The Berkeley Science Review: Articles

A Night at the Movies
A surprising effect of violent cinema (view PDF)
by Paul Crider

It's a popular pastime among public figures to decry violence in the media. In recent years, the ire has largely shifted to video games; for example, a recent bill introduced in the US Congress would require vendors to check customer IDs before selling mature-rated games. And after the Virginia Tech violence in 2007, the Federal Communications Commission issued a report calling for greater broadcast content regulation and expanding such regulation to cable and satellite channels.

Surely if less violence were consumed as entertainment, violent behavior in society would decrease. Well, not so fast. UC Berkeley Associate Professor of Economics Stefano DellaVigna, along with economics professor Gordon Dahl from UC San Diego, is working on a study that suggests the effects of media violence aren't so simple. DellaVigna and Dahl found that, on nights when violent movies open to wide audiences, violent crime actually decreases. Initially surprised by these results, DellaVigna recalls, "We tried to fix the bug in the data, but there was no bug."

The researchers determined audience size on particular weekends using box office data from the-numbers.com, a free site used by industry professionals to track movie information. They also used kids-in-mind.com, a website that rates movies on a ten point scale along three axes—sex/nudity, violence/gore, and profanity—to separate movies into strongly violent (Wanted), mildly violent (The Incredible Hulk), or non-violent (Wall-E). These were combined with crime data from the National Incident Based Reporting System to study the correlation between violent crime levels and attendance at violent films.

DellaVigna and Dahl found that for every one million theater attendees, violent crime—defined as reported assaults and intimidation—decreases by 1.1 percent for mildly violent films and 1.3 percent for strongly violent films between 6 pm and midnight. The effect per million viewers between midnight and 6 am (well after the theaters close) was larger: a decrease of 1.9 percent and 2.1 percent for mildly and strongly violent movies, respectively. This equates to about 1,000 fewer occurrences per weeken, and an estimated $695 million saved per year in avoided costs associated with violent crime. No statistically significant effect was observed for attendance at nonviolent movies.

Ideally, a scientist interested in the effects of movie violence might randomly assign half of a random group of people to watch a violent film and the other half to watch a nonviolent one and then observe what happens. The randomness of the groups averages out other possible explanations for any differences in observed behavior. "In the movie theater, we're not randomized. We choose to watch violent movies," explains DellaVigna. This makes analysis more difficult, and other possible influences on the data were considered and corrected for, like occasions of high television viewership (think Superbowl), possible seasonal effects, and inclement weather.

To explain the counterintuitive results, DellaVigna developed a model in which consumers choose between attending a violent movie, attending a nonviolent movie, or doing some alternative social activity. The model, once fit to the data, suggested that violence-prone individuals are disproportionately attracted to violent films, and while in the theater, they are "voluntarily incapacitated." That is, they're watching a movie and eating popcorn instead of engaging in other activities associated with violent behavior, such as drinking at a bar.

That violent crime decreased even more during the late night hours, after theaters are empty, was a more curious result. DellaVigna explains this effect as the result of a "substitution of foregone activities." In other words, even though violent individuals are no longer fixed on the edges of their seats, they have initiated a less raucous evening and may just go to bed after a late movie. And even if they do go out after the film, they simply have less time to drink themselves to belligerence. "It's the difference between drinking from 7 pm to 1 am and drinking from 11 pm to 1 am," says DellaVigna.

These findings may appear to dispute evidence from other research that does imply a relationship between media violence and crime. Indeed, laboratory studies in psychology have shown an increase in aggressiveness in individuals upon viewing violent video clips, although these studies don't address actual violent crime. Other studies show that survey respondents who view violent media are more likely to be involved in self-reported violent crime, although the direction of causation in these studies is unclear. DellaVigna himself found evidence supporting the claim that an increase in violent behavior does follow exposure to violent films relative to exposure to nonviolent films; the effect was just swamped in the short run (the evening and night after the movie). DellaVigna explains, "You've got two opposing effects. You have a movie like Hannibal causing more aggression, but it's also taking people out of bars. It's the second effect that dominates."

DellaVigna is quick to caution against jumping to political conclusions. Like the laboratory studies, this study cannot comment on the long-term effects of media violence. A better take-home message from the study is that any policy addressing a particular activity must address the next best activity as well. Perhaps violent video games increase aggression, but they also keep a potential offender fixed in his seat. Happily incapacitated, the young ne'er-do-well foregoes his next favorite activity, which could be cow tipping, or worse. "Can we think of activities for youth that are more attractive than committing crimes?" asks DellaVigna. "That's something we should keep in mind when designing policies on entertainment."

Paul Crider is a graduate student in physical chemistry.

Comments on this article? Drop us a line at with 'letter to the editor' in the subject!