Paternalism and Behavioral Economics
Coercive Paternalists, Libertarian Paternalists, and Antipaternalists
In the last fifty years or so, there has been an explosion of empirical work on how and when human beings depart from perfect rationality. This work has led, not surprisingly, to a rethinking of paternalism and its limits.
We now have three camps, more or less:
coercive paternalists, who urge that behavioral findings greatly strengthen arguments for mandates and bans (and leave John Stuart Mill in the dust, more or less);
libertarian paternalists, who urge that behavioral findings point to a host of freedom-preserving interventions, such as warnings, reminders, and automatic enrollment; and
antipaternalists, who urge that behavioral findings justify only, or at most, efforts to strengthen people’s capacities to make good choices.
It is important to see that each of the three views can be taken as a dogma, or a fighting faith, or instead as a presumption or an inclination.
For example, you could be a libertarian paternalist while also liking some mandates and bans (for example, compulsory seatbelt laws and social security laws). I like libertarian paternalism, but I certainly agree that there is a place for mandates and bans, even to protect people from their own mistakes. You could be an antipaternalist while also liking some nudges (for example, warnings about allergens). Still, presumptions and inclinations matter a lot.
A whole book could easily be written on the underlying debates. (I may have written one; who knows?) My main purpose here is far more modest. It is to put members of the three camps in the same room, so to speak, and to see what they might have to say to each other.
Libertarian Paternalism
Among many others, Richard Thaler and I like libertarian paternalism. We like GPS devices a lot. We would like to see their equivalents in many domains, helping people to get where they want to go. We see nudges as libertarian, in the sense that they preserve freedom of choice, and also as paternalistic, in the sense that they may make up for a lack of information or a behavioral bias on the part of those who are being nudged.
We think that nudges should be designed so as to make people better off, as judged by themselves (the AJBT criterion).
With respect to behavioral biases: People might have self-control problems. They might be present-biased. They are subject to inertia. They might show unrealistic optimism. They might have limited attention. They might use heuristics that lead to systematic errors. (On biases and blunders, a terrific resource, by Thaler and Imas, is this: https://www.amazon.com/Winners-Curse-Behavioral-Economics-Anomalies/dp/1982165111/ref=tmm_hrd_swatch_0?_encoding=UTF8&dib_tag=se&dib=eyJ2IjoiMSJ9.7tNr7a9Y_1tWZRv8HKrCM_4Ij-lknKvwUInhZy0Oe_mxrM7GKZvsBkNN_qdDRfyeca_ByVyPVZrBC5XfVQ7Q1jj8h56bZAH0MjQAKExfhq1I8Mp_BcnEkKFcA6TQ83iTki73pDtV3QLBuZpXn3j4Z7m9H7pHLK3WhnoW3tzSBGcBRk57jLJMAmtbJjQr1xzG.y5h5A5Ibn3UqObURr1JLEpQJxTCkRRS53z-P3CtEVj8&qid=1762019093&sr=8-1)
In response, some nudges are educative. Consider a warning (“cigarette smoking causes cancer”) or a reminder (“your bill is due”). A label is an educative nudge. Educative nudges might overcome unrealistic optimism, inertia, or present bias.
Some nudges are architectural. Consider automatic enrollment in savings plans, or design of a website, or of a store, so as to encourage healthy choices. All over the world, we are seeing educative and architectural nudges, informed by behavioral science.
To antipaternalists, nudgers say: You cannot wish nudging away. Some form of choice architecture is inevitable, and it will nudge. No website lacks a design. Where nudges are optional, they are often an excellent idea. They can make people’s lives easier and better. They can promote autonomy, sometimes through education (educative nudges, and return to the GPS device), and sometimes by allowing people not to focus on a topic that bores or irritates them (consider automatic enrollment).
To coercive paternalists, nudgers ask: Why take people’s freedom away? Who do you think you are? Nudgers think that if people reject a nudge, they may well have good or legitimate reasons for doing so. It is their life, after all, and they might well know best how to live it.
Still, nudgers ought not to be closed to reasonable arguments here. Perhaps we know that trans fats cause health problems. Perhaps we know that people would benefit from wearing seatbelts or from saving for retirement. If the evidence is very strong, nudgers might support mandates and bans, even if the goal is to protect people against their own mistakes. (For externalities, of course, nudges are not enough.)
Coercive Paternalism
Sarah Conly, Ryan Bubb, and Richard Pildes have defended coercive paternalism, and it seems that George Loewenstein and Nick Chater are in the same camp. The basic idea is simple: We know that people make a lot of mistakes. We also know that those mistakes can harm or even endanger their lives. Isn’t it therefore odd to insist, as libertarian paternalists do, on freedom of choice? Isn’t it ironic that many people who are informed by modern behavioral findings end up being so choice-friendly?
