In Soundview, in the East Bronx, the violent-crime rate is among the highest in the city — 8.4 major violent felonies a year for every 1,000 residents, 14th of 59 districts. It is, by the number everyone uses, a dangerous neighborhood. But its streets are among the safest in New York: pedestrians and cyclists are hurt there less often than almost anywhere — 50th of 59. Now cross the river to Stuyvesant Town and Turtle Bay, on Manhattan's calm East Side. Violent crime there is low — 43rd of 59, half Soundview's rate. Its streets are the 5th most dangerous in the city.
Which one is safe? The question doesn't have an answer, because "safe" isn't one thing. And it isn't only that the harms fall in different places — it's that they aren't even the same kind of harm. A killing, a broken arm, and a stolen car all get filed under "crime," as if they were points on one scale. They aren't. One is permanent.
This is the sixth chapter in a series measuring New York at the resolution that actually decides things. Chapter 1 measured the city in minutes, Chapter 2 in dollars over income, Chapter 3 in what the body absorbs by being there, Chapter 4 in whether the errands are enough, Chapter 5 in how much of the public realm you can have for free. This one is about harm — sorted, for once, by how permanent it is.
Why safety deserves its own chapter
Safety is the dimension people feel most and measure worst. It decides where you'll live, whether you'll let your kid walk to school, which subway car you'll pick, what time you'll come home — more viscerally than rent or commute time ever will. And almost all of that deciding runs on a single number: the crime rate. A neighborhood is "good" or "bad," "safe" or "sketchy," and the word points at one kind of event — the mugging, the shooting, the stranger.
But "the crime rate" is an average of things that should never be averaged. It adds a homicide to a stolen phone and reports the sum. It leaves out the car entirely. And it flattens the one distinction that matters most to an actual human being: whether the harm ends — a life, a body's use — or whether it's a bad week you recover from. This chapter refuses the average. It sorts harm into three tiers by permanence, and looks at the most permanent one first.
- Death — irreversible. A person is lost forever.
- Injury — bodily, serious, but usually survived and recovered from.
- Property loss — no body touched at all; a cost, not a wound.
Measured this way, the sources of harm pull apart — into different events, in different neighborhoods, falling on different people.
Death: the harm that doesn't heal
Start with the tier that can't be reversed. Over 2022–2024, New York averaged 394 murders a year and 279 traffic deaths a year — pedestrians, cyclists, and drivers combined. Those are the same order of magnitude. The road kills about seven New Yorkers for every ten who are murdered, and if you count only the deaths outside a car — the pedestrians and cyclists — it still kills 147 a year.
| Cause of death | Deaths per year | Over 2022–2024 |
|---|---|---|
| Murder (& non-negligent manslaughter) | 394 | 1,181 |
| All road deaths | 279 | 838 |
| — pedestrians | 122 | 366 |
| — cyclists | 25 | 74 |
| — drivers / passengers | 117 | 350 |
Two comparable tolls, and only one of them organizes where people live, walk, and let their kids out. But the deeper break is who dies — because the rate a neighborhood posts is not the risk a given person carries.
The rate is not your risk
Murder is not distributed like weather. It concentrates — in specific disputes and small, overlapping social networks — and its victims are overwhelmingly young men (below), the demographic signature of dispute- and group-involved violence. So the citywide murder rate badly overstates the risk to a New Yorker who isn't tied to those networks. (The public data carries no gang or relationship flag, so victim demographics are the only in-data proxy — a strong signal, not proof. Outside evidence agrees: the NYPD's chief of department has estimated 65–70% of the city's shootings involve a gang or crew — an official's estimate, not a published table, but pointing the same way.)
Traffic death is the opposite: close to democratic. Its victims skew older and less male, and the most over-represented group is the elderly — about a sixth of the population but a third of pedestrian and cyclist deaths, and almost none of the murder toll. A slow walk across a wide avenue is the risk, and everyone who crosses shares it. Nobody ages out of traffic; you age into it.
Who the dead are:
| Murder victims | Traffic deaths (ped + cyclist) | |
|---|---|---|
| Male | 82% | 65% |
| Under 45 | 72% | 34% |
| 65 or older | 6% | 32% |
So the fear is half-right. Murder is the city's single biggest killer here — fear isn't hallucinating it. But the murder rate is a poor guide to your danger: it falls on a small, identifiable slice of the city, while the risk nobody fears — the car — is spread across everyone, hitting the grandparent on the corner as readily as anyone. For most people, most of the time, the unfeared risk is the more personal one.
