I’ve been writing pieces of this post for more than a year, but parts go back to presentations I’ve given several years before. As I’ve written it it’s gotten longer and longer, so I broke it down into several sections and case studies in each. It’s mostly about road pricing but also touches on bottlenecks and congestion. It’s not meant to be exhaustive, but it may well be exhausting to read. Enjoy?

In 1996, Massachusetts Governor Bill Weld was running for Senate against John Kerry. In a last-minute ploy to gain votes, he suspended tolls on part of the Turnpike, making a show of taking out the tolling (seen here yukking it up with Big Dig podcast villain Jim Kerasiotes). In 2014, when the tolls made a valiant return, I wrote about how, for nearly two decades, they had incentivized people to take a slightly longer (by time, not distance) trip to avoid the tolls, adding traffic congestion to local streets. Jim and Bill didn’t exactly do the math: while people will generally pay tolls for a fast trip on a highway, if the tolls are easily avoidable, they’ll choose to avoid them.

Believe it or not, time has a value. 25¢ to save two minutes in 1997 is $7.50 per hour (about $16 today). Make it $1 for those same two minutes is $30 per hour ($64 today). People are actually rational individuals and, if given a substitute good at a reasonable price, will make a different decision. Yet it seems like many toll policymakers, whether in the public sector or private, have never heard of this newfangled field of “economics,” and make the same mistakes again and again.

So here is a 13,000 word blog post with more than you ever wanted to know about toll road policy. With a sprinkling of economics. (Disclaimer: I am not an economist, but I do know some basic math.)

I. Where tolls do and don’t work
—a. Where tolls do work
—b. Where tolls don’t work
II. One-way tolling
—a. Case Study: the Delaware
—b. Case Study: the Dumbarton Bridge
III. Parallel routes: substitute goods
—a. Case Study: Chicago Skyway
—b. Case Study: Indiana Toll Road
—c. Case Study: Louisville
—d. Case Study: Oregon
—e. Case Study: Vancouver
—f. Case Study: Bay City, Michigan
IV. Volume discounts, in-state discounts, marginal rates and the commerce clause
—a. Case Study: Louisville (again)
—b. Case study: Maine
—c. Case study: Massachusetts in-state discounts
V. AET, ORT and, uh … why is Maine still building toll booths?
—a. Case Study: New Hampshire
—b. Case Study: Sturbridge
—c. Case Study: Pennsylvania
—d. Case Study: Maine (again)
VI. So is tolling congestion pricing?
—a. Case Study: Congestion Pricing in Boston?

I. Where tolls do and don’t work
a. Where tolls work

Tolling roadways is an effective means to collect revenue from the use of a resource. In some cases, tolls are pricing a scarce resource, especially where a geographic constraint creates a bottleneck, often with higher-cost infrastructure (in other words, bridges and tunnels). Here, high tolls can be charged for the distance traveled, because the alternative, if it exists, usually requires a significantly longer journey. (The most severe outlier here may be crossing from one peninsula of Michigan to the other: if you ask Google to show you the route from St Ignace to Mackinac City avoiding tolls, it will still send you over the $4 toll bridge, rather than the 900-mile route around Lake Michigan or Huron. $4 is worth it for a pasty.)

In other cases, tolls are used to price a highway serving a major transportation axis with no easy roadway alternative. These tolls rarely use pricing to change demand, although there are some cases where tolls are higher at peak times to move demand away from the busiest periods. For roads which predate the Interstate Highway System, the first mover effect means that these roads acquired a natural monopoly to major transportation corridors and effectively rendered any competition moot (although there are some examples of where an untolled roadway parallels a tolled one), even though the roads may not be near capacity. 

Public policy means that these roads are usually, but not always, tolled below a revenue maximizing level (often, a revenue-maximizing toll would push traffic onto local, slower alternatives, which would be politically unpopular and create additional congestion). In many cases, the tolls are required to be spent only on the roadway itself, to pay off construction bonds and then to maintain the roadway. This had led to tolls which have dropped considerably since their implementation, as bonds were paid off. Even the Pennsylvania Turnpike, which is the most expensive long-distance roadway in the country, has seen its tolls drop since it opened, when adjusted for inflation.

	Opened	Toll	Adj inf	Current	Change
NY Thruway	1960	$5.60	$64.02	$26.29*	-59%
Maine	1947	$0.50	$7.79	$2.95**	-62%
Massachusetts	1964	$2.95	$31.98	$5.95	-81%
Pennsylvania	1940	$1.50	$36.15	$21.98	-31%

All tolls for in-state EZPass users, cash significantly higher (although still generally lower than historic tolls, although not Pennsylvania).
* Average of eastbound and westbound including one-way Tappan Zee toll.
** As little as $1.77 with 40+ transactions per month; original 1947 segment to Portland.

Tolls work best when they toll a section of highway with no reasonable alternative (or when the best alternative is a parallel transit line). If you’re going from Oakland to San Francisco, you can pay an $8.50 toll for a 10 mile trip, drive 80 miles and lose an hour of your life (or often quite a bit more) and pay for 70 miles of additional fuel, or take BART. When San Francisco went from a flat toll to a time-variable toll, some drivers shifted their trips to pay the lower rate (which should not surprise us, but given what we will see later, probably surprised some toll planners), and some probably shifted to transit (although the paper noted did not calculate that difference). 

Longer distance toll roads also make sense, as long as there aren’t parallel alternatives. The history of toll roads in the United States, aside from the 1800s turnpikes (which were quickly superseded by railroads) dates to the 1940s and 1950s, when, before significant federal investment, states would float bonds to build highways across the state. The first long-distance toll roads opened just before the war (the Merritt Parkway in 1938 and the Pennsylvania Turnpike in 1940) and were the models for states to build roads along their main demand corridors before there was national coordination (and sometimes no coordination; Kansas’s Turnpike ended in a field in Oklahoma for a time). The Federal-Aid Highway Act of 1956, which established 90% funding for the Interstate Highway system, all but ended toll road construction for a generation. Toll roads already built were grandfathered into the Interstate system (and states can earn credits for roads which provide a public good but don’t use with federal funding).

Suburban toll roads can be successful (at least in the sense of paying back their construction costs), although they may not start out as profit centers. SH130 outside of Austin was the butt of many jokes and the southern, privately-developed portion did underperform in early years and go through a bankruptcy, but increasing congestion on the free alternative and development along the tollway has led to an increase in traffic (albeit at a relatively high toll rate of about $20 for 90 miles). E470 in Colorado was lightly utilized until a new airport opened, and that, plus development, has made for a profitable roadway. There have been several toll roads in Southern California which have filled perceived gaps in the transportation network, or created sprawl. Or both. (And this is just the United States; many other countries have tolled roads as well.)

Over a long distance, having a limited access highway can save travelers significant time and make a toll worthwhile, as long as there is no free alternative. Chicago to Pittsburgh adds two hours avoiding tolls. From suburban Boston to Albany, where the Turnpike is the only reasonably-fast roadway through the topography of Western Mass, the $4.25 toll (plus a bit in New York) is a small price to pay for a trip nearly twice as fast.

Pennsylvania is an interesting case in the middle. From Pittsburgh to Philadelphia, avoiding the tolls adds more than an hour to a five-hour trip, and most drivers (although certainly not all) will pay the high toll to avoid the longer route. For the first 30 years of its operation, the Turnpike captured much of the east-west traffic in the state, although when the free I-80 opened, it siphoned traffic off not only by providing a faster route, but one without tolls.

 I-80 was originally planned to be built as a toll road, but once the interstate highway act was passed, the lure of “free” money from the federal government was too high, and the road was built using federal funds. Act 44 of 2007 required the Turnpike to make payments to fund other roadway and transit projects, anticipating participation in a federal pilot program to allow tolling of interstate highways. Tolling I-80 would have spread the toll burden across both east-west roadways (which have similar traffic volumes) and kept toll rates relatively low, while eliminating the free “shortcut” across the northern part of the state. 

