[EDIT: This post is getting linked a lot now for obvious reasons. Please remember it’s from 2014. There’s been a lot more research on this since then, which I have generally not followed closely. I don’t have a good impression of how the field has changed in the past six years, so this may or may not be out of date.]
Does the criminal justice system treat African-Americans fairly?
I always assumed it obviously didn’t. Then a while ago I read this harshly polemical but research-filled article claiming to prove it did. Then I found a huge review paper on the subject, written by a Harvard professor of sociology, which concluded after analyzing sixty pages of exquisitely-researched studies that:
Recognizing that research on criminal justice processing in the United States is complex and fraught with methodological problems, the weight of the evidence reviewed suggests the following. When restricted to index crimes, dozens of individual-level studies have shown that a simple direct influence of race on pretrial release, plea bargaining, conviction, sentence length, and the death penalty among adults is small to nonexistent once legally relevant variables (e.g. prior record) are controlled. For these crimes, racial differentials in sanctioning appear to match the large racial differences in criminal offending. Findings on the processing of adult index crimes therefore generally support the non-discrimination thesis.
Clearly this was more complicated than I thought. I decided to waste my precious free time reading seven zillion contradictory studies to figure out what was going on. Some people on Tumblr have demanded I report back, so here goes:
A. Encounter Rate
There are a lot of tiers to the criminal justice system, each of which will have to be analyzed individual. The first tier is – who does or doesn’t get stopped by the police?
One common point of discussion is traffic stops, leading to the popular joke that you can be stopped for a “DWB” (driving while black). Engel and Calnon (2006) seem to have done the definitive review in this area. Based on a national survey of citizens’ interactions with police, they find that 5% of whites and 11% of blacks have had their cars searched by police, with relatively similar results for other kinds of officer interactions. Therefore, blacks are about twice as likely to be searched as whites. Once you do a multiple regression controlling for other factors, like previous record, income, area stopped, et cetera, half of that difference goes away, leaving an unexplained relative risk of 1.5x.
These data admit to multiple possible interpretations. First, racist police officers could be unfairly targeting blacks. Second, blacks could be acting more suspiciously and police officers correctly picking up on this fact. Third, police officers could be racially profiling based on their past experience of more successful searches of black drivers.
One common method of disentangling these possibilities is search “success rate”. That is, if searching whites usually turns up more real crimes than searching blacks, then innocent blacks are being searched disproportionately often and the police are not just correctly responding to indicators of suspiciousness or past experiences.
Engel and Calnon review sixteen studies investigating this question. If we limit claims of dissimilarity to studies where one race is at least five percentage points higher than the other, there are eight studies with racial parity, six studies with higher white hit rates, and two studies with higher black hit rates.
In other words, in 62% of studies, police are not searching blacks disproportionately to the amount of crimes committed or presumed “indicators of suspiciousness”. In 38% of studies, they are. The differences may reflect either methodological differences (some studies finding effects others missed) or jurisdictionial differences (some studies done in areas where the police were racially biased, others done in areas where they weren’t)
The authors did their own analysis based on a national survey about citizens’ contact with the police, and found that 16% of whites searched and 8% of minorities searched reported that police had discovered contraband, a statistically significant difference. This contradicts the studies above, most of which found no difference and the others of which found much smaller differences.
One possible explanation the authors bring up is that previous research has shown black drivers who have received traffic violations are less likely than whites who have received traffic violations to admit to having received them on anonymous research surveys. For example, among North Carolina drivers known to have received tickets, 75% of whites admitted it on a survey compared to 66% of blacks (Pfaff-Wright, Tomaskovic-Devey, 2000). Comparisons of several different surveys of drug use find that “nonreporting of drug use is twice as common among blacks and Hispanics as among whites” (Mensch and Kandel). Since much of the “contraband” these surveys were asking about was, in fact, drugs, this seems pretty relevant. Overall different studies find different black-white reporting gaps (from the very small one in the traffic ticket study to the very large one on the drug use surveys). Plausibly this is related to severity of offense. Also plausibly, it relates to differential levels of trust in the system and worry about being found out – for poor black people, the possibility of (probably white) researchers being stooges who are going to send their supposedly confidential surveys to the local police station and get them locked up might be much more salient.
