ABSTRACT

I use event-level data from the Armed Conflict Location & Event Data Project (ACLED) to estimate the impact of Rodrigo Duterte’s “war on drugs” on political violence in the Philippines. Constructing a balanced panel of 80 provinces observed weekly from January 2016 through June 2017, I estimate event-study and difference-in-differences models that compare drug-war-tagged fatalities against a within-country placebo: non-drug-war political violence (battles, explosions, and non-drug violence against civilians). The headline finding is that the placebo series exhibits no discontinuity around either the election or the inauguration, while drug-war fatalities show a sharp, immediate spike. The violence increase is therefore specific to the drug war, rather than reflecting general conflict trends, ACLED source-coverage changes, or other simultaneous shocks. As a precondition for this comparison, I show that the relevant event date is the May 9 presidential election, not the June 30 inauguration. I discuss identification challenges, including anticipation effects, the endogeneity of ACLED’s tagging, and the absence of cross-regional treatment variation.