Does your robocar come home after it takes you to work?

Does your car come back here after it takes you to work?

This week we've looked at two issues regarding robocars in the city:

Now let's consider another common question. Will owners of private robocars send their car home after the commute to park, or serve other members of the family, particularly as an alternative to having 2 cars?

Yes, some people will continue to own cars, even when it's much more expensive

I see a lot of debate over whether people will own private robocars at all. Many people imagine the robotaxi proposition -- Uber style service at a price that matches or beats the overall cost of private car ownership -- is so compelling that they can't fathom why anybody would own a car still.

The proposition is compelling. Eventually use of robotaxis will be superior in almost every way, from a purely rational business. At the same time, some forces will push some people into ownership

  • Living or working in an area (such as rural areas) where fast robotaxi service is not available
  • Ability to keep a variety of things in the car, not just between stops, but always. (I admit to doing this myself.)
  • A guarantee that the car you like will always be available to you on short notice. No long waits, no surge pricing.
  • Ability to show off your wealth and status.
  • The private car may also offer manual driving and perhaps be sporty when you want it.
  • Particular tastes not served by the robotaxi market
  • Inertia and stubbornness -- ownership is how we've always done it. And many other irrational reasons.

Yes, private ownership will eventually cost thousands of dollars more per year than robotaxi use -- but even so, some will choose it and can afford to.

It does not matter that you don't think these reasons are compelling. There are others who do. So there will be private cars, even at higher cost.

Households with 2 or more driving age people will choose between having 2 private cars, having 1 private car and having none. Even for the car owner, the middle case may be the most compelling mix of the two worlds.

If both people commute, and there is only one car, they will have to choose who gets to use it in the commute. There are arguments in favour of either person. Whose commute is longer? Who needs a private car more often while at work? What does it cost to park near the workplace and how close can you park? Are the commutes at the same time? And how does this match with the kids' commute to school?

If only one person commutes, there can still be a choice about them using the car or not. Many one car families have the commuter use transit or carpool if they are convenient, to keep the family car at home for the person who does not commute, and for the kids' commute.

Let's consider a family, Taylor and Jordan, where Taylor has a traditional commute to a business district, and Jordan either has a commute at very different times, or does not commute at all, but does other trips and errands during the day. There may also be kids needing transport to and from school.

Does it make sense for Taylor to commute in the private car, then send it back for Jordan and the kids? I'm still looking to get a good handle on the incremental cost of extra vacant miles in a vehicle, but a decent rule of thumb is it's probably about half the total cost, which is the mature robotaxi price. So, for example, 50 cents/mile for robotaxi, and 25 cents/mile for incremental mile in owned car.

Free parking

If Taylor gets free parking at the workplace, the answer is to not send the car back. If Jordan's use is fewer miles than Taylor's commute, it's cheaper to use a local robotaxi rather than send the car back. If Jordan's use is greater, it flips -- but then it makes sense to just have Taylor commute by robotaxi and have Jordan keep the car, because that's cheaper. In fact, a simple way to see it is that the cost of Taylor just using a robotaxi at 50 cents a mile is equal to the cost of Taylor using the private car at 25 cents/mile but doubling the miles due to the return trip, with the downside to Jordan of need to use robotaxi during those round trips.

It gets even cheaper if Taylor has the option of robotransit which provides a trip similar in duration and comfort to a private ride, but at a much lower cost. In fact, if that's available, it's the clear choice for the commute, unless Taylor has regular need for the car while at work.

The vast majority of commuters actually get free parking at their workplace, and the ability of robocars to densely park "valet" style in 1/3rd the space means more probably will. It's also quite likely that employers, rather than provide parking, will subsidize employee commute rides.

Pay to park at work

The need to pay for parking at work will alter the equation. I estimate a cost for dense-packed robocar parking on a monthly basis of around $4/day in typical CBD regions, less in less-dense zones. Now, just looking at the cost, it can make sense to send the car back if Taylor's driving does not exceed Jordan's by 16 miles. The complete formula for the difference in cost between keeping the car at work minus the cost of sending it home is: J/4 + P - T/4 and thus to make it work sending the car home you need: T - J < 4P

This can get more complex because Taylor doesn't have to send the car all the way home. If it is send just one mile to where parking is $2, effectively the cost of P is now just $2.50, and now the margin must be within 10 miles. But it also moves the needle to where it again makes sense to have Taylor commute in the robotaxi (and pay no parking.)

