Estimated preventable deaths across tracked cities since their respective delay start dates
85% crash reduction (Kusano et al. 2024) × 10% VMT share (Fehr & Peers 2019). Counter ticks in real time. See methodology for assumptions and limitations.
Other US cities moved from testing to commercial driverless service in months. Blocked cities have been waiting years - with no end in sight.
| City | Status | Est. Preventable Deaths | Key Blocker |
|---|
All calculations are estimates with significant uncertainty. Full transparency on assumptions below.
For each city, we calculate cumulative traffic fatalities since the delay start date using linear interpolation within each year of reported data. We then apply:
Preventable Deaths = Cumulative Fatalities × 10% VMT share × 85% crash reduction
10% VMT share: Fehr & Peers (2019) found ride-hail accounts for 1–13% of city-core vehicle miles traveled across major US cities (DC: 6.9%, SF: 12.8%). We use 10% as a conservative proxy, assuming AVs would capture most existing ride-hail plus modest growth.
85% crash reduction: From Kusano et al. (2024/2025) peer-reviewed studies. Waymo demonstrated 85% fewer injury-reported crashes across 56.7M miles of operation vs. human drivers.
What this measures: The estimated opportunity cost of regulatory delay - if ride-hail-equivalent AVs had been operating in these cities, how many crashes would have been prevented? This is not a certainty; it is an estimate of what was likely preventable.
What this does NOT account for: Induced demand, non-Waymo AV providers, network effects, or variation in AV performance by city environment. City-level fatality data may be incomplete - TODOs are marked in the code where verification is needed.
The most common human-error causes of crashes are behaviors AVs simply do not exhibit. Source: NHTSA Risky Driving / NHTSA Traffic Safety Facts.
The safety comparison chart shows incident rates across providers, but the methodologies differ substantially:
Waymo: Uses police-reported crashes, peer-reviewed analysis. Serious injuries per million miles: 0.02. Source: Kusano et al. (2024) ↗
Uber (0.45 accidents/million miles): Self-reported figure from the Uber US Safety Report ↗. Self-reported data typically undercounts incidents compared to police-reported data.
Lyft (0.38 accidents/million miles): Self-reported figure from the Lyft Community Safety Report ↗. Same methodological caveat applies.
The direct comparison should be interpreted as directionally informative, not precise. The peer-reviewed Kusano et al. studies comparing Waymo to human drivers in matched conditions are the more rigorous source.