Understanding Paramilitary Violence Navin Bapat Lucia Bird Chelsea Estancona Kaisa Hinkkainen University of North Carolina at Chapel Hill University of Lincoln November 13, 2015 Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 1/26
The Logic of Paramilitary Violence Paramilitaries are a common feature of civil conflict. Examples: Sudan and Sierra Leone Usually conceptualized with principal-agent frameworks This raises three questions: Why do paramilitaries prefer fighting to negotiated settlements? Why do states support paramilitaries, risking decreased control? Why do states seek to keep their paramilitary support private? Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 2/26
What do we know about paramilitary involvement? Paramilitaries are present in a variety of intrastate conflicts (Jentzch, Kalyvas and Schubiger 2015) Scholars debate the level of state control over paramilitary violence (Stanton 2015; Mitchell, Carey and Butler 2014) States make efforts to keep ties to paramilitaries secret (Carey, Colaresi and Mitchell 2015) Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 3/26
Modeling Paramilitary Presence Rebels, Local Elites, and the Government 2 stage, 3 player game: Rebels, Local Elites, Government Stage 1: Government chooses to arm elites, fight conventionally, or do nothing Stage 2: Rebels challenge, and elites respond by ceding territory or fighting This process repeats. Power is endogenous: transfers from elites improve future rebel fighting power. Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 4/26
Modeling Paramilitary Presence Basic Game Tree: Second Stage Elites Offer x Accept Power shift, repeat Rebels Contest Rebels Reject Conflict ~Contest Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 5/26
Modeling Paramilitary Presence Discontinuous Shifts: Contraband Elites fight if ceding territory means a discontinuous increase in rebel power: Power shifts arise when territory has valuable lootable resources Occurs in weak states where the government does not defend conventionally H1: If the contested territory has contraband, high state capacity decreases the probability of paramilitary activity. Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 6/26
Modeling Paramilitary Presence Full Game Tree Elites Offer x Accept Power shift, repeat Rebels Contest Rebels Reject Conflict Aid Elites Gov. ~Contest No Conflict Conventional Conflict Defend Conventionally Cede Territory Rebels take territory, power shift away from Gov. Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 7/26
Modeling Paramilitary Presence No Contraband Commitment Problems Weak states benefit when elites resist but the elites cannot credibly commit do so. If the government arms elites that agree to fight, this improves the elites ability to combat the rebels. Elites lose the capacity to resist if they negotiate with rebels and the government pulls their support. Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 8/26
Modeling Paramilitary Presence Government Support Elites have an incentive to fight rebels immediately to avoid an unstable peace. Occurs when the government can strengthen elites but cannot fight rebels conventionally. H2: If the territory has no contraband, increasing state capacity has a curvilinear effect. Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 9/26
Modeling Paramilitary Presence Empirical Predictions Pr. Paramilitary Violence Weak Moderate High Government Capacity (wg) Territory with Contraband (Discontinuous Shift) Territory without Contraband (Continuous Shift) Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 10/26
Research Design Data Sources Pro-governmental Militia Data Set (PGMD) from Carey, Mitchell, Lowe (2013) Our analysis is limited to minor powers and states experiencing some degree of civil violence Variables Added: Fearon s contraband variable (2004) Logged GDP per capita (World Bank, 2005) Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 11/26
Research Design Dependent and Primary Independent Variables Dependent Variable: Informal Pro-government Militias Independent Variables: Presence of lootable natural resources (Contraband) State capacity (ln(gdp) per capita) Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 12/26
Empirical Results Table 1 : Logistic Regression: Paramilitary Presence (Key Variables) Model 1 ln(gdp) 6.10 (1.01) ln(gdp) 2 0.44 (0.07) Contraband 7.33 (1.92) GDP Contraband 1.03 (0.28) Num. obs. 523 Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 13/26
