David R. Henderson  

Does Econometrics Resolve Issues?

PRINT
Part-Time Jobs and Climbing Tr... A Controversial Issue Resolved...

Bryan Caplan's statement, "In social science, the best arguments prove more than the best studies," reminds me of a conversation I was in earlier this week. Economist Jeff Hummel said he couldn't think of even one controversial issue that had been resolved with econometrics. The other 4 economists present, including me, immediately started trying to think of counterexamples. The first one that came to my mind was Milton Friedman's consumption function. Jeff agreed that this had resolved an issue but pointed out that Friedman did it simply with data, not with econometrics. The other examples that the other economists came up with were similar: data had resolved the issue but it didn't require econometrics.

Question: Can you think of an issue that has been resolved with econometrics?

Ground Rule: It can't be simply that the person who did the econometrics convinced himself or herself. It has also to be that a number of people previously on the other side of the issue changed their minds in response to the econometrics.


Comments and Sharing


CATEGORIES: Economic Methods



COMMENTS (14 to date)
Doug writes:

Do you count finance as part of economics?

If so then Fama-French factors is a gold-standard examples of econometrics making an indisputable contribution to the field of finance. There are literally dozens of other examples where econometric applied to financial data proved new points.

nerdbound writes:

I should start by saying that even though I've been trained in econometrics, I'm absolutely stumped by the distinction you're drawing between learning from data and learning from econometrics. Maybe it's like: if a t distribution is used, that's just 'statistics', but if instrumental variables are used, that's 'econometrics'? If we draw a narrow enough circle around what counts as econometrics, then you're vacuously correct...

But more importantly, I don't think you're accurately describing the role of econometrics. The major goal of econometrics is not to drag people who disagree kicking and screaming into agreement, but to inform open-minded folks of approximate effect sizes. Like, I wouldn't a priori know how much Medicaid would improve health outcomes for poor people, but an informed reading of the econometric literature suggests it's a small positive. I wouldn't know a priori how large an effect minimum wages would have on employment, but an informed reading suggests that it's at most a small negative effect. Etc.

John Voorheis writes:

Isn't "controversial" doing all the heavy lifting in that sentence? And might the causality run the other way? That issues are only controversial where the econometrics is unclear (plagued by bad/incomplete data, perhaps)? Note this includes essentially all of macro.

Jim Rose writes:

econometrics showed that econometrics cannot provide sturdy inferences.

stephen writes:

To be fair, and this is beside the point to some extent, but econometrics, particularly 'low level' econometrics, are pretty useful in many business settings. At least in my profession. As far as the Big Questions? Probably not.

Peter St. Onge writes:

Not that we needed the metrics for it, but empirical data did resolve the socialist calculation debate.

John Palmer writes:

I don't know if it had much impact on others, but I used to be adamantly pro gun control, but reading John Lott's studies, and the followups by others, has moved me into the opposite camp.

I also recall having read (or possibly seen in a seminar, but damned if I can remember) a study using longitudinal data and showing that high school students attending larger high schools had slightly (but significantly) higher incomes than students who attended smaller high schools, ceteris paribus.

Eric Falkenstein writes:

Chris Sims demonstrated that large-scale Keynesian macro models were not better than far simpler VARs. Combined with the Lucas critique, this really diminished the popularity of macro vs. other kinds off economics. Keynesian models live on in politics, but it's a rather dead field intellectually, compared to all the hope pre-1970 that such models would really guide, steer, and predict macro policy.

Ken B writes:

Like some others I do not know where DRH draws the line but assuming it means something like "you need to do a lot of regressions to try to eliminate other generally accepted confounding factors and so end up with a conclusion dependent on a lot of modeling assumptions" than I suggest the deterrent effect of the death penalty as na example.

Bob Lince writes:

This may not concern econometrics, but I recall, in reading on the Monty Hall Problem, the story of a highly credentialed mathematician/economist who didn't believe there was any advantage in trading for the other curtain, until he ran a computer simulation, and only then saw the light.

SWaterton writes:

In the field of addiction, one could argue that econometrics supports the rational addiction model.

Devil's Advocate writes:

How about this one..."Minimum wage contributes to teen unemployment." Econometrics informs the discussion and helps solve the problem of understanding the causes of unemployment?

http://online.wsj.com/article/SB10001424052970203440104574402820278669840.html

Garett Jones writes:

First, I nominate the hypothesis that money surprises matter for output only as long as the surprise itself.

That hypothesis was refuted by Barro himself in a pair of econometrically straightforward articles. He found money surprises matter for too long for the Lucas-style money surprise channel to be the whole story.

So his work implicitly strengthened the monetarist/New Keynesian story that unexpected money changes have some influence on real output that lasts for a year or so, after which the effect of money changes show up just in prices.

The link (I'm writing from Munich, hope it works) finds a relevant quote from Blanchard and Fischer's Lectures on Macroeconomics; "nails in the coffin" is the key quote, page 360. Barro JPE 78 and Barro AER 77: together they (and the later VAR money shock literature) killed a promising blackboard model.

Second, I nominate conditional convergence models of economic growth. This hypothesis has been accepted partly because it's intuitive, to be sure, but it dominates the scene also because whenever you run a growth regression that controls for starting GDP per capita and even one or two reasonable controls (like education levels, lifespan, degree of capitalism), you're almost guaranteed to get a negative, statistically significant coefficient on starting GDP per capita.

So countries that start poor (conditioned on one or two proxies of long-run potential) grow faster. The poor grow rich faster: the opposite of common sense.

This repeated finding--the negative beta on starting GDP per capita--helped to kill a certain class of nation-level endogenous growth theories, which had claimed that countries that started off rich would grow faster forever.

And there's even a pseudo-constant: Countries (and US states!) converge to their long run potential at a rate of 2% per year. Some cites:

http://scholar.google.com/scholar?q=convergence+growth+"two+percent"

One can always disagree on what it means for an issue to be settled: there are phlogistoners sticking around in every field. And I've raised some issues with the conditional convergence result for the poorest countries in my paper "Cognitive Skill and Technology Diffusion". But that negative beta makes it extremely hard to claim that rich countries, qua rich countries, routinely grow faster.

David R. Henderson writes:

Thank you to Doug, nerdbound, John Palmer, Eric Falkenstein, Ken B, and Garett Jones for counterexamples. To SWaterton, I’m not up on the addiction literature.
I’ll have some good examples to run by Jeff Hummel when I next see him.

Comments for this entry have been closed
Return to top