This is a guest post by Peter S. Henne. Peter received his PhD from Georgetown University in May 2013, and was a Fellow at the Miller Center at the University of Virginia during 2012-2013; he is currently a postdoctoral researcher at the National Consortium for the Study of Terrorism and Responses to Terrorism. His research focuses on religion and foreign policy; he has also written on terrorism and religious conflict.
In his latest blog post on Foreign Policy, Stephen Walt calls for a re-evaluation of the United States’ approach to counter-terrorism. One statement–really a quick aside–caught my attention.
Walt claims that “opposition to foreign occupation and interference is one of the prime motivations behind terrorist activities.” Well he actually says: “Given that opposition to foreign occupation” causes terrorism, and then uses this assertion to justify calling for a reduction in US forces in Muslim countries. And then he specifically mentions “suicide bombing,” and links to Dying to Win, by Robert Pape.
Dying to Win is the book version of an article by Pape in the American Political Science Review, in which he argues that suicide bombing is a rational response to occupation. As I detailed in a blog post a few years ago, there are numerous problems with this argument:
- Pape defines “occupation” so broadly as to almost be useless. This includes not just military invasions, but separatist conflicts, the voluntary presence of foreign troops, and transnational identification with occupied peoples. Lumping so many different things together under “occupation” ends up not really telling us much about the outcome we’re trying to explain.
- Pape only looks at cases of suicide bombing, basically selecting on the dependent variable. No-variance research designs can tell us a lot about a phenomenon, but they cannot tell us what caused it.
- While most would agree with Pape that Islam does not cause suicide bombing, it’s hard to ignore the religious elements in attacks of groups like al-Qaeda. Religious ideology frames a struggle in cosmic terms, justifies extreme actions–like suicide bombing–and may even necessitate suicide bombings by constructing them as holy acts of martyrdom. Looking just at whether or not a country is “occupied” ignores these dynamics.
So there are issues with Pape’s thesis, and I think a lot of scholarship on terrorism and political violence has moved on. But why is this book still getting attention?
- The first reason has to do with the lack of satisfying answers to the question of what causes terrorism. The more research that is done into terrorism – and a lot has been done since 9/11 – the farther scholars get from a clear, simple answer to its causes. This is partly because terrorism is a tactic a diverse array of movements use in different contexts, partly because numerous causes combine in different ways to produce terrorism, and partly because some causes matter in certain contexts and others in different ones. Compare the preceding sentence to the one in Walt’s post I quoted above, and you’ll see why Pape’s thesis is tempting. But bloggers should avoid simplifying complex phenomena to score points in the blogosphere.
- The second reason is a bit more meta-: maybe the pleasing nature of Pape’s policy suggestions–that the United States should not invade any more countries–makes some more tolerant of the issues with his argument. It’s hard to prove a non-event, but I imagine bloggers–at least on Foreign Policy–would have been more swayed by critiques of his thesis if its implications were that the United States should nuke all possible adversaries.
So should scholars and pundits still cite Dying to Win, given the issues with this book? Does its value as a competing explanation for articles we write justify its continued use?
And if the answer is yes, what should we do in response to its deployment in the blogosphere? I guess I could keep writing blog posts like this one periodically, which I am happy to do. Or, when it comes to something as important as what the causes of terrorism prescribe for US foreign policy, we should all be a little modest in our claims.
Aren’t there any articles or books that do a good job of rerun the data while adding more granularity to occupation or the like?
I’ll try to keep in mind some of those caveats when discussing it in the future, but really, it’s Pape is going to be talked about until something better comes along. That “something better” might be a study that fails to reproduce the results in a fairly definitive manner. As you say, there may not be a simple overarching framework to use here.
Max Abrahms?
