Every year, the same jump
Crashes from 5 to 7 pm in the 3 weeks before the clocks fall back, and the 3 weeks after, in Jersey City, Hoboken and Hackensack.
- 20192019: 103 then 119+16%
- 20202020: 48 then 61+27%
- 20212021: 72 then 82+14%
- 20222022: 59 then 96+63%
- 20232023: 58 then 86+48%
- 20242024: 78 then 88+13%
- 20252025: 82 then 86+5%
3 weeks before3 weeks after
Up in 7 of 7 years. Every other hour of the same days: −2% over all 7. Source: Counselworthy analysis of NJDOT crash records.
Only when the clocks change
The same count, 5 to 7 pm in the 3 weeks after against the 3 weeks before, around Sundays when nothing changes. If the season alone pushed evening crashes up, these would rise too.
- 6 weeks earlier−5%
- 4 weeks earlier+6%
- The clock change+24%
- 4 weeks later−7%
- 6 weeks later−19%
Each fake date is a Sunday 4 or 6 weeks from the first Sunday of November, 2019 to 2025. None rose more than 6%. Source: Counselworthy analysis of NJDOT crash records.
In March, it mostly runs the other way
When clocks spring forward, 6 to 8 pm goes from dark to half daylight. Crashes in those hours fell 18%, 390 to 318, while every other hour rose 4%. But they fell in only 5 of 7 years, so it's a pattern, not a rule.
6 to 8 pm, 3 weeks before3903 weeks after318Years2019 to 2026, not 2020
- A crash in daylight
- A crash after dark
- Sunset
- 6 to 8 pm
City by city
Crashes from 5 to 7 pm, 3 weeks before and after the clocks fall back, 2019 to 2025.
| City | 3 weeks before | 3 weeks after | Change | Years it went up | Every other hour |
|---|---|---|---|---|---|
| Jersey City | 341 | 400 | +17% | 5 of 7 | 0% |
| Hoboken | 32 | 38 | +19% | 3 of 7 | 0% |
| Hackensack | 127 | 180 | +42% | 7 of 7 | −8% |
| All 3 | 500 | 618 | +24% | 7 of 7 | −2% |
A small count moves a lot by chance: Hoboken's went from 32 to 38. Source: Counselworthy analysis of NJDOT crash records.
Every finding, with its source
One sentence each. These are our own counts from public records: please credit Counselworthy and link to this page, or to the finding, when you use them.
In the 3 weeks after clocks fall back, crashes from 5 to 7 pm rise 24%: 618 against 500. (Counselworthy analysis of NJDOT crash records, 2019 to 2025)
Crashes at every other hour of those days fall 2%: 3,253 against 3,330. (Counselworthy analysis of NJDOT crash records, 2019 to 2025)
The 5 to 7 pm count in the 3 cities combined went up after the change in 7 of 7 years, 2019 to 2025. (Counselworthy analysis of NJDOT crash records, 2019 to 2025)
Before the change, 260 of the 500 crashes from 5 to 7 pm happened in daylight. After, none did. (Counselworthy analysis of NJDOT crash records, 2019 to 2025)
Sunset in Jersey City moves from about 5:51 pm to 4:49 pm overnight. (Counselworthy analysis of NJDOT crash records, 2019 to 2025)
Leave out Halloween, always in the weeks before, and crashes a day from 5 to 7 pm still rise 24%. (Counselworthy analysis of NJDOT crash records, 2019 to 2025)
Crashes from 5 to 7 pm that hurt someone went from 117 to 111; those with no injury reported, 383 to 507. (Counselworthy analysis of NJDOT crash records, 2019 to 2025)
Around dates 4 and 6 weeks from the change, the same count rose 6% at most. (Counselworthy analysis of NJDOT crash records, 2019 to 2025)
Jersey City: crashes from 5 to 7 pm rise 17% after clocks fall back, 400 against 341, up in 5 of 7 years. (Counselworthy analysis of NJDOT crash records, 2019 to 2025)
Hackensack: crashes from 5 to 7 pm rise 42% after clocks fall back, 180 against 127, up in 7 of 7 years. (Counselworthy analysis of NJDOT crash records, 2019 to 2025)
In the 3 weeks after clocks spring forward, crashes from 6 to 8 pm fall 18%: 318 against 390. (Counselworthy analysis of NJDOT crash records, 2019 to 2026)
That 6 to 8 pm count fell in 5 of 7 years counted, so it's a pattern, not a rule. (Counselworthy analysis of NJDOT crash records, 2019 to 2026)
What we can't say
- That more people got hurt. Crashes from 5 to 7 pm that hurt someone went from 117 to 111, so the rise is in crashes where police reported no injury.