Because of lobbying from certain parts of the state, the Federal Highway Administration disallowed the tolls, but the state had already planned for this revenue. This forced the Turnpike to take out bonds to pay the Act 44 funds (since reduced) and the toll for the Turnpike has become one of the highest in the world, a $62 for the full roadway, or close to 20 cents per mile, making it by far the highest-priced legacy, long-distance toll road in the country, and several times more than most other roadways which generally fall in the 3 to 7 cents per mile range. Even between Pittsburgh and Philadelphia, an untolled alternative only adds a bit more than an hour, so the state probably loses revenue to people bypassing the tolls, who in turn pay with their own time, leading to a loss of utility for everyone.

Select long-distance toll roads in the United States:

	Distance	Low toll	High toll	Low mile	High mile
NH (3/93)	44	$1.40	$4	$0.03	$0.09
NH (16)	33	$1.06	$3	$0.03	$0.09
Illinois (88)	140	$5.10	$10.20	$0.04	$0.07
Maine	102	$4.02	$8	$0.04	$0.08
Massachusetts	134	$5.95	$11.50	$0.04	$0.09
NY Thruway	495	$22.66	$39.66	$0.05	$0.08
Illionis (90)	77	$3.95	$7.90	$0.05	$0.10
Illinois (Tri-State)	84	$4.40	$8.80	$0.05	$0.10
Kansas	200	$11.66	$23.32	$0.06	$0.12
Okla (Will Rogers)	88	$5.40	$10.50	$0.06	$0.12
Okla (Turner)	86	$5.40	$10.50	$0.06	$0.12
Ohio	239	$16	$23.50	$0.07	$0.10
Okla (Indian)	105	$7.36	$16.74	$0.07	$0.16
Garden State	172	$12.38	$13.20	$0.07	$0.08
NH (95)	17	$1.40	$4	$0.08	$0.24
Atlantic City	44	$4.22	$12.98	$0.10	$0.30
West Virginia*	88	$8.81	$13.50	$0.10	$0.15
Indiana	156	$16.90	$16.90	$0.11	$0.11
New Jersey	117	$16.41	$22	$0.14	$0.19
Penna NE Ext	111	$19.44	$38.88	$0.18	$0.35
Penna Tpke	355	$62.54	$119.72	$0.18	$0.34

* West Virginia has an annual pass available for just $27, which covers use of the toll road for the entire year, its payback is just four trips across the roadway. The West Virginia EZPass is shown for the low rate here, regular road users, regardless of their state, can essentially use the road for free.

b. Where tolls don’t work

Where don’t tolls work?

Where there’s an easily-available substitute good.

My wife’s family lives near Philadelphia, and we too frequently wind up driving (thanks to $300 Amtrak fares) and pay our fair share of tolls along the way (enough that I’ve considered renting a car and paying the “all you can eat” toll transponder daily fee, since I would save quite a bit for a short trip). If traffic allows us to cross the George Washington Bridge going south, the fastest route is usually via the Henry Hudson Parkway, which has a $7 toll for out-of-region EZPass. However, there’s a route via Inwood or via the Major Deegan and Cross Bronx which costs us about 7 minutes, but saves $7. And, yes, I usually take it, increasing my travel time and externalities on city streets because my time is worth less than $60 per hour when driving on a trip for leisure. (And, yes, I should add a NY EZPass to my collection since the toll is less than half; more on that tomfoolery below.) Everyone loses: the MTA loses out on revenue, the City loses because of additional congestion, and I lose my time.

And yet, there are too many toll agencies which seem not to understand this—many of which are new agencies—and wind up with a variety of toll rates and policies which make no economic sense.

II. One-way tolling

Tolls work well on bridges because they are often natural bottlenecks without effective substitute goods. Many bridges like these have one-way tolling, which mostly dates back to when tolls required manual collection and collecting tolls in both directions would significantly increase the cost of collection, with the assumption that the substitute good was so costly that few people would balk at paying double the rate even if it made avoiding the toll more attractive. As long as these tolls are coordinated and there is a major barrier, this works well. For example, there are no untolled options to cross the Hudson River south of Albany, and as you go south, the tolls progressively get higher, but not to the point where there is an incentive to go out of your way to save money except for some edge cases near Albany itself (notably, it’s cheaper to take the untolled I-90 through the city than the New York Thruway around it, but this toll is collected in both directions and is relatively low).

If a tolled roadway only provides a modest time savings, a modest toll can be effective for most users (infrequent users, and those using rental cars, may still wish to avoid it because of additional fees). But implement a toll too close to a substitute good, and the calculus can change.

a. Case Study: the Delaware

One state west, the crossings of the Delaware River are tolled entering Pennsylvania (or Delaware) and free in the other direction, and traffic is balanced between the two directions, since there are no substitute good routes. 

Until Trenton. At this point, the Delaware River Joint Toll Bridge Commission operates “toll-supported” bridges, which are free bridges, generally with weight restrictions, and are supported by toll revenue elsewhere. The first of these, the “Trenton Makes” bridge, is roughly 500 feet upstream from the nearby one-way-toll bridge. Southbound traffic across this bridge (avoiding the toll): 11,907 vehicles per day. Northbound? 2251. 

There is a similar setup in New Hope a few miles upstream, with data for both bridges. The toll bridge has 4200 daily northbound users and 3200 daily southbound users. But we needn’t look far for the balance: the free bridge (I’m sorry, “toll-supported bridge”) in town has 5700 southbound users and 4500 northbound. More than 1000 people per day will go out of their way by 4 minutes to save $2, which makes sense; it’s a rate of $30 per hour. Simply making the toll bidirectional would make the 4 minutes of travel worth just $1, and assuming a linear time value of money, may increase toll revenues by $500 per day, or close to $200,000 per year. (The Bridge Authority could make more money in Trenton as well by taking in more toll revenue from the world with a two-way toll as well.)

b. Case Study: the Dumbarton Bridge

Another example is the Dumbarton Bridge, the southernmost crossing of San Francisco Bay. The Bay Bridge and San Mateo Bridge have one-way tolls, but the only substitute goods for using those bridges are BART and ¯\_(ツ)_/¯, respectively; if you’re driving, you pay the toll rather than going far out of your way. The Dumbarton Bridge is different. Going around the Bay to avoid the toll adds 18 minutes, nearly $30 per hour plus the additional cost of about 15 miles of driving. 

But there are more marginal trips where crossing the bridge might be slightly faster, and the toll is especially punitive. A trip from Newark to the Googleplex is 8 minutes faster via the bridge. A Google engineer is probably willing to pay $64 per hour to save this time, but a Googleplex cook or janitor may instead pay with their time. (Different users have different elasticities.) Again, this is a location where changing the now-electronic toll to a two-way toll would better balance these trips, and reduce the incentive to avoid the toll.

And the data confirm this (at least the data from 2011). 

	Eastbound	Westbound
Bay Bridge	123,000	125,000
San Mateo Bridge	51,000	46,000
Dumbarton Bridge	49,000	31,000

The Bay Bridge, which has basically no substitute good if you wish to drive across the bay, has as many toll-paying trips as untolled. The San Mateo Bridge, where a toll-free substitute is a relatively long drive, there are slightly fewer tolled journeys (53-47). But the Dumbarton Bridge has a 61-39 traffic split between eastbound and westbound traffic! There are something like 18,000 people who take the Dumbarton Bridge eastbound and then find another way in the other direction every day. If a bidirection toll brought half these people back to the bridge, adjusting for lower use on weekends, this would increase toll revenue by $12 million per year! Remember that this doesn’t cost people $12 million more per year, but the additional tollpayers would be saving at least $12 million worth of their time, and paying to take a faster route. 

There may be some edge case spillover from the San Mateo bridge, and it might make sense to make it, and the Bay Bridge, bidirectional tolls as well. It’s rare that I praise MassDOT, but in toll policy, I will do so multiple times. When tolls in Massachusetts were made electronic, the previously one-way bridge-and-tunnel tolls were recalibrated to be bidirectional (unfortunately I have not found any granular-enough data to see if this changed driving behavior). The Bay Area is leaving more than $10 million on the table by sticking to one-way tolls. And, no, people aren’t saving money. They’re paying with their time, and causing additional congestion and pollution from additional miles driven, so the overall cost to the community is even higher.

