Attacks on Civilians in Civil War: Targeting the Achilles Heel of Democratic Governments

Abstract

Previous research has indicated that democracy decreases the risk of armed conflict, while increasing the likelihood of terrorist attacks, but we know little about the effect of democracy on violence against civilians in ongoing civil conflicts. This study seeks to fill this empirical gap in the research on democracy and political violence, by examining all rebel groups involved in an armed conflict 1989–2004. Using different measures of democracy, the results demonstrate that rebels target more civilians when facing a democratic (or semi-democratic) government. Democracies are perceived as particularly vulnerable to attacks on the population, since civilians can hold the government accountable for failures to provide security, and this provides incentives for rebels to target civilians. At the same time, the openness of democratic societies provides opportunities for carrying out violent attacks. Thus, the strength of democracy—its accountability and openness—can become an Achilles heel during an internal armed conflict.

Keywords:

• civil war
• democracy
• violence against civilians

Acknowledgments

Several people have provided helpful comments on this paper; I would like to direct a special thanks to Hanne Fjelde, Scott Gates, Mats Hammarström, Håvard Hegre, Desirée Nilsson, Magnus Öberg, and the anonymous reviewers for their valuable input. Data and replication files are available at http://dvn.iq.harvard.edu/dvn/dv/internationalinteractions. All questions should be directed to the author.

Notes

¹The term civilians here refers to noncombatants (excluding unarmed official representatives for either of the warring parties).

² Bohara, Mitchell, and Nepal (2006) examine the effect of various aspects of democracy on insurgent political violence in Nepal, but they include both combatant and civilian deaths in their measure of violence. There are a few studies of violence against civilians that also include democracy as an independent variable (for example, Eck and Hultman 2007; Wood 2010). However, those are not modeled specifically to evaluate the effect of democracy on rebel behavior.

³Terrorism is usually defined as politically-motivated violence aimed at spreading fear among a larger audience and influencing a psychological target (for example, Hoffman 2006; Wilkinson 1974). It is common to think about terrorism as a tactic that can be used both within an insurgency and as an alternative to one. Moreover, terrorism is also often defined by the means of violence used: for example, suicide bombing often counts as terrorism regardless of whether the attack is directed against a military base or a civilian restaurant (see, for example, Pape 2005).

⁴Whether violence against civilians actually pays, is a slightly different question. Pape (2005) argues that suicide terrorism is increasing because the terrorist groups learn that governments give in to such tactics. Abrahms (2006) challenges these claims and contends that terrorism actually does not work as a coercive strategy. However, the efficiency of violence is not dealt with here.

⁵This finding seems to be quite consistent within studies on intrastate relations. In interstate wars, however, there is evidence that democratic states might even be more likely to engage in mass killings than nondemocracies (Downes 2007).

⁶This means that other groups that may potentially also target civilians, such as terrorist or criminal groups not involved in an armed conflict with the government, are not included in the analysis.

⁷A few actors have been redefined to match the actors in the OSV dataset used for the dependent variable, for example, rebel groups in Burundi have been collapsed into Hutu rebels for some years.

⁸Due to the low threshold for battle-deaths, many actors in low-intensity conflicts are included and these tend to vary in activity over the years. The strategic behavior of such actors is of interest also during the inactive years, especially since those actors could choose to target civilians instead of fighting. Therefore, groups are included for up to three years of inactivity. Moreover, to remove instances of coup d'états, groups have to be active for at least two consecutive years to be included.

⁹Note that the original dataset on OSV also includes data on other actors than those included in my dyadic dataset (such as governments, militias, and groups not active in conflict).

¹⁰For conflicts after 2000 I have used the original data to update the variable where possible.

¹¹In some cases, when information is not available for all years for a particular group, numbers on a group's troop size are used as a best estimate of the troop size for the following year.

¹²There is also a risk that this is to some degree a result of the Polity coding of democracy—that democracies with violence against civilians get lower democracy scores (Vreeland 2008).

¹³If the cutoff point for democracy is lowered to, for example, 3 on the Polity Scale, this new democracy dummy has a significant effect on violence against civilians, also without an anocracy dummy included in the model. This lends further support to the conclusion that it may be the semi-democratic states in the upper boundary of the anocracy category that are driving the results.

¹⁴Related arguments have been suggested by Kalyvas (2006) and Kalyvas and Kocher (2009), who propose that the degree of territorial contestation is the main determinant for when rebels seek to control the population through violence means. That factor is however difficult to control without more fine-grained subnational conflict data.