Table 2: Cross-sectional results for number of conflict events No. conflict events, GDELT (1) (2) (3) CoverageFB Coverage Observations District dummies Spatial controls Population controls No. conflict events, ACLED (4) (5) (6) 0.524 (0.500) −0.537 (0.495) −0.330 (0.710) 0.563 (0.779) −0.152 (0.946) 0.802 (0.952) 0.029 (0.030) −0.038 (0.029) 0.027 (0.023) −0.028 (0.025) 0.020 (0.023) −0.024 (0.022) 330 X 330 X X 330 X X X 330 X 330 X X 330 X X X ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. Robust standard errors clustered at district level are reported in parentheses. The dependent variable is number of conflict events per 1,000 population. Population controls: log population, log population density, dummy for below median urban rate, age (15–64 y.o.), population with no ID, population with electricity, mobile phone, landline phone, and internet at home. Spatial controls: 2nd order polynomials of distance to major town, distance to major road, distance to railway, distance to MPT transmitter, distance to other company’s transmitter, mean elevation, slope and aspect of the slope, variance of elevation and slope. The predictors are standardized. The results are consistent with the interpretation that incendiary content on social media may have played a smaller part than enhanced communication and coordination. Whereas the previous literature shows that propaganda and mass media tend to work in the intended direction, the literature on the role of communication technology in conflict situations presents mixed results.10 As availability of social media increases, occurrence of conflict becomes less likely. In a related study, Shapiro and Weidmann (2015) find that the expansion of cell phone infrastructure in Iraq decreased insurgent violence. The authors suggests that access to cell phones benefited counter-insurgents, for instance by making it easier to covertly inform security forces of militia activity. Table 2 presents results for number of conflict events per 1,000 people. The results suggest that Facebook availability has no effect on the intensive margin. The point estimates in columns (1)–(3) are also much more sensitive to controls than in the linear probability model. The point estimates in (4)–(6) very small, which reflects the low number of conflict events in the ACLED data. Next, Table 3 examines whether the treatment effect varies by conflict type. The outcome variables are indicators for different types of conflict events. The results show that social media availability and cell phone coverage influence different conflict types with varying intensity. Columns (1)–(4) show that, in GDELT data, the negative effect of Facebook on conflict is mostly due to its effect on assaults and fighting. Events categorized as assaults include abductions, physical assaults, and use of explosive devices. 10. See e.g. Pierskalla and Hollenbach (2013), Shapiro and Weidmann (2015), and Manacorda and Tesei (2020). 17