Let Coase Give Us a Ride Home. By Sebastian Galiani and Gustavo Torrens

The technology that will be needed to produce fully self-driving cars (FSDCs) is now within reach. Engineers are working on prototypes and, sooner rather than later, will have a commercial version ready. However, there is an institutional barrier that may interfere with the adoption of this important technology: Insurance companies will be reluctant to provide insurance for FSDCs. While insurance contracts for FSDCs will no doubt entail a number of provisions that will differ from those of traditional car insurance contracts, we argue that the key problem is the lack of a clear definition and allocation of property rights. In other words, there is as yet no clear liability rule that would establish who would be liable if an FSDC were to be involved in an accident. As a consequence, the uncertainty associated with providing insurance for an FSDC is artificially high, and this is likely to be reflected in high insurance premiums. Fortunately, Ronald Coase comes to our rescue. Following his 1960 insightful work, we know that in these cases governments need only to define and enforce a clear liability rule. Once this has been done, markets will do their job, and insurance premiums will properly reflect the risk associated with using an FSDC. In principle it should not really matter if automakers or owners are liable. The outcome should be efficient in either case. However, we briefly discuss some reasons why one or the other alternative might potentially be preferable.

There are two alternative solutions to this problem:

Alternative I: The manufacturer is responsible for any damage produced by the FSDC if it is involved in an accident. The manufacturer will buy insurance and, in the event of an accident, the insurance company will compensate the victims. The cost associated with the insurance policy will be passed on to final consumers, who will pay a price for their FSDCs that includes these costs.

Alternative II: Each owner of an FSDC is responsible for any damage produced by his/her car. Owners must buy an insurance policy. In the event of an accident, the insurance company will compensate the victims.

The Coasean solution does not necessarily mean that there will be FSDCs in equilibrium. It is possible that the technology will be too costly, or there will be no demand because consumers like to drive or because these vehicles are too prone to accidents. In all these cases, consumers will not buy FSDCs. The key point is that, when they are making this decision, they will be doing so on the basis of the right prices. For example, if the technology is too costly, then the cost of developing a fully functional FSDC will be very high and the final retail price will therefore be much higher than the price of a regular car. Consequently, few consumers will be willing to buy an FSDC. But what it is important here is that the price of FSDCs will be very high because they are costly to develop and/or produce, rather than because the insurance premium is too high owing to the absence of a clear liability rule that can be applied in the event of an accident.

Suppose that some people like to drive their own cars. Then, each consumer will weigh the change in welfare associated with not being the driver and the difference between the price of an FSDC and the price of a regular car. In such an environment, it is possible that old-fashioned consumers will not want to buy FSDCs. But, here again, the key point is that, in order to make the right decision, consumers must be dealing with the right price differential. In equilibrium, the marginal consumer who is indifferent between an FSDC and a regular car will be the Pareto-optimal marginal one only if the price of an FSDC relative to a regular car reflects the true marginal rate of transformation between an FSDC and a regular car rather than reflecting the fact that the property rights associated with an automobile accident are properly assigned in the case of a regular car but not in the case of an FSDC.

Finally, suppose that FSDCs are more prone to accidents than regular cars. Even so, it is possible that the use of FSDCs would nonetheless represent the most efficient allocation. This could be the case if FSDCs are more convenient and their benefits outweigh the cost of additional accidents. The key point, once again, is that consumers internalize these extra costs. But the definition of a clear liability rule and the consequent setting of insurance premiums –and, hence, the prices of FSDCs and regular cars– at correct levels will fully incorporate these differences. This implies that consumers will make the Pareto-optimal decision when choosing between FSDCs and regular cars.

Which alternative is better? This is a second-order point. First, there is a moral hazard issue. With regular cars, the probability of an accident is mainly determined by drivers. In contrast, with FSDCs, the probability of an accident will be primarily determined by the computer program developed by the manufacturer. In order to reduce moral hazard problems, it seems reasonable for the people who have more direct control over the probability of an accident to be liable. Second, insurance companies might have market power since they deal with a large number of consumers and there are relatively few insurance companies. It may be that, if they have to bargain with a small number of manufacturers when setting the levels of insurance premiums, their market power will be reduced. Third, there is a political economy issue. If FSDCs entail a much lower risk of accidents than regular cars do, then it is possible that insurance companies might make a strategic decision to block FSDCs because they know that, in the long run, this innovation could kill their market. At the extreme, if FSDCs are equated with zero accidents, the car insurance business becomes obsolete. If this were to be the case, then at the very outset, we should allow manufacturers to self-insure.

Another possibility is that there will be a mix of FSDCs and regular cars because some consumers will not adopt the new technology. The fact that FSDCs and regular cars may have different probabilities of being involved in an accident is not a problem: As things stand, there are cars and drivers with different probabilities of becoming involved in an accident and the insurance market deals with this situation simply by adjusting its premiums.

Finally, it is worth considering the possibility of multiple equilibria. Suppose that, with FSDCs, the probability of an accident is zero, but only if all cars are FSDCs. This might be the case, for example, if FSDCs need a complete inter-FSDC communication network in order to avoid crashes. In other words, under this assumption, we cannot mix FSDCs and regular cars, so there are multiple equilibria. Either all cars are FSDCs or, alternatively, everyone uses regular cars and nobody can introduce an FSDC. This is a problem that the Coasean solution cannot fully solve. However, the Coasean solution can be of help in reaching the new equilibrium. Suppose that there are two technologies for FSDCs: an expensive one that makes FSDCs operable even when there are regular cars on the streets, and a cheaper network-based technology that operates only when all cars are FSDCs. The best outcome is for everyone to use network-based FSDCs. However, we are starting from a ‘lock in’ situation in which everybody uses regular cars. Allocating property rights may induce companies to develop the more expensive technology and gradually introduce FSDCs. Once a sizeable number of consumers have switched, society can compensate the remainder so that they switch to the cheaper network-based technology. Thus, the Coasean solution can pave the way for a transition from a less desirable equilibrium to a better one.



Coase, R. (1960): The Problem of Social Cost, Journal of Law and Economics.