Congestion Pricing – New York City
What does data and research tell us about congestion pricing?
Thank you for reading our work! Nominal News is an email newsletter read by over 5,000 readers that focuses on the application of economic research on current issues. Subscribe for free to stay-up-to-date with Nominal News directly in your inbox:
Our updated goal for 2025 is to hit 10,000 subscribers:
If you would like to support us further with reaching our subscriber goal, please consider sharing and liking this article!
On January 5, New York City implemented cordon-based congestion pricing in New York City’s Central Business District (CBD), the first program of its kind in the US. Under this program, vehicles entering the CBD pay a fee – in most cases the fee is $9. Similar programs exist in other cities – most notably London and Stockholm.
The public and economists are interested in determining what impacts the implementation of congestion pricing has had on traffic. A recent study by Cook, Kreidieh, Vasserman, Allcott, Arora, van Sambeek, Tompkins and Turkel (2025) (“Cook et al.”) studied just that using an interesting econometric method – synthetic control.
The Challenge of Modeling
Determining how congestion pricing impacts the traffic in New York City is not an easy question to answer. It might be tempting to simply look at the number of cars entering the congestion zone and compare it to the time before congestion pricing started (before January 2025).
The problem with this simple approach is that each month could have different traffic patterns. For example, maybe there are fewer cars on the road in December and more in January on average This may skew the comparison making it seem like congestion pricing didn’t impact traffic flows.
It may be tempting then to compare the same months from the prior year – January 2025 vs January 2024. However, this approach could again be problematic as there may have been shifts in car behavior between 2024 and 2025. An example of such a shift could be the fact that more workplaces in New York required people to come into the office, which means more people need to commute to work, increasing the number of cars on the road year over year. Another issue could be weather patterns – if January 2025 was colder than January 2024, this can alter car usage independent of congestion pricing.
These external forces, such as weather patterns, return to office mandates, and holidays may all skew the results. A method that can control for any and all of these types of external causes is needed to be able answer the question of what is the impact of congestion pricing on traffic.
Experiment Approach
In medical trials, the way researchers control for these external causes is by using ‘placebo’ based experiments. Take two groups of people, give one group the ‘treatment’ and give the other group a ‘placebo’ (no treatment). If the number of participants in the experiment was sufficiently large and who was treated was randomly selected, it can be argued that any external differences between the ‘treated’ group and ‘placebo’ group (for example, age, gender, behavior, etc.) wash out. Basically, the two groups are identical to each other. Then any outcome difference between these two groups is investigated and the difference is attributed to the ‘treatment’, as all other external factors are controlled for.
In the congestion pricing context, the enactment of congestion pricing is the ‘treatment’. Thus, one could consider comparing New York (the city that got treated) to other cities like Philadelphia or Boston (cities that are not treated). However, can we really say that Philadelphia and Boston are more or less the same as New York? Probably not.
What would be great is we had another New York. Better yet, it would be great if we could have New York get congestion pricing and not get congestion pricing at the same time, in a multiverse type world. But that’s impossible… or is it?
Enter “Synthetic Controls”
What if we could create another New York that would not get congestion pricing. A statistical method called the synthetic control method does just that. In the most basic form, the synthetic control method creates an artificial New York that is a blend, or ‘synthesis’, of other cities.
Creating a Synthetic City
How do we create such a city? The ‘artificial’ New York should capture certain important features of real New York – for example, artificial New York should have similar population growth dynamics to real New York. Thus, to capture this feature, we would look at other cities like Boston and Philadelphia and determine that if we took 60% of Boston with its population growth and 40% of Philadelphia with its population growth, we would get New York's population growth.
This artificial New York – a blend that's 60% Boston and 40% Philadelphia – would be considered to be identical to New York.
Naturally, researchers would want to make artificial New York mimic not only real New York’s population growth, but other facts as well – weather, traffic behavior, economic growth, etc. This simply alters what percentage is placed on every other city – i.e. maybe New York is 30% Boston, 20% Washington DC, and 50% Philadelphia.
Cook et al. used the synthetic control method to determine the impact of congestion on New York. Cook et al. made an ‘artificial' or synthetically controlled New York, whose traffic patterns closely matched that of real New York prior to the implementation of congestion pricing.
Then, Cook et al. argue, since real New York got congestion pricing while synthetic New York didn't (as it is a mix of cities that did not get congestion pricing), any changes in traffic patterns between them must have been due to congestion pricing.
In a nutshell, had New York not implemented congestion pricing, the traffic patterns we would have observed in New York are the ones synthetic New York experienced.
