Thursday, October 09, 2003
Let me add a couple of things. (Full disclosure first: I have met Vigdor at a conference once and read another of his papers.) First, the root of Vigdor's argument is that there is customer discrimination towards black quarterbacks. Both network executives and sportswriters would be foolish not to be drawn by that preference to provide more TV coverage and more flattering articles. If nobody trusts buying a used car from a woman because they don't think she can look under the hood as well as a guy, the owner of the dealership rationally favors males in hiring used car salespersons. Limbaugh would not like this result because he thinks the majority of the public is conservative and color-blind. Maybe, maybe not.
Taranto also links to the article Vigdor co-authored on which his Prospect article is based ... only to find it gone but with a trace in the Yahoo cache. It is rather suspicious that it was up for a year and now gone. Moreover, as I read it, it doesn't seem to say exactly what is being said in the Prospect piece. Suppose that black TV viewership of Monday Night Football (this is the sample on which Vigdor was testing) is greater than white viewership. Suppose as well that members of each race prefer to see their own race playing on the field. Quarterback being a central position, it is possible to see "stacking" of own-race players in highly visible spots; hence, the greater share of black viewership of Monday Night Football. By Vigdor's own analysis, he can't tell for sure if it's stacking or preference for diversity. The paper tries to explain it away using other data and that the magnitude of the effect is too large for it to be stacking.
As I pointed out before, it's also possible that fans as well as sportswriters like to watch the effort of scrambling quarterbacks, who are disproportionately black. Vigdor's paper relied only on quarterback ratings to measure quarterback quality, so he did not control for the possibility that all viewers wanted to watch were QBs who run regardless of race. He and his co-authors rely (at pages 13-14) on the observations that (a) an index of racial tolerance from an unrelated survey is negatively related to time spent watching TV and (b) the audience for the ESPN games (where there's no evidence of a black QB effect) are people who watch a lot of TV and therefore either colorblind to QB race or worse, biased against black QBs. That's an awfully large stretch. He has no evidence from the Nielsen sample itself on racial attitudes of respondents.
This is not to say Vigdor is necessarily wrong, but to say that making strong policy statements to the press on the basis of a working paper is a risky venture. I don't know if, when he wrote this, he was trying to Rush-bash or advertise his own work as he builds his young career at Duke. The fact that he took the paper down after it was up a year makes it harder, though, to sustain the second hypothesis.