Since people first started talking about self-driving cars, some have asked the cynical question, "Hey, in order to save money on parking, won't people just tell their car to circle around the streets until their owner is ready to be picked up?"
Recently, a paper by Adam Millard-Ball of UC Santa Cruz offered deeper analysis of this question, and to my surprise, suggested that it might actually be economical for people to do this. If they did, it would clog roads and be highly antisocial, which is a matter of concern.
It's definitely worth examining all the ways in which people might abuse robocars, or what unintended consequences there might be. History shows that any new ability, even if highly positive, gets abused for a variety of reasons. This also has some similarities to the problem of roving robotic billboards I wrote about earlier.
I've written frequently that circling rather than parking is not likely to be a problem, but it's worth examining why Millard-Ball's reasoning that it will is flawed. His paper makes the following assumptions, which are most likely incorrect.
- He calculates that a car could drive around for an incremental cost of only 50 cents per hour. It is probably much more.
- In particular, he imagines cars might do this by driving extremely slowly -- less than 2mph -- by seeking out streets where this is easier to do.
- He presumes that the cost of parking in a robocar world will be similar to today's cost. It should actually be far less.
- He doesn't anticipate that if, for some reason it still makes economic sense to circle rather than park, that governments or public pressure will quickly stamp it out.
Any of these issues is enough on its own to suggest there will not be a problem with vacant cars "cruising" the streets.
The incremental cost of driving
This area is the most uncertain. There is much calculation of what the cost per mile of owning a car is -- typically around 50 to 80 cents according to major sources like Kelly Blue Book, Edmunds, AAA and others. But that's the total cost for typical car use. Cars lose value both with time and with use. Our only proxy for use at present is miles driven, but it's a flawed proxy. Cars suffer more wear per mile at low speeds, especially when accelerating and braking, than they do cruising on the highway at high speed. Gas cars even get better gas mileage on the freeway than in the city. Millard-Ball takes research suggestions ranging around 18 cents/mile (not including insurance,) increases them slightly for the low speed and concludes that if you can drive 2 miles/hour it might only cost 48 cents per hour to cruise slowly on the roads.
Stanford demonstrates automated parking in a lot in 2009
Credit: Brad TempletonThere is a stronger case that electric vehicles have a lower incremental cost. They get better mileage at low speed, though stop and go still wears on them (and their batteries) more than smooth driving. For gasoline cars, which would be idling much of the time at a slow speed, hours might well still be a better proxy for aging than miles.
Some of the cost of driving comes per mile, but much of it comes from other factors which are more connected to hours of operation than miles on the tires. Exposure to elements and road salt (where present) happens by the hour. Wear on batteries seems based on temperature and the stress of current spikes (found on both acceleration and deceleration in electric cars.) The interior wears per hour, but not in this case, because nobody is sitting in it.
Unaddressed in the study is the cost of accidents, which today we calculate from the cost of insurance. Most commonly, a number of around 6 cents per mile is used, and indeed there are "pay as you drive" insurance companies which sell insurance by the mile at such a rate. It is hoped the cost for the robocar will be less. This is hard to calculate. For humans, it's likely that risk is very much accumulated by the minute, varying with the complexity of the road driven and not simply the speed. Severity, however, does go up greatly with speed, as well as ability to avoid. For robocars, risk is based on a mix of factors, most of all traffic complexity. A packed street with many types of road users is much more difficult than an open highway. On the other hand, time to react is larger. The risk of software faults is probably better measured in hours than miles.
Most robocars won't buy insurance. Rather, their makers will self-insure against any incidents. So those makers will want to keep risk as low as possible.
The short summary is we don't know the answer here. We won't know until somebody takes an electric vehicle and drives it for a long time in slow traffic and we measure the wear on it. My intuitions are this will produce a number more than 48 cents/hour -- probably 3 to 5 times as much.
The legality of deliberate slow driving
Prof. Millard-Ball posits the vehicle could deliberately go very slowly. Really, really slowly. In most places, there are laws against doing that. In California, the state examined in the paper, vehicle code 22400 prohibits driving at "such a slow speed as to impede or block the normal and reasonable movement of traffic," except for safety reasons.
As such, cruising cars would need to go at around the average urban average speed of 12mph, raising the cost to $2 or more per hour even with these low incremental cost estimates, and probably more, not yet adding the risk cost.
Below, I'll discuss whether it would be practical to deliberately disobey that law.
The low cost of parking
Today, parking in dense districts costs from $3 to $5 per hour, though it is sometimes more for short durations because of the way lots are run. But that's the price with today's parking economics. The economics of robocar parking are vastly different, and I forecast the cost of parking will be extremely low in the future, well under $1/hour.
