What's the true incremental cost of driving a car?

Taxis wear out by the mile, not by the year

What is the incremental cost of operating a car? It's not very well understood and here I hope to, together with readers, come to some better understanding of it.

Many sites calculate the "total cost of ownership" for cars. You can go to sites like Edmunds True Cost to Own or Kelly Blue Book Total Cost and they will calculate the total cost of owning a car for its first five years, not including parking. The cost drops as the car gets older, and these sites use average figures, so if you are a good driver, your insurance will probably be less. They tend to base it on driving the car 15,000 miles per year, though the average is only 12,000. (Newer cars, however, tend to get driven more in the early part of their life and less later on.)

You may also like the AAA cost analysis for different classes of cars, which calculated a cost of 59.2 cents for the average sedan, and much more for an SUV.

But it's tricky. While you'll spend that much per mile on owning a car, that's not the additional cost of driving an additional mile -- or operating an additional hour.

The additional mile adds the following costs:

  • Your car depreciates -- it loses resale value as you add more miles to it. Around 11-20 cents per mile.
  • Your maintenance schedule is based on miles, so you speed that up, at about 6 cents per mile. Your tires wear out at about 1 cent/mile depending on what type you have.
  • Most insurance companies now offer different prices based on the miles you drive. While you may not be making use of that, liability and collision insurance really depend on miles, since you don't crash if you're not driving. Typical estimates are about 6 cents/mile. Metromile, a semi-pure "pay as you drive" company, advertises a price of $29/month plus 6 cents/mile.
  • Of course you burn gasoline or electricity as you drive. Anywhere from 4 cents in an electric car to 8 cents in a Prius to 22 cents in a guzzling premium SUV. Or more.
  • The more you wear out your car, the more repairs that get needed. Most repairs are under warranty in the first few years, but many are not, and none are after the warranty. A repair is anything you need to fix outside of schedule maintenance. Not counting any car accidents, which will add to the cost. Some things, like scratches and chips, are part of expected depreciation, but things beyond normal wear and tear cost you extra. Your interior is wearing out every hour you sit in it, as well.
  • As noted, you pay for parking and tolls the more you travel, but that's wildly variable.
  • Certain behaviours, like fast acceleration can put extra wear on cars, and on batteries. High speed charging puts extra wear on batteries. A highly degraded battery will reduce the value of an electric car far more than would be predicted from its mileage.

While the whole car pricing industry works on miles, because that's what we know, there's a strong argument that a lot of value loss should be measured in hours. Other vehicles, like boats and planes, are all aged in engine hours. This makes sense -- your car is actually wearing out a lot less going 60mph on the freeway than it does driving stop and go traffic in the city. It even uses less fuel per mile, if it's a gasoline car. The national average speed is around 40mph, which can be used as a very rough translation from miles to hours.

Your car is also costing you money just sitting there. It's depreciating every month, even if you don't drive it. There is a base price for insurance (including comprehensive insurance) that does not change with miles. You paid taxes when you got it and you pay a registration fee every year of a few hundred dollars. You've tied up money in the car and thus that money has a cost, even if you didn't get a car loan.

Yet you don't buy your car just to have it sit around. Every car buyer buys their car expecting it to serve them some base amount of miles. Their commute for example, and a few other things. It's only worth looking at incremental miles if you have a trip that definitely wasn't in the plan for the car, or you're comparing the cost of a drive to an alternative. When planning whether to buy a car or not, or looking at the economics of a business, you need to price miles to cover all the costs, even the monthly ones. When looking at a specific extra trip you didn't plan when you bought the car, you can look at the incremental cost.

There have been various attempts to calculate the incremental cost of miles. Barnes and Longworthy in 2004 calculated 15.3 cents/mile for cars, and 19.2 cents/SUVs. This did not include insurance or parking and was at $1.50/gallon fuel. They bumped the number to 19.1 cents for cars and 23.1 cents for urban driving. They added extra for poor pavement and rough driving. Professor Mallard-Ball (with whom I differed on whether robocars would circle around rather than park increased their estimate to 31 cents/mile for urban, but felt an electric car would be only 21 cents/mile.

Electric cars should cost less per mile to operate, and eventually to own. (Some would argue they already have gotten cheaper to own.) This depends on what they actually cost to maintain, and how much their batteries actually degrade and whether they can be recycled at higher value, among other things.

