Note: I have recently published a more detailed, more updated scholarly version of this in Transportation Research Part C: Emerging Technologies. “Downtown congestion pricing in practice” https://www.sciencedirect.com/science/article/pii/S0968090X18306983 I am going to clean up a pre-print version soon that will be open access.
Recently New York Governor Andrew Cuomo announced, “Congestion pricing is an idea whose time has come.” He is talking about proposals to charge tolls on driving into parts of Manhattan, such as the Move NY plan. Mayor de Blasio, on the other hand, “does not believe” in congestion pricing — hopefully in the sense of not supporting it, not denying its existence.
I wrote my Ph.D. thesis at UC Berkeley about congestion pricing in downtown areas like Manhattan, and I read a lot of news articles about its prospects. I have become concerned that news coverage does not pay enough attention to other cities’ experiences with pricing. The article on Cuomo’s support, for instance, simply says: “Cities across the world, like London and Stockholm, have adopted systems that have succeeded in reducing traffic and improving public transit.” But the lack of context isn’t surprising: the thousands of pages academics have written on the topic are almost all behind paywalls.
Unfortunately, ignoring other cities’ experiences can hobble clear thinking. One gets the impression that there is a coherent, monolithic system called “congestion pricing,” which either works or fails. For example, Donald Blinken writes to the New York Times:
After congestion pricing was introduced there some years ago, there was a decline in central London traffic. But as anyone who has been to London lately can tell you, the traffic situation seems even worse than before. A city government website confirms the recent increase.
In other words, congestion pricing has nothing meaningful to do with long-term traffic reduction but is simply an additional tax on motorists.
While London has become congested, the fact requires some context: First, London has undergone a sharp fall in road capacity. So, the London Congestion Charge has likely kept speeds from falling more than they have. Second, the London Congestion Charge has many exemptions which compromise its effectiveness but are not an essential part of pricing as an idea.
To clear things up, I’ve written the following history of downtown road pricing. In particular, I emphasize the stories of how the schemes came to be, politically and administratively, in keeping with the spirit of Bent Flyvberg’s theory of Phronetic Social Science: “It is a basic tenet of phronetic research that in so far as social and political situations become clear, they get clarified by detailed stories of who is doing what to whom.” I close with a list of open access references.
Here are some things to keep in mind while reading.
This has not been peer reviewed. If you find a mistake, please comment.
The scope is limited. First, I am talking about the recent, rare practice of tolling streets of large downtowns — not the old, common practice of tolls on bridges and major roads, nor tiny systems like Durham’s or Valetta’s. Second, I ignore schemes that were only proposed, with the exception of Hong Kong’s.
I use “road pricing,” “congestion pricing,” “charging” and “tolling” interchangeably, but I prefer “downtown road pricing” to be very broad. And by “scheme” I mean the British sense of the term: a specific program design — not a con job.
I have tried to translate quoted costs and prices into 2016 US dollars, but doing so is tricky and the figures are only meant to be a sort of ballpark estimate. The conversions are in parentheses, like “It costs 10 kr ($1.2).”
Road tolls have interested economists and civil engineers for centuries. In The Wealth of Nations, Adam Smith writes:
When the carriages which pass over a highway or a bridge, and the lighters which sail upon a navigable canal, pay toll in proportion to their weight or their tonnage, they pay for the maintenance of those public works exactly in proportion to the wear and tear which they occasion of them. It seems scarce possible to invent a more equitable way of maintaining such works.
What concerns Smith in this passage are explicit costs — the “wear and tear” on the road — and the “equitable” way to assign them. To Smith it is fair to put the burden on users — not taxpayers.
In 1920, Arthur Pigou published The Economics of Welfare — one of the most famous and influential economic treatises of all time. In one passage, Pigou describes a situation in which road tolls can be used to redirect traffic so as to cut the aggregate time spent traveling. This is a sea change in thought: now tolls are not merely an equitable way to pay for roads, but a way to efficiently manage them. You could draw an analogy to traffic signals: signals do ensure that travelers in orthogonal directions both get a chance at crossing an intersection (fairness), but it is also possible to coordinate signals to improve traffic (efficiency).
Since Pigou, most research on pricing has been theoretical, answering questions like: “What should the toll be if the macroscopic fundamental diagram theory of traffic flow applies?” But transportation economists have also always been deeply enmeshed — institutionally and by enthusiasm — in civil engineering, and have thought deeply about applications. For road pricing in downtowns — where there are thousands of intersections, and where streets are small and crowded — the chief problems are these: (i) how to check when and where a car is, and (ii) how to charge the driver. You can’t have many tollbooths in Lower Manhattan.
In a 1959 testimony to the US Congress, William Vickrey (future Nobel Laureate for his work on auctions) spelled out a wireless, electronic road pricing system for cities that would make tollbooths unnecessary. Having studied electrical engineering in college, he even built a prototype in his driveway and would show skeptics a list of his own comings and goings.
