Phasing in Road Pricing One Electric Vehicle at a Time

The traffic congestion you face is caused by other people. Those people did not think about the delay they imposed on you when they chose to travel. They didn’t even know about you, since they were already on the road before you were. They are ahead of you in the traffic stream.

Similarly as a driver, you, completely oblivious, impose congestion on those who follow. You never met them. You have no opportunity to stop and apologize, or even say excuse me, since that would cause even more congestion.

Congestion is a collective action problem.

Economists have long had a solution to this problem. It is called road pricing. Almost all economists support this in principal. Yet, it is implemented almost nowhere (Singapore is the best example, followed by Stockholm and London, but London does not vary prices by time of day except that it is on during the day and not at night), indicating there must be some problems with the way it has been presented, or the cure (road pricing) is perceived as worse than the disease (congestion).

Problem 1: The collection of revenue. Historically this was at tollbooths, which were sources of delay rather than source of delay reduction, so people would naturally be skeptical that putting tollbooths everywhere would be an improvement. Technology now permits toll collection at full speed, using in-vehicle transponders or license plate recognition.

Problem 2: Administrative costs. Putting a toll collection gantry out on a single facility is one thing. It’s not especially cheap, and must be more expensive than gas taxes.  Putting them everywhere is expensive. Using current electronic toll collection technologies that depend on readers and facility-based collection points does not scale to the system as a whole. Localized toll collection cannot in general solve the widespread congestion problem.

Problem 3 : Privacy and tracking. Surely the government will be monitoring whatever transponder or GPS device they put in the car. I have seen Law and Order as much as the next person, and I know what the police and prosecutors already do with EZ-Pass. Even if there are technical solutions (using a pre-paid unregistered cash card rather than a credit card) no-one will believe that the authorities aren’t tracking. The entire NSA scandal just makes people suspicious. While in my view privacy is mostly dead (and of your own doing so long as you carry a communications device with you or pay with credit cards), it is even deader on public roads, even without road pricing (since we have cameras, police have cameras, traffic managers have cameras, and sousveillance is everywhere). But people are still nervous, and we need to recognize that.

Problem 4: Implementability. Rolling this out and turning on the switch is a big shock to the system. Transportation is inherently a conservative field, people are comfortable with slow change. So deploying 200 million transponder devices and millions of readers across the network before turning them on was at best a foreboding task, and in all likelihood terribly unwise. What if it didn’t work properly?

Problem 5: Fairness. Tolling people is often perceived as unfair (which usually does not take in to account the distributional inequities in the existing road financing system). Everyone has the same amount of time, but rich people have more money. There are lots of solutions to this problem, but in the end, the fear is at least some individuals would be better off without the change.


Opportunity 1: Electric Vehicles are coming. In some sense they are already here. While their market share of hybrids is still low (still less than 3% of all new sales) and Battery and Plug-in EVs (still less than 1% of new sales), the latter category is growing rapidly.


US Electric Vehicle Market Shares
US Electric Vehicle Market Shares

Now extrapolation is dangerous, but we do have claims from some of those in the EV industry, namely Elon Musk of Tesla about achieving market share of about 13% by 2020. Further we have the history of technologies which show an S-shaped life-cycle dynamic. The tricky part is determining the ultimate market share (which I will assume to be 100%), and the rate of growth. Existing data allows us to estimate the rate of growth. Combining Hybrids and EVs, Figure 2 shows the best fit logistic (life-cycle) curve. A market share of 50% of new vehicles sold is achieved in 2022 or so. This is 8 years away. Eight years is a long time. Eight years ago there were no iPhones or Androids.



US Electric Vehicle Market Shares Extrapolated
US Electric Vehicle Market Shares Extrapolated


The main constraints have been limited consumer demand due to range anxiety and issues of charging location and speed. So we need to assume (1) The cars will get better over time, (2) Batteries will get better over time, (3) Electricity will get cheaper over time.

I believe all three of these are certain, the only question is the speed with which these things occur, and the degree to which batteries get better.

