Driverless cars, asset utilization and the peak demand problem

Will driverless cars be revolutionary? The Economist’s Free Exchange blog surveys some arguments for and against. I have a pretty high bar for revolutionary—I’m not at all convinced the internal combustion engine revolutionized transport, let alone society more broadly. But I think they’ll have a big impact, because they’ll let us use our transport assets much more intensively, and that frees up whole parts of the economy to work on other more useful things.

Right now, we don’t use transport infrastructure very intensively. Most people who own a car probably drive it a couple of hours a day, max. In two-car households, a second car may only make it out on weekends. It’s not just cars: hit the motorway at 3am and you’ll find our roads massively underused too.

This, in a nutshell, is the peak demand problem. Our demand—both as individuals and in aggregate—for different services varies throughout the day, week and year. Everyone wants to drive at 9am, the peak of demand. Nobody wants to drive at 3am, the trough. But we have to size our system for the peak. So we have roads built for 9am, which are empty at 3am. Similarly on the individual level, you own a whole car, even though all you really need is two-car-hours a day.

This phenomenon is not limited to transport: it applies to nearly everything.

Aggregate peaks are everywhere: we build our power grid to deal with the hottest day in summer or the coldest day in winter, when climate control causes demand to spike. We build commercial office space to handle the 9-5 peak, but most is virtually deserted through the night. This waste dwarfs just about anything else in the economy.

Individual peaks abound too: I own a bed, but only use it about eight hours a day. My laptop gets maybe ten hours of use a day. An expensive camera, maybe an hour a week.

With individual peaks, if an asset is costly enough, people will find ways of increasing its utilization. Factories often run 24/7, in three shifts. Ryanair has a turnaround time of just 25 minutes, in an effort to maximise aircraft utilisation. Expensive assets also spawn rental businesses, which can improve asset utilisation: holiday homes, cars, corporate jets, computational power—all these can be bought by the hour or day.

But sometimes the rental solution just doesn’t work, because transaction costs are too high relative to the cost of the asset. An expensive camera, I might well rent. I won’t rent a bed, because it would be rather difficult to get it in and out of my house each night, and owning a bed is just not that expensive. Not to mention that demand for beds has something of an aggregate peak anyway, at least in any given time zone. And flying beds around the world from time zone to time zone is getting really quite silly, as ideas go.

Aggregate peaks are harder to deal with. In some cases we can use clever pricing to create incentives for people to shift their demand in time (I didn’t fly home from the UK to Australia for Christmas; I’ll go home in April and save £1000.) In many cases—especially with public infrastructure—we can’t or won’t do this, and the result tends to be excess capacity most times, and occasional shortages at peak times (which manifests, for example, as traffic jams at rush hour, mail delays at Christmas, blackouts or ‘load shedding’ during extreme weather).

Driverless cars help address both individual and aggregate demand variability in our transport system.

On the individual side, they could significantly cut the transaction costs of renting a car, reducing the number of cars we actually need. I don’t own a car; I rent one by the hour from Zipcar (which has also been in the news today). But this solution doesn’t work for everyone. For one thing, the cars are located in fixed spaces. I happen to live across the road from a Zipcar, but if I had to walk fifteen minutes, it might not be worth it. A business like Zipcar depends on a positive feedback: the more members they have in an area, the more cars they can place there, which means additional marginal members can be recruited, and so on. So seemingly small transaction costs like that 15 minute walk get amplified.

Now if my Zipcar were driverless, I could book it online, and it would be outside my door in a couple of minutes: before I’d even put my coat on and locked the house. And with a lot more Zipcars around, the chance of one not being available when I needed it would drop to virtually nil. In a world like that, far fewer people would own cars, and the cars that were left would be used far more intensively. If the utilization of the average private car went up from, say, 10% to 50% (remember there’s still an aggregate peak related to day/night, which is hard to smooth) we could keep 80% fewer cars. Apparently there are around 1 billion cars on earth. Say half of them are private, and we cut that by 80%. That’s 400m fewer cars. That would free up raw materials, factories and auto workers to do something else useful.

On the aggregate side, by taking humans out of the picture we could shift some transport tasks to off-peak times. It’s hard to say how big an impact that would be: a lot of goods transportation probably already happens at quiet times. But it could still be significant. Moreover, although this is not a peak issue, by allowing cars to travel much closer together in convoy, going driverless should allow us to defer—possibly indefinitely—capacity upgrades to our road network. Perhaps we might even reclaim some of the space in our towns and cities that is currently given over to roads.

So all in all, I think the economic impacts of driverless cars could eventually be very significant. And that’s just the ones I can easily imagine (as the Economist blog points out).

As a final thought, consider that in this driverless future, some people might still want to drive. On congested roads, one of the major costs to you using a road is the lost opportunity for someone else to use it (and in an efficient world, all roads are near to congested at times—or else the road was built too big). Some cities have a crude congestion charging system to reflect this. Right now, if you drive a car you displace one other car. If you drive a semi-trailer, perhaps you displace five or six other cars. But in the driverless future, the driverless cars must leave a lot of safety margin around potentially-unpredictable human drivers. So even as the driver of a small car, you might displace eight or ten driverless cars.

London’s congestion charge is going to get a whole lot more expensive.