In the aftermath of George Zimmerman’s trial for the killing of Trayvon Martin, Richard Cohen wrote an editorial for the Washington Post arguing that Zimmerman’s profiling of Martin was more-or-less justified. “I don’t like what George Zimmerman did, and I hate that Trayvon Martin is dead,” he wrote. “But I also can understand why Zimmerman was suspicious and why he thought Martin was wearing a uniform we all recognize.” In describing the events culminating in Martin’s death, Cohen stated:

“There’s no doubt in my mind that Zimmerman profiled Martin and, braced by a gun, set off in quest of heroism. The result was a quintessentially American tragedy—the death of a young man understandably suspected because he was black and tragically dead for the same reason.”

In essence, Cohen believes that suspicion of young black men is warranted because these young men commit a disproportionately higher percentage of crimes. The editorial described above was not the first instance in which he let these beliefs be known; in the 1986 inaugural issue of the Washington Post Magazine, Cohen wrote an article defending jewelry store owners who refused entry to young black men out of fear of crime. Nor is he the only editorialist to have expressed this belief; fellow Washington Post columnist Kathleen Parker also defended Zimmerman’s suspicion of Martin on that night:

“[T]his is one of those rare instances in which everyone is right within his or her own experience…[I]t is wrong to presume that recognizing a racial characteristic is necessarily racist. It has been established that several burglaries in Zimmerman’s neighborhood primarily involved young black males.”

A similar argument has been used to justify data showing that over 50% of those stopped under New York City’s stop-and-frisk tactic are black, despite the fact that they represent about 25% of the city’s population. Mayor Bloomberg and Former Police Commissioner Ray Kelly have often responded to accusations of racial bias in the program by pointing to the fact that the city’s murder suspects are overwhelmingly black. Mayor Bloomberg has gone so far as to say that the city “disproportionately stop[s] whites too much and minorities too little.”

The interesting thing about this claim is that it can be tested empirically, and yet none of its proponents have attempted to demonstrate its truth through statistical analysis. Using data on crime distribution across race, age, and gender groups, it is possible to estimate how many more dangerous people are correctly identified as dangerous when profiling young black men as opposed to profiling people at random. If profiling young black men leads to significantly more accurate identifications of potentially dangerous people, then perhaps Cohen and the others are right to claim that would should be suspicious of young black men.

What follows is an analysis of crime data designed to answer these questions: What safety benefits follow from profiling young black men? Will suspecting young black men of being dangerous lead one to correctly identify dangerous people more often than just suspecting people at random? If so, how much more often? And how often will suspecting young black men of being dangerous lead one to incorrectly identify a benign person as dangerous? In other words, how often can one expect to make Zimmerman’s mistake?

I. Quantifying Dangerousness

The first step in this analysis is to operationalize the concept of dangerousness. In order to test the safety benefits of profiling young black men, we need to be able to measure the number of dangerous people who are young, black, and male and the number of dangerous people who are not young, black, and male. Criminality is an obvious candidate to serve as a proxy for dangerousness, but there are many different available measurements of criminality. There are plenty of incarceration data available, broken down by race, age, and gender, but there are a number of reasons to believe that such data are under-representative of the actual criminal population. Many crimes do not end in convictions, and many convictions do not end in incarceration sentences. It is generally agreed that results from the National Crime Victimization Survey provide a more accurate picture of the amount of crimes being committed in the United States; the results from these surveys suggest that over half of all crimes are not even reported to the police. But these victimization surveys do not collect data on the suspected perpetrators, so they provide no basis for determining who is committing these crimes.

Ultimately, I believe FBI arrest data provide the best available statistical representation of dangerousness both within the population at large and the young black male population specifically. There are certainly problems with arrest data—many crimes do not lead to an arrest, and many arrestees are in fact innocent. Furthermore, this data does not represent how many different people are arrested, only how many arrests are made total. If a small group of people get arrested many different times, then the number of arrests will greatly over-represent the number of criminals/dangerous people in the population. But in the end, using arrest data is likely to provide a more accurate estimation of the number of dangerous people than using incarceration data, which tends strongly toward under-estimation. Moreover, unlike victimization or crime reporting data, arrest data involves data on the perpetrators of crime, not the crimes themselves. Conviction data would be optimal, but such data has not been collected nationwide as of yet.

II. Statistical Analysis of Profiling

The Table 1 below includes the 2012 FBI national arrest data on offenses commonly associated with dangerousness. Unfortunately, the FBI does not provide direct data on the proportion of arrests involving young black males. What they do provide are separate tables on the proportion of crimes committed by males as opposed to females, whites as opposed to blacks, and young people as opposed to older people. But by beginning with number of arrests involving males and multiplying by the percentage of arrests involving black people and the percentage of arrests involving young people, we can approximate the amount of total arrests that are attributable to young black males as a group. This calculation presupposes that the percentage of crimes committed by black people aged 16-29 more-or-less holds constant for both males and females, but this is not an unreasonable assumption given the circumstances. According to these data, there were a total of 1,723,239 arrests made in the U.S. for dangerous crimes, and an estimated 225,223 of those arrests were of black males aged 16-29.