H1: Effect of Contraband No State Support 4 2-2 -4-6 -8-10 6 7 8 9 10 11 LGDP Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 14/26
H1: Effect of Contraband Percent Increase in Probability, No State Support Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 15/26
H2: Effect of State Capacity No Contraband 4 2 6 7 8 9 10 11 LGDP -2-4 Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 16/26
H2: Effect of State Capacity Percent Increase in Probability- no Contraband % Increase in Pr. Paramilitary 0.2.4.6 4 5 6 7 8 9 10 11 Government Capacity (LGDP) Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 17/26
Conclusion Results and Future Work Preliminary evidence in support of hypotheses: Lootable natural resources increase the probability of paramilitary activity in weak or moderately weak states. Paramilitaries arise in moderately powerful states without contraband. Future work: Why keep support secretive? Testing with microlevel data Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 18/26
Conclusion Thank you! Navin Bapat bapat@email.unc.edu Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 19/26
Appendix Table 2 : Description of Game Parameters Parameter Values Meaning x 0-1 Elites transfer to Rebels p 0-1 Rebel probability of victory wr 0-1 Rebel capacity wg 0-1 Government capacity β 0-B Cost to government of arming elites θ 0-1 Decrease in p with armed elites c 0-1 Cost of conventional fighting Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 20/26
Appendix Table 3 : Summary Statistics, DV: Informal Paramilitary Presence Frequency Percentage No Paramilitary Emergence 264 51% Paramilitary Emergence 259 49% Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 21/26
Appendix Table 4 : Summary Statistics, IV: Contraband Frequency Percentage No Contraband Present 370 71% Contraband Present 153 29% Table 5 : Summary Statistics, IV: ln(gdp per capita) Minimum Mean Standard Deviation Maximum ln(gdppc) 4.24 6.77 1.18 9.92 Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 22/26
Appendix Table 6 : Logit: Paramilitary Presence Estimate Std. Error Pr(> z ) (Intercept) -23.7436 3.5195 0.0000 ln(gdp) 6.0958 1.0141 0.0000 ln(gdp) 2-0.4406 0.0720 0.0000 Contraband 7.3303 1.9243 0.0001 GDP contraband -1.0289 0.2818 0.0003 Distance 0.0003 0.0001 0.0261 Polity2 0.1322 0.0228 0.0000 Polity2 2 0.0126 0.0043 0.0037 Ethnic fractionalization -0.7680 0.4869 0.1147 Dem. Aid GDP 0.0093 0.0036 0.0106 Aut. Aid GDP -0.0528 0.0447 0.2381 ln(area) 0.2042 0.0898 0.0230 N 523 AIC 631.05 BIC 682.16 Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 23/26
Appendix Table 7 : Probit: Paramilitary Presence Estimate Std. Error Pr(> z ) (Intercept) -14.17297 2.044321 0.000 ln(gdp) 3.660687.597108 0.000 ln(gdp) 2 -.2648403.0423717 0.000 Contraband 4.242742 1.140394 0.000 GDP contraband -.5962883.1676685 0.000 Distance.0001711.0000727 0.019 Polity2.0784443.0132507 0.000 Polity2 2.0069624.0025685 0.007 Ethnic fractionalization -.5007565.2875475 0.082 Dem. Aid GDP.0053449.0021334 0.012 Aut. Aid GDP -.028052.024979 0.261 ln(area).119687.0540717 0.027 N 523 Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 24/26
Appendix Table 8 : Logit (country-clustered standard errors) Estimate Std. Error Pr(> z ) (Intercept) -23.74359 7.457397 0.001 ln(gdp) 6.095782 2.1377 0.004 ln(gdp) 2 -.4405869.1504182 0.003 Contraband 7.330315 3.391101 0.031 GDP contraband -1.028915.5059483 0.042 Distance.0002739.0002027 0.177 Polity2.1321893.0454957 0.004 Polity2 2.0125556.0078566 0.110 Ethnic fractionalization -.7680098 1.168355 0.511 Dem. Aid GDP.0092525.0077422 0.232 Aut. Aid GDP -.0527618.0719103 0.463 ln(area).2041773.1955083 0.296 N 523 Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 25/26
Appendix Table 9 : Probit (country-clustered standard errors) Estimate Std. Error Pr(> z ) (Intercept) -14.17297 4.239232 0.001 ln(gdp) 3.660687 1.247406 0.003 ln(gdp) 2 -.2648403.0879848 0.003 Contraband 4.242742 2.033947 0.037 GDP contraband -.5962883.3068264 0.052 Distance.0001711.0001174 0.145 Polity2.0784443.0264926 0.003 Polity2 2.0069624.0045972 0.130 Ethnic fractionalization -.5007565.668406 0.454 Dem. Aid GDP.0053449.0045661 0.242 Aut. Aid GDP -.028052.0368838 0.447 ln(area).119687.1195323 0.317 N 523 Bapat, Bird, Estancona, Hinkkainen Understanding Paramilitary Violence 26/26