Fewer shootings, more assaults: thirty-five years of violence in New York

Closely. The shooting files give exactly the 561 incidents and 703 victims that the weekly report shows for the same period of 2025, 1 January to 4 October, and for January to June of this year the 322 incidents and 381 victims of the department’s first-half release. These are agreements of administrative counts, exact or within a stated tolerance, not sampling margins.

Felony assaults counted from the complaint files by the month they were reported come to 29,841 for 2025, against 29,838 in the department’s yearly table. For earlier years the two differ by about 1% (20,898 against 20,698 in 2019): the table was fixed early in the following year and the complaint file has been revised since, so the yearly series in the charts is the table’s, and the complaint files carry only the breakdowns by code, place and borough.

The State’s murder counts and the NYPD’s are equal in every year from 2001 to 2022, which is why the long murder line can join them. The shooting file dates a death by the day of the shooting, the yearly table by when it was recorded, so the share of murders that were shootings is approximate.

Yes, several times, and each change adds cases that would once have been counted as misdemeanours or other crimes. Strangulation has had a felony code of its own in the NYPD’s files since November 2010, and the NYPD counts it as felony assault. Assaulting a person aged 65 or over first appears among felony-assault arrests in 2011, and assaults on traffic agents, school safety agents and other public employees have had codes of their own since 2017; these are dates in the NYPD’s files, not the dates the laws took effect. Since 2019 the law has widened it three times, none with a code of its own in the files: to more transit workers, such as station customer assistants, fare collectors and the workers who maintain buses, tracks and stations, from late September 2022 (Chapter 233 of 2022); to traffic checkers and ferry workers from July 2024 (Part Z of the 2024 budget bill); and to retail workers in late 2024 (Part A of the same bill, in force 180 days after it was signed in April).

Of the rise since 2019 in the complaint files, the catch-all code carries 71%, strangulation 15%, assaults on police and peace officers 11% and the other public employees 3%, and each of the four groups rose. The catch-all code also holds cases the law has added, such as assaults on people aged 65 or over, which the complaint files don’t separate, so leaving out strangulation removes only one of the widenings.

Set each against its own level in 2019 and the three lines part: by 2025 felony assaults stood at 144, shootings at 89 and murders at 97, about where they were. Set them against 2006 instead, and shootings are less than half and murders about half, while felony assaults are about 70% higher.

No. If assaults had simply moved from the misdemeanour count to the felony one, misdemeanour assaults would have fallen. They rose too, by 8% from 2019 to 2025, far less than felonies, and they remain below their level of 2006. That rules out a simple swap, but it can’t measure how many cases the law’s widenings moved.

No. Per 100,000 residents, felony assaults rose from 247 in 2019 to 355 in 2024, and they were 335 in 2000. Murders fell from 30.7 per 100,000 in 1990 to 4.6 in 2024. The population figures are the ones the State uses for its own crime rates, which stop at 2024, so 2025 stays a count. They come from different vintages of the Census Bureau’s estimates, which don’t join exactly, so these rates show the scale of the change rather than measure it precisely.

All five. From 2019 to 2025 felony assaults rose 51% in the Bronx, 39% in Brooklyn, 34% in Manhattan, 43% in Queens and 64% on Staten Island. These are the boroughs’ totals; we don’t split each borough by place.

About half in homes: 42% in 2006, 52% in 2019 and 51% in 2025. The street’s share fell from more than a third in 2006 to a quarter, but since 2019 the two grew at about the same pace. On the subway, felony assaults rose from 377 in 2019 to 584 in 2025.

In homeless shelters, which have had a code of their own since 2018, they went from 157 in 2019 to 861 in 2025. The City’s daily census of its own shelters averaged about 47,000 people from March to December 2021, when its open file begins, and about 86,000 in 2025. There is no figure for 2019 in the file, and the complaints’ shelter code may include places the census leaves out, so this is a count, not a rate.

The files can’t say exactly. In 2025 the people wounded in shootings came to the equivalent of 2.4% of all felony assaults recorded, about one in forty. The files don’t link a shooting to its complaint, and they date them differently, the shooting by the day it happened and the complaint by the day it was reported, so this gives the scale, not a share. The complaint files don’t record the weapon, so the rest can’t be split by knife, blunt object or bare hands.

