How the NYC crime data misleads
Why New Yorkers don’t feel safe despite some data claiming they are
Following the horrific crime last weekend where a woman was set on fire on the F train, Nate Silver started another episode of the very common Twitter/X debate: “NYC is actually safer than other cities in America.”
The data Nate is putting out is technically correct. NYC has lower violent crime rates per capita than many other American cities and these rates have fallen significantly in recent decades. I’ve also been a (generally happy) resident here for 11 years and have never once been the victim of a violent crime. So I’m not here to argue that NYC is a very dangerous place to live/visit or other hyperbolic statements you may see on the internet.
But most New Yorkers know these stats don't tell the full story - and it's not just because they're overreacting to news headlines like Nate suggests. When I dug deeper into the data, residents’ concerns made much more sense. I found that:
Citywide crime rates are misleading due to crime's localized nature and the arbitrariness of city borders. Some NYC neighborhoods are very dangerous while others are not.
Looking at the data “per capita” masks the localized danger people reasonably perceive because of high population density. No one likes playing needless Russian Roulette on their city block.
Violent crime may be low in nicer areas of NYC, but these areas have high levels of non-violent crime and unreported crime.
You can’t buy your way to safety in NYC as much as other “more dangerous” cities, which makes people feel more helpless.
City-wide crime rates are meaningless given the localized nature of crime
One of the worst ways to assess the dangerousness of a city is to look at the overall violent crime rate and compare it to another city. If I convince you of nothing other than to stop linking to this data, then I will be happy.
Nate links to data showing NYC’s homicide rate is ~5 per 100k population, much lower than a number of other major cities.
But violent crime is highly localized and city borders are completely arbitrary. Violent crime is not evenly distributed throughout a city:
“[A]ccording to several analyses of cities of varying sizes, Weisburd and his colleagues have found that somewhere in the range of 4 percent of a city’s street segments will see somewhere in the range of 50 percent of that city’s total crime, and about 1.5 percent of a city’s street segments will see about 25 percent of that city’s crime.”
A map of gun murders in NYC1 shows they are concentrated in specific areas which makes the “overall NYC murder rate” a pretty meaningless statistic. Living in some suburbs of Queens or Staten Island, where there were 0 homicides from 2020-2023, is fundamentally a different experience than living on East 125th Street, where there were >40. Some parts of NYC have Tokyo-level murder rates, and others are some of the most dangerous places in the country or the world!
The geographic distribution and arbitrary nature of where you cut the borders makes comparisons between cities kind of silly. On paper, Washington D.C. has a murder rate nearly 10x that of NYC in 2023. Is it actually 10x more dangerous? It similarly depends on where you are: Northwest DC is extremely safe, and Southeast is possibly more dangerous than anywhere in NYC. But D.C. also excludes extremely low crime areas right near the border in Maryland (ex: Bethesda) and Virginia (ex: Arlington). Simply expanding Washington D.C borders to include parts of the metro area would make the crime rates look a lot better.
So looking at the “averages” of this data is pretty pointless and will mask the reality that many New Yorkers know: significant parts of NYC are unsafe. The fact that violent crime is low on Staten Island doesn’t change the reality of someone living in the Bronx.
“Per Capita” is misleading given NYC density
A common response to this might be “sure, NYC has some dangerous areas just like any other place in America. But most parts of it have low violent crime rates per capita!” Putting aside the silliness of the framing (everywhere in NYC should be a safe place to live!), it also misses the fact that “per capita” is doing a lot of the work here.
The more dense an area, the less the “per capita” data makes sense. Given NYC’s high population density, a single block could have 5k people whereas a large suburb in many areas could have less. If there’s a murder a block away, I don’t care that much whether 50 or 500 people live in my building even if “per-capita” data says the former scenario is 10x more dangerous. No one wants to play Russian Roulette everyday, even if their odds are good.
Less serious crime happens at higher rates, even in “nice” areas
The prevalence of violent crime can’t fully explain why many NYC residents feel unsafe. As I mentioned, most NYCers (especially people I know) don’t actually feel at risk of being murdered each day even if they recognize NYC can be a dangerous place.
