On Racial Bias in Police Shootings

Written by ideasanddata.wordpress.com on June 10, 2020. Posted in Current News

After the George Floyd killing and subsequent turmoil, there’s an awful lot being written about black people being unfairly killed by police. But what are the facts?

Such writings are normally always in response to a particular incident of this supposedly occurring. Of course, these are merely anecdotes and in a nation with 300 million people you can come up with a new anecdote every month for something that basically never happens.

You can also create a general impression of racial bias if stories about white people being killed by police are less sensationalized than stories featuring black people.

If we turn from anecdotes to data, this narrative very quickly falls apart. Numerous organizations provide estimates of the rates at which black people are killed by police.  Generally, these estimates are not too far of from each other.

For instance, The CDC says that around 27{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of people killed by police are black. A sociologist used data from Killedbypolice.net to argue that number should be 30{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117}. The FBI puts it at 32{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117}.

By contrast, Black Americans account for 13{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of the total population, 38{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} violent criminals, and 53{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of murders.

FBI (2017)

Black people also account for 40{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of those who murder police officers, and so probably instigate around 40{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of potentially lethal confrontations with the police (FBI, 2014).

Thus, black people are underrepresented among those killed by police relative to their representation among those who commit violent crimes, who commit murder, and who kill police officers.

Cesario et al. (2018) carried out a more thorough analysis of this using multiple sources for crime rates, including using estimates from victimization surveys. The paper also distinguished between everyone killed by police and those who were killed by police while unarmed and not aggressing. For the majority of estimates, white people are over-represented among such killings. In nearly all cases there was no evidence for significant anti-black bias.

This data is consistent with the implications of simulation experiments. These studies put police officers into as realistic of simulations as possible, give them a fake gun, and then see if police are more likely to shoot an unarmed black than and unarmed white. Each officer goes through a large number of simulations, over a several week period, which differ in many ways and are not told that this experiment is being used to analyze racial bias.

James et al. (2012) carried out such an experiment and found that police were 25 times more likely to shoot an unarmed person if they were white rather than black.

Similarly, James (2016) found that police officers took 1.09 seconds to shoot an armed and aggressive white suspect and 1.32 second to shoot a similar black suspect.

They are found that 14{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of non-aggressing white suspects were shot compared to 1{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of non-aggressing black suspects.

Thus, comparing police shooting rates with simulation experiments and crime benchmarks imply a pro-black or anti-white bias in police shootings.

A total lack of anti-black bias is also implied by research which compares the behavior of black and white police officers. For instance, Meinfeld et al. (2018) finds that black people account for 33{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of those killed by non-white police officers compared to only 28{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of those killed by white police officers.

Crotty et al. (2017) find a non-linear relationship between the proportion of an area’s police that are black and the rate at which police kill black people. However, there was no evidence that police forces which were almost entirely black had significantly lower rates of police killing black people than did police forces that were almost entirely white.

Finally, Johnson et al. (2019) analyzed data on those fatally shot by police in the year 2015. As shown in table 2, the race of police involved in fatal shootings is unrelated to the probability that the person being shot is black or Hispanic.

Thus, a straightforward reading of the empirical evidence does not support the idea that racial bias drives the rates at which black people are killed by police. Generally speaking, the left refuses to engage with empirical evidence on this topic, preferring to talk exclusively about anecdotes. However, a few attempts have been made to justify the left wing perspective using social science and it is worth explaining why these studies fail to do so.

On Ross (2015)

Ross (2015) is probably the study leftists are most likely to cite to show that their viewpoint is valid.

Ross looks at county level trends in police shooting unarmed suspects. He finds that, in the median county, black people are 3.49 times are likely than white people to be killed by police while unarmed. Moreover, this result is unchanged by controlling for crime rates.

So, why is the county level picture so different from the national level one? Probably because there are serious problems with the Ross paper with respect to the measures used, the way in which data were analyzed, and the conclusions that the analysis produced.

