Did congestion pricing change taxi and Uber/Lyft traffic?
Data: NYC TLC trip records, yellow taxis and “high-volume for-hire” cars (Uber, Lyft, Via), January 2023 to August 2026 (September was not out on 2026-10-05). Weekdays 6 am–8 pm by pickup time, federal holidays left out. Peak is 7–10 am and 4–7 pm. Pricing started 2025-01-05, so January is left out of every comparison. Run 2026-10-05.
Short answer
Section titled “Short answer”Trips: no clear drop. Uber/Lyft trips that start or end in the zone fell 1.4% in 2025 against 2024 (−3.3 to +0.5), and that is no more than the drift before pricing. Yellow taxi trips in the zone rose 13.5%. Together, for-hire trips touching the zone went from about 202,500 to 208,700 a weekday (+3%). These data cover taxis and app cars only, not private cars or trucks, so they say little about the traffic pricing was meant to cut. A $0.75 (taxi) or $1.50 (app car) fee per trip is also small beside the $9 charged to a private car.
Speeds: a small gain. Trips inside the zone ran about 2% faster in 2025 than in 2024, compared with trips above 60 St (taxis +2.1%, Uber/Lyft +1.9%). The year before pricing, the same measure fell 2.2% and 2.1%. In 2026 the gain was gone. In minutes: about 20 to 45 seconds saved on a 2-mile trip inside the zone, which takes about 17 minutes.
Trips that cross the zone edge took about 3% less time on the same zone pairs, against a control of pairs above 60 St (taxis −3.3%, Uber/Lyft −3.0%). By 2026 the gain had shrunk to −1.8% and −0.5%.
For buses we found +3.5% in 2025, so cars gained about as much as buses (a little less on the plain comparison, the same if the earlier slide is removed).
The zone
Section titled “The zone”TLC records only say which of 263 zones a trip began and ended in, not the street. We take the zones from the fee itself. In March 2025 files, among trips that start and end in the same zone, the share with a congestion fee falls into three groups:
| Class | Rule | Zones |
|---|---|---|
| In the zone (38) | fee on at least 60% of same-zone trips (lowest: Battery Park 73%; the rest 87–97%) | Alphabet City; Battery Park; Battery Park City; Chinatown; Clinton East; Clinton West; East Chelsea; East Village; Financial District North and South; Flatiron; Garment District; Gramercy; Greenwich Village North and South; Hudson Sq; Kips Bay; Little Italy/NoLiTa; Lower East Side; Meatpacking/West Village West; Midtown Center, East, North and South; Murray Hill; Penn Station/Madison Sq West; Seaport; SoHo; Stuy Town; Sutton Place/Turtle Bay North; Times Sq/Theatre District; TriBeCa/Civic Center; Two Bridges/Seward Park; UN/Turtle Bay South; Union Sq; West Chelsea/Hudson Yards; West Village; World Trade Center |
| Straddles 60 St (6), left out | fee on 5–60% | Central Park; Lenox Hill East and West; Lincoln Square East and West; Upper East Side South |
| Control: Manhattan above 60 St (19) | fee on under 5% | Upper East and West Side North, Yorkville, Manhattan Valley, Bloomingdale, Morningside Heights, Harlem, East Harlem, Hamilton Heights, Manhattanville, Washington Heights, Inwood and Inwood Hill Park, Marble Hill |
Everything else (other boroughs, airports, the islands) is “other”. The same list applies to every year. A trip “touches” the zone if either end is in it. A trip is “outside” if neither end is in it or in a straddling zone. Marble Hill is a Bronx neighbourhood that TLC files under Manhattan; it has very few trips and does not matter.
Method
Section titled “Method”Trip counts. Trips per weekday, for each calendar month, year on year: 2025 vs 2024, 2026 (Feb–Aug) vs 2024, and 2024 vs 2023 as the “placebo” (before pricing, so it should show nothing). Two controls: all trips outside the zone, and trips with both ends above 60 St. Intervals (95%) come from resampling days within each month 500 times.
Speeds. Trips with both ends in the zone against trips with both ends above 60 St. Speed is miles divided by trip time, with outliers cut (2–90 minutes, 0.3–30 miles, 1–60 mph). The effect is the change in the zone’s mean log speed minus the change above 60 St: a difference in differences. Intervals by resampling days. A second version fixes the mix: each zone pair is compared with itself, weighted by its 2024 trips (about 1,440 pairs in the zone, 280–360 above 60 St), with intervals by resampling pairs.
Trend adjustment. The placebo is not zero: in-zone trips were losing ground to the controls before pricing. “Trend-adjusted” subtracts the placebo from the 2025 result. It assumes the earlier slide would have continued; treat it as the upper end, not the answer. For 2026 the same step would extend the trend two years and is not shown.
