I'M waiting for someone to ask me what "sophistry" means. I'll pull out my copy of Operations Research magazine and say, "See for yourself!"
I'll be carrying the September-October, 1992, issue, and I'll point to an article called "America's Vietnam Casualties: Victims of a Class War?" The article was written by Arnold Barnett, a professor at the Massachusetts Institute of Technology's Sloan School of Management, and two recent graduates of the school, Timothy Stanley,and, Michael Shore. Operations Research is not a mass-circulation journal, and I am sure that most of the journalists who have written about this article never bothered to read what it actually said, as I'll explain in a moment. Still, the article is surprisingly important, for the impact it has already had on public discourse and for what it shows about the corruption of educated thought.
A brief digest of the article's conclusions was sent to news organizations last fall, and it generated a lot of coverage. The interest was natural: according to the summary, the article disproved one of the major planks of the conventional wisdom about the Vietnam War. "Everybody knows that Vietnam was a class war, whose burden was borne disproportionately by the inner-city and rural poor and minorities," Time magazine said in its story about the report. "Well, it seems that everybody is wrong." The Wall Street Journal ran a story with the headline "CLASS WALLS BREACHED BY DEATH IN VIETNAM." William F. Buckley Jr. wrote a column about the study, saying it proved that the war in Vietnam was an "all-American effort." Most other stories were similarly respectful of the MIT study and its startling claims.
If true, the MIT findings would obviously be important, both in changing the standard version of Vietnam War history and in raising questions about how we know what we "know." If Vietnam really was an "equal opportunity war," as Time said in its report on the study, how could so many people have believed the opposite for so long?
Someday we may have to ask that question. But not now. The MIT study is preposterous. It raises questions, all right, but they concern the academics who conduct such scholarship and the journalists who pass it on without checking the details.
Now the necessary disclaimer: I have one large bias, but not the one the MIT authors might suspect. Their study is presented largely as a rebuttal to an article I wrote eighteen years ago called "What Did You Do in the Class War, Daddy?" It appeared in The Washington Monthly, and it argued that because the sons of the nation's economic, professional, and political elite were generally spared the costs of the Vietnam War, the war went on longer than it otherwise would have. The MIT authors say that I was wrong: "In terms of the bereavement it brought to America, Vietnam was not a class war."
I don't mind the disagreement. My real bias is more primitive: I would prefer never to raise this subject in public again. For me it involves the Oprahlike spectacle of rehashing the way that I and people like me dodged serving in the war. The subject is also becoming the Baby Boom's version of the Rosenberg case: when aging cranks start haggling over the fine points of their old arguments, everyone else tries to get out of the room. But the reaction to Bill Clinton's and Dan Quayle's draft histories suggests that the inequities of service in Vietnam, perceived or real, still matter to many Americans, which is why the MIT study matters too.
HERE is how the study worked. Some 58,000 Americans died in Vietnam. From a list of the dead the MIT scholars made a random selection of 1,525 names, a sample easily large enough to achieve statistical significance. To determine whether any class bias was evident among the dead, the scholars decided to concentrate on income alone as an indicator of class. Money, after all, is a good rough measure of where people stand. But how could the scholars figure out the financial backgrounds of the casualties? These people died twenty to thirty years ago. America being what it is, their families have moved, dispersed, or died.
Facing huge obstacles, the MIT scholars came up with an approximation. From military records they determined each dead soldier's home town. From census data they determined the median income for each of those towns -- or at least any town with a population of 2,500 or more, the smallest unit for which the census reports median income. The scholars then assumed that each soldier's family income was the median income of his home town. If Bill Clinton had been drafted from Oxford or Yale Law School, then, he would have counted as a poor boy rather than as a member of the educated elite. From other census data the scholars determined the income distribution for all men of military age during the Vietnam War. Then, with all the data in place, they moved through an increasingly elaborate set of correlations and "disparity" calculations to find whether there was a sharp economic difference between the people who died and America as a whole. When all the computer runs were finished, the team discovered that the economic difference was surprisingly small.
What's wrong with this approach? The logical error is so grotesque that I'm almost embarrassed to point it out. The study assumes the very hypothesis that it is designed to test. That is, a study purporting to test whether casualties were representative rests on data that defines each casualty as representative. As a basic axiom of statistics, the variation between two large groups will almost always be smaller than the variation within either of the groups. The structure of the study limits the range of possible economic variation to the relatively small city-by-city differences across America, rather than to the much larger family-by-family inequalities within any city. The variations are limited more dramatically still by the assumption that each soldier who died came right from the middle of his home town's economic structure.
Let me make this specific: My home town, Redlands, California, had in those days a median income about 14 percent above the national average. The people I knew from Redlands who died in Vietnam included one Mexican-American who did not go to college, one white who went briefly to junior college, another who went into the army after high school, and a third who, anomalously for our town, went to an Eastern prep school and then to Harvard on an ROTC scholarship. Their stories illustrate both the chanciness of life -- the ROTC student was one of only twelve from Harvard College to die in Vietnam -- and certain larger sociological patterns. But for the purposes of the MIT study, these four people are all the same person, and each of them is 14 percent above the national average in income.
The MIT scholars would wave away this argument as "anecdotal," the kiss-of-death term social scientists use to dismiss any evidence that can't be reduced to mathematical models. The question is, What violates reality more grossly: my memory of one high school in one small town, which other people can test against their own experience and information, or the MIT model, which purports to be scientifically accurate while systematically misclassifying people like the ones I knew?