How could the perverse incentive to buy unsafe cars be mitigated without abandoning injury minimisation?
By deliberately blurring the rule — grouping vehicles into classes and ignoring safety differences within a class, so that buying a safer car within your class no longer costs you anything.
Whether such perverse incentives really would become a problem is not easy to answer in advance. One could counter them by departing partially from the goal of harm minimisation — deliberately making the rule slightly less optimal in order to keep the incentives pointing the right way.
The key observation is that the fix does not have to be all-or-nothing:
- Not every vehicle has to be treated as a "black box" (the extreme option, in which the car knows nothing about what it is hitting). More moderate deviations may be enough.
- One could, for example, sort vehicles into categories and ignore the safety differences between different models of the same class and roughly the same age.
- That would preserve the ability to distinguish an SUV from a small car, while within each class it would again make sense to maximise safety technology — because the extra safety no longer makes you a target.
Two honest caveats close the discussion: whether these and similar modifications could really solve the problem, and how much they would impair the effectiveness of injury-minimising programming, is a further question whose answer requires empirical data as well as argument. And whether it can be answered satisfactorily remains to be seen.
Tip: The general technique is worth keeping: when an optimal rule creates a bad incentive, coarsening the rule's inputs can restore the incentive at a small cost in optimality. Tax brackets, handicap systems and insurance risk pools all work this way.
Go deeper:
Wikipedia: Moral hazard — the insurance analogue, where deliberately coarsening the rule (deductibles, risk pools) is the standard remedy.