Which streets sped up after congestion pricing?
Data: the MTA’s bus speeds between timepoint stops, by route, direction, stop pair, hour and month: 2023–2024 and 2025 on, through August 2026. Weekdays, 6 am–8 pm. Pricing started 2025-01-05. Run 2026-10-05. Follows Congestion pricing and buses.
Short answer
Section titled “Short answer”Buses sped up on most downtown avenues and crosstown streets in 2025, and the gain faded in 2026 on some of them, not all. Madison Av and the 23 St and 42 St crosstowns gained about 6% and held it. 3 Av, 6 Av and E 23 St gained 4–5% and gave most of it back. Lexington Av and 1 Av never gained; Lexington lost about 6% and fell further in 2026.
The fade is real but smaller than the route-level note suggested, and it depends on how you draw the zone (below).
Method
Section titled “Method”1. Segments. The MTA publishes bus times between consecutive timepoint stops. A segment is one route, direction and stop pair. Its speed is miles × trips ÷ hours, totalled over the months with 30+ weekday trips in both years compared. Change is the log of the speed ratio, weighted by bus-miles (miles × trips).
2. Zone. A hand-drawn polygon around Manhattan below 60 St: the 60 St line runs through 59 St & 1 Av and 59 St & 11 Av, and the other edges follow the shoreline (polygon in the script). A segment is inside if both stops are in it, straddling if one is, outside if neither. Only routes with a segment in the zone are used (115 of 161 candidate routes), so the outside parts of those routes are the control. Counts, pooled over all years: 664 inside, 328 straddling and 842 outside segments.
3. Compare matching months. 2025 against 2024, 2026 (Jan–Aug) against 2024 Jan–Aug, and 2024 against 2023 as a placebo. Zone change minus outside change; 95% intervals from resampling segments. Resampling whole routes instead gives similar intervals.
4. Streets. An inside segment whose two stops share a street name (“W 34 ST/7 AV” to “W 34 ST/BROADWAY”) counts for that street, across routes and directions. A street’s change is net: each segment’s change minus its own route’s change outside the zone (routes with under 50,000 outside bus-miles get the average outside change).
Results
Section titled “Results”| Zone | Outside | Difference | 95% interval | Zone segments | |
|---|---|---|---|---|---|
| Both stops inside | |||||
| 2024 vs 2023 (placebo) | −2.2% | −0.4% | −1.9% | −2.8 to −1.0 | 448 |
| 2025 vs 2024 | +2.8% | 0.0% | +2.8% | +1.8 to +3.7 | 444 |
| 2026 (Jan–Aug) vs 2024 | +0.5% | −1.4% | +1.9% | +0.7 to +3.1 | 324 |
| One or both stops inside | |||||
| 2024 vs 2023 (placebo) | −1.0% | −0.4% | −0.6% | −1.2 to 0.0 | 677 |
| 2025 vs 2024 | +2.4% | 0.0% | +2.4% | +1.7 to +3.1 | 671 |
| 2026 (Jan–Aug) vs 2024 | −1.7% | −1.4% | −0.3% | −1.3 to +0.8 | 480 |
- The two zone definitions agree for 2025 (+2.8% and +2.4%, close to the route-level +3.5%) and disagree for 2026. Segments with one stop outside the zone are long and often run mostly outside it, such as the BM1 from Cortelyou Rd in Brooklyn to Trinity Pl. They slowed in 2026 (−1.3% net, interval −2.4 to 0.0) and pull the combined number to zero. Segments fully inside show +1.9%, close to the route-level +2.4%. I trust the fully inside number more, since it measures the zone itself.
- The placebo isn’t clean. Segments fully inside were already losing ground to the outside before pricing (−1.9%, interval −2.8 to −1.0). That matches the earlier note’s crosstown locals. Some of the 2025 gain is recovery from a weak 2024.
- The fade. Compare 2025 and 2026 on the same months (Jan–Aug) and the same segments. Net effect fell from +3.0% to +1.9% for fully inside segments: −1.1 points (−1.8 to −0.4). For all segments with a stop inside it fell from +2.9% to −0.3%, a drop of 3.1 points (−4.0 to −2.2).
- Spread of the fade. Across fully inside segments, net speed fell between 2025 and 2026 on 60% of bus-miles. The median change is −0.4 points; a quarter of segments lost more than 4 points and a quarter gained more than 1.6.
By street
Section titled “By street”Streets with 5,000+ bus-miles in 2024 and at least two segments, sorted by 2025 change. Changes are net of the routes’ change outside the zone. “Placebo” is the same measure for 2024 vs 2023: a street with a large placebo is not a clean test. “Fade” is 2026 minus 2025, both over Jan–Aug, in percentage points.
