Murder and misperception

Been thinking back to this National Review article again after recent tweets. e.g.

I think it's worth noting - as the National Review article does - that the MS-13 alone has had members hit with 207 murder charges betwen 2012 and the time of the article's publication vs. 138 killed in school shooting from 2014 til then1 (or 150 since 1999). i.e. MS-13 - just a single gang - has probably resulted in more deaths per year in recent years than school shooters.

As Gardner's thread observes, if one looks at the data student victimizaion of various sorts over the time period has seen substantial decreases in pretty much all measures. I also saw a number of references to Lisa Gilbert's twitter thread on school shooter trainings and thought this the most revealing:

To me the primary purpose of the drills that are being run is political theatre. I fall more in line with the perspective in The Atlantic's Active-Shooter Drills Are Tragically Misguided. Not only do such drills traumatize children over extremely unlikely events, but it seems that in the case of Parkland the killer may have used knowledge from past drills of how the school would react in order to stage his attack.

That's school shootings but I wonder how a similar dynamic of misperception plays out re: campaigns like Black Lives Matter? How accurate are people's risk perceptions there and are people accounting for the costs of the tactics deployed on the people involved? That's one though that'd come to mind when I'd encountered Black Lives Matter: The Wellbeing Cost of Racial Shootings in the United States a little while back. Haven't found a corresponding paper on this but here's the abstract from the talk:

This paper discusses the impact of racial shootings on Black well-being in the United States between 2008 and 2015. Using data from the Gallup Daily Polls, we first reveal a sudden and persistent drop in life satisfaction recorded by Blacks compared to similar Whites from early 2013. There is no similar effect for Hispanic individuals, and results are robust to the inclusion of state, time effects and household characteristics. To explain this finding, we exploit withinstate variation in public and media awareness of police-race interactions, following the Trayvon Martin case in 2012. Contrary to other groups, internet searches and reported cases of arrest-related deaths have a strong negative well-being effect within the Black community that can explain up to half the Black-White decline after 2013.

i.e. The Black Lives Matter movement seems to have resulted in a sudden, persistent relative drop in life satisfaction for African Americans relative to their peers - with the price paid for this by the African American community not very well accounted for I think. Personally I suspect that one of the big reasons for this effect is media-driven misperceptions of the degree to which American police officers kill African Americans.

The outrage over police shootings had prompted the Washington post to start collecting a database of them and they found a lot more than expected. What do the demographics look like though? Depending on how you account for the unknown and other categories, somewhere between 26-33% of those shot by police were African American.

Were people to be polled asking their estimate of the fraction of those shot by police I'd guess the general answer would be more like 90%. Even someone like John McWhorter who's written some not-particulary-politically-correct books on the topic of race - e.g. Losing the Race: Self-Sabotage in Black America had expressed surprised about this:

The heart of the indignation over these murders is a conviction that racist bias plays a decisive part in these encounters. That has seemed plausible to me, and I have recently challenged those who disagree to present a list of white people killed within the past few years under circumstances similar to those that so enrage us in cases such as what happened to Tamir Rice, John Crawford, Walter Scott, Sam Debose and others. The simple fact is that this list exists ...

Taking a raw figure like 26-33% and figure out how to treat it in public communication is more complicated. e.g.:

  • police shoot a disproportionate number of African Americans relative to their rate in the US population

  • police shootings of African Americans are roughly proportionate to what Gallup reports as their perceived rate in the US population

  • police shootings of African Americans are lower than the population fraction estimated to be African Americans reported by Tressie McMillan Cottom2 as the results of polls in the undergrad sociology classes she teaches.

Basically, as usual, the population is generally extremely bad at estimating the rates of things and, in this case, there's also interplay between rates of police shootings and the fraction of the population that's African American. ... and this hasn't even yet accounted for how this might different for other aspects of interaction between African Americans and the police, for which I'd expect you'd see a larger difference3, or issues like class.

At the moment it seems to me that the media is not unbiased but differently biased relative to population at large, and perceived by the public as such (Fox / Breitbart just being a bigger skew in a different direction than most).

I think it's much easier to convince someone that some stuff in the media is fishy, but the question of what fraction is fishy is different and much more complicated to assess. In general as well, "We overestimate what we worry about" which makes this even worse. It's the sort of thing that over years seems to have enabled Fox News to consolidate a viewer base that's over the years become increasingly out of touch with reality at time and similarly, I think, created fertile ground for someone like Trump.


  1. How to account for the precise number of MS-13 murders I'm not sure. e.g. I'd think multiple people who together committed a crime might be charged for the same murder (which may count as one or multiple charges) which might reduce the number of murders. That said, I'd that there've been more MS-13 murders than times MS-13 has been charged for it which would underestimate the murders. ↩︎

  2. Would consider McMillan Cottom's views fairly conventional as judged by her articles in The Atlantic↩︎

  3. See, e.g., Surprising New Evidence Shows Bias in Police Use of Force but Not in Shootings↩︎