There are of course many other forms of police stop. These tend to follow the same pattern as traffic stops – strong data that police more often stop black people, police making the claim that black people do more things that trigger their suspicion instinct (including live in higher-crime neighborhoods), and difficulty figuring out whether this is true or false.
Sampson and Lauritsen review several studies on police stops of pedestrians. I’ll be coming back to and citing sources from this Sampson and Lauritsen article many times during this discussion as it is one of the most rigorous and trustworthy analyses around – Sampson is Professor of Sociology at Harvard and winner of the Stockholm Prize in Criminology and his review is the most cited one on this topic I could find, so I assume he represents something like a mainstream position. After reviewing a few studies, most notably Smith (1986), they conclude these sorts of police stops demonstrate no direct effect of race – in any given neighborhood, black people and white people are treated equally – but that there is an indirect effect from neighborhood – that is, the police are nastier to everybody in black neighborhoods. Although they don’t say so, the most logical explanation to me would be that black neighborhoods are poorer and therefore higher crime, and so the police are more watchful and/or paranoid.
Summary: There is good data that police stop blacks more often, both on the road and in neighborhoods. Studies conflict over whether the extra stops are justifiable; likely this varies by jurisdiction. Extra neighborhood stops are most likely neighborhood-related effects rather than race-related per se, but the neighborhood effects do disproportionately target black people.
B. Arrest Rates For Violent Crimes
Police records consistently show that black people are arrested at disproportionally high rates (compared to their presence in the population) for violent crimes. For example, blacks are arrested eight times more often for homicide and fourteen times more often for robbery. Even less flashy crimes show the same pattern: forgery, fraud, and embezzlement all hover around a relative risk of four.
(White people are arrested at disproportionally high rates for things like driving drunk, and Asians are arrested at disproportionally high rates for things like illegal gambling, but these carry lower sentences and are less likely to lead to incarceration.)
Once again, there are two possible hypotheses here: either police are biased, or black people actually commit these crimes at higher rates than other groups.
The second hypothesis has been strongly supported by crime victimization surveys, which show that the percent of arrestees who are black matches very closely matches the percent of victims who say their assailant was black. This has been constant throughout across thirty years of crime victmization surveys.
While everybody is totally on board with attributing this to structural factors like black people being poorer and living in worse neighborhoods, anyone who tries to analyze higher black arrest and incarceration rates without taking this into account is going to end up extremely confused.
There were some attempts to cross-check police data and victim data against self-reports of criminality among different races, with various weird and wonderful results. Once again, after a while someone had the bright idea to check whether people who said they hadn’t committed any crimes actually hadn’t committed any crimes, and found that a lot of them had well-verified criminal records longer than War And Peace.
Sociologists learned an important lesson that day, which is that criminals sometimes lie about being criminals.
No one has had any better ideas for how to corroborate the crime victimization survey data, so it looks like probably that’s the best we will do.
Summary: Arrests for violent crimes are probably not racially biased.
C. Arrest Rates For Minor Crimes
Usually when people talk about racial disparities in arrest rates for minor crimes, they’re talking about drugs. The basic argument is that black people and white people use drugs at “similar rates”, but black people are four times more likely to get arrested for drug crime. You can find this argument on pretty much every major media outlet: NYT, Slate, Vox, HuffPo, USA Today, et cetera.
The Bureau of Justice has done their own analysis of this issue and finds it’s more complicated. For example, all of these “equally likely to have used drugs” claims turn out to be that blacks and whites are equally likely to have “used drugs in the past year”, but blacks are far more likely to have used drugs in the past week – that is, more whites are only occasional users. That gives blacks many more opportunities to be caught by the cops. Likewise, whites are more likely to use low-penalty drugs like hallucinogens, and blacks are more likely to use high-penalty drugs like crack cocaine. Further, blacks are more likely to live in the cities, where there is a heavy police shadow, and whites in the suburbs or country, where there is a lower one.
When you do the math and control for all those things, you halve the size of the gap to “twice as likely”.
The Bureau of Justice and another source I found in the Washington Post aren’t too sure about the remaining half, either. For example, anecdotal evidence suggests white people typically do their drug deals in the dealer’s private home, and black people typically do them on street corners. My personal discussions with black and white drug users have turned up pretty much the same thing. One of those localities is much more likely to be watched by police than the other.
