Friday 6 September 2019

National League of Cities report gets congestion pricing wrong

"The views contained in the opinion piece below are solely my own and should not in any way be attributed to my employer, D’Artagnan Consulting LLP"

As I said a few weeks ago, the United States has suddenly "discovered" congestion pricing over a decade after Europe and over two decades after Singapore, in part because it is actually going to happen in New York.  This is good news.

At last, there is policy debate about using pricing to address road congestion, rather than the debate about simply building more road capacity or building more capacity for alternatives. 

The not so good news is that some writing about the topic appears a bit rushed and this enthusiasm to produce research and analysis can sometimes mean that there are gaps and inaccuracies which come out. One of the latest efforts is a report published by the National League of Cities called "Making Space: Congestion Pricing in Cities" (PDF).  It's not as good as it could have been, and could have done with a bit more time and some research (e.g. simply going to sources on this blog) to better inform cities considering congestion pricing.

Let's be clear, I am very supportive of anyone seeking to learn more and to think more about, I fear that a rush to generate "knowledge" could actually hinder progress in pricing, because it may oversimplify concepts, or worse yet disseminate misinterpretations of experience and concepts that might encourage opposition to road pricing more generally.  For example, to talk of congestion pricing primarily in terms of generating new revenue can raise fear that it is just a new form of tax, rather than a tool that addressing a major urban transport policy problem, which can also generate revenue (which itself could be used to reduce other sources of revenue).

I've read the report end to end. Overall it has some useful points and is well structured, and I hope it generates interest in congestion pricing as a concept, but I don't think the report is necessarily as helpful as it could be for decision makers, policy advisors or advocates for better pricing of road use, because parts of it are just not correct, and some useful details are left out.  For example, the number one reason congestion pricing programmes across the world have not progressed (and quite a few have tried) is because of a lack of public and political acceptability.  Ignore this at your peril.

The main flaws of the report are that it:

• does not adequately define congestion pricing;

• does not describe all of the behavioural changes that congestion pricing can encourage;

• does not describe the technology used in ANY operating congestion pricing systems today (but depicts technology that is not used);

• misconstrues the purpose of the London scheme, doesn't note the significant limitations of the scheme today;

• incorrectly reports the use of revenue collected from the Singapore ALS and ERP schemes, and does not accurately describe how it works;

• takes a narrow view of equity in congestion pricing, essentially claiming that the sole mitigation is additional modal choice;

• concludes with a new (and tangential) point about electric and automated vehicles, giving the impression that the imperative to consider congestion pricing is about such vehicles, not congestion more generally.

Below is a thorough review of the report.

Detailed review of the report


It's important to get a good definition of congestion pricing for those who want clarity in explaining what it is.

Unfortunately, the first "definition" on page 3 is not particularly helpful as it start with road user charging.

"Road user charge systems require drivers to pay based on distance driven and, perhaps other costs of road use, such as wear and tear on roads, traffic congestion, and air pollution.”

This might confuse the efforts of some states to test distance based road charging to replace gas tax (as revenue to pay for road infrastructure) into congestion pricing, but they are different concepts, with different objectives, using different technologies.  It is important to separate out charging for infrastructure cost recovery and charging to manage demand, although these can be combined (as they are in so many other sectors of the economy).

This definition is cited as originating from Congressional Research Service. “Mileage-Based Road User Charges.” 2016. 

This is all very well, in context.  However, it is not necessarily appropriate for congestion pricing. Road user charging is applied to a network not a single road (a key difference from tolling). Road user charging may have charges that vary by vehicle type, configuration, time of day, location and environmental factors, which together are meant to represent infrastructure costs (which are more than wear and tear, as a significant proportion of road wear and tear is unrelated to road use).

You see congestion pricing isn't about charging vehicles to pay for network infrastructure costs.  The report implicitly acknowledges this by continuing with another definition treating congestion pricing as a subset of road user charging.

