CUNY Datathon 2026 · Phase 1
Beyond the Average
Mapping Traffic, Air Quality, and Environmental Justice After NYC Congestion Pricing
“Did traffic and air pollution change evenly across New York City after congestion pricing began, or does a neighborhood-level analysis reveal a more complicated pattern, especially in communities that already experience high environmental burdens?”
New York City's congestion pricing program is frequently analyzed using aggregate citywide statistics. This project moves below those citywide averages to examine neighborhood, corridor, and time-of-day variation — and to be explicit about what the evidence can, and cannot, support.
01 · Citywide Overview
Beyond one citywide number
To investigate whether change was uniform, the analysis compared matched 2024 and 2025 periods across traffic crossings and air-quality monitors citywide — before narrowing in on any single neighborhood.
How to read this map:
Thin dark outlines and labels mark the five NYC borough boundaries, dissolved directly from official U.S. Census tract geography. Triangles are MTA crossings (red = higher 2025 traffic, blue = lower). Circles are NYCCAS PM2.5 monitors, colored by season-matched change (red = higher, blue = lower in 2025). The orange/tan tract shading is official NYS-designated disadvantaged-community (DAC) geography — existing environmental-justice context, not something congestion pricing caused. The purple polygon is a schematic Congestion Relief Zone outline. The star marks the Cross Bronx case-study location.
Across the ten MTA bridge and tunnel locations included in the model, eight showed higher average daily traffic in 2025, while two — the Queens Midtown Tunnel and the Hugh L. Carey Tunnel — showed lower average traffic than the same matched period in 2024. Among 75 NYCCAS PM2.5 monitoring sites with usable season-matched data, 42 recorded lower readings in 2025 and 33 recorded higher readings. The median monitor-level change was approximately −0.166 µg/m³.
These traffic patterns are descriptive. An observed increase at a specific crossing does not, by itself, prove that congestion pricing caused vehicles to divert there. The important finding here is spatial variation — not a single universal increase or decrease across the city.
Understanding the PM2.5 numbers above
What is PM2.5?
PM2.5 refers to fine airborne particles with diameters of 2.5 micrometers or smaller. Because these particles are so small, they can be inhaled deep into the lungs. Exposure to PM2.5 is associated with respiratory and cardiovascular health effects.
Why use PM2.5 in this project?
PM2.5 was selected as the primary citywide air-quality measure because it is health-relevant and NYCCAS monitors it across neighborhoods throughout New York City, allowing spatial patterns to be compared across many locations. PM2.5 can come from multiple local and regional sources, so changes should not be interpreted as traffic effects alone.
How should the −0.166 µg/m³ median be interpreted?
The median monitor-level change was a small decrease of approximately −0.166 µg/m³. This does not mean every neighborhood experienced that decrease. Of the 75 monitoring sites analyzed, 42 had lower PM2.5 readings and 33 had higher readings. The more important finding is the variation across locations rather than the citywide median alone.
Sources: NYC Community Air Survey; U.S. Environmental Protection Agency.
02 · Time of Day
Daily averages can hide differences
A single daily average can mask sharply different behavior at different hours. Separating weekday traffic into commuter peak and off-peak periods reveals patterns a topline number would hide.
Definition note: Weekday commuter peak was defined for this analysis as 6:00–9:59 AM and 3:00–6:59 PM. This is my analytical definition for the project, not the MTA's official congestion-pricing toll-period definition.
Individual crossings behaved quite differently from the network-wide figures. At the Throgs Neck Bridge, peak traffic rose by approximately +1.1% while off-peak traffic fell by approximately −0.8% — two opposite signs at the same facility, invisible in a single daily average. The Bronx–Whitestone Bridge, Queens Midtown Tunnel, and Hugh L. Carey Tunnel each showed their own distinct peak/off-peak splits as well.
This is exactly why the project looks past one daily number: a single average can hide meaningfully different time-of-day dynamics at the same crossing.
03 · Air Quality
Air quality and the Congestion Relief Zone
Air-quality monitors don't all move together, and comparing them fairly across 2024 and 2025 required season matching and reference-site adjustment before geographic patterns could be compared more carefully.
- Season matching — comparing NYCCAS sampling periods from the same season in 2024 and 2025, and excluding sampling windows that straddled the start of congestion pricing.
- Reference-site temporal adjustment — using NYCCAS reference sites to account for citywide pollution shifts unrelated to any single policy or location.
