98

The Financial Times did some analysis of Donald Trump's controversial immigration restrictions (which were claimed to be about protecting Americans from terrorism).

They focus on analysing risk from refugees since (my emphasis):

President Donald Trump will indefinitely block Syrian refugees from resettling in the US and will temporarily suspend the entire US refugee programme

They put the refugee issue in context:

Those resettling in the US are interviewed and screened for terrorism and crime links by several US agencies, including the National Counterterrorism Center, the FBI, the Department of Homeland Security and the State Department.

Mr Trump cited terrorism risks as his reason for limiting the number of refugees the US takes. However, since the US refugee programme began in 1975, more than 3.2m refugees have entered the US and only three have carried out a deadly terrorist attack.

The FT analysis includes the chart below:

FT analysis

They claim that Americans are more at risk from death caused by vending machines than death caused by terrorist attacks committed by refugees.

Are their estimates of the risks credible?

Community
  • 1
matt_black
  • 56,186
  • 16
  • 175
  • 373
  • 12
    Please do not attempt answers here. Please do not post blatantly racist comments. Please do not try to explain statistics in 500 chars. Yes the claim should say "incidence" and not "probability". Let's move on. – Sklivvz Feb 03 '17 at 11:12
  • 1
    The premise of the question is somewhat arbitrary and ridiculous, why not compare the number of people killed by refugees to the number of people killed by spatulas, or chocolate chip muffins? – Robert S. Barnes Feb 04 '17 at 17:08
  • @RobertS.Barnes a fair point. Why not make the comparison? It would be an even better way to emphasis just how small the historic risk from refugee terrorists is. – matt_black Feb 05 '17 at 10:36
  • 6
    ´Are Americans more likely to be killed by vending machines than terrorist refugees?´ I'll go out on a limb and say **yes**. Refugees tend to interact much less with vending machines [citation needed], hence they have a lower risk of getting harmed by one. – xDaizu Mar 13 '17 at 15:13
  • @xDaizu very funny. – matt_black Sep 26 '19 at 11:24

3 Answers3

87

Yes those numbers appear to be correct. The Cato instituted published "Terrorism and Immigration A Risk Analysis" in September 2016. From that report:

Chance of Dying in an Attack by a Foreign-Born Terrorist

NPR article on the chance of winning the lottery from 2012 notes that it's more likely to be killed by a vending machine. The odds given : 1 in 112 million.

And that means it's more likely, at least judging from one sort-of-old but widely cited statistic, that you'll be crushed to death by a vending machine as you try to shake loose a stubborn candy bar.

The supposed odds of such a death? About 1 in 112 million.

The link to the source of their data is dead, but it seems likely that 1 in 3.6 billion odds are lower than a whole lot of very unlikely causes of death (struck by lightning while drowning: 1 in 183 million).

