Follow the evidence from collision patterns to adjusted models and validation. Start with the conclusions, then inspect the reasoning and exact results.
These studies analyse police-reported injury collisions. They do not measure the full insurance claims outcome. No study below establishes a validated policyholder rating factor.
20MPH PUBLICATION · PHASE 6
20mph limits: read the evidence
Compare measured journey duration and variation, review cost evidence, and test the casualty prevention needed to offset assumed delays. National net benefits remain unresolved.
Use the same six questions in every study. A completed analysis stage can still leave a hypothesis unsupported or unresolved.
1
Question
State the proposed signal and its possible use in pricing research.
2
Measure
Define the events, population, period and exposure before comparing rates.
3
Initial pattern
Inspect the descriptive gradient and its support.
4
Adjustment
Explain area and road context before testing an additional feature.
5
Validation
Challenge time, geography, calibration and exposure assumptions.
6
Decision
Record what survives, what fails and the evidence needed next.
A raw gradient suggests a question. Adjustment asks whether other characteristics explain it. Validation asks whether the remaining signal survives elsewhere and under different assumptions.
How the studies connect
Shared foundation: collision records and compatible exposure
STATS19 supplies events. DfT supplies aggregate mileage estimates. Local mileage allocation is an additional modelling assumption. Check each study’s numerator and denominator before comparing it with another.
Connections explain the research questions, not interchangeable results. Each study retains its own cohort, sources, assumptions and validation history.