Table 1

In combination with census data on the total amount of black males aged 16-29 that currently live in the U.S., we can use the above arrest data to project the safety benefits associated with targeting these men for suspicion. Table 2 below shows the results of this projection.

Table 2

The total black male population aged 16-29 is 4,378,690. Using the arrest data as a place-holder, 225,223 of these men, about 5% of the total, are considered dangerous for the purposes of the analysis. The total U.S. population is 308,745,538. Given that there were a total of 1,723,239 of arrests made altogether, we arrive at a base rate of about 0.5% for dangerousness in the general U.S. population.

If one were to suspect all young black males of being potentially dangerous, one would correctly identify 225,223 as dangerous. If one were to suspect a comparably sized random group of being potentially dangerous, this random group would have the same 0.5% base rate of dangerousness as the general population, meaning one would correctly identify only 21,206 people as dangerous. So the first conclusion that can be drawn from this analysis is that profiling young black men will lead one to more accurately identify strangers that are dangerous. In fact, one will be 10x times more accurate in one’s determinations of who poses a potential safety threat.

But the second conclusion that can be drawn from the data is that profiling young black men will still lead one to incorrectly identify who is and is not dangerous the vast majority of the time. A miniscule number multiplied by 10 is likely to still be miniscule; the jump from 0.5% accuracy to 5% accuracy is not as substantial as it might first seem. The rate at which one will commit Zimmerman’s mistake—assuming a young black man is dangerous when he in fact is not—is around 95%, which means that for every 1 young black man that is correctly suspected of being dangerous, about 19 young black men are incorrectly suspected of being dangerous. Given this observation, it is no wonder that 89% of those stopped in NYC under stop-and-frisk turn out to be completely innocent. Moreover, profiling young black men still leaves 1,498,016 dangerous people unsuspected.

III. Discussion

I would really hesitate to conclude from this analysis that profiling young black males yields any considerable safety benefits. While it is true that one will properly be on guard against approximately 10% more of the dangerous people by profiling these men as opposed to profiling at random, there are still around 1.5 million dangerous people out there who will catch one unaware. Imagine a scenario in which there are 17 potential threats to your safety, and you adequately protect yourself against 2 of them. With 89% of the potential threats still extant, can you really say you are that much safer than you were before?

But I suspect that many people would happily accept even this marginal increase in safety if there were no associated harms with profiling young black men. After all, small safety benefits are still better than no safety benefits. The problem with this line of thinking is that there are associated harms with profiling young black males. And while it is impossible to empirically weigh these associated harms against the safety benefits identified above, it appears obvious from even a cursory examination that the former substantially outweigh the latter.

For one, well over 90% of young black men have to deal with the stigma of being incorrectly assumed to be dangerous. The nominal safety benefits gained by profiling provide a blanket justification for unfairly subjecting these men to harassment from law enforcement and suspicion from private citizens. There are countless personal stories online in which young black men discuss the humiliation and shame of being wrongly accused of wrongdoing. In the end, treating these men with suspicion means treating them differently on the basis of their race and gender—two characteristics which we have a strong public policy against using as a basis for differential treatment.

But what’s more, profiling young black men functions as an implicit endorsement of the notion that there is some inherent link between blackness and criminality, which is clearly not what the data show. It is true that young black men are more likely than comparably sized populations to be dangerous; it is also true that dangerous young black men form an extremely tiny minority of all young black males, just as they do with all other comparable populations. Dangerousness is not the province of a single race or age group—it is the province of a tiny minority of all race and age groups. And yet, profiling young black men sends the false message that we need to worry about these young men because there is something wrong with them as a group.

Finally, the chance that law enforcement or a private citizen will mistakenly believe that a young black man is dangerous is incredibly high, and the results of such a mistake can be tragic. George Zimmerman’s shooting of Trayvon Martin provides an obvious example. When Zimmerman concluded that Martin was dangerous based solely on his appearance, the analysis above suggests that Zimmerman was about 19x more likely to be wrong than right. It was not at all surprising that Martin turned out to be doing nothing wrong; the data suggest that it was about 95% likely that this would be the case. Had Zimmerman understood this, I am sure he would thought twice before going after Martin. But he was blinded by the faulty assumption that because young black men commit more crimes than comparably sized population groups, there must be a lot of young black male criminals out there. Every time an armed person, including a police officer, makes a similar decision that someone is dangerous based solely on race, age, and gender categorizations, they put the lives of young black men in danger

IV. Conclusion

The statistical analysis suggests that while the safety benefits accrued by profiling young black men as dangerous is very minimal, the claim made by Cohen and others that there are benefits to be gained by profiling is not false. But these marginal benefits must be weighed against the stigmatization that profiling brings upon on young black men and the increased likelihood of tragic death resulting from mistakes like Zimmerman’s. It is hard to imagine anyone reasonably concluding that these nominal safety benefits are worth the terrible harm that profiling inflicts on these young men.