- Monitor-to-monitor variation — PM2.5 readings differ meaningfully across locations, so single-site changes are interpreted in the context of that variation, not in isolation.
- Congestion Relief Zone geography — the CRZ boundary is shown for geographic context, not as a proof of causal effect inside or outside the zone.
This map is a supporting, descriptive visual. Geographic overlap between a monitor's PM2.5 change and the Congestion Relief Zone is not, on its own, evidence that congestion pricing caused that change.
04 · Why the Cross Bronx?
Selecting a high-burden corridor for deeper study
The Cross Bronx Expressway was not selected because the citywide map had already shown that congestion pricing harmed it. It was selected because the surrounding community already carries an extremely high transportation and health burden — making it a place where scrutiny is warranted, regardless of what the citywide averages showed.
Putting the asthma burden in context
NYC: 66.4 per 10,000
NYC: 143.7 per 10,000
The tract-level percentile above shows that the Cross Bronx study area ranks among NYC’s highest-burden locations on the disadvantaged-community asthma indicator. A broader neighborhood measure tells a similar story: 2023 emergency-department data show High Bridge–Morrisania asthma visit rates well above the citywide average, for both adults and children ages 5–17, as shown above.
These emergency-department rates describe the broader High Bridge–Morrisania neighborhood, not Census Tract 227.01 specifically. They provide health context for the study area and should not be interpreted as evidence that congestion pricing caused asthma.
Source: NYC Department of Health and Mental Hygiene, Environment & Health Data Portal, Asthma and the Environment in High Bridge–Morrisania (2023).
These indicators show why even statistically uncertain pollution results matter in a community already carrying a substantial baseline burden. High environmental-justice burden is a reason for closer scrutiny — not proof that congestion pricing caused harm here.
05 · Cross Bronx Deep Dive
Traffic fell. Air quality was inconclusive.
At NYSDOT station 0191, traffic measurements were matched to the NYCCAS sampling windows, which were approximately two weeks long — the most temporally precise comparison in the project.
How the Cross Bronx traffic and PM2.5 windows align
- NYCCAS neighborhood PM2.5 sampling is periodic, not continuous at every site — each sample represents an approximately two-week integrated sampling window, not an hourly or daily reading.
- 2024 sampling windows were paired with the comparable seasonal period in 2025 at the same site (season matching).
- Any sampling window that straddled January 5, 2025 — the start of congestion pricing — was excluded, so no window mixes pre- and post-policy days.
- For the Cross Bronx case study, NYSDOT station 0191 hourly traffic was averaged over the same date window represented by each NYCCAS PM2.5 sample. That shared date range is how the traffic and PM2.5 periods were aligned — there is no separate, independent traffic time period.
- Hours flagged by NYSDOT quality-control checks were excluded from the traffic average for that window.
| NYCCAS PM2.5 window | Traffic data used | Status |
|---|---|---|
| Jan 22 – Feb 4 | NYSDOT 0191, same dates | Post-policy |
| Feb 19 – Mar 4 | NYSDOT 0191, same dates | Post-policy |
| Mar 19 – Apr 1 | NYSDOT 0191, same dates | Post-policy |
| Apr 16 – Apr 29 | NYSDOT 0191, same dates | Post-policy |
| May 14 – May 27 | NYSDOT 0191, same dates | Post-policy |
| Jun 11 – Jun 24 | NYSDOT 0191, same dates | Post-policy |
| Jul 9 – Jul 22 | NYSDOT 0191, same dates | Post-policy |
| Oct 1 – Oct 14 | NYSDOT 0191, same dates | Post-policy |
| Nov 26 – Dec 9 | NYSDOT 0191, same dates | Post-policy |
These nine roughly two-week windows are the only 2025 periods with usable, non-straddling PM2.5 samples at the Cross Bronx site — they do not represent continuous or year-round monitoring. An equivalent set of pre-policy 2024 windows (matched by season) was used for the “before” comparison shown below.
Citywide PM2.5 (Figures 1 and 3) uses season-matched observations across all 75 NYCCAS monitors with sufficient matched 2024–2025 data. The Cross Bronx case study below is narrower and more precise: it uses these nine specific ~2-week PM2.5 windows at one site, with NYSDOT 0191 traffic matched to those exact same dates.
Sources: NYC Community Air Survey sampling schedule; NYSDOT Traffic Data Viewer, Station 0191.