Ignored by Congress, Vickrey found a readier audience in the UK, where there was an appetite for less driving. A 1963 report, Traffic in Towns, conveys the British zeitgeist in a very British style:
It is impossible to spend any time on the study of the future of traffic in towns without at once being appalled by the magnitude of the emergency that is coming upon us. We are nourishing at immense cost a monster of great potential destructiveness, and yet we love him dearly. To refuse to accept the challenge it presents would be an act of defeatism…
The American policy of providing motorways for commuters can succeed, even in American conditions, only if there is a disregard for all considerations other than the free flow of traffic which seems sometimes to be almost ruthless. Our British cities are not only packed with buildings, they are also packed with history and to drive motorways through them on the American scale would inevitably destroy much that ought to be preserved.
Since Britain wasn’t about to nourish at immense cost a monster of great potential destructiveness, despite loving him dearly, the Ministry of Transport empaneled a team of engineers and economists — centered at Britain’s Road Research Laboratory — to study road pricing in cities. In 1964, the panel published “Road Pricing: The Economic and Technical Possibilities” (called the “Smeed Report” for the panel’s leader, Reuben Smeed).
None of the Smeed Report’s schemes were adopted in the 1960s, but the report advanced principles that would later blossom. Perhaps most importantly, it consistently embraces the notion of tolls for use of entire city zones, rather than trying to have, say, a different toll on Broadway than on 2nd Ave. The drawback, of course, is that zone-level pricing can only go part of the way toward “optimality”: there may be a nasty pocket of recurring congestion somewhere inside a zone.
1975: Singapore’s Area License System (ALS)
Singapore is an island nation a little smaller in area than New York City and with 5.6 million people. When it gained independence from Britain in 1965, it had a GDP per capita equal to about 1/6 of the United States’. Today its GDP per capita is about equal to ours, and its median household has twice the wealth of the median US household. Remarkably, for such a dense and rich country, growth has come with little congestion, thanks to plans adopted in the 1970s — including the world’s first downtown pricing scheme.
How it happened
Between 1960 and 1970, the number of private autos in Singapore more than doubled. Ordinary people were making more money, and the state was energetically building “New Towns”: massive public housing projects away from the Central Business District (CBD). Meanwhile, the volume of public roads rose only 35%, and there was no rail system nor a coordinated bus system.
The prospect of a flood of cars worried Singapore’s leaders: they nurtured a vision of the CBD as a hub of international business and thought traffic would strangle it. In 1971, the Singapore Concept Plan, a modeling and planning study, came out forecasting rapid growth in traffic and calling for a huge infrastructure campaign. Road-building began right away — road spending rose 6x from 1975–1980 and road density doubled from 1975 to 1985. But disputes over cost-benefit analysis delayed work on the Mass Rapid Transit rail system until 1982.
While pro-construction, the government was sober about whether infrastructure alone could balance supply and demand for road space. So in January 1974 the government convened the Road Transportation Action Committee (RTAC), with members from across the bureaucracy, to tackle demand.
Regarding car ownership, the RTAC aimed for no more than one car per twelve persons. To achieve this goal, the government raised the Additional Registration Fee (a sales tax on new cars) to 25% in 1972, 100% in 1975, and 175% by 1983. Since 1990, the government has run a complex system of quotas, taxes and rebates to make cars costly and clean.
The RTAC also did three things to discourage cars from driving downtown: (i) double downtown parking prices; (ii) improve bus service via commuter shuttles and a park-and-ride system; (iii) the Area License Scheme — the world’s first downtown pricing system.
The Area License Scheme (ALS) launched on June 2nd, 1975. At post offices, gas stations, convenience stores and roadside booths, a driver paid S$3 ($6.20) for a daily “license.” The license was a paper decal, to be placed in your windshield, that let you enter a 6.2 km² Restricted Zone (RZ), Monday through Saturday morning. For enforcement, wardens standing at 22 access points to the RZ inspected passing vehicles for licenses. The design loosely follows one proposed in the Smeed Report.
ALS immediately cut traffic into the Restricted Zone during charging hours by 44%— much more than the 30% anticipated. What happened to all this traffic?
Among commuters to jobs in the restricted zone, the share commuting in cars with less than four passengers dropped from 48 percent to 27 percent during the first few months of operation, while the combined modal shares of carpool and bus rose from 41 percent to 62 percent. (Small and Gomez-Ibanez, 1998, p. 216)
Downtown speeds rose 20%, but enough traffic switched to an overwhelmed ring road around the Restricted Zone to cause serious crowding — at least in the short run. Also, so many people switched to buses that bus travel times actually rose. So, all things considered, the initial benefits of ALS were ambiguous, and economists concluded it made travelers worse off.
