So, I ran the first meta-analysis of election violence.
For every predictor in 581 models, I code how often it is statistically significant in the direction the author expected using an approach used in electoral turnout and electoral integrity meta-analyses.
3 quick findings… 5/n