III. Parallel routes: substitute goods

The Delaware River may have a few untolled crossings parallel to tolled ones, but these at least are not the main crossings, and are weight-restricted and open only to cars. While this leads to incentives where people will use a longer, slower route in one direction to save a toll, there are both historical and political reasons where this makes some sense. But the data are clear: give people a substitute good, and they will use it. For more recent implementations of tolling, was this lesson learned? Rarely.

a. Case Study: Chicago Skyway

The Chicago Skyway is a case where the lesson was learned, eventually. The Skyway was originally built by the City of Chicago (which had the authority to build toll bridges, but not toll roads, so it is officially a bridge with long approach ramps) and never met traffic projects, because it is roughly parallel to an untolled route. The untolled route via I-94 is 6 miles longer and generally 5 to 15 minutes slower depending on traffic, but combined with the Indiana Tollway toll, it costs $15.90 to travel ($6.94 in Indiana, $8.10 in Illinois). Most drivers are willing to save $16 even if it means a somewhat longer trip.

The Skyway underperformed expectations from day one, and the City lost money on the roadway, initial tolls were just 25¢ and were $2 by 2000. In 2004, they found a solution: they sold the road. A multinational consortium bought the roadway and did the only thing they could figure out: they jacked up the price, today it costs $8 for the same trip (even adjusting for inflation, this is almost three times as much). Instead of trying to serve the greatest number of users, they are trying to make money. And they’re failing. I-94 averages nearly 200,000 vehicles per day, while the Skyway sees just 36,000, carrying just 1/6th as much traffic and making it the least busy highway in Chicago by far. In a sense, the private owner has created a bypass of traffic for some commuters, but the substitute good is close enough in geography and time that there is not any price where they can make a profit on the roadway, even if they are charging a revenue-maximizing price. Unlike Chicago’s parking meters, the City got a good up-front deal on the asset, and the owner is stuck with the underperforming roadway.

b. Case Study: Indiana Toll Road

The Indiana Toll Road, which the Skyway flows into, was never a distressed asset like the Skyway. It is the only conceivable route between Ohio and Chicago, and with I-80 and I-90 interlined across the state, the main truck route, with nearly one third of the traffic made up of heavy trucking (which pays a higher toll, and does more damage to the roadway). Cintra-Macquarie (same firm as the Skyway) bought out the toll road in 2005 for $3.8 billion (equivalent to $6 billion today) and, even managing an asset with significantly less competition than the Skyway (although a trip from Chicago to Detroit, while approximately equidistant via the Toll Road or I-94, will almost always use I-94 to avoid the tolls), overpaid and went belly-up. A new consortium has replaced the original one.

Tolls on the road in 1958 ranged from $1.95 for passenger cars ($22.84 adjusted for inflation) to $13.75 ($161) for six-axle trucks. The same tolls today range from $17 to $107, so white they have decreased, they are relatively high compared to other toll roads, and the concession operator will continue to raise rates, although these increases are limited by statute. While the concession has lost money, it does show that tolling a roadway without a direct substitute good is a means to raise capital funding for other projects, in Indiana’s case, these have been mostly used to repair and build roads in other parts of the state. The private sector tends to overpay.

c. Case Study: Louisville

Louisville sits on the south bank of the Ohio River, a wide-enough river that there are only limited crossings. Within 8 miles of the city, there are four roadway bridges which cross the river, two downtown and two more on a circumferential highway further afield (one of which opened within the past decade). The next nearest crossings are more than 30 miles away from the city. The only substitute goods are a bus every 30 to 60 minutes or a bike path across the Big Four Bridge. (Until 2013, the only bike/ped crossing was the sidewalks of the Clark Bridge; the Big Four had been the “Bridge that goes Nowhere” since it was taken out of rail service in 1968, and had no ramps for 40 years, hence the name.) The new highway bridge has a bike/ped facility; given its location in the suburbs, I doubt it is heavily used.

Here are the four bridges, from west to east:

• Sherman Minton (I-64, 1964, 6 lanes), 5 miles west of Downtown, mostly serving points west of Louisville (untolled), closed for several months in 2011 because of structural deficiencies.
• George Rogers Clark Bridge (1929, 4 lanes), in Downtown, serving local traffic, parallel to the …
• Kennedy (1969) and Lincoln (2015) bridges (I-65, 12 lanes), half a mile upstream from the Clark Bridge. Tolled since 2016. Serves north-south traffic into the city from I-65 and from northern suburban areas.
• Lewis and Clark Bridge (I-265, 2016, 4 lanes), 7 miles east of Downtown. Tolled.

So between 2015 and 2017, the region went from having 16 lanes across the river to 26. And, believe it or not, traffic disappeared. Some have hailed it as an example of tolling and congestion pricing, but, uh, it’s a better example of what happens when you double the capacity of river crossings. (“Just ten more lanes, bro!”) Notably, traffic on the bridge that was doubled in size has fallen by 50%. But overall traffic hasn’t decreased. Instead, it’s shifted to the free or more convenient alternatives, one of which, the older Clark Bridge, is all but adjacent to the newly-tolled bridge. (The new toll bridge further east of the city doesn’t really have a substitute good, since it reduces the distances traveled for its users by enough that they are unlikely to seek a free alternative.)

The toll to cross the bridge is $2.68, collected in both directions (half that rate if you cross 40 times per month; we’ll return to this later). Taking the parallel free bridge adds between 4 and 8 minutes to travel times, giving the toll a value of time of $10 to $40 per hour. 

Traffic on the bridge declined from 140,000 vehicles per day to 70,000. Some pay a toll to use the new East End Bridge (the current bridge was the substitute good for this bridge before the East End Bridge was built, since the next nearest bridge would be a 90 minute detour upstream), most of the rest take one of the free bridges, either the parallel one, or, if their trip was roughly equidistant, a slightly longer route to cross the bridge downstream. It’s likely that many of the users of the toll bridge are those who use it frequently enough to get the discount.

I am going to cite this Streetsblog/CityLab article which misinterprets this as successful “congestion pricing” when instead it is rational economic actors seeking an optimal solution. (Fun fact, an early version of this post, before it exploded into way too many case studies, was just going to dunk on this article.) It’s laughably simplistic (i.e. just plain wrong). They looked at data from just the tolled road without looking at the whole system which paints a completely different picture. As we can see, many toll planners seem to have the same grasp of basic economics. They assumed that all of the change came from high elasticity for trips, without the actual conclusion that while trips are not elastic, but the willingness to pay is. You know, since the trips didn’t actually go anywhere.

At least they wrote about it so we can pick on them.

If asked to pay for even a fraction of the cost of providing a road, half of all road users say, “No thanks, I’ll go somewhere else” or not take the trip.

The fact that Louisville residents would rather drive miles out of their way or sit in traffic for an extra 10 or 15 minutes to travel on a “free” road, rather than spend a dollar or two for a faster, more direct trip tells you the very low value that highway users attach to these extremely expensive roadways.

Yet this is not the case. People will still make the trip. They’ll just make it differently. Congestion pricing will reduce some trips in a city like New York, where the toll is high, unavoidable and alternatives are plentiful. In Louisville, people will just take another bridge.

Few if any trips take 10 or 15 additional minutes, most are either faster or just slightly slower. These drivers are either going elsewhere and paying a toll on the new bridge, or taking one of the free options. Or they are paying with their time for a slightly longer trip. Very few are actually not making the trip, or using a different mode. Users are generally making rational decisions based on time and costs, they are not reducing their trips entirely. The value is not “very low” but “rational.” Anyone who was making an interstate bridge by driving their $20,000 vehicle several miles each way has a purpose for making the trip and probably values the trip at more than the $2 toll. They just don’t necessarily value their time at $30 to $50 per hour. Economics makes sense.

Data show that traffic on the bridges as a whole dropped by 2% from 2013 to 2018, and has been mostly stable since then, going from 224,000 to 218,000 in ten years, a 2.7% total decline. In Clark County in Indiana, VMT has increased by 2% since 2015, which would suggest a 4.7% decrease on the bridge relative to prevailing regional conditions, but this is mostly due to a 44% increase in commercial VMT; non-commercial VMT has decreased by 3.6%. So relatively speaking, bridge traffic is somewhere between down 4.7% relative to local conditions to up 1.6%; averaging the two to 1.5% may be a reasonable estimate given that most of the new commercial traffic is likely local deliveries which are not using the interstate bridges.

