At Cafe Hayek, Don Boudreaux highlights a new paper by Jonathan Rothwell challenging the findings of David Autor et al that trade with China is harming American workers. The abstract of the paper sounds interesting, but I want to focus on one point in particular (Emphasis mine):
At the community level, Autor, Dorn and Hanson (2013) find that local areas have experienced slower job and wage growth and higher unemployment because of import competition with China. Upon analyzing their data, I conclude that their results are biased by the weaker macroeconomic performance of 2000-2007 relative to the 1990s. When I analyze inter-local area economic changes — rather analyzing changes within and across areas — I fail to reject the null hypotheses that import competition has no effect on wage or employment growth, except within the manufacturing sector during the most recent period, or that it has no effect on many other outcomes, including labor force participation, intergenerational mobility, and mortality.
There’s an interesting lesson to be learned here, beyond just what Rothwell finds:
Findings can depend on how one slices the data. To wit, Autor et al find significant negative effects when the data is within or across areas and Rothwell finds significant positive effects when the data is inter-local area. We see the same in minimum wage (time series vs panel data, etc).
Any statistician can tell you that regression models can change depending on how you cut and categorize the data: different “n” can give different outcomes, different controls and dummies can give different signs, etc. We try for robustness, but it is still at the end of the day a model.
Of course, none of this is to disparage the work of Autor et al or Rothwell, or even econometrics in general (an important field, if used correctly). But we need to fully understand its limitations and our own assumptions, and be very careful before tossing out theory.
Gordon Tullock, in his 1967 paper in the Western Economic Journal, demonstrates exactly this. Tullock begins with a conversation regarding welfare costs from monopolies and tariffs, citing recent research that finds these welfare losses are pretty minimal. In fact, they’re so small that Tullock finds:
Judging from conversations with graduate students, a number of younger economists are in fact drawing the conclusion that tariffs and monopolies are not of much importance. This view is now beginning to appear in the literature.
Does this mean our theory about trade and tariffs are wrong? Does this mean tariffs can be helpful, or at least not substantially harmful? Does this mean microeconomists spend too much time focusing on tariffs at the expense of other topics? Or is it a measurement issue and the theory is fine? Tullock explores this issue and finds it is a measurement issue, not a theoretical issue. In other words, our tools not theory were incomplete. Tullock explains in the article the need to factor in lobbying costs which do not show up in the standard welfare analysis but are nonetheless substantial (read the article for yourself to see his argument. It’s short, 9 pages, and not technical at all).
Had Tullock not looked beyond the initial challenge to trade theory, had he (and other economists) just thrown off the theory based upon the small welfare losses, the world would be a far worse place. As it is, his (and Jim Buchanan’s) explorations eventually lead to the field of Public Choice and provided us with a cleaner understanding on the theory of trade, tariffs, monopolies, politics, and the costs associated therefrom.
The story of Gordon Tullock in the 1960’s is why anyone should be weary of claims that theory of any kind is “mistaken” or “proven wrong” by this or that study. We see this all the time with minimum wage. The good economist (or scientist) will ask the question, as Tullock (and Mundell) did back in the 60’s: Is the theory invalid, or our tools? It may be the theory is (such as with the case of geo-centrism) or our measurement tools are lacking. In fact, we see this with regards to minimum wage: measurable job losses may be minimal, but there are many other margins firms adjust along, not all of whom are measured. It would be mistaken to toss out the theory.
Economics is still a young science. I suspect, as has already happened, some of our theories will be tossed out as we gain more insight and knowledge. But we musn’t be too hasty in doing so (especially when there is political pressure to do so), lest we sacrifice knowledge for convenience and insight for what my professor Thomas Startmann calls “naive analysis.”