The reality of police violence in the US

I wrote a post a few months ago that generated a lot of controversy and was heavily criticized. I wrote it after the officer who killed Philando Castile was acquitted, which prompted everyone to lament the so-called epidemic of unarmed black men being killed by the police. Using data about police shootings in the US compiled by the Washington Post, I showed that a black man who isn’t carrying a weapon was literally more likely to be struck by lighting than killed by the police. My point was that, given how incredibly rare these shootings are, it’s completely irrational to be afraid that you’re going to get killed by the police if you’re black and just going about your business. Yet, as I also pointed out, it’s something you often hear people say, although it’s usually hard to tell if they’re sincere or just virtue signaling. I also noted that liberals often make that very point when people are freaking out because of terrorism, but that somehow they don’t see the same thing is true about police killings of unarmed black men. It was a pretty limited point, but after I posted it, all hell broke loose. I was called a racist, people faulted me for not listening to what black people were saying, etc.

Most of the criticisms I received were really bad, so I’m not going to waste my time responding to those. For instance, one person claimed that, by the kind of logic my argument rested on, since it’s very unlikely that one is going to be struck by lightning, one might as well dance outside during a storm. But nothing I said implies such a ridiculous thing and it’s not my fault if people think it does. My argument is that if something is overwhelmingly unlikely, such as being struck by lightning or killed by the police while just going about your business, then you shouldn’t be afraid that it’s going to happen to you. Saying that does not in any way imply that, if something is overwhelmingly unlikely to happen, then it makes sense to do something that will make it more likely, such as dancing outside during a storm or brandishing a gun to a cop’s face. There is no doubt that, if people are so rarely struck by lightning or killed by the police, it’s partly because they don’t do that kind of things.

Other people raised objections that were explicitly addressed in the post, which apparently didn’t deter them. It seems that, when it comes to race, people are not only incapable of thinking rationally, but they can’t even read anymore. For instance, people claimed that my comparison with people who are struck by lightning was fallacious, since it makes it sound as if police killings were a force of nature and not something that resulted from human decisions. But I anticipated this objection and explained why it wasn’t a problem for the point I was making. My point wasn’t that we couldn’t do anything to reduce the number of people who are killed by the police, but that given how incredibly rare it already is, it’s completely irrational to be afraid it’s going to happen to you. Again, when someone makes the exact same point about terrorism, you don’t hear liberals screaming for blood. In fact, not only are they not screaming for blood, but more often than not it’s them who make that argument and, with one qualification I plan to discuss in another post, they are absolutely right.

Before I talk more about the reality of police violence in the US, I want to briefly discuss another objection some people have raised against my post, which I have to admit I didn’t anticipate when I wrote it. Indeed, some people said that, while it’s true that black people are overwhelmingly unlikely to be killed by the police, the kind of hyperbole I criticized in my post is useful to generate support for the broader cause of fighting racism in the criminal justice system. It’s amazing that people apparently can’t see the double standard this response illustrates. If a conservative said that people risk their life every time they go out because of crime, liberals would rightly criticize that hyperbole, even if the person in question objected that he only resorted to it to shed lights on the problems created by crime, which by the way are far worse than those created by police violence. Perhaps more importantly, this argument ignores another point I made in my post, namely that even from a liberal point of view, the energy spent fighting police violence against blacks would do a lot more good if people spent it on other causes, because racism in the criminal justice system doesn’t explain nearly as much as people think. On the other hand, it would not leave as much room for virtue signaling, which may be the real issue here.

But the more interesting criticism of my post is that it focused exclusively on uses of force by the police that proved lethal, but ignored other, less extreme forms of police brutality. Several people claimed that, although I had shown that unarmed black men were overwhelmingly unlikely to be killed by the police, it was still rational for them to fear it, because they are kicked, punched, etc. on a regular basis by the police. I say it’s the most interesting criticism of my post I have encountered not because it has much merit, but rather because it will allow me to dispel some widespread misconceptions about the prevalence of police violence against blacks in the US. As this reaction to my post illustrates, many people in the US, especially on the left but not only, are convinced that blacks are constantly stopped and routinely brutalized by the police. But, as the rest of this post will show, this is demonstrably false. The reality is that, even if you look at every type of police violence and not just cases in which it proved lethal, police violence against black people is overwhelmingly rare. Of course, it’s not as rare as being struck by lightning, but it’s still rare enough to make the amount of attention this issue gets and the hyperbolic statements it prompts completely irrational.

In order to show that, I’m going to use the data from the Police-Public Contact Survey (PPCS), which provides detailed information about contacts between police and the public. It’s conducted by the Bureau of Justice Statistics (BJS), as a supplement to the National Crime Victimization Survey (NCVS), based on a nationally representative sample of US residents age 16 or older. As the NCVS, the PPCS aims to collect data directly from the population, instead of relying on data provided by law enforcement agencies. Respondents are asked whether they had a contact with the police during the past 12 months and, if they did, have to answer a battery of questions about the nature of their interaction with the police during the last contact. In particular, they are asked questions about whether the police used or threatened to use force during that contact and, if so, what the police did or threatened to do exactly. Since the respondents are also asked questions about their age, race, gender, etc., it makes it possible to calculate the prevalence of police violence for various demographic groups. Moreover, since it’s entirely based on what respondents say and doesn’t rely on police reports, it eliminates the possibility of bias from law enforcement agencies.

The most recent survey was conducted in 2011, but they changed the questionnaire in a way that makes it difficult to compare with previous years, so I’m only going to use the data from the surveys conducted in 2005 and 2008. (The data are freely available online and you can find the R code I used for my calculations here.) In total, for 2005 and 2008, more than 150,000 people were reached by the survey (including 16,078 blacks, 107,915 whites and 20,274 hispanics) and the sample was weighted so as to take into account that not every demographic group had the same probability of being reached, which makes it by far the best source of information we have on police violence in the US. But the picture it paints of police violence, especially against black people, is completely at odds with the dominant narrative. Since the narrative in question claims that it’s mostly black men who are the victim of police brutality, I will restrict my discussion to males, who are more often targeted by the police no matter their race. (This is already what I did in the post where I showed that unarmed black men were more likely to be struck by thunder than killed by the police.) If you make the same calculations for black in people in general, regardless of their sex, the results are essentially the same, except that police violence comes out as even more uncommon.

For instance, the media is full of stories about black people who are stopped dozens of time a year by the police, for no other reason than because they’re black. However, as you can see on this graph, the probability that a man, regardless of his race, has at least one contact with the police in any given year is quite small.Thus, among men, only 20.7% of whites, 17.5% of blacks and 17.1% of hispanics have at least one contact with the police in any given year. White people are actually more likely to have a contact with the police than black/hispanic people. There are plenty of possible explanations for this, e. g. perhaps it results from the fact that white people are less reluctant to call the police for help, but it’s still at odds with the dominant narrative.

As you can see on this graph, the same thing is true if you look at the average number of contacts per year. White men have on average 0.35 contacts with the police every year, compared to 0.32 for black men and 0.27 for hispanics. Thus, despite how ubiquitous stories about black people who are constantly stopped by the police are in the media, they are not representative of reality.

It could be that, while on average black men have less contacts with the police than white men, they are significantly more likely to have a large number of contacts with the police. As this graph shows, this seems to be the case to some extent, but the difference is hardly impressive.Among men, the probability of having more than 3 contacts with the police per year is only 1.2% for whites, 1.5% for blacks and 0.8% for hispanics. So black men do seem to be more likely to have frequent contacts with the police, but the difference is not very large. Moreover, for reasons I will explain later, it’s unclear that bias has anything to do with it.

Of course, not every contact with the police results in the use of force, so we turn to this issue next. This graph shows the probability that, for men who had at least one contact with the police during the past 12 months, it used force or threatened to use it against them.Among men who had at least one contact with the police, the probability that force was used or threatened is 1.9% for whites, 5.5% for blacks and 3.8% for hispanics.

It should be noted that, although there are substantial differences between groups, they are exaggerated by the fact that, with the method I used for the calculation, the probability for white people is lower than it really is. Indeed, when the PPCS is administered, interviewers only ask respondents whether force was use or threatened during the last contact. Since some people had more than one contact with the police during the past 12 months, using the data from the survey to calculate the probability that force was used or threatened presumably results in underestimating it. Now, you would think that it would do so equally for every group, but it doesn’t seem to be the case. In fact, according to the BJS (see p. 7 of this report), asking respondents whether they had experienced threat or use of force during the last contact or during any encounter with the police in the past 12 months didn’t make any difference for blacks and hispanics, but it did increase the probability that force was used or threatened against whites by approximately 35%. If I applied this correction to my estimate for white men, the probability would be approximately 2.5% instead of 1.9%. However, since I want to err on the side of caution, I decided not to correct any of the estimates I computed for white men. Beside, even if I did, there would still remain substantial differences between the groups. However, I want to insist that it’s unclear to what extent these differences result from bias on the part of the police, for reasons that will be discussed below.

I just gave the probability for each racial group that force was used or threatened against men who had at least one contact with the police in this group. As we have seen, although there are differences between groups, the probability is not very high no matter what group we consider. But  they are still much higher than the probability that force was used or threatened against a random man, because as we have seen, most people don’t have any contact with the police. This graph shows the probability that force was used or threatened against a random man for each racial group.The probability that force is used or threatened against a random white man is approximately 0.4%, whereas it’s approximately 1% for black men and 0.6% for hispanic men. Again, we observe differences between groups, but the probability that force is used or threatened against a random person is very small no matter his race.

The figures I have just given include not only the actual use of force, both verbal and physical, but also the mere threat to use it. Verbal force is defined as cursing or shouting, while physical force is defined as pushing/grabbing, kicking/hitting, using a pepper spray or pointing a gun at someone. Let’s have a look at the probability that purely verbal force was used against someone who had at least one encounter with the police. Among men who had at least one contact with the police, purely verbal violence was used against 1.3% of whites, 4.2% of blacks and 2.8% of hispanics.

If we calculate the probability that a random man was victim of purely verbal force, this is what we get.Thus, the probability that purely verbal force was used against a random white man was approximately 0.3%, while for a random black man it was 0.7% and 0.5% for a random hispanic.

Now let’s have a look at the probability that physical force was used against men who had at least one contact with the police in the past 12 months.For white men, the probability is approximately 0.9%, compared to 3.4% for black men and 1.6% for hispanic men.

When you calculate the probability that physical forced was used against a random man, here is what you obtain.In this case, the probability was approximately 0.2% for a random white man, 0.6% for a random black man and 0.3% for a random hispanic man. As you no doubt noticed, we keep observing the same pattern. The rate is highest for blacks, intermediate for hispanics and lowest for whites. The temptation to infer that bias is responsible will probably be irresistible to many people, but as I already noted, it should be resisted for reasons that will soon become clear.