But over time, Singapore was able to adapt the ALS to keep traffic restrained in the face of precipitously rising employment, vehicle ownership and incomes. Charges were increased and differentiated by vehicle type, the boundaries of the Restricted Zone expanded, exemptions ended and new charging periods added. See the table below. The most substantial changes were to exemptions: originally, ALS exempted taxis, motorcycles, commercial vehicles, carpools and public transit, but the taxi exemption was ended within three weeks, as were the others, except for transit, in 1989. As we will see below, exemptions are among the most consequential parameters of a pricing scheme.
1982: Hong Kong Electronic Road Pricing (ERP)
In the late 1979, Hong Kong found itself in the same spot as Singapore ten years earlier: during the 1970’s, real per capita income and the number of registered vehicles both doubled, but the length of roads rose only 17%. In 1981, there were 282 registered vehicles per km of road space, the highest ratio in the world, even though only 17% of the population owned cars.
How it happened
Faced with worsening congestion, in 1979 the government released a White Paper endorsing specific expressways and rail transit as well as vaguer calls to reserve road space for transit and goods-moving vehicles. By 1982, the annual rate of growth in the vehicle population had reached 12%, prompting the Secretary for Transport, Alan James Scott, to announce draconian “fiscal measures”: the government tripled annual license fees for private cars, raised the initial registration fees on cars to 70–90% of the purchase price and more than doubled the gas tax.
The fiscal measures were effective in one sense: the number of registered private autos fell 42% from 1981–1984 (a recession helped). But traffic worsened anyway, because households bought vans instead of cars and registered them as “commercial” vehicles; and, more fundamentally, because it is vehicle driving, not vehicle ownership, that clogs the roads. So in March 1983, Hong Kong contracted with consultants from Britain’s Road Research Laboratory (origin of the Smeed Report) to design and test a wireless system called Electronic Road Pricing (ERP), similar to Vickrey’s earlier proposal. The study lasted from July 1983-March 1985.
On the study’s strategy side, consultants carried out modeling studies to compare various cordon locations and charging structures. These were much more complicated than anything that has ever actually been implemented, having over 100 toll sites, multiple cordons and directionally-dependent prices.
On the technical side, the ERP worked like this: A device called an Electronic Number Plate (ENP) was mounted under the vehicle. When a car passed over an inductive loop in the pavement, the ENP became energized and returned a vehicle ID, which a modem would route to a central computer. Every month, a user was mailed a bill listing the time and place she had crossed zone boundaries during charging hours. Field trials involving 2,600 equipped vehicles and 18 tolling sites showed the technology to be highly reliable.
Despite successful field trials and modeling that suggested significant traffic improvements, Hong Kong’s government decided not to adopt ERP when the study ended in 1985. Hau (1990) offers ten reasons for the failure-to-adopt, but two stand out: First, congestion had been ameliorated, temporarily, by outside events: a new expressway, a new rail line, the fiscal measures of 1982 and a recession. Second, in December 1984 Britain had agreed to deliver Hong Kong to China, and while it might have been a well-intentioned show of transparency, the bill listing one’s movements was destined to exacerbate worries about communist surveillance.
1998: Singapore Electronic Road Pricing (ERP)
By the 1990s, Singapore’s Area License Scheme had become unwieldy:
About 60 enforcement personnel and another 60 officers at dedicated license sales booths were required each day. The enforcement duties were demanding, given the long hours spent under the sun and rain, not to mention the dust and the noise. In addition, there were 16 different types of licenses in use at its peak, and much concentration by the enforcement officers was required to ensure that they identified them correctly. — Chin (2010)
Beyond the manpower needs, there were fundamental limits to a system based on paper licenses:
- It couldn’t charge per entry.
- It couldn’t change prices in increments smaller than a few hours.
- It was easy for people to cheat by switching licenses among vehicles.
Since Hong Kong’s field trials vouched for the wireless option, Singapore solicited bids for an electronic system, and in October 1995 awarded a contract for its development. After extensive trials and publicity, the system, also called Electronic Road Pricing (ERP), launched in September 1998.
Singapore’s ERP can be understood in terms of two technologies:
- The In-vehicle Unit (IU), a device that sits on dashboards (or handlebars for motorcycles). The IU has a slot for a “Cash Card,’’ a stored value card to be load with money at 7-Eleven or post offices. There are IU’s for six vehicle classes (see below), because the toll charged to a vehicle is a multiple of its “Passenger Car Unit’’ (PCU), an index of vehicle size— e.g., a bus pays twice as much as a car, which pays twice as much as a motorcycle (see below). The IU costs S$150 ($110).
2. Gantries, the big arches below. Gantries work in pairs. When a vehicle passes, the first gantry fires a radio signal to the vehicle’s IU, telling it to charge the CashCard appropriately. An optical sensor on the second gantry confirms the vehicle is the type attested by the IU. If there is an error, a camera on the first gantry captures the vehicle’s rear number plate.