Mallard-Ball's 21 cents plus insurance and parking, or 27 cents/mile plus parking, might be taken as an experimental number. Or 37 cents for the gasoline sedan.

There are arguments for lower numbers. As I noted, not all insurance companies vary the cost greatly with miles driven, and none yet charge purely by the mile even though this makes sense. The price of gasoline and electricity varies a lot. Repair and maintenance costs for more modern electric vehicles may be much lower -- there are arguments that they will be, but data is still limited. Battery technology is also getting better.

As such, the web sites above might do well expressing car ownership in terms of a cost per month plus a cost per mile. If we look at AAA's 59 cents/mile for the average sedan, and it drives 15K miles/year, they are expressing the car ownership and operation as about $280 per month plus 37 cents/mile. The real cost per month would decline on a decay curve until the end of life for the vehicle, and the depreciation portion would also decline, but energy, insurance and maintenance would remain stable while repairs would increase -- in fact the end of life of the vehicle is generally the time when repair costs rise faster than depreciation slows, though it is also when the vehicle simply looks unacceptable or is too unreliable. There is also a new factor for robocars -- when their technology becomes obsolete and non-competitive, even if it still functions well.

This analysis alters some of my earlier examinations on the economics of some things, such as whether people will want to hire out their private robocars through taxi services. (Think UberX but you don't have to drive it.) It makes it more likely if those miles can be viewed as purely incremental. Of course, hiring out your car does more than wear it out, it also makes it unavailable to you, requiring you to either face inconvenience or hire back a robotaxi. (I suspect robotaxi companies that use private cars will offer you lower-cost robotaxi if they can't get your car back for you when you need it, possibly at a wash price. I still suspect that most use of private cars by robotaxi services will be during premium "surge" times when the rate they are willing to pay goes up. More examination of this math is needed.

Not so when running a professional taxi fleet. You must recover all costs, so your base or average price must cover that 59 cents/mile (or whatever the total cost divided by the total miles is.) You can charge some customers less if you charge other customers more, but unlike a personal car, a taxi wears out almost entirely with use. A typical taxi will drive 60,000-100,000 miles per year and be fully worn out in under 5 years with current lifetime engineering. Leaving your taxi idle is relatively inexpensive. This means there may be little point to using it at a lower rate off-peak compared to preserving it to serve more hours on-peak.

Comments

This is a key figure. I think a company has two choices here, but that it can't get the best of both worlds. The first choice is to run things like a typical NYC taxi. That is, get 60,000-100,000 miles per year for the average vehicle, and have little idle time during the week day and weekend night. This would require charging enough money so that people don't switch en masse to using robotaxis for commuting. It will also likely require a lot of "dead miles" where the car is driving from a place where people are arriving to a place where people are departing.

The other alternative is do things very differently from today's taxi companies. That is, not to worry about idle time, but in turn not get anywhere near the number of miles per year of the other scenario. This will probably be the way to go for places where parking costs are low. Unlike a typical taxi company, there's no driver to pay by the hour for sitting around doing nothing.

This second alternative will also facilitate surge pricing, where surges aren't necessarily the peak demand times, but are times when the location of the demand doesn't match the location where the cars are sitting idle. The surge price will pay for the extra miles involved, and if the surge is big enough might also pay for certain guarantees the company could offer to regular users ("we'll pick you up within 5 minutes or your next three rides are free").

I still don't see this being very attractive to commuters. Not in areas where they can park for free both at home and at work.

I think we'll see a lot less surge pricing. You are right that respositioning will have an effect on this. A company eager for business will predictively position vehicles to be able to offer competitive wait times, but they could elect to not do so if rates drop. Or rather, add a charge to move the vehicle.

But I don't think customers will like, "There are no vehicles near you, you have to wait 15 minutes and pay $4 for us to move the vehicle to you."

Generally, though, surge pricing should not be a big thing. You have your fleet. You can't "get more vehicles on the street" by raising the prices, as Uber does. If a vehicle is sitting idle, and there is a customer ready to pay the regular price, there is no reason not to serve the customer unless, as you say, they are far away. (Some vehicles will be out of service due to low charge in off-peak hours.)

Though this is where using private cars could come in. If your fleet is too small for demand (this is only at peak) you can either pass business to a partner/competitor, or you can recruit a private car. You might need to 'surge' to entice the private car.