What the city spends to prevent each
If two harms kill comparable numbers, you might expect comparable effort against them. On the money spent specifically to prevent each, they're actually close. New York's Vision Zero street-safety program runs about $260 million a year in operating funds on top of a $4.7 billion ten-year capital plan for redesigns and protected lanes — roughly $470 million a year of construction. The city's flagship anti-violence effort outside the police — the Crisis Management System of community "violence interrupters" — runs about $100 million a year, inside a broader Office of Neighborhood Safety of around $190 million. Same order of magnitude. By the dollars aimed directly at prevention, traffic and violence are funded comparably.
The asymmetry is somewhere else, and it's enormous. The city's dominant answer to violence isn't the interrupters; it's the NYPD — roughly $5.8 billion a year in operating budget alone, and closer to $12 billion once pensions and benefits are counted. Nothing remotely that size is pointed at the road. Two comparable death tolls: one is met by a twelve-billion-dollar institution, the other by a few hundred million in street redesigns. The comparison is deliberately unfair — the NYPD does far more than prevent homicide, there's no clean "homicide-prevention" line to isolate from it, and one-time capital dollars aren't annual operating dollars — but the order-of-magnitude gap between how the city answers the two harms is real, and the body count doesn't explain it.
Injury: the harm we recover from
Below death sits the tier that fills most of the "crime rate" and almost none of the fear budget: injury. Here the numbers invert. Against ~394 murders, the city logs roughly 47,000 violent felonies a year — the assaults and robberies that hurt without killing. Against 279 road deaths, it logs about 13,000 pedestrians and cyclists injured a year. Injury is where traffic stops being a footnote and becomes a flood.
And injury is where the geography turns strange — where the map of crime and the map of danger stop matching. The map below carries all three harms as neighborhood rates. The Fear vs. harm tab colors each district by which corner it lands in; Violent, Traffic, and Property are the underlying rates. Flip between them and watch the bright districts move around the city.
Violence pools in the South Bronx and central Brooklyn — the historically disinvested districts, where it tracks concentrated disadvantage. Traffic harm pools in the dense Manhattan core, where it tracks exposure: the more people on foot and the more cars turning through them, the more casualties, so the crowded, walkable center is the most dangerous and the car-dependent outer edges are the safest. Property crime pools in the commercial cores, where there's more to take. Three harms, three forces, three different maps.
One test makes it concrete. If violence and traffic were the same danger, the districts would line up on a diagonal — high on one, high on the other. Put violent crime across the bottom and traffic harm up the side, and instead they scatter (Spearman ρ = +0.48).
The upper-left quadrant — low crime, dangerous streets — is not empty. Ten districts live there: the calm, well-regarded parts of Manhattan and brownstone Brooklyn where the reputation is "safe" and the crosswalks are anything but. And the far upper corner, alone, is Midtown, off the chart for the same reason it's off every chart in this series — almost nobody lives there, so its handful of residents carry the crash toll of the millions who pass through. (Even normalized per street-mile instead of per resident, Midtown is still the most dangerous district in the city, so it's the crowd, not the arithmetic.)
Property: harm to your stuff, not your body
The third tier barely belongs in the same word. Burglary, grand larceny, and car theft take your things, not your body — a cost you insure against and absorb, not a wound. It's the largest of the three by sheer count (about 79,000 major property felonies a year), and it has its own geography again: it pools in the commercial cores and downtowns, where there's the most worth taking, and it's the one harm that's often highest in the districts that are lowest on violence and traffic death. A neighborhood can be the safest place in the city to keep your body and a middling place to keep your car. It's carried here as the third axis on the map above, and then mostly set aside — because the chapter is about harm to people, and this is harm to property.
Five neighborhoods
Five community districts, one per borough, at the corners of the argument. Violent is major violent felonies per 1,000 residents a year; traffic is pedestrians and cyclists hurt per 1,000; property is major property felonies per 1,000. Two of these — Brownsville and Bayside — anchored Chapter 2 too; the series keeps circling the same neighborhoods.
| CD | Neighborhood | Boro | Violent /1k | Traffic /1k | Property /1k | The story |
|---|---|---|---|---|---|---|
| 209 | Soundview / Parkchester | BX | 8.4 | 1.1 | 11.1 | Feared, not deadly |
| 106 | Stuyvesant Town / Turtle Bay | MN | 3.2 | 2.5 | 9.9 | Calm, but deadly |
| 316 | Brownsville | BK | 10.5 | 1.7 | 7.1 | The feared archetype |
| 411 | Bayside / Little Neck | QN | 1.4 | 0.7 | 9.0 | Safe both ways |
| 503 | Tottenville / Great Kills | SI | 1.0 | 0.6 | 3.7 | Safest in the city |
Use the tabs below to explore — selecting an anchor highlights it on the map and surfaces its detail.