Crossing trips. Trips with one end in the zone, the same origin and destination zone pair before and after (pairs with at least 20 trips in both periods), change in log minutes. Control: pairs entirely above 60 St.
Trips per weekday, 6 am–8 pm, trips touching the zone.
| 2024 | 2025 | Change | 95% interval | Same measure, 2024 vs 2023 | |
|---|---|---|---|---|---|
| Yellow taxi | 59,900 | 68,000 | +13.5% | +11.2 to +15.8 | +5.0% |
| Uber/Lyft/Via | 142,600 | 140,700 | −1.4% | −3.3 to +0.5 | −2.1% |
| Both | 202,500 | 208,700 | +3.0% |
For 2026 (Feb–Aug) against the same months of 2024: taxi +4.8% (+2.1 to +7.4), Uber/Lyft −5.6% (−7.9 to −3.4).
Trips into the zone (start outside, end inside) show the same: taxi 14,100 → 16,300 (+16%), Uber/Lyft 38,000 → 37,400 (−1.5%, −3.0 to −0.1), about 580 fewer a day. Peak hours are the same: Uber/Lyft −1.3% (−3.3 to +0.7), taxi +15.1% (+12.4 to +17.8).
Against a control, the answer depends on the control. Change in zone-touching trips minus change in the control, 2025 vs 2024:
| Control: all outside | Control: above 60 St | |
|---|---|---|
| Taxi, 2025 | −12.9% (−14.3 to −11.6) | +1.1% (−0.4 to +2.7) |
| Taxi, placebo | −1.6% (−3.3 to 0.0) | −4.6% (−6.2 to −3.1) |
| Taxi, trend-adjusted | −11.5% (−13.9 to −9.1) | +6.1% (+3.3 to +9.1) |
| Uber/Lyft, 2025 | −7.0% (−8.6 to −5.5) | −4.0% (−5.2 to −2.7) |
| Uber/Lyft, placebo | −7.8% (−9.3 to −6.3) | −5.5% (−6.6 to −4.3) |
| Uber/Lyft, trend-adjusted | +0.8% (−2.0 to +3.7) | +1.7% (−0.5 to +3.8) |
- Uber/Lyft: the 2025 gap against either control matches the gap in 2024, before pricing. Once the earlier drift is removed there is nothing left (+0.8%, +1.7%). By 2026 the raw gap is wider (−17% against all outside) but so is two years of drift; adjusted for it, −2.5% (−7.0 to +2.1).
- Taxi: against trips above 60 St there is no effect. Against all outside trips it looks like −13%, but taxi trips outside the zone rose 30% in 2025 (airports and other boroughs, which we did not split out), so that control is poor. We lean on the above-60 St comparison for taxis.
- Against all outside trips, every one of the 11 months of 2025 shows the zone doing worse, for taxis and for Uber/Lyft. That fits the control problem above, not a charge effect: the placebo year shows the same gap for Uber/Lyft.
Translation. Fewer daily trips into the zone: for Uber/Lyft, about 580 a weekday on the raw count (1.5%), and none once the earlier drift is allowed for (the adjusted figure is a gain of 90 to 390 a day depending on the control). For taxis, none: 2,300 more a day. We cannot pin any drop in taxi or app trips on the charge.
Speeds
Section titled “Speeds”Weekdays 6 am–8 pm, trips with both ends in the zone against both ends above 60 St.
| In-zone mph | Above 60 St mph | Difference in differences | 95% interval | Trend-adjusted | Same zone pairs | |
|---|---|---|---|---|---|---|
| Taxi 2024 vs 2023 (placebo) | 7.4 → 7.1 | 9.3 → 9.2 | −2.2% | −3.5 to −1.0 | −2.5% (−2.8 to −2.0) | |
| Taxi 2025 vs 2024 | 7.1 → 7.2 | 9.2 → 9.1 | +2.1% | +0.7 to +3.5 | +4.4% (+2.0 to +6.9) | +2.1% (+1.5 to +2.8) |
| Taxi 2026 vs 2024 | 7.4 → 6.9 | 9.4 → 9.0 | −1.1% | −2.8 to +0.5 | −1.8% (−2.6 to −0.9) | |
| Uber/Lyft 2024 vs 2023 (placebo) | 7.8 → 7.6 | 10.1 → 10.1 | −2.1% | −3.0 to −1.2 | −1.9% (−2.1 to −1.6) | |
| Uber/Lyft 2025 vs 2024 | 7.6 → 7.6 | 10.1 → 10.0 | +1.9% | +0.9 to +3.0 | +4.1% (+2.4 to +6.0) | +2.0% (+1.6 to +2.4) |
| Uber/Lyft 2026 vs 2024 | 7.8 → 7.4 | 10.3 → 9.9 | −1.0% | −2.2 to +0.2 | −1.5% (−2.0 to −0.9) |
(mph is total miles over total hours, for the months compared; the difference is on mean log speed.)