For those interested, I had a piece in IO a few years back that looks at this. It’s mainly about an argument I make about adoption capacity, but it also fixes the methodological problem Pape creates by only looking at those groups that use suicide bombing. The results show that occupation is unrelated to suicide bombing: https://journals.cambridge.org/download.php?file=%2F62837_80224488EB1EC0549A5F7E6E8A93C7ED_journals__INO_INO64_01_S0020818309990233a.pdf&cover=Y&code=c4f11a3a5b79548f0e7fa9d615b280fd
Bottom lines:
1. The decision to use suicide bombing is a tactical choice by groups that has much to do with organizational issues (e.g. younger, more flexible groups are more likely to adopt) and network effects (religious networks)
2. By only focusing on those groups that adopted suicide bombing, Pape obscured the fact that the vast majority of terrorist groups felt like they were occupied and did *not* adopt suicide bombing. It’s a great example of selection on the DV 101 (check this APSR article for more on that: https://www.international.ucla.edu/cms/files/acmr.pdf), because his decision to select on the DV *actually* influences the outcome. Occupation predicts resistance to occupation, not suicide bombing.
All that being said, I think Peter’s post is correct in outlining some of the reasons people continue to cite Pape. And I would encourage everyone to check out Peter’s work on these topics!
The relevant IO is one of my favorite MCH articles. — PM
Neither of the links in this comment seem to work.
This link to the second article should work: https://www.international.ucla.edu/cms/files/acmr.pdf
Can’t help with the first one I’m afraid.
Interesting! Regarding the second point, though, on Pape’s assertion that occupation causes suicide bombing and how this might be an artifact of selection bias:
If Pape were claiming that ONLY occupation -> suicide bombing, then sure, selection bias. But if the point is only to suggest that it (sometimes) does, then that’s a different claim and selection bias doesn’t need to enter the picture… thus making it a case of what you call “telling us a lot about a phenomenon” and showing how one factor (occupation) can lead to an outcome (suicide bombing.)
I think this stands even taking into account MCH’s second bottom line, below. What MCH’s findings suggest, however, is that in the broad spectrum of causes/cases of suicide bombing, occupation seems to be a rather rare trigger.
Of the problems with Pape’s argument, bullet point 1 is legitimately arguable, bullet point 2 is a valid criticism (although GM is also right on this), and bullet point 3 doesn’t seem to stand up at all.
I mean, can we really criticise Pape for lack of subtlety in variable design and then resort to “it’s pretty obvious religion is involved” as a counter-argument? I don’t think Pape does “ignore” these dynamics. Instead, he addresses them directly. Regardless of causation effects, his work is useful at least partly because a fuller empirical picture of suicide bombing shows that it is not as simple as ‘religious groups adopt suicide bombing tactics’. Yes, religion may frame things cosmically to justify extreme sacrifice, but then the combination of grand cause and extreme measures can also be found in nationalism, anti-occupation ideology, fascism, radical egalitarianism and so on.
As for the argument that competing arguments means that a prominent work should no longer be cited, well, it all seems rather uncharitable. I’d be interested in reading strong counter-arguments (by the way MCH’s IO link is broken, at least for me), but why not led that dialogue take place at the usual level (academic articles, opinion pieces in FP or wherever, your own blog posts?).
Full citations for the two links in MCH’s comments would be most helpful. Neither link seems to work.
Yes, we’ve moved on. Mostly due to bullet point 2. https://www.international.ucla.edu/cms/files/acmr.pdf
There are many rival suicide terrorism theories (e.g., Bloom, Lankford, Moghadam, Pedazhur, and others). Few have been empirically tested. We (Mike Findley and I) have tested Bloom’s and find qualified support under certain contexts. I don’t think explanations are the problem, we need to figure out which ones explain the world the best and under what conditions.
Additionally, his theory is about suicide terrorism, not terrorism generally. While it is a deadly form, it is one subtype and far from the majority of attacks. We’ve made some progress in understanding why it happens cross-nationally, where it is likely to happen, how institutions promote or inhibit its use. We have much to do to explain individual reasons for joining groups, using it as a form of contention, etc.
Also, Wade and Reiter (https://www.sagepub.com/martin3study/articles/Wade.pdf) have a paper that fixes some of the research design problems in Pape and reaches different conclusions.
