Odds of Being Arrested by Sex, Race and Class

The somewhat surprising result of the previous post — that racial bias in police shootings is not due to crime rate differentials but rather due to class bias in police shootings together with interracial income disparities — is robust. A clear bias against African-Americans in fatal police shootings is evident after we control for population:

FatalAdjPop

The racial bias attenuates but does not vanish once we control for homicide.

FatalAdjPopHom

But the racial bias in fatal police shootings vanishes once we also control for median household income.

FatalAdjPopHomMedian

These results are based on the Negative Binomial model of police shooting fatalities fitted on state-race aggregates. But if the model captures the reality of American law enforcement, we should find the same class bias in other metrics. Here we make a preliminary investigation of the model-free odds of getting arrested. For this purpose, we interrogate data from the National Longitudinal Survey of Youth (NLSY97).

Sex differences in criminality are so pronounced that the odds ratio for men and women sets a high benchmark for other categories. Simply knowing the sex of an individual radically changes the odds of ever being arrested: conditional on being female, the odds are cut to a third (OR = 0.32, P < 0.001); equivalently, conditional on being male, the odds rise by a factor of three (OR = 3.12, P < 0.001).

SexOR

Table 1. Odds of Being Arrested by Sex.
odds ratio log OR std error z-score
Female 0.32 -1.137 0.048 -23.6
Male 3.12 1.137 0.048 23.6
Source: National Longitudinal Study of Youth. N=8,964 participants aged 12 to 16 as of December 31, 1996. Unadjusted odds of being arrested at any time in 1997-2018. Estimates in bold are significant at the 5 percent level.

Interracial odds ratios are somewhat tamer by comparison. The odds ratio (OR) for non-Hispanic whites is 0.76 (P < 0.001) so they face three-fourths the odds faced by everyone else. Black OR is 1.43 (P < 0.001) so that the odds of being arrested for blacks is almost half as much again as non-blacks. The odds ratio for Hispanics is 0.98, which is not significantly different from 1, P = 0.384. Same for individuals of mixed race: OR = 1.26, P = 0.155.

RaceOR

Table 2. Odds of Being Arrested by Race.
OR log OR std error z-score
White 0.76 -0.280 0.045 -6.2
Black 1.43 0.357 0.050 7.1
Hispanic 0.98 -0.016 0.055 -0.3
Mixed 1.26 0.232 0.228 1.0
Source: National Longitudinal Study of Youth. N=8,964 participants aged 12 to 16 as of December 31, 1996. Unadjusted odds of being arrested at any time in 1997-2018. Estimates in bold are significant at the 5 percent level.

How do these odds for blacks and whites compare with those associated with the class gradient? If we proxy class by educational attainment, we find a very strong class gradient in the odds of being arrested by the police. ClasORedu

Table 3. Odds of being arrested by Educational Attainment.
OR log OR std error z-score
College degree 0.32 -1.135 0.061 -18.5
High school diploma 1.16 0.145 0.047 3.1
No diploma 3.84 1.345 0.063 21.2
Source: National Longitudinal Study of Youth. N=8,964 participants aged 12 to 16 as of December 31, 1996. Unadjusted odds of being arrested at any time in 1997-2018. Estimates in bold are significant at the 5 percent level.

The odds ratio for college graduates — potential members of the professional middle-class — is very small: OR = 0.32, P < 0.001. This is astonishing: college graduates face the same odds of being arrested relative to non-graduates as women face relative to men! Those with a high school diploma but no college degree — our operational definition of the working-class — face higher odds: OR = 1.16, P = 0.001. Those without even a high school diploma — the true underclass — face dramatically higher odds. People without a high school diploma are four times as likely to be arrested as people with one: OR = 3.84, P < 0.001. Incredibly, these odds are even more skewed than those for the sexes.

Instead of using educational attainment as our proxy for class, if we use parental household income, the odds temper but not by all that much. We choose household income thresholds corresponding to the 25th and 75th percentile to divvy up the population into low, mid and high income households. Individuals from relatively high income families face lower odds of being arrested: OR = 0.69, P = 0.005. Individuals from families with middling income face higher odds: OR = 1.43, P < 0.001. Incidentally, these odds are comparable to the relative odds faces by whites and blacks. Finally, individuals from lower income families are twice as likely to be arrested as those not from lower income families: OR = 2.09, P < 0.001.

ClasORincome

Table 4. Odds of being arrested by Household Income.
OR log OR std error z-score
High 0.69 -0.372 0.145 -2.6
Mid 1.43 0.361 0.094 3.9
Low 2.09 0.739 0.126 5.9
Source: National Longitudinal Study of Youth. N=8,964 participants aged 12 to 16 as of December 31, 1996. Unadjusted odds of being arrested at any time in 1997-2018. Estimates in bold are significant at the 5 percent level.

We can see that educational attainment is a much stronger risk factor than household income. It may be suggested that educational attainment, at least as captured by years of education under your belt, reflects differences in attitudes and behavior rather than class. We can get around this problem by proxying class by mean parental educational attainment. If, as we have argued, class is passed on at your parents’ dinner table, then mean parental educational attainment can serve as a proxy for that inheritance. Educational attainment is predicted by mean parental attainment: r = 0.45, P <0.001.