Different vehicles, and hiring out

Indeed, vehicle type plays a role in this math as well. Buying and operating larger vehicles will cost twice as much as single person commuter vehicles. Few families today buy a single-person commuter car, but this is the most efficient and economical choice for a solo commute and should be a common robotaxi offering.

It may very well be that the commute can be done in a 30 cent/mile robotaxi meant for one person, leaving the more expensive 4-seat private car at home. While the private car's incremental cost is less than the robotaxi, it's not much less, and the robotaxi continues to be the winning choice.

Another entrant in this equation is hiring out the commute car as a robotaxi while at work. This eliminates some or all of the parking cost, and possibly brings in some income on the vehicle. While doing this, Taylor may not be able to use that car for an unexpected ride, but a robotaxi would be available.


It should be noted that sending a car home is probably not a very convenient style of use. It means Taylor can't use it during the day and must use robotaxis. It means the car is not available to Jordan until 9:30 or 10am, and not after 4:30pm. Thus it may not be available for dropping kids at school, though it might do pick-up.

Once again, the decision may be based on factors other than the math, like those listed above. If Taylor has a very short commute, then there is not that much inconvenience to Jordan, and even though that short commute is no more expensive in a robotaxi (due to the round trip) the desire to have one's own car may win out.

Social factors

The deeper question will be about the social consequences of these decisions. If it's an economic wash, it's not a social one. The empty round trips add some congestion to the roads -- though it is mostly, but not entirely in the reverse commute direction. (Some places have high demand in both directions.) The empty round trip also creates emissions from its energy use, whose externalities are not priced into the energy cost. And while it is priced into those per-mile prices, the extra driving creates extra risk of accidents. If the choice is to not create the risk at a slightly higher cost, or have the accidents and use the money to fix them, it's obvious what choice we want.

The only reason this question even comes up is that when people buy a car, their ownership of it consists of 3 components:

  1. The time-based depreciation of the car, based on its age in months, as well as other non-usage based costs such as licence fees, interest etc. These happen if you drive or not.
  2. The usage-based costs of the car, based on miles or operating hours.
  3. Parking.

A robotaxi company also has all these costs (with less parking cost because they spend more time in circulation, working) but has to bill customers strictly by usage for their share of those time-based costs. The car owner, paying the monthly costs no matter how much they drive, is wasting their resource if they don't drive, and thus looks at the lower incremental cost in doing their math. This pushes them to do vacant car moves which would be wasteful with a taxi.

It could be that governments will wish to nudge this. One way they could do this is with road usage charges. As soon as there are road usage charges, they can discourage wasteful vacant operations, and bring the balance between use costs and ownership costs more in line with social goals.


Your analysis is interesting but omits one important factor: my time. You factor all sorts of costs into the ownership or rental, but you value my time at $0!. Even if my economic status is very low, the cost to me, which may not be monetary at all, in having to extend time for an auto use can easily overwhelm all other dollar figures. Not everything can or should be reduced to dollars.

How does your time factor into this? We are talking about the moving of empty vehicles with no human in them. The only human time involved is the time that your vehicle is not accessible to you as it reverses the commute, and the human is probably busy at other things during that.

That was a really compelling argument, but it all relies on the cost of a robotaxi being only twice the incremental cost of operating a car.

I do expect to see governments move toward road usage fees. Remember that most people (at least in the United States) *already* pay road usage fees, in the form of gas taxes (over 47 cents a gallon, on average, which comes out to about 2 cents a mile, on average). It might make the most sense to charge road usage fees per hour, once governments are forced to move away from gas taxes, as people stop using gasoline. I just looked it up, and AAA said that Americans drove on average 10,900 miles in 294 hours in 2015. So 37 miles per hour, on average, which means that governments (state and federal combined) would have to charge about 75 cents an hour in road usage fees to equal the revenue they're getting from gas taxes.

Interestingly, if governments charged 75 cents an hour in road usage fees (which, again, would average no more than what people are already paying in gasoline taxes), it'd definitely not make sense for people to drive around on the roads in order to save 50 cents an hour in parking fees. :D

And Millard-Ball's thesis was that this is the only solution. It is a solution (as are other types of road management.) And it may be a good idea. However, I don't believe it is necessary for this problem. It's good for other problems.