"Congestion pricing is a type of road user charge system in which a flat or variable rate fee is charged to vehicles that drive in a specified area or zone within a city." (pg 3)

This definition is fundamentally flawed because it ignores the crucial element, time of day.

Congestion pricing does not charge vehicles 24/7, it charges only at times of peak demand (otherwise it isn't charging to reduce congestion, it becomes a toll).

To understand the importance of flexibility in pricing policy, congestion pricing isn't just about vehicles driving "in" a specified area (an area charge), but for crossing into (and perhaps out) of an area. 

It also need not just be an area, it could be a corridor (much of Singapore's network is corridor charging). So, this definition is both too wide (ignoring time of day) and too narrow (ignoring different types of charges). To make it a subset of road user charging (which is defined previously as distance charging), is mistaken, as congestion pricing need not have any distance element at all. It continues...

"With variable pricing, the goal is for congestion charges to rise in accordance with increased traffic congestion, thereby pushing some drivers off the road and making traffic flow more smoothly" (pg 3)

Well I understand the point being made, but it gives a very negative impression to say that charging rises according with increased congestion, but rather increased demand.

The key idea being that a driver pays for a more consistent level of service, so congestion does not rise, but rather demand does, so pricing does to ensure congestion remains stable.

It's also a bit coarse to say it would "push drivers off the road" (and antagonistic to drivers), but rather pricing ought to encourage a variety of changes in behaviour, which not only includes not driving, but also driving at different times, different route (if possible) or driving less frequently.

The next big mistake is on technology, where this report makes a big blunder and just gets it plain wrong.

"These programs rely on tracking drivers either through manual odometer readings or onboard devices." (pg 3)

No. They do NOT rely on "tracking drivers" at all.  Besides the subtle point that it is vehicles (not drivers) that are detected, there are no congestion pricing schemes in operation that "track" trips.  Because there is no need to do so for many types of pricing.  All current systems do is detect a vehicle crossing a specific point at a time when it is charged.  

Equally importantly, the technology reported here is simply wrong.

NO congestion pricing schemes anywhere use manual odometer readings. How could they? What value is an odometer reading in identifying where or when a vehicle travelled?  Why would an odometer help?

Furthermore, of current operating schemes, only Singapore's use onboard devices (which are only used to deduct a prepaid stored value smartcard to protect privacy). The key technology used is number plate recognition. That isn't mentioned. Yes, Singapore is planning to replace its devices with new GNSS based one that could track movements more generally, but again that isn't even mentioned in the Singapore case study.  

So let's look at the case studies.

London (pg. 6-7)

London is an important case study and much of the information is reasonably accurate, but to claim that London took a "divergent path" to New York in funding its "dated transportation infrastructure" inferred that the congestion charge was a central part of transport funding.

However it wasn't, and it isn't, as it was not set up primarily to raise revenue. This is a point admitted at the end of the case study when it states that it only raises 8.5% of Transport for London's revenue (although in the 2018-2019 annual report TfL's revenues show gross revenue of £229.9m from the congestion charge out of total revenues of £5.6562b, that's 4%.)  Unlike New York, the congestion charge in London was not about revenue.

The report continues to say...

"The results so far appear to be extremely positive"

"Extremely positive" is a bit of a hyperbole, but it is fair to say it was positive for some years, but nobody would claim that it made a dramatic difference to London traffic, although it eased congestion in the centre, most of London remained and remains heavily congested.

Most notably, the London Assembly report in 2016 noted that the Congestion Charge is no longer fit for purpose. There has been an 11% increase in light delivery traffic in the three years before that report, and a 70% increase in private hire vehicle (including Uber) traffic as well. The claim that air quality is improving is based on an 11 year old report. Since then, a previous UK government policy of encouraging diesel vehicle ownership has reversed some of those gains, and the issue of air quality has been a big issue for the current Mayor.

