Timothy Lee over at Vox.com presents the case for the taxi model of car use in a driverless car world. This is a subject that we’ve covered before on this blog, and my opinions haven’t changed very much. But let’s review the topic from a slightly different angle.
Proponents of the taxi model of car use tend to present (and overstate) the benefits of that model, and ignore the improvements in value that other models have in a driverless car world. Ultimately, I think that a lot of it is going to come down to how expensive a driverless car ends up being.
As Lee mentions, cars sit unused 90% of the time. However, let’s be clear: that doesn’t mean that 10 people can share a single car. Demand for cars is not constant. Most people want cars during their morning and evening commutes, sometimes during lunch time, etc. Almost all cars are parked at 4am, and indeed at 2pm (on workdays at least), but that’s not going to change in a driverless car world — there’s just no demand for the cars then.
How much the utilization of a car could increase in a dominant taxi model depends a lot on how much customers are willing to accept rationing during high-demand times. Basically, you could probably get by with 1/4 of the cars that we have now, if people are willing to somewhat regularly wait for 30 minutes to get their vehicle during rush hour, and to have virtually no chance of getting a vehicle on Thanksgiving and whatnot.
Compare an ownership model: your car is always available for you (or perhaps you compete with your spouse), but you don’t get the cost savings of implicitly sharing with other people.
So, now to some numbers. Let’s say that a taxi model ends up cutting the number of cars down to 1/3 what it used to be. So basically each person is paying for 1/3 of a car. The taxi company needs to make some profit: let’s say their gross margins are a pretty narrow 10%. So if X is the cost of a car, then a person pays about: X * 0.33 * 1.1 = 0.363X. In contrast, an owner pays X.
What is X?
A pretty normal car today costs $25k new. If you put 100k miles on it and then sell it, you’ll probably recover about $10k of its value at time of sale. So it costs you $15k during the time that you own it. It probably takes you 7 or 8 years to put 100k miles on it (most people drive 12-15k miles per year), so let’s say that it costs you $2k per year (just the ownership: this doesn’t take into account gas and maintenance). Cost conscious people cut that maybe in half down to $1k per year (or less) by a combination of buying a cheaper car and keeping their cars longer. Rich people may pay vastly more for more expensive cars.
So this suggests that if a driverless car magically costs the same amount as a normal car does today, and you go with the taxi model, you save almost $1300 compared to the owner. Maybe down to $650 per year if both you and the owner are cost-conscious people.
But driverless cars won’t cost the same as normal cars. They’re going to need a bunch of probably pretty expensive sensor systems that normal cars don’t need. How much those systems will cost is really up in the air. I’ve heard the claim that Google’s current LIDAR array costs about $80k — but it’s not clear that LIDAR will ever be the sensor system that you can go into production with. And presumably there would be economies of scale involved in the mass production of driverless cars.
There’s also a question of how much a sensor system’s value will depreciate over time. If the sensor system costs $20k, but when you sell the car at the end of its lifetime it’s still basically worth $20k (so, on a $25k base vehicle like we were talking about above, the new cost would be $40k, and the cost you could sell it for after 100k miles would be $30k), then the sensor system is to some degree costless. Or maybe it’ll be effectively worthless at the end of its life. For the purposes of basic sanity, I’ll assume that it depreciates in a similar way to the rest of the car.
Okay, so let’s look at a few options:
Sensor system costs $10k (and depreciates to $4k)
Sensor system costs $20k (and depreciates to $8k).
Sensor system costs $50k (and depreciates to $20k).
The $10k sensor system makes a $25k car’s ownership cost go from $15k to $21k. Spread out over 7 years, that’s $3k per year. The owner pays all of that, the taxi guy pays just over $1k a year, saving $2k.
The $20k sensor system creates an ownership cost of $27k, about $3.6k a year. Taxi guy pays $1.3k a year, saving $2.3k.
The $50k system creates an ownership cost of $45k, about $6k a year. Taxi guy pays about $2.1k a year, saving $3.9k.
For the $50k system, a cost-conscious consumer has to note that even a driverless taxi system is still more expensive than an owned, driven car. Now, it’s not a lot more expensive, and maybe the advantages of a driverless system are worth the extra price. But keep that in mind when you’re suggesting that the major motivator of the ownership model is purely price.
What does an owner get that a taxi guy doesn’t?
- He has exclusive use of his car during high-demand times.
- He can do things like leave his stuff in the car (probably especially relevant to parents who may be toting around a lot of stuff)
- He can keep his car in a low-expense, timely way if he, say, goes way out into the sticks and has dinner with a friend outside the normal radius of the taxi fleet.
Is that worth probably somewhere between $50 a month to $500 a month? I think it is to some people, and not to others. Obviously, $50 a month will make it a lot more attractive than $500 a month. Bear in mind that huge numbers of people voluntarily pay generally $80 to $120 a month for the benefits of a smartphone — which, don’t get me wrong, I like smartphones, but they’re hardly necessities. People will pay in at least the lower range of all these values for convenience.
So basically: at a first pass, it seems likely that the dominance of the taxi model will depend a lot on how much a sensor system costs. If it’s pretty cheap (and/or doesn’t lose much value over time), then I suspect the ownership model will stay strong. If it’s fairly expensive, then I’d expect more people to migrate to a taxi model. If it’s super-expensive, then we might not see very much uptake of driverless cars at all.
At second pass, things get more complicated. Driverless cars would change society. If transportation gets easy, would people again flee city cores, choosing to live more suburban or ex-urban lives (with the advantage that getting into the city core would be easier and more convenient than ever before)? If so, that will diminish the attractiveness of the taxi model. Will cities reclaim spaces currently used for parking? If so, that will diminish the attractiveness of the ownership model (because your car would have to go way far away to park itself, and then maybe it would actually be faster for you to get a taxi to come pick you up than for your car to do it). What will happen to the core costs of cars in the (potentially long) time before driverless cars are really ready to go? Etc.
One thing you can be confident of: it’s more complicated than most people are currently making out.