And yes, the range of $1/hour is probably pretty reasonable, even low. Enforcing this on human driven cars is challenging though -- it requires some infrastructure and Orwell risk. Enforcing it on robocars is actually much easier, because there is no need for retrofit, and they already track their travels. Big companies are much easier to enforce the law on, because if one car is caught cheating, all cars from that company are suspect, and will get in trouble. If one human is caught cheating, that's all that happens.

As you know, gas taxes are a political 3rd rail. Though the move to electric cars is starting to change the landscape a bit. The non-electric drivers want the electric cars to pay the equivalent of gas taxes. However, for emissions reasons this will be resisted, as the truth is gasoline burners are not paying the external costs of their actions.

As for robotaxi cost, it should be (when mature) close to the all-in retail cost of car ownership and use today (which is 50-70 cents/mile) but it will be lower for cheap robotaxis meant for one person and urban travel, as they can be made for half the price.

Private cars, however, will not be cheap single person vehicles. Owners will, as they do today, gravitate to 4-5 passenger sedan/SUV/minivan styles for family vehicles. These tend to be more expensive. So their incremental cost will be probably more than half the cost of a larger robotaxi and only a bit less less (if not more than) the all-in cost of the solo urban car.

Why would robotaxi cost be close to the all-in retail cost of car ownership and use today?

The all-in cost of a car today includes not just the actual cost of making the car and operating it, but all the retail margins, marketing costs, channel costs etc. The auto industry and related industries are profitable selling people cars and services at these costs.

The robotaxi has some extra costs over the robocar. Fleet management, cleaning, service. But it has a lot of savings by vertical integration. Self-insured. In-house maintenance and repairs with no middlemen. Wholesale costs of energy.

And it's putting on miles much faster than a private car. As you've enjoyed discussing, incremental miles are cheaper than lifetime miles. So a car that operates as a taxi and does 60,000 miles in a year has a lower cost per mile than a private car that does 10,000.

If a cheap robotaxi meant for one person and urban travel can be made for half the price, Maybe a two-adult household will switch from two cars to three. One single-person robocar for each adult, which they drive to work every day, and one multi-person robocar for the family, which they keep in the garage most of the time and only use for whole-family travel.

And the cost of those three cars will be less than the cost of the two cars they have now. Equal in cost for the car itself (1 + 1/2 + 1/2), but much less in operating costs.

For now, I am suspecting that people don't like to buy the commute cars. Today people almost never buy single person cars in the USA. (They are pretty common in some other countries and in fact in some places, single person vehicles, including motorcycles, are the bulk of vehicles on the road.)

A one purpose car is great as a taxi. It's less attractive as a family asset because of exactly that. It's only good for that one purpose. Though "single person travel" is a pretty broad purpose, I will agree.

Generally, if you are looking at a car that will only be for the work commute, and do 6,000 miles (average for a commuter) odds are that the taxi is a cheaper and better choice, with no parking or other ownership hassles. Unless you want to store stuff in it, or impress people at work (who are also all arriving in these) it's hard to justify owning it.

It's easier to justify the vehicle that can do most family needs, including commutes and family trips.

As I note, there are places where ownership is good. But since we're talking about the cost of ownership being a mix of the monthly, non-use-dependent cost, and the use cost, more cars is more expensive, as each is used less and sits around more. A car that costs half as much will have a slightly higher than half monthly cost. Depreciation will be half, but licence fees, non-use insurance etc. will be slightly more.

Also two one seaters are very nice when you want to go 2 places, but not when 4 of you have to go somewhere. Two two-seaters can be good, as an inline or F2F 2 seater is not a lot bigger than a one seater.

As far as I can tell, there are no single-person (or two-person) commuter cars available in the USA at half the cost of an energy efficient subcompact 5-seater. To qualify as a commuter car I would require it to be enclosed, to have heating and A/C, and to be street legal for all roads. If there is one out there at half or less than the cost of, say, a Toyota Prius C (about $21,000), I'd love to hear about it and I might even buy one.