You might ask: “Why does Singapore require everyone to have this big piece of equipment in their car, instead of a little EZ Pass-type RFID device?” The answer is privacy: the gantry tells the IU to charge the card some amount (rather than the IU signaling the gantry your account ID), so the system only records your passage if there’s a mistake or your card is cashed. Recall that Singapore designed its ERP only a few years after privacy concerns had sunk Hong Kong’s.
The gantries form rings (“cordons”) to enclose certain areas with certain prices. ERP started in 1998 with 33 gantries that reproduced ALS’s Restricted Zone around the Central Business District. Beginning in 1999, gantries were added to create an “Outer Cordon,” and they have since a little sub-cordon around the Orchard Road shopping area. Wikipedia says there is a route that will cost S$15 ($11) at the peak. Today, there are around 80 gantries in operation. Check out gantry locations and prices here.
Prices at every gantry vary over the day. The figure above illustrates a weekday toll schedule for the CBD cordon in different years. Since 2003, whenever tolls would otherwise change by more than S$1, there is a five-minute interval in which tolls rise or fall by half the amount of the change, to keep cars from dangerously waiting or speeding to save money. The Land Transport Authority (LTA) updates the schedule every three months to maintain speeds of 45–65 kph on expressways and 20–30 kph in the RZ — the speeds engineers think maximize flow.
From news articles it looks like revenues are around S$150 million per year. Revenues are not hypothecated; they flow into the government’s general budget, of which they form an insignificant fraction. However, ERP is not designed to be a cash cow. The switch from ALS to ERP actually cost about 30–40% of revenue because, being more precise, ERP tolls could be cheaper than an ALS license and still achieve traffic targets. Also, since the government charges very high taxes on gasoline and ERP reduces travel and idling, the overall effect on the budget is ambiguous.
Implementation cost S$197 million ($172 million) in 1998, of which S$100 million paid for IU’s and S$97 million for the gantries and computing. Annual operating costs have measured about 20–30 percent of revenues.
2003: London Congestion Charge (LCC)
Singapore and Hong Kong are quasi-city states with fairly light car ownership. Thus, they could put a detection device on nearly every vehicle that might reasonably enter on a given day. By contrast, cars enter London from all over Britain and even Europe, and London can’t easily require they all have a charging device installed. Consequently, although British academics have always played an outsized role in transportation economics and engineering, London had to wait on pricing until Automatic Number Plate Recognition (ANPR) became reliable and cheap in the 1990’s. London is also an interesting case study for congestion pricing, because — unlike Hong Kong or Stockholm — no body of water isolates its busiest areas.
How it happened
After the Smeed Report, studies of different schemes for road pricing in London were commissioned and carried out constantly from the 1970s through the mid-1990s. But after every study, decision-makers shelved the study’s proposal until a later date — in large part due to uncertainty over technology.
The ball finally got rolling in a real, albeit roundabout, way in 1997. That year, the Labour Party won power after a long hiatus and published a Green Paper proposing a directly-elected Mayor of London and Assembly to govern the Greater London area. The next year, the people of Greater London voted overwhelmingly for the Mayor/Assembly idea.
As it happened, Labour had proposed that one of the Mayor’s powers be the ability to enact downtown pricing, and so the Government established a working group of experts to study the matter one more time. The group was called Road Charging Options for London, or ROCOL. In 1999, obeying the referendum, Parliament passed the Greater London Authority Act, which among other things set out ground rules about road pricing — e.g., that the first ten years of net revenues fund transport improvements.
In March 2000 ROCOL published its findings as a big report, which endorsed using ANPR. Two months later, Ken Livingstone was elected Mayor of London with a promise to consult the public on pricing. Following a public consultation, Livingstone released a concrete plan in July 2001 that softened the original ROCOL plan — e.g., by offering discounts to residents of the Charging Zone.
The London Congestion Charge (LCC) launched on February 17, 2003 as a £5 ($10) license to travel or park on the street within the 22 km² Charging Zone (CZ) (shown below) between 7 AM and 6:30 PM on weekdays. The scheme is enforced by Automatic Number Plate Recognition.
The LCC has several distinctive features:
- It is the only downtown pricing scheme that charges for all use of road space inside a zone — not traversals of a cordon. To do so, at launch, the system utilized 500 cameras positioned throughout the CZ.
- It does not vary by time-of-day.
- Transactions are the user’s responsibility: a driver without an Auto Pay account must pay online, by phone or in participating stores before midnight on the day of entry to avoid a fine.
- It is expensive. The toll was raised from £5 to £8 in July 2005, £10 in January 2011 and £11.50 ($15) in June 2014. Why so expensive? I give two reasons: First, the Charging Zone is well served by transit and easy to avoid; hence, much chargeable traffic there consists of visitors, deliveries, millionaires, etc. So the LCC targets a very price-inelastic market segment. Second, there are so many exemptions the toll needs to press hard on chargeable traffic.