| Street | Segments | Bus-miles 2024 | mph 2024 | mph 2025 | mph 2026 | Change 2025 (95% interval) | Change 2026 | Placebo | Fade |
|---|---|---|---|---|---|---|---|---|---|
| Madison Av | 10 | 98,967 | 4.7 | 4.9 | 5.0 | +6.7% (+5.5 to +10.2) | +4.8% | −4.5% | −0.1 |
| 8 Av | 2 | 8,935 | 4.0 | 4.3 | 4.4 | +6.0% (+5.8 to +9.5) | +6.2% | −4.3% | −3.3 |
| W 42 St | 7 | 37,372 | 4.5 | 4.7 | 4.7 | +6.0% (+1.9 to +7.2) | +6.2% | +4.8% | −1.8 |
| W 23 St | 2 | 61,774 | 5.2 | 5.5 | 5.5 | +5.6% (+0.6 to +11.4) | +5.6% | −2.8% | −0.8 |
| 3 Av | 25 | 61,277 | 5.8 | 6.2 | 6.0 | +4.9% (+2.8 to +7.6) | +0.9% | +0.4% | −7.4 |
| W 34 St | 2 | 7,072 | 5.3 | 5.5 | – | +4.3% (+3.9 to +4.8) | – | −1.7% | – |
| E 23 St | 5 | 10,731 | 5.5 | 5.7 | 5.4 | +3.6% (+0.9 to +4.9) | +0.2% | +1.9% | −5.0 |
| 6 Av | 6 | 38,658 | 4.8 | 5.0 | 4.9 | +3.5% (−2.3 to +10.2) | −0.8% | −1.4% | −6.2 |
| Broadway | 18 | 91,318 | 6.3 | 6.3 | 6.3 | +2.1% (+0.5 to +4.0) | +2.9% | +0.8% | +0.3 |
| 5 Av | 9 | 57,193 | 6.3 | 6.5 | 6.5 | +1.5% (+0.4 to +2.5) | +1.2% | +0.1% | −0.9 |
| 11 Av | 2 | 11,320 | 6.2 | 6.3 | 6.2 | +1.1% (−1.1 to +4.1) | +0.6% | −2.1% | −2.2 |
| E Houston St | 4 | 16,740 | 5.4 | 5.5 | 5.4 | +0.7% (−1.0 to +1.6) | +2.9% | +2.0% | +0.9 |
| 2 Av | 2 | 9,102 | 5.3 | 5.5 | – | −0.5% (−3.7 to +3.0) | – | +4.1% | – |
| West St | 3 | 6,678 | 5.3 | 5.2 | 5.4 | −0.8% (−1.4 to −0.2) | +5.6% | −2.2% | +4.2 |
| 1 Av | 5 | 60,717 | 6.2 | 6.2 | 6.0 | −4.8% (−11.2 to +0.2) | −2.7% | +0.5% | −1.2 |
| Lexington Av | 6 | 54,995 | 5.0 | 4.8 | 4.8 | −5.9% (−7.2 to −2.9) | −9.0% | −5.6% | −3.9 |
- Sped up and held: Madison Av, W 42 St, W 23 St, Broadway (small gain, no fade). Madison’s gain is the most solid: 10 segments, 99,000 bus-miles, interval +5.5 to +10.2.
- Sped up, then faded: 3 Av (+4.9% to +0.9%), E 23 St, 6 Av. 3 Av is the best-measured of these.
- Never sped up: Lexington Av is clearly slower, and slower again in 2026. 1 Av is slower, within the noise.
- Only 16 streets pass the filter, 4 of them negative, so there is no top 10 of streets that slowed. After Lexington, the weakest are 1 Av (−4.8%, interval just touches zero), then West St and 2 Av, both near zero.
- Left out: Hudson St (+52%) and Pearl St (+11%). Hudson’s M11 and M14A buses ran 4.4 and 3.2 mph in 2024, against 6.2 and 4.5 in 2023, then recovered in 2025: a 2024 slump (placebo −30%), not pricing. Pearl St’s placebo is +24% and its stop pairs vanish in 2026.
Caveats
Section titled “Caveats”- Streets cover little. Timepoint stops are 1–2 miles apart and only 28% of fully inside bus-miles (753,000 of 2.7 million) have both stops on the same street. Many streets have one or two segments, and the intervals are rough where there are few.
- The zone edge is hand-drawn, not the official boundary. Stops near 60 St, the FDR and the Hudson could land on the wrong side. Segments straddling the edge are the least reliable group (see above).
- Segments change over time. The MTA moves and renames timepoints. Of 721 inside segments in 2024, 678 appear in 2025 but only 483 in 2026. The fade numbers use the same segments in both years, so this doesn’t drive them, but 2026 covers fewer streets (W 34 St, 2 Av and Pearl St drop out).
- The 2024 baseline is shaky. Segments fully inside were already slowing against the outside in 2024, and single streets (Hudson) had odd years. A different baseline year would change the street ranking.
- Net change assumes the outside part of a route is a fair control. If pricing also changed traffic outside the zone, that biases every number. For routes with little outside part (crosstown locals), the control is the average outside change instead.
- Speed includes time at stops and lights; it doesn’t isolate traffic. Averages over a month hide days when traffic was bad. The 2026 months are Jan–Aug only; the 2025 comparison uses all 12 months, except in the fade rows.
- The headline here (+2.4% to +2.8% in 2025) is lower than the route-level +3.5%, and the 2026 value depends on the zone definition (−0.3% or +1.9%, against +2.4% route-level). The two datasets weight routes differently, and the route-level one counts miles inside the zone exactly rather than by stop pair. I did not reconcile them further.