Could somebody write a version of these things that we nonspecialists could write and assign to undergrads? Pape gets a lot of attention because he’s nicely anthologized.
The continued invocation of research design to discredit Pape’s findings I sometimes find frustrating. The argument in APSR pointed out that if Pape were looking for the best correlations with terrorism or the best predictors of terrorism, then a no-variance research design is inappropriate. Everyone agrees with this. Pape’s rejoinder is that he is not interested in that: he is adopting a necessary conditions-based understanding of causation. Most agree that if one is interested in necessary conditions, then generally, variance on the DV is irrelevant.
The necessary-conditions based argument has a number of defenders in political science, most noticeably the QCA folks.
In short, the issue is not research design but the philosophy of causation. If one buys necessary conditions, then the critiques completely miss the mark. I have seen few claims to the contrary, at least yet.
The first of Peter’s arguments is a pretty crushing indictment of Pape’s research design, in that Pape has effectively said that conflicts involving insurgent parties predict suicide bombing. Insofar as this is based on a conditional understanding of causality, it brings to mind some criticisms of Mackie, such as that his INUS model doesn’t distinguish conditions. By this logic, ‘involves homo sapiens’ is as significant a causal condition as ‘occupation’. Necessarily (har) we have some more particular requirements as to what counts as a good explanation.
Pape’s findings imply that there is some sort of tactical advantage to suicide bombings and that this advantage is appealing to militarised insurgent movements. This is valuable for those to whom it might not otherwise be obvious (do such people exist?). But really, ‘suicide bombing is caused by the advantage it brings to the people who use it’ is simply a restatement of the axiom of folk psychology.
I am not really a Pape defender, so I do not want to get into the other details of his argument. This isn’t really what I understood point 1 to be. I thought point 1 was a criticism of the measure of occupation. Pape’s theory, if memory serves, is that the perception of foreign occupation is causally relevant to terrorism. The perception of foreign occupation, however, can mean different things in different societies. Because the measure is ambiguous, its difficult to tell what the measure is for a particular case.
This is a reasonable criticism, especially if you favor observational data. But, really, if his theory is :perception of foreign occupation,” then I have trouble figuring out how observation data will be sufficient to tell what types of cases ex ante will trigger that perception.
I was referring to pt 2, which brings up the INUS arguments you discuss. I realize that some likely disagree that necessary conditions are causes. My argument is that if one believes that the identification of necessary conditions are relevant to causal explanation, then there may be reasonable reasons to select cases on the DV because tests of necessary conditions are different in kind from tests rooted in regressions.
Unfortunately, much of the work that criticizes Pape acknowledges only briefly that he is interested in necessary conditions, and ignores this central argument.
Also, I am not quite persuaded by the argument that foreign occupations may be trivial, like oxygen is to living, for a couple of reasons. The first is I am not yet aware of anyone running the recommended tests for triviality on Pape’s data, that would be interesting to see the results of. https://www.braumoeller.info/wp-content/uploads/2012/12/Methodology-Necessary-Conditions.pdf. Therefore, we do not know if the perception of foreign occupation really is trivial.
The second reason is that a central philosophical complaint against trivial conditions is that they cannot be manipulated (we cannot eliminate oxygen in order to avoid without killing all people) or that they are uninteresting (oxygen explains war, boy, I was looking for something else). But, even if Pape’s conditions are trivial, it would show that some trivial necessary conditions are important because they can be manipulated. If foreign occupation is the problem, well, there is something that one might be able to do about that. Its technical triviality doesn’t seem especially problematic because its politically salient.
Again, my beef though is simply with assuming that all political science works on the same template, and imposes foreign standards on a piece of scholarship.
