Pedestrian Fatal Crash News ObservatoryNEWS-DETECTED EVENTS · 2016–2025
PUBLIC AGGREGATE VIEWLOCAL DERIVED

A decade of people killed walking

13,426 pedestrians were killed in 12,878 crashes reported across 51 states between 2016 and 2025. Where the reporting says, about half the drivers left them there, and close to two in three were struck in the dark, with the toll climbing every December as the evening commute moves into the night.

13,426pedestrians killed, 2016–2025
50%hit-and-run: 5,968 of 12,023 crashes where reporting says either way
64%struck in dark conditions: 5,930 of 9,211 where light condition is stated

UNITED STATES · 2016–2025
12,878 EVENTS IN VIEW

Crash events
12,878Deduplicated physical events. The only unit that counts crashes.
Linked articles
18,841News records linked to an event. More articles is not more crashes.
Independent stories
18,061Near-duplicates collapsed. Syndication is not corroboration.
Multi-source
2,834Events with more than one independent account.

The decadeClick any year to filter the whole dashboard to it. Click it again to clear.

United States · 2016–2025

event
01,0002,000'16'17'18'19'20'21'22'23'24'25
Canonical eventsOf which multi-source

A rise means more covered crashes were detected: real change blended with shifts in how much news coverage existed and was indexed. It is not a fatality trend.

The statesEvery state is one equal cell, so a small state stays as readable as a large one. Click a cell to filter the dashboard.

Equal-cell cartogram

event
FEWERMORE

Counts, not rates. No exposure denominator exists, so a populous state looks large partly because it is populous.

When people are killedClick any cell or year label to filter the dashboard to that year.

The one circumstance this corpus knows exactly, with no model and no geocoding, and the one with a strong, known signal.

event
Every month of the decade · national
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
44169events per month
Winter multiplier
1.52×

Dec (1,363) against Apr (894) in the United States.

The winter peak is the darkness effect: after the autumn clock change the evening commute happens in the dark. It is the strongest and most consistent signal in this corpus, and it needs no model: event_date alone gives it.

Monthly profile · national
Jan: 1,113JFeb: 988FMar: 954MApr: 894AMay: 993MJun: 934JJul: 1,022JAug: 1,042ASep: 1,140SOct: 1,208ONov: 1,227NDec: 1,363D

Y-axis floors at the trough, not zero, because a zero baseline hides a real swing.

Day of week · national

Saturday carries the most, Wednesday the fewest.

Mon1,885
Tue1,702
Wed1,695
Thu1,880
Fri1,957
Sat2,014
Sun1,745