Now, as we have seen, the use of physical force includes relatively mild forms of violence such as pushing or grabbing. To be sure, this can be really unpleasant, but it’s a far cry from being kicked in the ribs or punched in the face… So we should also try to find out how common serious uses of force by the police are. One way to do that, which is made possible by the PPCS, is to calculate the probability that a man who had at least one contact with the police was injured as a result of it.As this graph shows, the probability in question is approximately 0.39% for white men, 0.46% for black men and 0.19% for hispanic men. In that case, the probability is higher for whites than for hispanics, but I wouldn’t put too much stock in that result, because injuries resulting from a contact with the police are so rare that it could easily be measurement or sampling error.

Now, this is already not a lot, but keep in mind that it’s the probability that a man who had at least one contact with the police during the past 12 months was injured as a result of it. Since most people didn’t have any contact with the police, the unconditional probability is even smaller.According to the PPCS, the probability that a random white man is injured by the police in any given year is approximately 0.08%, which is about the same as for a random black man, For a random hispanic man, on the other hand, it’s only 0.03%. Again, I wouldn’t make too much of the exact figures, because those rates are so low that measurement or sampling error make the estimates very imprecise.

However, if it’s possible that measurement or sampling error play a large role, it’s precisely because the probabilities are incredibly low. For instance, despite the fact that it interviewed 3182 black men in 2005, the BJS only found 2 who had been injured in a contact with the police during the past 12 months. By comparison, according to the National Hospital Ambulatory Medical Care Survey (NHAMCS), in 2014 (the last year for which data were publicly available), the emergency department visit rate for motor vehicle traffic injuries was 3.7% for black men. (The data are freely accessible online and you can find the R code I used for my calculations here.) Assuming no one had to go to the emergency department more than once for this type of injury, which presumably is approximately true, it means that a random black man is almost 44 times more likely to suffer a traffic-related injury than to be injured by the police.Moreover, although there are no doubt injuries that are not justified and for which the officers responsible should be punished, that’s probably not the case for most of them. Even if we assume that 1/3 of the injuries caused by the police are unjustified, which strikes me as implausible, it would mean that a black man is almost 131 times more likely to suffer a traffic-related injury than to be injured without justification by the police. Just let that sink in for a minute.

Again, my estimate is probably very imprecise because of measurement and sampling error, but whatever the real figure is, there is no doubt that it’s ridiculously low. We can make a similar comparison with the rate at which the police use physical force against black men. (Remember that physical force includes something as mild as pushing and grabbing, which is why most uses of physical force by the police don’t result in any injury.) Since this is significantly more common, the estimates are more precise, because measurement and sampling error have less influence. My calculations show that, even though most of the times, the use of physical force by the police is pretty limited and doesn’t result in any injury, a random black man is still 6 times more likely to suffer a traffic motor vehicle injury than to be on the receiving end of physical force from the police.Thus, if we assume that 1/3 of the uses of physical force by the police are unjustified, a random black man is approximately 18 times more likely to suffer a traffic-related injury than to be a victim of unjustified physical force by the police.

Now, it’s possible that, while on average police violence against black men is extremely rare, there is a subset of black men, presumably those who live in heavily policed areas, against whom the police routinely uses force. By looking only at the average, I might have concluded that police violence against black men is uniformly rare, when in fact there are places where it’s pretty common. As we shall see, there is some truth to this hypothesis, but not a lot of it and not enough to save the dominant narrative. Before I show that, however,  I want to point out that, even if I had only shown that police violence is overwhelming rare for the vast majority of black men but relatively common for a subgroup of this population, it would already be a serious blow to the narrative. Indeed, if you listen to what the sophisticates say, the mere fact of being male and black puts you at risk of experiencing police brutality no matter where you live, how much you make, etc. So if you’re trying to resist my conclusions and defend the narrative against my criticisms by arguing that, although police violence is overwhelmingly unlikely for most black men, it’s quite common for a subset of them, you’re in effect moving the goalposts. It doesn’t mean that what you say could not be true, but we should still take note of this fact. In any case, as I will now argue, this isn’t really what’s going on anyway.

The data from the PPCS don’t include any information about the areas where the respondents live, but there is a way around the problem, which I think makes it possible to test the hypothesis that, while for most black men police violence is extremely rare, it’s still pretty common for a non-negligible subset of them. Presumably, the people who live in heavily policed areas and are most at risk of experiencing police violence are also those who have the most contacts with the police, so we can just look at them and see how common police violence is for people in that subset of the population. For instance, we can look at black men who had more than 3 contacts with the police during the past 12 months, which as we have seen is only 1.5% of them. The probability that physical force was used against a black men in that subgroup was only 3.4%, which is about the same as for black men who had any contact with the police. (Since interviewers only asked respondents about the use of force during their last contact with the police, it’s probably a bit higher, but remember that when the BJS tried to ask respondents who had at least one contact if the police used force against them at any time during the past 12 months, it found no statistically significant difference.) In other words, even among the 1.5% of black men who are most frequently in contact with the police, one is still more likely to suffer a traffic-related injury than to experience physical force on the part of the police, although the difference is probably not significant.

In fact, since the probability that physical force was used conditional on contact seems to be the same for this subgroup as for black men who had at least one contact with the police, we may as well focus on the latter subgroup. As we have seen,  only 17.5% of black men have any contact with the police in any given year, so what this means is that, even among this group, one is about as likely to experience physical force at the hands of the police as to suffer a traffic-related injury.Thus, the probability of experiencing police violence is significantly higher among the subset of black men who have at least one contact with the police, but it’s still relatively low and we’re talking about a small minority of black men.

Now, you may think it’s not so rare and that it’s not such a small minority, but there are several things to keep in mind here. First, there is no doubt that many of the 17.5% of black men who have at least one contact with the police in any given year are not people who live in disadvantaged, heavily policed areas. Surely, there must be plenty of middle-class black men who interact with the police not because they live in such a neighborhood, but for the same reasons that middle-class white men do, i. e. because they were victim of a crime and called the police, were stopped by the police for speeding on their way to work, etc. Even if we assume that half of the black men who have at least one contact with the police in any given year live in disadvantaged neighborhoods that are heavily policed, which strikes me as utterly implausible (it would mean that middle-class black men are ridiculously less likely than their white counterparts to interact with the police), it would mean that only 8.8% of the black men who have at least one contact with the police in any given year live in that kind of area.

But the proportion of black men who live in heavily policed areas is no doubt much higher than that. Just consider that, according to the US Census Bureau, more than 25% of black men live under the federal poverty level. The vast majority of them probably live in areas that are heavily policed and so do many black men who don’t live under the federal poverty level. Indeed, it’s a well-established fact that, even among white and black people with similar income, the latter are significantly more likely to live in bad neighborhoods. (I plan to explore the reasons for this state of affairs when I review Douglas Massey and Nancy Denton’s excellent book on housing segregation in the US.) Even if we assume that only 25% of black men live in disadvantaged areas with a heavy police presence, it would still mean that, in any given year, 2/3 of them don’t have any contact with the police. Thus, according to this back-of-the-envelope calculation, a random black man who live in such a neighborhood is 3 times more likely to suffer a traffic-related injury than to experience physical force at the hands of the police. To be sure, this is just a back-of-the-envelope calculation and it would be nice to have better data, but I think it’s more than enough to establish that, even if we focus on black men who live in heavily policed areas, unjustified police violence is still very rare.

Moreover, remember that physical force includes a lot of pretty mild stuff, such as pushing and grabbing. As we have seen, even among black men who had at least one contact with the police in the past 12 months, only 0.46% were injured by the police, which is 8 times less frequent than traffic-related injuries among black men in general. (Again, don’t take the exact figure too seriously, for it’s so rare to be injured by the police that measurement and sampling error probably make the estimate very imprecise.) Perhaps more importantly, a lot of that is no doubt justified, especially against the people who are most often stopped by the police. Presumably, many of them, if not most of them, are people the police has good reasons to stop and who often behave in ways that justify the use of force. Keep in mind that, according to the Bureau of Justice Statistics, there were about 1.68 million black men under state or federal justice supervision in 2014, which is approximately 9.5% of the black male population. So we’re talking about a population where criminal behavior is relatively common. Thus, I think it’s safe to conclude that, even among black men who live in disadvantaged areas where the police is heavily present, unjustified police violence is very rare.

Of course, I know that even one case of unjustified police violence is one too many, but let’s be honest for a second. I don’t think anyone can look at these figures and tell me that the amount of attention this problem receives in the media is even remotely justified given how rare police violence actually is. (To be clear, this claim doesn’t rest on the assumption that a police-induced injury is no worse than a traffic-related injury of the same type, only that it’s not so much worse that it justifies the amount of attention police violence against black men receives compared to other problems that have far worse consequences but receive only a fraction of that attention. Some of these other problems, such as poverty, disproportionately affect black people but also plenty of other people.) The truth is that, if there is so much hysteria about this, it’s because the American sophisticates live in a parallel universe where blacks are routinely abused by the police. Back in the real world, however, blacks are almost never victim of police violence. This myth is kept alive by a bunch of demagogues such as Al Sharpton, who derive power and money from perpetuating it, as well as by the media, which regularly turns up the heat on this issue by publicizing cases before the facts are known and almost systematically assume the worst about the police.

For instance, as I pointed out a few months ago, the Department of Justice investigation about the death of Michael Brown proved beyond a doubt that Darren Wilson’s decision to shoot was justified and that early witness accounts which had been uncritically repeated in the media were mendacious. But almost nobody knows that, because the media almost completely ignored the conclusion of the investigation, as it systematically does when the facts contradict the narrative. For another example of the way in which the media often distorts the truth in order to make incidents fit the narrative better, although in that case the police wasn’t directly involved, just remember how NBC dishonestly edited the tape of George Zimmerman’s 911 call in a way that made him sound racist. To be clear, I’m not saying that journalists don’t really believe that police brutality against black is routine, I’m sure most of them do. But it affects the way in which they cover that kind of incidents and, even when they don’t lie about what the evidence shows, they will often not give a lot of publicity to the fact that contradict the narrative about the prevalence of racism in the criminal justice system and the frequency of police brutality against black people. The fact that black men are very unlikely to be victim of unjustified police violence is a good news. If people really want to help black people, they should focus their energy on more pressing issues that disproportionately affect them, such as poverty, segregation, etc.

