Strategic rebels: a spatial econometric approach to rebel fighting durations in civil wars

ABSTRACT

Recent research on multi-actor civil wars highlights that rebel organizations condition their conflict behavior on that of other rebel organizations, with competition and free-riding constituting the core theoretical mechanisms. We provide a new actor-centric approach to explicitly model strategic interdependence in multi-actor civil wars. We argue that rebel organizations have incentives to remain mobilized until the end of a conflict to maintain their power to negotiate, power to spoil, power to enforce, and power to protect. This induces strategic complements that dominate duration dynamics in multi-actor conflicts. Based on a network game-theoretic model, we derive a spatial econometric framework that allows for a direct test of strategic interdependence. We find that the estimated duration interdependence is positive but partially offset in secessionist conflicts where the public goods nature of the incompatibility also induces strategic substitution effects.

KEYWORDS:

• Civil war
• duration analysis; network games; spatial econometrics; strategic interaction

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

¹ Increasing awareness to these interdependencies has been addressed not only by network approaches (e.g. Gade, Hafez, and Gabbay 2019; König et al. 2017; Metternich et al. 2013) but also random effect models (see Cunningham and Sawyer 2017).

² Note that because coefficients will be biased, clustering standard errors will not solve this problem, i.e., incorrect inference will prevail. See also Poast (2010).

³ See ‘power to resist’ in Cunningham, Gleditsch, and Salehyan (2009) and Conrad, Greene, Walsh, and Whitaker (2019).

⁴ ${\bar{y}}_i$ is obtained from the first-order condition by equating marginal benefits with marginal costs, i.e.: $\frac{\mathrm{\partial }{U}_i}{\mathrm{\partial }{y}_i}=b^{\prime }(y)-c_i=0$.:

⁵ Note that a pure strategy equilibrium hinges on an exogenous upper bound on $y_i$ when $\rho <0$. Empirically, this upper bound exists if finite endowments or resources exist.

⁶ These are calculated as $y_i^{\ast }=\frac{1}{1-\rho {\sum }_j{y}_j}{\bar{y}}_i$.

⁷ A further obstacle originates from the fact that these models cannot easily deal with censored observations. Our proposed solution avoids dealing with censoring (or TVCs for that matter) and relies on non-censored observations only.

⁸ Ideally we would endogenize both waiting and fighting times through simultaneous equations (see Hays, Schilling, and Boehmke [2015]). However, to do so we require suitable instruments for identification, i.e. at least one variable that is associated with either outcome, but not the other. Unfortunately, we are not able to plausibly identify such a variable. This implies that our results, in this regard, are vulnerable to potential endogeneity concerns.

⁹ To accommodate the log specification (which does not allow for values of zero) we added one day to each conflict entry time. Choosing even smaller values did not alter the results.

¹⁰ Testing for the difference between coefficients according to Clogg, Petkova, and Haritou (1995), we also find no difference between the two estimated $\rho$s.

¹¹ This project benefited greatly from the open-source R community. Estimation relies on ‘maxLik’ package developed by Henningsen and Toomet (2011). Presentation of results is supported by ‘texreg’ (Leifeld 2013).

Additional information

Funding

This work was supported by the Economic and Social Research Council [ES/L011506/1];Gerda Henkel Foundation [AZ 07/KF/13];Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung [#116795].