There are minor errors, for example taxis (known as black cabs) are still exempt, but minicabs are not (a distinction I don't necessarily expect US researchers to understand). The report hardly mentions the Western extension of the scheme, and why it failed, which is an object lesson in communications and politics - basically those who had to pay didn't think they got value from it, and businesses within the zone believed they were harmed by it.

The report says that the lesson learned is that a plan must be adaptable and flexible, yet there is not much evidence that the London Congestion Charge has been that flexible. Yes, the rate has increased somewhat and the discount/exemption categories have been tightened, but as an area charge it is fundamentally inflexible. A single charge applies to drive into and within the zone for any amount of travel between 0700-1800 weekdays. A single crossing by a car costs as much as a truck circulating for over a hundred miles within the zone. Indeed, the removal of the private hire vehicle exemption is estimated to make a 1% difference to traffic levels, because a vehicle could enter and exit the zone all day, and could spread the cost of the charge over multiple paying trips.

Yes, traffic in central London is down historically, but it has fallen by one-third since 1986 for multiple reasons, so it cannot be simply attributed to the congestion charge (it could be said that it has helped accelerate that trend of course).  Only a small proportion of trips into central London are by car (<15%), because London has grown over a railway and underground network (and complementary bus network), most of which was built a century or more ago.  Finally, London has always used automatic number plate recognition technology, which isn't mentioned.


The case study on Stockholm is largely accurate, although the claim that it is different from others because it has charges that vary by time of day, is only true if you don't mention Singapore and Gothenburg. Of course, Stockholm uses number plate recognition technology, like London too.


Singapore's long standing congestion pricing system from the ALS in 1975 to ERP since 1998 was always about an integrated strategy to discourage car use and encourage use of public transport. However, it is simply untrue to say it was introduced to improve public transport.  It was introduced to reduce congestion, and this supported increasing public transport patronage and the efficiency of the bus system. Yet the implication is that the revenue from the Singapore ERP system is dedicated for any transport spending, which is not true.  It is not used to fund public transport, it is simply general government revenue. It might be worth noting that Singapore doesn't subsidise the operations of its public transport network, with fares covering rail and bus operating costs.

"Rates vary from $0 - $3 depending on the time of day, the road type and local traffic conditions. The ability of the system to respond in real-time" (pg.10)

This is only partially accurate. Yes rates vary depending on time of day, road type and conditions, but they also vary by location and by vehicle type. Motorcycles pay half the price of cars, and buses and trucks pay multiples of 2 or 3 times the rate of cars. Furthermore, the system does NOT respond in real time. Prices are reviewed and varied every three months, and may increase or decrease, or change in time of operation, according to the specific local conditions of each charging point. This sophistication is key to understanding why Singapore is the most efficient and effective system in operation, but it is not real time pricing.  That doesn't exist outside toll lanes seen in some States.

New York

The good news is that this part of the report is largely fair and it is welcome that it noted the risk of exemptions undermining the effectiveness of a scheme. It is perhaps worth adding that exemptions provide an opportunity for fraud as well.

Equity and Congestion Pricing

The report starts by claiming...

"A key question they should ask themselves is: does our current transportation system enable residents from all economic backgrounds to access jobs and city amenities without cars?" (p.16)

No, they shouldn't ask that question, because this would be pretty much impossible. That's because the questions of location and time of day need to be carefully considered.

Congestion pricing doesn't require all residents to abandon driving to jobs or other services.  For a start it wouldn't be imposed on all roads for all jobs at all times (although bearing in mind the definition at the start of the report didn't mention time of day, you might be forgiven for ignoring this).  Pricing should only be introduced at locations and at times when there is severe congestion, this is not the case for many trips.  The assumption is that people need to be priced out of driving to work for congestion pricing to succeed, when there need not be a significant shift in vehicle numbers to make a meaningful impact on congestion.

A reduction at peak times of between 5-15% of trips should significantly ease delays, and some trips at peaks are discretionary, so it is expected that these may be priced off the network at that specific time. What is needed is data about what trip purposes and origins/destinations are to understand what the road use profile is. Bear in mind that some commuters may travel earlier or later, to pay less or not at all.