- Inside the zone, speeds did not rise; they stopped falling (taxi +0.3%, Uber/Lyft −0.2%). Above 60 St they slowed by about 2%. The gain is the gap.
- Ten of 11 months in 2025 are positive for both, from −2.5% to +6.3% for taxis, −1.7% to +5.3% for Uber/Lyft. Single months swing a lot.
- Peak hours: taxi +2.5% (+1.2 to +3.8), Uber/Lyft +1.8% (+0.8 to +2.8).
- 2026 against 2024 is no gain on the plain comparison (peak hours for taxis: +0.1%).
Minutes saved on a 2-mile trip inside the zone (2025 vs 2024; such a trip took 17.3 min by taxi and 17.5 min by Uber/Lyft in 2024):
| Plain gain | Seconds | Trend-adjusted | Seconds | Looking only at 2-mile trips | Seconds | |
|---|---|---|---|---|---|---|
| Taxi | +2.1% | 21 | +4.4% | 44 | −4.4% in time (−5.9 to −3.0) | 46 |
| Uber/Lyft | +1.9% | 20 | +4.1% | 41 | −1.9% in time (−3.1 to −0.7) | 20 |
So about 20 to 45 seconds. That is the rough size of what traffic could give a trip on these streets; most of a 2-mile Manhattan trip is waiting at lights and loading.
Trips that cross the zone edge
Section titled “Trips that cross the zone edge”Same zone pair before and after, one end in the zone. Change in trip time minus the same change for pairs above 60 St (negative means faster), 6 am–8 pm.
| 2025 vs 2024 | 2024 vs 2023 (placebo) | 2026 vs 2024 | |
|---|---|---|---|
| Taxi, all crossing (3.4M trips) | −3.3% (−3.9 to −2.8); raw −1.0% | +0.9% | −1.8% (−2.5 to −1.1) |
| Uber/Lyft, all crossing (14.1M) | −3.0% (−3.4 to −2.6); raw −2.2% | +1.2% | −0.5% (−1.2 to 0.0) |
| Uber/Lyft, to/from Manhattan above 60 St | −1.5% (−1.8 to −1.1) | +1.5% | +2.0% |
| Uber/Lyft, to/from other boroughs | −4.2% (−4.6 to −3.9) | +0.8% | −2.2% |
| Uber/Lyft, to/from airports | −2.0% (−2.5 to −1.6) | +1.6% | +0.5% |
Raw, the same pair took 1.0% (taxi) and 2.2% (Uber/Lyft) less time. A 30-minute trip gains about a minute. Trips to and from other boroughs gain most. The same pattern holds at peak hours. These intervals come from resampling pairs and leave out day-to-day shocks that hit every pair, so they are narrower than they should be.
Caveats
Section titled “Caveats”- Cars in these data are for-hire cars only. Private cars and trucks, the bulk of what the charge targets, are not in the TLC files. For the charge’s effect on traffic, use the MTA’s vehicle entry counts.
- The fee on these trips is small ($0.75 a taxi trip, $1.50 for Uber/Lyft, per trip that starts, ends or passes through the zone), so little change is expected from price alone.
- Taxi trips rose 14–16% across all of 2025 in these files (3.58M rows in March 2024, 4.15M in March 2025). We did not find a cause. TLC’s data have changed before (new reporting rules, trips added late); check against TLC’s own monthly totals before leaning on the taxi counts.
- The placebo is not flat. In-zone trips were losing speed and share against the controls before pricing, so the plain and trend-adjusted figures bracket the answer; neither is a clean cause.
- Control choice matters for taxi counts (−13% vs +1%). Trips outside the zone include airport and outer-borough trips with their own trends.
- TLC gives zones, not streets. A “trip in the zone” can run on the FDR Drive or West Side Highway. The six straddling zones are dropped, which loses some trips that did enter the zone.
- Speeds are crude. Trip time includes waiting for the passenger and the dropoff, and distance is the meter’s. We cut trips under 2 minutes, over 90, under 0.3 miles, over 30, and under 1 or over 60 mph (6% of in-zone taxi trips and 0.4% of in-zone Uber/Lyft trips are cut). Driver and trip mix can change: the same-pair version holds the mix of zone pairs fixed but not trip length within a pair.
- Day-level intervals understate the real uncertainty because weather, events and construction hit whole weeks. The by-month spreads above are a better guide.
- Holidays other than federal ones (the day after Thanksgiving, Christmas week) stay in.
- 2025 January and 2026 months. January is left out of all comparisons. 2026 covers February to August only, set against the same months of 2024.
- Fee data. Not every in-zone trip carries the fee in the data (fee shares were 87–97%), so the zone list rests on shares, not on a flag per trip.
- Uber/Lyft pooled. The file mixes Uber, Lyft and Via. Not split out.