Liberty Matters

Systemic Racism in the American Criminal Justice System – a Serious Problem That Does Not Exist

“Systemic racism,” the Wise Men say, is everywhere.
 Academics like the best-selling author Ibram Kendi have argued for years that essentially all large gaps in performance that we see between racial and regional groups are due to some form of bigotry “within systems,” however subtle and hidden. By this point, this claim has been made across contexts that range from group income differences to test scoring patterns to trends in hiring, as discussed by such authors as Charles Murray in his Facing Reality. Dr. Kendi, in fact, would likely argue that anyone who disagrees with him re this point is a genetic racist. Per one of his most famous interviews, and many others from him and other scholars, the only possible explanations for human performance differences like those we see across almost every arena are “racism” and “inferiority.”
The arena of crime, and of citizen/police interactions more generally, may be the ultimate example of an American domain within which claims of subtle, systematized racism are constantly made. The core argument of the 2012-2021 Black Lives Matter Movement, of course, was that Black citizens are killed very disproportionately – and almost at random – by White policemen and vigilantes. Activist Cherno Biko once claimed on Fox News in prime-time that an innocent, presumably unarmed, Black man is “murdered” by American police about once daily. Benjamin Crump, Trayvon Martin’s family counsel and one of the best lawyers in the USA, wrote a completely unironic 300-odd page book titled Open Season: the Legalized Genocide of Colored People.
The logic here is purely the logic of Kendi: the sheer existence of the gap in rates of police shooting demonstrates the existence of the bias which must have created it. Much the same logic is often extended well beyond this single topic. We often hear that Black/White differences in arrest rate, incarceration rate, sentencing rate and average sentence length, and so one all demonstrate racism. Indeed, books like Michelle Alexander’s The New Jim Crow often take a practical next step: the existence of these differences indicates that the ‘carceral system’ has replaced caste-ism as a tool for the intentional control of the USA’s Blacks (and, to a lesser extent, Hispanics, Natives, poor White “hillbillies,” and other undesirables).
 However, to a striking extent, these arguments collapse whenever they are empirically tested using contemporary social science tools such as multi-variate regression. It is worth breaking this technique, which is decades, if not centuries, old and the backbone of the contemporary quantitative social sciences, down in some detail. In almost every case, population groups that vary in terms of a highly significant characteristic like race/ethnicity or political orientation also vary in terms of other characteristics like social class, region of residence, and personal behavior. As I may have noted a time or two before, the modal average age for a Black American is 27, while that for a White American is more than 30 years older – currently stable at 58.
Patterns like this are common. It is thus not possible to establish the existence of bias simply by pointing to a univariate difference between two groups (much less between two small survey samples of individuals taken from those groups). In all serious, quantitative academic papers – and this is the point – the influence of race or racism on a dependent variable like arrest rates is tested using a computerized mathematical analysis within which all other variables that might be relevant, like those given above, are held constant at their median.
When this is actually done, race often proves to have little or no remaining influence. The fact that Black Americans are more likely than White American to be shot by police, for example, appears to be a pure artifact of the fact that the Black crime rate is roughly 2.4 times the White rate – something you can calculate using Table 14 of this recent National Crime Victimization Survey report – and this alone predicts high Black rates of arrest and of hostile police encounters. And, in fact, police shooting rates track those “hostile encounters” almost exactly. In 2022, 225 of 1,096 police shooting ‘victims’ were African American, with this 20+% representation for a group compromising 12-13% of the national population largely matching (in fact slightly training) national criminal demographic patterns.
An even more in-depth multivariate examination of the police shootings question, conducted by Harvard University’s Roland Fryer, produced the same result several years ago. Using sources of data like the Police-Public Contact Survey, and including controls for the behavior of suspects alongside other “important context,” Fryer concluded that: “On the most extreme use of force – officer involved shootings – we find no racial differences.” Indeed, White officers were quite a bit (about 28%) more likely to shoot White suspects than equivalent Black suspects.[1] Fryer’s explanation for this “unexpected” finding is simplicity itself: police officers are normal citizens who don’t really want to kill anyone – and also tend to be “utility maximizers” who would prefer not to be accused of Ku Klux Klan-level racism specifically for killing Black citizens.
