urban planner | educator | maker
js6676@columbia.edu
cv  


projects
urban planningurban technologytransportation
climate
urban design
master’s thesis



blog 
about 





















© Jui Shih, 2023 - 2026




Investigating Central Business District Tolling Program in NYC
One Year Later




Subject(s)
Transportation Planning, Urban Planning, Public Transportation, Congestion Pricing

Course
Advanced Spatial Analysis

Tools
ArcGIS Pro, Python, Adobe Illustrator, Adobe InDesign

Project Location    
New York, NY

Completed Period
May, 2026





This is a group final project investigating New York City’s Central Business District Tolling Program one year after it went into effect. This is the first congestion pricing scheme in the United States and has received intense public attention and contestation since its proposal. We examined the program’s impact across three dimensions: traffic volume, modal shift, and air quality. We sourced datasets on automated traffic counts, MTA subway and Citi Bike Ridership, taxi trips, and satellite-measured NO2 and applied various spatial analysis tools, e.g. Global Moran’s I, LISA clustering, and Getis-Ord Gi& hotspot analysis, to trace where congestion pricing’s effects landed and where they did not. 
























Background


Historically, New York City's transportation network is one of the most heavily utilized urban mobility systems in the world, serving millions of residents, commuters, and visitors daily. Simultaneously its also been known as the most congested city in the United States and one of the slowest in the world. In 2024, New York was ranked as the slowest city in the U.S. by Tom Tom Traffic Index. According to the Department of Transportation, the average speeds in Manhattan Central Business District (CBD) crawled between 7-9 mph during business hours, barely faster than a brisk walk. This gridlock has environmental and economic impact. 

For nearly twenty years, the congestion relief zone tolling program has been in various stages of planning and finally after decades of delay, the Central Business District Toll Program (CBDTP) was implemented and went into effect on January 5, 2025. This traffic management system charges a variable fee to vehicles entering Manhattan from the 60th Street line down to the Battery, excluding West Side Highway on the western border and FDR on the east. The program’s goal is to reduce gridlock, lower carbon emissions, and generate revenue specifically for the MTA to upgrade New York’s aging subway, bus, and rail infrastructure. 

Early analysis of the NYC program found CBD road speeds increased roughly 15 percent in the first months, with speed gains extending to roads outside the zone that share high traffic co-occurrence with the CBD (Cook et al., 2025). The Regional Plan Association (2025) found delay reductions across the broader metro region, including in the Bronx and Bergen County, NJ. These studies focus on aggregate effects; less is known about the spatial distribution of those effects at the neighborhood level, or about whether any rerouting placed disproportionate burden on lower-income areas with fewer transit options.

Project Description

While our project consists of four members sharing three main components, this webpage focuses on what I was responsible of. My portion of the project tackled the traffic volume analysis, which is the entry point into measusuring what actually changed on the streets. Specifically, I was to identify traffic hot spots before (2024) and after the implementation of the Central Business District Tolling Program (CBDTP) in NYC from January 2025 onwards. However upon further data exploration and cleaning, I discovered that the data containing the recorded traffic is inconsistent. The locations recorded by the automated traffic recorders (implemented by DOT) do not have the same recordings for both 2024 and 2025 and each recorded location has different numbers of sampling. These discrepancies made it hard to do a direct comparison of traffic volume before and after the CBDTP implementation. As such, I sourced other datasets such as MTA subway ridership to find the volumes of other transportation modes and investigate if there were any changes at each subway station. 



Untangling NYS DOT Automated Traffic Volume Data Count


As we discovered that traffic detections are not conducted uniformly at the same spot in the same frequency, even with a large dataset spanning full 2 years (Jan 2024 - Dec 2025), it was extremely difficult to find differences or volume changes. However, I decided to keep the data as is and visualize the locations availabe, despite that they may not line up or overlap. This is intentional, as it sheds light on the limitation of data. Blue dots are vehicle volume detected in 2024, the year before congestion pricing went into effect, and green dots are vehicle volume detected in 2025, a full year after. I isolated 4 different time periods and visualized them using the same method- proportional symbols of both years, overlaying the locations that recorded in both years. The points with both years were calculated to render  % change of the daily vehicle volume between 2024 & 2025. 


Rush Hour AM weekdayRush Hour PM weekday




Weekday Average daily traffic volume Weekend Average daily traffic volume



Before we even get into the “before vs. after” comparisons, the most  important structural fact is that only 9 of 210 total segments within the selected two year span dataset were measured in both periods (2024 vs. 2025). There were 96 segment monitors in 2024 and 105 new segments in 2025 post Congestion Pricing. This means the change calculated could not be representative of traffic behavior as there are only 9 matching segments that are the only true before/after data comprison. 

Despite this, I believe there is still something to explore, i.e. targeting specific hours in the day such as rush hours, late nights etc to inform some kind of pattern. The following are a few findings that are worth noting. 

  • Rush hour in the evenings have almost two times more vehicle volume than the morning rush. This makes sense as people commute to Manhattan in the morning but disperse across the 5 boroughs in the evening, concentratin outbound pressure on local roads. 

  • Adding on, Queens stands out for late night travel (10PM - 5AM), which reflects the borough’s role as a gateway for overnight freight, airport traffic, and expressway through traffic. Brooklyn showed elevated night time.

  • The Brooklyn Queens Expressway is the highest volume corridor in the data, with highest raw count of any segment at around 62K - 67K vehicles per weekday.

The 9 matched detection points show some decline on key corridors:

  • Williamsburg Bridge approaches fell -6% ~ 7% in rush hour volume

  • West End Avenue decreased -4.9% in hour hour but its all day weekday volume actually increased +7.8% - does this suggest traffic shifted out of peak hours? 