This shouldn’t have been a surprise. When tolls were implemented in Norfolk, Virginia in 2013, a significant chunk of traffic moved to parallel bridges. The total number of vehicles crossing didn’t change, but where they were crossing did. This case study was also not in a location where additional road space was added, nor where there was a directly parallel free bridge. 

I’m not about to run an economic analysis of how the changes in traffic compare to before the bridges, tolls, new bridges which may have increased trip demand and other factors, but will go out on a limb and say that tolling did not reduce trip demand by 50%; Of the 140,000 people using the trip, about 30,000 were given a faster, if tolled, new option, and most of the rest took the parallel, untolled option. Maybe a few dozen took the bus.

These drivers are making a very rational decision because they value their time at $40 per hour which, under most conditions, means that it is “cheaper” to take the free alternative even if it takes longer since the value of time for transportation is only about $20 per hour. The real question is who the drivers left on the tolled bridge actually are. Some probably make the trip frequently enough to pay the lower toll, others have higher values of their time, still others are infrequent users who would pay a higher rate (and for whom the detour is even more lucrative). This doesn’t however, indicate that Louisvillians are somehow irrational actors driving far out of their way, but instead rational actors attempting to avoid paying a toll valued more than their time.

The two states spent $1 billion doubling the size of I-65, only to have half as many people use the bridge. That money was wasted. Nothing more clearly illustrates the utter folly of highway expansions.

But the same number of people cross the river, which is the outcome of having bridges. They’re just making a calculation of how much their time is worth given the options available. If the toll was spread across all four bridges (in which case it could be cut in half since it would be paid by twice as many users) I very much doubt we’d see a 50% decrease in traffic. It would more likely be close to zero, unless the toll were much higher. 

I do agree that this money was likely wasted given the oversized bridges, but it doesn’t have to do entirely with highway expansions. The East End Crossing made the difference, shifting a lot of this traffic. Having the parallel Clark bridge, even one that might become congested, took care of most of the rest. This does not make the argument for congestion pricing, but does show that, given a reasonably-priced alternative, people will adapt to it. And they have, more so than pre-project expectations. Traffic is very slightly lower overall, but it’s hard to say whether that is because of the tolling reducing trips, or overall changes in travel behavior. The East End Bridge would have been enough, without spending billions on a new bridge to add many more lanes.

In any case, it’s a clear case of toll policy setters refusing to talk to an economist.

(The answer of “why can’t they just toll every bridge” is fair, and probably comes down to federal funding issues: the tolls could only be used for bridges which were built using the toll money, even if the bridges otherwise act as a system and are almost perfect substitute goods for each other. No, this is not a rational policy. And while the tolls may be the best deal that the state can get to fund the bridges, it’s not congestion pricing if it doubles the number of lanes and creates congestion on an older, narrower bridge.)

d. Case Study: Oregon

The bridges across the river into Oregon from Vancouver, Washington are somewhat similar to the Ohio crossings near Louisville. Oregon may try to take the same approach as Louisville, although it may be a bit more compatible with creating congestion pricing. Like Louisville, Portland is bordered to the north by the Columbia River which is almost the exact same size (by average annual discharge) as the Ohio (the Columbia is a bit wider). Here there are really only two bridges, the Interstate Bridge (I-5) and the newer (1977)  I-205 Bridge further east. Neither is tolled.

The Interstate Bridge is old. The older span dates to 1917, and it was paralleled in 1958. Each is three lanes wide, and it is the only mainline interstate bridge with a movable span for ships to pass. The bridge was tolled in the 1960s to pay off expenses from this expansion. Given the age of the bridge, it needs to be replaced, and plans are underway to do so, with a price tag reaching well above $10 billion, which will include four lanes in each direction and light rail tracks. The span, which predates the Interstate system, will be tolled starting in 2028 in order to raise funds for this. 

The bridge carries about 140,000 vehicles per day, so each $1 toll rate will raise about $50 million in annual revenue. Even with a $5 toll each way, this is far less than the bridge will cost to replace. And a toll set too high will drive more people to use the free alternative on I-205, a few miles east, although it may discourage some traffic from coming through downtown Portland. That said, a high-enough toll may move enough traffic off of I-5 that the Rose Quarter high project could be precluded or at least slimmed down, especially if better transit were eventually offered across the bridge. While less egregious compared to the Louisville example, it still seems like a missed opportunity to toll both crossings in a way which raises enough revenue and reduces traffic enough to act as congestion pricing. It will raise a significant portion of the revenue needed for the new bridge, but the availability of a parallel bridge means that the revenue maximizing toll is relatively low.

But people think that tolls will eliminate a third of traffic. Let’s just say that this is … unlikely. (High enough tolls may shift a lot of this traffic to a different route, though.) People will continue making the trip. A few may shift to buses. Most will probably shift to the other bridge. For someone making trip from Vancouver to Portland (or the airport) a relatively low toll will not cause them to not make the trip at all. Most will either pay the toll ($5 is a small portion of, for example, the cost to park at the airport and fly to another city, and there is no airport nearly in Washington, and only a small portion of the average hourly wage someone may make by commuting across the river) or pay with their time on 205. If the toll is set to $20, a Chicago Skyway situation will develop: most people will opt for the untolled parallel option, even if it’s longer.

e. Case Study: Vancouver

Let’s journey north of the border to find out another example. In Vancouver (British Columbia, this time), the crossings of the Fraser River were mostly free until around 2010, when two bridges (one new, one rebuilt) with tolls were put into operation. What happened? Traffic on the tolled bridges did not meet expectations, while traffic on parallel bridges increased. When tolls were removed, the traffic returned to equilibrium. Traffic on the newly-untolled bridges increased, while traffic on parallel bridges declined. Far from the Streetsblog argument that people do not place value on their trips, people are willing to pay for their trip if it makes the trip significantly faster. But when it doesn’t, they’ll find a free alternative. And if tolls seem arbitrary and punitive, eventually people will be willing to pay higher overall taxes to remove them altogether.

(Really, someone with the time and training could take all of these examples and run a regression to find out how much drivers value their time. You’d think the toll concessionaires would hire someone with an economics degree to do this before blowing billions of dollars on toll roads with substitutes.)

f. Case Study: Bay City, Michigan

Perhaps the most ridiculous toll scheme is the privatization of two of the bridges in Bay City, Michigan. Never heard of Bay City? Hold up your hand. Point to the “first dorsal web space” between your thumb and index finger. Now you know where Bay City is. As to why it has toll bridges, well, that’s a longer answer where I can’t just point to my hand.

Bay City is spread across the two banks of the Saginaw River where it empties into Saginaw Bay. In a three mile span of the river, there are four bridges. The two northern bridges are owned by the city, the two southern ones are owned by the state. The city-owned bridges needed tens of millions of dollars of repair, and without the funds to repair the bridges, the city leased them out for 75 years with the stipulation that United Bridge Partners to rebuild them and then levy a toll to cover the upfront costs.

I have many questions about United Bridge Partners, which is part of a private equity fund investing in transportation projects which seems to have looked at Cintra Macquarie and said “well, that’s an intriguing business model, what if we did it worse?” They have a portfolio of bridges where they either buy a distressed asset to toll it or build a new bridge the public wasn’t going to build, but they seem to mostly buy bridges which don’t have much of a market. Their bridge ownership ranges from “sure, why not” to “I can not comprehend how this makes any sense.” 

In the first category we have 

• The Houbolt Bridge in Illinois, which connects I-80 to major rail-truck intermodal yards, and truck drivers and logistics companies may be much more willing to pay $8 to save a few minutes or miles. 
• The Dominion Bridge in Virginia, where the City got them to build a $345 million new bridge for a $1 toll on a relatively high-demand route, when alternate routes add about 8 minutes of travel time, so most people will pay the toll.