Now, although police violence against blacks is extremely rare, it’s true that it seems to be more common than police violence against whites. (Again, I want to caution against taking the comparison that can be made from the figures I calculated too seriously, because the rates are so low for some types of outcomes that measurement or sampling error presumably results in a small signal-to-noise ratio. Moreover, as I explained above, my calculations probably underestimate the rates for white people and therefore exaggerate the racial disparity.) I already know that, in order to save the narrative, people who are committed to the view that white supremacy is a defining characteristic of the US will claim that it proves cops are biased against blacks. In a sense, this is not a very interesting claim, because there obviously is some bias against blacks in the police. A more interesting question is how much of the disparities we observe in the rates I calculated above can be explained by such a bias. I can’t answer that question precisely here, but I can explain why it’s very likely that bias only explains a relatively small part of the difference.

The reason is that, although many people can’t help equating disparities between groups with bias, this inference is fallacious because, for whatever reasons, there are large behavioral differences between groups. For instance, men are much more likely than women to be arrested for violent crimes, yet it doesn’t occur to anyone to say that bias plays anything other than a minor role in that disparity. Everybody understands that men just commit violent crimes at several times the rate of women. This probably explains why, as this graph shows, the police is also also far more likely to use physical force against men.To be specific, the probability that physical force was used by the police against a random man is approximately 0.24%, whereas it’s only 0.07% for women. But you’re never going to hear anyone blaming that disparity on bias. It’s not that bias doesn’t play any role in that disparity, which it probably does to some extent (although it’s not obvious such a bias is irrational, for reasons I explained elsewhere), but differences in behavior no doubt play a much larger role. Similarly, according to the NHAMCS, black men are almost 3 times more likely than white men to suffer a traffic-related injury, but I don’t think anyone is going to say that cars are racist.

Thus, while there seems to be differences in the rates at which force is used by the police against people of different races, part of those disparities probably have nothing to do with bias. Indeed, it’s just a fact that hispanics and, even more so, blacks commit violent crimes at much higher rates than whites. This can be shown by using the NCVS, which is another survey conducted by the BJS every year to assess the reality of victimization rates among US residents, without having to rely on police data that could be biased. The latest data currently available are from 2015, when the BJS reached more than 189,000 people for the NCVS. Interviewers ask respondents if they have been victim of a crime in the past 12 months and, if they have, they ask them a lot of questions about the nature of the incidents. In particular, for violent crimes, they ask them about the race or ethnicity of their offenders. This makes it possible to get a pretty accurate picture of differences between the rates at which people of different races commit that kind of crimes. Any difference that we observe cannot be ascribed to bias in the criminal justice system, since the data come directly from the victims.

According to my calculations, in 2015, black men were 3 times more likely to commit violent crimes than white men, while hispanic men were just 1.1 more likely to do so. (The data are freely available only and you can find the R code I used for my calculations here. I didn’t take into account crimes that were committed by multiple offenders, because the way in which the data were recorded make it difficult if you’re only interested in crimes that were perpetrated by men, but doing the calculations for both men and women suggest it would increase both the black/white and the hispanic/white disparity.) To the extent that cops are more likely to use force against people who commit violent crimes, which they surely are, this could easily explain most if not all of the disparities we have observed above in the rates at which the police uses force against people of different races. Again, that’s not to say that bias plays no role, which I’m sure it does. But it’s unlikely to explain a very large part of the discrepancy. Similarly, as I have argued elsewhere using recent meta-analyses of studies about bias in the criminal justice system, discrimination probably explains only a small part of the black/white disparity in incarceration rates, which is mostly the result of the fact that black people offend at much higher rates than white people. As I have argued in the post where I criticized people who talk about the so-called epidemic of unarmed black men shootings by the police, if liberals really want to help black people, they should focus on changing the conditions that explain why black people are more likely to engage in criminal behavior.

Note that, even if bias explained the entire difference in the rates at which the police uses force against people of different races, it would still not justify the hysteria on that issue. Indeed, while this would be a problem, it would still be the case that police violence is overwhelmingly rare even against black men. Black people in the US, at least a large segment of that group (don’t forget that most of them are doing pretty good), are without a doubt the victim of injustice. Nobody chooses to grow up in a ghetto and I often wish conservatives showed more understanding for the impact this can have on someone’s life and took issues such as housing segregation more seriously. But liberals need to understand that we are no longer in 1921 and that racism doesn’t have the explanatory power it used to have. Right now, they basically live in a fantasy world, where momentous developments such as the end of Jim Crow and the advent of civil rights were only superficial changes that didn’t really affect the fabric of the country, which remains based on white supremacy. This is the kind of quasi-religious mumbo-jumbo that makes you popular among the sophisticates, who constantly remind everyone how much they love Ta-Nehisi Coates’s deranged nonsense, but it bears no connection whatsoever to reality. Black people in the US have many problems, but with a few exceptions, police violence is not one of them.

Finally, I want to respond to one last objection, which I know is going to be raised against me because it always is. People will say that, as a white man (who to make things worse is a foreigner and didn’t grow up in the US), I should just shut up and defer to what black people have to say about this, because they’re supposed to have some kind of privileged epistemic access to facts about police violence and how it affects them. It’s really one of the most disconcerting fact about the contemporary intellectual landscape that such patently obscurantist nonsense has become so popular on the left. The issue of how prevalent police violence against black people is cannot be settled by listening to what black people have to say about this. We can only figure this out by looking at systematic evidence, such as victimization surveys, which is what I have done in this post. By doing that, I am listening to what black people have to say about police violence, but I’m doing so in a scientifically responsible way. There is no doubt that, as this poll shows, black people are much more likely than white people to think that police violence against minorities is very common. (However, as the same poll shows, they are also significantly more likely to say they wish the police were more present in their neighborhood.) But we can’t infer that, just because they’re black, they are more likely to be correct. In fact, as we have seen, white people are closer to the truth, although even them probably overestimate the prevalence of police violence against minorities.

Black people don’t have a magical radar in their head that allows them to know how common police violence is across the country or even in their neighborhood. Since they have direct knowledge of what happened to them personally, you can trust them about that. But not when it comes to larger social phenomena, for their beliefs about that are influenced by far more than just their personal experience, such as the media which constantly hypes police violence against black people. The fact is that, if you ask them about how often the police uses force against, not them personally but black people in general, many of them (though by no means all of them) will tell you that it happens all the time. But as we have seen, if you draw a representative sample of the population and ask each black person in that sample how often the police has used force against them personally, then use their answers to determine how common police violence against black people is, you find that it’s extremely rare. Black people are just wrong about how common police violence against minorities is, not in spite of the fact that they’re black, but probably because they are. Sometimes being more directly concerned by a phenomenon makes you less, not more, likely to be right about it.

Note that, on many issues, liberals have no problem recognizing that fact. For instance, there is a cottage industry of articles deploring the fact that, although crime has fallen spectacularly in the US since the 1990’s, most Americans mistakenly believe it has increased. (Incidentally, in order to show that, these articles typically rely on figures derived from the NCVS, so you should think twice if you’re inclined to criticize me for using it.) Liberals are absolutely right to point out this misperception of the evolution of crime in the US, but if people of any color can be wrong about this, there is no reason to think black people can’t be wrong about the prevalence of police violence against minorities. In fact, as we have seen, they clearly are. It’s also not the only example of this phenomenon. For instance, a recently published study based on a representative sample of US adults showed that approximately 68% of black people never or rarely experienced discrimination, while about 26.5% sometimes did and only 5.5% declared experiencing discrimination often. Yet I have no doubt that, if you asked black people how common discrimination against people in their group is, a much larger proportion of them would say that it’s very common. They just think it happens to other black people. The fact is that people of every color are often victim of misperception on politically sensitive issues. Liberals have no difficulty, on the contrary, accepting that fact when it comes to white people, yet they often talk as if black people were somehow immune to this phenomenon. But they are not.

EDIT: Some people have suggested that, although police violence against black men is extremely rare on average, it may be relatively common among a subset of black men, such as those who live in heavily policed areas. This point was made by Simon Penner in a comment below and several other people raised it on Facebook. The worry is that, by only looking at the rates at which the police use force against black men in general, I may have obfuscated the fact that, among a subgroup of black men, police violence is pretty routine. I replied to Simon in a comment, but I’m not sure it was very clear and, since I suspect most people won’t read the comments, I edited the post to add a discussion of this objection. I argue that, if we take a closer look at the data from the PPCS, we see that although there is some truth to that idea, there isn’t enough of it to invalidate my conclusions and salvage the dominant narrative about the prevalence of police brutality against black men.

ANOTHER EDIT: I just realized that I had made a mistake in the calculation of the probability that a man had more than 3 contacts with the police in the past 12 months, which resulted in a significant overestimation of that probability, so I edited the post to give the correct rates and replaced the graph with one that is accurate.

46 thoughts

  1. This has been discussed frequently on “The Glenn Show”. One response I heard from John McWhorter if I’m remembering correctly is that there is a symbolic element to the response to those high profile cases, of unfair treatment, of being followed, in suspicion and stopped that wouldn’t necessarily show up in incident reports; that there’s a failure of policing even if studies like Fryer et al. don’t find a racial bias in shootings or physical force;

    Looking at BLM it’s not only about policing, it’s also about mass incarceration, about anger and frustration regarding the economic and social place of black people in the 21st century. Conservative leaning democrats like Glenn Loury and John McWhorter will disagree with BLM and Ta-Nehisi Coates about causes and the extent that racism figures into it but I think they agree that there’s something there to be discussed which does include issues of law and order.

  2. I am very sympathetic to your point of view, generally, but lately I have been trying to be more charitable towards my ideological opponents. If the numbers say one thing, and people consistently say another, I am inclined to believe the numbers, but there is usually still a reasonable reason for why people say what they say.

    You state above (and I am, again, inclined to agree) a candidate explanation: essentially, media lying for self-serving purposes. But this can’t be the whole story. If the media was lying but it was obvious to everyone that the media was lying, then it wouldn’t convince anyone. But it does convince lots of people! So I think there must be an extra hidden variable we’re missing

    I have a candidate explanation and it might be an interesting avenue for you to follow up on, if you’re so inclined. It’s an idea I’ve taken to calling “Weaponized Simpson’s Paradox” (note: need better name). The idea is thusly: One aggregates statistics over a multimodal distribution in such a way as to generate summary statistics that are mathematically correct but don’t meaningfully reflect the reality of either distribution. Note, I’m not accusing you of doing this, but instead suggesting that this might be what you have accidentally done, due to the intrinsic properties of summary statistics over the United States.