Finally, all of this is based on self-reported data about drug use. Remember from a couple paragraphs ago how studies showed that black people were twice as likely to fail to self-report their drug use? And you notice here that black people are twice as likely to be arrested for drug use as their self-reports suggest? That’s certainly an interesting coincidence.
The Bureau of Justice takes this possibility very seriously and adds:
Although arrested whites and arrested blacks were about equally likely to be drug use deniers, these results nevertheless have implications for the SAMHSA survey. A larger fraction of the black population than the white population consists of criminally active persons and, therefore, a larger fraction of the black population than the white population would consist of criminally active persons who use drugs but deny it. Consequently, the SAMHSA survey would probably understate the difference between whites and blacks in terms of drug use. Whether the effect of such drug use denial among criminally active persons is large enough to account for the unexplained 13% is not known, but research on the topic should pursue this possibility.
It should be noted that a study investigating this methodology gave random urine drug tests to some of the people who had filled out this survey, and found that half of the actual drug users had reported on the survey that they were squeaky clean. There were no racial data associated with this investigation, which is too bad.
Summary: Blacks appear to be arrested for drug use at a rate four times that of whites. Adjusting for known confounds reduces their rate to twice that of whites. However, other theorized confounders could mean that the real relative risk is anywhere between two and parity. Never trust the media to give you any number more complicated than today’s date..
D. Police Shootings
A topical issue these days. Once again, the same dynamic at play. We know black people are affected disproportionately to their representation in the population, but is a result of police racism or disproportionate criminality?
Mother Jones magazine has an unexpectedly beautiful presentation of the data for us:
The fourth bar seems like what we’re looking for. You could go with the fifth bar, but then you’re just adding noise of who did or didn’t duck out of the way fast enough.
As you can see, a person shot at by a police officer is more than twice as likely to be black as the average member of the general population. But, crucially, they are less likely to be black than the average violent shooter or the average person who shoots at the police.
We assume that the reason an officer shoots a suspect is because that officer believes the suspect is about to shoot or attack the officer. So if the officer were perfectly unbiased, then the racial distribution of people shot by officers would look exactly like the distribution of dangerous attackers. If it’s blacker than the distribution of dangerous attackers, the police are misidentifying blacks as dangerous attackers.
But In fact, the people shot by police are less black than the people shooting police or the violent shooters police are presumably worried about. This provides very strong evidence that, at least in New York, the police are not disproportionately shooting black people and appear to be making a special effort to avoid it.
For some reason most of the studies I could get here were pretty old, but with that caveat, this is also the conclusion of Milton (1977) looking at police departments in general, and Fyfe (1978), who analyzes older New York City data and comes to the same conclusion. However, the same researcher analyzes police shootings in Memphis and finds that these do show clear evidence of anti-minority bias, sometimes up to a 6x greater risk for blacks even after adjusting for likely confounders. The big difference seems to be that NYC officers are trained to fire only to protect their own lives from armed and dangerous suspects, but Memphis officers are (were? the study looks at data from 1970) allowed to shoot property crime suspects attempting to flee. The latter seems a lot more problematic and probably allows more room for officer bias to get through.
[EDIT: A commenter pointed out to me that Tennessee vs. Garner banned this practice in the late 1980s, meaning Memphis’ shooting rate should be lower and possibly less biased now]
The same guy looks at the race of officers involved and finds that “the data do not clearly support the contention that white [officers] had little regard for the lives of minorities”. In fact, most studies find white officers are disproportionately more likely to shoot white suspects, and black officers disproportionately more likely to shoot black suspects. This makes sense since officers are often assigned to race-congruent neighborhoods, but sure screws up the relevant narrative.
Summary: New York City data suggests no bias of officers towards shooting black suspects compared with their representation among dangerous police encounters, and if anything the reverse effect. Data from Memphis in 1970 suggests a strong bias towards shooting black suspects, probably because they shoot fleeing suspects in addition to potentially dangerous suspects, but this practice has since stopped. Older national data skews more toward the New York City side with little evidence of racial bias, but I don’t know of any recent studies which have compared the race of shooting victims to the race of dangerous attackers on a national level. There is no support for the contention that white officers are more likely than officers of other races to shoot black suspects.
