Looking at the data for the different names from the previous post, a few things come to mind.
There is a great variation even within a category. What is the reason for that? Do people like Kristen so much better than Emily? Ebony over Aisha? What’s wrong with Neil? What is Brad doing right (other than sharing his name with Brad DeLong)? Can the authors explain it? Is the difference between blacks and whites due to choosing these specific subset of all names instead of race? What if the researchers had chosen all white names with results like Neil and all black ones like Jermaine? Is that even possible? Here is what the authors say:
Not surprisingly, we find variation in callback rates across names. Chance along would produce such variation because of the rather small number of observations in each cell. We therefore formally test the hypothesis that the names within each sex-race category produce the same effect. We estimate a probit regression of the call back dummy on all the personal first names, allowing for clustering of the observations at the employment ad level. For all but one sex-race category, we cannot reject the hypothesis that all the first name effects are the same. Only for African American female names do we reject this null at a significant level. Five out of nine female African American names (Aisha, Keisha, Tamika, Lakisha and Tanisha) do worse than the worst female White name. The last four female names (Latoya, Kenya, Latonya and Ebony) perform only slightly below the average White female name.
We investigated two possible explanations for these name specific effects among African American females. First, we considered the possibility that employers might be relatively less familiar with some of the worst performing names and it is this lack of familiarity that motivates their callback behavior. However, we found no obvious correlation between name-specific call back rates and the relative frequency of each of the names, at least in the Massachusetts birth certificates. Second, we considered the possibility that the name fixed effects reflect differences in social class or economic background. To assess the relevance of this interpretation, we used some limited Massachusetts birth certificates data on mothers’ education. More specifically, for each of the five most common African American female names in our sample (Aisha, Ebony, Keisha, Tamika and Tanisha), we were able to obtain information on the fraction of mothers having completed high school. We found no obvious relationship between the name-specific callback rates and mothers’ education. (Footnote: Female names by ascending mother education are: Tamika, Keisha, Tanisha, Ebony and Aisha.)
A number of black names in the study (Aisha, Rasheed, Kareem, Jamal, Hakim) seem like Arab in origin. Did that affect the results in any way? We see that Aisha, Rasheed and Kareem do badly but Jamal and Hakim don’t.
Bertrand and Mullainathan selected these names based on three factors: the popularity of these names among babies born in Massachusetts between 1974 and 1979; a high ratio of frequency in one race to the frequency in the other race; and as a result of a public survey which asked people to guess race and other features just based on a first name. If I was doing this study, I would include popular names that are equally popular among blacks and whites as well. That would give me something to measure my results against. Also, I would like to see the results of the public survey. Did they ask about education, income, intelligence, etc.? For example, was there a correlation between the call-back rates and any of the features they asked on the public survey?
Even though last names used were also race-specific, they do not analyze the effects of last names. “Because sample sizes are not large enough to separately consider each first and last name combination, we focus on first names.”
I should also mention Bertrand and Mullainathan found that the observed differences in callback rates for blacks and whites are statistically significant.
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.