But customers hate the unpredictability of surge prices and may stop using you if you do that to them. So more likely you eat the extra cost of the private car to meet promises to customers.

I shouldn't have called it surge pricing. Uber shouldn't have done it either. No, customers wouldn't like, and companies won't say, "you have to pay $4 extra for us to move the vehicle to you." I think companies will adjust prices according to supply and demand, though, and those extra fees during peak demand will pay for things like the extra moving around that has to be done when the surplus of cars standing by waiting is lower.

And no, it won't be that a single car will drive 30 miles to serve a passenger. But what will happen is more like this: If the company tries to have 5-10 cars on standby at the casino, and suddenly finds itself with only 2, it might move some cars from their standby positions at the airport, and might shift some cars from their standby positions near the office park in order to cover the airport. In times of extreme demand that move might cascade far away from the original surge location. I'm reminded of the shifting that fire departments did on September 11, 2001, where stations 100 miles away from Manhattan found themselves covering stations slightly closer to ground zero, as the surge in demand cascaded all the way to them.

You suggest yourself that there will be a lower charge during off-peak hours. Don't call it "surge pricing," certainly not to the customers, if that scares them. But fees are going to go up and down depending on demand.

Customers who want to lock in prices ahead of time can, of course, make reservations. Know you're going to need a car at the Chicago airport when your flight arrives? Let the company know your flight number, and they can give you a discount for making their job at guessing demand easier.

(And no, I don't mean that there will be a specific car earmarked for you. But maybe they'll keep 6 cars on standby at the airport, when they see your flight has arrived, instead of 5.)

Yes, every serious fleet manager has been work for some time on their algorithms to predictively reposition vehicles. I started initial investigations to that at Waymo 7 years ago, I have no idea where they are now. There have also been several academic simulations of how many empty vehicle miles are needed to run a fleet, based on either real world data (if you're Uber or have other sources) or the NHTS or other academic data sources. I have seen estimates as low as 5%, I expect a bit more in the 10-15% range. It depends how many vehicles you want to put in your fleet, and the cost of having them idle. (Which is parking and capital, both small, and obsolescence which is larger.)

In New York, Taxis do 38% of their miles with no passenger, but they do a lot of wasteful cruising.

It also depends a lot on how people use the service. If you guess it's going to be (or use real world data based on) similar to current usage patterns, you'll get one answer, but if you guess that the usage patterns are going to be drastically different (such as you'll have if people start giving up their personal vehicles), you'll get a very different answer.

And one good thing about surge pricing (but don't call it that!) is that it gives a free market incentive for people to use the system in the ways that are most efficient for that equation.

Unfortunately, I'm not sure there's a lot of useful simulation that can be done there, as it's a very complicated interaction, and also will rely on a lot of factors that we can't really know right now. But I think it's safe to say that altering price based on current conditions will improve efficiency, and that it's even safer to say that doing so for that reason and others (price discrimination) will increase profits.

There might be variable prices, but I think they would be much less common than for things like Uber where human labour is the scarce resource.

The question is, will cars be scarce or plentiful? Will there be a shortage of cars that are close enough to serve you, and do you need to pay extra money to be one of those allocated a vehicle during the shortage?

There may be different answers depending on whether there is competition in your area, and how much competition there is. If there is a local monopoly, a shortage with higher prices is more probable.

With competition, there is an incentive for fleets to grow to make shortages rare. That's because it does not cost that much to grow your fleet. Since your vehicles are wearing out by the mile, having a larger fleet with the same demand just means your vehicles last slightly longer, and you pay a little bit more to store them and a little more interest on your money. (I have analysis of this elsewhere.) The reward of growing the fleet is you provide better service than your competitor because you tend to have more vehicles available, which means shorter wait times. This reward is more than the cost of growing the fleet, so you do it.

But there are also two answers if a shortage approaches. One is to recruit private car owners to offer their cars in the shortage. This may cost a little more, though possibly not if the private owners think the incremental miles only cost them half the robotaxi fee.

The other is to start enforcing carpooling. For while there may be a shortage of vehicles there probably is not a shortage of seats. So in this case you might well see the system declare, "Private rides are 50% more now. Pool rides are still cheap."

