Hmmm. Just read OpinionJournal – Best of the Web Today, who comments on the NY Times story on race and hiring I comment on below:
Employers were more likely to ask the “applicants” with “white names” in for an interview than those with the “black names.” But something’s wrong here. A chart that accompanies the print version of the Times story but doesn’t appear online shows the frequency with which people with “white” names and “black” names got called for interviews:
“White” names
Kristen: 13.6%
Carrie: 13.1%
Laurie: 10.8%
Meredith: 10.6%
Sarah: 9.8%
Allison: 9.4%
Jill: 9.3%
Anne: 9.0%
Emily: 8.3%
“Black” names
Ebony: 10.5%
Latonya: 9.1%
Kenya: 9.1%
Latoya: 8.8%
Tanisha: 6.3%
Lakisha: 5.5%
Tamika: 5.4%
Keisha: 3.8%
Aisha: 2.2%
Now, what’s “white” about names like Laurie and Jill? Wouldn’t a fair comparison have included some odd-sounding white names, like Dweezil or Moon Unit? And if employers discriminate against people with “black” names, how come Latonyas and Latoyas were more likely to get called back than Emilys were?
Uh-oh…better go to the actual statistics before I go making claims about this…
[Update: did a quick run of the numbers through Excel, and got a statistically significant difference between the two sets, so there is something going on. I’ve got a friend who’s trying to get both the ‘shooting’ research below and this as original papers.]
The NYT article had some info on methodology. I was unable to find the original paper on the Web.
The fact that Laurie and Jill could be either black or white is immaterial, I think. Basically, if you don’t know the person’s race then you have a 10% chance of getting called.
However, the black names are clearly black. What’s more, the sample size was reasonably large and the difference they found was quite significant: 10% vs. 7% That’s a lot.
There might be another explanation, but I think the obvious one takes precedence unless someone comes up with something better. Plus, it jibes with a lot of other research in this same area.
And there’s one other thing in its favor: the WSJ doesn’t like it. Their editorial page is the most routinely dishonest writing I’ve ever seen, and it’s almost a certain bet that anything they disapprove of is actually true.
>> There might be another explanation, but I think the obvious one takes precedence unless someone comes up with something better.
Yes, people with goofy names might be more trouble.
There are at least two things that distinguish the unpopular names. On what basis do we that one of them is responsible for their unpopularity?
All the talk about statistical significance is missing a very important point–that there is more spread within the two groups than there is between the groups. So there’s something more than just racism going on here, but just what isn’t at all clear.
In any case, Excel isn’t the place to be turning here–a proper statistics textbook is. It isn’t enough just to declare that 3 percentage points is “a lot.” The analysis needs to be quite a bit more subtle than that, given that the numbers are so small and the spread within groups is so dramatic.
(If it were 100% vs. 70%, it would be much clearer, but the intra-group spread would still be a mystery).
Seems to be a bit of racism in the assumption about what are “black” names and what are “white” names. The dark skinned Martha, Mona and the two Cheryls I know might take exception… can they be black and not have a “black” name? How about Caroline, Pam, Peggy and Stephanie, what are they…..white you say? …NO, ASIAN! See how silly? Obviously there an assumption that the names on that list weren’t interviewed because of the name. Remember assume: ASS U Me? Not ME, I’m not ready to make that leap, how about U? Back to lurk mode.
The argument that there are no “white names” is immaterial, as Kevin wisely points out below.
* * * * *
That said, since I’m working on validating a statistical software package and armed with tons of statistical texts, I copied the data from the article into the software and was prepared to lecture Gore/Lieberman-style about the equality of variance in the samples and how even if you assume equal variance or not, the samples prove that there is a definitely a difference, even if it might be smaller even than the incorrect 3%.
Then I realized I was just making the different mistakes with a tool that required a higher licensing fee.
Truth be told, quantitative data is given higher authority, but numbers can always be spun the same as any other story.
That’s why in cases like this I’d just assume go the qualitative route. Do I know of situations where people have been affected negatively during a hiring process because of their race? Yes. Then there’s still a problem. End of study.
Pollyanna…probably. But better that than needing numbers to accept that racial inequality still exists. It’s those people who don’t live in a real world.
You need to look at the methodology used by the hiring companies. Maybe, the resumes are being read first by a screening computer. Maybe the programming logic is to ignore applicants with too similar resumes? Maybe it picks the first one it comes across that meets certain criteria and ignores subsequent, which means the order in which resumes were submitted are a problem. I would say in the raw numbers there is statistical significance, but peel away the onion and it could be a different story.
Lastly, I might add the study does not verify who is actually hired. What if after all the interviews a significant number of black applicants were actually hired?
Mike, the question is not whether any racism exists in the world, that’s an easy question and the answer is well know. The question is the degree and abundance of racist attitudes in the world and in this case the degree to which having a name which is perceived as “black” will result in unfair prejudice in attempting to get a job interview relative to having a name which is perceived as “white”.
First, so far as I can tell (and I haven’t read the entire paper yet) the researches simply created their own set of “white” and “black” names, and did little testing on the assumption of the “blackness” or “whiteness” of the names. Second, the spread between the individual names is very large, and that makes it difficult to make a conclusive statement that there is serious, persistent racism with respect to resume call-backs. For example, apparently it’s better to have a “black” name like Latoya, Kenya, Ebony, Leroy, or Jermaine, than it is to have a “white” name like Emily, Anne, Neil, Geoffrey, or Brett. And it’s seemingly much better to be named Jermaine rather than Rasheed or Brad rather than Neil than it is to be named Greg rather than Hakim.
I don’t think the researches have proved their main point at all with this data, especially considering the lack of research on the perceoptions of the names themselves independent of resume call-back potential. Considering how critical the precise selection of the names is to the conclusions of the study I think the fact that the names were not picked randomly is fatal to the conclusion of the report. For example, if they had chosen a list of names such as: Emily, Anne, Jill, Allison, Sarah (or Meridith, or Laurie), Neil, Brendan, Matthew, and Todd (or Brett) for the “white” names and Latoya, Latonya, Kenya, Ebony, Jamal, Hakim, Leroy, and Jermaine for the “black” names then the “white” / “black” call back difference completely evaporates. That seems like a huge flaw considering that they did in fact choose the names rather than select them randomly. It would have been better if they had created a large pool of names, done a study (in the same cities) to determine the perception of the degree of “blackness” or “whiteness” for the the names, chosen a random sampling of the names for the resume study and then determined the correlation (and certainty of correlation) between the “blackness” or “whiteness” of a name and its chances of resume call back. With the data from this study there’s not much more that you can say about it other than “maybe it shows racial prejudice, or maybe it’s just statistical weirdness, the error bars are far too wide to draw firm conclusions.”
Additionally, the average call back % for the top 4 names from each of the “black” and “white” lists vs. the bottom 4+4 names (“black” + “white”), i.e. lists balanced in “blackness” and “whiteness” of names is greater than the difference between the “black” and “white” lists alone, for both males and females. It would be interesting to see the results of the relative resume call back rates between random lists of names and how that correlates with the “black” / “white” % of each list.
BTW – Those aren’t Black names, they’re Mormon names.
http://www.wesclark.com/ubn/