This is clearly a case where "correlation is not causation" must be considered. Suppose that refugees love Facebook. If so, then places with refugees will have

This is clearly a case where "correlation is not causation" must be considered. Suppose that refugees love Facebook. If so, then places with refugees will have more Facebook usage. Anti-refugee violence can only occur where there are refugees. We thus would expect that Facebook usage would be higher in locations with more anti-refugee violence. Another explanation is that old people are less likely to engage in violence and also less likely to use Facebook.