More adventures in the reliability of fixed traits
| Gabriel |
[Update: the finding is not robust. See Hannon and Defina]
Last year I described a memo by Gary Gates about some implications of coding errors for gender.
The new Social Problems has an article by Saperstein and Penner that’s comparably weird (and an impressive effort in secondary data analysis). They use data from NLSY on self-reported race over many waves and find that after a prison term an appreciable number of people who always or almost always had called themselves “white” will start calling themselves “black.” The finding seems to be a real effect rather than just random reliability problems as it’s often a sustained shift in identity over subsequent waves and the rate is strongly in one direction after the prison shock. You can’t really demonstrate validity for this kind of thing in a brief summary, but suffice it to say that if you read the article you get the very strong impression that the authors were both careful from the get go and faced demanding peer reviewers.
It’s obviously a weird finding but it makes some sense once you buy that a) identity is somewhat plastic and that b) the racial associations of crime and criminal justice could have an impact on this. One way to understand the results is that many people have multiple ancestry which they can invoke in kind of an ancestral toolkit. For instance Wentworth Miller ’95 (who given his most famous two performances could be seen as the personification of this article) might decide to emphasize his whiteness, Jewishness, blackness, or Arabness, whereas I am limited to the much narrower gamut of identifying as white or Jewish. Largely as a function of with whom I was associating at the time, I have in fact at times self-identified as mostly Irish or as mostly Jewish, but I’d always imagined that I lack the option of plausibly considering myself black. The really weird thing is that this intuition seems to be mistaken as the SP article finds that the effect is not limited to people who the authors could identify as being Hispanic or mixed-race but extends to many non-Hispanic whites with no known black ancestry.
One complication to keep in mind is that almost all American blacks have a lot of white (and to a lesser extent, American Indian) ancestry but we still tend to consider them “black” rather than “mixed race” unless they have a very high proportion of white ancestry and/or very white phenotype. That is the American identity=f(ancestry) is pretty bizarre and it’s hard to puzzle how that affects findings like in the SP piece. For this and other reasons I’d love to see this kind of study replicated in a society with racial schemas other than the “one-drop” tradition.