In the second category:

• The South Norfolk-Jordan Bridge in Virginia has closer replacement goods, and is underperforming with only a fraction of previous traffic (parallel to the previously-mentioned Norfolk bridges, but closer to the replacement good bridge). 
• The Memorial Bridge between Ohio and Parkersburg, West Virginia, another case where there is a small market for toll users. Parkersburg is a city where the population has declined by 50%, and the city was planning to tear down the mostly redundant bridge, until the private equity showed up. 
• The Cline Avenue Bridge in Indiana. This bridge, also IN-912, was originally a multilane highway bridge which had severe corrosion and was demolished between 2009 and 2013. The state put out a bid for a private operator, and UBP built a new, two-lane, undivided highway in the same location, at a cost of well over $150 million, as a private, tolled roadway.
• Bay City …

In the later category, it seems to be no path for them to ever recoup these costs. On the Cline Bridge, UBP says that it services over 1 million crossings per year, which sounds like a big number, with a $2.90 toll for each. That’s about 3000 per day (a lightly-used rural road) or 10% of the traffic of the old bridge (and one third of their expected traffic). The toll on the parallel Indiana toll road is less expensive, and the untolled surface options for local traffic only add a couple of minutes. The business model seems to be based on truck traffic (which creates additional maintenance issues) but the question is quite open as to whether the company has any hope to recoup its investment with a parallel roadway nearby. But still, to finance a $140 million investment, they won’t even come close to repaying the interest at 3000 vehicles per day especially given parallel alternatives.

Their lesson from Cline and Jordan to them was to … do something similar in Bay City. There, two tolled crossings are easily replaced by free crossings which add just a couple of minutes for most users, and while they are currently free for in-city residents, that discount will disappear in 2028. Plus it’s not a huge market: Bay City has about 30,000 people and the population has dropped by 50% since its peak (so the 14 lanes across the river should be plenty). 

The 30,000 people in Bay City are angry about the tolls, and are proposing to boycott the toll bridges—and they should! For most users, a different route will only add a few minutes, and it’s quite possible that a well-publicized boycott could reduce traffic on these bridges close to zero, and if so, good for them. (Maybe some people will even take to walking or biking across the tolled bridges, which is free.) At some point, basic economics has to come into play. And while there has been some additional traffic along the free bridges, there should be capacity for the community to stick it to private equity and, potentially, force their hand into selling back the asset at a loss. (I’d imagine the community putting up homemade detour signs warning everyone off the bridges.) Power to the people! Don’t let private equity get you down.

So, good for Bay City, which is just the latest example of a city government fleecing a private equity toll group which failed to hire an economist.

IV. Volume discounts, in-state discounts, marginal rates and the commerce clause

Many toll agencies have discounts for in-state users, others only discriminate between EZPass users and those requiring an invoice from a plate (some don’t; I see you, Pennsylvania, and I appreciate you). Louisville and Maine have gone one step further and each taken the same harebrained approach to frequent travelers: if you take a certain number of trips, all of your trips for the month are discounted. (Hello! Marginal rates would like a word.)

a. Case Study: Louisville (again)

Crossing the Ohio, that magic number is 40. If you take 39 trips, you pay $104.52. If you take 40, you pay $53.60. Anywhere near 40 trips and you have a strong incentive to make several loops of the bridge to cut your bill. For example, someone crossing 34 times per month ($91.12) would only have to make 3 round-trips (about 30 minutes) to save $37, or $74 per hour. We already know that a lot of people value their time at less than that. 

(How many people do this? I’m not sure, but I have freedom of information requests out to both agencies to see if I can find out.)

The idea is that 40 trips gives regular commuters a break. Which is on its own reasonably bad policy (encouraging driving), but even worse because of the implementation. Given remote work, vacations and holidays, there are probably significant numbers of people who wind up in the 30-40 trips per month range and are incentivized to push it up to 40 to get the discount. This is an economic loss for the agency and for the user, who is forced to spend time circling the bridge to reach the required number of trips. I’d imagine if I tried to sketch the deadweight loss on a surplus/loss diagram, I’d want jam the pencil into my temple instead.

The problem is that the American driver (and apparently toll policymaker) is, by and large, a moron, and has the same understanding of toll rates that Kelsey Plum has of marginal tax rates. A sensible policy would be to apply some sort of discount to additional trips beyond a certain level, say, 10% after 10 trips, 20% after 20, etc, or even 1% for each trip taken, which would decrease the marginal cost of each trip without any weird incentives. But that, apparently, is too difficult for the American driver’s mind, and certain the average toll agency’s policy group, to grasp.

b. Case study: Maine

Maine has a similar scheme: a 20% discount after 20 trips, 40% after 40 (beyond a significant discount for using an in-state transponder, which is why I now have three transponders in various faraday bags in my glove box). What’s interesting about Maine’s system is that by entering and exiting in Wells, Kennebunk and Biddeford, you can trigger four toll transactions without paying an additional penny and with only minimal additional time (a couple of minutes per transaction). For a driver trying to hit an arbitrary trip cutoff, it makes it easy to trigger. 

In Maine, the volume discount is only available for in-state accounts, so outsiders and vacationers pay the higher fees, even when they use an out-of-state EZPass. Notably, the 40 trip discount is per account, so I am working with family to get all of our Maine trips onto one account, since we may well hit 40 transactions per month (especially with some creative trips on and off at the same exit to increase our transaction count). Down the road, New Hampshire will double its tolls next year, but only on out-of-staters. The $2 Hampton Toll—where 70% of transactions are out-of-staters—will go to $4, while in-staters, or those with an in-state EZPass, will pay just $1.20. Which means, yes, I’ll be fumbling between EZPass transponders and faraday bags as I make this trip. (I should really create a device which houses multiple EZPass transponders against the windshield and shifts the faraday block between them.)

But wait! Is this legal? Is this not an infringement of the commerce clause? If I’m engaged in Interstate Commerce, should I really have my rights trampled if I don’t go out of my way to acquire multiple devices? The courts, unfortunately, say “yes,” although I would argue that their opinion is long-since outdated and the right argument could come to a different conclusion. 

There are two cases here, Doran v Mass Turnpike (and the parallel, subsequent Yerger) and Angus v MTA. The TL;DR seems to be that they fail to establish an undue burden against the toll authorities because a) transponders are available to anyone no matter their state of residence and b) there is no undue burden to “swap and stuff” transponders when crossing state laws. 

The legal test here is the “Pike test” based not on a turnpike but on Pike v Bruce Church and whether a law or policy places an undue burden relative to its local benefits. In Doran, which is the base for subsequent cases, it was found that the “swap and stuff” procedure was not an undue burden for someone using multiple toll agencies, nor was setting up and paying multiple accounts. But Doran was decided early on in electronic tolling in 2003, at a time when Maine used an incompatible system, New Hampshire didn’t have electronic tolls at all, and fewer states had in-state discounts. Today, a commute from Salisbury, Massachusetts to Wells, Maine through New Hampshire would cost more than $1000 per year in additional fees without the swap-and-stuff maneuver. It was also decided at a time when we knew little about distracted driving (no one had a smart phone) and swapping requires taking your eyes off of the road to move the transponders while traveling at a high rate of speed. 

After 23 years, it seems ripe to be retried: the facts have changed. If it were, it could also force the 36 separate tolling agencies in EZPass to come to consolidate operations, IT and billing into a single entity. I’m not sure how much we pay for this sort of unnecessary redundancy, but I’m sure it’s not nothing. (Alas, I’m not holding my breath.)

c. Case study: Massachusetts in-state discounts

There is an interesting case in Massachusetts where some people are given toll discounts which are not available to outsiders: residents of four neighborhoods—Chelsea, the North End, Charlestown and East Boston—are given discounts for the bridges and tunnels starting and ending in their neighborhoods. Since these facilities predate federal highway funding and are tolled, the state can give these residents lower tolls for the crossings. 

On the one hand, giving a discount to local users in areas with good alternatives (the Blue Line and 111 bus, which provide frequent transit service) seems like questionable policy. On the other hand, it is, in a sense, redistributing toll funds from the users of the roadway to the communities affected by the roadway, albeit in a sort of backwards manner. There are significant externalities to having a major highway cut through an urban area, this allows the people most affected by the externalities to use the roadway at a discount. Of course, a better policy may be to give these residents discounted transit passes or improve the quality of transit service, rather than give drivers a discount.

By expanding the concept, some amount of road pricing could be returned to the communities through which the road passes to compensate the community for the negative externality. Perhaps communities could decide how the funds were best distributed, whether through toll discounts, a cash rebate to residents, discounted or free transit passes, improved transit service, or some other mechanism. Residents of these neighborhoods do not get such a choice, and instead the only beneficiaries are people who drive, even if many households in the neighborhoods don’t drive (or even own cars, especially in the North End).