- Anything about people on foot. Crashes from 5 to 7 pm with someone on foot went from 29 to 35: too few to tell a change from chance.
- That every city shows it every year. It's clearest in the 3 cities combined. In Jersey City alone, the 5 to 7 pm count rose 17%, up in 5 of 7 years; Hoboken was up in 3 of 7 and Hackensack in 7 of 7.
- Why, for certain. Most work and school hours run by the clock, so the 5 to 7 pm rush likely keeps its hour when the clocks change while the light moves. But we don't count how many people drove each evening, so this can't say any one trip got riskier.
- That it's the same every March. Springing forward, the 6 to 8 pm count fell in 5 of 7 years. That's a pattern, not a rule.
Sources and how we counted
Sources
- NJDOT Crash Data Portal (police crash reports, NJTR-1) · saved October 7, 2026
How we counted
- Every crash NJDOT's Crash Data Portal lists for Jersey City, Hoboken and Hackensack, 2019 to 2026-09, the same records as each city's crash data page. A crash with no time on its report is left out.
- Each clock change: the first Sunday of November (fall back) and the second Sunday of March (spring forward). The 21 days from that Sunday on are "after", the 21 days before it "before", in every year whose whole 42 days we have.
- The evening: 5 to 7 pm when clocks fall back, the two hours that go from half daylight to dark. 6 to 8 pm when they spring forward, the two hours that go from dark to half daylight. Every other hour of the same days is the comparison.
- Daylight and dark: each crash against sunset on its own date in its own city, worked out with the same sunrise equation as the crash data pages. In Jersey City, sunset moves from about 5:51 pm to 4:49 pm the night clocks fall back, and from about 5:57 pm to 6:58 pm when they spring forward.
- A year is set aside when crashes at every other hour moved by more than a third across its change, because something besides the clock changed the roads that month. That's March 2020: the rest of the day fell 41%.
- The fake-date test: the same count around Sundays 4 and 6 weeks before and after the real one, when nothing changes.
- Halloween always falls in the 3 weeks before. Leaving it out shortens that window by a day each year, so we compare crashes a day: the 5 to 7 pm rise is still 24%.
- Before a finding goes on this page it has to hold every one of these: the evening moves at least 10% the way the dark predicts, the rest of the day moves less than half as much, at most 1 year goes the other way, no fake date moves it as much, and chance alone would give a split that lopsided less than 1 time in 100. Falling back passes all of them. Springing forward misses one, so it's here as a pattern.
- NJDOT lists far fewer Hoboken crashes from February 2023 on, so Hoboken adds less to the later years. Each year's before and after come from the same records, so a gap thins both sides alike.
- NJDOT calls these records preliminary: crashes can take weeks to appear and can change. These are police reports, so a crash nobody reported isn't here.
- Crash numbers never change a law firm's score.
Use these numbers
Free to reuse under CC BY 4.0: credit Counselworthy and link to this page, or to the finding you use. Reporters and researchers can email hello@counselworthy.com.
Download the counts (CSV)Counselworthy, "Daylight saving time and car crashes in New Jersey," updated November 4, 2026, from NJDOT crash records. counselworthy.com/nj/research/daylight-saving-time-crashes/