Taxi Crashes in NYPD Data: When and Where They Happen

In short: NYPD's collision data lists 2,608 crashes in 2025 where the first vehicle was a "Taxi" — about 3.0% of the 85,546 crashes recorded citywide that year. Those crashes injured 1,805 people and killed 4 (3 pedestrians and 1 motorist). Manhattan had more than half of the crashes with a borough recorded; Saturday and Sunday were the busiest days; and the most common cause police recorded for the taxi was driver inattention or distraction. App cars (Uber, Lyft) mostly don't show up as "Taxi" in this data, so read it as a picture of yellow and green cabs, not all for-hire vehicles.

These figures come from our own query of NYC Open Data's Motor Vehicle Collisions – Crashes dataset (h9gi-nx95) on 2026-10-12; the latest crash in the data then was dated 2026-06-11. Method and limits are at the bottom.

By hour of day

Crashes climb through the afternoon and stay high into the night: 7 PM (159) and 10 PM (141) are the busiest hours after midnight, and 5 AM (48) is the quietest. The midnight bar looks high partly because 52 crashes are stamped exactly 0:00 — a value that can stand in for an unknown time — so don't read it as the most dangerous hour.

Part of the pattern simply follows traffic: more taxis are working in the evening. This data has no trip or mileage counts, so it can't tell you the crash rate per hour.

By day of the week

Saturday and Sunday each had 411 crashes; Monday had the fewest (336).

By borough

Of the 2,231 crashes with a borough, Manhattan had 53%, Brooklyn 21%, the Bronx 13% and Queens 12%. Yellow taxi pickups are far more concentrated: in TLC trip records for September 2025 – August 2026, about 87% of yellow taxi pickups were in Manhattan (dataset trips-pickups-by-borough, below). Crashes happen anywhere along a route, and the "Taxi" label in NYPD data may include green cabs and other vehicles, so this gap is not proof that outer-borough driving is riskier.

What police recorded as the cause

Each crash record has a "contributing factor" for each vehicle. For the taxi (vehicle 1), the most common in 2025:

A recorded factor is the reporting officer's entry, not a court finding of fault. Still, the top entries map onto rules TLC drivers already know: distraction (see the phone rule), failure to yield, tailgating and speed (see NYC speed limits) are all TLC hazardous moving violations with points.

Over the years

Taxi crashes dropped sharply in 2020 and have stayed between about 2,550 and 2,750 a year since 2022. All crashes in the dataset fell in 2020 too, and the taxi share went from 4.6% in 2019 to 2.5–3.3% since. This dataset alone can't say why. TLC's own crash data shows a similar step down in 2020:

What this data can't tell you

• Uber and Lyft cars are hidden. Police record vehicle types such as "Sedan" or "Station Wagon/Sport Utility Vehicle". Most app cars are recorded that way, so "Taxi" mostly means yellow and green cabs. For all TLC vehicles, use TLC crash statistics.
• Vehicle 1 isn't "the one at fault". We counted crashes where the taxi is listed first, to match the factor recorded for it. Counting a taxi anywhere in the crash gives 3,692 for 2025.
• No rates. There are no miles or trips in this dataset, so counts can't be turned into risk per trip.
• Missing and default values. 14% of 2025 taxi crashes have no borough; 52 have a 0:00 time stamp.
• Lag and revisions. The dataset is updated over time, so recent months can change. We use full-year 2025 only.

Method. Socrata API queries on h9gi-nx95 with crash_date in 2025 and upper(vehicle_type_code1) = 'TAXI' (one record used lowercase "taxi"); hour from crash_time; weekday from crash_date. Year totals use the same filter for 2016–2025.

If you are in a crash, the reporting steps are in crash reporting for TLC drivers, the deadlines in our crash deadline planner, and the leaving-the-scene rule at 80-13(a)(4).