Links - Incentives

They're essential yet incomplete.

I talk a lot about incentives because they are extremely important, often underappreciated, and sadly easily misaligned.

Creating the framework of the right incentives can be very difficult. It requires defining the goals properly as well as incorporating the relevant knowledge into the system—“internalizing” it.

Mike Munger explains in this piece how two fundamental concepts, two Is, are required to work in concert for good outcomes. His example is using gold as a roofing material.

We all know why we don’t use gold, certainly the superior material for building roofs, and instead use wood and other composites. And yet we don’t exactly know why we use the alternatives instead of gold. That is the virtue and beauty of the market—called capitalism.

A political system is very different: resource allocation is decided by voting rather than pricing. But the incentives in voting are all perverse. Remember, almost everyone thinks that gold is a better roofing material than wood. And, to be fair, gold is a better roofing material, if you ignore all the other things gold can be used for! But if I don’t have gold, I have no way of obtaining it, because no one is going to vote to give me the gold; they all want it for themselves.

To claim it is obvious that you don’t use gold because it is too expensive—relative to alternatives—is to give the game away. Why is it more expensive? You really both do and do not know. Yes, it has alternative uses that are more valuable, but now why is that. And how do you know that?

We need information to answer these endless questions. And since resources are finite, we need a system that best automates the accumulation and transmission of that information.

And even that isn’t completely sufficient, as many well-meaning people might object . . .

Adam Mastroianni identifies this additional layer to our problem. Incentives work—that is both the key to solutions as well as a source of problems.

The problem with cynical theories of human behavior is that they do work, kind of, but in a stupid way.

When you rejigger incentives in the hopes of changing behavior, you attract the people who are most motivated by the incentives themselves, and these are the people you want to attract the least. Incentive-hunters are bent on Goodharting you, that is, doing exactly what it takes to extract the reward, even at the expense of what you actually wanted them to do.

In his piece he gives an overall strong argument and great assessment of how incentives, while powerful, are neither sufficient nor sufficiently desirable. Power cannot be contained without good incentives, but it will not be contained by incentives alone. We should strive for a world with systems and structures that use good incentives to foster and augment the good rather than for restraint of the bad and reliance to bring good.

Markets fail, use markets.