More on the Height of Bravery
By Arnold Kling
It should be real easy to determine whether the above sociological jibberish holds any merit: simply compare wealthy Americans (all of whom presumably have decent healthcare) to wealthy Europeans and poor Americans (who presumably have bad healthcare) to poor Europeans (who supposedly have the same health care as the wealthy). If the wealthy Americans and Europeans are the same height and the poor Americans are shorter than the poor Europeans, then we could maybe make a case for a difference in health care. If both rich and poor Europeans are taller than their American counterparts, then it’s tough to make a case for a healthcare difference causing the problem.
The authors do look at this. And what do they find?
The lagging U.S. height performance is not caused by a long left tail in the height distribution. Heights are normally distributed and the whole U.S. height distribution is shifted to the left. In other words, rich Americans are shorter than rich western Europeans and poor white Americans are shorter than poor western Europeans.
Case closed. Or as Samuel Johnson would have said, “There’s an end on’t.” It ain’t a difference in the healthcare system.
Eades goes on to propose increased consumption of soft drinks, either by itself or as a displacement for milk, as the culprit. That is plausible, but unproven speculation.
My point in bringing this up is that the original study commits a classic fallacy, which is to label the unexplained in a regression. When you use regression analysis, you include some variables and you exclude others. The latter are excluded because you cannot measure them or you do not know what they are. They are called the “omitted variables” or “error term” in regression terminology.
It is a fallacy to interpret the error term as representing exactly one omitted variable–health care in this case. Another classic example of this fallacy was committed by Christopher Jencks, who found that his measures of IQ and schooling together explained less than 50 percent of the variation in earnings, and decided that the remainder had to be “luck.” Again, he was labeling his ignorance, not making a valid inference.
What I find disturbing is that Paul Krugman devoted a column to trumpeting this research. I cannot imagine in a million years that he would have been as laudatory of an equally shoddy paper that reached a conclusion in support of American capitalism.
I, too, am guilty of wanting to give the benefit of the doubt to papers that support my viewpoint. But I would like to think that I approach all research with at least some methodological skepticism, rather than assume that if the results come out to my taste it must be good work. In any event, when I link to a paper, I strongly urge you to read it yourself and to think about possible flaws.