Fair do’s! I share your frustration with the notion that selecting on the DV is always to be avoided. My main issue with Pape’s argument is that what counts as occupation is so amorphous as to apply to most ‘asymmetrical’ conflicts, meaning that we have not ended up with an interesting finding. I’m not sure I need to apply any additional tests to conclude that this finding is trivial, because I was never in any doubt that suicide bombing was a force-multiplier for weak parties. We only need to think about what suicide bombing involves, tactically, to conclude that it’s going to be appealing mainly to ‘insurgents/terrorists’. None of the interesting questions, on organisational infrastructure, conflict geography, cultural context, access to technicians, etc, get answered by this finding. Even the question of whether there is some special link between Islam and suicide bombing* is not answered here, because there is only one conflict in which combatants systematically engaged in suicide bombing and weren’t Muslims.**
*for the record, I don’t think there is
**There have been a couple Christian suicide bombers in Israel-Palestine, but that’s all
A couple of things to add:
1. Selecting on the DV can be a defensible choice for generating hypotheses, it can be less pernicious when the DV is either truncated, or when we can potentially identify the expected bias involved in this choice. In Pape’s case, none of these conditions hold. It is a no variance (not a truncated DV, all values of the DV are suicide terrorism) design, thus we cannot draw inferences. The formal exposition is in the paper I cited above. So I think folks are right to question rejecting Pape “just because he selected on the DV” but based on his stated purpose and the consequences of his selection, there really is no defense of his position on this point.
2. In Pape’s data as well as other databases, there are plenty of campaigns and attacks that are committed by non-muslims, which is part of what makes his argument appealing to folks who know the data. This is also a potential explanation for the persistence of his argument. We know the LTTE, Marxists in Palestine, the PKK, and other non-religious non-Muslim groups have used suicide terrorism, thus we need a story that that does not rely on the simplistic religion/Islam = suicide terrorism.
Besides those of the LTTE and a couple PFLP bombers of Christian background, as far as I know all suicide bombers whose attacks are part of systematic campaigns have been Muslim. So it seems false to say that ‘plenty of campaigns’ have been committed by non-Muslims. This is what drives that simplistic ‘it’s got something to do with Islam’ argument.
Selecting on the DV can do a lot more than generate hypotheses. It can allow you to examine processes and sequences of mechanisms linking causes to effects. It’s just that this requires something more than associational analyses.
Simon,
I think we are in agreement on point 1. There are places where on can defend selection on the DV. Just not in Pape’s case.
On point 2, as I note and should probably be more clear about, there are many attacks and a few campaigns (a coordinated series of attacks over time) that have involved non-Muslims. PKK in Turkey, LTTE in Sri Lanka, DHKP in Turkey, PFLP (Christian and non-Religious), Barbar Khalsa International(BKI) in India, etc. There was even an attack by a tax protester in the US (Joseph Stack). Also important to note, that individual goals differ widely from group goals. In Iraq, for example, individual reasons for attacks have varied from shame, money, revenge, religion, etc. The group may have a goal while the individual that perpetrates has quite a different one. Reductive to say all or most (with some caveats) of suicide attacks attributed to one religion or religion generally.
I think I’m not being precise enough with my language. By ‘Muslims’ I really mean ‘members of communities which are traditionally adherents of Islam’. This includes the PKK (Kurdish communities are traditionally Sunni, as I understand it) and, I presume, the DHKP. I wasn’t aware of the BKI though. How many suicide bombings did they carry out? If I recall, the PFLP carried out two bombings.*
This is actually another good example of the ‘selecting on the DV’ problem: the fact that most systematic suicide bombing campaigns seem to be carried out by people who come from Muslims communities* doesn’t mean that confessional identity is an important factor. It is not so much reductive to attribute suicide attacks to one religion as it is simply false, I think.
The point about the difference between group goals and individual goals is interesting. I’m not sure I’d phrase it in those terms though. I think I’d sooner phrase it as certain political platforms, around which militant social movements mobilise, give moral cover and technological capacity for individuals to serve more prosaic and parochial personal goals. I don’t think this means that there is a different between personal ‘group’ goals though; I don’t imagine individual suicide bombers would say that group goals are incidental or that they don’t share these goals, but merely that they are not and immediate or overarching impetus for the action. I’m sure you’d know better than me though.
