3.3 Chapter Summary
4.1.4 Measuring the Control Variables
A number of other factors, in addition to properties of interim government, can affect commitment problems of warring parties and thus explain the stability of post-interim peace. The existing literature has particularly argued for the role of several characteristics of (1) a foregoing intrastate armed conflict, (2) a previous regime, and (3) the post-interim period; and one may additionally argue that (4) other aspects of interim governments than the ones captured by my independent variables affect the stability of post-interim peace. In my main analysis – and thus also in the subsequent discussion of this section – I follow Achen (2005) and Clarke (2005) who caution against over-specified “garbage can” regression models with too many controls, and I thus focus on five key control variables belonging to these four categories just listed; control variables that have been proposed as important by the vast majority of recent literature. In my robustness checks in section 4.2.2, I fit further sets of controls that I explain “on the go” and list in the codebook in Appendix A.
Firstly, many scholars reason that ethnic conflict is less conducive for stable peace in the aftermath of war as compared to intrastate conflict over non-ethnic issues. In ethnic conflicts, rebel groups typically fight in the name of a specific, previously marginalized ethnic group or a coalition of several ethnic groups. This impedes the chances for peace because violence exacerbated by ethnic divisions makes peaceful coexistence unlikely if minorities do not receive credible guarantees that they will not be marginalized in future politics (Doyle and Sambanis, 2000; Lake and Rothchild, 1996; Wucherpfennig et al., 2012). To capture the ethnic conflict control variable, I rely on previously collected data by Walter (2004) and Kreutz (2010), who code all cases as conflict over ethnicity and with a value of 1 where “combatants broke down along ethnic lines, or a faction defined itself as a separate ethnic group” (Walter, 2004, p. 376), such as in Burundi’s civil war between ethnic Hutus and Tutsis (Kuperman, 2015). All other cases are coded as 0. In my sample, 31 interim governments (or exactly 50 per cent) follow periods of conflict over ethnicity.
Secondly, intrastate conflict over territorial issues has been found to result in more stable peace as compared to conflict over government, for instance because a government may be more willing to bargain and make concessions to a rebel group if the fight concerns not the entire national territory, but only smaller parts of it (Flores and Nooruddin, 2012; Svensson, 2009). I understand conflict over government as one with an incompatibility “concerning type of political system, the replacement of the central government or the change of its compo- sition,” while a conflict over territory includes an incompatibility “concerning the status of a specified territory,” such as secession or autonomy (Uppsala Con- flict Data Program, 2015a). To capture the variable conflict incompatibility, I use the UCDP/PRIO Armed Conflict Dataset that codes territorial conflicts as 1 and government conflicts as 2 (Gleditsch, Wallensteen, et al., 2002; Pet- tersson and Wallensteen, 2015). In my sample, 49 interim governments follow
governmental conflicts, while 13 follow territorial conflicts.
Thirdly, past research argues that conflict intensity affects the stability of peace, because intrastate conflicts that result in more battle-related deaths re- duce the parties’ ability to reconcile (Doyle and Sambanis, 2000; Jarstad and Nilsson, 2008; Mason and Fett, 1996).39 I follow previous studies that have cap- tured armed conflict intensity as a dichotomous measure (e.g. Reid, 2015) and include a dummy for conflict intensity in my models that captures if a preced- ing conflict exceeded 1000 battle-related deaths over the entire conflict period (“civil wars” coded as 1) or not (0), using data in the UCDP/PRIO Armed Conflict Dataset. 33 interim governments followed high-intensity civil wars.
Fourthly, qualities of the post-interim period are possible explanations for post-interim peace. Past research has found the level of economic development and population size to impact the stability of peace. Low levels of economic development decrease employment opportunities for young men; and a pool of impoverished, unemployed men reduces the costs of recruiting for warring parties (Bigombe et al., 2000; Collier, Elliott, et al., 2003).40 I capture economic development as the level of GDP per capita in the post-interim period and log-transform the variable for my models due to a right-skewed distribution. Robustness checks additionally include a measure of annual GDP per capita growth as well as infant mortality rate (per 1,000 births) to capture the quality of post-interim socio-economic livelihoods. In addition, the very definition of intrastate armed conflict “which classifies armed conflict as a civil war only if there is a high threshold of deaths, implies that civil wars are more likely to occur in populous countries” (Hegre and Sambanis, 2006, 514f.), and I thus also include the logged value of a country’s population size. For all GDP, infant mortality, and population statistics, I rely on the World Bank (2014)’s World Development Indicators.
39Others have argued that more severe conflicts with higher numbers of deaths reveal more
private information and thus correlate with a lower risk of renewed violence (Fearon, 1995b).
40Generally, it would also be reasonable to argue that institutional aspects of the post-
interim period, such as the level of democracy or type of electoral system impacts the stability of peace. Yet, these political variables may be directly affected by properties of the interim government. For instance, electoral systems (or any other institutional qualities of a post- interim political system) are likely to be negotiated by those elites that are involved in decision- making in an interim government. Moreover, and related to a point I formulated in the introduction, because my definition of interim government includes the criterion that such governments end in elections, any variable measuring post-interim levels of democracy (a concept that includes elections as a minimal criterion) would be biased. Thus, all variables capturing political institutions of the post-interim period are likely to be intervening, instead of confounding variables, and should thus not be controlled for (King, Keohane, et al., 1994; Ray, 2003). By definition, a confounding variable is an antecedent factor that correlates with both predictor and outcome. A variable Z however intervenes in the relationship between predictor X and outcome Y: For instance, the design of post-interim institutions is likely to be an intervening variable if it correlates with the stability of peace (Y), but is the result of a property of interim government (X). This is likely the case if post-interim institutions are designed, for instance, by parties in power-sharing rule. Controlling for Z may make me discover intermediate links in the relationship between X and Y, but no confounding factors that discredits my hypothesis about a link between X and Y (Ray, 2003).