    I think that this idea applies to a _lot_ of elements of US society. The US is gigantic, and there are dramatic disparities between different subsets of the country. So, for a random example: national averages on the subject of costs of living and specifically housing prices are, as best I can tell, essentially meaningless. Because you can roughly partition the US into two relatively homogenous subsets: high-cost-of-living urban coastal areas, and low-cost-of-living suburbs and interior states. The cost to purchase a decent-sized single family home in, say, Kansas, is (I am estimating) maybe $150k. The cost to purchase that exact same home, where I live, in the SF Bay Area, is somewhere between $1M and $2M. If we were to generate (simplified for example’s sake) average house prices for the nation, we would say that the average house costs (1.5M + 150k) / 2 = $825k. This number would be literally, mathematically, and statistically correct, but it is meaningless for all practical purposes. This number does not meaningfully reflect the actual cost of housing for anybody.

    I suspect that what is happening is that the US is like this. It’s not a nation with the very very low police violence stats that you have presented here. It’s a nation of EXTREMELY low police violence stats, punctuated with regions with extremely elevated (relative to your analysis) levels of police violence. Given the dramatic disconnect between (to use a local example) East Oakland, which is incredibly sketchy, and Alameda, a middle-class city on an island less than 5 miles away, this seems plausible to me.

    Hypothetically, if this was true, you would see an interesting effect. People who both don’t live in the dangerous areas and don’t have any social ties to them, will agree with your article but they won’t think twice about it. They’ll see reports in the media about how terrible the police are, they will think “wellp I’m sure glad I don’t live in Oakland”, and then they’ll disconnect and go on with their lives. They will in a sense know that the media, reporting national level statistics, are not accurate to their lives, but they won’t care. You, as a social scientist, will never hear from them.

    On the other hand, people who live in these areas, as well as people with social ties to them (this encompasses everything from people whose friends/family live there, to liberals who use the existence of those areas to virtue signal), will see these exaggerated statistics reported in the media, and they will agree that these (mathematically false/misleading) statistics are accurate. This is because they are not looking at the big picture of the nation, they are hyper-focused on the worst parts, and the reported statistics appear to accurately reflect those parts of the country that these people are focused on.

    To restate and summarize: If the distribution of police violence in the country is very lopsided, and the media reports on police violence in a way that dramatically overestimates the rate of violence relative to the reported statistics, there will be a lot of people for whom the reports are obviously false (and they will ignore them), and a smaller but more vocal number of people for whom the reports are obviously true (and they will proselytize them). The end result of this process is that you, an outsider, who is trying to draw conclusions from the data, will see the following:

    1) The data says the level of violence is X
    2) The media reports the level of violence as Y > X
    ====> the media is lying
    3) People believe that the level of violence is Y
    ====> The media has succeeded in deceiving people about the level of violence

    All of this longwinded comment is to say: It would be an interesting follow-up investigation to break out these statistics by geographic area, and see if it paints a more nuanced picture. It is my suspicion that doing that would show data that suggests somewhat different conclusions than the ones you’ve put forward here.


    Overall: Thank you for writing this. Your blog is generally great and I appreciate the time and effort you put into researching these issues. I believe the US is pretty innumerate, in general, and this allows motivated actors to manipulate policy into directions that are not appropriate for the facts on the ground. Blog posts like this fight against it. Everything above shouldn’t be taken as an argument, or an accusation, or even as a criticism. I’m offering it as a suggestion for a interesting direction you could take if you wished to do so.

    1. A fun, silly analogy/example that a friend came up with. It’s like the following:

      California is on fire.

      Houston is flooded.

      The national statistics average over this and say “actually, America is totally fine. Not too hot, not too cold”

      The media, who only care about California, report this as “America has a fire problem”

      You are looking at (1) the media; and (2) the national statistics, and say “well the media says there’s a fire problem, but the statistics confirm: the media is wrong. Everyone, you should chill out. There is no fire problem.

      What you are not looking at is either California specifically or Houston specifically. This is not really your fault though! You’re looking at the data, which is the only thing you have direct, reliable access to.

    2. Thanks for this thoughtful comment, I think it’s good pushback.

      First, I want to note that, if you listen to what people say in the media, they don’t just say that police violence is a major issue in a few places. We hear famous black people who live in areas such as Beverly Hills claim they feel threatened by the police all the time. I don’t think anything you say implies this isn’t the case, but I want to point that out anyway, to prevent any attempt to move the goalposts by someone else who might be tempted to claim that the narrative was always only about a small part of the black male population.

      I also want to say that I don’t think the media is lying about this or, at least, I think it’s more complicated than that. I have no doubt that most people in the media genuinely believe that black people are constantly stopped by the police, routinely brutalized, etc. They only lie when the facts don’t support this narrative, but I’m sure they think it ultimately serves the truth, and it’s typically about the details of a particular case.

      For instance, NBC dishonestly edited the tape of George Zimmerman’s 911 call to make him look bad, because it served the narrative about Trayvon Martin’s death. But I don’t think they lie when, for example, they claim that police violence against blacks is ubiquitous. Of course, I think it’s false, but I don’t think they know it and in order for something to be a lie the intention to deceive must be present.

      I now come to the main point you make in your comment. It’s a good point, and there is some truth to it, but I don’t think there is enough of it to invalidate the conclusions I draw from the data. There is actually a way we can address this, at least to some extent, with the data from the PPCS. We can just look at the subset of people who have a lot of contacts with the police.

      For instance, we could just look at black men who have more than 5 contacts per year, which means on average they had a contact with the police every other month. You could think that it’s a decent proxy for people who live in areas that are heavily policed. However, as we shall see, it only captures a very small fraction of this population

      Indeed, only 0.5% of black men have more than 5 contacts with the police in any given year, which is a pretty small number. Moreover, according to the PPCS, the police used physical force against 6.8% of them. In other words, even if we look at the 0.5% of black men who have the most contacts with the police, they are only 1.8 times more likely to be injured by the police than to suffer a traffic-related injury.

      The problem is that so few black men have more than 5 contacts with the police that measurement and sampling error probably make the estimate extremely imprecise. One way to address this problem is to use a larger subset of the sample, by looking at black men who had more than one contact with the police.

      According to the PPCS, that’s only 5.6% of black men, which confirms that, for the overwhelming majority of them, interaction with the police is pretty exceptional. Black men in that subgroup are slightly more likely to suffer a traffic-related injury than to be injured by the police.

      Now, you may think it’s not so rare anymore (although I think it’s still pretty rare), but there are several things to keep in mind here. First, physical force includes a lot of pretty mild stuff, such as pushing and grabbing. Perhaps more importantly, a lot of it is no doubt justified, especially against the people who are most often stopped by the police. Presumably, many of them, if not most of them, are people the police has good reasons to stop and against whom the use of force is often justified.

      Moreover, there are far more than 5.6% of blacks in the US who live in heavily policed areas, where under your hypothesis police violence is supposed to be routine. Compared to whites, blacks in the US are heavily concentrated in urban areas, and 27% of them live under the federal poverty level. Since both poverty and a city environment are strongly correlated to crime and police presence, the proportion of black men who live in heavily policed areas must come to significantly more than 5.6%.

      So most of the black men who live in heavily policed areas have no doubt significantly less contacts with the police and the police no doubt uses force less often against them than against the 5.6% I looked at to address your concern. (Keep in mind that only 17.5% of black men have any contact with the police in any given year.) These 5.6% are very atypical in terms of how often they come in contact with the police and, presumably, also in terms of behavior.

      So when you put all that together, I think it’s safe to say that, while looking only at the average somewhat reduces the rate at which the police uses violence against black men, it doesn’t hide the fact that a substantial part of the black male population is routinely abused by the police, which is simply not the case.

      There is a very small share of the black male population that experience police violence more often, though even for them it’s a stretch to say that it’s routine, but this has probably a lot to do with criminal behavior in that subgroup.

      When you think about it, this isn’t really surprising, even if you just look at the data I give in my post. On average, police violence is so incredibly rare and, compared to how many black live in areas that are heavily policed, there are so few of them who have any contact with the police, that it would have been really surprising if the rates I calculated were hiding the fact that police violence is actually more or less routine against a smaller but still quite large subset of black men who live in heavily policed areas.

      In short, there is some truth to what you’re saying, but I really don’t think there is enough of it to salvage the narrative. I certainly agree that it would be nice to have better data, but unless you make utterly implausible assumptions, I really don’t see how better data could significantly affect my conclusions.

      EDIT: I initially mixed up some of the numbers, so I edited the comment to correct the mistake, sorry about that. While I was at it, I also added a few observations, which I hope clarify the issue.

      ANOTHER EDIT: As I explain in a note I added to the post, I had made a mistake in calculating the probability that a man had more than 3 contacts with the police in the past 12 months, which I corrected. Since I was using the same code to calculate the proportion of black men who had more than 5 contacts with the police in this comment, the figure I gave for that probability was also mistaken. The proportion of black men who have at least 5 contacts with the police in any given year is actually much smaller than what this comment originally indicated, so I made another edit to correct the mistake. Needless to say, unlike the previous mistake, this one just strengthens my case.

      1. Thanks for this prompt reply. I appreciate the work you’re putting into this.

        Your numbers seem to suggest that my hypothetical is not correct, and this makes sense. It does leave me with this lingering feeling of “how can so many people believe this narrative is correct, when almost everyones’ personal experience must necessarily contradict this”

        On the other hand, the media has somehow successfully memed half the country into thinking the other half are nazis and need to be murdered in the streets, so maybe I rationality of human beings

    3. Of course different areas have different levels of policing. Policing is almost perfectly correlated with the amount of violent crime. What method of policing would you prefer?

      If the flip-side were true, and policing was essentially equal everywhere, it would be called racist to let certain dangerous areas go “under-policed.”

  3. Thanks for this!

    Towards the end, you argued that the different rates of police force use on different races is most likely due to different rates of violent crime between the races. I found this argument inconclusive.

    Only a small portion of the police’s interactions with civilians are interactions with people who are currently commiting a violent crime. If these are driving the use of force rates, then your argument is more persuasive.

    But the rates may be driven by interactions with folks who haven’t committed violent crimes.

    Seems like we need further analysis to come to any conclusion regarding how much of the different rates of police force are due to bias.

    1. I agree that further analysis is needed to determine how much of the disparity bias explains and I think I say so in the post. But I also think you misunderstood my argument.

      I’m not talking about the rates at which people of difference races commit violent crimes because I think a large proportion of the police’s interactions with civilians are interactions with people who are currently commiting a violent crime. (I don’t really know how common or rare that is.)

      My assumption is rather that a difference in the rate at which people of different races commit violent crimes is a decent proxy for differences in behaviors that increase the likelihood that force will be used against someone.

      But the point I really want to make is that a mere disparity is really poor evidence of bias, precisely because there are well-documented, large difference in behavior between racial/ethnic groups.

      1. Thanks for the clarification!

        Using your assumption, I still don’t see how to connect the fact that a black man is three times more likely than a white man to commit a violent crime to the fact that random black man is three times more likely than a random white man to experience police force, even if the relative rates suspiciously look the same.
        There may be further assumptions you make that I’m having trouble figuring out.

