Pages

Saturday, July 16, 2016

Black Lives Matter paradox

January 2017 update: With the Chicago Facebook attack a couple days ago, a number of political and media voices have tried to obfuscate the data and situation to fit their own scheme. It is undoubtedly neither easy nor safe being a Black civilian in America. Even today as many people see their social mobility climb, many Blacks are poor and ill, and fearful of their neighbors. Yet there are more than Blacks living on earth and they don’t have exclusivity on facing human trials; we must therefore rely on the actual mortality statistics to guide all of us from a justly macro perspective. And those probabilities couldn't be more clear, and should be just as acknowledged instead of running away with imaginary conversations about the much greater problem.  Both White and Blacks are many times more likely to be killed by someone in their own race then the other way around.  However when dealing with the other way around, we should note that BOTH Whites and police are MANY TIMES more likely to die from a Black(s), then the other way around.

Given recent distressing killings between Blacks and police -in the past week - it is worth poring into the broad statistics that underlie homicides in this country, by race.  The subject here has become far more emotional of late due to three reasons: (A) a lack of recent official national data covering homicides by race and by the police, (B) activists have somewhat diverted our attention away from the broader statistics and instead focused us on a few edited videos of murders under the premise that this is what all Whites are about, and (C) we have an election season where both sides of the aisle want a voice on this topic as it comports with their "broader" socioeconomic policies.  It is as important to get the truth out there in terms of where the threat lies, as it is to be compassionate to the many thousands of murdered innocent Americans (of all stripes) each year but that happen disproportionately by one race.  It is also worth reminding that the vast majority of initial deplorable headlines, about Blacks killed by police, have never later led to convictions by a balanced jury of peers who saw more concerning evidence than the media showed.  Clearly there are some false positives, though there are far many more false negatives.

Let’s start by dissecting some headlines recently by both Former Mayor Giuliani, and also blogger Nate Silver.  Both have made slightly different points, though both are factual in some sense, one is generally closer aligned to the broad statistics.
  • In the U.S. Blacks are killed at nearly 8x the rates of Whites. (based on this Silver is pointing to Rwanda, though bringing this up in this past week only conflates and misdiagnosis the situation)
Now let’s dig into these deaths, by race:
  • If you are Black, then you were nearly 9x as likely to have been killed by another Black versus by the police.  And there was a very little chance at all that a White civilian killed you. (based on this Giuliani is wrong only in the accuracy of his stated probability, during a conversation about Black Lives Matter, but he is certainly close)
  • If you are instead White, then you were nearly 3x as likely to have been killed by another White versus by either the police or a Black civilian (this too would have supported Giuliani’s point, though he chose to never make it)
  • To drive home the point of where the threat is, something critical should be retold as we will below, that the 8x murder rate for Blacks versus Whites would rise to 8.5x if there were no police murders at all. (this reflects how much more impactful police killings are on White people then they truly have on Blacks)
Now in the past year of data there have been 5,556 homicides of Americans by members of their own race, or by police.  The Washington Post and The Guardian have unofficial estimates, and I averaged their databases together and organizing it the method of Bayes' mathematics.  See the split of these murders, which are now precisely sized proportionally by the number of homicides in each of these four categories below.


We show this same information, charted chart two different ways so that the most crucial storyline can be reinforced.  First we should state that while there are nearly 5.5x as many Whites as Blacks in the U.S., it is disquieting that Blacks still kill their own almost as much as Whites do (using the left chart above: 45% for the left portion versus 55% for the right portion)!  Additionally, within the left-side’s number of Blacks killed, it’s actually a smaller portion are killed by police versus the similar comparison for Whites.  This is despite the fact that Whites are living in lower-crime neighborhoods (and yes they too are under economic duress just as Blacks are) to begin with.  If Whites are actually killing each other at a lower rate, it also must imply those neighborhood have less violent criminals!  For more information, recheck the bullets highlighted above describing Former Mayor Giuliani’s comments.

Now when looking at who is committing these dire murders, it is clearly more often than not being done by “everyday” neighborhood civilians, and not by the police.  We prove this by portioning the same 5,556 murders on the right chart above.  We see that the right portion of that ride-side’s chart shows that civilian murders are nearly equal (47% are Blacks versus 53% are Whites).  Yet the only 1 in 7 annual number of all homicides are due to police (the left portion of the right-side’s chart) and not civilians.  When the police kill, 2/3 of the time it is Whites being killed!  We can ask where is the furor over that, when Whites don’t even account for >2/3 of the same-race civilian killings (as a reminder, it was only 53%)?

To be thorough, it is wistfully accurate that police too are victims of unnecessary murders by American civilians (these shootings are up >50% versus 2015's year-to-date statistic).  This type of murder happens at 8% the level of civilians killed by police, and 1% of the overall 5,556 level of annual killings in total we’re discussing here.  So this may appear to be a small number, but it is certainly not insignificant at all, to those we put to work in law enforcement.  Who’s killing these police?  We dejectedly note that relative to Whites, Black civilians kill police at an astonishing 4-5x (Blacks are the same ones who are killing each other at 8x the national rate and paradoxically expecting law enforcement to stop that, but then shriek when the police do).  The Dallas sniper murders last week (and possibly the Baton Rouge killings today) have not done anyone favors, and numerically it will inexpugnably and instantaneously spike the number of police killed by Blacks, by ~15%!  None of this is to say that we advocate any Blacks being stalked and hunted anywhere, but the broad information we hear in the media doesn't offer proof that this is what is happening nor proof of economic discrimination.  It is also a gloomy reality that Black civilians are also killing Whites at a significantly higher rate than the other way around.

