I like the data organized this way. For each want ad, four resumes were sent, two Whites and two Blacks.
|
Equal Treatment 87.37% |
No Call-back 82.56% |
1W+1B 3.46% |
2W+2B 1.35% |
|
Whites Favored 8.87% |
1W+0B 5.93% |
2W+0B 1.50% |
2W+1B 1.43% |
|
Blacks Favored 3.76% |
1B+0W 2.78% |
2B+0W 0.45% |
2B+1W 0.53% |
Another interesting breakdown is by occupation and industry.
| Occupation | % of Ads | White callback rate | Black callback rate | Ratio |
| Executive and Managerial | 14.5% | 7.91% | 5.95% | 1.33 |
| Administrative supervisors | 7.7% | 9.57% | 5.85% | 1.64 |
| Sales representatives | 15.2% | 8.04% | 5.09% | 1.58 |
| Sales workers, retail and personal services | 16.8% | 10.46% | 7.05% | 1.48 |
| Secretaries | 33.9% | 10.49% | 6.63% | 1.58 |
| Clerical workers, admin. support | 11.9% | 13.75% | 9.96% | 1.38 |
I would have expected similar results. At the highest level (executive and managerial) and the lowest (clerical workers, admin. support) discrimination is lowest.
| Industry | % of Ads | White callback rate | Black callback rate | Ratio |
| Manufacturing | 8.3% | 6.93% | 3.96% | 1.75 |
| Transportation and communication | 3.0% | 12.16% | 14.86% | 0.82 |
| Wholesale and retail trade | 21.5% | 8.76% | 5.71% | 1.53 |
| Finance, insurance and real estate | 8.5% | 10.63% | 4.35% | 2.44 |
| Business and personal services | 26.8% | 11.30% | 6.71% | 1.68 |
| Health, educational and social services | 15.5% | 12.14% | 9.50% | 1.28 |
| Other/unknown | 16.4% | 8.71% | 6.47% | 1.35 |
Let’s now consider what the authors have to say about the hypothesis that the observed differences can be due to perceived social class rather than race of the applicants:
Second, perhaps employers are inferring more than just race from applicants’ names. More specifically, maybe employers are inferring social class. When employers read a name like “Tyrone” or “Latoya,” they may associate that name with the ghetto or other disadvantaged social background. Of course, because African Americans on average do in fact come from poorer backgrounds than Whites, this argument would have to be sharpened. These names would need to be more reflective of economic background than being African American already is. While plausible, several of our results are inconsistent with this interpretation. First, recall that the African American sounding names we use are not as atypical as they may seem. In fact. as Appendix 1 shows, they are quite common among African Americans. Second, for the subset of African American female names where we had access to data on social background (mother’s education to be precise), we found no correlation between social background and callback rates. Finally, and perhaps most telling, in Table 7, we found that African Americans are not helped more than Whites by living in more White or more-educated neighborhoods. If the African American names were mostly to signal negative social background, one might have expected a better address to yield greater returns for the African American names than for White names.
The authors make a plausible case, but not an air-tight one in my opinion. I think a batch of neutral-sounding names, i.e. common names equally popular among whites and blacks, would have given a good comparison. Also, the authors did not have access to much social/economic data; only data about mothers’ high school education for a small subset is not enough. I would guess that there is a bigger jump in social/economic status if the mother is college-educated as opposed to the difference between “completed high school” vs “not completed high school.” (I do not have any data on this, but would appreciate if someone could point me to it.)
In the end, I have no doubt that there is racial discrimination in hiring, though definitely has gone down quite a lot over the years. The authors do make a better case than a lot of the previous studies, but their case is all about discrimination among first names. There is definitely correlation with race, but is it causation? I am not completely convinced.
NOTE: All the tables and quotes belong to Bertrand and Mullainathan and are from their paper “Are Emily and Brendan More Employable than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination.” All copyrights belong to the authors or to the publisher of their paper.
I have an idea about a possible control. In the next test, the resumes should all have “white” sounding names (as defined by the researcher) but should vary by neighborhood. The neighborhoods should be readily recognizable to potential employers as being (1) upper middle class white; (2) lower middle class white; (3) working class white; (4) middle class black; (5) working class black; and (6) black ghetto. If there are any white ghettos or upper middle class black neighborhoods (e.g. Laurelton in Queens) in the area, they should also be included.
This should allow the researchers to achieve at least a rough approximation of whether class trumps race or vice versa. For instance, the statistics might show that Alison from the Upper West Side does better than Alison from West Harlem, but that Alison from West Harlem does about as well as Alison from Bensonhurst.
Of course, a survey that varies by neighborhood probably won’t clear up the issue by itself. It will probably take a number of single-variable tests before the results of the name survey can be placed in context.
Jonathan, they do vary the neighborhood randomly in this study. Since the applicants are supposed to live in the same city as the employers, therefore, the quality of the neighborhood they live in should be obvious. What the researchers found was that all applicants were helped by living in a more affluent, more educated and whiter neighborhood. Black and white sounding names were helped equally. So, the difference between the callback rates remains between the races.