But what everyone sees are the less serious crimes that are more evenly distributed throughout the city (even in nice areas!) and create a low-level sense of ambient danger. This is borne out in the data2: while Manhattan (south of 59th St) had less violent crime (murders, felony assault) than the Bronx, it had more “less serious” crime like grand larceny, petty theft, and transit crime.3
But a lot of the crime and disorder I’m discussing is not reflected in the data at all. Take the NYC subway. Nate notes in a follow-up piece that violent crime on the subway is pretty low:
“That includes the subway, where violent incidents have increased but are very low in absolute numbers: 5 homicides in 2023 out of 1.15 billion recorded rides: your risk of being murdered on the subway in a given year in New York is of roughly the same magnitude as being struck by lightning.”
This isn’t the right framework to think about things. As someone who rides the subway daily, it’s nearly a weekly occurrence at this point that I get on a subway, identify an obviously deranged person and need to focus on them to ensure they don’t do anything crazy. Most of the time nothing has happened to me, but like many I’ve had my fair share of close calls: fist-fights starting next to me, people getting in my face and threatening to attack me, and one guy giving a long preamble suggesting he was about to carry out some attack and that he’s tired of being out on the street (sound familiar?). Telling myself “the statistical probability of murder right now is pretty low” doesn’t really make me feel particularly relaxed.
This is not even to mention the large swath of other anti-social behavior you see on the subway from lewd acts, drug usage, and people just sprawled out over the floor of the car. The data doesn’t capture any of this because no one calls the police on these people given it clearly does nothing. And this has gotten much worse in the last decade.
What particularly frustrates New Yorkers is Nate’s basic framing. Murder on the subway is very low, but why does it exist at all? In a city with the highest taxes outside of California in the US, why do people have to put up with this Russian Roulette in our public spaces? Why do dangerous people roam our subway system and people are just expected to put up with it? Why benchmark ourselves to St. Louis instead of Tokyo or Madrid? It’s the fact that these crimes are eminently preventable, not just the amount of crime, that matters most to people.
Crime in NYC is harder to escape than other cities
The last thing that makes New Yorkers feel unsafe is that they are more helpless to escape all the crimes discussed. In a lot of other cities, even ones with “higher” murder rates like Baltimore or Detroit, you can spend money to avoid crime. You may live in a nice suburb near the city, exclusively drive everywhere, and hang out in mostly private spaces (corporate offices, other people’s homes, restaurants, etc.). The lower density also means crime is geographically spread out and more easily avoidable.
NYC doesn’t have this. You can live in a safer neighborhood, but even then you are likely taking the subway, walking a lot, and congregating in more public spaces like Central Park. This, plus the density of NYC and the feeling that a lot of crime in NYC is random4, means the general crime and disorder are harder to avoid here.5
I don’t think NYC should aspire to be like these other cities. The public spaces, walkability, and lack of need for a car are what make NYC great! And more importantly, safety shouldn’t be something you need to buy. The goal should be to make crime lower for everyone, not just avoidable for people with money. But if you don’t make the public areas safe, people will look to buy their way out.
My hopeful take is that this will create more pressure to fix crime in NYC than in other places where people can (unfortunately) ignore crime more easily. However, the first step to solving the problem is acknowledging that the data people usually show is not reflective of the reality on the ground.
Guns were the weapon of choice in 54% of murders in 2023 and this map looks representative of similar total homicide maps I’ve seen.
Based on NYPD data covering the Bronx and “Manhattan South” precincts.
Note: I look at absolute numbers because it’s hard to normalize them for each area. While the Bronx has more residents (1.4m vs ~1m in this area of Manhattan) there is a huge influx of commuters and shoppers (likely >1m) into this area of Manhattan each day that make it more populated during the day.
The average person cares a lot more about random violent crime vs crime due to people engaged in illicit activity or something similar. If the murders in a city were mostly drug deals gone bad, the average person wouldn’t worry that much (for better or worse).
The data I found is inconclusive on whether crime in NYC is more random than other places. 27% of murders last year were from strangers, vs 19% for the USA in 2019. But there are enough high-profile random attacks every week that make residents feel like it is and adds to the feeling of “Russian Roulette”.
In SF, I think it’s even worse: I see multi-millionaires on Twitter/X complain about getting their homes or cars broken into regularly no matter where they live (I’m not kidding).
Thoughtful as always!
What's your preferred crime metric? Mine is "expected crime exposure": person-weighted expected crimes within 10 minutes' travel.