  • 9th Avenue decreased -54% in rush hours and -47% in total weekday volume. With such high figure, this might just be a dramatic outlier, reflecting a localized change e.g. road work, closure etc rather than congestion pricing.

  • University Heights Bridge (crossing Bronx-Manhattan) actually grew +16% in AM rush hour specifically, possibly absorbing diverted traffic from priced crossings.
  • East 58th Street grew 10.5%, perhaps another diversion effect from other areas. 




Limitations


As stated, only 9 locations have records in both 2024 and 2025. The other recorded data segments are entirely different street segments with no overlap. As I understand through further research is that NYC DOT’s automated traffic counters are not permanently installed sensors, but portable counters that get deployed to different locations each counting cycle. Additionally, the sensors may not have been placed at the same location in both years for the 9 that matched up! 

The records also show uneven distributation across boroughs with Brooklyn having the most entries, then Manhattan, Queens, Bronx, and Staten Island. There are high traffic corridors in the Bronx and Staten Island that are not represented.



Subway Ridership before and after Congestion Pricing

While traffic volume on the road was rather difficult to analyze, I seeked other datasets that may capture volume in other ways. I explored several other datasets e.g. Bus Ridership, Annual Average Daily Traffic, (AATD), and finally landed on Subway Ridership datasets as they appear to have robust and consistent collection across stations and time periods. To understand how mobility patterns may have shifted, I compared average daily subway ridership before and after the Congestion Relief Zone Toll. Using station-level ridership records, I separated weekday and weekend averages and calculated the percentage change between the two years. First I intentionally kept the subway stations where they are without aggregating them onto a separate spatial unit. The maps on the left visualize ridership change using graduated color classifications with positive % change in red and negative % change in blue. The maps on the right were created using the same dataset, where I joined station ridership changes onto 1/4 square mile hexagonal tessellations. 

Several neighborhoods outside the Congestion Relief Zone, particularrly in parts of the Bronx, Queens, and Brooklyn experienced notable increases in subway ridership, suggesting potential shifts in travel behavior and transit reliance after the toll took effect.  




Weekday Ridership % Change
Weekday Ridership by tessellation





Weekend Ridership % ChangeWeekend Ridership by tessellation




When I pivoted to explore this data, my concern for proocessing the large dataset had me starting with a small batch, spanning only the months of April in 2024 and 2025. After cleaning the dataset, I found interesting phenonmenon/patterns with certain amount of shifts that took place across MTA Subway. Considering the limited time available, I decided to continue on with the dataset I have, of which I have cleaned, rather than downloading a new dataset to be cleaned and filtered at the risk of not being able to process them all in time. 

At a glance, of the total 428 stations, the average subway ridership during the week increased 11.9% versus 5% increase over the weekend. Almost every  station gained riders on weekdays. The maps show on the left side represents weekday and weekend changes, and ones on the right are daily average ridership tabulated and statistically summarized within hexagons using tessellation tool.  To visualize the different digits (increases shown with + , and negative shifts with - ), I used red for the  highest increase and blue for the negative increase and yellow-orange for the middle. 





What the maps illustrate:

Top/Left - Some  clusters showing increase ridership in the Bronx and Upper Manhattan covering neighborhoods in East Harlem, West Concourse/Mount Hope, Woodlawn/Williamsbridge Olinville, Hunts Point/ Crotaona park east; Queens stations such as Point-Sunnyside-West- Queens 

Bottom/Left: The increases are seen in the neighbohoods of Central Harlem, Huns Point, and Manhattan; a bit patchier and smaller.

Top/Right: Some concentrations along the northern and eastern corridors.  The few blue hexes representing decreases are mostly concentrated in specific spots possible tied to service suspensions rather than congestion pricing.  

Bottom/Right: Decline in ridership appear sharpest during the weekends as seen in the light blue/blue hexagons representing negative figures, possibly deterring casual day-trippers. 

Stations with extreme decreases are most likely reflective of service suspensions at those station rather than congestion pricing effects. Similarly, the +250% increases are likely as a result of station service restoration that was previously closed. 

The most visible shift after the toll went into effect is the weekday ridership surge system (+1-21%, as illustrated in the light yellow hexagons) 

Additional Notes



This chart on the right compares total transit ridership counts by day of the week between 2024 & 2025 across multiple transportation modes in NYC.  After exploring the subway ridership dataset, I was curious about the overall trend in subway, bus, communter rail, and bridge /tunnel usage before and after congestion pricing implementation. 

Subway ridership dominates all other transit modes, consistently recording the highest daily volumes in both years. Weekday ridership in 2025 exceeds 4 million riders on most weekdays
     


Made in Python, this chart on the right shows traffic volume entering the toll zone by days and time period since Congestion Pricing began. This data set was only made available after toll was implemented and therefore do not have prior years to compare with. 
   



Other Datasets to consider


Another set of data that might have been interesting to explore is  the “Annual Average Daily Traffic (AADT) that the New York State Dept of Transportation recorded and maintained, but has since year 2020 been discontinued. It is an estimate of the average daily traffic along a defined segment of roadway, which is calculated and statistically derived from short term counts taken along the same section of highway.  However, it is heavily reliant on statistical applications to determine valid forecast and estimates, so afterall it may not be as suitable. 

Data Sources: 


  • Metropolitan Transportation Authority
  • New York State Department of Transportation
  • NYC Department of City Planning





Project Deliverables // Individual Memo






Project Deliverable // Final Group Poster



BACK