I was only talking here about comparing the cost of a family owning two normal cars vs. a family owning one normal car and two half-price one-seater commuter cars. The latter is clearly a win unless the family needs to use both normal cars at the same time, since they pay the same price for the cars but less in ongoing fees. In many cases the family could switch to one normal car plus one one-seater commuter car (letting the commuter who drives farther take the one-seater to work), and save even more.

This all assumes that one-seaters costing half the price of a five-seater exist, though. Based on my quick search, I couldn't find any. A Smart EQ forTwo costs more than a Toyota Prius C.

And no, at this point I wasn't comparing to using robotaxis at all. I think your estimates of the retail price per mile to hire a robotaxi (for commuting, which will presumably be peak time if lots of people abandon car ownership) vs. the incremental cost per mile to drive a car, are way off. You estimate 2x. I think it's more like 5x. But there's not much more to say about that, unless you want to give numbers. My guess is that the cost of a robotaxi during peak hours will be similar to the cost of an Uber (during non-surge pricing) minus the net profits earned by the Uber driver. The cost of ownership and operation of a vehicle is only a fraction of the cost of running a robotaxi company, and perhaps more importantly, the retail price of a service has very little relation to the average cost of providing that service. (Economic theory suggests that marginal costs tend to equal marginal revenues, at a point where marginal costs are rising; average costs tend to be lower than marginal costs; in fact they have to be lower or else the company will be unprofitable; and I think it's quite clear that average costs will be *much* lower than marginal costs when it comes to providing cars for commuting, as that is a peak time when the marginal cost of adding a new customer is equivalent to the marginal cost of buying a new car for just that one customer -- all the other commuter cars are already in use during rush hour).

Several questions here.

Half-width enclosed cars are mostly made by small manufacturers, and outside the USA. If you have a narrow stance you either can't go fast (no good in USA) or you must be able to bank. There are a variety of banked wheel 4 and 3 wheelers out there, mostly not enclosed. The Toyota i-Road is a good example of an enclosed design. Of course, for a pure city car, going fast is not always necessary, but it's still less stable to have a narrow car. I drove a narrow car a friend had that was scary that way, in spite of having a bunch of lead in the floor.

Old article on narrow cars

As for the cost see Robotaxi economics

The cost of an uber, less the driver: Actually, that's not too far off. Right now, the cost of an Uber, less the driver's profit, is the incremental cost of driving that car plus Uber's share. Except, right now, Uber pays almost its whole share back to drivers, in a money-losing attempt to grow. So you are actually predicting the taxi will be much cheaper than I am predicting!

It won't be that cheap. But how much cost should the fleet operator add? They don't have much in the way of per mile or per ride costs, above the costs I cite. Today, they have huge software development costs. In the future, it's almost all NRE which is not much per ride if you get big enough.

Back to our favourite topic, incremental cost. I've done more research and found papers putting it closer to 20 cents, plus insurance. So I used, in this article, 25 cents which is half the typical all-in cost.

But the question of "incremental cost" is a misleading one. Is there truly an incremental cost? In some cases yes, but others no.

When you justify buying a car, you expect to drive it. Nobody buys a car to leave it in the garage all year. The reality is, you plan for a certain level of driving, and you justify the cost of the car based on that level of driving. An incremental mile is actually more validly a mile you didn't factor in as part of that equation.

So there's a case that sending a car back home is an incremental mile, but a case that it might have been part of your plan for the car.

If you bought a car for yourself, and decided to start driving Uber, then those Uber miles are incremental miles.

One could argue that "any mile you're examining the cost of" is an incremental mile, which would negate this. But I'm not sure that's right. So if you bought a car to commute, I think you should consider the full cost as the cost of your commute. But if deciding to send the car home, if you didn't plan for that, it's incremental. Confused yet?

By the driver's net profit I would mean the net profit they'd report on their Schedule C if they elected to report actual expenses instead of a per mile rate. That's mostly the cost of the car, but it's not merely incremental cost, and it includes "dead miles." It also includes some non-car expenses, like the costs of a cell phone (or in some cases multiple cell phones). But it's mostly car expenses.

And "Uber's share" would have to mean everything paid by the passenger minus the gross commission. Base fare plus per mile plus per minute plus service fees. Plus tips if you're including them in driver profit. Or ignore tips if you're not including them in driver profit. I calculate much more than 50 cents a mile. What do you get?