E. Prosecution And Conviction Rates
Conviction rates of blacks have generally found to be less than than conviction rates of whites (Burke and Turk 1975, Petersilia 1983, Wilbanks 1987). I don’t know why so many of these studies are from the 70s and 80s, but a more recent Bureau of Justice Statistics finds that 66% of accused blacks get prosecuted compared to 69% of accused whites; 75% of prosecuted blacks get convicted compared to 78% of prosecuted whites.
The 1975 study suggested this was confounded by type of crime – for example, maybe blacks are charged more often with serious crimes for which the burden of proof is higher. The 1993 study isn’t so sure; it breaks crimes down by category and finds that if anything the pro-black bias becomes stronger. For example, 51% of blacks charged with rape are acquitted, compared to only 25% of whites. 24% of blacks charged with drug dealing are acquitted, compared to only 14% of whites. Of fourteen major crime categories, blacks have higher acquittal rates in twelve of them (whites win only in “felony traffic offenses” and “other”).
The optimistic interpretation is that there definitely isn’t any sign of bias against black people here. The pessimistic interpretation is that this would be consistent with more frivolous cases involving black people coming to the courts (ie police arrest blacks at the drop of a hat, and prosecutors and juries end up with a bunch of stupid cases without any evidence that they throw out).
There was a much talked-about study recently that found that “juries were equally likely to convict black and white offenders when there was at least one black in the jury pool, but more likely to convict blacks when there wasn’t.” This is consistent with previous studies. Jury pools contain twenty-seven members; the probability that there will be at least one black jury pool member in the trial of a black subject (who of course is most likely to live in a predominantly black area) is high. The study’s “equally likely to convict black and white offenders” was actually “2% more likely to convict white offenders than black offenders”, which was probably not statistically significant with its small sample size but is consistent with the small pro-black effects found elsewhere.
Summary: Prosecution and conviction rates favor blacks over whites, significance unclear.
Older studies of sentencing tend to find no or almost no discrepancies between blacks and whites. This was the conclusion of most of the papers reviewed in Sampson and Lauritsen. The gist here seems to be that there were “four waves” of studies in this area. The first wave, in the 1960s, was naive and poorly controlled and found that there was a lot of racial bias. The second wave, in the 1980s, controlled for more things (especially prior convictions) and found there wasn’t. The third wave was really complicated, and the writers sum it up as saying it represented:
…a shift away from the non-discrimination thesis to the idea that there is some discrimination, some of the time, in some places. These contingencies undermine the broad reach of the thesis, but the damage is not fatal to the basic argument that race discrimination is not pervasive or systemic in criminal justice processing.
The fourth wave expands on this and finds discrimination in some areas that hadn’t been studied before, such as plea bargaining. However, it continues to find that on the whole, and especially in the largest and best-designed studies there is very little evidence of discrimination. The article concludes:
Langan’s interpretation matches those of other scholars such as Petersilia (1985) and Wilbanks (1987) in suggesting that systemic discrimination does not exist. Zatz (1987) is more sympathetic to the thesis of discrimination in the form of indirect effects and subtle racism. But the proponents of this line of reasoning face a considerable burden. If the effects of race are so contingent, interactive, and indirect in a way that to date has not proved replicable, how can one allege that the “system” is discriminatory?
A more recent (fifth wave?) review adds some problems to this generally rosy picture, saying that “Of the [thirty-two studies containing ninety-five different] estimates of the direct effect of race on sentencing at the state level, 43.2% indicated harsher sentences for blacks…at the federal level 68.2% of the [eight studies containing twenty-two different] estimates of the direct effect of race on sentencing indicated harsher sentences for blacks”. The majority of estimates that did not find this were race-neutral, although six did show some bias against whites. They conclude:
Racial discrimination in sentencing in the United States today is neither invariable nor universal, nor is it as overt as it was even thirty years ago. As will be described below, while the situation has improved in some ways, racially discriminatory sentencing today is far more insidious than in the past, and treating a racial or ethnic group as a unitary body can mask the presence of discrimination.
I really like how you can make a large decrease in the level of a bad thing sound like a problem by saying it is becoming “more insidious”.
Even more recent studies have found even larger gaps. A study by the US Sentencing Commission investigating the effect of new guidelines found that blacks’ sentences were 20% longer than those of similar whites; a later methodological update reduced the gap to a still-large 14.5% and a a different recent study says just under 10%. Although the particular effect of these new guidelines is a matter of HORRIBLE SUPER-COMPLICATED DEBATE, neither side seems to deny the disparities themselves – only whether they are getting larger.
