Toronto's Speed Cameras

Where each number comes from

Why 250 metres

How far from a camera does its influence extend? The literature uses a range, and our 250-metre choice sits in the middle of it and matches our own New York analysis, which allows direct comparison between the two cities.

The result does not depend on that choice. We repeated the analysis at six distances and out to a kilometre, and the effect is present and significant at every one, attenuating with distance as a spatial deterrence model would predict.

Why the police comparison runs at neighbourhood level

The Toronto Police "Tickets Issued" file records 769,678 speeding tickets from 2014 through 2024 by neighbourhood, using the City's 158-neighbourhood classification plus a residual code, and it carries no coordinates and no intersection, while the collision and camera files are geocoded to the metre.

So we can ask whether camera effectiveness differs between neighbourhoods with heavy and light police enforcement, which is the question we answer, and we cannot ask whether an officer and a camera worked the same corner. A neighbourhood is a coarse unit for this, since a camera appears to influence roughly 250 metres and a Toronto neighbourhood is many times that, so genuinely local substitution could be present and invisible to us. No evidence of substitution is a statement about what these data can detect.

The Waterloo 2024 vintage

Waterloo Regional Police changed its publication format from CSV to XLSX in 2024, and the data show the signature of a coding-threshold change rather than a change on the roads.

Records jumped 38 percent while the injury rate halved, and the hit-and-run share rose from 13.4 to 17.4 percent. More records at lower average severity is the signature of a threshold change in what gets coded as a collision. Our primary specification therefore excludes 2024. The contrast that circulates most, Waterloo at 129 against Toronto at 85, takes that vintage, and we do not lean on it.

Limits we cannot argue away

We think readers are best served by the limits stated as plainly as the findings.

A single effect size. The estimate shifts with the research design, from roughly zero to −10 percent. Our judgment of −4 to −7 percent is a judgment, and we have tried to show the work rather than declare a number.

How much of the post-removal rise was camera-specific. The direction is clear; the magnitude is entangled with the post-pandemic recovery and cannot be separated from it with these data.

How much belongs to the camera alone. At 193 sites, or 37 percent, a radar sign's deployment overlapped the camera's, so what we call the camera effect there is strictly a bundle. An exploratory split on the busier half of the network puts cameras without a sign at −9.2 percent against bundled sites at −7.9, statistically indistinguishable. No cross-site split can separate the constants: school-zone signage, markings and zone speed limits persist at nearly every site.

Direct severe-injury effects. KSI events near camera sites are too rare to test, so those figures are imputed from the injury effect.

Risk per kilometre driven. Traffic-volume data for the period do not exist in public form. Our estimates are crash counts; a proxy adjustment using radar-sign traffic counts left the estimate essentially unchanged, at −11.1 against −10.9 percent.

That reporting practices held still. About 16 percent of collision records lack usable coordinates and differ modestly from those that have them, and Waterloo's 2024 data show signs of a coding change. Both are named and bounded; neither can be fully corrected.

Which sites will respond next time. Site-level outcomes are dominated by noise at the crash counts involved. Our best cross-validated model explains about 11 percent of the site-to-site variation.