It is not a binary choice. When congestion pricing is introduced, most of those who drive will still do so at peak times, but will be paying and saving time and fuel. Those who change behaviour may change modes, they may change travel times, or they may choose to not travel at all (consolidating multiple discretionary trips into fewer trips). It is critical to understand that choice is not "drive or use public transport", it is drive as you did before, drive later or earlier, drive less frequently, drive on a different route or use public transport (or active modes).

Discussion of equity ought to consider the work done by the OECD, and consider analysis of congestion pricing compared to alternatives, not just for charging for road use (including parking charges), but the cost of congestion both in time, fuel and emissions. Congestion pricing needs to take into account its effect on people accessing employment and essential social services, but to minimise negative impacts it needs to be designed to target congestion when and where it happens.  Just implementing blunt zones is not the way to do this.


There are some useful conclusions here, but there is a tangent here that really wasn't explored earlier in the report....

"exploring congestion pricing may become more important as we move toward a future where both electric and self-driving vehicles are ubiquitous on city streets." (p.18)

Why? Why would motive power matter?  It is MORE important whilst most vehicles emit noxious fumes, and the issue of congestion is distinctively different to motive power and automation. The electric vehicle point is not explained at all.

"Once Uber and others roll out autonomous vehicle fleets, calling a car will be cheaper and more competitive" (p.18)

Notwithstanding the hype, the reality is that this is many years, possibly decades away. Why will it be cheaper (and cheaper than what)? Is it assumed the labour cost savings more than offsets the capital costs of autonomous vehicles? The assumptions around this are considerable and worthy of a separate discussion.

"Furthermore, when not in use, autonomous vehicles need to go somewhere. There are three options: go back home, park somewhere or circle around. Most likely, these cars will endlessly circle the streets rather than parking and paying fees." (p.18)

Currently when not in use, vehicles need to go somewhere.  However, why assume that autonomous vehicles will circle endlessly? Will the cost of energy be so cheap (in this case likely electricity) that this would be cheaper than parking (and charging)?  This is a big assumption largely based on conjecture.

More importantly, this is a diversion from the key point. Congestion is an issue now, it is likely that on-demand transport services have exacerbated this, and to claim that cities ought to wait until electric or even autonomous vehicles become commonplace is simply wrong. What are the objectives here?

Fortunately, the report returns to some focus...

"Congestion pricing can serve as a market-based regulator that gets the right number of cars on the street at a given time." (p.18)

YES that is exactly right (but also taxis and trucks at least).

"Local leaders resoundingly want cities for people, not cars."

Maybe so, but people use cars too. Yes, it is obvious that in almost all major cities there is never enough capacity to accommodate private motoring if the price to access roads (a scarce resource) is too low. However, the future is likely to be more choice. Car use will remain valid for many people for many trips, but they will change when and where they drive. Some may use other modes more, particularly walking and cycling for short trips (both of which are often underappreciated by some politicians who see more publicity in opening new fixed infrastructure for public transport or cars).

My conclusion

The National League of Cities should be congratulated for being interested in congestion pricing, but this report doesn't quite grasp what congestion pricing is, how it can be implemented and most importantly, the biggest challenges in doing so. The report doesn't consider the cities that have tried but failed to introduce congestion pricing. The UK, for example, had a rush of activity by some cities to study and proceed with congestion pricing programmes after London's apparent success, but all of them failed. Edinburgh and Manchester are notable examples from which there are lessons to be learned. Copenhagen also failed to proceed, and is often touted as a role model as a sustainable city.

Most US cities are very different from the cities that have implemented congestion pricing to date, but it can still learn from them and from the vast amount of experience obtained from other cities that have been studying congestion pricing. They would be well served by research that is more in-depth, accurate and appears less rushed.

No comments:

Post a Comment