- It offers an extraordinary number of discounts and exemptions. (See the details of discounts here). The most consequential are for taxis and private hire vehicles. [Note: A “blue badge” is a Europe-wide handicap designation. Private hire vehicles are vehicles licensed to do pre-arranged pickups, whereas taxis are hailed in the street.]
Being relatively new, the LCC has undergone only one really substantial change in its short history: In February 2007, Transport for London added a 19 km² area around Hyde Park called the “Western Extension.” But data didn’t suggest the Extension was making a big difference, and the public opposed it, so London’s second Mayor, Boris Johnson, abolished it in January 2011.
The LCC prompted a sharp, sustained decline in entries to the Charging Zone by chargeable trips (e.g., cars/minicabs, lorries (trucks) and vans) and a sharp increase by exempt trips, especially taxis.
Unfortunately, the shift to exempt traffic has diluted the scheme’s effectiveness. Transport for London (p. 12, 2007) notes:
Analysis of vehicle movements (four or more wheels) into charged area within a charging day indicates that about 40% are made by vehicles for which a full charge has been paid. A further 4% or so of incoming movements are made by residents’ vehicles; 7% by buses; 23% by taxis; 10% by London licensed private hire vehicles; 5% carrying a notified blue badge holder; 2% exempt emergency service vehicles; around 4% potentially liable to a penalty charge; and 5% by other exempt or 100% discounted vehicles.
This is a buried lede: in 2007, only 40% of entering vehicles with 4+ wheels paid the full charge! And 33% were licensed cabs and private hire vehicles. There is no compelling social benefit for these trips, like there is for emergency and handicap vehicles.
Reducing entries is only a means to several ends — one being lower travel times. At first, the LCC did speed up traffic, but by 2007 all the original gains had been lost, and bus speeds had fallen 8 percent. The chart below shows travel time trends. In the figure, the light-colored part of a bar shows how many minutes it would take to go one km in free-flow conditions, while the dark-colored part shows the extra time attributable to congestion. The lower the better.
Bleak as this picture seems, it needs to be taken with a grain of salt. All of London, not merely the congestion zone, has undergone a long trend of falling traffic speeds. Transport for London blames a deterioration in road capacity.
There are two reasons roads have lost capacity. First, constant construction — especially roadwork and utilities work. (See this INRIX report). Second, authorities have reallocated capacity from cars to pedestrians, buses and cyclists: Transport for London (p. 2, 2007) attributes
- “widespread use of traffic control and safety-related measures on major and minor roads”
- “continuing allocation of the road carriageway to bus services, including bus lanes and bus priorities at traffic signals”
- “enhanced provision for cyclists and pedestrians at junctions”
- “enhancements of the ‘public realm’ with increased space for pedestrians”
The same report also raises an intriguing possibility — that this is happening because of congestion charging:
The implementation of capacity re-allocation measures may have accelerated in the post-charging environment and will have contributed to the continuing decline in speeds. It is probable that some of these measures have been enabled by charging and would not have happened had charging not reduced traffic levels in the centre of London. (p. 3)
In addition, of course, the Charge suffers from too many exemptions.
ROCOL had predicted setup costs would be about £30–50 million ($60–100 million), annual operating costs in the same range and £230–270 million ($460–540 million) in annual revenue. Unfortunately, costs have been much higher than expected, while revenue in the central charging zone was lower than expected until recently. See the table below. One reason for the revenue shortfall is the number of exemptions. Note that penalties (for the years when such data are available) accounted for substantial revenue.
Net revenues have mainly funded bus service.
Green, Heywood and Navarro (2016) finds the LCC has made inner London dramatically safer: it has reduced the number of car crashes, the number of fatalities and the rate at which cars crash (coverage here, ungated, earlier version here). Consider the figure below. Dots show the number of accidents per month. The lighter-colored dots are London and the darker ones the average for the next 20 largest cities in the UK. Both trend downward, but there is a sudden jump downward for London when charging is introduced that isn’t there for the other cities. The oscillations are seasonal effects.
2006: Stockholm Congestion Tax
Stockholm is the largest city in Sweden, with about 940,000 inhabitants in the City proper and 2.4 million in the metro. The City is about as dense as Boston and has about 50% more area — although only one third of the area and people are located within the Congestion Tax area. It is a natural place for downtown pricing, because its downtown is isolated by water.
How it happened
Like the London Congestion Charge, the Stockholm Congestion Tax, enacted on a permanent basis in 2007, is the culmination of a long fight. The Tax’s earliest ancestor was an unrealized proposal floated in local politics throughout the 1980’s: a single pass that you could either use as a transit day pass or put in the windshield of your car to drive downtown. This was not exactly sound in a theoretical sense; there is no reason to expect the costs of operating a transit system to coincide with the delays you cause by driving downtown. But it has an understandable cognitive appeal: “you pay the one price to ride downtown, whatever your mode.”
