State-highway speed reversal · Nelson–Marlborough
SH6 Blenheim to Nelson: before, during and after
SH6 north of Nelson dropped from 100 km/h to lower limits in December 2020, then went back to 100 km/h in June 2025. It is the only Nelson-Tasman state highway that was lowered and then raised again.
119 crashes in the 5 years before (≈24/yr) fell to ≈13/yr over the 4½ lower-limit years (rate ratio 0.53, 95% CI 0.38–0.73, p = <0.001).
Death and serious-injury crashes fell from 13 before (including 4 fatal) to 3 during the lower limits. The fall is statistically clear (rate ratio 0.26, 95% CI 0.05–0.94, p = 0.025).
Limits only went back to 100 km/h in June 2025. The 1 serious crash and 17 crashes since are too few to read anything into, and 2026 records are still incomplete.
01 / The pattern
What the corridor record shows
The three periods run 5 years, 4½ years and about 1 year, so the fair comparison is crashes per year. On that measure every type of crash was lower under the reduced limits.
02 / The test
Is the fall more than chance?
An exact Poisson rate-ratio test compares the lower-limit period with the 5 years before, accounting for the different period lengths. Totals and deaths-and-serious-injuries clear the usual significance threshold; injuries fall short of it on these small numbers.
| Outcome | Before (n) | per yr | Lower (n) | per yr | Change | 95% CI on ratio | p |
|---|---|---|---|---|---|---|---|
| All crashes | 119 | 23.8 | 56 | 12.5 | 47% fewer | 0.38–0.73 | <0.001 |
| Injury crashes | 44 | 8.8 | 25 | 5.6 | 36% fewer | 0.37–1.06 | 0.071 |
| Deaths & serious injuries | 13 | 2.6 | 3 | 0.7 | 74% fewer | 0.05–0.94 | 0.025 |
| Serious-injury crashes | 9 | 1.8 | 3 | 0.7 | 63% fewer | 0.07–1.50 | 0.154 |
| Fatal crashes | 4 | 0.8 | 0 | 0.0 | 100% fewer | 0.00–1.70 | 0.127 |
Counts are crashes (not casualties), combined across the three sections, from NZTA’s OIA release. “Change” is the annual-rate difference; a 95% interval that stays below 0% means a statistically clear reduction.
03 / Stress tests
Is it just less traffic, or COVID?
A crash count can fall simply because fewer vehicles use the road. Three checks say that is not what happened here.
Traffic barely changed
-6%Change in corridor traffic during the lower-limit years (NZTA telemetry at Hira and Atawhai). With traffic this flat, the drop is not just fewer vehicles.
Per vehicle-km, still clear
44% fewer (p = <0.001)All crashes, once traffic is divided out. Death and serious injuries: 73% fewer (p = 0.042). Both stay statistically clear.
Not a COVID effect
60% lowerCounting only 2022–24, after traffic returned to normal, against 2018–19: 10 versus 26 crashes a year. The drop outlasts the COVID years.
One more sign: the share of crashes that killed or seriously hurt someone fell from 11% before to 5% during the lower limits. That points to less severe crashes whatever the traffic, though on so few crashes the shift on its own is not significant.
04 / On the map
Every reported crash on the corridor
The 127 crashes the CAS download places on this corridor, 2018 to 2026. Colour is the speed-limit period; bigger ringed dots are crashes that killed or seriously hurt someone. Toggle the periods, or click a dot for its date.
The dot counts sit a little below NZTA’s official totals: the download is rough at the urban Atawhai end and stops at the Nelson-Marlborough boundary. The pattern is the point, not the exact count.
05 / Annual trend
The raw download agrees
Counted straight from the CAS download, not from NZTA’s tables, the yearly totals show the same shape: high before 2021, a long drop under the lower limits, and only a part-year since the reversal.
06 / By section
The three sections
Each section was lowered on the same dates but to a different limit. Deaths and serious injuries, then all crashes, by period.
| Section | DSI before | DSI lower | DSI raised | All before | All lower | All raised |
|---|---|---|---|---|---|---|
| Rai Valley to Whangamoa Saddle16.85 km · lowered to 80 km/h | 4 | 0 | 0 | 50 | 22 | 11 |
| Whangamoa Saddle to Hira7.82 km · lowered to 60 km/h | 7 | 2 | 0 | 53 | 19 | 2 |
| Hira to Atawhai6.95 km · lowered to 80 km/h | 2 | 1 | 1 | 16 | 15 | 4 |
07 / How to read this
What this can and cannot say
One rural corridor with few crashes can only show so much. Five things to keep in mind.
- One corridor, not a region. NZTA confirmed no other state highway in Nelson-Tasman was lowered then raised, so there is no comparison road and no pooling. The death-and-serious-injury result rests on 13 crashes versus 3.
- Speed was not the only change. Over the same years NZTA also added guardrails, rumble strips, wider shoulders and motorcycle barriers on SH6 north of Nelson. The lower crash rate reflects the limit and these works together, and on one corridor the two cannot be pulled apart. Only the shift to less severe crashes points to speed on its own, and that signal is weak.
- Traffic and COVID are already accounted for. Adjusting for traffic barely moves the numbers, and the drop holds using only the post-COVID years. Neither explains it (see the stress tests above).
- The reversal is too recent to judge. Limits returned to 100 km/h on 6 June 2025. Just over a year of data is in, and 2026 is still incomplete in CAS, so the post-reversal counts are far too small for a reliable comparison.
- The numbers are NZTA’s own. Period counts come from the OIA release, which used precise corridor geometry. The CAS download only backs up the trend.