Executive Summary

In 2019, congestion in Manhattan’s Central Business District (CBD) reached its worst levels on record, with average daytime traffic speeds falling to 7 mph from 9.1 mph in 2010. This is no surprise. Overuse of Manhattan’s roads is an entirely predictable outcome: a “tragedy of the commons”—of treating the roads as an unpriced common-pool resource.[i]

The situation should improve with the implementation of tolls for Manhattan’s CBD that were authorized in New York State’s 2019 budget. The tolls are intended not only to reduce traffic but also to raise enough revenue to support $15 billion in bonds for the Metropolitan Transit Authority’s 2020–24 capital budget, with “any additional revenues . . . available for any successor programs.”[ii]

Though the tolls have been authorized by the state legislature, the precise details and structure of the policy have not yet been formulated. The city and state, therefore, have a unique opportunity to devise an efficient congestion-pricing system using dynamic tolling. This system would raise revenue by narrowly tailoring variable tolls to actual congestion levels, using the “invisible hand” of variable price signals to achieve consistent travel speeds in the CBD.”

The Regional Plan Association (RPA), a prominent NYC-area urban-policy think tank, recently issued a report with recommendations for congestion pricing in New York City. In this paper, I will build on its analysis and propose an alternative policy that more narrowly tailors tolls in order to achieve target traffic speeds in the CBD. While tolling of any sort is an improvement from the unpriced status quo, the more responsive the tolling method is to traffic, the more likely prices are to match supply and demand for road space at any given time. Otherwise, prices will be unnecessarily high when traffic is light, or too low to prevent gridlock when it is heavy.

The goal of NYC’s congestion-pricing program should be to reduce traffic congestion with narrowly tailored tolls that nonetheless raise enough revenue to satisfy the authorizing legislation’s requirement. While policymakers will ultimately have to decide on the minimum target speeds and the requisite congestion charges necessary to achieve them, this report focuses on the impact of two different tolling scenarios: one in which the maximum peak cordon toll is capped at our best estimate of the “economically optimal dynamic toll”[iii] and another in which it is capped at a lower but perhaps more politically feasible price.

With a projected target speed of 10 mph in the CBD, an optimal dynamic toll would briefly go up to a peak of $26 in each direction during rush hours. Mid-day tolls between the morning and evening rushes would average $1.75 each way or $3.50 round-trip, while overnight congestion tolls should be at or near zero. A peak toll of $26 in each direction would mean a maximum round-trip toll of $52 but most drivers would pay far less: the weighted average round-trip toll would be about $18. Drawing on the experience with high-occupancy toll (HOT) lanes elsewhere in the U.S., the city should implement real-time toll discounts such that drivers pay only the toll necessary to achieve the projected target policy speed, up to the maximum cap of $26.

This report also considers a peak toll that is considerably lower than the economic optimum and consistent with a slower target speed: a peak toll of $15 in each direction (which yields a weighted average round-trip toll of about $4) and a rush-hour speed target of 8mph—illustrating the model’s best estimate of the policy tradeoffs of a lower but perhaps more politically feasible peak toll. There is a prudential case to be made for careful, incremental change with respect to tolling. The tolling framework presented here relies on flexible, automatically responsive pricing to accommodate unanticipated reductions in congestion—including prolonged reductions like the one that the city is now experiencing due to the Covid-19 pandemic.

It is also true that the costs of current road policy are extremely high. These costs include travel time losses (analyzed in this paper) as well as costs that are outside this paper’s scope, such as lengthy emergency vehicle response times or unpriced vehicle exhaust emissions in the nation’s most densely populated area. While a policy response—and tolling—commensurate with the scale of New York’s congestion problem may be advisable and achievable in the long run, it would be unwise to overreach on such a solution before a broad public consensus is achieved.[iv]  

Regardless of the toll levels chosen, dynamic tolls on each crossing into Manhattan should float independently. For example, if traffic is moving fast on the Queensboro but slow on the Queens-Midtown Tunnel, the tolls on each crossing should temporarily readjust to maintain optimal traffic volumes on each crossing. Because the city already tracks the location of for-hire vehicles (FHVs) such as taxis and Ubers, it should adopt a more efficient per-mile or per-minute congestion fee for FHVs, in lieu of subjecting them to the cordon toll. Such a fee would also obviate less-efficient restrictive regulations on FHVs, such as the onerous cap on the number of FHV drivers.

Dynamic tolling—adjustment of prices in three- or six-minute intervals to achieve a particular target speed—will charge the minimum necessary toll to achieve the target speed. Rather than using estimates based on historical averages, dynamic pricing will automatically begin relieving tolls the moment that traffic volumes begin to abate for any reason. Transit improvements, recessions, holidays, gas price shocks—anything that causes traffic volumes & congestion to decrease for any reason, for a given hour on a given day or over a long period of time, will automatically be reflected in lower tolls.

Furthermore, as long as the toll is permitted to float high enough during rush hour, this approach will raise more revenue than the law’s approximately $1 billion annual minimum target. Under the economically optimal dynamic tolling scenario with a target speed of 10mph and a weighted average round-trip toll of about $18, revenues exceed $5 billion annually in a commonly used transportation model even though overnight tolls are zero.[v] This revenue figure would be a bit lower if, as this paper recommends, there is broad crediting for upstream tolls (such as those paid by drivers crossing the George Washington bridge prior to entering the CBD). The RPA estimates that upstream toll credits cost about $120 million annually. In the lower-toll scenario with a rush-hour speed target of 8MPH and a weighted average toll of about $4.37, estimated revenues are roughly $2 billion annually before accounting for broader toll credits.