One idea would be to establish some sort of “community impact fee” for all highways in a jurisdiction, charging a small price for each mile driven (perhaps as little as 1 to 2 cents per mile). More than a third of vehicle miles traveled take place on major highways which concentrates externalities in specific neighborhoods, and some portion of this road pricing could then be returned to cities and towns based on the traffic volume and the number of people living near the roadway. So a lightly-traveled highway in a rural area would receive a relatively small sum (although perhaps a large amount per capita) while a busy highway through an urban area would receive more (although perhaps the same amount per capita as a busy highway in a more rural area). In Indiana, some of the proceeds of the toll road sale have been returned to the communities. Maybe people would be more amenable to increased tolls if some of the funds were returned to their community.

V. AET, ORT and, uh … why is Maine still building toll booths?

Maine doesn’t just stop with bad toll policy. They’re building bad toll infrastructure, too. We’ve seen what happens when toll booths are eliminated but backwards toll policy remains. But Maine takes it to the next level: they have backwards toll policy and are doubling down on outdated, expensive infrastructure.

More praise for MassDOT, when they went to full all-electronic tolling in 2016, they were the first major toll road in the country to do away with tolls (several newer roadways had opened with electronic tolling) and many others have done this since. MassDOT was somehow on the leading edge of toll infrastructure (they have to be on the leading edge of something, I guess).

But some states still build toll booths. This costs more to build, more to maintain, pose a significant safety risk, and can then wind up creating additional traffic congestion. How’s that for a lose-lose-lose-lose?

A note on safety: this has been known since toll booths first popped up. Lane changes, speed changes, physical barriers: none of these are good. Tolls in Connecticut were removed in part because of a horrific accident at a toll plaza. It did not help that Connecticut’s toll road was unique in that rather than being built with infrequent exits and toll booths mostly on the slower exit ramps, it was tolled entirely by mainline toll booths across the entire roadway. The NTSB has faulted toll booths in deadly accidents like this one, but there is no legislation or policy to encourage adoption of electronic tolling, like creating a unified billing system, mandating RFID chips in license plates for toll payment and detection (maybe some privacy issues there, maybe we could chip plates and ban flock cameras) or requiring states to report on scofflaw drivers who avoid tolls out-of-state. This last point is especially fraught; even states with EZPass can’t seem to agree, and states lose millions of dollars to out-of-state toll scofflaws who don’t pay tolls in other states. As this article shows, a Pennsylvania driver may use an EZPass in their own state, but then remove it and rack up pay-by-plate tolls elsewhere, knowing there would be little mechanism to enforce the tolls in their home state. As usual, toll policy makes no sense.

a. Case Study: New Hampshire

Before Maine kept building toll booths, New Hampshire was the first to screw up open road tolling (ORT), and not by policy, but by physical geometry. Their experiment in open road tolling was hailed as the future when it was implemented in 2010, and it was, in theory, better than the previous state of a 23-lane toll booth with EZPass lanes on either side and cones which could be adjusted for volume in the middle. (New Hampshire was late to the EZPass party, too, not joining until 2006, with the half-price, unadvertised tokens available for 50 years before then.) Prognosticators hailed it as “a miracle of modern technology” and “completely state-of-the-art” installed by people who “were very concerned about these legendary backups.” (In this case, the prognosticator was the Boston Globe’s now-editor, Brian McGrory, around the time he was hating on bikes. So, wrong.)

One problem, Brian: they didn’t actually solve the backups, and the open road tolling made it worse. 

Some terminology before we go further: Open Road Tolling (ORT) versus All-Electronic Tolling (AET). The former has untolled through lanes flanked by toll booths. The latter has a gantry across the entire roadway with photographs used for vehicles without transponders. ORT is more expensive to install, operate and maintain, less safe, and the only benefit is that you can still take cash, so you don’t have to bill people without transponders. 

For a later implementation, New Hampshire would commission a report to study the difference between ORT and AET and the conclusion was that “AET yields positive net revenue compared to ORT in virtually every scenario” (Page 48), not to mention environmental concerns of a larger footprint for toll booths and safety improvements without merges and slowdowns. The caveat is that this is in relation to a “clean slate” implementation, where either an AET or ORT would need to be built, since AET is significantly less expensive, like at the Hampton Tolls nearby. However, it is clear that any toll agency moving from a toll booth system to something else should choose AET.

So converting the Hampton tolls to an AET system now may not have a positive payback given the infrastructure required (it certainly would have been better had it been implemented initially). But the question is whether the Hampton Tolls are actually a completely state-of-the-art miracle of modern technology. It turns out, traffic winds up moving faster through the cash lanes on busy days. One state official said that “it’s seasonal and as a prior DOT employee once told him, ‘You don’t build a church for Easter Sunday.’” The problem is that, to stretch this analogy, they have a church built for Easter Sunday, but didn’t build enough doors.

The issue for New Hampshire is not that they overbuilt the road, but instead that they wildly underbuilt the toll plaza through lanes. Designing an ORT system in the late 2000s, when there was significantly less-precise toll photography, is defensible. Designing one where only 50% of the roadway is dedicated to the open-road tolls is not. The congestion occurs because if the roadway is near capacity, and more than 50% of the transactions want to use EZPass, it creates a completely unnecessary bottleneck. There are other implementations of ORT where the roadway doesn’t narrow and the toll booths are off to the side (which can still induce congestion as people move across the roadway to reach the desired lanes). But to build a too-narrow roadway, and do so without the ability to easily widen it, is ridiculous. At this point, 90% of transactions are electronic, to the point where even at major toll booths the state leaves them unstaffed overnight and uses photo tolling for the few users without a transponder.

A further issue is the location of the toll plaza between two major highway merges. Seven miles north of the Hampton Tolls, the roadway splits into the Spaulding Turnpike north into New Hampshire and I-95 across the bridge to Maine. Seven miles to the south the road splits I-95 south to Boston and I-495 southwest to Lawrence, Lowell and Worcester. The four-lane Turnpike feeds into five lanes going north, and six going south. In theory, traffic should be relatively free-flowing on the four-lane section of the road, since it will have to merge from higher capacity roadways upstream. But the multiple sorting locations along the roadway cause significant congestion.

When the toll plaza was simply two dozen booths spread across the highway, there was no need for drivers to move back and forth across to find a desired lane. Motorists had 14 miles between the Spaulding Turnpike and 495 to filter across the roadway. For instance, someone coming from Maine to 495 would need to move from the left to the right, and someone coming from the Spaulding to 95 would move from right to left. This does create some congestion, especially when traffic is added from Route 101 (especially in the afternoon of summer days when the beaches empty), but EZPass made it worse, and ORT worse still.

When it was originally implemented, the EZPass lanes were placed in two banks so that for each direction of travel, EZPass users would not have to shift to one side of the road or the other. ORT moved EZPass users to the center of the roadway, meaning that many users now have to cross the roadway twice in the 14 mile stretch between merging and diverging. This doubles the number of what I call “sorts” and halves the distance between them. Even if the ORT lanes were widened to accommodate additional traffic, simply requiring people to move back and forth across the roadway creates traffic.

With the state set to double the toll for cash (and out-of-state transponders: reminder that anyone can get an in-state transponder), meaning even more users will move away from cash (and, hopefully, to NH transponders, to stick it to them). This will only exacerbate the issue at the Hampton Tolls. Maybe the additional revenue will allow the state to convert it to a full gantry system once and for all, moving the merge-and-sort back to one sort in 14 miles.

b. Case Study: Sturbridge

Even good AET can not make up for roadway deficiency. On Easter in 2009, a toll-taker shortage created an 8 mile backup on the Mass Pike approaching Sturbridge. Rather than waive tolls to increase throughput, the Turnpike authority elected to continue to collect tolls, and the backup ensued. Yet even with changes in policy and, a few years later, the elimination of the toll booths entirely, backups at Sturbridge still routinely occur in both directions.

