Microfoundations of Civil Conflict Reconciliation: Ethnicity and Context

Abstract

Comparative work on reconstruction and peace building in war-torn countries is dominated by a macro-oriented approach, focusing on structural political reforms, legal issues, disarmament, demobilization and reintegration of (rebel) soldiers, and repatriation of the displaced. This article offers a different perspective, examining micro-level determinants of reconciliation. Earlier research indicates that political attitudes in post–ethnic conflict societies are shaped by ethnic affinity. A large literature on the importance of contextual conditions for human behavior would suggest that ethnic composition of the local population and physical proximity to the conflict zone also should affect individual support for peace and reconciliation. To test these propositions, we draw on a geo-referenced survey of the Macedonian population that measures respondents' perception of the 2001 civil conflict. Contrary to expectations, the spatial and demographic setting exerts only feeble impacts on individuals' support for the Framework Agreement. Several years after the conflict was settled, the survey data reveal a strongly divided Macedonian society where ethnicity trumps all other individual and contextual factors in explaining the respondents' preferences.

Keywords:

• ethnic conflict
• Macedonia
• nationalism
• peace agreement
• reconciliation

Acknowledgments

We thank the editor, three anonymous reviewers, and colleagues for helpful comments, and Andreas F. Tollefsen for excellent research assistance. An earlier draft of this article was presented at the 6th Pan-European International Relations Conference in Turin, Italy, 12–15 September 2007. Replication data are available at http://dvn.iq.harvard.edu/dvn/dv/internationalinteractions.

Notes

¹The limited number of attitudinal surveys in post-conflict societies prevents a comprehensive cross-national analysis of individual attitudes toward peace agreements. That said, Macedonia is an interesting case for several reasons. The brief conflict represents the most recent armed conflict in Europe outside the Caucasus. Moreover, the conflict was limited both in time and space and never developed into full civil war. It is likely that societal polarization and traumatization increase with the intensity and longevity of the conflict; thus, the relatively limited fighting in our case reduces the possibility that our findings are product of extreme circumstances and therefore not transferable to other contexts.

²Where not otherwise specified, the term “Macedonian” refers to ethnic Macedonians, while “Albanian” refers to ethnic Albanians of Macedonian nationality.

³For an overview of different concepts of ethnicity and ethnic violence, see Oberschall (2000).

⁴Though see Bakke et al. (2009), who report that people who live close to areas of violence in the Caucasus are more disposed to forgiving than those who live farther away from the war zone.

⁵We thank one anonymous reviewer for bringing our attention to this point.

⁶The last item about amnesty does not strictly ask about minority rights, but correlates very well with the other variables, and is therefore included in the minority rights scale. Scales with and without this items have a correlation of r = .99.

⁷As an additional validity test, all analyses were run on the individual indicators separately. This does not alter our main results, nor do these models indicate that the scales are not one-dimensional.

⁸The conflict event data were coded as part of a joint project led by Ola Listhaug, NTNU, and are available through the ACLED database (Raleigh, Linke, Hegre, and Karlsen 2010).

⁹A number of other municipality characteristics were also tested, including the population share with secondary education, percentage of the population employed in the service sector, the number of internally displaced persons, a dummy variable for the capital Skopje, and population size. With the exception of population size, none of them turned out to be statistically significant, nor did they otherwise alter the results reported here.

¹⁰We tried a number of other predictors on the individual level, but none of these revealed any additional significant effects. Moreover, some individual-level indicators could not be included because they were highly collinear with ethnicity or were related to our dependent variables in ways where causal direction could not be assumed.

¹¹Due to highly skewed distributions of the distance measures and because one might expect a decaying effect of distance we also tested log-transformed variants. However, measures of fit indicated that the reported, non-transformed variables contribute to more efficient models, if only by a fraction.