The cars will become better. Already Tesla produces the best car (Model S) in the US according to Consumers Reports. It is of course pricey. On the other hand, you need to discount the price some because you will not need gas ($3/gallon at 15000 miles per year at 30 mpg, which is about $1500, or $15000 over the life of the car). The price will also drop with true mass production.

You can’t beat free: Many have understood for a while (see this 2007 post e.g., and this from earlier in 2014) the solar cost curve is bending and will become cheaper than alternative sources of energy soon.

Soon is basically here. My dad in Arizona has solar panels. There is a house on my commute with solar panels.

Solar production is growing at about 41% per year.  As demand increases, even more resources to improve the technology and drive the cost downward in a magic bullet fashion will arrive.

Solar energy panels do have a fixed cost, but the variable cost per unit of electricity drops to approximately zero. This means you are replacing the cost of gasoline with about nothing, if you have solar panels on your roof generating more electricity than you would otherwise use. There is the alternative of selling the excess back to the grid, but one imagines once everyone starts doing this, the grid isn’t going to pay much, if anything for excess power. We have heard “Too Cheap to Meter” before, about Nuclear. Unfortunately we did not implement that successfully. Solar is a much more grassroots rather than top-down process, and more likely to succeed.

The difficulty is energy storage. Batteries are getting better, doubling energy density about every 10 years (or 20) – which is of course a big difference. So even if solar is cheaper than the grid, the sun isn’t always on (you know, the rotation of the earth etc.), so batteries are required at home as well in the car.

But we don’t need batteries to store a year’s worth of energy, we need them to store enough to be competitive with cars, i.e. to be good enough and cheaper, so that they can either be charged fast or swapped out fast. MP3s don’t have the fidelity of analog music, but they were good enough. Cell phones don’t have the sound quality of land lines, but they were good enough.


From a transportation funding perspective, the most important implication is that EVs don’t pay gas taxes. If they become widespread, there will be a not just  the slow decline of gas tax revenue we see already due to peak travel and better fuel economy, but an actual crash.


Opportunity 2: Congestion remains a problem

Question: If pizza were free, how much pizza would be available at dinner time in the dorm?

Answer: None.

Question: So when roads appear free, how much surplus road space do you have during rush hour?

Answer: None.

Congestion should not be a surprise, it is what you get when you underprice a good. While it is not getting especially worse in most of the US, it is not getting especially better either. Time is still money, and this problem will remain until we actively do something about.

Opportunity 3: Still roads require some funding.

Roads don’t plow themselves. Roads and bridges don’t repair themselves. Roads don’t repave themselves. Bridges don’t erect themselves. The money for these things must come from somewhere, and people (and their machines) must be paid to do these things. The best source for these funds are the people who directly benefit from the existence of these public works – the users themselves. Our system in the US is a combination of funding from users directly, and non-user beneficiaries, as well as the general public (which usually fall into the first two categories).

Opportunity 4: Traffic is self-organizing.

While the theoretically perfect, first best, solution would charge a unique price for each link for each time of day, that is far more detail than we actually need to have an effective system. We trade-off between the additional efficiency from more time and place specificity against the additional administrative complexity and decrease consumer acceptance from such a fine-grained system. Most priced systems are much simpler than the ideal because of these practical concerns.

Fortunately, to a first-order approximation, we don’t really need to know which road people are traveling on, just the time.  Wardrop’s Principle of User Equilibrium (not strictly true, but good enough for the moment) says all used routes have the same travel time. Which is to say, when traveling between A and B at a given time, if there are multiple routes you might use, their travel times are equal, and if one is higher, you won’t use it. And this holds for everyone. Traffic spreads out in a regular way to exploit available routes. So while there might be some advantages to tolling one route more than another (because their marginal costs differ), that introduces a lot of complexity for a relatively small system-wide gain. My estimates of the spatial Price of Anarchy on the Twin Cities network is that there is only a small loss (less than 2%) due to letting people route themselves rather than the Central Planner allocation. In short, the main problem is temporal (peaking) rather than spatial (routing).