If rush-hour traffic proves to be more responsive to pricing than currently expected, the target policy speed can be increased to ensure that the overall scheme yields enough revenue. If rush-hour traffic is less responsive to pricing, a preset maximum toll cap will provide some pricing certainty during rush hour and help to avoid the legislative blowback that would be likely to result from uncapped surge pricing.

In short, the plans presented in this paper will allow the city to maintain target traffic speeds in the CBD, cap the maximum toll charged for entry into the CBD, automatically charge drivers lower or zero tolls during periods of reduced congestion, and raise additional revenue that, if used appropriately (see Appendix II), would allow the MTA to make critical improvements to transit services for New Yorkers.

This paper’s model was written for and calibrated to pre-COVID pandemic traffic volumes and speeds. As such, the annual revenue projections apply only to a medium-term return to normal. But the lower revenues consistent with any given target speed during this period of reduced economic activity are a feature, not a bug of this paper’s “congestion reduction over revenue” approach: Daytime average tolls calibrated to traffic speeds automatically fall in a deep recession and only recover when traffic volumes recover and begin to drive travel speeds below the target once again[vi].


[i] Road pricing is a special thing in public finance. It raises revenue without imposing a new net cost on society, as higher income or sales taxes would do. The total price of driving in Manhattan is already high, whether roads are priced with money or not—see Appendix to see how high the current travel time losses are in typical traffic conditions. Drivers can pay with time or with money, but wasted time is a pure loss to society, whereas tolls create transferable revenue that can be put to some other useful purpose, like transit investment. This is why business groups like the Partnership for NYC support road pricing. Residents and businesses are already losing billions of dollars’ worth of time to congestion, so converting those time losses into toll revenue is a very attractive alternative to new income or sales taxes.

[ii] The Laws of New York, Title 8, Article 44-C, Section 1704-A: Central Business District Tolling Program.

[iii] By ”optimal Pigouvian toll”, I mean the toll that is exactly equal to the marginal time cost imposed on other drivers by the average trip within the CBD, similar to the pro forma speed-volume schedule in the appendix. It is ”optimal“ in the “partial equilibrium” sense that drivers will only take trips when they value that trip in excess of the average delay cost imposed on the grid.

[iv] This paper relies on the Hayekian logic of dynamic pricing to build in some robustness to error in our expectation of the responsiveness of traffic to price—much more so than a fixed toll schedule that policymakers would scramble ex post to change in the event of an economic depression or other unexpected negative exogenous shock to traffic volumes. But intellectual humility is warranted insofar as the headline speed target is still subject to some model uncertainty.

Specifically: The speed-volume curve tells us exactly how much to charge the next car to travel a mile for any given number of hourly vehicle-miles of travel in Manhattan. For any given volume of travel, we know how much the next vehicle mile will slow down all the other cars already on the road, and can then reliably convert those minutes of aggregate travel time loss into a dollar value of delay per mile without knowing anything about human behavior in response to tolls. We know, for example, that when hourly traffic volumes hit the PM peak of a little over ~170,000 VMT in Manhattan, corresponding to 6MPH on average, then the next mile trip causes about ~$26 in slowdown to all the other cars combined. That’s why we set the ideal maximum toll cap at $26. Such a toll is “optimal” in the sense that trips valued in excess of the delay cost per mile should keep happening, while trips valued less than that should not.

But federal legislation allowing dynamic HOT lanes on federally funded Interstates requires the additional step of setting a concrete minimum policy speed target, not merely charging the abstract volume-conditional Pigouvian toll schedule. So we know that roughly $26 is the maximum correct price to charge per mile, according to the speed-volume curve and our knowledge of the average value of travel time, when traffic is at its average hourly worst in the pre-Pandemic baseline during the PM rush hour. But we must then translate the known $26 per-mile Pigouvian toll from the speed-volume curve into an estimated speed target achieved through a cordon toll. To do this the BTA model has to commit to an estimate of travel time and price elasticities and average travel distance per cordon entry by different vehicle and trip types—and the BTA happens to expect the human response to the $26 Pigouvian peak rush hour toll to coincide with roughly a 10MPH speed target. So in sum: We can be confident from the baseline speed-volume curve that $26 is the maximum necessary per-mile toll, but whether that optimum coincides with the 10MPH rush hour speed target in equilibrium relies upon the empirical accuracy of the BTA model. Less-responsive traffic would result in the $26 peak toll lasting longer throughout the day than currently projected. More responsive traffic could yield a higher optimal speed target than currently projected to result from the $26 optimal cap.

[v] Charles Komanoff, “Balanced Transportation Analyzer.” This model is used by RPA in its analysis of scheduled variable tolls.

[vi] This paper’s recommendations for the use of revenue in excess of the $1 billion annual target are similarly robust to surprise shocks: By using any excess revenue to restructure and retire existing debt and thereby indirectly relieve the operating budget, no particular capital project is at risk. Prepayment of outstanding debt, even if “lumpy” within the next few years of economic recovery, will provide operating budget relief in the near term and eventually stable borrowing capacity for next MTA capital plan.