An impressive academic peer-reviewed paper, which I spent considerable time reviewing as the associate editor, was well-referenced earlier this year (for example written up in the Wall Street Journal). It explores threat factors associated with police who shoot at a scene (including critical insights such as Black officers are >3x more likely to shoot versus Whites when both confronted as a team within the same situation).  The author and the renowned American Statistical Association journal were a vital probability and statistics citation for the recently viral Harvard University/National Bureau of Economic Research economics working draft on police aggression (and showing that shooting isn’t even the largest component of police aggression to begin with!)

Here is the topical research article we made free: http://www.tandfonline.com/doi/full/10.1080/2330443X.2015.1129918

To summarize, we have a tapestry of racial homicide information that demonstrates that within the U.S., Black lives are committing crimes at an awfully-high disproportionate rate (again we can isolate the Blacks-on-Blacks homicide numbers above).  Additionally, Black civilians account for a disproportionately higher number of all murders of police, who do care about Black lives and are trying to reduce violent crimes in Black neighborhoods.  And while police kill fewer Blacks than Whites, Black police are >3x likelier to shoot versus White police.  Whites tend to murder anyone at a lower late, and this is despite the glorification in the media of a fraction of incidences that don't represent the righteous struggle for many.

It is clear where the threat is where it comes to violent crimes, and so too why the recent movement against Whites is an awkward paradox.  Whites certainly have their flaws and discriminations and certainly struggling to make ends meet just like everyone else, but we also see that Black lives do matter to Whites.  And all innocent lives should matter.  The question is, why do Black lives (or any innocent lives) not matter to Blacks who disproportionately murder?  This makes the appeal that Whites don’t care about anything, hard to absorb.  

Let’s put this all a little differently.  If we were in a situation in America, where we had an equal proportions of Blacks as Whites, in this hypothetical the number of butchered Americans in the hands of Blacks would burgeon (and astonishingly by police would fall) uncontrollably and our entire society would be left effectively unbalanced.  That would not be a stellar result.

6 comments:

  1. Thank you Salil for your statistical analysis. Apologies for the multiple posts as it was not populating correctly.

    Question on the "Who's killing these police?" = 4/5 are Black civilians. The 2003-2014 FBI data reflects roughly 52% of the alleged perpetrators were white. Can you advise or cite the source for the 80% black civilian claim so I can reconcile the difference? Thank you.

    ReplyDelete
    Replies
    1. Thanks much Kange, and sorry for the delay (busy week before I head down to Washington this weekend to teach)! I meant to write a couple points, and since edited the text above. The number of police shot and killed by civilians is up nearly double in 2016 YTD, versus 2015 YTD. More precisely, the vast majority of that increase are from Blacks. Additionally in general, Blacks are 4-5 times as likely to kill a police officer, versus Whites (also same source as you are using). Not only are these statistics staggering to anyone who notices, but it clearly comports again with what we saw in Kansas City this week. Last, you might love the fact that this article was read over 80,000 times, and has had a hefty >275 social media engagements (yours being one of them!) Please see: https://twitter.com/salilstatistics/status/755498707215273984

      Incidentally someone by the name of Matt also wrote about the Baton Rouge ambush, which happened moments after this article was published. This was another distressing reminder (as noted in the article links) of the underappreciated risks that police and their families face, simply confronting criminals each day for society's benefit!

      Delete
  2. Salil:

    Great post to create debate. A colleague and I were discussing your visualization. How does the distribution of race over the entire US population impact the analysis? To make it more clear what I'm asking, the 2010 US Census estimates whites as 72% of the US population and blacks as 13%. Returning to the 5000+ murders in the US, and the split between race, how does the analysis change when you consider the amount of murders per race over the entire country when also considering the race distribution of the US population? Is that a relevant piece of info to include? More importantly, what is a good visualization for that?

    ReplyDelete
    Replies
    1. Hi Linwood, great to hear from you, and I see your LinkedIn essays are doing marvelous! This is of course a great and complex question; which I hope you will take a shot at trying to disentangle (using Bayesian techniques likely). One thing you can do is size the diagram boxes based on the proportion of Blacks (13%) and Whites (72%), and then show the amount of box filled in based on the amount of relative murders for that box. Or you can smartly invert this, and take the chart I have above, but instead recast the amount of filled in space (black or white) based upon the proportion of their racial population. See my lottery article for an example, using the U.S. map ( http://statisticalideas.blogspot.com/2016/04/a-losers-lottery.html) Also don't forget to see my sequel article if you like (http://statisticalideas.blogspot.com/2016/07/whose-lives-matter.html) and add your thoughts there on other topics that should be considered since there is a lot of room for additional dialogue, from all sides of this topic.

      Delete
  3. Sorry, forgot the US census source: http://www.census.gov/2010census/news/releases/operations/cb11-cn125.html

    ReplyDelete
    Replies
    1. Hey Linwood. On a recent tweet (@salilstatistics) we showed additional sources beyond Census (notably HUD, CDC, FBI, DJS, CEA, etc.) And motivated by the spirit that Black lives do matter -but more critically- ALL innocent lives should matter, in a way they haven't before. This sequel research was also popularly shared on ZeroHedge and worth a look: http://www.zerohedge.com/news/2016-07-30/whose-lives-matter

      Delete