by Mario Rizzo
I am both intrigued and annoyed by Robert Barro’s recent opinion piece in the Wall Street Journal. He adduces empirical (econometric) evidence to support the view that the fiscal stimulus package has done very little good in the short run and will do harm in the long run. I do not want here to discuss technical data or identification problems. Instead, I want to discuss something both more elementary and more profound.
The world of econometric estimation is a world shrouded in mystery, even for some econometricians like Edward Leamer of UCLA, for example.
In my opinion this is because this attempt to make economic theory have contact with data often requires adjustments both in the theory and the application that make their relationship to real-world people (thinking as most cognitively limited minds are likely to do) very unclear. All sorts of equations are solved, knowledge is had or gained at rapid speeds, modes of reasoning unknown to mere humanity are postulated, and so forth. This is so much the stock and trade of modern macro theory that most economists don’t give a second thought to it.
Barro’s analysis, by far not the worse example of this manner of doing economics, nevetheless is presented with very little concern for a good story about how his econometric results can be rationalized in terms of human action.
In a letter I recently emailed to the Wall Street Journal, I said:
Robert Barro’s empirical work on the effects of the stimulus package is to be applauded because it is a systematic, not casual, analysis. Stories about people hired do not constitute science. However, I am disturbed that he does not try to present a plausible story or mechanism about why he gets his results. Such stories are valuable checks on empirical work. We gain additional insight if the story seems plausible or not.
Does Barro get his short-term results because of some traditional crowding out effect? For that, interest rates would have to be higher today than otherwise. Is he willing to argue that? Or is it because the stimulus spending draws on resources that are not currently unemployed so output elsewhere is contracting? Or is it due to the “Barro-effect” that the current deficit actually increases saving (as people save to pay future taxes)? This would imply an offset in aggregate spending.
If we knew the postulated mechanism we would have additional scope to test the reasonableness of his results. Right now, I can only say, as the late Nobel laureate George Stigler used to say: “I want to see the econometric results that were thrown in the waste basket.”
Of course, I am not accusing Barro of any impropriety. But in the absence of a “good story” it is hard to know what to make of the results. How much did he and his research assistants have to “massage” the data to come up with what they did? That could be instructive on its own.
In point of fact — regarding how economics proceeds as a discipline — if a rationale for empirical results is not offered, or is not convincing, the results will not be believed, at least not by anyone not predisposed to believe them in the first place.