In a forthcoming QJE paper, Gordon Dahl (UCSD) and Stefano DellaVigna (Berkeley) test whether violence in the movies influence violence in real life. Using data on cinema releases and attendance, the authors look for correlation between violent-movie attendance and violent crime.
Intuitively, we might imagine that violent movies spur violence. When one leaves the a movie with a lot of fighting, that person might go looking for a fight, right? Surprisingly, the paper finds that violent movie attendance actually causes violent crime to decrease. This may be because those who commit violent crimes like to watch violent movies. When a new movie comes out, they go to the cinema instead of on their "crime spree" (or instead of going to the bar, which otherwise would result in them getting drunk which makes them more likely to commit a violent crime).
They estimate that a violent movie results in 1000 fewer assaults on any given weekend. Admittedly, however, the authors are unable to test for long-run affects of violent movies. Although a violent movie results in less crime in the short run, could increased violence on the big screen result in a more violent society over the course of many years?
Link to the paper on the NBER site
Or, try to download a PDF directly