I have written several articles on the future of Robocar parking that you might like to read for more details. However, I will summarize the key points below.
Drivers need parking that's a short walk from their destination. As such, there are only a few lots that compete for a driver's business. You can't just go to the lot half a mile away for a cheaper price. But a robocar would. In fact, almost every parking space within a few miles will be competing for a robocar's business, and even today, many of them have lots of empty space.
Parking space is special type of product. If you don't rent it, the value vanishes. The cost of taking in a parking customer is close to zero, and thus only the opportunity cost of having a better paying customer come along later, and the "market depression" cost of making customers expect they can get super discounts. But if you don't rent the space, you get nothing. Once you figure that opportunity cost, if anybody comes in offering more, you should take it.
But robocars have almost no opportunity cost. That's because they never "park" -- they are always able to move. So your lot can offer space to a robocar on the condition that it either move or match the price if a better offer comes. If you fill a floor of your parking lot with robocars paying 30 cents/hour, and a human driver shows up willing to pay $3/hour to park close to their office, you just kick some robocars out. Electric robocars, silent and emission free, using automated gates and payment systems, effectively cost you nothing. And so you'll rent them space at almost any price they will pay.
To top it off, they will use up the unpopular space at the far end of your lot, and they'll park in far less space. Today's typical parking lots for humans use 300 to 350 square feet per stall -- they need lanes for cars to get in and out. Robocars can park more densely than any valet. You don't even need to open the doors for the valet to get out. A typical medium size car takes up just 100 square feet when packed densely. You are parking 3 times as many of them in your lot than regular cars. If, as I have forecast, many city dwellers opt to commute in single person city cars that use 55 square feet, you can get 5 of them in the space for one regular car.
So, perfect customers who park at the back of the lot, will leave if you get a better offer and who take up only 1/5th of the space of regular cars -- what will the parking lot charge them in a highly competitive market where they can chose from any one of 100 different parking lots? It's not a lot. I suspect it's even less than the optimistic 50 cents/hour cost of cruising Prof Millard-Ball proposes.
The parking glut
The above analysis is based on our current parking demand. Robocars change that a lot too. As Uber and Lyft have already shown, robotaxi services will be one of the most popular ways to drive in cities. Those using robotaxis don't park at all. The taxi may have to take short pauses, but mostly it goes to serve another customer. As this grows, demand for parking will drop. (The convenience of taking your private robocar downtown with cheap parking will provide a counter-force, but a lesser one.) Even when the taxis are "parking," there will be only one car where there used to be 4 to 10 serving those people. I also forecast robotaxi service will be cheaper than having your own private car, and thus be very popular in cities.
In addition to the valet-dense parking in lots, robocars can also park a lot of places regular cars can't. That's because they don't park, they "stand." They can leave at any time. So they can be allowed to stand at the entrance of non-busy driveways or in front of fire hydrants. They can also double park and triple park where it makes sense to do so, because they can get out of the way of any car they have blocked in, as soon as that car turns on its turn signal.
With all these factors, I predict a giant glut of parking. Such a glut that the prices drop too low, and many owners of parking lots work to redevelop them to other purposes. I hope that some of the parking will get turned into green parks -- though in reality more condo towers are the most likely outcome. Either way it's going to be a land bonanza in cities, some of which allocate 1/4 or more of their land to parking.
Once the glut is redeveloped, the price of parking will rise again, but probably not to the level where it makes sense to circle.
The antisocial nature of it
Cruising around, blocking traffic and putting people at (small) risk just to (not) save money on parking is a pretty antisocial thing to do. So it's hard to see any developer of robocars implementing such a feature in a car. The public backlash would be nasty. No car from any significant provider is going to have such a feature. The interfaces of these cars will mostly let you tell your vacant car to either go to a destination (like your home, or a place to pick somebody up) or to go find a place to wait/park. When you ask it to wait, it (or rather the software services you use) will go into the "spot market" where people do bids and asks on parking spaces. It will find the best value space and go to it, and it may move in the meantime to get better prices if that makes sense. Then it will come back to pick you up when commanded. It won't have a "circle around" button.
Indeed, while the first antisocial element we might identify is clogging roads, the reality is that idle driving for no purpose also creates a small but real risk of accidents. We tolerate a lot of risk on the road, but it's doubtful we would tolerate any risk just to be incredibly stingy on paying for parking. The producers of these cars will, as I wrote, be self-insuring and thus taking on this risk themselves. The last thing they want is somebody hurt by their car in an accident while it was cruising the roads to save parking costs.