Though never implemented, the proposal led to thinking about electronic tolls, which were becoming a popular way to pay for infrastructure in Norway at the time. (You can read on Google Books this excellent review of Norwegian urban tolls, which are designed to raise money.) In 1990, Stockholm’s three major political parties negotiated an infrastructure package called the “Dennis Agreement” to get cars out of central Stockholm. To this end, the Agreement involved both transit and road improvements. An electronic cordon toll around Stockholm was supposed to provide much of the funding. In 1997, the Dennis Agreement fell apart for confusing political reasons having to do with conflicts between environmentalists and moderates, but the idea of downtown pricing survived.
After the 2002 round of Swedish elections, the Social Democratic Party needed the Green Party to join a coalition government. The Greens made their support contingent on a trial of downtown pricing in Stockholm. This was awkward for the Social Democrats, since the Social Democrat candidate for Mayor of Stockholm had stated on TV: “My message to the voters of Stockholm is that there will be no road charging during our next term of office, and that is a manifesto pledge on our part.” Consequently, the Social Democrats decided to put some distance between themselves and the policy by promising that the trial would be followed by a referendum on making pricing permanent. Meanwhile, the conservative parties in Sweden formed a group called The Alliance, and made opposition to tolls a party plank.
The trial lasted from January 3, 2006 to July 31st, 2006. Partly due to the Social Democrats breaking their campaign pledge, the public and media started out very hostile to the Tax. Gunnar Söderholm, the head of the congestion charging office, famously called the trial ‘‘the most expensive way ever devised to commit political suicide.’’ But once the charges were established, polls and reporting showed rapid acceptance. There is now a great body of analysis on the question of how and why attitudes changed, but the chief lesson seems to be that people are loss averse: before charges, they were averse to losing free roads; after charges, they were averse to losing fast roads. See Eliasson (2014) and Harsman and Quigley (2011) for good expositions of how opinions turned around.
The referendum took place in September 2006 in the form of a question on the general election ballot. 52 percent of Stockholm residents voted to make the Congestion Tax permanent. At the same time, some towns in Stockholm County where the Tax was especially unpopular held superfluous ballot referenda to demonstrate their dislike, but these measures were not binding nor representative of opinion in the County generally, since only the towns where the Tax was most hated felt motivated to hold their ballot measures.
At the same election, The Alliance won power from the Social-Democrats/Green coalition. Despite having opposed the tolls, The Alliance accepted the referendum result and bargained a massive $12 billion (total, not annual) infrastructure package for Stockholm, whereby revenues from the congestion tax were matched with national funding. Similar to the Dennis Agreement, the money bought roadworks and tunnels to keep cars out of Stockholm. The largest project is the Förbifart Stockholm (Stockholm Bypass), a series of road tunnels west of Stockholm, as shown in the figure below.
The Congestion Tax is a time-variable toll on weekday trips, in both directions, across a cordon around central Stockholm. Prices are the same in both directions. Because Stockholm occupies a sort of archipelago, just 18 toll points are necessary to form a cordon. The cordon is shown below. (I don’t know why they skip numbers in the control points.) Note the green line in the figure; this is the Essinge bypass. Even though the bypass goes through the cordon, until January 1st 2016 there were no tolls to drive on it — only to exit it at a ramp onto city streets. Now you have to pay to drive on it, because the City is growing quickly and needs money for more infrastructure.
You are charged for each crossing of the cordon up to a maximum daily charge, which was 60 kr (~$7.30) at the launch and became 110 kr (~$13.50) on 1/1/2016. Price schedules for Stockholm and Gothenburg appear below.
As in London, the Tax is enforced by ANPR cameras, but these are not scattered throughout the zone like they are in London. Transport Styrelsen, the Swedish Transport Agency, explains how the charging infrastructure works in a figure that is suspiciously reminiscent of IKEA instructions:
The Stockholm Congestion Tax has many fewer exemptions than London. Alternative-fuel vehicles were exempt at first, but not after August 2012. Taxis were only exempt during the trial. Today the only real exemptions are for large buses, motorcycles/mopeds and the “Lidingö exception”: Lidingö is an island to the east of Stockholm (you can see it on the right side of the map above) that can only be accessed from mainland Sweden by passing through the cordon. To make things easy on the islanders, if you drive into the cordon and then come out at the bridge to Lidingö within 30 minutes, then you don’t have to pay the Tax.
As the figure below shows, the Tax effected a sustained fall in cordon crossings of about 20%. Interestingly, the dashed line (which shows measurements take in the months between the end of the trial in Summer 2007 and the return of pricing in September 2007) suggests people drove less even when the tolls were turned off temporarily.
Travel time savings were substantial, as shown in the figure below, which shows proportionate increase in travel time over free-flow travel time. Measurements are from October, so 2005 and 2006 are without charges and 2007 and 2008 are with. “Orbital” routes are outside the cordon for going around Stockholm.
Finally, here is a figure showing how people adapted to the charges. Among commuters who stopped driving in response to the Tax, nearly all switched to transit. Discretionary and “professional” (e.g., deliveries) trips tended to “disappear,” meaning they cancelled trips, combined multiple trips into one trip or went somewhere else than central Stockholm.
