Security Studies Program Seminar

Terrorism, Civil War Outcomes, and Post-War Stability: Hypotheses and (Very) Preliminary Findings

Page Fortna
Associate Professor of Political Science
Columbia University

September 17, 2008


Q. Can you determine when a group began to adopt terrorist tactics? Maybe they only began to adopt terrorism once they knew they were losing.

A. Stanton, who coded these cases, reports that there are only a few cases in which groups turn to terrorism later in their lives (PKK in Turkey). In general, groups tend do not change their original strategies. When they do, it happens when things are getting better, not worse.

Q. Does the 87-case universe worry you?

A. It is the universe of cases — a sample of 87 would be worrisome. Missing data (7-8 cases) appears not to be systematic, ongoing wars aside. In general having a small N should bias against finding anything, so in that sense the positive findings are a strength. It does also mean that particular cases are more likely to drive the results. For instance, the PKK was originally classified as terroristic for its entire duration, and correcting this coding modified the results somewhat.

Q. Are you controlling for the other things these groups are doing? How do we know it is not cleansing, control, or destabilization doing the work?

A. Not yet. That is a follow-up project — looking at the effects of these other strategies on the stability of peace. Maybe groups using these other tactics are nasty ones the government does not trust, further explaining why terrorism itself does not appear to impact peace durability.

Q. You say terrorism does not work, and that groups do not change their strategy. The bigger implication: These groups are not strategic actors in the traditional sense (they do not learn), and are driven more by ideology. That is a strong claim.

A. I am not trying to explain who is using the strategy. Maybe part of the answer is that terrorists use terrorism for reasons other than to gain concessions — to raise money, to outbid other groups, to sustain the organization, etc. Maybe groups think terrorism is effective but it is not. Maybe some are not being strategic, or are being strategic initially but do not switch when it is not working. Maybe they are blocked from using other strategies. It would be interesting to look at the ongoing cases — some of them go on and on, so why are they not adapting? Maybe if they change the strategy, the group would fall apart — so what is good strategically is not good for the leaders.

Q. I am skeptical about the bright line between terrorism and insurgency. In South Asia , every civil war since 1989 has involved attacks on civilians. How was the coding done?

A. It is not either/or — all the groups are insurgent, and some are also terrorist. The categories are not mutually exclusive; you can use terrorism and cleansing. Under Stanton's coding, terrorism is defined as a campaign, not one or two isolated attacks (we want to make sure it is not a rogue part of the group carrying out the attacks). The place where the distinction is the trickiest is between destabilization and coercion; you can attack a railroad to disrupt infrastructure, or to injure/kill passengers, or both. Some of the destabilization cases are those in which there was a deliberate attempt to minimize civilian casualties (for instance, attacking factories at night, when all the workers had left).

Q. Regarding the stability of peace: The similarity between terrorism and non-terrorism is striking. You could switch focus to the conditions for maintaining peace, with an eye toward controllable, policy-relevant conditions. This would avoid relying on a research design that makes a continuous variable (terrorism) dichotomous.

A. My book is on the stability of peace, and whether peacekeeping affects it.

Q. These cases are really interesting, and qualitative studies would help. Going into the texture of how negotiations actually happened, I am not sure how a lot of it fits in with the variables you were focusing on in the quantitative studies.

A. The value of qualitative work is that other variables show up.

Q. There is an obvious endogeneity between terrorism and rebel group strength, which you will never be able to control for. I am concerned that on the scale of interesting to provable, the research design is on the interesting side. It is all cross-sectional — we would like to know differences of differences (how well these groups did over time, versus how well they did in the end).

A. The current measures of group strength are weak. Data is forthcoming on the size of government forces versus rebel group forces. It is hard to say if that is relevant to the cases, since strength is private information. Stanton does not find convincingly that terrorism is a weapon of the weak, using that government forces versus rebel size data.

Q. Within these groups, what is the mix of terrorist activity versus other kinds of insurgency?

A. People used to ask what role terrorism should play in larger insurgent movements. The successful terrorists could also have a good guerilla army. Once you get past the “terrorism works” debate, you want to understand the mix of tactics. That may or may not matter for duration as well.

Q. Is terrorism instrumental, or is it expressive? What are the relative weights of instrumental and expressive goals? You would not limit the context to civil wars if you were more concerned with expression. These results reinforce the expressive story.

A. I am starting from the assumption that terrorism is at least in part instrumental, so a partial explanation for my findings is that it is expressive. I am not sure how you would know. The problem with studying expressive motivations at the transnational level is that it is not clear what the alternative to terrorism is. Passive resistance? Staying home? Stanton has direct evidence of groups saying they are using terrorism to coerce the government. So it is at least partly instrumental, although we do not know the extent.

Q. You say there are zero cases in which terrorism has coincided with rebel victory. The database is from 1989-1999 — so you do not actually have the universe of cases. Is there something about this time period that is different?

A. The only justification for the time range is data availability. The eventual time-range goal of the data collection project is 1816 onward.

Q. Regarding aims: What did different rebel groups say they wanted to do at time 1, and what did they say they had accomplished at time 2? You might find a set of aims that look very different from what you have coded — such as protecting a narcotics business. If you expand cases into post-colonial struggles, you might find that terrorism is working (e.g., Algeria ).

A. Very few people are going to say, “We attacked civilians to protect our drug business.” Some aims will be un-measurable. If they are talking about motivations readily perceived as both criminal and immoral, they are not going to be honest.

Q. The reason you do not think terrorism is effective is a credible commitment problem. But that problem should apply to the other tactics you talk about. Why would that mechanism just make terrorism less effective — why would it not hold across the board?

A. It should, particularly for the destabilization and cleansing tactics (less so for control).

Q. Maybe use “discrimination” (of violence) as an independent variable, then. (Highly non-discriminatory groups should face more significant commitment problems.)

A. The results should be much stronger if I used this variable.


Rapporteur: Nathan Black

back to Wednesday Seminar Series, Fall 2008