Forty-nine percent of black men have been arrested at least once by age 23, and so have over 40 percent of all men.
Those are the big takeaways from a study in the April issue of Crime & Delinquency that has gained new relevance in light of the events unfolding in Ferguson, Missouri. University of South Carolina's Robert Brame, University of Albany's Shawn Bushway, University of Maryland - College Park's Ray Paternoster, and University of North Carolina - Charlotte's Michael Turner estimated "cumulative arrest prevalence" by race and gender using the National Longitudinal Survey of Youth (NLSY). They focused on the 1997 cohort, born between 1980 and 1984. The respondents were surveyed at ages 18 and 23, which would have been between 1998-2002 and 2003-2007, respectively.
The authors group together all non-black, non-Hispanic people, a category that includes white, mixed race, Asian-American, and American Indian people. Disturbingly, if not surprisingly, black men are significantly more likely to have been arrested at both age 18 and age 23 than non-black, non-Hispanic men. Hispanic men are somewhere in between, but the difference between them and both black and non-black, non-Hispanic men isn't statistically significant:
The results for women are more mixed. At age 18, there are barely any racial disparities in arrest prevalence at all, and at age 23 white women are somewhat more likely to have been arrested, though none of the differences are statistically significant. Keep in mind, when interpreting this graph, how different the Y axis is from the male graph. Regardless of race, women are less likely to have been arrested by age 18 or 23 than men:
One caveat to keep in mind is that a substantial number of people targeted for the survey did not reply, and some who responded at age 18 did not respond again at 23. The authors assume these people are "missing at random." That is, that "individuals with missing information in a particular race and sex cell have the same arrest rate as the individuals who provided information on their arrest experiences." The authors concede that this is a "strong and untestable assumption."
So they also analyzed the data as if either all the missing respondents in a given gender or racial category had been arrested, or none of them had been, to see how off, in theory, the numbers could be. No conclusions about racial or gender differences in arrests hold up in the most extreme cases of error. That doesn't necessarily mean the overall findings are wrong. It seems fairly unlikely that, say, every missing non-black, non-Hispanic respondent had been arrested but every missing black respondent hadn't been, such that the racial arrest gap disappears. But it's a good reminder of the limits of even the best survey research.
Hat-tip to David Wessel, Gary Fields, and John Emshwiller.