Fudging Stats On The Dangers Of Getting Killed By Muslim Terrorists: Politics First; Facts, Dead Last
Truly terrific post by ev psych doctoral student Jesse Marczyk at Psychology Today digging into confirmation bias and the dishonest way stats are presented to make an argument that doesn't hold water:
The first article on the chopping block was published on the New York Times website in June of last year. The article is entitled, "Homegrown extremists tied to deadlier toll than Jihadists in U.S. since 9/11," and it attempts to persuade the reader that we, as a nation, are all too worried about the threat Islamic terrorism poses. In other words, American fears of terrorism are wildly out of proportion to the actual threat it presents. This article attempted to highlight the fact that, in terms of the number of bodies, right-wing, anti-government violence was twice as dangerous as Jihadist attacks in the US since 9/11 (48 deaths from non-Muslims; 26 by Jihadists). Since we seem to dedicate more psychological worry to Islam, something was wrong there There are three important parts of that claim to be considered: first, a very important word in that last sentence is "was," as the body count evened out by early December (link is external) in that year (currently at 48 to 45). This updated statistic yields some interesting questions: were those people who feared both types of attacks equally (if they existed) being rational or not on December 1st? Were those who feared right-wing attacks more than Muslim ones suddenly being irrational on the 2nd? The idea these questions are targeting is whether or not fears can only be viewed as proportionate (or rational) with the aid of hindsight. If that's the case, rather than saying that some fears are overblown or irrational, a more accurate statement would be that such fears "have not yet been founded." Unless those fears have a specific cut-off date (e.g., the fear of being killed in a terrorist attack during a given time period), making claims about their validity is something that one cannot do particularly well.
The second important point of the article to consider is that the count begins one day after a Muslim attack that killed over 3,000 people (immediately; that doesn't count those who were injured or later died as a consequence of the events). Accordingly, if that count is set back just slightly, the fear of being killed by a Muslim terrorist attack would be much more statistically founded, at least in a very general sense. This naturally raises the question of why the count starts when it does. The first explanation that comes to mind is that the people doing the counting (and reporting about the counting) are interested in presenting a rather selective and limited view of the facts that support their case. They want to denigrate the viewpoints of their political rivals first, and so they select the information that helps them do that while subtly brushing aside the information that does not.
...Saving the largest for last, the final important point of the article to consider is that it appears to neglect the matter of base rates entirely. The attacks labeled as "right-wing" left a greater absolute number of bodies (at least at the time it was written), but that does not mean we learned right-wing attacks (or individuals) are more dangerous. To see why, we need to consider another question: how many bodies should we have expected? The answer to that question is by no means simple, but we can do a (very) rough calculation. In the US, approximately 42% of the population self-identifies as Republican (our right-wing population), while about 1% identifies as Muslim. If both groups were equally likely to kill others, then we should expect that the right-wing terrorist groups leave 42 bodies for every 1 that the Muslim group do.
He points out more dishonest arguing in a New Yorker piece by Lawrence Krauss:
Lawrence goes on to say, the average Paris resident is about as likely to have been killed in a car accident during any given year than to have been killed during the mass shooting.
...This point about cars is yet another fine example of an author failing to account for base rates. Looking at the raw body count is not enough, as people in Paris likely interact with hundreds (or perhaps even thousands; I don't have any real sense for that number) of cars every day for extended periods of time. By contrast, I would imagine Paris residents interact markedly less frequently with Muslim extremists. Per unit of time spent around cars, they would pose what is likely a much, much lower threat of death than Muslim extremists.