Robotaxis aren't going to be cheap. For most people they won't be cheaper than owning a car. They'll be wonderful for situations where you can't or don't want to use your own car, though. Which I guess will be less and less often. For going to the bar, you can take your own personal robocar. For driving to the airport from home, you can take your robocar.

Maybe there will even be an Airbnb type service to allow you to rent out your personal robocar, when you're not using it, to people who can't or don't want to use their own car. Hopefully the regulators don't prevent something like that.

Uber's share is the commission and any fees that go to Uber. If you are an Uber driver, the passenger pays $2/mile approximately (leave out special fees)

You get 70-80% of that, but then Uber gives its commission back right now in driver incentives, it is reported.

Your costs are your dead time, as said, and the incremental cost per mile of your vehicle (about 20 cents, depending on vehicle.) The rest you keep. You may well use the IRS vehicle rate on your taxes, why does that matter.

The idle time is in my spreadsheet, but it's different for a robotaxi. There is no human who needs compensation for that time, so the only cost of idle time is any parking fees, and the cost of capital to sit there. It's the idle miles (diving between rides) which are a real cost, and estimates run that at around 15% but I have seen some suggest lower.

The other way to ask this is, "what should Uber's commission be, in a competitive and mature world?" What are the costs of what Uber does? It's rather low since it's all automated. The only real cost is the insurance. They do pay for marketing (but only to recruit drivers.) They have some customer support. The rest is NRE which should become small. Right now the commission is about 50 cents/mile. I think it should be perhaps 5 cents, plus insurance.

What should Uber's commission be, in a competitive and mature world? Higher than what it currently is, because Uber right now is not making a profit.

The non-electric drivers want the electric cars to pay the equivalent of gas taxes.

Already happens here in Missouri; EVs must buy an "alternative fuels" decal every year that is similar to the cost of the gas taxes they would have paid.

One problem we found when the kids were young was how difficult it was to properly install the bases for the rear-facing infant seats. Not all of them would work in every car, and some cars could only have them in certain seats, and others didn't have room for multiple infant seats and booster seats. Even in a vehicle I was familiar with, when I had to reinstall one, it took a noticeable amount of time, even without cranky kids supervising. If I were in that situation and considering a robocar, I'd want to own one so that the seats I needed were always properly installed for taking a trip on a moment's notice.

Now that the kids are old enough for standard seats, I'd probably prefer to use a robotaxi, although I do have things that I'm used to just keeping in the car. I think I could learn to adjust.

Yes, I think some people with kids might have this concern. However, there are so many people with kids that there will be standard configurations of taxi you can order that come with kid seats. Seats that an adult can't use at all. Facing the parent, as well, at the front where the driver used to sit.

All children not able to use adult seats would sit rear facing. All seat belts would be adjustable to all child sizes. It is probably possible to reduce the number of seat types to two, maybe even to one -- imagine a configuration with dedicated airbags or motorized configurable seat belts to fit the child, and ditto for side walls. I have not done an examination of the rules and principles to see what's possible.

So you would find cars configured for one adult plus one child, and 1-2 adults plus 2 children, and a smaller volume configured for 3 children. You would probably be able to summon those quickly as they will be in large supply. Rarer configurations would take longer and might make people want to own. Rather than booster seats, I would expect just better seat belt configurations.

But nobody's tried to try to solve this problem yet. While a universal seat, which uses a combination of motorized belts and airbags, might be an expensive item, it would be cheaper than buying a vehicle I think.

In a few places you mention insurance. Will the owner/operator of a robocar even need insurance? Ultimately it'll be up to legislators to decide, but in the vast majority of the cases of crashes involving a robocar, the owner/operator will not be the one at fault. Either the other driver will be at fault, or the manufacturer of the robocar will be at fault. The owner might have joint and several liability simply for the fact that they own the car, but that would only come into play if the manufacturer were insolvent, and manufacturers will likely have the deep pockets to cover claims.

I'm not sure the law will require the owner/operator of a robocar, who doesn't install any non-approved after-market software or hardware, to have an insurance policy at all. They might want a comprehensive policy (to cover against theft, fire, natural disasters, etc.), or an uninsured/underinsured motorist policy, but these are, and likely will continue to be, optional.

If the law does require a liability policy it'll likely be extremely cheap, since the owner/operator will almost never be liable for anything.