Going eastbound, the backups make sense. The roadway is fed by the three-lane I-84 from Hartford and the two-lane Turnpike and feeds into three lanes. In other words: 3 + 2 = 3. (Well, really, 2 + 2 = 3; the onramp from 84 to the Turnpike is a bottleneck but still feeds more traffic into the main roadway than there is capacity.) If the upstream supply is anywhere near capacity, the system will break down.

But going Westbound, we wouldn’t intrinsically expect the same traffic, but it can be worse. Three westbound lanes feed into a two-lane onramp to Sturbridge and eventually the three-lane I-84, and the two-lane Turnpike westbound. In this case, 3 – 2 = 2, which should leave enough capacity, right? 

If only it were that simple. One issue is that there is more demand towards Hartford than Springfield, approximately 60% of traffic exits in Sturbridge to go south towards Connecticut and New York City. But the roadway is three lanes wide, and ⅗ is less than ⅔, so this should be manageable. The larger issue is that the majority of traffic has to sort itself into the two right-most lanes to make the exit, and when the roadway is near capacity, this perturbation can cause delays which send the roadway over capacity and cause congestion, which then builds a traffic jam. This is exacerbated by the nearby Charlton service plaza, where entering traffic needs to move across the roadway to stay on the mainline (at least the highest demand is to keep right). 

This is a case where “just one more lane, bro” might actually make a difference, a fourth lane from the service plaza west, and most importantly a third lane on the exit ramp to I-84, could alleviate some of this congestion. The current configuration of the roadway going west is two lanes to stay on I-90 and two onto I-84, while the traffic split is 60/40. Expanding the ramp would allow the lane assignments from left to right to be 90–90/84– 84–84, better allocating the lanes for the traffic demand. (Going east is harder, a third lane on the ramp would necessitate a firth lane on the main line and an eventual merge, although allowing I-84 traffic to maintain a higher speed on the steep uphill past the interchange, where trucks accelerate slowly, may also help.)

c. Case Study: Pennsylvania

The change from cash to electronics did seem to come quickly, and agencies struggled to adapt at varying rates. For many years, the Pennsylvania Turnpike had a toll booth just east of the Delaware River Bridge. In 2016, with a connection to I-95 finally underway (this was a decades-long ordeal) and would mean that drivers coming off the bridge would have to enter the ticketed toll road, and then immediately exit onto 95. This would mean, for some drivers, three toll booths in about 10 miles: one to exit the New Jersey Turnpike, another to cross the bridge and enter the Pennsylvania Turnpike, and then yet another to exit onto I-95. (There are, or at least were, a lot of double booths near state lines; one to take payment for completing one state and the next to give a ticket for the next state. As far as I know, no states ever combined the facilities into one where the toll-taker would process one state’s transaction and give out a ticket for the next, because that would make too much sense.)

So the Delaware Bridge management (equally the Pennsylvania and New Jersey turnpikes, different from the earlier Delaware Bridge toll authority) decided instead to go electronic. But they still needed to give out tickets and retrieve tolls from the rest of the Turnpike, so they built an entirely new toll booth a few miles away which opened in 2016, the same year Massachusetts was removing all of its toll booths. The cost? $85 million, not much less than Massachusetts’s full conversion. This toll booth was in operation for less than four years. The pandemic led the state to shift more quickly to all-electronic tolling, and it quickly became permanent. It’s one eerie to drive through a brand-new, and mostly abandoned, toll facility.

d. Case Study: Maine (again)

So given all of this information, toll agencies wouldn’t set out to build new ORT in 2025, would they?

Let’s go one state north from New Hampshire to Maine to find out.

Maine’s first foray into open road tolling was the New Gloucester toll booth which gives us another example of how people value their time. A trip on the Turnpike takes about 8 minutes less than using the parallel Route 100, and at the time of a Maine Turnpike-commissioned study, cost $1.25 less (with a Maine EZPass, the highway is somewhat less expensive now, especially for frequent users). At the time, the $9.37 per hour savings (about $12 today) was enough for about 10 percent of drivers to take the longer route around the toll. 

The toll booth there was replaced with ORT in 2013, and ORT was, at the time, somewhat defensible, in a sense because the sensing for license plate images was not as good as it is today. (It also doesn’t cause traffic backups like in New Hampshire because the AADT of the roadway is only about 10,000, so volume is never high enough to cause issues, even with a single through lane.) Maine’s main toll booth in York, like the castle in Monty Python, was sinking into a swamp. It needed to be replaced and relocated. New Hampshire knew that replacing a toll booth with anything but AET did not make sense.

With the current information and analyses of toll infrastructure clearly favoring AET, what did Maine choose for its more recent implementations? They went with ORT. They had a study to back this decision up, and it is quite possibly the most questionable such study I’ve seen (and I’ve read a bunch of, uh, interesting studies—that’s a Midwest “interesting” for those paying attention). It makes the assumption that since one third of vehicles at the York tolls are from out-of-state, they will have trouble recouping unpaid tolls, and there will be little incentive for infrequent drivers to switch to EZPass just to save time in Maine. This leaves out not just reciprocity with New Hampshire and Massachusetts, but that nearly every user of the York tolls will have already had to pay a toll on the adjacent roadway in New Hampshire, and most coming from Massachusetts or further would pay tolls in those states as well. And 90% of New Hampshire toll users are using EZPass, so this assumption is just completely wrong, since it is, for all intents and purposes, the same road. This creates an assumption of a very high proportion of “leakage” of unpaid tolls.

The study notes that in order to recoup this leakage, they would need to add a $3 surcharge to cash tolls, which would push more drivers to the parallel roadway, suggesting that something like 40% of cash-paying traffic would choose a different route if the surcharge were applied. Yet this seems unreasonable. As we saw earlier, about 10% of people will skip a toll if it costs them $10 per hour. Their $6 toll here would compare with $1.50 to enter the highway upstream in Kennebunk, a $4.50 differential, but even under the best of conditions this would cost about 15 minutes, although with traffic on Route 1, it’s often closer to 20 to 30. So under the best of conditions, this would be $14 per hour, under typical conditions, $10.50 and when Route 1 is particularly busy more like $7. Yet their analysis assumed that 40% of people would make this decision, which would add to congestion and travel time on Route 1. There is no citation given for this, the consultants seemed to make the number up. (Consulting: getting paid the big bucks to make numbers up!) And even with AET having a better NPV, the state squinted at the study and went with ORT.

They didn’t fool everyone. Nearby residents pushed the state to adopt AET, but the state refused to do so, and in 2021 opened a new, $39 million ORT toll booth. Within a year it was clear that it was obsolete. From the end of this article is Turnpike Authority chair Peter Mills attempting to save face:

Mills disagrees that the new plaza will become obsolete that quickly. Even if it stops accepting cash, cash lanes can still collect electronic tolls. That provides a crucial stopgap if there is a malfunction of the mainline electronic tolling lanes, he said.

Last year, vehicles were routed through the cash lanes at the New Gloucester toll plaza because of damage to underground sensors in the electronic lanes.

“If our lanes all went down for some period of time, that is a financial casualty of the first order,” Mills said. “Having an extra lane here and there is trivial if it provides the protection we need.”

(Obligatory note that Peter Mills, a Republican, is the sister of Governor Janet Mills, a Democrat. Maine is weird.)

The state spent $40 million to build a white elephant, and is now justifying that decision by saying “well if the toll gantries malfunction, we can force everyone through these auxiliary lanes, which happened once somewhere else. I’ve driven through this toll booth recently, with a constant stream of traffic in the through lanes and not a single vehicle at the toll booths. What a waste. (Maine continues to build ramp toll booths. ¯\_(ツ)_/¯)

VI. So is tolling congestion pricing?

No … and yes.

There are prerequisites for congestion charging, which include high congestion, good transit alternatives, and a relatively compact, contiguous area. Bay City may create congestion, but it’s not congestion pricing. The Pennsylvania and Jersey Turnpikes may have high tolls, but they’re not pricing congestion.

Congestion charges are basically a toll spread across a boundary. Depending on the geography of a city, these can be easy to implement, or considerably more difficult. A high enough congestion charge at a porous border would have a host of issues. Applying it only to highways would push traffic onto local roads through neighborhoods. Applying it in an area with significant local vehicle traffic would arbitrarily charge these local trips. (One critique of Manhattan’s successful congestion charge is that it arbitrarily tolls a trip from 66th Street to 56th Street but not 71st street, which is easily refuted by “there is really no reason to be driving from 66th Street to anywhere, and if you are, you can probably afford the charge.”)

