“The War Will Come to Your Street”: Explaining Geographic Variation in Terrorism by Rebel Groups

ABSTRACT

Geographic variation in rebels’ use of terrorism is not well understood. This article explains the use of terrorism in civil conflict through examining geographic variation in terrorist attacks across first–level administrative regions. Two explanations are tested using data on 47 groups in 21 countries: that terrorism is intended to punish supporters of counterinsurgency efforts or to destabilize regions of the country that are both outside of rebels’ military reach and have substantial grievances against the regime. Results show that terrorism is most prevalent in national capitals and regions that are more deprived. The findings suggest that rebel groups face multiple incentives for violence beyond zones of direct military confrontation with the government, using both highly visible attacks against the center of power and attacks intended to geographically expand the rebellion. The findings imply maximizing public service provision and minimizing economic inequality may reduce the breadth of rebels’ potential expansion.

KEYWORDS:

• Geography
• inequality
• insurgency
• terrorism

Acknowledgments

The author wishes to thank James Forest, David Lindsey, R. Blake McMahon, and two anonymous reviewers for their helpful comments.

Supplemental Material

Supplemental data for this article can be accessed on the publisher’s website.

Notes

¹ Due to functional restrictions in operationalizing terrorism, the definition is narrowed to exclude violence against the military and police in the empirical section of the article.

² Targeted violence that, though seemingly arbitrary, is conducted against specific individuals within an occupied population (Kalyvas 1999).

³ However, rebel use of terrorism seems to also make negotiated settlement less likely (Fortna 2015).

⁴ Punishment also differs from Kydd and Walter’s “intimidation” strategy, which targets potential rebel constituents as a way of punishing disobedience within the rebellion’s base constituencies.

⁵ Stanton’s (2013) “control” strategy, while similar, is not contingent on government repression.

⁶ Included areas may be the most opportune for conveying a message and excluded places may be poorly defended, among many potential reasons.

⁷ Participating in an armed conflict against the government that results in at least 25 battle deaths in a given year.

⁸ The GTD classifies attacks as “intentional…[entailing] a level or threat of violence…carried out by sub–national actors” and two of the three following criteria: “aimed at attaining a political, economic,religious or social goal,” “evidence of an intention to coerce, intimidate, or convey some other message to a larger audience (or audiences) than the immediate victims” or “outside the context of legitimate warfare activities.”

⁹ Groups that do not commit any terrorist attacks would drop out of group–level fixed–effects models altogether.

¹⁰ Some groups were collapsed when the GTD did not assign group names to some perpetrators and employed ethnic or ideological markers. For example, the Communist Party of India — Maoist, were often labeled as ‘‘Maoists,’’ the Chechen Republic of Ichkeria was assigned the term “Chechens’’ or the various groups associated with an independent Sikh state of Khalistan were designated “Sikh Extremists.’’ Accordingly, some like–minded groups were merged into “groupings.”

¹¹ Attacks by rebel groups in countries other than those where they began their rebellion, such as the Lord’s Resistance Army’s (LRA) attacks in Sudan and the Central African Republic were considered international terrorist attacks and were not included. Groups that were in countries/territories where government control was so weak that it was not possible to discern between rebels and the state, such as al–Shabab in Somalia, were not included. Hamas and the Palestinian Liberation Organization (PLO) in Palestine, which both targeted an occupying force and were, either in whole or in part, the government of the occupied territories, were also excluded as they did not clearly fit the mold of other rebel groups using terrorism groups.

¹² Data on attacks in 1993, which are missing from the GTD at the sub–national level, were imputed.

¹³ While the GTD also provides latitudes and longitudes for attacks, these coordinates proved somewhat unreliable when compared to the cities listed for each attack — many were often in the geographic center of the country regardless of the stated location — and the regional coding technique was utilized instead.

¹⁴ Composites are created from daily images of outdoor lights, fires and gas flares between 20:00 and 21:30 local time over a calendar year, dropping images obstructed by cloud cover or solar glare and one–time events, such as fires.

¹⁵ Images from military satellites began to be digitized in 1992.

¹⁶ A discussion of why the square root of population density is preferable and a comparison of the measure to one without square roots of mean densities is presented in the Supplemental Appendix.

¹⁷ Geo–positioned gridded (at 1 square kilometer) population density data was obtained from the Gridded Population of the World (GPW) data–set from the Socioeconomic Data and Application Center (SEDAC) at Columbia University (University Center for International Earth Science Information Network 2011).

¹⁸ Although Kuhn and Weidmann (2015) find between a 0.57 and 0.42 correlation between income and electrification inequality, suggesting a large amount of unexplained variation.

¹⁹ The typology is similar to Carter (2016), but varies in that attacks against only military targets, rather than any government institution, were considered to be insurgent tactics.

²⁰ There is divergent evidence as to whether the presence of competing groups will deter attacks by a particular group (Findley and Young 2012a) or increase them (Bloom 2004), thus the effect is worth accounting for.

²¹ Taken from the the Global Digital Elevation Model, which captures elevation values along a 30 square kilometer resolution global grid (Survey 2006).

²² In line with the perceived interaction between changes in patterns of violence against civilians across the span of a civil war (Raleigh and Choi 2017).