Professor Brown points out this Washington Post article, “In Boardrooms and in Courtrooms, Diversity Makes a Difference” and comments that it may be relevant to much of the seminar, especially to the “Social Comparison of Abilities at an Elite College: Feeling Outclassed with 1350 SATs,” by Matthew B. Kugler and George R. Goethals, our assigned reading for Monday.

When the Rev. Martin Luther King Jr. called on America to open the doors of opportunity to people of color, the civil rights leader was making a moral argument.

Cedric Herring recently decided to take things one step further. Given that discussions about morality are often divisive, the sociologist decided to take a more scientific approach. In other words, beyond the question of whether diversity is a good thing, is there evidence that it makes a difference?

Herring has just completed his study. He found that companies that are more diverse have more customers, a larger share of their markets and greater profitability. In fact, when Herring puts his numbers on a graph, he finds a linear relationship between diversity and business success, meaning that as diversity increases, those business indicators increase in step.

“Those companies that have very low levels of racial and ethnic minorities have the lowest profits and the lowest market share and the lowest number of customers,” he said. “Those that have medium levels do better, and those that have the highest levels do the best.”

Herring got his results by obtaining data about diversity levels and business performance from about 250 companies. He verified the information with independent statistics from Dun & Bradstreet Corp. and documents filed with the federal government. The 250 companies are representative of all U.S. businesses with more than 10 employees — from the restaurant down the street that employs a dozen people to multinational corporations with thousands of workers. Herring found the same relationship between diversity and business success whether a company was large or small.

Not to be too cynical, but that seems like a highly suspect result to me. I can’t find a copy of the working paper on Herring’s website, so perhaps these concerns are answered therein.

1) How were these 250 companies selected? Of course, the answer had better involve some sort of random process otherwise selection bias could easily tilt the reults one way or the other.

2) How are these 250 companies “representative” of all US business with more than 10 employees? If they really are a random sample from this universe, then the vast majority of them would be small. (There are, obviously, many more companies with 15 employees than with 15,000.) I doubt that the sample was random.

3) What does it mean that Herring “verified” the data? Small companies do not have to make much data public and, as a rule, therefore don’t. You think Dunn & Bradstreet has a file on every company in the US with more than 10 employees? Hah! There is probably a federal file on every such company since they all need to get a Federal Employer Identification Number, but I do not think that large amounts of such data are publicly available.

4) I am probably being too hard on Herring. It isn’t his fault that the Post writer can’t accurately describe the study. But I have seen too much junky social science research to simply assume that everything is well-done, especially since there is no working paper. I will e-mail Professor Herring and invite him to comment.

5) Even if Herring somehow got a more or less random sample of 250 firms, I can’t imagine how he got the rest of this data. Assume that one of the firms was, say, a landscaper in Des Moines. How did Herring find out what the racial composition of this company, either its workforce as a whole or just the leadership? I do not think that such data is available anywhere.

6) My guess is that the writer has not accurately described the sample. Perhaps Herring just looked at firms in the Chicago area and gathered relevant statistics by hand. Perhaps he just looked at large, publicly traded companies. One could, by studying websites, come up with a sense of the racial diversity of the leadership of such firms.

7) But even in this best case scenario, I still have my doubts. How does Herring measure, for example, market share? In the real world, this is very hard to do. What is the total market for, say, landscaping in Des Moines? Does the market include just Des Moines or also the surrounding suburbs? How about the surrounding counties? Each different measure of the total market size will lead to a different measure of market share for the landscaper in our sample. To come up with useful data on this, for 250 firms in different industries, is hugely difficult. Now, Dunn & Bradstreet might have some sorts of market share measures for large companies, but even then, the strength of the results, as described in the article, leave me suspicious. The other example cited in the article, work by a Tufts psychology professor, seems equally suspect.

Herring, who works at the University of Illinois at Chicago, says he agrees with advocates who have long argued that workers with different backgrounds make companies more responsive to customers. This model suggests that racial diversity is a marker for diverse ideas, attitudes and life experiences, and that having a range of perspectives can alert a company to threats and opportunities.

New research from psychology, however, suggests that this information model might only partially explain diversity’s impact. Something more subtle — and intriguing — also seems to happen when people of color join groups that were formerly all white: The entire group starts to think in new ways. Minorities, in other words, not only bring new perspectives to the table but also seem to catalyze new thinking among others.

Give me a break. Let’s all just gather round the camp fire after a fun afternoon of trust-building exercises and sing Kumbaya! Although it is a pleasing and not-implausible hypothesis that racially diverse groups do better than racially homogeneous ones (all else equal), there is no good evidence for this that I have seen. Nor is there any evidence that such groups do worse, although folks like Steve Sailer argue otherwise.

Again, perhaps I have been too harsh here. Those with conflicting views should speak up!

UPDATE: A copy of the paper is available here. Although it appears that the sample is more random than I initially thought, the whole exercise is fairly sloppy. Although the source used by Herring does seem to provide a reasonable 1,000 company sample, there is no reason to believe that the 250 companies he focuses are a random sample of those 1,000. Nor would I put much faith in, for example, the market share data, which seems to be self-reported. I suspect that the entire project is skewed by the fact that very small firms are, obviously, less diverse than big firms. There is no way to stay 100% white if you have 1,000 employees. Diversity doesn’t cause bigness. Bigness causes diversity. Want more details? Let me know!