(via Shakesville) Steven Levitt reports in Monday’s Freakonomics blog about an urban policy researcher, Jeffrey Grogger, at UChicago who did some analysis of the correlation between wages and speech, more particularly the quality of “sounding black”. Levitt:
His main finding: blacks who “sound black” earn salaries that are 10 percent lower than blacks who do not “sound black,” even after controlling for measures of intelligence, experience in the work force, and other factors that influence how much people earn. (For what it is worth, whites who “sound black” earn 6 percent lower than other whites.)…
Grogger asked multiple listeners to rate each [sample voice collected from telephone surveys] and assigned the voice either to a distinctly white or black category (if the listeners all tended to agree on the race), or an indistinct category if there was disagreement.
Then he put this measure of whether a voice sounded black into a regression (the standard statistical tool that economists use for estimating things), and came up with the finding that blacks who “sound black” earn almost 10 percent less, even after taking into account other factors that could influence earnings. One piece of interesting good news is that blacks who do not “sound black” earn essentially the same as whites.
This is interesting, but not really surprising; it’s reminiscent of Purnell, Idsardi, and Baugh (1999) about housing discrimination. I looked at the actual report by Grogger, and I confess that he’s using some statistical wage modeling techniques that lost me at the word “signal,” so any more ambitious readers should read that and let me know if there are any red (or other-color) flags. There are some basic issues with the method in terms of generalizability - he was using speech samples from adolescents (age 12-16) with little work history, and his listeners who were identifying the speech were disproportionately white and female - but from what I can tell he basically followed a categorization task that’s become fairly common in perception research (like that used by Cynthia Clopper and David Pisoni), and he cites lots of relevant sociolinguistic literature.
Anyway, what I wanted to point out is not really related to the implications of Grogger’s findings, but rather to the way they are presented by Levitt. Several commenters on the NYT blog have pointed out that Levitt seems to be couching these findings within a preference for linguistic assimilation rather than an appreciation of linguistic difference, and this strikes me as particularly problematic. For instance:
If one believes Grogger’s effects are causal, then investing in the ability to not “sound black” looks to have a huge return — roughly of the same magnitude as getting one more year of schooling.
Yeah, except that since these are really probably discrepancies based on race as it’s perceived through language, inequalities will probably persist in some other way unless the underlying discriminatory attitudes are addressed. He also writes:
Of course, there is the issue of one’s identity. There may be personal costs associated with being black and not sounding black. But these costs would have to be pretty large. (When I have Asian Ph.D. students go on the job market in the United States, I tell them that I think there is rampant discrimination against non-English speakers and encourage them to adopt Americanized first names for the job market. Very few of my students choose to do so — either a testimony to the identity cost of pretending to be someone you aren’t, or possibly their lack of faith in my assessment of the amount of discrimination.)
The “identity” costs of “not sounding black” would have to be “pretty large” in order to make people refuse to “give up” their “blackness” for the sake of higher wages? That’s just a bizarre thing to say. Also, everyone should adopt American names? Way to band-aid the problem!
The issue is this: if what you are measuring is the relationship between “how black someone sounds” and X variable, you are really measuring the relationship between someone’s assessment of blackness and X variable. It is not in the end about sounding black; it is about being black, and sound is one way that gets perceived. The problem is that racial discrimination is being made on the basis of language, which is to say that the problem is racial discrimination, not language use. This is all pretty obvious, I guess, but when you talk about issues like this in terms of what the discriminated-against group can/should do to stop being discriminated against (i.e., people who ’sound black’ should work on ’sounding white’ and Asians should get names that make them sound American), you are doing absolutely nothing to fix the discrimination. You are in fact providing a frame in which the discrimination can persist and be attributed to someone’s failure to [assimilate/work on it/change/recognize that the financial costs outweigh the identity costs] rather than systemic power differentials.