ventsyv
  • 7,136
  • 2
  • 27
  • 44
  • 26
    Why don't they count the 1993 World Trade Center bombing under "asylum". Why don't they include the 11 September 2011 killing of 3 people by the Boston Marathon bomber? They also leave out one of the police officers killed by the Marathon bombers after the initial bombing. – DavePhD Jan 31 '17 at 15:58
  • 35
    Likely they used the visa status used to enter the country: the Tsarnaevs entered on a tourist visa, then applied for asylum after the fact (and one of them later became a citizen and so might not be included in the above table anyway, depending on how they considered it). Either way I think it's reasonable to remove those two from the "refugee/asylum" category - given they were here for almost fifteen years, and with no evidence they were bad actors until very near to the event. – Joe Jan 31 '17 at 16:10
  • 27
    I can't find any example of someone dying from a vending machine since 1998. http://darwinawards.com/darwin/darwin2001-25.html And the 1998 death was in Canada. Do people really still die from vending machines, or has the design been changed. Are the statistics limited to the United States? – DavePhD Jan 31 '17 at 16:16
  • 19
    Some context is always nice. You're more likely to die from lightning than a Terrorist Attack at all (let alone form a refugee), or a public shooting. However, also be mindful that this data is being wielded for political story control, and is designed to make one side seem preposterous. – SnakeDoc Jan 31 '17 at 16:18
  • 5
    Please remember the Be Nice policy, and make sure your comments aren't going to be misconstrued. – Oddthinking Feb 01 '17 at 02:00
  • The CATO study appears to ignore country of origin, which is what people are all up in arms about in the past few months. – apnorton Feb 02 '17 at 16:32
  • 2
    Please limits your comments what concerns this post and especially leave any racist quip out of here. Finally, comments are not for "explaining how statistical science works in 500 characters". If there is a statistical mistake in *this answer* point it out and move on. If you think there is a statistical mistake in *the claim* find a reputable reference that says so and post an answer. – Sklivvz Feb 02 '17 at 17:56
  • 2
    Wikipedia: "The Cato Institute is an American **libertarian** think tank" I think that for a good post in skeptics SE you should not make an answer based on a one source with a political agenda. Statistics are easy to manipulate, which can be seen by that many other answers do have found single cases which would prove these statistics wrong. – user3644640 Feb 03 '17 at 11:50
  • @user3644640 While that's certainly true, the Cato paper itself is very well sourced. – ventsyv Feb 03 '17 at 14:55
  • 3
    @ventsyv The point is also how you define a refugee, it is not lying if you do not define a person asking for a refugee status as a refugee. If there is a political agenda, like increasing the number of refugee quotas, disregarding common sense how a normal layman would classify a refugee,n you would like to classify such people as tourists or something else. Asylum~Refugee, for a layman they are mostly the same thing, because they do not need to deal with bureaucracy about such things. – user3644640 Feb 03 '17 at 15:12
  • @user3644640 That's why going by visa category is the most unbiased way to present the data and let people draw their own conclusions. In that regard, the original claim is a bit misleading, but even if you combine asylum and refugees together, you'll still get ~ 1:1.5 billion chance which is still lower than winning the lottery. – ventsyv Feb 03 '17 at 15:36
  • 1
    @Joe How are you coming up with "almost fifteen years, and with no evidence they were bad actors until very near to the event"? Tamerlan Tsarnaev arrived in 2004, punched someone in the face for not being Muslim in 2007, attended a radical mosque by 2008, was arrested for aggravated assault in 2009 and was killing Jews by 2011. https://en.wikipedia.org/wiki/Tamerlan_Tsarnaev – DavePhD Feb 03 '17 at 16:13
  • 2
    @DavePhD: The assault in 2009 was domestic violence (not really a hint for later terrorism). The involvement in the triple homicide in 2011 is alleged and under investigation. Over a span of 12 years that looks more like a criminal law enforcement than an immigration issue to me. At this point I'm more concerned about "why didn't we notice and catch this guy earlier even though his extreme views were well known" than "why did we let him in in the first place"? – David Foerster Feb 04 '17 at 11:44
  • @DavidFoerster I don't think the 2011 killings are still under investigation. Both the perpetrators are already dead so there can't be a trial. The other perpetrator told investigators that they did it, then attacked the investigators, so they killed him. And I'm not saying refugees shouldn't be allowed. – DavePhD Feb 04 '17 at 14:57
  • There's something wrong with these statistics - there were 2996 deaths ( not including over 6000 injuries ) in the 9/11 attacks alone. Also, it would be fair to include attacks made by those who where native born, but radicalized by foreign born residents. One should also be looking at injuries, not just deaths, as many times the number of injured are many multiples of those killed. The whole construction of the question is flawed, and as it's said, trash in, trash out. – Robert S. Barnes Feb 05 '17 at 12:05
  • 1
    @RobertS.Barnes In all fairness, the question specifically asks about being *killed*, so while number of injuries might be an interesting data point, it's not really relevant to the claim discussed in the question. Yes, answers could discuss peripheral claims, but it's not an invalid answer because it *doesn't* discuss peripheral claims. – user Feb 06 '17 at 15:32
  • 1
    WTC 1993 and Boston were included - Cato paper: _"Terrorists who were asylum seekers killed four people in terrorist attacks, three of them in the Boston Marathon bombing on April 15, 2013, carried out by the Tsarnaev brothers. The brothers entered the United States as young children and later became terrorists. Ramzi Yousef, who helped plan the 1993 World Trade Center bombing that killed six people, was the other asylum seeker. Because Yousef planned and carried out those attacks as a member of a six-person team, this report considers him to be responsible for one of the six murders."_ – bain Nov 01 '17 at 22:21
78

According to US Citizenship and Immigration Service webpage Refugees & Asylum, last updated in 2015:

Asylum status is a form of protection available to people who:
•Meet the definition of refugee

In other words, not all refugees are asylees, but all asylees are refugees.

By not including asylees in the definition of "refugees" the Financial Times disregarded attacks such as the Boston Marathon Bombings.

Tamerlan and Dzhokhar Tsarnaev (the Boston Marathon bombers) were refugees, but the FT did not included them.

Ibragim Todashev, who allegedly helped Tamerlan Tsarnaev kill 3 Jewish men on 11 September 2011, was a refugee, but the FT did not include him.

1993 World Trade Center bomber Ramzi Yousef was a refugee, but the FT did not include him.

The FT left out 15 terrorism deaths: 7 from the 1993 World Trade Center bombing, 3 from the 11 September 2011 killings, and 5 from the Boston Bombing an aftermath.

So instead of the 3 deaths that FT counted based on the source Terrorism and Immigration, there were really 18.