If the law does require the operator of a robocar to have insurance, how would they possibly enforce that, anyway? The operator is, presumably, the person who presses the button telling the robocar to go. But the police will pretty much never observe, let alone be able to prove, who pressed that button.

You can, of course, require the owner of the robocar to have insurance. But if the car is leased, then the owner isn't the same as the person riding in the car. And the leasing company can always self-insure, especially considering that owner liability is going to be virtually nothing for a car manufactured by a deep-pocket manufacturer.

Of course it will be logged who pressed the button. The idea of "the person who pressed the button" is a temporary kludge in the law that was created to move to law that makes more sense. The vehicle codes are full of references to the driver, and we needed a simple way to say who the law should say that is in a robotaxi. But it's actually silly, and I expect it to be replaced with a normalized set of rules, liabilities and penalties for a company operating a taxi fleet.

I do expect self-insurance. The builders of robocars are large companies, and they know more about the risk of their own cars than any actuary.

Logged how? I'm talking about a personally owned robocar. I don't think the government would have any proof of who pressed the button. In most states, traffic violations, such as driving without insurance, have to be proven beyond a reasonable doubt. I don't see any way to do that. Certainly any traffic lawyer could get the case thrown out. Traffic camera cases get thrown out all the time, and they come with video evidence. Some log entry in a log file wouldn't even be admissible as evidence if it was properly challenged.

In the longer term, you're right that legislatures will have to address this. They'll have to address it in large part because these cases will be unenforceable (in those states where traffic camera violations are enforceable, it's invariably because the legislature passed a law specifically making them enforceable). At that point I am arguing that it makes no sense to require occupants of a robocar (whether they own the car or the car is owned by a robotaxi company) to have liability insurance. I suppose legislatures can and regularly do enact stupid laws, but this (requiring someone highly unlikely to ever be liable for anything to carry liability insurance) would be an incredibly stupid one.

If legal liability goes to the person who activated the vehicle (which I don't think will actually be the case for personally owned robocars) then there may well be a requirement to log who it is. Because if somebody doesn't log it, or somebody erases the log, then liability will probably fall on them.

I don't think traffic violations need reasonable doubt. That's the criminal standard. And non-moving violations don't need a person at all, they can be filed on the owner of a car.

There will be a requirement for insurance, actually, but it will be satisfied by the fact the vendor of the car is providing it. You won't have to buy it from an insurance company, unless your car allows human driving, in which case there will be policies that cover only that. However, you also want, perhaps, comprehensive coverage.

I don't think legal liability will go to the person who activated the vehicle either, so why are we even talking about that?

The standard of proof for traffic violations differs state by state. In California, unless it has changed very recently, it's proof beyond a reasonable doubt. Most states that I know about also use that standard, but a quick search wasn't able to find how many out of the 50 use that standard.

I don't think there will be a requirement for insurance, unless your car allows human driving.

"Living or working in an area (such as rural areas) where fast robotaxi service is not available"

I'm trying to reconcile this with the recent statement that we will never see Level 5 autonomous cars. Absent Level 5, what robocars are out there in the boonies, far beyond any geofencing limits?

The word "geofencing" is a bad choice, and I wish people didn't use it. What they really mean is mapped. The car is constrained to the roads that you map, which is (mostly) the roads that are economic to map. There is no reason it can't be economic to map rural roads. While it is more lucrative to map urban roads, it's not that expensive that people won't do rural roads too. There are some other limits that might reduce rural operation such as dirt roads, but otherwise rural roads are often easier than urban ones.

The other main economics of rural roads is you can't have the same density of cars. You won't have 500 cars per square mile. But then, you don't need to have that. It will take a bit longer to get a pickup, probably, due to the lower density, but maybe not even that, because rural roads are faster.

Related to the child-seat argument, I think one reason people will want their own car is so that they can stock it with things they might need only very rarely (such as extra water or an emergency blanket) or might forget to bring.

If you're planning to stop at the gym on the way home from work, it might be very convenient to leave your gym bag in the car, rather than carrying it into the office and then having to carry it back out again. (Can put it in the car the night before, no chance of forgetting it at the office, hands free for the other stuff you need to carry, etc.) But that doesn't work with a robotaxi.

Do you refer to some desire to store things other than that described at the start of the article?

Add new comment