Research · Proposal · Updated Oct 5, 2026
Lanes on the 12 streets where buses lose the most time, and have little or no lane, would save riders about 720 hours each weekday (440 to 1,070), worth $3.5 million a year. Flatbush Av and Broadway lead, close enough to call a tie. A half-hour ride would get about a minute shorter. This is an estimate of what lanes could do, built on the gain past lanes gave, not a forecast for any one street.
For each street, we take the time buses spend each weekday on its blocks without a lane, cut it by 3.2% (the time a 3.3% speed gain saves, the average for new lanes in 2023–25), and multiply by 14, the riders aboard a typical bus. On Flatbush Av, buses spend about 269 hours a weekday on blocks with no lane: 269 × 3.2% ≈ 8.6 bus-hours, × 14 riders ≈ 120 rider-hours. The 12 streets are the ones, of 237 we measured, where buses lose the most time against the street’s own fastest hour and less than half the route-miles have a lane.
Rider-hours saved each weekday if the blocks of each street without a lane got one, at the typical gain of 3.3%. Flatbush Av and Broadway save the most in all, close enough to call a tie. Per mile of new lane, Madison Av is far ahead: 17 routes share its blocks.
Rider-hours saved each weekday by a lane on the blocks that lack one, 12 slow New York City streets, 7 am to 7 pm, 2026
Source: MTA bus segment speeds; MTA bus ridership by route; NYC DOT bus lanes
Under each street: the share of its route-miles already on a lane, and the miles of street that would need one. Weekdays, 7 am to 7 pm. Streets within a fifth of the leader’s hours count as tied. The 2% to 5% range moves every street up or down together, so it can’t change the order; streets close together, like the top two, could swap if their own gains differ.
The fewer the streets, the more each mile of lane gives back; the more streets, the bigger the total. How well the lanes work matters as much as how many there are.
| Version | Lane-miles | Rider-hours a weekday | Rider-hours a year | Riders’ time a year | Bus service a year | Per lane-mile |
|---|
Buses ran 3.3% faster on 53 stretches that got a new lane from 2023 to 2025 than on similar stretches without one, over the same months (95% interval 1.9% to 4.7%). Before the lanes opened, the two groups moved together; the gap opens the month the lane does and holds on average over the year, though single months late in the year are less sure. Dropping any one corridor leaves the answer between 2.7% and 3.7%, though single streets range from 1.4% (East Gun Hill Rd) to 8.3% (East 96 St). It is the same kind of test as on the congestion pricing evaluation: compare what changed with what didn’t, over the same months.
Weekday bus speed, 53 stretches given a lane in 2023–25 minus similar stretches without, months from opening, %, with likely ranges (95%)
Source: MTA bus segment speeds; NYC DOT bus lanes
Bus speed on stretches that got a new lane, minus matched stretches with no lane, by month from the lane’s opening. Each month is measured against the six months before opening; the opening month itself is left out. Weekdays, 6 am to 8 pm; 53 stretches on 23 routes.
The biggest doubt is that every street gets the same gain; any one street could save from 0.4 to 2.5 times as much (1.4% or 8.3% against 3.3%). What the model leaves out cuts both ways: it misses weekend riders, but also what a lane costs drivers and deliveries.
For each street, we take the time buses spend on it each weekday, keep the part on blocks with no bus lane, and cut it by the speed gain new lanes have delivered. Then we turn bus-hours into riders’ hours and dollars.
The street table’s riders are boardings on every route that touches a street, not riders on that street, so they can’t replace the 14 riders aboard. Bus speeds on these streets are averages over six months. The caveats that matter most are under Where this works and where it doesn’t.
Two things would tighten this ranking: a street-by-street check of what the method would have predicted for the lanes opened in 2023–25, against what they gave, and rider counts by street in place of the citywide 14. To try a lane of your own, use Draw a bus lane. Every number here is in bus-lanes-proposal.json, so you can check our work.