        To my mind, it’s still possible that bias makes a large contribution to the different rates of police force, which I think you disagreed with in the post.

        But I’m totally on board with the larger point: we shouldn’t infer bias from discrepancy.

        1. In other words, dependong on how large a percentage of the population are violent criminal, your explanation of the gap in force rates that has a small role for bias might require P(force/civilian violent criminal) >>> P(force/not violent criminal).

          Whether such a large difference is plausible seems to depend on how interactions that result in force tend to arise, likw if they arise more often from routine patrolling and speeding stops.

        2. I’m assuming that the rate at which people in a group commit violent crimes is correlated to the rate at which they engage in the sort of behavior that results in the use of force by the police during contacts, such as resisting arrest, attempting to run away, etc. If that is the case, then from the fact that black men are more likely to commit violent crimes than white men, it follows that they are also more likely to engage in the kind of behavior that prompts a violent response from the police.

          This strikes me as extremely plausible because, presumably, the same kind of characteristics that make you prone to commit violent crimes, such as impulsiveness, also make you prone to engage in the kind of behavior that prompt the police to use force against you.

          I’m not saying that bias doesn’t play any role, I’m sure it explains part of the disparity, but when you put that fact together with the fact that, as I point out in the post, the evidence strongly suggests that bias explains only a very small part of the racial disparities at other points of the criminal justice system (arrest, decision to convict, length of sentence, etc.), I think it’s unlikely to explain much of it.

          1. re the relationship of crime rate to police interactions

            wouldn’t police tend to police more/be called out more to neighborhoods that have higher crime rates? and therefore tend to encounter/stop more people of that particular community? they will also be stopping people and questioning people who are related to the reported crimes. if more suspects are reported as black, they will be looking out for black guys rather than say, NE asians guys.

            if community A has a higher crime rate than community B, then police will be called out more to community A and tend to interact more with people of community A than B, but it (police interactions) still be a function of crime rate rather than racism towards people in community A.

            this isn’t to say that there’s no racism, only that a difference in crime rate should be expected to drive a disparity in police brutality towards different groups. this doesn’t seem to bother us when it’s men versus women…

          2. as i couldn’t edit my post, i wanted to be more clear…

            i’m talking about two effects that would strongly relate crime rate to police violence even without racism,

            an intercommunity effect:

            if police are called out more to community A (higher crime) than community B (lower crime), they will interact more with people from community A than B, which provides more opportunities for police violence to occur.

            an intracommunity effect:

            if victims are reporting suspect of type A more than type B, police are going to stop and question more people of type A than B, again providing more opportunities for police violence to occur.

            these are two effects to add to Philippe’s.

            so even if we assumed no racism and assumed a constant police violence rate regardless of the race of an individual he encounters we should still expect to see large disparities as a function of different crime rate alone.

  4. @Simon

    There is some evidence that crime is very concentrated in certain counties. It would be interesting to see a breakdown of this data between say the top 10% and bottom 10% of counties by crime.


    I suspect your police violence numbers are inflated. People who have had contact with the police are more likely to be criminals. Police who have had violent contact with the police are much more likely to be criminals. Criminals are more likely to see police action in the worst possible light. What you and I might interpret as a gentle shoulder squeeze to calm a person down they might see as a rough grab and a push. This is just speculation though and I don’t know of a good way of getting rid of this bias. We could try to use data from the police but that is going to have its own biases embedded in it.

    — PG.

  5. I have no reason to doubt your statistics, but while it is obvious why black organizations might be motivated to exaggerate the amount of police violence against blacks, why would mainstream liberal white media and organizations do so? What is their motivation in your opinion?

    I would like to point out (if readers are not already aware) that Glenn Loury and John MacWhorter, described above as black conservative democrats, can also be described as black no-nonsense contrarians. Their conversations can be found in Bloggingheads TV. Loury, who is an Economics Professor at Brown, also interviews other progressive and generally white intellectuals.
    He might be interested in your work outlined above.

    1. I don’t think left-wing media organization think they exaggerate anything, because I’m sure they really believe anti-black police brutality is ubiquitous. What I think is that, because they believe this to be the case, they often distort the truth about particular cases and/or give undue emphasis to incidents that are presented as representative of the experience of the typical black man in the US, when in fact it’s exceptional.

      I agree with your characterization of Loury and McWhorter. I actually like both of them, especially Loury who I think is very smart and courageous. He wrote a response to students who questioned his credential as a “scholar of color” (what a ridiculous expression) on Facebook a few years ago that was absolutely priceless. I hadn’t thought about emailing him about this post to see what he thinks, but it’s a good idea and I will probably do it.

      1. This post seems like the kind of research which interests Loury. I’m sure that he would welcome reading it and of course, he has the background in statistics to appreciate it or even criticize it.

  6. Thanks Philippe! I see now exactly what you mean by “proxies.” Sure, a group’s rate of non-compliance with the police may be correlated with its rate of violent crime.

    It’s not clear to me what this establishes about how much the disparity in police force is due to bias, unless we know other things: the strength of the correlation; how much compliant behavior contributes to the chance the one experiences police force. (This isn’t an objection as much as a statement of my own confusion.)

    But I do wonder whether a finding of that Bayesian analysis of lethal force in PLOS puts pressure on the story you are trying to run in this thread. They found that “There is no relationship between county-level racial bias in police shootings and crime rates (even race-specific crime rates)…”. If differences in the behaviors that increase the likelihood force is used against you explain a significant amount of the disparity force, it’s strange we don’t see this in lethal force.

    I enjoyed reading your other posts, thanks!

    1. @Johm. Good, thanks! I can see how statistical discrimination can happen without racism.

      I agree that it seems plausible, without much empirical data, that it contributes to the disparities in force. I wish I knew how to use data to more directly determine the role it plays.

      One thing that surprised me was that the probability for a black man having at least one contact with the police was slightly smaller than the same probability for a white man. The gap between the force rates for black and white men was determined by the gap in force rates conditional on having contact. So I initially thought that bias possibly explains a large share of the discrepancy in force rates.

      1. well, unfortunately this dataset is less than ideal so don’t put too much stock into this analysis all by itself.

        for now i’d like to see us as a society approach this matter in a sober and objective way.

        first, we must dispel this notion that a differences in representation must necessarily equal bias…that must actually be proved, not just assumed. not just when it comes to police bias, but across the board there is this sloppy tendency to assume that a difference in representation can only be because of discrimination.

        that said, this is not a trivial matter to analyze, there are a multitude of factors involved, bias just being one of them, and this is in aggregate, we must resist the urge to generalize about cops as a whole. an over-reaction to perceived racism can be harmful too. crime-ridden areas actually need policing the most and hostility towards the police IN GENERAL (as opposed to those few bad cops) is sure to be counter-productive.

      2. to add another thought i just had

        the bias can run both ways, and there could be an interaction here…

        blacks could be generalizing about the police too and as a result being particularly hostile and uncooperative with the police when they do encounter them, and that could be a big factor in driving the disproportionate violence used against them by the police.

      3. on the other hand, some people who generally fear the police might comply more for fear of being hurt.

        so what makes the difference? what factors determine how somebody might respond when stopped by the police?

        * a culture of defiance and belligerence, maybe, you would expect to find americans resist arrest more than chinese who seem to have a high cultural regard for authority whereas americans are more skeptical of authority.

        * prior or present criminal behavior, even if the cops aren’t stopping you for that, it might make you jumpy around cops. a cop could want to stop you for something totally separate but you’d just been dealing drugs like that so you run.

        * testosterone could be a big factor. high-T men don’t like being pushed around and act impulsively.

        * IQ could be another big factor.

        is there good data on blacks and resisting arrest?

        anyway, these are some of the variables we’d need to correct for to try and tease out just what role racism is playing here.

        and another question, what are the predominant ages of the people experiencing police violence? if they are young men that would make the cops not only sexist, but ageist too. it would be a bit too coincidental that the groups that police are most “biased” towards also are the most criminal…

        that said, there probably is bias involved too, we over-generalize, we’re human, cops included, but that claim must be proved, not just assumed on the basis of a group disparity.

        ok, rambling off 🙂

  7. I feel that the media, while it may not lie, does sesationalize anything that will sell more papers, increase ratings or lead to more attention for the reporter/anchorperson, etc..”If it bleeds it leads” has been used to explain the tendency and the competion between various outlets has gotten so intense that trying to treat an incident in an intelligent, non-sensational manner leads to being ignored by the audience. No one wants to hear about dog bites man but man bites dog is big news; in other words, if police arrested a black male criminal today and he wasn’t beaten or shot it gets no attention and probably won’t make front page or the tv news, unless the crime itself is spectacular or heinous. But, if a suggestion of improper police violence can be introduced into the encounter or even exaggerated,it now becomes much more notable and will attract a bigger audience.

    In the case of some, not all, police killings the dead person’s family will ratchet up the intensity; which, leads to more media coverage, because they are immediately on tv with one or more lawyers looking to sue for damages.The whole media/family/lawyer perpetual motion engine begins generating blame and greed.

    I don’t have statistics to back up any of this but I lived in St.Louis during Ferguson and in Baton Rouge during the Sterling killing and am old enough to have seen several more of these cases and I find the media to be guilty of enflaming the public to meet their own goals.

  8. If these charts are right, then police contacts are significantly more threatening and at least a little more dangerous for black men then they are for other demographic groups. I don’t know if the comparison to lightning strikes is accurate, but I consider lightning storms to be somewhat dangerous events and I modify my behavior in non-trivial ways during lightning storms in order to ensure my safety.

    But lightning storms are chaotic patterns of atmospheric pressure and moisture levels. Police officers are moral agents and police departments are agencies of municipal, county, and state governments. When a police officer threatens you in his/her capacity as a police officer, he/she is threatening you on behalf of the government she represents. And I think it makes a lot of sense to be particularly concerned about how your government treats you and people like you–in ways that would not apply to, say, the weather, or traffic accidents. Especially when various governments at the local, state, and federal levels have long histories of treating members of your demographic especially poorly.

  9. I’m curious about two things. First, you use the term “the media” several times. Maybe we’re just supposed to know what this means. I don’t and as far as I can tell you don’t define it or contextualize it (against, say, other depictions of blacks in the media). This seems critical to your case, as you direct your response to some kind of dominant narrative. You’re an empirical guy, however, so I fear I’ve missed something. A second question: the overall probability of police violence doesn’t seem to be relevant. It’s the probability of of violence in certain places (such as in those high crime areas). You get close to this point when you say “But the proportion of black men who live in heavily policed areas is no doubt much higher than that .. .” but you miss the opportunity to hit upon (what I would take to be) a critical point.

