How Much Does It Cost You in Wages if You Sound Black?

Wednesday, July 9th, 2008

I don’t think this headline conveys the right idea, but it describes an interesting study. How Much Does It Cost You in Wages if You Sound Black?

As part of a large longitudinal study called the National Longitudinal Survey of Youth, follow-up validation interviews were conducted over the phone and recorded.

Grogger was able to take these phone interviews, purge them of any identifying information, and then ask people to try to identify the voices as to whether the speaker was black or white. The listeners were pretty good at distinguishing race through voices: 98 percent of the time they got the gender of the speaker right, 84 percent of white speakers were correctly identified as white, and 77 percent of black speakers were correctly identified as black.

Grogger asked multiple listeners to rate each voice 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.

(It turns out you don’t want to sound southern, either. Although pretty imprecisely estimated, it is almost as bad for your wages to sound southern as it is to sound black, even controlling for whether you live in the south.)

The headline implies a certain causality, but that’s not necessarily the case at all:

It is possible that there are many other characteristics that differ between blacks who do or do not “sound black” that Grogger cannot control for in his regressions. It does seem likely that the biases at work would make his estimate an upper bound.

Leave a Reply