These are similar to issues with tolling highways, if a congestion fee arbitrarily makes winners and losers or is easy to avoid, then people will avoid it.

• Stockholm’s congestion area is mostly on a series of islands, and crossing onto the island creates natural boundaries and bottlenecks for tolling. 
• Manhattan is somewhat similar as an island, although rather than tolling the entirety of Manhattan, a line is drawn at 60th street. Charges will increase to $15 by 2031. Per-trip fees for taxicabs and ride shares are between 75¢ and $1.50.
• Singapore has more of a toll system, with a series of gantries around the center of the country which doesn’t have significant internal physical boundaries. There is a maximum charge of $8 for foreign vehicles, local vehicles pay a distance-based fee based on satellite surveillance. Göteborg and Oslo have similar systems
• London’s charging zone has the fewest natural boundaries, Milan is also relatively porous. New York’s congestion zone covers about 9 square miles and has about 18 roadways entering, since most of the boundary is made up of rivers with few crossings. London’s is only slightly larger (about 11 square miles) but has several hundred entrances. This sort of system requires a dense enough core with good enough transit that there are few local road trips which are affected. London and New York have significantly higher charges than other cities as well, although London has only about one tenth as many trips as New York, which raises by far the most revenue.

City	Area (sqmi)	Charge (day max)	Vehicles/day
Stockholm	12	$1-$5 / $13	100,000
New York	9	$9*	500,000
London	11	$24	40,000
Milan	3.2	$8.50	81,000
Göteborg	No defined boundary	$7**	28,000
Singapore	No defined boundary	$8**	300,000

* Will increase in coming years.
** Maximum charge within a defined time period, lower charges if only some toll gantries or waypoints are passed.

If we look at the United States, there are only a few cities where congestion pricing would be feasible, i.e. where there is good-enough transit and a high-density area with high demand. (In other cities, like Atlanta or Los Angeles, it would be more akin to road pricing since demand is more spread out and there are fewer non-car alternatives.) Each may have to rely on a different model if it were to implement congestion pricing.

• San Francisco already has tolled bridges from two approaches with significant geographic features to the third (San Bruno Mountain), a New York-style system would work, although this would not necessarily help as much with local traffic congestion in the city itself.
• Chicago is laid out on a grid and has no bottleneck features. A Chicago-Halsted-Roosevelt zone would have several dozen crossings, and a high-enough fee might see people parking or being dropped off just outside the zone. A Singapore-style system, with lower tolls on outlying expressways, might make more sense.
• Boston is sort of a hybrid of Chicago and San Francisco. Like London, it has a web of roadways with a river through the center of the most congested area, but unlike London, it has highways through the downtown area, some of which are tolled, and some of which aren’t. The untolled highways pass through relatively large geographic features (namely the parks they were pushed through in the 1950s) and tolling these roadways, and potentially some subset of river crossings, would generate considerable revenue, even if the charges were not particularly high.
• Washington, D.C. has a large network of roads and some natural boundaries, with fewer highways through the city.

There are two main benefits of congestion pricing. One is, obviously, to reduce congestion, which has been observed where it has been implemented. The other is to raise revenue. Congestion pricing need not be revenue-maximizing to be effective: a revenue-maximizing scheme may raise more revenue, but may not be politically feasible. For a city like Boston, a relatively low toll similar to Stockholm, which ranges from $1 to $4.50 based on the time of day, may be plenty to generate revenue to improve the transit system parallel to congested roadways, even if the charges only have a marginal effect on roadway congestion.

a. Case Study: Congestion Pricing in Boston?

Relatively low tolls still have high effects, although this likely has do more to raising revenue than to reducing congestion. We’ve explored many of the pitfalls of poor toll policy and bad economics, so how could congestion pricing be implemented in a city like Boston? 

In Boston, there are about 300,000 tolled trips into the city per day (on the Turnpike and the harbor bridge and tunnel) and about 400,000 untolled highway trips on I-93 north and south of the city. Two cordons could capture most of these untolled trips. A southern cordon could run along the Neponset River, through Milton and across the Blue Hills (with perhaps a portion of Milton where a driver would have to cross two lines so as not to penalize local traffic, even if most Miltonites could afford it) and another spanning the Middlesex Fells. (The Department of Conservation and Recreation, which originally gave up the land for these highways, should also toll its own roads, which run through parkland along the Charles River; the busiest of which—when it’s not decapitating trucks—provides a free alternative to the tolled Turnpike. The cash-strapped agency could solve its budget problems overnight.)

What about the border tolls proposed by ex-MassDOT Secretary in 2024? Aside from being odd optics, it’s bad policy, because the border is easy to circumvent. Depending on the highway, circumventing the toll would require between 3 and 7 minutes, so even at a $1 toll, this would mean a value of time of $9 to $20, the higher end of which would mean a lot of diverting traffic. (Avoiding the Fells or Blue Hills/Neponset would cost on the order of 15 to 20 minutes of increased travel time, so $2 to $3 tolls would likely dissuade only a few drivers.) It would also do little to target congestion where it is worst, especially since there is only a minimal viable transit alternative along most of these corridors (and a lot of the traffic crossing the border isn’t going to Boston anyway). Again, we need proposals grounded in economic reality, not in sound bites.

If the congestion charge for passing through the Fells and Blue Hills were set to be equivalent to the Turnpike and bridge tolls and all were adjusted for inflation (tolls haven’t been increased since 2016), it would generate somewhere on the order of $500 million per year. This funding could be funneled towards material improvements to regional transit—full accessibility, grade crossing elimination at busy roadways, electrification, rail improvements for Western Mass—to provide a better alternative to already crowded highways (also a substitute good!). 

It should be noted that tolls on the Turnpike are quite low, and have gotten lower over time. When it opened, a trip from Weston to Boston cost $2.40. Adjusted for inflation, that would be nearly $30 today, yet it costs $4.25. The “extension” from 128 to Boston originally cost 50¢, which would be equivalent to $5.37 today; the current toll is $1.70. So even as traffic has increased, since 2016, toll revenue has only increased by 10%, but adjusted for inflation, revenue is down 22%, or more than $100 million per year.

Raising tolls on the Turnpike is fraught, because state law requires that funds from the Turnpike be used to maintain the Turnpike, so raising tolls to pay for investment parallel mass transit would require a change to statute. Because of this statute, the Turnpike remains one of the cheapest toll roads, per mile, in the country, even if it is one of the only such roads which leads directly into the main city in its state (most other legacy toll roads skirt the main city in the state; the 1965 Turnpike Extension brought the road into the city, for better or worse). 

Massachusetts transportation funding flowchart is already a web of arrows pointing in various different directions. And federal statute gives states wide latitude on how they can deploy their federal transportation funding, especially by “flexing” federal highway funding to cover transit projects or operation. Massachusetts has historically transferred about 3% of its highway funding to transit (at least in recent years). This is in the middle of the pack in comparison to other states, but well behind California (10%) and New Jersey (close to 15%). Congestion pricing would allow these funds to shift around while maintaining funding for aging roads and bridges. With more toll revenue, it would be easier to justify flexing more federal funding to improve transit.

Should we raise tolls on the Turnpike to at least keep up with inflation, especially if we used the surplus to pay for (and leverage potential federal funding for) parallel mass transit improvements? Probably! But there is a political argument that additional tolls on drivers coming from the west (who are mostly Massachusetts residents) while drivers from the north and south (a higher makeup of whom are from New Hampshire and Rhode Island) get a free ride isn’t unfair, or, at least, politically popular.

Road pricing is clearly a reasonably way to raise revenue and change behavior, but is too frequently mangled by policymakers who seem not to understand basic economics. So any such increase and change would likely have to be paired with additional tolling or congestion pricing, and would require a deft politician with an understanding of basic math, something which, alas, is in short supply in the Commonwealth. Massachusetts was a leader in creating a safer tolling system and has mostly avoided the pitfalls of poorly-thought through economic measures. But it is a laggard in using that system to benefit the traveling public.