  10. Interesting analysis. Would you kindly include error bars for your figures to promote interpretation of the findings?

    1. Thanks. You are right that I really should have, and to be honest, I don’t really have a good excuse for not having done so. The only reason I didn’t is because the data is weighted and I was too lazy to learn how to calculate confidence intervals for a weighted sample. I don’t think it’s really essential for the conclusions I draw, but it would still be worth doing, so perhaps I will do it later when I have some time.

  11. Thank you for the statistical analysis. I will be concise. 1) I don’t see the white male victims in the media. I think this further skews the perception that black men are disproportionately brutalized. Given the larger number of white males in this country compared to black men, why don’t we see any reporting from FOX or other conservative media sources on these events?

    You stated that white men are less likely to report conflicts with the police. Also, you stated that black men are more likely to be involved in violent crime. However, these statistics do not explain the zero reporting on police violence against white men.

    This abscene in reporting skews the data and fuels the perception that you have addressed.

    Perhaps, the spin media machines have struck again. Your feedback is appreciated.
    ( It’s difficult to determine tone from a message. I am not angry, upset, or trying to dispute or endorse your statistics. I am simply stating what I noticed missing when it comes to police brutality in this nation. If it brutality only impacts a small minority called black man why doesn’t it impact a larger majority called white males?)

  12. Are you sure your baseline comparison (motor vehicle traffic injuries) is correct? NHTSA reports 2.4 million traffic injuries in 2015, so about 0.75% of the population. ( It’s conceivable that black men are injured 5 times more than the average person, but that seems quite unlikely to me. Also, your view that 0.6% a year of some subpopulation getting roughed up is small seems odd to me. As a member of a community, I would think about that kind of thing over a 5-10 year time frame. If 3-6% of my friends and family were getting manhandled by the police, it would have a strong affect on how I thought about the world. Also, I think a comparison with other intentional human threats would have been much more germane.

    Here’s how I would read your data if I were an African-American. I’m reasonably likely to encounter the police during the year, and it will probably happen to me in the next three years. When that contact occurs, the chance it will be unpleasant is far greater than other human contact I have. 5-6% of the time I’ll be threatened. When I’m threatened, I will almost never get killed if I don’t have a gun, and significant injuries are rare. However, I’m about as likely to get manhandled by the police as I am to get injured in an accident when the encounter comes (and I think quite a bit more likely if your injury stats are wrong). I’m about as likely to get manhandled by police as to get burgled or vandalized, so it’s a pretty substantial risk. The risk of physical debilitation is relatively low, but I know that I or a friend or family member will see this happen.

    It’s really hard to know what percentage of the time this happens to someone who either hasn’t committed a crime or who has committed a minor offense. But you’re suggesting it’s rare, and African-Americans tell me it’s a regular part of life. Your data suggests to me that they’re right.

    Also, I can’t think of a good reason why you ignored the Fryer paper. It’s quite a bit more serious as academic work goes than what you’ve done, although I did benefit from seeing your statistics. As he indicates, it’s only preliminary work on a sample of cities, but it’s quite a bit better start on whether the incidence of violence is greater for African-Americans. Isn’t his work something you’d need to need to at least address if you want to be taken seriously?

    1. As I explain in the post, to calculate the traffic-related injury rate, I used the data from the NHAMCS. The discrepancy with the number you quote, which comes from the NHTSA, could simply be explained by the fact that black men are more likely to be injured in a traffic accident than other people. According to the data I used, they are almost 3 times more likely to suffer a traffic-related injury than white men, and since men are also significantly more likely than women to be injured in a traffic accident, I don’t think 5x is a stretch. But it’s also possible that the discrepancy comes from the fact that the source of the data is not the same. The NHTSA seems to rely on police reports, whereas the NHAMCS relies on emergency department records, so I’m inclined to think the latter is more reliable because presumably the police doesn’t know about a lot of people who are injured in a traffic-related accident.

      As for the claim that 0.6% of black men are “roughed up” by the police very year, let’s keep in mind that this includes pretty mild stuff such as being pushed and grabbed, which is not what most people have in mind when they hear about someone being “roughed up”. If we’re looking at the rate at which serious force is used against black men, it’s a lot more uncommon. Moreover, as I note in the post, keep in mind that, most of the times when the police uses force, it’s probably justified. Your calculation of the probability over a 5-10 year time frame implicitly rests on the assumption that experiencing violence on the part of the police in year t is probabilistically independent from experiencing it some other year, but this is clearly false. Indeed, people who engage in criminal behavior, no matter their race, are presumably much more likely than other people to experience police violence. And, as I also point out in my post, about 9.5% of the black male population is under justice supervision at any given time.

      So when you put all that together, you can see that, for a random law-abiding black man, the probability of experiencing serious police violence must be much smaller than 3-6%. I think it’s clear that, by any reasonable standard, it’s overwhelmingly rare. In particular, it’s certainly not true that a random law-abiding black man can know that he, a friend of his friends of a member of his family will experience this. Now, if someone wants to insist that it’s nevertheless not rare, there is not much I can say because “rare” is a context-dependent and subjective term, but at least I can insist that we use precise, numerical statements. If I can only convince people of that, that’s good enough for me. But I can’t help thinking there is quite a bit of goalpost moving involved here, because I’m pretty sure that if I had asked people before I looked at the data how rare unjustified police violence must be for it not to be a major concern, people would have given me much higher rates.

      I didn’t discuss Fryer’s paper, although I think it’s a good and important paper, because as you note yourself it can’t tell us much about bias in the use of force for the US in general, but even more importantly because my post is not about bias and that’s why I only talk about it incidentally, because I know that people will raise the issue. The data from the PPCS aren’t really suited to determine how much bias explains the black/white disparity. I just want people to understand that one can’t infer bias from the mere existence of a disparity. I have no doubt that bias is part of the story, but I doubt it’s most of it.

      Note that even Fryer, who in addition to the PPCS used data that are much better suited for detecting bias, although they come from the police so there probably is some bias in reporting, doesn’t find huge differences conditional on circumstances of the stop. We’re talking about odds ratios of 1.17 or something like that when fixed effects and all the controls are in the model for the Stop and Frisk data. (He finds a much larger effect of race when he uses the PPCS, but as I said, compared to the Stop and Frisk data, the PPCS doesn’t allow for very good controls.) So it really wouldn’t fundamentally change the picture I paint in my post.

      But again my point is not about bias except incidentally and because I know that people are obsessed with that. What I’m trying to convince people of is that police violence against black men is rare enough that, compared to other problems (such as poverty or indeed crime) that also disproportionately affect black people and have far worse consequence, the amount of attention the issue of black violence against black receives is unjustified, because that attention could do a lot more good if it were directed at these other issues. This would still be true, in my view, even if the whole black/white disparity in the use of force resulted from bias, which is evidently not the case.

      1. Thanks for your response, I appreciate it.

        I still think you need to reevaluate use of the NHAMCS data. The CDC data on injury suggests 50% higher rate of injury for African-Americans not a factor of 3. Also, I think your argument about the ratio of males to females mistakes who drives for who gets injured. CDC data suggests slightly more injuries for women. ( Could the difference result from multiple visits for an injury?

        With respect to independence in suggesting 3-6%, the 3% was to account for that. You suggested that multiple police contacts within a year occurred for about 1.5% of African-Americans, so that seems a good starting point across years. 4-5% seems about right to me across a decade, and a community’s memory will be cumulative. I would agree that police use of force is justified in very many cases, but that seems like the kind of thing data should shed light on. Why presume we know the answer until we get better data? I don’t think that reflecting on what “most people” think helps. People are generally dreadful at those types of estimates. You’re right that if asked in advance they would get the answer wrong, but they answer most such questions wrong. For the group of relatively well educated people concerned about the subject, my guess is that a large percentage know that the largest part of the problem is people being verbally abused or manhandled in a way that doesn’t lead to extensive physical damage and a small part of the problem is people getting beaten to a pulp. The average non-fatal viral video involving police abuse doesn’t lead to lasting physical damage (often not any at all). However, most people who put themselves in that situation find it horrifying and see it as a major problem. Your “overwhelmingly rare” conclusion relies on such incidents being considered not serious.

        I agree with you that the majority of police interactions involving police violence have some justification on the part of the police. But I think you likely underestimate the percentage that involve innocent citizens. As you indicate, the survey sample leaves out the 9.5% of African-American men under judicial system supervision, and parolees are likely under-sampled as well. What percentage of innocent people experiencing manhandling would be too large? 5% would definitely reach that threshold for me, and probably even 1-2%. I also think segmenting the issue into law-abiding vs. non law-abiding is too binary. Many inappropriate incidents of police violence probably involve situations where police begin with legitimate concerns based on past experience. A fair number probably involve situations where poor attitudes on the part of a police officer or a suspect at initial contact lead to escalation and there is shared blame.

        I think your writing on incidence of bias strongly leads readers to infer that it isn’t a significant problem. I think it’s helpful that you argued it’s unreasonable to assume the 3:1 ratio difference doesn’t automatically indicate bias. But given strong evidence from the Fryer paper that bias plays a significant role in police departments with good data and past records, why not indicate it is a serious concern? Argumentation that minimizes any conceivable concern on the part of opponents convinces the already convinced quite strongly. It also creates significant resistance in those who recognize the evasion.

        I realize that prevalence of violence, not incidence of bias, is your concern. And some of the people most worried about police violence would probably do a great deal of good by recognizing that much criminality results from poor personal choices and that in communities strongly affected by criminality police will need to use force and it’s not helpful to make them nervous to do that when it’s the right choice. But there’s not nearly enough evidence from this survey data to conclude that police violence isn’t an issue. You don’t say so explicitly, but the only reasonable conclusion from accepting your argument would be that African-Americans have imagined a police violence problem that doesn’t exist, and that where it appears to exist the cause is justified police response to African-American criminality. If you’re wrong and police violence does harm a moderately large number of people, and if this makes it more difficult for communities to cooperate to stop criminality, you’ll have done a great disservice to anyone who accepts your conclusions.

        1. The conclusion I’ve drawn, from his two main posts on police violence, is that the standard narrative (upon which BLM, the MSM, and others have relied) is, if not utterly false, grossly and disturbingly overstated.

          If that’s right, then everyone was done a great service, and we can only hope many others are persuaded. If it’s false, then it shouldn’t be difficult to disprove given the education levels of standard narrative true believers. Moreover, it shouldn’t be difficult to hold onto the standard narrative given the number of people who espouse it and its dominance within the media and culture at large.

          If this sort of “you have to consider all the potential consequences of stating quite possibly true things” response were made against an argument that subverted a widely-held conservative narrative, then it would be roundly mocked.