*In contrast to individuals who carry out suicide attacks outside of the systematic frameworks of recruitment, training, and deployment that feature in examples such as the LTTE, which until the past few years accounted for over half of all suicide attacks if I recall correctly?
**Calling the PFLP ‘Christian and non-religious’ seems no different from identifying the PKK as ‘Muslim and non-religious’.
Holy typos batman! My apologies; phone typing is messy business.
Should this post be called a Pape smear?
Pape’s thesis generally lacks empirical support, but the more general notion mentioned by Walt – that foreign “interference” is associated with terrorism – finds support in a number of articles. Most focus on the U.S. involvement abroad. Examples include Sobek and Braithwaite, Savun and Phillips, several articles by Plümper and Neumayer, Choi and Salehyan, and a few others.
Research on interstate alliances and rivalries and their ties to terrorism also gets at this idea, although possibly in a less direct way (Conrad; Findley, Piazza, and Young; Plümper and Neumayer again, among others).
thanks for all the comments everyone. I just wanted to respond to a few.
-I didn’t do my usual list of hyperlinked academic articles, but I guess I should have. Some of the relevant citations from this discussion:
Michael Horowitz, “Nonstate actors and the diffusion of innovations,” International Organization 64.1 (2010)
Max Abrahms, “The Political Effectiveness of Terrorism Revisited,” Comparative Political Studies 45.3 (2010), as well as his 2006 IS piece which has been included in an intro to IR reader
Assaf Moghadam, “Motives for Martyrdom,” International Security 33.3 (2009)
and one of my articles, “The Ancient Fire,” Terrorism and Political Violence 24.1 (2012)
-on the religion point, religious ideologies don’t always lead to suicide terrorism/not all suicide terrorism is by religious groups, but the social movement dynamics involved in mobilizing religion–religious framing, threat attribution, mobilizing structures–can contribute to the use of suicide terrorism. In my TPV piece I found that suicide bombings by religious terrorist groups were more violent than those by non-religious ones, indicating religion has some effect.
-I second MCH’s and Joseph Young’s critique of the research design. I’d go further, though. The necessary-and-sufficient defense felt a bit like a post hoc justification for Pape’s work. Set-theoretic explanations are certainly valid, but it’s not like Pape used QCA, method of sequence elaboration, or any other qualitative method that draws on necessary and sufficient causation. So I still think there are issues.
That’s more reasonable. The claim I was concerned with in the original post was different: “No-variance research designs can tell us a lot about
a phenomenon, but they cannot tell us what caused it.”
I’d also point you to James Jones critique in “Blood that cries out from the earth” as another concern about the characterization / classification of group types.
Apologies that the link was broken to my IO piece. If you email me directly at horom (at) sas.upenn.edu, I’ll email you a copy. I’ll also post one on my website tonight (presuming bluehost comes back online).
On the substance – the point of my article, to be clear, was not a takedown of Pape. That was incidental. Turned out, when you fixed the research design problem, his findings went away.
The point was the argument about how group organizational factors and network effects drive the *diffusion* of suicide bombing.
BTW – as a few people mentioned, Pape’s “you are occupied if you feel occupied” definition was pretty problematic, especially given his dismissal of religion (not in an essentialist way, in a “this could be a variable, among many, that influences how people behave” way). For example, AQ thinks that the U.S. is “occupying” Saudi Arabia in the 1990s so they launch suicide attacks. That’s an example of occupation -> suicide bombing for Pape.
But why does AQ think that the U.S. is “occupying”, in that case? What’s different about it than, say, U.S. military bases in Germany? If you believe AQ, it’s because their Salafist beliefs make them think that the deployment of foreign troops in the cradle of Islam is offensive (now one could say that AQ says one thing and really believes another, but that’s not the point that Pape is making there).
My main critique of Pape’s Dying to Win is an article titled “Suicide Terrorism, Occupation, and the Globalizaiton of Martyrdom: A Critique of ‘Dying to Win,’ published in Studies in Conflict and Terrorism. It can be accessed here for free: https://www.tandfonline.com/doi/pdf/10.1080/10576100600561907