It’s not clear to me why there’s such a difference between the earlier studies (which found little evidence of disparity), the middle studies (which were about half-and-half), and these later studies (which show strong evidence of disparity). I guess one side of a HORRIBLE SUPER-COMPLICATED DEBATE would say it has to do with changes in sentencing during that time which replace mandatory sentences with “judicial discretion”. If you’re mandated to give a particular sentence for a particular crime, there’s a lot less opportunity to let bias slip in then if you can do whatever you want. There is some evidence that different judges treat different races differently, although the study has no way of proving whether this is anti-black bias, anti-white bias, or an equal mix of both in different people. Unfortunately, there is also concern that mandatory minimum sentencing is itself racist.
Capital punishment is in its own category, and pretty much all studies, old, new, anything agree it is racist as heck (Sampson and Lauritsen cite Bowers & Pierce 1980; Radelet 1981; Paternoster 1984; Keil and Vito 1989; Aguirre and Baker 1990; Baldus Woodward & Pulaski 1990 – there’s no way I’m reading through all of them so I will trust they say what the review says they say). This seems to consist not only in black suspects being more at risk, but in white victims’ deaths being more likely to get their offenders a death sentence.
Summary: Most recent studies suggest a racial sentencing disparity of about 15%, contradicting previous studies that showed lower or no disparity. Changes in sentencing guidelines are one possible explanation; poorly understood methodological differences are a second. Capital punishment still sucks.
There seems to be a strong racial bias in capital punishment and a moderate racial bias in sentence length and decision to jail.
There is ambiguity over the level of racial bias, depending on whose studies you want to believe and how strictly you define “racial bias”, in police stops, police shootings in certain jurisdictions, and arrests for minor drug offenses.
There seems to be little or no racial bias in arrests for serious violent crime, police shootings in most jurisdictions, prosecutions, or convictions.
Overall I disagree with the City Journal claim that there is no evidence of racial bias in the justice system.
But I also disagree with the people who say things like “Every part of America’s criminal justice is systemically racist by design” or “White people can get away with murder but black people are constantly persecuted for any minor infraction,” or “Every black person has to live in fear of the police all the time in a way no white person can possibly understand”. The actual level of bias is limited and detectable only through statistical aggregation of hundreds or thousands of cases, is only unambiguously present in sentencing, and there only at a level of 10-20%, and that only if you believe the most damning studies.
(except that you should probably stay out of Memphis)
It would be nice to say that this shows the criminal justice system is not disproportionately harming blacks, but unfortunately it doesn’t come anywhere close to showing anything of the sort. There are still many ways it can indirectly harm blacks without being explicitly racist. Anatole France famously said that “the law, in its majestic equality, forbids rich as well as poor people from begging for bread and sleeping under bridges”, and in the same way that the laws France cites, be they enforced ever so fairly, would still disproportionately target poor people, so other laws can, even when fairly enforced, target black people. The classic example of this is crack cocaine – a predominantly black drug – carrying a higher sentence than other whiter drugs. Even if the police are scrupulously fair in giving the same sentence to black and white cokeheads, the law will still have a disproportionate effect.
There are also entire classes of laws that are much easier on rich people than poor people – for example, any you can get out of by having a good lawyer – and entire classes of police work that are harsher on poor neighborhoods than rich neighborhoods. If the average black is poorer than the average white, then these laws would have disproportionate racial effects.
For more information on this, I would recommend Tonry and Melewski’s Malign Neglect: Race, Crime, and Punishment in America. They begin by saying everything above is true – the system mostly avoids direct racist bias against black people – and go on to say argue quite consistently that we still have a system where (their words) “recent punishment policies have replaced the urban ghetto, Jim Crow laws, and slavery as a mechanism for maintaining white dominance over blacks in the United States”. If you want something that makes the strongest case for the justice system harming blacks, written by real criminologists who know what they’re talking about, there’s your best bet.
(warning: I haven’t read the book. I did read a review article by the same people, which the book is partially based on)
Some police officers say the reason they are harsher in poor urban neighborhoods is that the expectation of high levels of unruly behavior necessitates unusually strong countermeasures. For the same reason, I am screening all comments for the next few days. If you post one, expect it to show up eventually or perhaps disappear into the aether.