Speed Enforcement Effect — hills, cameras and the speeds drivers actually choose

Same hill, same limit — one has a camera.
Drivers choose very different speeds.

NSW fixed speed cameras × connected-vehicle Operating Speed data (June 2026) × matched-grade controls. Descriptive speed association — not a crash or causal study.

Flat Rock Dr — no camera 6.1% grade

p85 76

steepest section, 60 zone — tail 16 over; medians up to 68

Eastern Arterial — camera 6.2% grade

p85 58

whole distribution under the same 60 limit

At-camera point dip 17 strict pairs

+0.0

median gap vs same road 1 km away — corridors are uniform

The naive-baseline trap exhibit

+12

selection artefact, published as a warning — not an effect

Where there's no #Camera, gravity wins!

What a 6% hill does to speed compliance - and what one camera changes.

Two Sydney hills. Both ~6% grade. Both 60 km/h zones. One has a fixed #speed_camera.

On Eastern Arterial Road (camera, Gordon): 85th-percentile operating speed 55-58 km/h. The whole distribution sits UNDER the limit - on a 6% descent.

On Flat Rock Drive (no camera, Naremburn-Northbridge): p85 61-63 along the road, but on the steepest section the tail hits 76 with medians around 68 - in month-aggregated #Operating_Speed data that smooths peaks (spot speeds run higher).

Same geometry. Same limit. The difference is what #road_users get away with.

This is a descriptive matched pair, not a causal study - and the wider picture is consistent: across 17 strict same-road camera/control pairs we find NO localised dip at cameras (median gap +0.0 km/h). Corridors with cameras read uniformly paced; steep corridors without them grow a fat, fast tail downhill.

Interactive evidence, imagery, the bias register, and an agent that only answers what it can verify: https://enforcement.crashstats.com.au

#RoadSafety #DriverBehaviour #SpeedManagement #SpeedCameras #OperatingSpeed #VisionZero #SafeFutureLab #AI_Agent

The matched-grade pair mart_extreme_pair · Operating Speed data · Jun-26 · OSM · DEM

site	camera	grade	limit	median	p85 (tail)
Flat Rock Drive, Naremburn→Northbridge (steepest section)	no	~6.1%	60	68	76
Flat Rock Drive (whole road, by direction)	no	~6.1%	60	47–54	61–63
Eastern Arterial Rd, Gordon (camera between Koola & Nicholson)	yes	~6.2%	60	49–52	55–58

Month-aggregated connected-vehicle tiles; edge aggregation damps peaks — individual observed speeds on Flat Rock Drive exceed these values. Descriptive comparison of one matched pair; it cannot by itself attribute cause.

Street View — Flat Rock Drive descent — Flat Rock Drive — the no-camera descent (Google Street View)

Street View — Eastern Arterial Road descent — Eastern Arterial Rd — the camera hill (Google Street View)

Satellite — Flat Rock Drive — Flat Rock Dr — satellite (red = steepest section)

Satellite — Eastern Arterial Road — Eastern Arterial Rd — satellite (blue = camera zone)

stage	n	rule
TfNSW fixed-camera list	123	open data (CC-BY)
GPS-locatable	103	coordinates parseable
paired	60	CV features ≤200 m + control band 0.6–1.4 km
non-school-zone	41	time-variable limits excluded
strict same-road pairs	17	identical road + limit, grade-similar

stage

rule

TfNSW fixed-camera list

123

open data (CC-BY)

GPS-locatable

103

coordinates parseable

paired

CV features ≤200 m + control band 0.6–1.4 km

non-school-zone

time-variable limits excluded

strict same-road pairs

identical road + limit, grade-similar

Same hill, same limit — one has a camera.
Drivers choose very different speeds.

The matched-grade pair mart_extreme_pair · Operating Speed data · Jun-26 · OSM · DEM

Hilly vs flat corridors mart_hills

What we did NOT find mart_summary_e1

The trap we refused mart_baseline_e2 · exhibit

Gates — what survived mart_attrition

Ask the evidence