Today, the Stockholm Congestion Tax costs about 220 million kr (~$27 million) to operate each year. In 2015 (the most recent year with complete statistics) it raised about 818 million kr (~$100 million). So overall you’re looking at about $80 million per year in profit.
The question of how much it cost to set up the Stockholm Congestion Tax is not cut-and-dry. It cost 1.9 billion kr ($230 million) to run the trial in 2006. However, only 1.05 billion kr (~$120 million) was spent prior to the trial — an amount that included all the “implementation” infrastructure like the gantries, cameras and the computer system. The rest was spent on operation costs during the trial and support.
Implementation costs were higher than hoped for, partly due to political considerations. For one thing, the fact of it being a six month trial meant that everything had to go perfectly with little testing, and the contractor for the job claims it could actually have all been done for about half the cost if it weren’t a trial. Also, the scheme was dealt a costly blow early on:
Midway in the procurement process, the legal status of the congestion charge changed from a “municipal environmental charge” to a state tax (a legal investigation concluded that it was illegal for a city to charge moving vehicles on existing roads). This had many effects, including that the responsibility of the procurement had to be changed from the City of Stockholm to the national government. This increased the cost for establishing the system considerably. Eliasson (2014)
2008, 2012: Milan EcoPass and Area C
Two facts frame the Milan experience.
First, Italian cities commonly operate “limited traffic zones” or ZTL’s. These are central, historic neighborhoods where vehicle access is restricted. A ZTL, for instance, might ban private vehicles, except for the zone’s residents, during the working day. Here is a website with ZTL’s for several Italian cities. Generally, the ZTL’s are enforced by ANPR, so the physical infrastructure needed to charge tolls was already up and running before pricing came to Milan. And, just as importantly I think, the polity already understood the idea of rationing road space.
Second, Milan has bad particulate matter air pollution. The city has among the highest rates of car and motorcycle ownership in Europe, and it sits in the Po Valley — a windless and densely-populated region subject to temperature inversions.
The two Milan pricing schemes have applied to the Cherchia dei Bastioni, a ring of 16th century fortifications around the city center. During his 2002–2007 term, Mayor Gabriele Albertini discussed some form of charging for access to this area. But it was Albertini’s predecessor, Letizia Moratti, who took up the cause in earnest following her election in 2006. An ANPR-enforced daily license called “Ecopass” was implemented January 2008 with the specific goal of curbing access by high-emission vehicles.
While other systems have provided exemptions for a single class of clean vehicles, Ecopass has been the only scheme designed to finely discriminate among vehicles for their emissions, begetting a complicated menu of prices.
Ecopass was also riddled with exemptions and discounts:
There is a 50% rebate for the first 50 entries per year and a 40% rebate for the subsequent 50 entries. There are no rebates for accesses exceeding 100-per-year. Discounts are also available for residents in the tolled area. A number of categories are exempted. These include motorcycles and scooters, public transport, vehicles for handicapped people, army and police (state and local) vehicles, vehicles used for public services, ambulances and, from 10 a.m. to 16 p.m., vehicles transporting exclusively perishable and refrigerated food products, provided a permit is purchased from the municipality. (Rotaris et al, 2010)
As you might expect, the charging structure and liberal exemptions (e.g., vehicles transporting perishable and refrigerated food) totally recomposed traffic toward the exempt vehicles. In 2009, 42% of entering vehicles paid the toll; by 2010 that proportion was down to 15%.
Unfortunately, the effect on air pollution not as large as hoped for. Mattioli et al. (2012) provides an interesting account of how supporters and opponents of Ecopass used and misused air quality statistics. But overall it isn’t surprising that Ecopass’ impact was muted: air pollution is a regional problem, while the Cherchia dei Bastioni has an area of only 8.2 km² (3.2 mi²). Moreover, the pattern of exemptions was also focused on the types and ages of car engines, which are only proxies for actual emissions.
2012: Area C
In 2010, local activists organized a petition drive for several environmental and transportation referenda, including a strengthening of Ecopass. Voting took place in July 2011, and all referenda passed by large margins: support was 79 percent for the Ecopass one. A few weeks later, Mayor Letizia Moratti was unexpectedly replaced by the far left Giuliano Pisapia.
Supported by the referenda, Mayor Pisapia and the new leftist majority on the city council set about to remake Ecopass into a more traditional downtown pricing scheme — one focused on traffic congestion. Ecopass was rebranded as “Area C,” with the C standing for “Centre” and “Congestion.” The redesigned scheme launched on January 26th, 2012.
The new pricing structure appears below. As shown, there is still an emphasis on clean air in the sense that high-emission vehicles are prohibited from the charging zone, while low-emission vehicles are exempt from charge. Wikipedia says that taxis are also exempt but I have not found confirmation of whether this is still true. Residents receive 40 free days of entry per year. Charging is in place 7:30 AM-7:30 PM. Like London, the toll is the same all day long. Apparently there are plans to end the exemption of hybrid and electric vehicles, but the date for doing so has been pushed back several times.
