Automated speed enforcement reduced vehicle speeds in school zones in Toronto: a prospective quasi-experimental study

The study addresses whether automated speed enforcement (ASE) effectively reduces vehicle speeds in urban school zones, where child pedestrians are vulnerable and injury risk is strongly determined by vehicle speed. The context is a global public health burden from road traffic crashes and WHO recommendations for low speed limits (≤30 km/h) in areas with mixed traffic and pedestrians. Prior evidence shows ASE reduces speeding and crashes primarily on highways, but urban applications, particularly in low-speed pedestrian environments like school zones, are under-reported and previous studies have methodological limitations. Toronto implemented ASE within its Vision Zero programme to protect vulnerable road users; this study evaluates the impact of ASE on speeding in these school zones using a quasi-experimental design.

Systematic reviews in highway settings report ASE reduces speeding (14%–65%), average speeds (1%–15%), crashes (8%–49%), overall injuries (12%–65%), and fatal/severe crashes (11%–44%), though with noted limitations such as weak control groups, confounding, and measurement issues. Urban applications are less documented, focusing on urban freeways or pedestrian areas; effects are harder to isolate when speed limits change concurrently. Prior urban arterial studies showed modest reductions in 85th percentile speed (e.g., ~0.77 mph), and a small school-zone study found reductions between warning and ticketing phases but lacked true pre-ASE speed measurements. Child pedestrian injuries concentrate near schools, highlighting the need for effective speed management in school zones.

Design: Prospective quasi-experimental study at 250 ASE deployment sites across Toronto school zones, with five sequential deployments between July 2020 and December 2022. Unit of analysis was the road segment where ASE was installed. Intervention: 50 mobile ASE cameras (two per ward) deployed for several months per site; cameras moved to new sites within the ward after each deployment. Site selection: Data-driven by City of Toronto Transportation Services considering child collision history, vehicle volumes, provincial guidelines, planned roadwork, and on-site feasibility. Data collection periods: (1) Pre-ASE: Vehicle speed and volume collected using pneumatic road tubes over three consecutive weekdays between 2018 and the intervention date. (2) Warning period: Provincial regulation required ≥90 days of warning signage before ticketing. For the first deployment (Jan–Jul 2020), 50 ASE cameras were present and recorded speeds but did not issue tickets; for subsequent deployments, warning signs were posted without cameras and data were collected at 44 locations via pneumatic tubes over three weekdays. (3) ASE intervention period: Cameras operational for 101–179 days per site (average 132 days), measuring speeds and volumes; simultaneous pneumatic tube measurements were collected at 24 sites to assess measurement agreement. (4) Post-ASE: At 38 locations, pneumatic tube speed/volume data were collected 28–45 days after camera removal over two weekdays. Cointerventions: Speed limit changes and traffic calming were tracked; sites with cointerventions were excluded from modelling. Outcomes: Primary outcome was proportion of vehicles speeding (any amount and thresholds of ≥5, ≥10, ≥15, ≥20 km/h above the limit). Secondary outcome was 85th percentile speed. Analysis: Descriptive and graphical analyses stratified by posted speed limits. Regression modelling included fixed-effects negative binomial regression with robust SEs via GEE for counts of speeding vehicles, with offset for total vehicles, fixed site effects, seasonality (Summer: May–Oct; Winter: Nov–Apr), and exclusion of weekend observations to match pre-ASE collection. Separate models estimated effects at each speeding threshold. For 85th percentile speed, linear regression with similar covariates (site fixed effects, seasonality) was used. Sensitivity analyses: Assessed regression to the mean and warning-only effects by including categorical periods (pre, warning, ASE) at sites with all three phases; evaluated post-ASE effects at sites with pre, ASE, post data; compared measurement agreement between ASE cameras and pneumatic tubes at 24 sites using Bland-Altman plots.

• Proportion of vehicles speeding reduced by 45% during ASE intervention compared with pre-ASE (RR 0.55, 95% CI 0.49–0.61). - Reductions were greater at higher speeding thresholds: ≥5 km/h RR 0.38 (95% CI 0.32–0.44), ≥10 km/h RR 0.26 (95% CI 0.19–0.33), ≥15 km/h RR 0.16 (95% CI 0.13–0.19), ≥20 km/h RR 0.12 (95% CI 0.09–0.15). - 85th percentile speed decreased by 10.68 km/h overall (95% CI −11.48 to −9.88). Stratified reductions: 30 km/h roads −11.5 km/h (95% CI −14.1 to −8.9), 40 km/h roads −9.3 km/h (95% CI −10.3 to −8.3), 50 km/h roads −10.6 km/h (95% CI −13.6 to −7.6). - Sensitivity analyses: Warning-only periods showed minimal change vs pre-ASE (e.g., overall excess speeding RR 0.92, 95% CI 0.86–0.98), whereas ASE ticketing showed substantial reductions (e.g., overall RR 0.64, 95% CI 0.56–0.72). Post-ASE periods showed speeding proportions returning toward pre-ASE levels (e.g., post-ASE overall RR 0.87, 95% CI 0.69–1.06), indicating limited persistence of effect after camera removal. - Measurement agreement between ASE units and pneumatic tubes was good, with Bland-Altman analyses showing no meaningful bias sufficient to explain observed effects. - Models included 187 locations after exclusions for incomplete pre-ASE data (44 sites) and cointerventions (speed limit changes 15 sites; speed humps 12 sites; other traffic calming 8 sites).

ASE implementation in Toronto school zones led to substantial reductions in speeding and 85th percentile speeds, addressing the core research question that enforcement can effectively lower vehicle speeds in pedestrian-rich urban areas. The greater effect at higher speeding thresholds suggests ASE particularly curtails the most hazardous behaviors, which should translate into fewer and less severe pedestrian injuries by reducing kinetic energy and improving driver reaction opportunities. The study design accounted for key biases: regression to the mean was unlikely given minimal changes during warning-only and post-ASE periods; cointerventions were documented and excluded; measurement bias was assessed via simultaneous pneumatic tube and ASE measurements showing good agreement. The fixed-effects models controlled for time-invariant site characteristics and seasonality, though potential spatial correlation between nearby sites was not modeled. Results support integrating ASE with built environment changes and regulatory approaches as part of comprehensive Vision Zero strategies.

Automated speed enforcement in Toronto school zones produced a 45% reduction in the proportion of vehicles speeding and a 10.68 km/h decrease in 85th percentile speeds, with larger effects at higher speeding thresholds. ASE is a pragmatic, effective tool to reduce vehicle speeds and potential injury risk among child pedestrians in urban school zones. Future research should evaluate injury outcomes once sufficient data accrue and examine broader network effects, potential displacement, and integration with built environment and regulatory interventions.

• Injury outcomes were not measured; speed was used as a surrogate for injury risk. - The intervention period coincided with the COVID-19 pandemic, with disruptions to schooling and traffic patterns; although analyses argue ASE drove observed effects, residual confounding is possible. - Lack of routine city-wide speed/volume data limited assessment of secular trends and spillover/displacement effects to nearby streets. - Spatial correlation between proximate intervention sites was not explicitly modeled. - Cost constraints limited data collection to intervention sites. - Post-ASE effects suggested limited persistence after camera removal. - Some sites were excluded due to cointerventions or incomplete pre-ASE data, potentially affecting generalizability.