Friday, March 12, 2010
Here's an interesting question: how big is the effect of the ARRA stimulus? You can't measure it by collecting data, because you have nothing to compare your data against. Would we have had less total output without it? More? The same? Did it shift the makeup of output in some way? There's no way to answer any of these questions without knowing what would have happened in the absence of stimulus -- something we cannot observe, because it didn't happen. The same issue holds for unemployment, for the time pattern of production, and other matters of interest.
The only way to get at such a question is to postulate a theory of how such spending influences economic activity. But even that isn't enough; a theory simply tells, at best, the directions of effects; it cannot identify magnitudes. Hence some sort of measurement of something related, perhaps based on numbers from micro data, is required to calibrate the model. But you cannot measure the effects of interest, you can only compare reality against against a hypothetical world in which the cause whose effects you are trying to guess did not occur. The whole exercise rests on theory. It involves what Frederic Bastiat called "that which is seen and that which is not seen." Your hope your theoretical underpinnings are relatively sound, so that you've made an educated guess rather than a WAG. But your estimate of the effects is theory laden, and won't systematically be any better than your theory.
As "we" here at Unforeseen Contingencies understand it, that's essentially the approach that is used in macroeconomics these days. And if you're going to ask such questions, there's probably no other way to get an answer. And this seems to be exactly what the CBO did.
As mentioned in the previous post, there's been a little brouhaha on the Cafe Hayek & Econbrowser blogs over the nature of the CBO results. Russ Roberts has now clarified his position, and argues that the effects of ARRA haven't been measured, might not be measurable, and the CBO has not produced estimates of these effects. It should be clear from the above that I disagree. The effects are impossible to measure, but can be estimated. The estimate will be theory laden, and in macro, there are almost as many theories as economists.
"We here at Unforeseen Contingencies have, of course, already developed our own theoretical analysis of stimulus and multipliers, which we commend to our readers who could use a brushup on the issues. For some more technical skepticism of multipliers, see Cwik, Cogan, Taylor, & Wieland.
Given all this, though, there's a sense in which this discussion isn't as important as it might seem. Can government interventions boost GDP? It's hard to imagine that there'd be no effect at all. As mentioned in the previous post (and many times elsewhere in the blog) the Austrian theory of the business cycle is built on the recognition that artificial stimulus can generate an unsustainable boom. For that matter, the Soviet economy was able to get high rates of employment. Hence even if stimulus is successful in pushing GDP back to potential, if the spending is on things that are ultimately unproductive or negative value added, it's for naught... something I managed to get Menzie Chinn essentially to concede back in February 2009. (Scroll down to find his response to my comment.)
Photo: Empirical proof that ARRA has stimulated the economy...evidence of a boom in the "under construction" sign building industry.