Let’s start with measures. First, it’s worth noting that Ross chose to count everyone who was shot by police, regardless of whether they died. So this is a study on police shootings, not police caused deaths.

To count those shot by police, Ross utilized a crowd sourced website. That is, his databases included police shootings that users of this website happened to log. There is no good reason to assume that this data will be accurate, especially at a small level of aggregation such as counties. Ross could have used government data to do this, but he didn’t because he thinks such data would be biased, a proposition for which he provided no evidence.

To measure violent crime, Ross used assault and weapon related arrest rates. This is a seemingly poor proxy since people can be arrested for weapon related incidents that don’t involve violence (e.g. having a stolen gun), and because black people are more over-represented among violent crimes like murder than they are assault. There is no obvious reason to not include a more traditional set of crimes to measure violent crime.

Then there’s the way in which the analysis was done. Ross calculated the probability of black people, white people, and Hispanic people, getting shot by police while unarmed in each US county. He then compared these to calculate black people’s relative odds of being shot, and then looked at the median result.

This initially didn’t work because in most US counties unarmed people of any race are never shot by police and so the median relative probabilities were zero divided by zero. So, Ross through out all the counties in which no one was killed and took the median of the remaining results.

Of course, this still did not solve the problem. Ross is dividing the probability of unarmed blacks being shot by police by the probability of unarmed whites being shot by police. Even in many counties where an unarmed black was shot, no unarmed white is, and so the denominator of the fraction will still be zero. To avoid this, Ross changes the zeros to be closer to the average figure. This process, to some extent, divorces the analysis from reality, and could have any number of unintended effects.

Taking the median also seems like a problematic procedure since this figure is not weighted by the actual number of people killed in each county. It is possible that a few counties exist in which many people were killed by police and in which there is little racial differences in the probability of being shot. While I do not know that this is so, such counties could mean that, if we looked at all the data at once, the actual odds of being shot for blacks relative to whites would be very different than the median result.

Then there are the results of the study. The head line result is that black people are more likely to be shot than white people even after controlling for differences in crime rates. The most interesting aspect of this result is this: “there is no consistent relationship between the race-specific crime proxies (neither assault-related nor weapons-related arrest rates) and racial bias in police shootings”. This result is so absurd that I feel we can safely discard this analysis as being totally invalid, probably due to one or more of the problems I have already specified.

To see why this result is ridiculous, consider that shootings by police can be divided into two categories: just shootings and unjust shootings. Even if you think that police engage in the latter an awful lot of the time, and even if you think that Blacks are disproportionately the victims of it, shooting as a response to violent crime obviously occurs as well. Given this, places with more violent crime on the part of Blacks should obviously have more police killings. The only way for this to not be true is if police officers only shoot Blacks for unjust reasons and don’t shoot Blacks when they are actually engaged in dangerous violent crime. This is roughly as implausible a result as is imaginable and seriously undermines the entire analysis.

The finding that police shootings are unrelated to crime rates has also been contradicted by later research. As I noted earlier, Johnson et al. (2019) analyzed data on those fatally shot by police in the year 2015 and found that black and white cops are equally likely to shoot a black citizen. The paper also showed that the proportion of local crime committed by blacks or Hispanics strongly impacted the probability that a person shot by police in that region would be black or Hispanic. This analysis is to be preferred as it used a more conventional measure of violent crime, and because it counted police shootings by combining several existing databases and then directly contacting hundreds of police departments.

For all these reasons, the Ross paper is not good evidence for racial bias.

Correlating Racism and Police Shootings

Another empirical strategy employed by the left is to try to show that local levels of racism predict local rates at which black people are shot by police. Hehman et al. (2017)  is the best example of this literature.

Herman et al. analyzed data on a massive sample of 2,156,053 Americans divided into 196 regional areas. They measured local rates of crime, income, population density, and education, as well as implicit and explicit racial bias by race. Of these variables, the only one significantly related to the degree to which black people are disproportionate killed by police was how white people scored on the implicit bias test. Even this association was barely significant (p=.031), implying that it is not very reliable.

