Thursday, December 14, 2006

Malcolm Gladwell gets a course in statistics   posted by p-ter @ 12/14/2006 09:08:00 PM

There's been a somewhat ill-tempered back-and-forth between Steve Sailer and Malcolm Gladwell on the subject of race and racism in the last few days (you can start in the middle with this post and move forwards and backwards from there to get the whole story). The crux of the debate is this story:
In Blink, I tell the story of a study done by the law professor Ian Ayres. Ayres put togother of group of young men and women--half white and half black--and sent them to 242 car dealerships all around Chicago. All were attractive, well dressed, and well-educated. All had the same cover story: that they were professionals from a wealthy part of Chicago. All pointed to the lowest-priced car on the floor and said--"I'm interested in buying this car." Ayres's question was--all other things being equal, how does skin color and gender affect the initial price quoted by a car salesman? His results: white men, on average, got quoted a price $725 above invoice, white women got quoted a price $935 above invoice, black women $1195 above invoice, and black men $1687 above invoice.

Is this discrimination (in the pejorative sense of the word) or is this price discrimination (a morally neutral action: charging each group what it's willing to pay)? Gladwell claims the former, Sailer the latter. The debate is at the links above, so go read it yourself. What I want to focus on is this comment from Gladwell:
Let's go back to the study. The male and female, black and white testers who Ayres sent out to car dealerships all gave the salesmen the same set of facts. They were all roughly the same age (late twenties). They all drove the same kind of car into the lot. They all dressed neatly and conservatively. They identified themselves as college-educated professionals (sample job: systems analyst at a bank). And they said they lived in the upper-income Chicago neighborhood of Streeterville. The car salesman, then, has several pieces of data from which to create his stereotype. He has the gender, race, age, occupation, educational level, and class (or at least a class proxy) of his potential customer. And what did he do? With the black men, he zeroed in on age and race, and ignored everything else.
Why [are the car dealers] so intent on zeroing in on what is only one of many available and relevant facts about the customer? The short answer to that question, I think, is that this is what racial prejudice is: it is the irrational elevation of race-based considerations over other, equally or more relevant factors.
It should be obvious that the imputation of racial prejudice here is absolutely absurd (as was immediately pointed out in the comments). Here's why:

Let's take a look at all the factors Gladwell claims are available for the car salesman to consider: gender, race, age, occupation, educational level, and class. Now let's consider the sample he has to work with (limiting ourselves to black males, as Gladwell does: "with the black men..."). Sex: all male. Age: all late 20s. Occupation: all professionals. Educational level: all college educated. Class: all well dressed and wealthy.

In fact, the only factor that varies in this analysis is race (where he gets the thing about age, I'm not sure). This isn't an accident; the study was explicitly designed that way. So is it at all surprising that race is the only factor that influences anything? Of course not! In fact, when you include women in the analysis, you find that gender plays a role as well (and you see a possibly interesting differential role of gender in the two races). If other variables were included in the sample, they too would likely have been found to play a role in dealer pricing. But they were held constant*. That's why they call it a variable, my friend--it has to vary.

*It occurrs to me that I worded this very poorly originally, though I suppose everyone rearranged the sentence for me unconsciously.