Four conditions that make me more comfortable with paternalism

Let’s talk about paternalism. Paternalism and behavioral economics are deeply intertwined, and here is why. If you believe that people are perfectly rational, the conclusion is simple: let them make whatever decisions they want. After all, every one of their decisions is rational, so who are we to interfere? But the moment you start saying that we are irrational — that we are not designed to think long-term, that we are easily tempted, that computing compound interest is not something our brains do well — you have opened a different door. If people are predictably irrational, then maybe we also need to help improve their decisions. Maybe, sometimes, we need to limit their options.

And of course, in many areas of life we already accept paternalism without blinking. We accept traffic laws; you can’t just drive wherever you want, at whatever speed you want. You want to use heroin? No, you can’t do that either. And yet you can invest your money almost anywhere you want, in almost anything, with almost no one standing in your way. So the interesting question is not whether paternalism is good or bad in general. The interesting question is: when do we feel more comfortable being paternalistic, and when should we insist on leaving people alone?

After thinking about this question for a while, here are the conditions that, for me, make paternalism more acceptable.

Condition one: the decision is big and consequential. Think about retirement savings. A mistake there is not a bad lunch; it’s a mistake that shapes decades of your life.

Condition two: the decision is hard to reverse. This is connected to size, but it’s not the same thing. If you get your retirement wrong, you discover it at age 65, and there is no version of “oops, let’s start over.” Some mistakes come with an undo button, and some don’t. The ones that don’t deserve more protection.

Condition three: it is hard to learn from experience. If you ask me whether I prefer broccoli to cabbage, I don’t need anyone’s help — I can try both, and over time I’ll figure it out. Trial and error works beautifully for vegetables. But there are domains where we simply don’t get a good learning process. Much of medicine is like this: as individuals, we don’t get enough tries, the feedback is slow and noisy, and by the time we’ve “learned,” the lesson may have cost us dearly. Where learning is impossible, the case for letting everyone learn on their own collapses.

Condition four: there is a large information gap. I’m buying a house; the person selling it knows a great deal about it, and I know very little. I walk into a doctor’s office; my doctor knows a lot, and I know almost nothing. When one side of a decision holds most of the cards, “let the individual decide freely” is a nicer-sounding arrangement than it is a fair one.

Put these four together — a big decision, hard to reverse, resistant to learning, with deep information asymmetry — and under those conditions I feel much more comfortable saying: let’s not make everyone figure this out alone. Let’s do the work collectively, figure out what is good for most people, and make that the standard. Not a prison, a standard.

What is the right path for retirement savings? Should every person work it out for themselves, or should we say: big decision, impossible to learn from experience — let’s set sensible defaults? What should we do with people who got injured and are trying to make their way back to work? Should each of them navigate that maze alone, or are there some things we should do together?

Notice that the same four conditions also tell us when to keep our hands off. Small decision? Reversible? Easy to learn from? Both sides equally informed? Then leave people alone — that is exactly the territory where freedom to make our own mistakes is how we learn, and where paternalism has no business.

Now it’s your turn. If you have other conditions that you think should be on this list, I want to hear them. And if you think there are behaviors where I’m wrong — where we should be less paternalistic even though they rank high on all four conditions — let me know.