27

As reported in The Guardian, the Australian TV host Sonia Kruger has recently kicked up a storm by claiming that

There is a correlation between the number of people in a country who are Muslim and the number of terrorist attacks

Putting aside whether or not it's even relevant, is she right?

lemon
  • 931
  • 8
  • 13
  • 5
    The moderator doesn't seem to like the answer to this question. Since this can't be answered without responses being deleted, I'm voting to close. – quant Jul 19 '16 at 21:53
  • @quant While I don't agree that GIS's answer should be deleted, I am in agreement with Oddthinking that your answer fails to answer the question. – March Ho Jul 20 '16 at 11:16
  • 1
    For somebody with access to a statistics package, [here is a reasonable measure of terrorist activity](https://en.wikipedia.org/wiki/Global_Terrorism_Index). Number of Muslims in a country should be easy to calculate, so it's just a case of crunching the numbers. – DJClayworth Sep 23 '16 at 12:59
  • 3
    I'd bet this is more tautology than an actual claim. "There is a correlation between the number of Muslims in a country and [Islam related] terrorism." In today's conservative media world, terrorism is practically synonymous with "islam related terrorism". –  Apr 21 '17 at 04:52
  • 2
    @fredsbend I disagree. A correlation between Islamic terrorists and [Islam related] terrorist attacks, on the other hand, would be tautological. – lemon Apr 21 '17 at 08:06
  • 5
    During the terror campaign by the Provisional IRA in the UK people in Ireland would have asked the question: "are we talking about catholic muslims or protestant muslims?" ;-) – matt_black Apr 21 '17 at 11:44
  • 3
    I'm a little mystified by the deletion of the answer by GIS as original research. The answer sought to visualise accessible data and not to *generate* original data. I'd have thought this was a good way to deal with questions like this. – matt_black Apr 21 '17 at 11:51
  • 1
    As demonstrated here: http://www.tylervigen.com/spurious-correlations you can correlate any two variables. The real question is what is the meaning of that? Because Mrs Kruger did not go as far as to draw a conclusion, I think the question is unanswerable. – ventsyv Apr 21 '17 at 17:55
  • 1
    @ventsyv that shows picking two random variables that correlate and look funny together, if someone's trying to show a correlation where there isn't one they usually have to log an axis or screw with their definitions. – daniel Apr 21 '17 at 20:50
  • https://trends.google.com/trends/explore?q=anime,culture – airstrike Apr 23 '17 at 16:11
  • I have an answer, but can's post due to protection. The answer is no, and there are clear statistics that show that as the number of Muslims increased in some countries, the number of terrorist incidents declined. – dont_shog_me_bro Sep 09 '17 at 16:28
  • @ゼーロ You need to earn just 10 more rep to post an answer (e.g. post an answer to an unprotected question here and I'll upvote it ;)). Alternatively you could provide a link to a relevant resource and someone else may turn it into an answer. – lemon Sep 09 '17 at 16:39
  • So that we know more precisely what is being claimed, can someone include in this question what was being claimed by columnist Andrew Bolt? – Andrew Grimm Dec 03 '17 at 10:38

1 Answers1

6

A study by Fish et al (2010) concludes:

Relying on cross-national analysis, the authors find no evidence of a correlation between the proportion of a country’s population that is made up of Muslims and deaths in episodes of large-scale political violence in the postwar period.

Caveats:

  • Data is from 1946-2007, so is comprehensive, but over 10 years old now.
  • Only considers "large-scale acts of domestic political violence" (ie, YMMV).
  • Just one study, though the author does have a book out (2011) reviewing more research on the subject.

On the other hand, it looks at both total number of acts and people killed, predominantly Muslim countries vs proportion of Muslims in countries, attempts to control for socioeconomic factors, outliers, ambiguity in the data, and other factors. Using a variety of different models and criteria to see if any of them affect results, the conclusions are pretty much the same.

Arnon Weinberg
  • 842
  • 8
  • 12
  • "Control for socioeconomic factors" are dubious. Perhaps being islamic lead to certain socio economic factors. Perhaps, for example, being islamic means the state is not secular, and that leads to more violent. If you control based on "secularism" then of course you see no data. –  Dec 22 '18 at 21:35