If your vehicles last slightly longer, in addition to paying slightly more to store them and paying a little more interest on your money, you'll also have vehicles that are a lot more out of date (less safe, less luxurious, less energy efficient) for a short amount of time.

But yeah, I think shortages will be rare. In the global sense, I think they'll be nearly nonexistent. There will always be a car available somewhere for a passenger. The only question is how far away it is. But I also think they'll be rare in a local sense, because there aren't really that many surprises as to when and where demand will pop-up. But those times when there are surprises, there will be price surges, and all companies servicing that area will have to choose between hiking price, turning customers away, or making customers wait a very long time *and* operating at a loss.

The real trade-off will not be between shortages and having cars sitting idle. The trade-off will be between dead-miles and having cars sitting idle. And on that trade-off I think companies will choose to have cars sitting idle a lot. In fact, I think cars will sit idle so much that they won't "wear out by the mile." At least, I think this is what will happen in places where parking is cheap (and I think in the suburbs "cheap" will be essentially free - homes and businesses will allow robotaxis to park on their property for no charge simply as a way to get faster service).

The unanswered question is how many customers "pick a brand" and tend to use it, or even subscribe to it with a package deal, vs. how many shop each time for who has the lowest prices or best wait time. (Or if they are allowed to, use a meta-app which shops for them.)

I suspect that a brand will advertise, as a feature, "no variable pricing." Which means they eat the loss in a local shortage. If they are rare, that's doable.

Non-business users hate surge pricing. Sometimes I have taken Uber on the 50 mile trip to San Francisco when it makes sense. But if I open the app that morning and it says "Surge -- $200" it causes a strong reaction. "You can't depend on Uber" it says, because that blew up the math that had told me to take it rather than pay for $50/night parking at a hotel up there. So, I switch to my car, because I have one.

I suspect people will be willing to pay a small bit more for predictable prices. Others might not. Or rather, they might not have to pay more, because the larger fleet offers predictable prices as part of their economies of scale.

I'm sure there will be options for people who prefer consistent pricing.

I don't see why there would a large percentage surge fee on a 50 mile trip. That makes sense for Uber, maybe, as the driver probably has to drive all the way back those 50 miles (probably without a passenger most of the way back). With a self driving vehicle there's not as much urgency to get back, so there's more opportunity to pick up fares on the way back. In fact, in most cases you don't even need to send back the same car in the other direction.

On the other hand, I think it does make sense to charge more to the person going from a place where there is a shortage of vehicles to a place where there is a surplus of vehicles, than to someone going in the opposite direction. Especially if the passengers are booking the trip at the last moment the moment. People don't like the way airlines charge their passengers, but an airline that tried to give passengers what they claim they want (flat prices per mile for all passengers) would likely go out of business, as the only passengers they'd get are the ones that lose them money.

And I am sure there are many models of pricing that will be explored that have not been imagined yet. Though I do think the public likes predictability of pricing. In fact, it's one of the reasons (though not one of the top ones) that people like car ownership and subscription services of all kinds. Most pay TV it seems is sold in flat rate bundles via Netflix or Cable compared to a la carte, even though the latter is cheaper for many users.

It would be curious what would happen if a transit agency had variable pricing. I think there are some that have slightly more expensive rush hour tickets. I have not heard of one that heavily discounts low usage routes, or surcharges the most crowded vehicles.

Airlines are, as suggested, the masters at this. But their prices are far higher, and they have found ways to exploit every kind of price discrimination they can get their hands on. It also is complex when two types of pricing compete. If everybody flat rate prices, it works. If there is a mixture, then of course everybody goes to the flat rate service at the peaks, and to the variable service off-peak, and neither makes money. Unless they have more to compete on than price.

Right now, Uber feels it has to offer every driver in a surge zone the surge price. It could, in theory tell drivers, "You can accept short rides at surge price, but long rides come at non-surge price" and perhaps they would be open to both, at least at some ratio. The attraction of short rides at surge price is you stay in the surge zone. But long rides make you more per hour, which is mostly what humans care about. I don't know if the companies have done this math. They could offer "we'll pay you the same per hour, based on these assumptions about what your car costs to operate."

Robots don't care so much about this. The taxi company just wants maximum total profit per day, revenues minus cost of operating the fleet. Customers want simplicity of pricing, but reality suggests the cost of operation on highway and in stop and go city vary, both in cost per mile and cost per hour.