ClasORPedu

Table 5. Odds of being arrested by Mean Parental Education Attainment.
OR log OR std error z-score
High 0.60 -0.515 0.052 -9.9
Mid 1.20 0.185 0.045 4.1
Low 1.45 0.370 0.057 6.4
Source: National Longitudinal Study of Youth. N=8,964 participants aged 12 to 16 as of December 31, 1996. Unadjusted odds of being arrested at any time in 1997-2018. Estimates in bold are significant at the 5 percent level.

We find a very significant class gradient. Kids with highly educated parents face dramatically lower odds of being arrested: OR = 0.60, P < 0.001. Individuals from working class families in the middling half of America, where educational attainment is more limited, face worse odds: OR = 1.20, P < 0.001. And those born to the least educated quarter of American parents face much worse odds still: OR = 1.45, P < 0.001. Again, these odds are very similar to those found for blacks and whites.

The bottom line from the analysis of the unadjusted numbers is that the class bias in arrests is much greater than the racial bias and almost as large as the sex bias. Moreover, educational attainment emerges as the most important risk factor for ever being arrested.

EduEverArrested

Obviously, there are clear interracial differences in educational attainment.

RageGrade

We should therefore ask whether interracial differences in odds of getting arrested survive once we control for educational attainment. They answer is clear — they do not. The entire disparity between blacks and whites vanishes once we control for our main proxy of class.

ArrestsControllingforEdu

On the other hand, sex differences survive even after we control for educational attainment.

ArrestsControllingforEduBySex

These results reinforce our earlier findings. The racial bias we find in fatal police shootings and arrests is entirely accounted for by the class bias in policing together with African-American overrepresentation in the working-class and the underclass. The heavy hand of the law falls disproportionately on the poor and the poorly-educated. Depending on your point of view, American law enforcement is a machine for class oppression or serves to keep the rabble in line. What it is not is biased against racialized minorities per se. Rather, it is incidentally so biased as a result of educational and economic deprivation in African-American communities.

While the message may not find much favor among antiracists, our findings do open up a very promising line of attack. For if race factors through class, as documented here, then the problem reduces to the class bias of American institutions. This opens up a path forward towards a resolution of the impasse of elite-mass relations: Prioritizing class over race is not only politically necessary to forge a progressive-working class alliance, but also physically sufficient in that helping the working-class as a whole is the surest route to helping black lives in America.

4 thoughts on “Odds of Being Arrested by Sex, Race and Class

  1. This may be true from a statistical point of view, but politically is problematic. If you want black people to vote, you have to address their racial concerns directly. “Educating” them on why class is more important than race (based on statistical analysis) will be experienced as condescending. The language of “race” and the language of “class” will appeal differently depending on what population you are attempting to connect with on a political/ social change level. Also, do you think a good argument can be made that addressing directly the problems of the black working class specifically will accrue benefit to all members of the working class, including whites? (to kind of flip your conclusion…)

    1. What this sort of statistical analysis is useful for is not persuading the working-class, black or white. It is also not quite ready for an antiracist crowd. Rather, the purpose of such an exercise is to leash our perceptions to reality more tightly. It is of obvious relevance to policy design. And it behooves us to pay attention to empirical reality when we engage in political discourse. The odds ratios and the patterns I have documented demand to be understood. Once we make sense of them, it might open up hitherto unsuspected domains of social possibility. At least that’s the hope in the realist business. And, yes, I imagine that ‘black working-class’ would probably play better than the unqualified working-class — at least for the woke wing of the Democrats. I can equally well-imagine that would it backfire in the force-field of class-partisan politics. Even if that cannot be avoided — who can convince antiracist activists of the merits of paying attention to the facts? — making the police a ‘folk devil’ goes too far. ‘Defund the police’ is opposed by two-thirds of Americans. It is a recipe for political disaster. As I said elsewhere: ‘It is particularly blatant in this case because by portraying policemen and police unions as the ‘folk devil’ in the professional middle-class discourse, one is directly attacking one of the few functioning institutions left in Small Town, USA. To be a soldier, a fireman, a policeman, or a miner may not mean much to woke college-educated kids, but these are some of the few ladders to a life of responsibility and social respect left for working-class kids. The portrayal of miners, soldiers, and policemen as ‘folk devils’ in elite media gripped by moral panics seems like a pretty good way to piss off a lot of people in Flyover Country.’

  2. Hi Policy Tensor,

    Thank you for following up the previous post with this excellent analysis. With regards to the current post, am I right to interpret that the arrests rate disparity between African Americans and Whites completely disappears controlling for educational attainment, without crime rates being controlled for? I’m trying to understand whether class (rather than just race) can account for any disparity in crime/homicide rate differences among whites and African-Americans. This seems to be alluded to, if the the arrest rate disparity vanishes once class alone is taken into account (unless I’m misinterpreting something, or committing some sort of logical fallacy here).

    1. Hi Luther:

      Yes, the black-white gap in the probability of arrest vanishes once we control for educational attainment by itself. We can think of arrests as a proxy of crime. So in that sense, it is true that the black-white gap in crime vanishes once we control for years of schooling completed. I’m now looking at drug offenses — we don’t have data for homicides in this survey (what are you going to ask? did you kill someone since our last interview? Haha). Let’s keep an open mind. But I won’t lie. It does seem likely that any black-white gap in the probability of using or peddling marijuana or hard drugs will vanish once we control for class.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s