The scatter plot of the relationship inspires even less confidence, as it appears that the regression line is not mapping onto any real linear relationship. 

The same paper carried out another analysis this time adding a measure of the degree to which whites in the area associated black people with guns. This subset of the implicit association test was more significant than the entire test (p=.001) and once it was controlled for the rest of the test no longer predicted the rate at which black people were shot by police.

However, the scatter plot once again makes it hard to believe that there is any real linear association here.

And unfortunately, the paper analyzed income, education, and bias, by race, but did not analyze crime rates by race. Because of this, we cannot rule out the possibility that areas in which there are larger racial gaps in criminal behavior cause white people to form locally accurate stereotypes and so more strongly associate black people with crime, thus producing more biased scores on tests of implicit association.

Another problem with this analysis is the very use of implicit association tests, which have questionable validity for predicting behavior. We can test whether implicit biases manifest in behavior by examining the correlation between IAT score and how people act in formal experimental tests of discrimination that measure the degree to which they treat black people differently than white people when other differences between people are equalized. Carlsson and Agerstrom (2016) have meta-analyzed this data, producing a statistically insignificant correlation of .03. In other words, IAT scores have no relation with racist behavior.

For these reasons, studies linking local levels of racism to police violence are not convincing, especially when paired against the various lines of evidence suggesting that black people are not over-represented in unjust police killings.

On Propaganda

Since black people are under-represented among those killed by police relative to their relevant crime rates, and they are less likely to be shot by police in simulation experiments, and police of all races are just as likely to shoot a black person, the empirical evidence really does suggest that there is no anti-black bias in police shootings. Given this, it is noteworthy that so many people believe in the left’s narrative. This fact is explained by a propaganda campaign carried out by the media. The tactics involved in this propaganda are worth spelling out.

First, the media makes it seem like unarmed black men are shot by police so frequently that it is rational for black people to live in fear of this happening to them. In fact the weather, accidentally falling down, accidentally drowning, cocaine, intraracial homicide, and HIV, are more common causes of death for African Americans than is being killed by a cop while unarmed.

Yet no one thinks African Americans should fear these things to the degree that they are encouraged to fear police. These shootings are very rare and even if they were due to bias it would be absurd for them to receive the amount of attention that they do.

Ignoring White Victims

Of course, these things don’t only happen to black people, and another consistent pattern in media propaganda is the ignoring of whites who are the victims of crime or shot by police.

The fact is that more whites than blacks are shot by police each year. It’s true that black people are over-represented among such shootings relative to their population size, but if media coverage reflected this over-representation accurately the majority of big profile stories still would not be about blacks.

And this lack of coverage isn’t for lack of stories. For instance, the media coverage of the Mike Brown shooting might have been matched by the police shooting of Daniel Shaver, an unarmed white man killed, on video, by police because he moved his hands towards his waistband after begging the police not to kill him.

Similarly, the case of Philando Castile could have been matched with that of Justice Diamond. In 2016, Diamond, a white woman, called the police to report that she heard a female nearby screaming and thought she might be in trouble. When Diamond came outside to talk to the officers she had called, she “startled” police officer Mohamed Noor, a black male, and so he fatally shot her as she approached the vehicle.

But in reality nearly all big profile stories about police shootings are about black men being shot by police. Whites shot by police are relatively ignored. So are the whites who are killed by black citizens.

In total, police kill around 215 black people per year. This includes black people who were attacking police officers. By contrast, between 1976 and 2005, on average each year there were 981 black-on-white murders and 375 white-on-black murders.

BJS (P. 68)

These numbers imply that for every ten black people that are killed by police, there are 45 white people killed by black people. These murders almost never turn into national stories.