Impact of Congestion Pricing
Inside the Zone
After creating synthetic New York, Cook et al. analyzed what occurred in actual New York. Speeds in the CBD (the area impacted by congestion pricing) increased by 15%. Overall, speeds pre-congestion pricing in the CBD were 8.2 miles per hour, and 9.7mph after. It is worth noting that this is an 18% increase, but the authors adjusted for the fact that speeds in other cities (that made up synthetic New York) also slightly went up – by around 3%.
Outside the Zone
Cook et al. also looked at roads outside of the congestion pricing zone. Typically, congestion pricing is assumed to push traffic out to other roads.1 Any reduction in traffic in the congestion pricing area could be offset by increased traffic in other areas.
Cook et al. partitioned the roads into several categories by co-occurrence with the congestion pricing area. A road with 80% co-occurrence, for example, means that 80% of the car trips on that road end up going into the congestion pricing zone.
Cook et al. found that speeds on the highest co-occurrence (80% to 100%) roads went up by 16%. Even roads with 0% to 20% co-occurrence saw speed increases, albeit smaller at 4%. Overall, Cook et al. found that no type of road or route, whether inside in the congestion pricing area or outside of it, saw any speed reduction. That is, it appears that there has been no negative spillover.
Congestion Pricing – a Policy that Works
Congestion pricing is often a contentious policy as many, especially drivers, view it as an extra cost without any benefits. The Cook et al. research shows that congestion pricing has increased travel speeds within the zone and some parts outside of the zone significantly. This benefits car drivers as well as they have to spend less time in traffic. Even car accidents appear to have fallen:
Additionally, although not directly measured, Cook et al. estimated that due to the slightly higher speeds, emissions of carbon dioxide in the wider New York City area have most likely fallen.
Intuition for Congestion Pricing
To see the intuition as to why congestion pricing as a policy improves outcomes, I saw the following example:2
Assume you are going out at night. You can be on the street or go to a dance club. The value from being on the street is zero. Suppose you personally value dancing at $50. Entry into the dance club is free. Thus, people will keep entering the dance club, until it is over-crowded and no one can dance, resulting in $0 value to everyone inside the dance club. Everyone is indifferent between being in the dance club or outside on the street, as the value of either choice is zero.
Now suppose the dance club charges a cover charge/entry fee of $20. Certain people will not value going to the dance club at $20 and thus choose to stay outside, receiving $0 of value. The people who do go into the dance club, value dancing at least at $20. Since there are now fewer people, everyone in the club can dance, meaning there are at least some people in the club who have positive surplus (they value dancing at more than $20 like you at $50).
It is worth noting that in the latter scenario, no one is worse off and someone is better off (this is called in economics a “Pareto Improvement”). The people outside the club have $0 value, but they also had $0 value when the club was free, while some people inside the club now have positive value.
Lastly, we have not even considered the revenue from the cover charge/entry fee. The amount of money raised from the cover charge is irrelevant in terms of the Pareto improving outcome. The money raised is just a further benefit, as the club may have more money for improvements for example.
Congestion pricing acts similarly to a cover charge, which is why it is generally seen as a “Pareto” improving policy (i.e. no one is worse off but at least one person is better off).
Interesting Reads from the Week
April US Labor Market Update:
and give us the US labor market update – here and here. The labor market appears to finally be in balance right now. But any downward trajectory from here will be worrisome.- goes over what economists mean when they discuss ‘rationality’. And it is quoting the book we recently recommended! Deeponomics has several interesting articles on economics and finance – also from an academic angle.
- discusses some of his work from home research and how work from home increases mobility, which translates to better outcomes:
“When people know that they must only go to work 3 days a week, many will choose to live further from where they work. This is a type of “Rebound Effect”. As people search for housing and the right community using a larger commuting radius, this increases their feasible choice set. Let me explain. Suppose that you can commute at 30 miles per hour and you have to go to work 5 days a week. A person will spend 10 hours a week commuting if he lives within 30 miles of where he works.
If this person must only go to work 3 times a week, then the person can live more than 40 miles from work and commute less than 10 hours a week. This increases in the commuting radius opens all sorts of new opportunities in terms of being close to activities for one’s children, better schools, and recreation opportunities.”
We have often discussed why WFH appears to be broadly beneficial and (if) any negative impacts can be mitigated with different management approaches.
Because of the geography of New York City, I've always considered the congestion pricing area as a final destination for people rather than a through-road. In London, for example, the congestion pricing area can be used to traverse across the city. In New York, my hunch is that few people willingly travel through Manhattan.
Unfortunately I was unable to find the original social media post of this example; If you have seen it, please let me know and I will add the attribution.