Racial sentencing gaps turn out to be subject to the same sort of Statistics 300-level methodological adjustments as gaps in police shooting rates and indeed crime rates (recall that age difference between White and Black males). Of course, it initially “looks bad” to say that one racial group is up to 5x as likely as another to go to state or federal prison, and tends to remain there longer. However, as we have also noted, it is also the case that some groups commit crimes and get arrested much more than others do. And, it also seems fair to hypothesize that members of high-crime groups might on average have longer criminal records than members of lower-crime groups, potentially explaining lengthier sentences for any particular crime.
 Indeed, again, that is exactly what multivariate tests that adjust for things like “extensive prior record” find. During just the past few weeks, Stetson U’s Christopher Ferguson and Sven Smith conducted what may be the best ‘meta-analysis’ of justice system outcomes so far – including no less than 51 distinct past studies – and found that there appears to be little if any actual “systemic racism” operating inside the American criminal justice system. Overall, the pair concluded that “[no] class nor race biases for criminal adjudications for either violent or property crimes could be reliably detected.”
 Some influence of race and class was – in the interest of frank honesty – found in the context of drug offenses. However, all of the relevant “effect sizes” were “very small:” a poor Black defendant in a drug case might be a few percentage points more likely to lose a criminal case than a more advantaged offender. And, even this disparity (the authors say) may well be due mostly to the poor quality of many previous studies. Overall, per Ferguson and Smith: “Narratives of ‘systemic racism’ as relates to the criminal justice system do not appear to be a constructive framework from which to understand this nuanced issue.”
It is hard to disagree with, and almost impossible to ‘debunk,’ research this sweeping and comprehensive. So: why is there such a massive gap between (probable) reality and mainstream, upper-middle class perception? Dr. Ferguson’s line about poor study design provides a clue. Simply put, many academics who research race have historically opted – intentionally or not – to not adjust for obvious factor variables which influence performance gaps between Whites and Blacks, and which would reduce or eliminate them when taken into account.
Ferguson and Smith, speaking with exquisite politeness, point out that the field of criminal justice has historically been no exception to this general trend. As I noted in a short analysis of their original piece for National Review, “In a surprising number even of the better-known studies in their field… ‘control variables are comparatively lacking – many studies that examine race don’t control for class.’ ‘Most’ studies control for ‘age and prior criminal record,’ but not all.” There is evidence, even beyond that suggested by common sense, that ideological bias plays a role here: “Studies with ‘citation bias’ toward one side of the debate about whether racism exists in the modern criminal justice system – generally, a bias in a liberal direction – were far more likely to find racism than were unbiased fair-test studies.”
While this is not the focus of this short article, the same unraveling of narrative just seen in criminal justice happens basically everywhere, when we properly adjust for the other things that obviously vary among different-age members of different races (or men and women, etc.). As economist J. O’Neill famously noted back in the 1990s, the 15-18% income differential between adult men of different races – something almost universally attributed to racism – drops to roughly 1% after adjustments for variables like mean age, region of residence, and aptitude test scores.
On a related note, per the Sociologists Darolia and Koedel, the claim of the List Experiment that job applicants with stereotypically Black names are dealt with more harshly than those with more White-sounding names seems at best dubious when social class is adjusted for. Simply put, “Jamarrian” might have a harder time getting on-boarded than “Billy,” but there is little evidence that “Marcus Blackman” or “John Lopez” would.[2] Even the famous Black/White gap in aptitude test scoring seems increasingly unlikely to be attributable to racial bias (or genetic factors): the hardly conservative Brookings Institute recently pointed out that Asian American school-children study roughly twice as much as White Americans and three times as much as Black Americans – and these patterns track extremely well with trends in later SAT scoring.
Simply put, racism – or, more specifically, discrimination: racism as praxis – can be said to exist where and only where one of two otherwise identical persons is treated differently from the other because of the sole factor variable of race. Simply finding a gap in performance, or even in treatment, between two different people proves nothing. In the arena of American crime and criminal justice, what we see today is almost entirely the second thing.
[1] To be fair, officers displayed some biases in both directions, and were up to 50% more likely to go “hands on” with both Black and Caucasian Hispanic suspects. As far as I can tell from my read of Fryer (2017) no extreme – i.e. highly statistically significant - patterns of racialized police behavior toward either white or POC suspect populations were on display.
[2] These names are very similar to, but not identical to, those used in Darolia and Koedel’s study.