Race relation-omics?

The Forum's public opinion survey on race relations released late last year found race relations in SE Wisconsin are poor and not perceived as getting better. But we also found that in areas of interpersonal relations, such as marriage, adoption, friendship, etc., attitudes toward people of another race have improved across the board since we last surveyed on the issue in 1991. How can one survey find such disparate results? The answer may lie in economics.

A recent column by Tim Harford in Forbes magazine highlights the research economists have done on racism and concludes:
Economists tend to assume that people respond to the incentives they face. If that's true, we have to face up to the fact that young black Americans are facing some lousy choices. There is a lot of work for all of us to do.

Turns out economists are making a distinction between "taste-based" discrimination and "statistical" discrimination. The former is when discrimination occurs because of a dislike of minorities, the latter is when race is used as a marker for a trait to be selected out. Harford explains:
Non-racial statistical discrimination is actually rather common: An insurer will consider your age and your sex when deciding how much to charge for auto insurance. Why? Because the stereotypes, however crude and however unfair to individuals, contain a bit of extra information.

Harford's pessimistic conclusion arises from the realization that statistical racial discrimination by employers could be an economically reasonable position, if race is, in fact, a relatively reliable marker for something such as the quality of school attended. In that case an employer will respond to the positive incentive resulting from discrimination:
A racist who turns down workers even though he knows them to be competent will take a hit to the bottom line, but statistical discrimination could improve profits, which makes it harder to stamp out. As long as an employer can learn something extra from an applicant's race that he can't learn from looking at a résumé...then the worrying possibility for profitable discrimination exists.

So, the longer most minority children are attending the country's worst schools, the more reliable race is as a marker for quality of schooling. Even more troubling, however, is evidence that statistical racial discrimination based on an unreliable marker can cause the marker to become more accurate.
[Roland] Fryer and two colleagues, Jacob Goeree and Charles Holt, showed how statistical discrimination could easily lead to a vicious circle. They used computer-based classroom games that assigned students the role of employers, "purple" workers and "green" workers. Students in the role of employers quickly jumped to the mistaken conclusion that purple workers were uneducated, and that view became self-fulfilling, as purple workers abandoned hope of getting hired and stopped paying for education. Once the downward spiral set in, a color-blind employer would actually lose money.
"I was amazed," recalls Fryer. "The kids were really angry. The purple workers would say, 'I'm not investing [in an education] because you won't hire me', and the employers would respond 'I didn't hire you because you weren't investing.' The initial asymmetries came about because of chance, but people would hang onto them and wouldn't let them go."

Thus, if people feel that the instances in which they make distinctions based on race are not due to dislike of that racial group, but to a racially-relevant factor, we can find individuals reporting that their own attitudes about race have improved while race relations overall continue to be dismal. Because, if you're on the receiving end of the discrimination, does it really matter what label an economist has given it?