Abstract
Scholars have argued that autonomy, federalism, and other forms of territorial self-governance for ethnic groups are both a successful tool to manage ethnic conflict and a Trojan horse that leads to separatism. This paper contributes to the debate by identifying autonomy as a dual-use technology: under conditions of peace, autonomy acts to maintain a unified state; under large-scale political violence (civil war), autonomy increases the probability of an ethnic group's region emerging as an independent state. I present a cross-national, statistical test of this theory during the post-1945 period.
Notes
1 This paper uses decentralization as defined by Wolff: ‘the delegation of executive and administrative powers to local levels of government’ which differs from fiscal and other forms decentralization (Lyon, Citation2016).
2 It is possible that some of these groups could pursue independence, a point raised by Henry Hale (Citation2004) in his discussion of core ethnic groups. I did repeat the below results leaving those dominant, monopoly, and senior partner groups in the dataset and the results are not impacted in any significant way.
3 The Indonesia-East Timor civil war is in the World Bank dataset used in the second part of the analysis below, but the World Bank's definition of civil war restricts the case to 1975–1982.
4 There are some controversial cases included in the World Bank dataset, including Israel's separation from Palestine and Pakistan's separation from India as war outcomes. I elected to leave these cases in the dataset and acknowledge the fact that these are cases at the margins. Since neither Israel nor Pakistan had autonomy before the start of the war and yet both achieved independence as a war outcome, the cases go against my theory and therefore leaving them in the dataset should not bias the results in favour. Other non-standard cases include Russia's second civil war with Chechnya that began in 1999 or Indonesia's civil war with East Timor that began in 1975. Should these cases be excluded and treated more as territorial annexation that are not relevant for the current research question? I ran the below results with these cases included and excluded, and the statistical significance of the autonomy variable remained statistically significant.