This decline is important because it challenges a simple expectation: if traffic volume were the dominant driver of local air pollution, a decline in traffic might be expected to produce a detectable improvement. The adjusted air-quality models show why that relationship is not so simple.
The adjusted air-quality results were statistically inconclusive.
All three point estimates were positive, but every 95% confidence interval included zero. This means the data do not provide statistically conclusive evidence of either a policy-related increase or decrease in NO₂, PM2.5, or black carbon at this location. The available evidence indicates that matched-window traffic declined, while the adjusted air-quality signal remained inconclusive — not that pollution definitely rose, definitely fell, or that congestion pricing caused any change here.
Why can lower traffic coexist with an inconclusive air-quality result?
Several alternative explanations and data limitations were investigated rather than assumed away. None of the sensitivity checks provided evidence that changes the central conclusion: the local air-quality results remain statistically inconclusive.
Weather was different — and already controlled for
- −1.07°C temperature
- +0.33 m/s wind speed
- −0.70 mm/day precipitation
- Rainy days: about the same
The adjusted models already control for temperature, wind, precipitation, and season, so these 2025-vs-2024 weather differences were accounted for — not ignored.
Broader NO₂ conditions did not explain the local pattern
- Raw Cross Bronx: +1.58 ppb
- Reference-adjusted: +2.36 ppb
- Reference sites: −0.32 ppb
Reference-site NO₂ declined slightly over the same period, so a general rise in citywide background NO₂ does not appear to explain the Cross Bronx pattern. This does not make the Cross Bronx change causal.
Wildfire smoke was explicitly tested
- Pre-policy: 50 smoke days
- Post-policy: 26 smoke days
- Pre: 5 medium + 1 heavy day
- Post: light overlap only
| Original | +0.359 µg/m³ |
| With smoke-day control | +0.330 µg/m³ |
| With weighted smoke-intensity control | +0.321 µg/m³ |
| Excluding medium/heavy-smoke windows | +0.447 µg/m³ |
Every 95% confidence interval across these sensitivity models still included zero. NOAA HMS identifies satellite-observed smoke overhead, not direct proof of ground-level exposure — this analysis does not support increased 2025 wildfire smoke as the explanation for the inconclusive PM2.5 result.
Vegetation change was investigated
A Sentinel-2 NDVI vegetation-greenness sensitivity analysis — not a direct tree-count or canopy-loss measurement. Four seasonally matched summer image pairs were compared between 2024 and 2025.
- Mean NDVI change: ≈ 0
- Veg. cover >0.20 NDVI: +0.39 pp
- Mean NDVI change: −0.005
- Veg. cover >0.20 NDVI: −1.05 pp
- 3 of 4 pairs lower in 2025
This suggests essentially stable vegetation very near the monitor and a small, mixed decline over the broader 1-km area — not evidence of a large or consistent vegetation decline capable of clearly explaining the air-quality result.
Traffic volume is not the same as emissions
NYSDOT station 0191 records total traffic, direction, lanes, and time — but not vehicle class. The approximately 4.5% decline in matched traffic volume cannot be assumed to represent a 4.5% decline in emissions. Changes in vehicle composition are a plausible mechanism, but this dataset cannot test them.
Pollution sampling is relatively sparse
- NO₂: 11 + 11 = 22 obs.
- PM2.5: 10 + 9 = 19 obs.
- Black carbon: 10 + 9 = 19 obs.
NO₂ samples represent roughly 15.6-day integrated windows on average; PM2.5 and black-carbon samples represent 13-day windows. These are periodic integrated samples, not continuous measurements — which limits statistical precision compared with the much denser hourly traffic data.
The result is not “air quality did not change.” The result is that the available data do not establish a statistically conclusive local improvement or worsening.
PM2.5 robustness check: removing each observation one at a time produced 41 alternative models. All 41 still had 95% confidence intervals containing zero, and 0 of 41 had p < 0.05 — indicating the inconclusive PM2.5 result was not driven by a single unusual observation.
Sources: NYC Community Air Survey; NYSDOT Station 0191; NOAA Hazard Mapping System smoke polygons; Sentinel-2 Level-2A imagery via Earth Search. Sensitivity analyses by author.
06 · Methodology
How I tested the question
The analysis moved from broad, exploratory citywide patterns to a more rigorous, explanatory model of a single corridor.