In any case many nudges do little or no good. Coercive paternalists ask: Why tinker around the margins? Why not actually solve problems?
Coercive paternalism might seem easiest to justify if we are utilitarians or welfarists, assessing interventions by reference to their costs and benefits. Indeed, Conly asks for use of a cost-benefit test. If the benefits of paternalism outweigh the costs, she says: Paternalism is fine, and it might even be mandatory. (I am simplifying her argument.)
It is true that if we believe in personal autonomy, we might have problems with coercive paternalism. But Conly is a philosopher, and a superb one, and she believes that it is not possible, in the end, to use the idea of choice as a general trump card against coercive paternalism. (Here’s her book: https://www.amazon.com/Against-Autonomy-Justifying-Coercive-Paternalism/dp/1107024846)
To libertarian paternalists, coercive paternalists say: You are trimming your sails! (That is what Bubb and Pildes say.) Or: You are cowards, or temporizers, bowing to political winds. You are incrementalists. You are not doing much good. You lack the courage of your convictions. Having focused on human mistakes, you do not draw the logical conclusion, which is that coercive paternalism is mandatory, not forbidden.
To antipaternalists, coercive paternalists say: What world are you living on? Are you in some kind of cult? We know that people make choices that impair or even ruin their lives. Do you want people lying in ditches? Do you know nothing about behavioral research?
Antipaternalism
The antipaternalists include Ralph Hertwig, Riccardo Rebonato, and Mario Rizzo. Some antipaternalists argue that the behavioral findings do not prove what they are said to prove. On one view, people have complex utility functions, and what outsiders see as a mistake might be nothing of the sort. Maybe heuristics work well, and should be seen as a solution, not a problem. Maybe people really like brownies and fudge; it is not a matter of present bias. (This can be seen as a broadly Hayekian point.)
On another view, antipaternalism is justified by reference to the constant risk of government error. Maybe - some antipaternalists would agree - paternalism would be acceptable if public officials could be trusted. But they cannot. Often they do not know enough (Hayek again). Often their motivations are not at all good. (Consider the role of powerful private groups.)
Some antipaternalists point to the interest in liberty. They urge that people have a right to err, even if their errors can produce a lot of trouble. They add that people learn from their own mistakes. A population that is nudged is a population of sheep.
Some antipaternalists urge that what they call “boosts” are often better than nudges. In their view, the goal of behaviorally informed policy should be to improve people’s capacities to make decisions for themselves.
For example, financial literacy might be preferred to automatic enrollment in pension plans. We might teach statistical literacy more broadly. Instead of nudging people to make better food choices, we might teach people about how to choose better at restaurants or grocery stores. We need Boost Units, not Nudge Units.
To be sure, many people find it challenging to distinguish between boosts, favored by antipaternalists, and educative nudges, favored by libertarian paternalists. Those who explore boosts respond that they are not speaking of (for example) fuel economy or energy efficiency labels. They mean to improve people’s general capacities. Fact boxes can be counted as boosts.
To libertarian paternalists, antipaternalists tend to say: You suffer from hubris. You do not take freedom and agency seriously enough. Architectural nudges, such as default rules, might be clueless. They might fail to respect heterogeneity. They treat people as infants or as objects.
To coercive paternalists, antipaternalists say: You threaten both autonomy and welfare. If freedom can be overridden whenever government officials think that cost-benefits calls for overriding freedom, we are all in a world of trouble. We are fully aware that behavioral findings suggest that people sometimes err. But you are overreading those findings. In any case, the right response to error is not coercion, but help, in the form of efforts to boost people’s capacities.
Paths Forward
How should we choose among the three approaches, or (better) how should we identify the circumstances in which one or another approach makes best sense? One answer, favored by Hertwig (in some of his writings), and also by Conly, Bubb, and Pildes, and also the present writer, would be to ask about human welfare. What do we know about which approach would make people’s lives better, in context?
Another answer, favored by Jeremy Waldron (I think), would be to ask about the right conception of liberty or autonomy, and to explore whether one or another approach would be consistent with it. Kantians would be inclined to be skeptical of coercive paternalism, though they might well be amenable to libertarian paternalism, depending on its nature and scope.
One path forward, in my view, is to insist on a distinction between first-order agency and second-order agency. People make choices; but they also make choices about whether to make choices. Sometimes choice-making is not a lot of fun, which casts a new light (I think) on how to think about paternalism. If people would prefer not to make a choice, we should be inclined to respect that second-order preference, not because we are paternalistic but because we are not paternalistic.
This is of course a very large topic. What I have tried to do here is to sketch, very briefly, the grounds on which libertarian paternalists, coercive paternalists, and antipaternalists disagree with one another, with the hope of getting clarity on the contexts in which, and the assumptions on which, one or another approach might make best sense.