It’s worth noting that this same pattern in evidence in how we are taught history. Roughly 3,500 blacks were lynched in the United States between 1882 and 1968. That’s a rate of 40 per year. Obviously, 40 lynchings per year is 40 too many. But a rate of 40 lynchings per year suggests that for each black person lynched something like 10 – 25 whites were murdered by black citizens, and it is not at all obvious why a black person being lynched is worse than 10 – 25 white people being murdered. Yet that is the value judgement implicit in how we are taught to think about American history.

The systematic ignoring of white victims has given people a false impression about who is and who is not killed both by police and by their fellow citizens both in the present and in the past.

Assuming Racial Motives

Another tendency of this propaganda is to assume racial motives when there is literally no evidence that such a motive exists. When stories of police officers mistreating blacks go viral it is often assumed that the police are mistreating these people because they are black. There is almost never any evidence at all that this is true.

Since police mistreat people of all races, the more parsimonious explanation for police mistreating black people is that there is a general motivation which can account for police mistreating both blacks and whites, and we should not assume otherwise unless we have good reason to do so. This is all obvious, but never adhered to.

Deceptive Editing

False impressions are also instilled in viewers when the media alters evidence to make it fit the establishment narrative.

For instance, in 1992, the public were repeatedly shown the last few minutes of the video of the Rodney King beating. King had been driving while intoxicated and attempted to outrun the police when they tried to pull him over. Once pulled over, King repeatedly resisted arrest, physically removing from himself the officers that were trying to subdue him. It was only after a prolonged period of several attempts to subdue him that the police beat him so that he would stop getting back up and it is only this action that is shown in the viral clip that spawned the LA riots.

Even more deception occurred involving the killing of Trayvon Martin. This famous shooting occurred in 2012. George Zimmerman was on the lookout for people who had been committing crimes in his neighborhood and so followed Trayvon Martin, a 17 year old black male, as he traveled through the area. Zimmerman called the police and shortly after the phone call ended a fight broke out between Zimmerman and Martin. The fight ended when Zimmerman shot Martin, ending Martin’s life. Zimmerman claims that Trayvon assaulted him and that he shot him because he feared for his life once he was forced onto the ground and Martin was bashing his head in.

Following this incident, the media engaged in a smear campaign against Zimmerman which was so dishonest it’s almost comical. When Zimmerman’s phone call to police was released, CNN falsely accused Zimmerman of calling Martin a “coon”, something which CNN later admitted he did not do. Even more ridiculously, NBC released an edited version of the call in which Zimmerman appears to start talking about Trayvon’s race unprompted when in reality the police had asked for his race and NBC simply removed this portion of the audio. And ABC News initially lied about the video of Zimmerman after the incident, claiming that there were no visible injuries on his head, an assertion they later retracted.

These are two examples of a more general pattern: the news systematically lies to people about these incidents in an effort to make black people appear as victims and white people appear as evil wrong doers.

Bearing False Witness

Yet another common feature of this propaganda concerns the tendency of people to lie about police shootings they witness.

Much of the controversy surrounding Taryvon Martin’s death stemmed from people who lived near by, and a friend who Trayvon was on the phone with, initially claiming that Trayvon did not start the fight. All these people later admitted that they didn’t actually witness the incident and so could not possibly know who started the altercation.

We saw the same thing with the Mike Brown case, where several witnesses later admitting to simply making things up.

Assuming Black People Can’t Fight

Another consistent feature of this propaganda is to pretend that black men don’t know how to fight. This is the implicit assumption when leftists argue that shooting a black male cannot be justified unless that black male also has a weapon. This is obviously stupid, but it comes up enough to be worth mentioning.

Pretending That Crime is Unrelated to Character (and Past Crime)

The left also has a tendency to pretend that facts about the people involved in an altercation with police have no bearing on the probability that the relevant person started the altercation. This is obviously wrong because different sorts of people differ in the probability that they would commit different sorts of crime. Generally violent and criminal people are more likely to attack police officers than are peaceful law abiding people and so such information is relevant to determining how probable it is that someone attacked a police officer given any set of evidence. The left pretends to not understand this but, at the same time, they often seek to make every black male who is shot by police appear to be an upstanding citizen who was about to get their life together and who is best represented by a picture of them when they were small children.

