Ours is certainly not an age of civil peace. At this moment, neighbors are killing neighbors, in organized groups, in ongoing civil conflicts spanning at least 36 separate countries. Such violence is shocking in its brutality; but through our revulsion we tend to forget that in each of these conflicts, the lines of animosity were not simply given by nature. They were actively produced. Humans are not born knowing the difference between a ‘Serb’ police officer and a ‘Croat’ police officer, or the difference between a ‘Sunni’ mosque and a ‘Shia’ mosque. The participants had to be taught how to hate, and who to kill. In other words, the production of collective violence is always preceded by the production of a certain kind of collective idea: the idea that it is justified, or even necessary, for ‘us’ to kill ‘them.’
Increasing evidence indicates that the chances of success for such an idea can be powerfully influenced by the topology of the communication infrastructure present in a country. In previous work I presented global data showing that mass communication technologies, such as newspapers, radios, and televisions, have tended to favor state integration and stability. In that work I argued that by offering large nationwide loudspeakers to entrepreneurs of political ideas, such technologies incentivize the production of big ideas, which resonate on a national scale, and thereby facilitate the generation of unified national attachments. As a consequence, states with higher levels of mass media accessibility have been far less likely to experience violent internal divisions, as it has been more difficult for political entrepreneurs in these states to succeed on the basis of appeals to narrow sectarian loyalties, in the face of competition from larger-scale producers. That is, we might say that mass communication technologies function to strengthen economies of scale in the marketplace of ideas.
At first glance, one might think that the same would be true for social communication technologies, such as cellphones, which instead facilitate private interactions between individuals. The problem is that linkages facilitated by social media technologies tend to be connections between friends and acquaintances, who tend to be quite similar to each other in terms of ethnicity, religion, socioeconomic status, and any number of other factors. As a result of this tendency toward “homophily”, social media technologies tend to promote communications that flow along preexisting social and political divisions, rather than across them.
Such technologies, rather than generating pressures towards breadth and unity, may therefore incentivize political entrepreneurs to play to the local crowd, by telling them what all local crowds want to hear: that they are special, and righteous, and that their problems are caused by outsiders and heretics who are fundamentally different from themselves. In other words, it seems plausible that the segregated nature of social communication technologies will function to weaken economies of scale in the marketplace of ideas, making it easier for smaller-scale producers to succeed in promoting radicalized justifications for collective violence, even in the face of counter-narratives promoting unity and stability.
However, in order to test this conception of ideational competition, I needed a new kind of data. While existing work in this area had made great strides with sub-national measurements focusing on a single country or a single technology, this test would require sub-national measurements of media penetration that spanned multiple countries and multiple technological forms.
To accomplish this, I relied on surveys collected by the DHS Program, which report geo-coded household characteristics, across 24 African states, with survey dates ranging from 2005 to 2010. In particular, I focused on the ownership of a radio receiver as an indicator of mass media penetration, and ownership of a cellphone as an indicator of social media penetration. These measurements were then converted into continuous estimates of media penetration over space, utilizing spatially explicit interpolation (i.e. ‘kriging’), with parameters optimized through empirical cross-validation. Finally, to avoid the difficulties associated with regression on arbitrarily-sized spatial units, I estimated point process models, which generate statistical inferences by comparing covariate values at the locations of observed events to values measured at randomly simulated control points.
The results of the optimized kriging estimation can be seen in Figure 1. Radio penetration is displayed in purple, and cellular penetration in green, with absolute levels shown in the left panels, and relative levels in the right panel. The plot makes clear that the African context contains an enormous diversity of historical trajectories, with strong variation in technological penetration both within states and across states.
Taking advantage of this diversity, I used the point process modeling approach to estimatethe impact of each form of media penetration on the production of collective violence, based on the violent event locations measured by the UCDP GED database (shown as red dots in the figure).
As predicted, the results indicate that locations characterized by higher levels of radio penetration tend to generate lower rates of collective violence, while locations characterized by higher levels of cellular penetration tend to generate higher rates of collective violence. They also point to the intriguing possibility of an interactive effect between these media forms, in which the prior presence of radio infrastructure dampens the violence-promoting effects of cellular infrastructure, demonstrating the strong effects that can arise through historical path-dependence.
The results thus provide further evidence that the dynamics of collective violence are driven not only by the material forces of armed coercion, but also by the normative forces of communication.
Camber Warren is an Assistant Professor at the Department of Defense Analysis, Naval Postgraduate School.