Table 3: Cross-sectional results for probability of conflict: by type of conflict CoverageFB Coverage Data Mean(Y) District FE Controls Coerce (1) Assault (2) Fight (3) Mass violence (4) Battle (5) Civilians (6) Explosion (7) −0.034 (0.111) 0.167 (0.116) −0.277∗∗ (0.127) 0.325∗∗∗ (0.108) −0.253∗∗ (0.099) 0.214∗ (0.112) −0.054 (0.059) 0.033 (0.046) −0.120∗∗ (0.060) 0.036 (0.051) 0.001 (0.051) 0.026 (0.040) −0.005 (0.031) −0.036 (0.035) GDELT 0.51 X X GDELT 0.31 X X GDELT 0.52 X X GDELT 0.07 X X ACLED 0.14 X X ACLED 0.08 X X ACLED 0.09 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 an indicator for conflict of particular type in a township between June 2016 until end of August 2017. All regressions include population controls and spatial controls. 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. There are 330 observations. Fight includes most forms of conventional military force. Based on ACLED, the negative effect on conflict is driven by a decrease in battles. ACLED (2019) defines a battle as “violent interaction between two politically organized armed groups at a particular time and location.” Although the event classifications are somewhat different between GDELT and ACLED, the pattern is quite similar. 6.2. Panel Estimates I now turn to the difference-in-differences model. The dependent variable in columns (1)–(2) of Table 4 is an indicator for conflict events, and in columns (3)–(4) the number of conflict events per 1,000 population. All specifications include township fixed effects and month-year fixed effects. The coefficients on the interaction terms correspond to a one standard deviation change in CoverageFB. Treat and Post are indicators for the treatment period and post-treatment period, respectively. Because the measure of cell phone coverage is time invariant, it is captured by the township fixed effects. The coefficients on the interaction terms show that on average, during or after the treatment period, higher MPT cell phone coverage is not associated with any change in probability of conflict or number of conflict events per capita. These results show no systematic change in conflict occurrence over time that is associated with the Facebook campaign, although occurrence of conflict is on average more likely during the treatment period. Table 5 examines the effect of Facebook availability on different conflict types. The outcomes are dummy variables that take value one if there was at least one conflict event of that type in the township in a particular month. The estimates show that there is some variation between conflict types, although the coefficients are imprecise. The point 18