There's lots more left to model here.

Uber is definitely very limited as to what types of pricing it can do, compared to a robotaxi company.

One thing you said that I have to somewhat disagree with, is that you said if there's a mixture of flat rate service and variable rate service, that neither company will make money. I don't think that's true. During the peaks, the variable rate service will automatically adjust its prices to fill its capacity (that might be slightly more, slightly less, or much more than the flat rate service, depending on the relative capacities). During non-peak, most people will use the variable rate service (which probably won't fill up because it's non-peak). This will happen for a little while. Then the flat rate service will change its pricing model, get special favors from the government, or go out of business. (Note that this is assuming that the flat rate service isn't actually some sort of subscription service. If the flat rate service gives incentives for people to sign up with them and use them pretty much exclusively, then they certainly will be able to survive.)

I think some sort of subscription model will work well for robotaxi companies. Maybe not a pure monthly fee, but something like $X/month plus a small incremental cost per mile. It's what people are used to, and it makes a lot of sense. The monthly cost for the most common package would likely be a little bit less than what people would pay for an equivalent lease of a mid-sized car. The incremental cost per mile might be somewhat more than the incremental cost per mile of an equivalent personally owned robocar, but instead of having to pay the incremental cost for mid-sized car all the time you'd only have to pay the incremental cost for whatever sized car you needed for that particular trip. There will probably be some sort of family plan, but you won't save much per month over buying an individual plan if you need multiple cars at the same time (especially if you need multiple cars at the same time during peak hours). If the need for multiple cars at the same time is rare (a working spouse and a stay-at-home spouse) the monthly fee might be significantly less than buying two cars. More than just a single subscription, as this will increase the need for the company to keep cars on standby in residential neighborhoods during the day, but maybe not a whole more. (If this couple tried to game the system and share a single subscription plan it would become obvious fairly quickly, even without any facial recognition, as the location of the pick-up would hop around so much.)

There are still some cost savings to having your own car, though. One that seems kind of difficult to fix is that it's very efficient to be able to charge your car in your own garage at night while you sleep (as opposed to having it drive to some central charging station). Electricity distribution has a cost, but it's surely cheaper to send the electricity to the car than to send the car to the electricity. One possible way to fix this would be if the electricity company would be willing to install a meter at certain residences and charge the robotaxi company directly for the electricity it uses (having the homeowner resell the electricity might run into public utility regulations). In the longer term I guess people will move away from even having driveways or garages, and then robotaxi companies will just have to buy up land somewhere relatively close to residential neighborhoods. But it's going to take a very long time for already established housing developments to get rid of their driveways and garages.

Bus and train companies definitely charge different prices per mile for different routes (although that's often more a matter of price discrimination than reflecting actual costs - taking a bus from or to a rich neighborhood costs more than taking a bus from and to a cheap neighborhood - in fact one trick I once learned about Greyhound was that it is sometimes actually cheaper to buy two tickets for a connecting route than to buy a single ticket, even though the connection has to be made either way). Bus and train companies also generally charge different prices for peak and off-peak. At least one bus company I know of (Megabus) charges drastically different prices to different passengers on the same bus (they sell the first few seats on every route for $1, and prices go up from there with every seat that is sold). Local transit (and commuter rail) is often subsidized by, highly regulated by, or even run by the government. Maybe robotaxi service will be too. I hope not, and I kind of doubt it will be.

I think robotaxi markets will likely be much more competitive than they are in the bus and airline industries, at least once robocar software technologies become fairly widespread. (I definitely think that will happen quicker than many people, especially people investing in Waymo, expect. Not long after one company solves the problem, many others will too, both due to legal and "extralegal" means. Regulators might get in the way though, inadvertently or intentionally limiting competition under the guise of safety regulations.) In any event, more competition will lessen the ability of the companies to engage in price discrimination, so I don't think it'll be nearly as bad as the airline industry. (Then again, who knows, maybe the TSA will get involved. Robotaxi companies will actually be much more susceptible to terrorist attacks than other forms of mass transit, though the types of attacks will be extremely different.)

Yes, long I ago I itemized all the pricing models I could think of. And indeed, matching leasing was the one I thought might work best -- monthly fee, plus a fee for packs of 300 miles. That's most like how we pay today. We buy miles one fill-up at a time, maintenance every 6 months, the rest annually.

