I finally got around to listening to Russ Roberts’s podcast with Steve Fazzari about the stimulus package. While it was encouraging to hear a civilized conversation on the matter, I have come away with a pessimistic view about any likelihood of forging a consensus on the role of fiscal stimulus.
This pessimism is further enhanced by examining the evidence presented in the blogosphere. For example, Paul Krugman recently argued that fiscal stimulus wasn’t even tried:
What’s extraordinary about all this is that stimulus can’t have failed, because it never happened. Once you take state and local cutbacks into account, there was no surge of government spending.
Meanwhile, John Taylor presents a graph that shows:
…that state and local governments did not increase their purchases of goods and services—including infrastructure—even though they received large grants in aid from the federal government. Instead they used the grants largely to reduce the amount of their borrowing as the following graph dramatically shows.
Each of them are looking at similar data and coming to vastly different conclusions. The reason for the differing conclusions is the result of the identification problem. For example, it is possible that borrowing by states declined as a result of the fact that the federal government increased grants to states. However, it is also possible that the grants to the states offset planned reductions in borrowing. In the former case, the stimulus was ineffective. In the latter case, the grants possibly prevented a worse outcome. An abstract look at the data, however, is not sufficient to draw a conclusion.
It would be a mistake, however, to equate my pessimism about a consensus with pessimism about the process of evaluating the stimulus. I predicted in February of 2009 that the stimulus would fail and I believe that I have been proven correct.
Given the identification problems involved, how can I possibly think that I am correct? First, it is important (and necessary) to judge the stimulus by a particular criteria. I, for example, choose to judge the stimulus by the criteria outlined by Christina Romer and Jared Bernstein. The reason that I think the stimulus should be judged based on this criteria is because it is based on an explicit model, it produces specific predictions about the effects of the stimulus, and it represents the views of the policymakers who passed it. With the criteria chosen, it is then possible to evaluate the effects of the stimulus.
I don’t think that I need to go into much detail to suggest that the stimulus failed with the chosen criteria. Romer and Bernstein predicted that without the stimulus the unemployment rate would peak around 9% in 2010. However, with the stimulus the unemployment rate was projected to peak around 8% in the third quarter of 2009. Casual observation demonstrates that this projection was false. Based on this criteria, the stimulus failed.
Critics of this type of evaluation make two charges: (1) they argue that the economy was worse than expected and thus the effectiveness of the stimulus was less than expected as well; and (2) that things would have been worse without the stimulus.
Was the economy worse than realized? After the fact we might say so, but what is our evidence that the economy was actually worse than we thought at the time? It is true that Bernstein-Romer’s forecast of the peak unemployment rate with the stimulus was lower than the actual unemployment rate, but is that prima facie evidence that they under-forecast what unemployment would have been in the absence of the stimulus? No. The identification problem rears its ugly head again. It might be true that the model simply under-forecast unemployment. However, it also might be true that the stimulus was ineffective and the model’s predictions about the effectiveness of the stimulus were wrong. What’s worse is that in either case the model is significantly flawed and therefore not particularly useful for policy analysis.
Others claim that things would have been worse without the stimulus. Whether true or not, this point is irrelevant. There is no predictable content in that statement. In addition, the fact that things could have been worse is not in and of itself a strong justification for stimulus. It is important to understand how much worse it could have been. For example, would it have been worth it to spend $800 billion to reduce unemployment by 0.1%? By 0.5%? In other words, it is important to consider the fact that while there are potential benefits to stimulus, there are also costs — both monetary and non-monetary. The idea that it could have been worse says nothing about whether the benefits exceed the costs.
Prior to the recession, I thought that there was an emerging consensus on the role of fiscal stimulus (one fellow blogger even referred to my original post on the stimulus as a short lesson in Ph.D. macro). Perhaps there is largely a consensus within academia that seems smaller due to the voices that propagate the blogosphere and policy circles. Or perhaps that consensus is only imagined. Regardless, it is clear that influential minds still believe in the power of fiscal stimulus and are not persuaded by the evidence presented above. Nonetheless, the evaluation of fiscal stimulus cannot be made in the abstract. The only reasonable means to evaluate stimulus is to do so in light of ex ante predictions about the effects of stimulus derived from an explicit framework. Based on this criteria, the stimulus has failed.