Soundview / Parkchester (CD 209) is the chapter's cold open and its cleanest inversion. Violent crime is high — 8.4 per 1,000, 14th of 59 — the number that would get it called a "bad neighborhood." But its streets are nearly the safest in the city: 1.1 traffic casualties per 1,000, 50th of 59. It's a transit-and-walking Bronx district where relatively few residents drive and the crossings are calmer than the reputation. The danger here is real and it is the danger people mean. It just isn't the only kind, and on the other kind this "dangerous" neighborhood is one of the safest places in New York.
The contradiction: fear is aimed at the wrong risk
Stack the tiers back up and the mismatch is stark. Fear is trained on violent crime — a harm that is real, concentrated, and, at its most permanent, the city's leading killer. But fear treats the whole category as one thing, so it can't tell the killing from the stolen car, and it can't see the harm that lives in the next tier over: the road, which injures thousands and kills hundreds, and does it in the calm, walkable, low-crime districts that read as safe.
There's a reason for the blind spot, and it isn't stupidity. Violent crime is authored — someone did it to you, on purpose — and the mind files intentional harm as danger far more readily than an accident. A car making a legal turn through a crosswalk doesn't feel like a threat the way a stranger stepping toward you does, even when the odds say otherwise. So we sort neighborhoods by the harm that feels like danger and move toward the quiet, high-traffic districts — toward the exact risk we've stopped seeing.
The emptiest corner of the map makes the point. The very safest districts — Tottenville, Bayside — are safe on every tier at once, and what they share isn't good policing or street design. It's low density: few people, few crowds, few pedestrians. They're safe in large part because they're empty — the same emptiness that gave them the worst transit and the longest trip to a carton of milk in the whole series. The safest-feeling New York is the least urban New York, and it buys that feeling with a car in every driveway — its own kind of risk, just one that happens on the highway instead of the block.
Which is why this chapter ships no single safety score. Averaging a death, an injury, and a theft into one number is exactly the move that produced "the crime rate" — the flattening this whole chapter is trying to undo.
What this means for a person
- "Safe neighborhood" is a question, not an answer — ask safe from what? The crime rate answers one tier of one harm. If you walk, bike, or have kids or aging parents who do, it can point you toward traffic danger, not away from it. Look up both maps; they're rarely the same place.
- In a calm, dense district, the likeliest way to be seriously hurt is the crossing, not the stranger — a street-design and speed problem, the kind Vision Zero targets, not a policing one.
What this measure misses
- Reported ≠ occurred. Crime here is reported crime; under-reporting varies with how much a community trusts the police, so a low violent-crime rate can partly reflect low reporting. Traffic harm is far less prone to this — a hospitalization is recorded regardless of trust.
- No gang or relationship flag. The public data can't say whether a killing was a stranger attack or a dispute between people who knew each other. The victim-demographic concentration is a proxy for that distinction, not a measurement of it.
- Deaths are counted citywide, not mapped. With ~147 pedestrian/cyclist deaths a year spread over 59 districts, a per-neighborhood death rate would be mostly noise — so the neighborhood maps use the broader killed-or-injured measure, and the death comparison is kept citywide.
- Injured, not "severely." The public crash data has no injury-severity grade, so the traffic tier counts pedestrians and cyclists killed or injured, not the stricter "killed or severely injured" standard Vision Zero uses.
- Residents aren't the exposed population. Per-resident rates overstate danger in commercial cores like Midtown, where the daytime crowd dwarfs the few thousand residents. The chapter cross-checks traffic per street-mile, where the dense-core danger holds up anyway.
A fuller accounting — the exact filters, the seven-felony definitions, and the traffic-exposure math — is in the methodology.
Reach for the next chapter
Safety, sorted by permanence, turns out to be three harms wearing one word — and the places safe from the harm we fear are often the most exposed to the harm we ignore. Chapter 7 turns to the engine under all of it: economic opportunity — who can reach the jobs, build the income, and actually climb, and whether the neighborhoods with the most to fear are also the ones with the least room to move.