The vending machine statistic is based upon data before a 1995 campaign to improve the safety of vending machines. It does not indicate the current odds of being killed by a vending machine. 37 people are known to have died due to vending machine accidents from 1978-1995. Also, there is no indication of how many of the 37 were in the United States. In fact, according to the Guardian article The (mainly) men who have fallen under sway of drinks vending machines about Dr. Michael Q. Cosio who was the main researcher of this issue, in his initial study of vending machine injuries and deaths:

The alarming incidents all happened at American military bases in Germany and Korea.

Also, though the OP labels the graph "per billion people", the data, at least for terrorism, is per billion people per year. On the same basis the odds of winning the mega millions jackpot is much higher than what the OP graph says. 8 people per year with a population of 320,000,000 is 25 per billion people per year, for example.

DavePhD
  • 103,432
  • 24
  • 436
  • 464
  • Removed obsolete comments. – Sklivvz Feb 03 '17 at 14:28
  • 1
    If it's not terrorism, it's not relevant. The claim is very specific in that regard, thus I removed the paragraph that talks about not-terror related murders. – ventsyv Feb 03 '17 at 15:03
  • It's probably worth pointing what the odds are if you combine refugees + asylees. – ventsyv Feb 03 '17 at 15:14
  • 2
    Also the Tsarnaev brothers were in the country for over 10 years before they got radicalized, their asylee status has nothing to do with what they've done. – ventsyv Feb 03 '17 at 15:40
  • 1
    @ventsyv Tamerlan arrived in 2004 and by 2007 he had already punched someone in the face for not being a Muslim. By 2008 he was attending a radical mosque. By 2011 he was killing Jews to mark the 10th anniversary of the 9-11 attacks. – DavePhD Feb 03 '17 at 19:49
  • 4
    @DavePhD Which is basically what ventsyv said. They were homegrown terrorists, not terrorist refugees. They became radicalized _after_ arriving in the U.S., not before. – aroth Feb 04 '17 at 01:56
  • The cited [Cato Institute paper](https://www.cato.org/publications/policy-analysis/terrorism-immigration-risk-analysis) explains they are using the US legal classification in which refugee and asylum seeker are two distinct immigration categories: _"Unlike refugees, asylum seekers must apply in person at the border and are often detained before being granted asylum."_ [US Citizenship and Immigration Service](https://www.uscis.gov/humanitarian/refugees-asylum/refugees) agree with that distinction _"Under United States law, a refugee is someone who: Is located outside of the United States"_. – bain Nov 01 '17 at 22:01
  • 2
    @bain The law is very clear: "the alien may be granted asylum in the discretion of the Attorney General if the Attorney General determines that such alien **is a refugee** within the meaning of section 101(a)(42)(A)" https://www.gpo.gov/fdsys/pkg/STATUTE-94/pdf/STATUTE-94-Pg102.pdf To be granted asylum one must meet the definition of "refugee" and be physically present inside the United States or at a land border. – DavePhD Nov 02 '17 at 13:02
  • Yes, one definition includes meeting the definition of another, but asylum seeker and refugee are still two different immigration processes, one requires being in the US or at a land border, and the other requires being outside the US. [How it works: The U.S. Refugee Resettlement Program](http://www.pewglobal.org/2017/10/12/u-s-resettles-fewer-refugees-even-as-global-number-of-displaced-people-grows/#how-it-works-the-u-s-refugee-resettlement-program): _"All refugees are processed and approved outside of the United States."_ – bain Nov 03 '17 at 07:46
  • 2
    @bain I agree there are two different processes, one ("asylum procedure", section 208) for refugees in or at the land border of the United States, and another (section 207) for refugees outside the United States. Pew is directly contradicting the law, which says "The Attorney General shall establish a procedure for an alien **physically present in the United States** ... if the Attorney General determines that such alien **is a refugee**" – DavePhD Nov 03 '17 at 10:58
  • The quote was from the Cato Institute, which in turn cites the US Citizenship and Immigration Service web site i.e. [this paragraph](https://www.cato.org/publications/policy-analysis/terrorism-immigration-risk-analysis#sec-5.7) links to citation 48 which links to [this page on USCIS](https://www.uscis.gov/humanitarian/refugees-asylum/refugees). If they are wrong, then it is ultimately the US Citizenship and Immigration Service who are contradicting the law when they say _"Under United States law, a refugee is someone who: Is located outside of the United States"_. – bain Nov 06 '17 at 17:49
  • 2
    @bain and contradicting themselves, as they say "Asylum status is a form of protection available to people who: •Meet the definition of refugee •Are already in the United States ..." https://www.uscis.gov/humanitarian/refugees-asylum – DavePhD Nov 07 '17 at 19:22
6

According to this analysis on CrossValidated.SE, we don't really know.

