage, which I will excerpt at length and then translate:
CBO’s current estimates do not reflect any change in the agency’s assessment of the effect that each dollar of spending increase or revenue decrease has on output and employment…. CBO has also examined incoming data on output and employment during the period since ARRA’s enactment. However, those data are not as helpful in determining ARRA’s economic effects as might be supposed, because isolating the effects would require knowing what path the economy would have taken in the absence of the law. Because that path cannot be observed, the new data add only limited information about ARRA’s impact. Economic output and employment in the spring and summer of 2009 were lower than CBO had projected at the beginning of the year. But in CBO’s judgment, that outcome reflects greater-than-projected weakness in the underlying economy rather than lower-than-expected effects of ARRA. [Emphasis added.]
So the CBO is still using the same multipliers that it used in its original forecast of the effects of the stimulus. In other words, data from after the stimulus was enacted has not, and could not have, informed the CBO’s estimates of the multipliers and consequently the stimulus’s effects on output or employment. Instead, the CBO made a judgment call that similar multipliers applied in the current situation. While it is reasonable for the CBO to make that judgment call (and in fact it seems like the right call to me), note that there are serious economists whose judgment would differ.
And also note that the selection of the multipliers, in turn, was a judgment call. Those multipliers were suggested by academic and private economists, mostly from similar Keynesian models. While this range of multipliers reflects the prevailing wisdom among economists, it is not settled science.
The CBO spells out very clearly the three sources they use for determining multipliers. The first source, macroeconomic forecasting models, is taken from two economic consulting firms and a model used at the Federal Reserve — basically standard Keynesian models. The second is general equilibrium models, which are models developed at “freshwater” economic schools like Chicago, and include important assumptions like rational expectations. The CBO seems to discount these slightly, explaining that “…the research results appear to be too dependent on particular assumptions for CBO to rely on them heavily.” The last source is extrapolations from direct study of historical data. Again, these extrapolations have widely varying estimates of multipliers depending on which studies are included. But we can only assume that the CBO chose a judicious representation of studies and ranges of multipliers.
The important thing is to note that the assumptions from these sources regarding multipliers and the effects of stimulus on employment and GDP have not changed since the CBO’s February forecast (pdf) of the effects of the stimulus. The range of predictions generated using these models was a dismal failure: the low estimate of employment in the fourth quarter of 2009 without the stimulus was 142.4 million, and the baseline was 141.6 million. The latest BLS report put employment well below either of these numbers, at 138.2 million.
If the models underlying these assumptions could yield an initial prediction that was so far off the mark, why should we unquestioningly use the exact same models for the key assumptions in re-creating the counterfactual baseline, as CBO has done?
I understand that CBO has come up with the best model they can. Obviously, it is not a pointless endeavor to try to provide some estimates of policy effects.
But those estimates should be viewed with appropriate skepticism, and the authors of the stimulus should not use them to make unwarranted claims about the brilliance of their handiwork for electoral purposes.
UPDATE: Edited for clarity.