Share |

Outcome of the Stimulus and the Burden of Proof

e21 | December 2, 2010

At e21, we have observed that arguments in favor of fiscal stimulus are often predicated on mechanistic assumptions regarding the role of government spending in increasing employment levels. We have called for a more empirical approach that uses real-world data to assess the impact of the 2009 ARRA fiscal stimulus legislation.

On the tax cut portion of stimulus legislation, we have highlighted research by Claudia Sahm, Matthew Shapiro, and Joel Slemrod that drew on survey data suggesting that only 13% of households reported higher spending levels due to the one-time tax cuts in the fiscal stimulus. This casts doubt on the models used by the CEA and the CBO to assess the impact of ARRA, and on various private forecasting models that relied on the same set of assumptions.

A new study by Daniel Wilson at the San Francisco Fed calls into question the idea that the stimulus legislation as a whole — including the state transfers and direct spending portion — failed to generate the promised improvements in employment.

It is difficult to properly calculate the effects of the 2009 ARRA bill, as it was a nation-wide program. Though employment and growth failed to respond to ARRA as the Administration had suggested, fiscal stimulus advocates have argued that employment levels would have been lower still without the program.

Wilson’s study makes an important contribution to this debate by focusing on state-by-state comparisons. A large portion of stimulus funding at the state level was based on criteria that were entirely independent of the economic situation that states faced. For example, the number of existing highway miles was used to calculate additional transportation spending.

The study uses this resulting variation in state-level stimulus funding to determine what impact ARRA funding had on employment — including both the direct impact of workers hired to complete planned projects, as well as any broader spillover effects resulting from greater government spending. Administration economists have repeatedly emphasized the importance of this indirect employment growth in driving economic recovery.

The results suggest that though the program did result in 2 million jobs “created or saved” by March 2010, net job creation was statistically indistinguishable from zero by August of this year. Taken at face value, this would suggest that the stimulus program (with an overall cost of $814 billion) worked only to generate temporary jobs at a cost of over $400,000 per worker. Even if the stimulus had in fact generated this level of employment as a durable outcome, it would still have been an extremely expensive way to generate employment.

Interestingly, federal assistance to state Medicaid programs appears to have decreased local and state government employment. One possibility is that requirements to maintain full Medicaid benefits in order to receive federal aid proved sufficiently expensive that state governments pushed though additional rounds of layoffs in non-health related areas. This finding may suggest a potential pitfall with the Wyden-Brown proposal to decentralize health reform efforts at the state level: if comprehensive insurance requirements are retained, the net effect of reform may only shift safety-net spending towards healthcare and away from other urgent priorities such as education or welfare assistance.

The results of this one study should not be seen as definitive. As Wilson emphasizes, the results only apply to the variation caused by additional state-level spending. It is possible that the stimulus did generate a certain level of base employment growth to all states — or that the stimulus “crowded out” private investment on a nation-wide basis.

It is also difficult to determine the counterfactual employment growth that would have resulted in the absence of the fiscal stimulus law. To address this issue, Wilson includes other variables predictive of future employment growth. However, it is possible that employment would indeed have been worse in all states without a stimulus. It is also possible that employment would have been better than projected — for instance, if the Fed or Treasury had responded to higher unemployment through their own interventions.

Still, this result should be taken seriously, as it represents one of the few actual analyses of the stimulus program that does not rely on outdated multiplier estimates that assume their result.

Importantly, the results are also consistent with another recent analysis of government spending during Great Depression by economists Price Fishback and Valentina Kachanovskaya. During a period in which unemployment was extremely high and the costs of implementing a public works program were far lower than today, one might expect that fiscal stimulus might have proven more effective. Yet Fishback and Kachanovskaya find that a similar state-by-state analysis suggests that fiscal stimulus during the Great Depression failed to yield durable employment gains.

The burden of proof is now increasingly on the side of fiscal stimulus advocates. It is easy to point out possible flaws in each of the studies mentioned here — though the biases may end up either exaggerating or diminishing the estimates of the effects of the stimulus. But where is the evidence that the 2009 ARRA fiscal stimulus enhanced employment recovery in a cost-effective and sustainable manner?