Political ads often tout candidates as tough on crime while painting their opponents as ineffectual wimps on the issue. And for seemingly good reason: sometimes it’s obvious that concerns about crime affect election outcomes. After the sky-high murder rates of the early 1990s, for example, deep-blue New York City elected Republican and independent mayors for two decades straight.
But is there a more general relationship between crime rates and election results? In a new working paper, Boston University’s Arjun Vishwanath digs into the data and, surprisingly, comes up emptyhanded. When crime goes up, voters don’t systematically vote out incumbent parties or candidates, side against the sitting president’s party, or move toward the tougher-on-crime GOP.
Vishwanath’s analysis is one of only a few studies to tackle this question, and easily the most comprehensive. His data extend from 1960 to 2019 (with some of his analyses starting in the 1990s instead), and they encompass numerous executive-branch and law-enforcement offices whose occupants make public-safety decisions: the president, state governors and attorneys general, plus local mayors, county executives, sheriffs, and prosecutors.
He also looks at not one, not two, but three different crime indexes: total crime, property crime, and violent crime. These indexes are dominated by relatively minor, more common crimes—even the violent-crime index includes simple assault with no weapon or significant injury—and are also limited to crimes that are known to the police. However, Vishwanath shows the results for several other crime measures in an appendix, including homicide rates, where underreporting is less of a concern (but low sample sizes are more of a concern, as small jurisdictions can go years without a killing).
All these variations multiply out to a whole lot of models, each designed to test whether changes in crime rates predict changes in vote shares from one election to the next. For example, one model tests whether voters move toward Republican law-enforcement officials when property crime goes up, and another asks whether voters punish incumbent governors of both parties when violent crime does.
The results are overwhelmingly tiny and statistically insignificant, and sometimes they even lean in the “wrong” direction. The same remains true when, instead of analyzing changes in vote shares between elections, Vishwanath looks at differences in how areas responded to the same elections—asking, for example, whether counties with different crime trends see different shifts in their gubernatorial votes.
The bottom line is that, while some people may well vote based on crime concerns, voters in general don’t systematically move in any particular way when crime rates change.
In an interesting aside, Vishwanath explores the question of whether cops “juke the stats” in response to electoral pressures, which could throw off his analysis but is also important in itself. While there are certainly individual cases where this has happened, he fails to find unusual patterns—such as declining crime, more rapes classified as “unfounded,” or more assaults classified as “simple” instead of “aggravated”—as elections approach.
The analysis does come with some caveats. Agency-level crime data is a bit of a disaster in the U.S., and Vishwanath opted to clean those data (which means lots of judgment calls about which entries are errors); aggregate them to monthly totals by state, county, and municipality; and impute missing numbers himself, rather than rely on existing datasets that (quite crudely) attempt similar fixes. Any remaining measurement error—and there will most certainly be some—can weaken crime’s relationships with other variables.
In future work, I wouldn’t mind seeing a narrower analysis that, say, focused on elections in larger jurisdictions where murder rates, specifically, had changed by an unusually large amount.1 Murder is a crime that’s measured particularly well (not only by the police, but also by public-health officials), and large changes in the murder rate are something voters might be especially likely to notice.
More routine changes in more common crimes, though, seem to have quite small effects on elections. To the extent these trends are random or outside politicians’ control, that might not be entirely unfair. It also, however, implies a certain amount of forgiveness for the mismanagement of public safety.
From the Manhattan Institute
Daniel Di Martino provides an interactive tool detailing the tax benefits of moving from one state to another.
Other Work of Note
Cool new collection of articles about illegal markets in the Proceedings of the National Academy of Sciences. Featuring drugs, sex, guns, cigarettes, smuggled wildlife, and so much more!
Big new paper on the “economics of a shrinking world”: “modernity itself,” the authors contend, “makes a third child expensive and childlessness cheap.” On a smaller scale, the Colorado Family Planning Initiative also seems to have reduced fertility.
How are workers sorted by geography in a “superstar economy”?
Young Americans are increasingly worried about AI, survey data show. But what happens to people who go into fields that later decline?
Comparing AI development in China and the U.S.
Not sure if you should appeal your property taxes? Asking a chatbot might help.
Credit cards seem like a pretty profitable line of business.
Which federal databases have been recently terminated?
Which cities have the most vegetation?
Christmas came early this year, by which I mean the FBI released 2025 crime statistics before we made it two-thirds of the way through 2026. As you’d expect if you follow these trends, crime is way down again, including an 18% murder drop.
Interviewing prisoners after a prison riot.
Do defendants with college degrees receive lighter sentences?
How do sexual offenders try to avoid detection?
Suicide is contagious among teens. And some findings on suicide hotlines.
When federal courts started publishing lists of judges with overdue matters, judges got the lead out of their you-know-whats.
Does minimum wage hurt the low-income?
Why American workers’ annual raises don’t make up for unexpected inflation.
Final published version of a much-discussed paper presenting disappointing results on universal basic income.
In the appendix, to be fair, the paper separately looks at large localities, homicide rates, and especially large crime swings—it just doesn’t combine them.
Might there be evidence that people vote with their feet? Hard to imagine NYC population changes had nothing to do with crime. Comparing voting across election cycles doesn’t measure that,
"Political ads often tout candidates as tough on crime while painting their opponents as ineffectual wimps on the issue."
Yeah, that's how docility is maintained in the American electorate. Convince them to be afraid of crime, regardless of whether it's going up or down, and the fact that the US already has the highest incarceration rate in the world and yet the overwhelming political message is "Let's do what we're already doing... but harder!"
Americans are controlled and kept obedient through the non stop coordinated Epstein-class captured media feeding the public a never ending dose of fear and a declaration of who is to blame.