Research · Evaluation · Updated Oct 7, 2026
For a year from September 2023, one bus route in each borough ran free. Weekday riders on them rose 30%, against 6% on all buses, but only 12% were new to the route. Nearby routes lost about 5% of their paying riders (range 0.4% to 8.7%), an upper bound, since fare evasion there rose faster than elsewhere. The free routes didn’t speed up; they slowed 1.4%. It cost $16.5 million, about $4 an extra trip.
The +30% and 12% are the MTA’s counts and survey, with no range; our own figures have one. Confidence: high that riders rose, since the gain is five times the citywide one; medium on the nearby loss and the slowdown, since fare evasion and other changes on those routes could explain part of each.
The 2023 state budget told the MTA to make one bus route in each borough free for six to twelve months. The state capped the cost at $15 million. The MTA chose the Bx18A/B, B60, M116, Q4 and S46/S96.
Neither the MTA nor the law set a target, such as a number of new riders. We quote the MTA’s 1-year evaluation (June 2025) and the city’s Independent Budget Office (June 2026), not the budget law itself.
Free fares did what free fares usually do: more trips, mostly from people already riding. Service got a little worse, assaults on drivers fell, and the bill came in above the cap. The MTA did not report on rider satisfaction.
| Route | Riders before | Weekdays | Weekends | Cost | Per extra trip | Driver assaults | Speed |
|---|
Riders: weekday riders, September 2022–May 2023, and the change in the same months of the pilot, from the MTA’s passenger counters. Cost: the MTA’s full cost by route. Per extra trip: that cost over the MTA’s extra riders a day times the days of the pilot (our sum, rough). Driver assaults: verbal and physical, a year before and during. Speed: ours, against similar routes that kept fares.
Weekday paid riders, 19 routes beside the five free routes minus similar routes farther away, Oct 2021–May 2025, %, with likely ranges (95%)
Source: MTA bus hourly ridership
Weekday bus speed, 19 routes beside the five free routes minus similar routes farther away, Oct 2021–May 2025, %, with likely ranges (95%)
Source: MTA bus speeds
| Route | Next to | Paid riders before | Paid riders | Speed |
|---|
The 19 routes with at least a fifth of their path within 400 m of a free route. Each is compared with up to five routes in its borough that run nowhere near a free route and were closest to it in speed and paid riders. Paid riders before: a weekday, October 2022–August 2023. Changes: October 2023–August 2024 against a year before, minus the change on the comparison routes.
Riders no longer stopped to pay, but more of them got on. Against similar routes in the same boroughs that kept fares, the free routes slowed by 1.4% (range 0.3% to 2.3%), about half of it in the summer of 2024. That gap didn’t close when fares came back, so we can’t pin it all on free fares. What the data do rule out is a speed-up.
Weekday bus speed, five fare-free routes and similar routes with fares, 2022–23 vs 2023–24, mph; below, monthly gap, %; likely ranges (95%)
Each month against October 2022–August 2023. Average of the five routes.
Source: MTA bus speeds
Fare evasion would shrink the loss on nearby routes, and other changes on the free routes would shrink the share of the slowdown that free fares caused. The rest could move the answer either way.
The riders on the free routes, the survey, the costs and the service figures are the MTA’s. The free routes’ own fare taps fell to almost nothing while they were free, so public data can’t count their riders in that year. Our own work covers what public data can show: paid riders and speed on nearby routes, and speed on the free routes.
The caveats that matter most are under What could change this answer.
A 2025 bill to expand the pilot (S4421) did not pass. If New York tries free buses again, watch the measures on this page: riders new to the route, not just trips; speed against routes that kept fares; and paid riders on routes nearby. Every number here is in fare-free-pilot.json, so you can check our work.