- 1Explore citywide traffic & PM2.5 patterns
- 2Match comparable dates & seasons
- 3Adjust using NYCCAS reference sites
- 4Select a high-burden study corridor
- 5Match traffic to NYCCAS sampling windows
- 6Add a Van Wyck comparison corridor
- 7Run a Difference-in-Differences-style model
- 8Add weather & seasonal controls
- 9Test the pre-policy trend diagnostic
- 10Report confidence intervals
- 11Interpret cautiously
In plain terms: the project first looked broadly across the whole city without assuming where an effect might appear. After identifying the Cross Bronx as a location worth deeper scrutiny, the analysis added the Van Wyck Expressway as a comparison/control corridor and asked whether the change after the policy period at the Cross Bronx differed from the change at that comparison location — a Difference-in-Differences-style design. The Van Wyck was used to help separate broader, citywide temporal changes (weather, season, general trends) from anything more specific to the Cross Bronx; NYC Health has similarly used the Van Wyck as a comparison/control site in its own environmental surveillance work, in part because its traffic was not expected to change as a direct result of congestion pricing. Weather and seasonal controls were added because pollution can vary with meteorology even when traffic stays constant. Before trusting that design, the pre-policy trends between the two corridors were checked for similarity.
That check did not go well: the Cross Bronx and Van Wyck pre-policy trends were not strongly parallel. This means the Van Wyck comparison helps rule out some broad citywide explanations, but it does not provide definitive causal proof for the Cross Bronx results — see Limitations below.
Expert-Informed Methodology
Feedback encouraged moving beyond a simple before/after comparison — incorporating a comparison geography, weather and non-policy controls, a pre-policy trend test, and overall causal caution.
Feedback emphasized that traffic spillover and diversion can be difficult to establish causally, and should not simply be assumed from an increase observed at individual crossings.
Guidance shaped the overall workflow: explore broadly → test rigorously → select defensible findings → explain the “so what?”
07 · Limitations
What this analysis can — and cannot — show
- Citywide MTA traffic changes are descriptive. They cannot independently prove that congestion pricing caused a particular crossing to gain or lose traffic.
- Geographic overlap between traffic change, pollution change, and disadvantaged-community geography is not proof of causation.
- NYCCAS sampling is relatively sparse and seasonal — many observations represent roughly two-week windows, which limits temporal resolution.
- The Cross Bronx / Van Wyck pre-policy parallel-trends diagnostic was weak, which weakens a strict causal Difference-in-Differences interpretation.
- For this reason, the Difference-in-Differences-style estimates are presented as adjusted but statistically uncertain — not as definitive causal effects of congestion pricing.
- Confidence intervals are used throughout to communicate that uncertainty rather than to imply a precise, settled answer.
08 · Why This Matters
A single statistic cannot tell the whole story
A single citywide statistic cannot tell us where changes happened, when they happened, or how they interact with existing environmental burdens. The Cross Bronx remains a critical site for future attention — not because Phase 1 proved harm, but because the community already carries an extremely high baseline burden.
“The average tells us whether the city changed.
The neighborhood tells us who experienced that change.”
09 · Looking Ahead
Phase 2 — reimagining the Cross Bronx
Phase 1 changes the question that matters most for the next phase of work.
“What changed across NYC after congestion pricing?”
“What interventions could meaningfully improve conditions in communities where transportation and environmental burdens remain high?”
The Cross Bronx provides a strong case study for this question. Potential future interventions to evaluate could include:
- Highway capping
- Boulevard conversion
- New greenspace
- Freight-management changes
- Improved pedestrian connections
Any future redesign should be evaluated across at least three outcomes: traffic redistribution, pollution exposure, and environmental equity — with the goal of avoiding simply shifting one neighborhood's burden onto another.
Phase 1 did not simulate or test these interventions. They represent a direction for future work, not a result of this analysis.
Data Sources
Sources
- Metropolitan Transportation Authority — Bridges and Tunnels Hourly Crossings
- New York State Department of Transportation — Traffic Data Viewer
- New York City Community Air Survey (NYCCAS)
- New York State Climate Justice Working Group — disadvantaged-community data
- NOAA weather observations
- New York City Department of Health and Mental Hygiene — Environment & Health Data Portal
- NOAA Hazard Mapping System — smoke polygons
- Sentinel-2 Level-2A imagery via Earth Search
- U.S. Census Bureau
- U.S. Environmental Protection Agency