The data can’t separate more violence from more reporting or more recording. Complaints are what victims and witnesses report and the police record, and arrests follow where officers are and what they arrest for. Felony-assault arrests rose too, from 15,208 in 2019 to 22,964 in 2025, and arrests charged under the section of second-degree assault for injuring someone with a deadly weapon or dangerous instrument, the law’s phrase for anything from a knife to a bottle, from 4,501 to 7,974. Assaults on police officers are among the groups that grew.

What we can say is that the rise is long and broad: it shows in each borough, in each kind of place we can count and in each of the four offence groups.

No. Shootings have fallen since 2021 and felony assaults have risen since 2020, so both run across more than one administration, and many things changed at once, from the courts to the pandemic’s aftermath to policing and programs. The counts can’t test any one of them, and we don’t try.

The NYPD’s open files begin in 2006. The department’s own records go back further: its weekly report puts this year’s shootings 88% below the same weeks of 1993. In January to June of this year, the latest the open files reach, 48 people under eighteen were shot.

The NYPD publishes its yearly table from 2000. For the murders of the 1990s we use the State’s count from the Division of Criminal Justice Services, which follows the FBI’s definitions; it begins in 1990 with 2,245 murders, against the NYPD’s own complete-year figure of 2,262 for that year in its weekly report. The two series are equal in every year from 2001 to 2022, so we switch at 2001. They differ from 2023 (404 against 391 that year, 357 against 382 in 2024), and the State’s dataset page gives no reason, so we treat those years as a break and use the NYPD’s count throughout the 2020s.

For felony assaults before 2000 we use the complete-year figures in the NYPD’s weekly report. Where it overlaps the yearly table they can differ: for 2001 the report gives 23,020 against the table’s 23,453, about 2% less. The 2025 figure is the same in both, so the comparison with 1998, which the report draws itself, stands.

The State’s aggravated assault follows the FBI’s definition rather than New York’s penal law, and from 2000 to 2022 it counted about half again as many as the NYPD’s felony assault, so it never shares a line with it.

Shootings. The NYPD’s Shootings and Shooting Victims files (NYC Open Data), January 2006 to 30 June 2026. An incident is a shooting in which a bullet struck someone; victims join to incidents on the incident key. The victims file repeats 164 of the 381 victims of January to June 2026 as exact copies, and two of its rows have no incident in the incident file (one of them blank): we keep each victim once and drop the two. In the incident file’s rows for 2006 to 2025 the latitude and longitude columns are swapped, and from 2026 the Bronx is coded as two patrol boroughs; neither affects a count, and both are handled in the reader.

This year. The CompStat citywide report for the week of 28 September to 4 October 2026, saved as published and archived by the Internet Archive on 5 October: this year to date, 1 January to 4 October, against the same dates of 2025, preliminary. Its historical table gives complete years.

The long series. The NYPD’s seven major felonies and its misdemeanours by year, 2000 to 2025, under New York’s penal law; for murders before 2001, the State’s county counts summed over the five boroughs; for rates, the State’s population figures.

Felony assaults. The complaint files, counted on the City’s server by the month each complaint was reported (as the NYPD’s own tables count) and by offence code, premises and borough; the arrest files by the section of the penal law charged. “Without strangulation” leaves out the strangulation code from 2006; strangulation was not a felony of its own before November 2010, so the NYPD’s table for 2000 to 2005 needs no adjustment. Complaints whose premises is unrecorded count as “everywhere else” (393 in 2025, none in 2019); complaints with no borough are left out of the borough figures (1 in 2019, 2 in 2025). We compare with 2019, the last full year before the pandemic.

Chance. These are counts of events, not samples. A change is called a rise or a fall only when an approximate 90% interval for the ratio of the two counts excludes no change, under a simple model that treats each count as an independent Poisson count over the same span. That model leaves out changes in reporting, classification and population, and any extra year-to-year variation, so it understates the uncertainty and is not a prediction band. Under it: felony assaults from 2019 to 2025, 42% to 46% more; murders this year against last, 32% to 7% fewer; shooting incidents, 20% to 2% fewer; felony assaults, between 1% fewer and 2% more, about the same; the Bronx’s shootings, 252 against 218, from 1% fewer to 35% more. Single-year steps (a fall from 2023 to 2024, say) are not claimed; we name only the longer changes that clear the interval.

Code and checks. One script, analysis.py (linked below), produces every figure and chart from these files and checks each figure it quotes against the saved copy of its report. The dataset cards behind it (nypd_shootings, nypd_crime_history, dcjs_index_crimes) list the traps.