If some major vendor did want to put an antisocial button in their car, and public pressure didn't stop them, it's easy for the law to do so. It's hard for traffic law to be enforced because there are millions of drivers, and you can't watch them all. It's very different for companies. There will rarely be more than a score of different robocar platforms in a town, if that. Perhaps only a handful. It's very easy to watch them all. It's very easy to pass a traffic rule and make sure they are all following it. They won't dare get caught violating those rules, and risk not merely a traffic ticket, but the right to operate in a city.
Truth is, there's not even much need for laws. Governments will be able to get the reps from all robocar platforms in a room and talk things over. You can't get every driver in one room, and they won't obey what you say if you do. A simple meeting where it's agreed that such activity is a bad idea will be enough, I suspect.
The one thing this would not stop would be small players making add-on applications which control a car's API. You could make a program that acts like the owner's phone. First it tells a car to drive to location A. It watches, and before it gets to location A it says to redirect to location B, and so on. This would make the car wander. It could not wander below the speed of traffic, since cars will generally not allow illegal actions during unmanned operation. (There is even debate if they will obey illegal commands from occupants.)
You could do this, but not without the car's systems being aware you are doing it. One or two iterations of this will look nothing like normal human use patterns. The car vendor, who is motivated for the reasons above to stop you from doing this, can stop you.
That is to say if it made sense financially. But it's not going to make sense to put $3/hour of cost on a car to save 50 cents/hour of parking.
As I noted above, there is one type of antisocial driving the cities probably will need to ban, namely having a robotic billboard drive around all day to do advertising. That will be easy to ban and I suspect it will. I suspect vacant vehicle operation will be regulated, allowed only in the service of real operations for passengers and cargo, and not permitted for advertising, vending or wandering.
Free parking
All this talk of cheap parking may make one wonder about free parking. If there is free parking to be had, robocars will of course move to take that, and it could even make sense for them to drive around searching for it. Not for a long time (due to the cost) but for a while. After all, humans do that, though they are much more keen on getting parking close to their destination. It's estimated that in some areas, as many as 30% of the cars are hunting for street parking.
Many urban planners now believe that free parking was a giant mistake. The prophet of that movement is Donald Shoup and his book The High Cost of Free Parking. Shoup has convinced many that offering free parking is bad for cities, and I concur with many of his views.
Cities and buildings nonetheless have large amounts of free and unmanaged parking. Fortunately, the near-universal deployment of the smartphone offers a way to manage all these spaces, and make them cheap, but not free. For a cost far less than having parking meters, cities could move to smartphone app based parking allocation. They could even take surplus used smartphones (which are available in huge supply) and give them data plans only usable for the parking app, then give them to anybody who can't afford a smartphone. There would be no more hunting for parking even by human drivers -- they would tell the phone where they want to stop, and get a spot allocated to them as they approach the area. The robocars, of course, would handle this directly and move quickly to their allocated spots. In a car, there is always power for a smartphone. New arrivals, or those whose phone is broken could pop into an convenience store, or go to the old parking kiosk machines, to be allocated and pay for a spot.
They would not pay a lot. Parking in formerly free areas could be any price a city likes, even a nickel an hour. Retailers could still subsidize it if they wished. The main goal would be allocation, to eliminate the tragedy of the commons that comes with completely free parking.
If we're not ready to do this, it's also possible to just declare that vacant robocars may not make use of free parking. A robocar found in a free space would be ticketed. Which really means the developers would program their cars to never park there.
Parking as a proxy for road payment
The story of the arrival of cheap parking actually scares some urban planners. Today, the very high cost of parking in some regions is what regulates traffic. People are discouraged from driving downtown because parking will be so expensive. This pushes them to transit, which the planners like -- or to taxi/Uber/Lyft, which they don't. The good news is, there's much opportunity in the new world of computerized transportation to manage our roads in other ways, and keep the congestion down. But that's for another article.
I offered the Professor a chance to offer a short response to this critique. His more detailed points are made in his paper, but I sum up his response as follows, paraphrasing his words:
Professor Millard-Ball, in response to this critique, believes his estimates for the marginal costs of travel are still reasonable, and that while parking prices will fall somewhat, the urban land market will respond and redevelop newly unprofitable parking lots. Cruising at 50 cents per hour will still be cheaper than parking in city centers where land is scarce and valuable. Instead of regulating cruising, he believes that congestion pricing is a simpler, more effective and more efficient policy solution, and one that also confronts the root cause of the problem - underpriced (or unpriced) road space. Congestion pricing would also address the environmental, public health and other consequences of increased car use, once parking pricing no longer acts as a major constraint.
Update: There is one circumstance where people may want to circle, and that's during the "pick-up" rush, when everybody wants to leave at 5pm and they are all summoning cars. We see this already with Uber and Lyft. I have a special article about this problem.