The Consequences of Propaganda

This propaganda matters for many reasons, one of which is that it inspires anti-white violence and racism among blacks. Since the civil rights era, there have been over 20 violent race riots in the United States inspired by establishment narratives of black victim hood. In these riots, hundreds have died and thousands have been injured.

Setting group action like riots aside, narratives of black victim-hood have also inspired lone murders. Considering just the years 2016 – 2017,  In 2016 Micah Xavier murdered 5 Dallas police officers and injured 11 others in protest of police shootings of black men. He was explicit about his racial motive, stating that he was looking to kill white people. That same year, Gavin Long murdered three Baton Rouge police officers and injured three others, also in protest of the supposed racism of police. The next year, Emanuel Samson opened fire on a white church in Antioch, TN, killing one person and injuring 7 others. This was done in retaliation for the 2015 Charleston church shooting. Also in 2017, Kori Muhammad killed four whites in Fresno, CA, after being inspired by the shooting of Dallas police officers, and his general black supremacist ideology.

The names of Trayvon Martin, Mike Brown, and Rodney King, are all well known. Yet nearly no one knows the names of any of the hundreds of people killed in the name of the establishment’s racial narrative. One set of victims is made famous while another is not, and that tells us something about the biases of our society.

Today, most black people support Black Lives Matter, a movement associated with several deadly race riots and several anti-white mass murders.

Pew (2016)

To state the obvious, whites who support movements centered on white interests, and addressing perceived injustices committed against white people, are considered racists even when those movements are not associated with as large of a death count as is Black Lives Matter.

Back in 1992, 92{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of black people thought that the police in the Rodney King case should have been found guilty, and 1 in 4 blacks said yes when asked if the violence of the LA riots was justified. Perhaps surprisingly, 73{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of whites also favored a guilty verdict, and one in five whites also thought the LA riots were justified (Gallup, 2016). Recall that the LA riots involved dozens of people being murdered. It is disturbing to think that as many as one in four blacks, and one in five whites, thought that this was justified.

Black Americans also have an astounding propensity to deny the obvious guilt of black criminals. It wasn’t until the 2010s that most Black Americans came to admit that OJ Simpson was guilty of murdering his wife, a conclusion as obvious as any could be.

Bialik (2016)

Most Blacks also wanted George Zimmerman to be found guilty in the Tayvon Martin case despite there being literally no evidence suggesting that he was guilty.

Pew (2013)

The same was true in the case of Mike Brown, an incident even more obvious than Martin’s. Eric Gardner’s case was less clear, and this is reflected by the fact that most white Americans did not think that it was right that the policeman who killed Garner was found not-guilty.

Pew (2014)

More generally,  meta-analyses of the degree to which people exhibit an in-group bias when deciding if a suspect is guilty find that whites have nearly no bias in such decisions while the black people exhibit an in-group bias that is 15 times larger than the minuscule bias seen among whites.

Mitchell et al. (2005)

Conclusion

In conclusion, the relevant empirical evidence does not justify the view that there is an anti-black bias in police shootings.

Concerning the consequences of the narrative the left has spread, it is unlikely that in-group bias among blacks is entirely explained by the narratives pushed by the left, but it also seems unlikely that such narratives don’t inflate black people’s pre-existing biases. This narrative has also inflamed anti-white biases among white liberals. Ultimately, all of this serves to make peaceful coexistence between white and black people more difficult.

Read more at ideasanddata.wordpress.com

PRINCIPIA SCIENTIFIC INTERNATIONAL, legally registered in the UK as a company incorporated for charitable purposes. Head Office: 27 Old Gloucester Street, London WC1N 3AX. 

Please DONATE TODAY To Help Our Non-Profit Mission To Defend The Scientific Method.

Trackback from your site.