          1. With excellent data, I think you have a point. Academia leans left and there are plenty of people with the ability to analyze the data and the motivation to make the case. What you see in the Fryer paper regarding incidence of bias in policing is that evidence suggested one widely held view hasn’t been confirmed from data (no evidence for more bias in fatal shootings), one received more support from the evidence (bias in other applications of police violence) and a good model developed for evaluating the issue. The circumstances (willing police forces who wanted the data analyzed) would suggest there might be bias in fatal shootings where the costs to the shooter are lower, but we shouldn’t believe that until there’s more evidence. They also suggest that where the costs of bias in other applications of police violence are lower, there might be more bias. What you call the “standard narrative” got a needed correction on one count and a substantial correction on the other.

            These posts see the level of physical violence described as low. My first point was that I don’t I agree, and I think many others wouldn’t when they think about what the data means. If 17.5% of a population has police contact, 10% of those who do experience some kind of hostility, and for 3.4% of them it’s manhandling, I think most would say, “That sounds bad, but maybe not as bad as I thought.” I don’t think it grossly and disturbingly overstates the situation. If I’m right that it’s about as common as an injury in a car accident (we disagree about the numbers), then physical violence from police is about as common as the 2nd leading cause of non-fatal injury in most age groups. That’s not inconsequential. If, undeserved, black men have an extra risk factor in their lives about as common as car accidents, then it’s well worth media attention. If the author is right that most of the people who experienced this deserved it, then, his point deserves a greater hearing.

            But that’s where his data goes weak. He mentions the DOJ investigation of the Michael Brown. That’s important. It’s a huge problem that the media didn’t publicize their error. But he doesn’t mention the part of the DOJ study that demonstrated constant harassment-type interactions with police and local government very consistent with the type of non-life threatening but potentially demeaning police violence he doesn’t see as a problem. The action of Ferguson officials over time likely did a lot to cause mistrust of police that caused residents not to give the police the benefit of the doubt when they otherwise should have, and turned (with egregious media error) a manageable problem into a large one.

            But from that point forward, there isn’t much real data. The author suggests, and it’s certainly a hypothesis that has to be included in analysis until ruled out, that differences in behavior by African-Americans could explain any discrepancy in violence. Better evidence (the Fryer study) suggests that’s not true. But how is pointing out that what the author views as a small amount of police violence, and what I see as a rather large amount, which quality data analysis has yet to assign responsibility for, a “you have to consider all the potential consequences of stating quite possibly true things” response? I’m just pointing out the absence of evidence supporting a hypothesis in this point of the investigation.

          2. I’m really swamped right now, so this reply will be brief and will address only what I take to be the most important points, sorry about that.

            There may better data on traffic-related injuries out there, but I don’t know any and I really don’t think the CDC data should be preferred. They come from the NEISS, which relies on a sample of 100 hospitals that does not seem to have been updated since 200, whereas the data from the NHAMCS rely on a sample that is almost 4 times as large and is updated every year. Moreover, on the website you linked to, they say that injury rates by race were not calculated because it was often unknown, so there is really no doubt that the NHAMCS dataset is much superior for this. But it really doesn’t matter, since what I say is true no matter what what estimate you use for the rate at which black men suffer traffic-related injuries.

            The fact is that, according to the PPCS, only 0.08% of black men are injured by the police every year. This estimate is pretty imprecise, but it also includes all the people who were injured by the police in a contact where the use of force was justified. Sure, we can’t know what proportion of these injuries result from a justified use of force, but I’m having a hard time taking seriously anyone who thinks more than half are. Anyway, even if every injury resulting from a contact with the police were unjustified, and we accepted the way in which you try to account for the fact that such incidents are not probabilistically independent, it would still mean that the probability that a random black man is injured by the police in a decade is only 0.4%. This is still far more unlikely than experiencing a traffic-related injury, no matter what dataset you use to calculate the traffic-related injury rate. It also doesn’t even account for the fact that debilitating injuries are probably far more common for traffic-related injuries than for police-induced injuries.

            Of course, the truth is that we have no way to calculate the cumulative probability of being injured by the police over several years, because we simply don’t have enough data for that. What is clear is that, in any given year, a black man is far more likely to be injured in a traffic accident than in a contact with the police. This is true no matter how you estimate the rate of traffic-related injury and it’s even more true if you consider that only a fraction of police-induced injuries resulted from unjustified use of force. It’s also true even if we count every use of physical force, including those which do not result in any injury and largely consist of relatively minor things like pushing or grabbing. Again, this is true no matter what dataset you use to compute the traffic-related injury rate for black men, because the probability that a black man experiences any kind of physical force on the part of the police is still only 0.6%, which is almost half the traffic-related injury rate for black men according to the CDC.

            Several other things you say are incorrect. For instance, it’s not the case that the PPCS leaves out 9.5% of the black male population, you misunderstood what I said. The only black men that were excluded are those who are incarcerated, but they make up less than 1/3 of the black male population under justice supervision. The majority are either on probation or parole and they were not excluded from the PPCS. I only cited this figure because it shows that criminal behavior is pretty common in the black male population, especially compared to the white male population, so it’s yet another fact which makes it unlikely that bias is a major factor to explain the disparity in the use of force, because presumably justified use of force is significantly more likely against criminals.

            You also keep repeating that Fryer’s study contradicts what I’m saying about bias. Again, my post is not really about bias, but in any case Fryer’s paper does not contradict what I say about this in my post. In fact, not only does it not contradict it, but it supports it. For the Stop and Frisk data, the odds ratio for black goes from 1.53 to 1.17 when you add the controls and the fixed effects, which suggests that even for this notoriously aggressive program, most of the white/black gap in use of force doesn’t result from bias. (The odds ratio remains quite high for the PPCS, but as I have already noted, this dataset really doesn’t allow for much relevant controls as far as behavior during the contact is concerned.) This is consistent with what is observed at every stage of the criminal justice system, where relevant legal variables, not bias against black men, explains the vast majority of the black/white disparities.

            So really I just don’t see how anyone could in good faith deny that unjustified police violence, especially the use of serious physical force, is exceptional compared to something like traffic-related injuries or even violent crime. Yet the issue of unjustified police violence receives several orders of magnitude more attention than the risk of traffic-related injuries, so even if I agree that unjustified police violence is not morally equivalent, I don’t see how people can deny that the hysteria about this is complexly irrational. There are just many other issues that have far worse consequences and don’t receive a fraction of the attention unjustified police violence does. And what really annoys me is that, no matter how unlikely I show that one type of police misconduct is, people always move the goalposts and say that it still deserves the attention it gets because there is this other kind of police misconduct, which is not what they were talking about at first and isn’t nearly as bad, which is perhaps not as rare as what I was talking about. It just never ends.

            Anyway, this ended up being much longer than I thought, but I think I’ll leave it at that, because if this still doesn’t convince you that the issue of unjustified police violence is overblown compared to other, more pressing issues such as poverty, I don’t think anything I can say will convince you. I think it’s better to just leave you the last word and let people who are reading this decide if they still think unjustified police violence deserves the attention it receives.

  13. Nothing is about the pawn who does the dirty work, the puppets (Black Muslims) controlling the pawns (criminals) have worked for years as a hierarchy through religious channels and non-profits, to get the pawns to do terrible things and sacrifice and put themselves in public view and the possibility of being shot by defying authority (cops) to re-frame the “what it looks like” and exploit the incident and use it as a Black Muslim platform. The bottom line is that the highest level of the hierarchy is after cash for associated non-profits (below I list what has been gotten so far), and is pushing for an increase of foreign blacks to enter the USA to increase the black population here. If you cannot accept that information, I suggest that you review the many videos available of New Yorker, Farrakhan. The federal government has been keeping tabs on him for years. WATCH: America does not oppress black people. The cop in St. Louis is a bad cop and there are many, but the bad cops don’t represent the whole. There are very big problems with criminals who do bad things, and FBI has public records that show how many people of each race, gender, and type of crime that is committed, it also shows that White people get shot for crimes, but certain people want to CENSOR the facts because then it shows that the problem of cops shooting people isn’t aimed at one race. There is a marketing campaign to promote underachievers and CENSOR high achievers, because it means billions of perpetual tax dollar investment funds towards African American colleges, entertainment, businesses, politics, housing, development programs and preferential hiring. If America has factually oppressed black people in that past 37-years we wouldn’t have allowed black Senators, House Representatives, Mayors, Governors, a President, lawyers, doctors, dentists, engineers, police officers, designers, or race based magazines such as Ebony, Gospel Truth, Black MBA Magazine, Essence, Negro Digest, Jet, Sister, and the other 50+ magazines.. Or all black movies and TV shows such as the Jeffersons, Different Strokes, Soul Train, Cosby Show, and the thousands of other TV shows and movies, and American Black Film Festival. We wouldn’t have black singers on labels such as Usher, Beyonce, Rihanna, Stevie Nicks, Carl Ray and thousands of others. We wouldn’t have 35 African-American owned banks and credit unions in the United States, or 107 HBCU all black colleges which receive tax dollars from white and black people in the USA, and 74% of all black high school students go on to college as of 2016 some on scholarships designated for African Americans, which teach folks how to start their own business as well even though 77% of ALL AMERICAN BUSINESS OWNERS do not posses a college degree at all. FACT there are over 2.6 million black firms in the USA today. Many African American college students become engineers, doctors, dentists, lawyers, and other high paying occupations. Blacks also receive Affirmative Action Hiring Quotas preference at ALL government agencies, that is why most government offices have a high percentage of black government workers, should I mention that government workers receive pensions for retirement funding? FACT: only 99 companies in America in 2015 still offered pension plans. In government contracting minority owned businesses receive preferential contracts on ALL government construction projects. For those that are from low income they get free or reduced college, the families are eligible for free or reduced housing, food stamps, transportation. Those facts get CENSORED by certain groups too, that feel that the idea of “disadvantaged” would be taken away. So no, I do not agree with you or anyone else on white people keeping black people in perpetual poverty. But does discrimination exist? YES, but it is not one-sided. Go to YouTube you will find plenty of videos of crimes committed against white people, especially the elderly, and because of the media and certain groups making constant claims that white people are the reason for certain people living in poverty, THAT is the reason the criminals give to excuse abuse against white people. So NO, I am not going to be part of the perpetuating white-guilt that exists today in 2017. By volume more people of Caucasian race contribute to the tax base which is spent on improving the lives of black people and all other races as well. And by volume Caucasian people purchase more black entertainment shown in the theaters and on stage. Amazon was even accused falsely by a writer who censored information in his story about how Amazon didn’t deliver to certain neighborhoods because the areas were primarily black. NOT TRUE, they didn’t deliver to those areas because of the HIGH BURGLARY RATE, in fact there were 27,000 burglaries, thefts, and auto theft in ONE YEAR. And FYI, many of the Black Muslim Leaders that are pushing this “white people are keeping black people down” idea are not from America, they are from the country Africa, and other surrounding areas. There are many videos of religious sermons taped here in America by Tingba Muhammad and Louis Eugene Wolcott of New York, (aka Minister Louis Farrakhan, his parents Sarah Mae Manning and Percival Clark were from the Caribbean), where you can listen and watch them say such. FYI, Africans pushed the BEE Organization agenda which imposed racial hiring that fired 95% of all Caucasians from their jobs and replaced them with black employees. No company was allowed to hire any Caucasians if it exceeded 5% of the total employees. Caucasians in Africa lived in deep poverty as a result. To worsen the problem, Caucasian assets (all their belongings including their houses, cars, furniture, farms, businesses) were “REALLOCATED” by the government to black people. Those are the folks telling American black people that white people are keeping them down, and in some religious sermons telling them that if they don’t injure white people and disobey all authority of the American government, that other black Muslims will pressure THEM. I’m certain that they don’t want that information in the forefront either, so they CENSOR access to it.
    Media is bombarding viewers with messages that imply “All Blacks Live In Poverty”. Well here are some real numbers to consider that will surprise FACT: 2,511,340 U.S. BLACKS EARN MORE THAN $100,000 A YEAR

