Robotaxis Promise Safer Streets. They May Bring More Traffic.

A Waymo, Google's driverless taxi service, drops a passenger on a Greenwich Village block and does not leave. It idles at the curb, sometimes for minutes, waiting for the next rider, while the next vehicle in the queue circles the block to take its spot. The pattern looks benign in a single block, but a new Open Plans report says the same dynamics, multiplied across a city that has not yet approved commercial robotaxi service, can degrade the streets those vehicles are meant to serve. And the city still has the per-mile dead-head fees, drop-off geofencing, shared-ride minimums, and curb permits that can intervene before any commercial rollout.

The report, written by the Sam Schwartz Transportation Research Program (SSTRP) at Roosevelt House Public Policy Institute at Hunter College, projects what would happen if for-hire autonomous vehicles, meaning robotaxis operated by companies like Waymo rather than personally owned self-driving cars, launched in New York. Its core finding is a paradox: vehicles that almost never crash can still clog a block.

The safety data is the part the AV industry has been waiting for. Initial fleet numbers show Waymo vehicles involved in fewer injury-causing crashes overall than the city's existing for-hire fleet of taxis, Ubers, and Lyfts. But the report flags a sharper point: Waymo's rate of crashes involving serious injury or fatality is higher than that of the city's professional drivers, according to a Yahoo News summary of the underlying comparison. Waymo's standard framing compares its safety record to the average American driver. The Open Plans authors argue that benchmark is the wrong one for New York, where for-hire drivers spend their working hours in dense traffic and post collision rates well below the national average. Gothamist has covered the same dispute.

The livability data is where the report spends most of its room. Robotaxi trips do not end at drop-off. They end when the vehicle finds the next fare, and every minute between is a minute of empty vehicle miles traveled (VMT), a minute of curb consumed, and a minute in which a public street is being staged for private use. The pattern is familiar from ride-hail: a Streetsblog column on the report calls the resulting VMT growth a "Way no" to the assumption that driverless cars will shrink traffic. Multiply that across a fleet that is, by design, always repositioning for the next passenger, and the curb fills with empty vehicles while buses, bikes, and emergency access get pushed to the margins.

The report's parking analysis makes the same point from a different angle. Curb space in Manhattan is the most expensive parking real estate in the country, and converting even a small share of it into robotaxi pickup and idling zones would displace the parking that delivery trucks, paratransit, and residents already compete for. The specific displacement figures in the underlying PDF report require verification before they should be quoted as load-bearing numbers, but the directional finding is clear: a curb built around private vehicle staging is a curb that does less public work.

The city is not without levers. The Open Plans authors, like a number of urbanists tracking the rollout in San Francisco, Phoenix, and Los Angeles, point to a small set of moves that can intervene before commercial scale. Per-mile dead-head fees, which are charges for miles driven without a passenger, make empty repositioning expensive. Drop-off geofencing keeps robotaxis out of residential blocks and school zones during pickup peaks. Shared-ride minimums push operators toward higher-occupancy trips. Curb permits, already the city's main tool for managing food trucks, film shoots, and construction staging, can extend to robotaxi idling zones the same way. None of these is hypothetical; San Francisco and Phoenix have piloted versions of all of them.

Waymo's New York test permit expired in March 2026 without renewal, and the Curbed coverage of the pause frames it as the city's clearest piece of negotiating leverage. The company has spent millions on state and city lobbying for legalization, per the report, and a commercial permit would unlock the same fleet dynamics the report projects. The next twelve months of negotiation will set the terms other cities either adopt or resist, and the Open Plans report is, in effect, the policy menu the city can order from.

The unresolved question is not whether robotaxis are safe enough to operate. On the narrowest collision metric, they may already be. The question is what a city is willing to give up to find out, and whether the permits, fees, and curb rules are in place before the vehicles arrive at scale rather than after.