The first problem with this data is non-stationariness. These numbers hold only if the number of refugees and vending machines stay constant. These 3 deaths are not "deaths per 1M of refugees", it's a total. Let 10 times more refugees in, and you'll have to multiply that number by 10.

Taking that chart data to the extreme, I could say that illegal immigrants are 3 times less dangerous than refugees, so why have border control at all?

The second (and IMO more important problem) is statistical significance. 3 deaths over 35 years is simply too little to predict similar deaths in the future.

However, the author acknowledges that the original article's point may still be valid, despite flawed statistical proof.

There are legitimate issues with the chart, but the FT's broader point is correct that terrorism in the U.S. is quite rare. Your chance of being killed by a foreign born terrorist in the United States is close to zero.

Community
  • 1
Dmitry Grigoryev
  • 2,427
  • 12
  • 21
  • 4
    While correct, this is off-topic for Skeptics.SE. On-topic is citations for the numbers. Off-topic is criticizing (or agreeing with) the argument being made. – Brythan Feb 02 '17 at 18:03
  • Pointing out a hidden assumption in a claim does not disprove the claim, or prove it. It just contextualizes it. – Sklivvz Feb 03 '17 at 02:39
  • @Brythan Would a reference to a book/article explaining this kind of statistical fallacy help? Or asking the question on CrossValidated, since I doubt any real researcher would bother to point out fallacies in a newspaper article. – Dmitry Grigoryev Feb 03 '17 at 08:42
  • @DmitryGrigoryev I'm sorry but we consider this kind of answers off topic. You can't simply point out a *perceived* logical fault and theoretically disprove an implied claim. They are making claims about the current incidences. If they change, the old values will be incorrect. This applies in general but it doesn't affect the validity of what they are saying. – Sklivvz Feb 03 '17 at 10:46
  • [See FAQ: What are theoretical answers?](http://meta.skeptics.stackexchange.com/questions/2929/faq-what-are-theoretical-answers) – Sklivvz Feb 03 '17 at 11:08
  • @DmitryGrigoryev This covers 40 years of data, millions of refugees. Since the US admits between 70,000 - 90,000 refugees a year, it will take decades for number of refugees to double, let alone increase 10 fold. – ventsyv Feb 03 '17 at 15:28
  • 2
    @Sklivvz then the question itself must similarly be ruled off topic, as the question is *about* a theoretical model. Without a model, no *conclusion* whatsoever can be drawn from the numbers. – Chris Stratton Feb 04 '17 at 17:59
  • 1
    @ChrisStratton the question is about some incidence numbers, which are measurable facts, and certainly not what we think the implication of showing those numbers is. – Sklivvz Feb 05 '17 at 00:00
  • 3
    No. "Likely" from the very title, and "analysis" and "estimates" from the body are not facts, but rather the conclusions of a *model*, and cannot be determined without considering the *validity* and *applicability* of that model. Your mistake may be common, but it is still a mistake, and severe one. The question is *not* just about historical facts, and it is disingenuous to try to claim that it is when the very wording makes it obviously otherwise. – Chris Stratton Feb 05 '17 at 01:22
  • 1
    @ChrisStratton You make a more important error in delving into the detail. The point of the chart is not to posit a complicated risk model. It is to challenge the widespread public perception that refugees are a major terror risk. The simple historic numbers are a perfectly good way of doing that. – matt_black Feb 06 '17 at 10:24
  • @ventsyv That is true. The main point however is that all it takes is a single refugee to conduct a successful terrorist attack for those 40 years of data to become statistically invalid. – Dmitry Grigoryev Feb 06 '17 at 11:02
  • Yet it in over 40 years it hasn't happened which just goes to show you that the veting and the integration in the wider society works fine. If you look post 9-11 incidents the numbers are even better – ventsyv Feb 06 '17 at 13:42
  • 1
    @ventsyv That's t the whole point: statistics cannot be applied to something that hasn't happened. For example, I lived in several big cities for over 30 years, seen literally hundreds of people each day, and none of them was struck by lightning. How would you estimate the lightning strike probability with this data? – Dmitry Grigoryev Feb 06 '17 at 14:15
  • Of course you didn't. Average odds of getting struck by lightning 1:960000. Over 30 years 1:32000. Did you observe 32000 people 24/7/365 for 30 years?? Oh you had to sleep? I see... – ventsyv Feb 06 '17 at 14:53
  • Not to mention that the risk of being struck by lightning in the city is probably way lower than the average. – ventsyv Feb 06 '17 at 14:54
  • @ventsyv So you agree that my personal data on lightning strikes will not be sufficient unless I observe a statistically significant number of such events? That's the point about refugees: unless they commit enough acts of terror, it's impossible to predict how often they do that. – Dmitry Grigoryev Feb 06 '17 at 15:31