In 2012, operation costs were €7 million and revenues were €20 million, with the €12 million in net revenue doing to public transit as well as pedestrian and bicycle infrastructure.
In July 2012, a surprise court order suspended Area C for 8 weeks. Since the timing and duration of the suspension were not announced, Gibson and Carnovale (2015) use the suspension as a natural experiment and find that Area C reduces entries by 14.5 percent and air pollution by 6 to 17 percent.
2013: Gothenburg Congestion Tax
Gothenburg, Sweden’s second largest city, is an interesting case study of downtown pricing: It is small (~500,000 persons in the urban area) and has relatively low transit usage. In 2012 public transit accounted for 26% of trips among OD pairs involving the Congestion Tax in Gothenburg, whereas the statistic for Stockholm is 77%. This City Lab piece gives a good summary of Börjesson and Kristoffersson (2015), whose results are also in this open-access Power Point.
How it happened
Above, Stockholm inked a deal with the national government of Sweden in which pricing revenues would be matched with national funding. The deal marked a major shift, as most infrastructure in Sweden had previously been funded nationally. In early 2009 the Swedish Transport Administration (in charge of long-term planning, not operation) announced that, for the next decade’s National Transport Plan, priority would be given to projects with local matching funds. Acting quickly, local leaders from Gothenburg negotiated, in the space of a few months, the West Swedish Agreement. The wonderful video below unpacks the Agreement.
The goal of the Agreement is to take traffic off Gothenburg’s streets. The Agreement is worth €3.4 billion ($4 billion), including €2.0 billion for an 8-km rail tunnel under Gothenburg called The West Link, €500 million for public transport, and €400 million for a road tunnel called The Marieholmstunnel. Funding is split evenly between the national government and Gothenburg’s local governments, with the local share mainly coming from downtown pricing.
Unfortunately, because the process was rushed in order to get the Agreement into the National Transport Plan, project selection was not careful. Subsequent analysis showed the West Link — the biggest item in the Agreement — to have a Benefit-Cost ratio of just 0.45.
In September 2013, a non-binding public consultation showed 57% of Gothenburg voters opposed the Tax, but the scheme has been kept in order to fulfill the terms of the infrastructure agreement. Börjesson and Kristoffersson (2015) attribute the unpopularity of the Congestion Tax to the unpopularity of the infrastructure package.
The Gothenburg Congestion Tax launched in January 2013. It essentially copies the design of the Stockholm Congestion Tax: it is enforced by cameras, applies both in-bound and out-bound, and varies throughout the day in manner very similar to Stockholm’s (see the toll schedule already pasted above). The Tax is also turned off during the July holidays. Motorcycles are exempt. Gothenburg also has a somewhat unusual rule: if you pass multiple toll points within 60 minutes, you only pay one toll — the highest one. (As above, this site explains all the details.) Since the CBD is not on a peninsula, Gothenburg has 38 access points — 20 more than Stockholm — and several must be situated in residential neighborhoods.
The figures below show impact on speeds and entries just before and just after the charge was introduced.
Traffic over the cordon has fallen by about 12% throughout the day. It is interesting that the decline was similar in the peak and off-peak, since the tolls are very different at these times. Adaptation strategies resembled those in Stockholm.
The tax raised about 930 million kr ($115 million) in 2015. This is more than Stockholm despite Gothenburg being smaller, since Gothenburg is more auto-dependent. In 2013, operating costs were about 117 million kr ($14 million). This Bachelor’s thesis from U Gothenburg says setup costs were 761 million kr ($93 million).
If you are interested, DM me on Twitter and I will send you a compressed package of gated papers on downtown pricing. This section lists a few Open Access materials you can use to investigate on your own.
Eliasson (2014). The Stockholm congestion charges: an overview.
Harman and Quigley (2010). Political and Public Acceptability of Congestion Pricing: Ideology and Self-Interest in Sweden. RIP Quigley.
Börjesson and Kristoffersson. Effects of charges comparison : Gothenburg vs. Stockholm. (power point)
Börjesson and Kristoffersson (2017). The Swedish congestion charges: ten years on.
Gibson and Carnovale (2015). The effects of road pricing on driver behavior and air pollution.
Mattioli et al (2012). Milan’s pollution charge: sustainable transport and the politics of evidence.
London Congestion Charging Monitoring Program. (Click on “Monitoring and research”).
Green, Heywood and Navarro (2015). Traffic Accidents and the London Congestion Charge.
Transport for London (2007). Central London Congestion Charging Scheme: ex-post evaluation of the quantified impacts of the original scheme.
Chin, Kian-Keong (2010). The Singapore experience: The evolution of technologies, costs and benefits, and lessons learnt.