    Population: Blacks 42,000,000 *Persons not in the Labor Force include retired, children, disabled, homemakers, and those not employed.
    Civilian Labor Force = 19,318,000 Black persons, 123,607,00 White persons / Considering that at the start of the American Recession in 2009, with over 19 million black workers in the labor force, black households earned $836 billion dollars and spent an estimated $507 billion in 27 product and services categories. That’s an increase of 16.6% over the $435 billion spent in 2008. African-Americans’ total earned income for 2009 is estimated at $836 billion.
    Average Income of Labor Force:
    Blacks $37,090 Average.
    155,922 people earned over $1,600,000 last year in personal income.
    405,678 Blacks earned over $200,000
    540,904 Blacks earned $150,000-$199,999 last year.
    1,564,758 Blacks earned $100,000-$149,999 last year.
    1,603,394 Blacks earned $75,000-$99,999 last year.
    2,917,018 Blacks earned $50,000-$74,999 last year.
    2,801,110 Blacks earned $35,000-$49,999 last year.

    2,434,068 Blacks earned $25,000-$34,999 last year.
    (Some High School or HS Graduate)
    2,762,474 Blacks earned $15,000-$24,999 last year. Includes Part-Time Workers
    4,307,914 Blacks earned Under $15,000 last year. Includes Part-Time Workers

    In 2014, 71% of Black High School graduates enrolled in college, 67% of White High School graduates enrolled in college.

    Depending on the city, 8-24% of Blacks dropout of of High School in 10th grade, while 4-6% of whites do. The “status dropout rate” represents the percentage of 16- through 24-year-olds who are not enrolled in school and have not earned a high school credential
    Learn how the U.S. Census Bureau serves America as the leading source of quality data about our people,…

  14. Thanks for your response Philippe. I made my comparisons to getting manhandled by the police in response to a comparison you made at one point in the post. You started with a comparison to injuries, and that’s a more apt comparison. I should have taken that into account. I find your response on the accident data very unpersuasive. You’re saying that you will ignore the data from the US government agency whose job it is to determine injury frequency (and you refer to like it’s some random website), because: 1) the sample size is larger; 2) the hospitals in the sample were adjusted more recently; 3) something about injury rates by race. For 1, clearly with n that large, the sample size won’t matter one way or another. For 2, the sample for NEISS was chosen for representativeness. It’s not that unlikely it still serves that purpose (and insuring consistency for their longitudinal analysis makes sense). For injury rates by race they say that “Because of the relatively high proportion of cases with unknown/unspecified race/ethnicity (approximately 17% of overall cases), WISQARS Nonfatal does not provide injury rates for selected race/ethnicity groups.” They don’t say that they don’t provide them by race, and the data they do provide is consistent with US demography. More importantly, NEISS collects and analyzes the data for the intended purpose of this analysis. NHAMCS doesn’t. NEISS says, “At the end of each day, a NEISS hospital coordinator reviews all emergency department records for the day, selecting those that meet the criteria for inclusion in NEISS. The NEISS coordinator abstracts pertinent data from the selected emergency department record and transcribes it into coded form and onto a NEISS coding sheet, using rules described in a NEISS Coding Manual.” That’s got to matter more than a large, less representative, but more recently adjusted sample for providing good injury data. You’re left saying that somehow the US government agency tasked with doing the job miscounts injury data by a factor of 4-5 and that 3.7% of African-American men are injured in accidents every year. I had a hard time finding good data on how many accidents occurred in the US, but it appeared like 4% was about the total accident rate for major and minor accidents in the year. I made a more serious error for the purpose of having a good discussion in not adequately representing the argument you intended regarding injury. Are you sure you aren’t being a little stubborn here about reevaluating the quality of your work? I find it hard to take people seriously when there’s a reasonably large chance they’ve erred and they would rather justify themselves than make sure they are right. It’s almost certainly not true that there is no doubt that the NHAMCS data is better. Can you even rule out that many of the injuries included are repeat injuries for the same person?

    With respect to the PPCS survey, I misunderstood what the 9.5% referred to, but not your intent. I realize you’re trying to say that physical violence is likely justified because of a propensity to criminality among African-American males. My point was that you’ve underestimated the degree to which the PPCS sample is law-abiding. By misunderstanding the 9.5% reference, I’m less right than I thought. But parolees and those on probation are almost certainly significantly undersampled, so I think the argument is still merits a response, which you didn’t provide. With respect to your larger argument on this subject, I did recognize and respond to it by saying, “I think it’s helpful that you argued it’s unreasonable to assume the 3:1 ratio difference doesn’t automatically indicate bias.” But in the absence of higher quality data, there’s a whole lot we don’t know about the degree to which the difference results from bias. There’s still plenty of room for 20%, 50% or even more of the difference to result from bias. Or you could be right that it’s negligible. But for the purpose of policymaking, your eyeball estimates of criminality and the degree of justification for bias aren’t nearly good enough. Is there really a data-based argument that you can rule out a 20% greater use of unjustified police physical violence? Until there’s better data, why not just say, “We don’t know enough to answer the question yet, and it’s important to get that data?” Why would you want your audience to rely on your hunch if there was a potential problem and believing you might keep people from finding it?

    With respect to the purpose of your argument, I said “I realize that prevalence of violence, not incidence of bias, is your concern.” And as I indicated earlier, I said in the prior post that I think you’re making an important point when demonstrating that much of the apparent difference in physical violence is explained by other variables. But you didn’t acknowledge in your original piece that the best data on the subject at present does indicate bias as an explanation for part of the physical violence, and that both the author’s theoretical model and the nature of his data set (he says this explicitly) suggest violence might be greater in other samples. You’re certainly right that Fryer’s work supports one line of argument you want to make about incidence of violence. However, it does contradict the impression your work leaves that there is unlikely to be any meaningful disparity as a result of bias. His work is excellent and pertinent for two important things you write about so I can’t see any good reason it wasn’t included.

    With respect to what a good faith response would indicate, in my initial response, I said I would look at your data this way. “When I’m threatened, I will almost never get killed if I don’t have a gun, and significant injuries are rare. However, I’m about as likely to get manhandled by the police as I am to get injured in an accident when the encounter comes (and I think quite a bit more likely if your injury stats are wrong). I’m about as likely to get manhandled by police as to get burgled or vandalized, so it’s a pretty substantial risk. The risk of physical debilitation is relatively low …” Isn’t that consistent with what you’ve argued here? You consistently responded to me as if I wasn’t paying attention to what you wrote, which I don’t understand.

    In a quick look at the FBI arrest data, from what I can tell, the risk of physical violence (and yes a fair amount is justified, but we aren’t sure how much) is higher than very serious things like murder and a bit less common than things like aggravated assault. (You’d correctly point out that for the more violent types of crime, the effects are much greater than police violence). The risk of getting manhandled is similar to the risk of burglary/larceny/property crime. It’s not inconsequential. You suggest that it’s irrational because the media devotes resources to police violence in a way that they don’t to traffic injury. But society does devote substantial resources to reducing traffic injury. They don’t devote much media attention because societal resources are already invested. Media attention tends to focus on debates about where resources should be focused, and the attention is most strongly directed toward situations where people are perceived to be the cause of injury, especially if they have political power. We tend to have very intense scrutiny that well exceeds the proportional societal effects for a period, make some progress with the issue and then see it drop out of the public eye for a long time. Conservative media features analysis of specific environmental regulations that affect very few people as well, and those on the left accuse them of hysteria too. It often is a little hysterical, but overstating the case, making some progress, and then losing sight of an issue seems pretty common. The sooner conservatives recognize and agree to eliminate any bias that does exist, the more likely we can move on to some other issue that will get far more attention than it merits.

    I appreciate you giving me the venue to respond, and I appreciate the time you took.

    1. I just want to make a quick point on the NEISS vs. NHAMCS issue and then I will really leave things at that. I still disagree with the rest of what you say, but I have already said my piece about this and, as I noted before, I don’t think anything I could add would make any difference on anyone’s mind at this point.

      First of all, you are right that I read the warning about injury rates by race on the CDC’s website incorrectly, so I’m happy to grant you that. But you are still wrong that the NEISS is a more reliable dataset than the NHAMCS.

      Indeed, despite what you are suggesting, the sample used by the NHAMCS is representative in addition to being much larger. In fact, I have read the methodology for both surveys and it seems clear to me that, on the contrary, the sampling strategy used for the NHAMCS is far more sophisticated. (See the codebook for the dataset for more details on the methodology.) It also includes ambulatory surgery centers which, as far as I can tell, are excluded from the sample used by the NEISS. I have no idea what makes you think the sample used for the NHAMCS isn’t representative, but that is clearly not the case.

      It also seems that most data about ethnicity (i. e. hispanic origin) in the NEISS are imputed, which as far as I know is not the case for the NHAMCS. This is important because, when I use the PPCS, I’m looking at non-hispanic black men. I don’t think this makes a huge difference, but that’s still another thing which makes the NHAMCS more accurate for what I’m doing.

      You also seem to under the impression that, unlike the NEISS, the NHAMCS is not conducted by the US government and is therefore less reliable. But this is also incorrect, for the NHAMCS is conducted by the National Center for Health Statistics, which is part of the CDC. On the whole, even though again this isn’t really essential to my argument, I think it’s very clear that the NHAMCS is a better dataset than the NEISS for what I’m doing here.

Comments are closed.