Taxi-Meter: We want to keep the structure as simple as possible. Imagine an in-vehicle taxi-meter, with a per-minute charge. We can have as many different rates as we want, but we should start with a few (more than zero, otherwise it is not going to affect time of day people travel at all, more than one  if you want to avoid too much boundary effect of people not leaving until the rate changes. I suggest three different prices for starters. Once people get used to the idea, the rates can be adjusted. It is much easier to go from 3 rates to 4 or 6 than to go from zero rates to 1 or 2.

Rate Structure: For instance, imagine a price structure like this:

  • Peak 6 hours per day (~50% of current traffic) each traveler pays  [T]
  • Shoulder of peak discount (~25% of traffic) (50% discount) each travelers pays  [0.5T]
  • Offpeak discount (~25% of traffic) (90% discount) each traveler pays [0.1T]

Second, we establish the base rate for the peak times, and everything else is a discount (think about movie theaters and restaurants, which have the matinee and early bird special) rather than having an unwelcome “surge” pricing phenomenon. This is I think a more positive framing. Also since the non-peak rates are lower, the fraction can remain fixed, and there is only one base toll rate for policy to regularly adjust, and then a fraction of that rate associated with day-parts, which is adjusted less frequently.

Note, we are only tracking when you travel, not where you travel, and perhaps your residence (since rates will vary by jurisdiction of residence).

System Members: We want this to be as fair as possible. Fairness means lots of things to lots of people. However having rich people pay more rather than poor people pay more is a fairer way to start. At this stage of history, early adopters like people buying brand new EVs undoubtedly have above average incomes (though I don’t have actual statistics to verify this). Making everyone pay for roads, instead of just people with gasoline powered cars, is also fairer.

We want to phase this in to avoid a big-bang implementation disaster (like the botched roll-out of Obamacare). Fortunately for this system, most people don’t have EVs now. Also fortunately, we anticipate many people will in the coming decades.

So I suggest the membership in this system should be automatic (starting the model year after next)  for all new EVs, Hybrid EVs, other Alternative Fuel Vehicles sold. All such vehicles would get rebate on general tolls, local property taxes, other general revenue sources of road funding, as well as any gas tax paid as well (such as for Hybrids). This gives the automakers more than a year to implement the device into a small fraction of their cars. It ensure bugs and difficulties are discovered early and inconvenience only a small portion of the population. It gives the federal government a year to set up a revenue collection system that can ramp up over time to a larger share of the fleet, and one imagines, eventually to the entire fleet, either as EVs and other Alternative Fuel Vehicles come to dominate, or as it is imposed at some point on all new cars.

As more and more vehicles become non-gas powered, this system membership grows and it becomes more and more effective.

Opt-ins: Gasoline powered cars can voluntarily opt-in to this system, which for many travelers would be a cost savings and provide incentives that might be easily exploited to the betterment of all. We could further allow an opt-in location tracking, which would give a discount in exchange for rates which varied locally.

Surplus: If there is a surplus at the end of the year, above  the members’ share for the cost of roads and rebating for other taxes, every member of the system gets a dividend. A check in the mail, that they can use for whatever they want.

An Illustration

Currently 1 hour of travel at 30 mpg and 30 mph uses 1 gallon of gas, which is about $0.50 of state+federal gas tax [depending on where you are].

Note: This should be about $1.00 to $1.50 to cover the cost of all road infrastructure (not including externalities), depending on how you count. Other taxes cover a large share of road expenses, including property taxes, vehicle sales taxes, and so on.

T  = $2.50/hour (= $0.0417/minute)

[cf Cost per minute]

Assuming share of travel did not change by daypart, average revenue per vehicle hour would be about:


We would expect that share of travel would change by daypart, so that the average revenue per vehicle would be lower, and more in-line with system costs. Actual elasticity of demand with respect to the toll rates is an empirical question that can only be firmly established with experience, though we can make some estimates.


This extends the idea from my post: Road Pricing When? Old Fleet, New Fleet.