Constraints and military coordination: How ICTs shape the intensity of rebel violence

ABSTRACT

Why and how do information and communication technologies (ICTs) shape the intensity of rebel violence? Recent studies find that ICTs can both increase and decrease such violence. We argue that, during civil wars, this effect depends on the type of ICTs. Mobile phones give rebels better military coordination to organize violence. In contrast, the internet increases the constraints of rebel groups to use violence. On the one hand, the internet increases the visibility of rebel groups forcing them to moderate their levels of violence. On the other hand, the internet gives rebels’ opponents better tools to limit the levels of rebel violence. We test these two arguments empirically with panel data of rebel violence in countries experiencing a civil war from 1989 to 2007. Through a series of negative binomial regressions, we find general support for our hypotheses. For the case of the internet, we specifically find that its effects are concentrated in more recent periods when social media became more widespread.

¿Por qué y cómo influyen las tecnologías de la información y la comunicación (TIC) en la intensidad de la violencia de los rebeldes? Estudios recientes han demostrado que las TIC pueden tanto aumentar como disminuir dicha violencia. Nosotros sostenemos que, durante las guerras civiles, este efecto depende del tipo de TIC. Los teléfonos móviles proporcionan a los rebeldes una mejor coordinación militar para organizar la violencia. En cambio, Internet aumenta las limitaciones de los grupos rebeldes para utilizar la violencia. Por un lado, Internet aumenta la visibilidad de los grupos rebeldes y los obliga a moderar sus niveles de violencia. Por otro lado, Internet ofrece a los oponentes de los rebeldes mejores herramientas para limitar sus niveles de violencia. Probamos estos dos argumentos empíricamente con datos de panel de la violencia de rebeldes en países en los que hubo una guerra civil entre 1989 y 2007. Mediante una serie de regresiones binomiales negativas, encontramos un apoyo general a nuestras hipótesis. En el caso de Internet, descubrimos específicamente que sus efectos se concentran en los períodos más recientes cuando el uso de las redes sociales se generalizó.

Pourquoi et comment les technologies de l’information et de la communication (TIC) façonnent-elles l’intensité de la violence des rebelles? De récentes études ont permis de constater que les TIC pouvaient à la fois accroître et réduire ce type de violence. Nous soutenons que, lors de guerres civiles, cet effet dépend du type de TIC. Les téléphones portables permettent une meilleure coordination militaire des rebelles pour organizer la violence. À l’inverse, Internet accroît les contraintes auxquelles les groupes rebelles sont confrontés pour faire usage de la violence. D’une part, Internet augmente la visibilité des groupes rebelles, ce qui les force à modérer leurs niveaux de violence. D’autre part, Internet dote les opposants des rebelles de meilleurs outils pour limiter les niveaux de violence des rebelles. Nous avons mis ces deux arguments à l’épreuve d’une manière empirique en utilisant les données d’un panel de données sur la violence des rebelles de pays qui ont subi une guerre civile entre 1989 et 2007. Nous sommes passés par une série de régressions binomiales négatives et nous avons constaté que les données étayaient généralement nos hypothèses. Pour ce qui est d’Internet, nous avons plus précisément observé que ses effets se concentraient sur les périodes plus récentes durant lesquelles les réseaux sociaux sont devenus de plus en plus répandus.

KEYWORDS:

• ICT
• internet
• rebel violence
• civil war
• cellphones

Notes

¹ Intensity of violence refers to the number of killings perpetrated by armed groups. See Weinstein (2006, 202) for a similar measurement strategy.

² We define violence as the deliberate act of killing using the number of deaths as a systematic observable outcome of armed-group activity. However, we acknowledge that violence committed by armed groups is not limited only to assassinations (Weinstein 2007, 199–200).

³ To account for the influence of natural resources, we control for the access to natural resources by rebel groups in our statistical models. More details in Section 3.

⁴ Importantly, we suggest that rebels limit their levels of violence, rather than they stop using violence and change to nonviolent tactics.

⁵ We use the period from 1989 to 2007 to differentiate the use of cell phones only for basic communication purposes such as telephony and non-online texting. Further details in Subsection 3.4.

⁶ We made this decision because our theory addresses cases with sustained interactions between armed groups and their targets. We also excluded informal groups who use violence only sporadically because our theory explains the behavior of organized ones. Besides, any reduction of violence of informal groups could be endogenous to their lack of permanent organization for combat. Then, we only included violence produced by formally organized armed groups following the criteria of Sundberg, Eck, and Kreutz (2012).

⁷ The counts of deaths of government figures and deaths of other rebels come from battle deaths. Even though we cannot know from these data who began the attack, we assume that if one rebel group produces more casualties it was because it had better military coordination.

⁸ In Section 3.4, we provide further details of how, with our empirical strategy, we account for these dynamics.

⁹ In previous versions of this manuscript, we also included the GDP per capita as a control variable. However, as suggested by an anonymous reviewer, this variable could be correlated to the speed of ICT penetration in a country. Therefore, we decided to exclude it from the final models (signs and significance of coefficients hold). Results including GDP per capita as a covariate are available upon request.

¹⁰ We do not include the lagged of the dependent variable for estimating changes in violence. We included it to control for the previous level of violence observed in the country that can explain the violence observed in the current period given dynamics of revenge that lead to spirals of violence (Balcells 2017). We also estimated models without the lagged of the dependent variable as a covariate and including a 3-year moving average of the dependent variable instead. Significance and signs of the coefficients of interest hold for the three specifications. Results are available in the online appendix.

¹¹ In the online appendix, we show the results of models including all years with information available (1989–2015).

¹² To estimate the effect of internet after 2005 we add the coefficient of % of internet users (t-1) and the coefficient of the interaction term Internet × I[after 2005]. To estimate whether this effect is significant, we calculate a joint significance chi-squared test. Results suggest that the effect is significant at p < 0.10.