The NBA has the best record in professional sports in hiring black coaches. But the New York Times got that record wrong late last month in a 2,200-word front-page story.
The Times’ study of the past 15 years found evidence of discrimination against black coaches. A white coach, the paper claimed, typically gets to coach “50 percent longer [than a black coach] and has most of an extra season to prove himself.” Discrimination is abhorrent, and the Times’ evidence of discrimination caused real, understandable anger in sports pages across the nation.
Yet there are serious problems with the evidence. For starters, the Times authors, David Leonhardt and Ford Fessenden, selectively threw out data (such as not looking at current coaches). In fact, in one difficult-to-understand sentence buried deep in the article, they admit that if these data are included, “the gap between white and black coaches was nearly identical.” Yet they offered no explanation for why they threw that data out.
And even with selective use of data, the differences they found are not statistically significant — in other words, their “evidence” doesn’t show anything.
As one expert that the Times consulted, Harvard Professor Larry Katz, noted to us, the best research on this topic (by Larry Kahn at Cornell University) “finds no significant race differences” between black and white NBA coaches using “almost any reasonable set of controls.” Yet the Times made no mention of any opposing evidence.
The Times refuses to share the data used in its published stories, making it more difficult for others to analyze and criticize the newspaper’s statistical study in a timely fashion.
The Times reported that from 1990 to 2004, black coaches lasted an average of 1.6 seasons, while white coaches kept their jobs for an average of 2.4 seasons. The study excluded coaches who had served more than 10 years, coaches who had coached any time prior to 1990 and active coaches. Using data that we gathered independently with the same rules outlined by the Times, we found that black coaches lasted for 1.7 seasons compared to 2.57 seasons for white coaches. More than a third of the coaches were black, with all but two teams having employed at least one black coach.
But any empirical work explaining a coach’s length of service should account for things like a coach’s win rate and experience, among other factors.
In particular, a proper study must also note general changes in coaching durations over time. Coaching tenures are getting ever shorter even as black coaches have become more common; ignoring these trends would lead to a spurious correlation between race and coaching length of service.
While we looked at a data set similar to what the Times said it selected, it seems most appropriate to concentrate on all the coaches serving after 1990, even those still coaching. It turns out that — contrary to the Times’ study — white and black coaches have similar lengths of service.
Winning substantially helps job security (each one percentage-point increase in the win rate increases a coach’s job duration by at least 8 percent) — and it seems to help white and black coaches equally. But for some reason, if all else is equal, NBA teams win more often with a white coach — teams with white coaches had an 8 percentage point higher win rates. (That is, a .08 difference in the three-digit “winning percentage” seen on most sports pages.) The Times article focuses on anecdotal evidence supporting its contention that black coaches face discrimination, but there are at least as many examples to demonstrate that black coaches in the NBA are treated fairly.
For example, Doc Rivers lasted over 4 seasons in Orlando even though his win rate (.504) was less than the win rate for the three previous white coaches who averaged just 1.9 seasons of service.
Even Paul Silas, the black coach that the Times notes was treated poorly in Cleveland, spent 4.4 years with the Charlotte/New Orleans Hornets, as compared to Dave Cowens’ 2.1 years — even though his .573 win rate was lower than Cowens’ .609 win rate.
Mr. Leonhardt suggests that concerns about issues such as selectively dropping data and statistical significance aren’t really relevant because the Times is “not an academic journal, it is a newspaper.” Yet the fact remains that the Times — a newspaper with millions of readers — alleges discrimination on the basis of evidence that simply doesn’t hold up. That’s the equivalent of yelling “fire” in a crowded movie theater.
Stephen G. Bronars is the Regents professor of economics at the University of Texas. John R. Lott Jr. is a resident scholar at the American Enterprise Institute.