While per mile prices will be available, people are not used to them yet. Perhaps in time they will. If you charged me $50 for a round trip to San Francisco, I might be less inclined to go up there for dinner, and people don't like that. Though at $25 (ie. 25 cents/mile) it's not as big a deal.

As for where to charge -- installing charging in homes can actually be quite expensive, and a large fraction of people don't own a parking space they can put charging in. At my house, a charging station costs $5,000 because, like for many people, a panel upgrade is needed. So yes, it's cheaper to have it go off to a neighbourhood lot, next to the transformer station, with chargers it can just drive over. (Plug in, not inductive.)

If you already have all you need at your house, sure, the install will be perhaps $1K. Though that covers a lot of trips to the local parking lot that was converted to charging. The average private car goes 40 miles/day, which can be served in one hour by a 10kw level 2 charger. Which means 10-12 cars can use one every evening. More if you are willing to charge faster, which may be coming at no battery harm in the future. Once a station can do 50 in a night, it's cheaper than home stations.

So we have some people who buy by the mile and shop brands every trip. You have some who subscribe and never go off-brand. (Their subscription service subcontracts for them if they run out of cars, part of the deal.)

For the shoppers, it may well be variable demand based rates. But only to a degree because as you said, shortages will be rare, and mainly they only cost the price of repositioning.

Oh well, I guess you're right that the logistics of converting people's driveways into robotaxi charging stations would be too much. And now that I think about, there are plenty of other parking spaces close to my house that aren't used at night. The parking lot at the neighborhood park, the supermarket, the school, etc. The nearby gas stations could also be easily adapted to squeeze in some charging stations. Most of them have parking for the associated convenience store anyway.

The wasted (and ugly) space of all these driveways (if we're going to get rid of our personal vehicles and exclusively use robotaxis) is annoying, but I guess that'll just stay annoying until we eventually figure out a way to get rid of these driveways. Hopefully the zoning boards will be nicer about letting people convert their garages into more useful spaces for their personal-car-free lives. They probably won't.

Well, some people will like a driveway for being picked up. I have proposed changing the garage to the "room of requirement." You need a guest room? Robots come and deliver one -- they have the code to the door opener. Guest gone and you want some tools in a workshop? Over it comes. Needs fancier robots than we have today.

So most of them turn into storage. Like most of them already did...

My vision of charging is a bit different. I imagine the cars having a set of electrical contact plates on the bottom, with a small retractable cover. The car positions itself precisely over a small box on the ground. That box has retractable arms which lift up to press contacts against the plates. (Alternately the arms come down from the car into plates in the box.) Simple connection, made solid by the weight of the vehicle. Main issue keeping it clean. Cheap and automatic. Makes use of the fact the car is a robot and will position itself to the cm in just the right place.

Now that I think about it, another option will be to use the cars in the city for a while, racking up a lot of miles, and then let them retire to the suburbs after a couple years, at which point they can rack up a lot less mileage, and spend more time in driveways and parking lots.

It'd be interesting to see how that might work out. Give the cars 80,000 miles a year in the city for two years, then 30,000 miles a year in the suburbs for three years. For a total of 250,000 miles over 5 years. Could be a good deal for commuters in the suburbs. Or you could lease them out for those last three years, with the lease agreement including your company handling the operations.

I would guess suburban cars actually put on more miles -- in the world where people give up car ownership to live the robotaxi life. Suburban owned cars do more miles than urban ones. Taxis in the suburbs don't get much work because today, car ownership is close to mandatory out there.

Fewer cars in the city with more miles per car; more cars in the suburbs with fewer miles per car.

Of course many cars will regularly commute between and serve both areas. Maybe even all of them will rotate through this duty.

Well, they're not really a true scam but they will try to suck more money out of you where they can. I suggest doing your research to understand how they try to nickel and dime you. Your premiums and cost per mile will increase at each renewal. Their tracking device will drain your batter over time, which will cause it to lose signal and they will penalize you 150 miles a day until you fix it. If you have them, find someone else. They are not worth the trouble.

They happen to be the local promoter of pay as you drive. I don't endorse them. Payd, on the other hand, makes a lot of sense. I see no reason liability and collision insurance should not be charged by the mile. No risk is incurred when not driving.

Add new comment