Libertarian Paternalism

What it is

The phrase is a deliberate contradiction. The libertarian half is the insistence that people should be free to do as they like, so long as they are not harming others, and free to opt out of arrangements they find undesirable. Thaler and Sunstein use the word as a modifier meaning liberty-preserving, and nothing stronger. The paternalist half is the claim that it is legitimate for those who design choices to influence behaviour in order to make people's lives longer, healthier and better.

The weight of the position rests on a single qualification: better as judged by the choosers themselves. The authors describe the result as a paternalism of means rather than of ends. The designer does not decide what a good life is. The designer tries to help a person reach what that person would have chosen with full attention, complete information, unlimited cognitive ability and complete self-control, conditions that no real chooser meets. That standard is easy to state and hard to verify in any particular case, which is where most of the argument about it takes place.

Thaler and Sunstein published a paper under the title Libertarian Paternalism in the American Economic Review in 2003, with a longer companion in a law review the same year. The behavioural picture behind it, set out five years later in Nudge, is that people err in predictable ways, so some presentation of options is unavoidable and the choice is between designing it carefully and designing it by accident. The position links two ideas that are usually kept apart: the case for choice architecture, and a limit on how that architecture may be used.

The 2021 Final Edition of Nudge changes the standing of the word libertarian. In earlier discussion it described the whole approach. In the revised book it becomes a test that a particular nudge either meets or fails. The authors concede that opting out can be difficult in practice, especially where there is sludge, and say that a nudge that does not meet the ideal is a form of sludge whose cost belongs in any assessment of the policy. They report that they nearly gave the book a different title built on the idea of one-click paternalism, which names the test: leaving should take about one click. That makes the position more specific and easier to criticise than the 2008 version, and it is the authors' own change.

The same chapter lists what they concede alongside it. Choice architects, public or private, are not always smart or knowledgeable, nor always well motivated. All of them are fallible and some may be malicious, and the research can be used by bad actors.

In effect

Organ donation, where the authors turned the test on themselves

The clearest application is one the authors made against a policy many readers assumed they favoured. Presumed consent treats everyone as a donor unless they opt out. In the Final Edition they say they do not support it, and the reason comes from the position itself. Some people strongly oppose anything being done to their bodies after death, yet may not know the policy exists. Inaction can reflect low salience, because people did not know they had a choice, or sludge, because opting out costs effort. A default whose exit is unnoticed is liberty-preserving in name only.

The pension default

Automatic enrolment is the textbook example of a default that keeps the exit open. Brigitte Madrian and Dennis Shea studied a large company that moved from opt-in to automatic enrolment and found that participation rose sharply. Each employee could still decline. Whether a given enrolment scheme passes the one-click test depends on how easy leaving is in practice, which is a fact about that scheme and not about the principle.

Commercial defaults

Pre-ticked renewals and privacy settings that start at maximum sharing use the same device. The principle supplies a way to sort them. Where leaving is easy and the default serves the user's own judgement, it fits the position. Where leaving is hard or the default serves the seller, the authors' own vocabulary calls it sludge.

What it does not say

It does not say that influence is harmless. The authors are explicit that designers are fallible and that some are malicious, and that the research can be misused.

It does not define the target in a way that can be measured. The standard is what a person would choose with full attention and complete self-control, which no one observes directly, so applying it always involves a judgement by the designer. The position is argued, not tested.

It does not rest on the size of nudge effects. Because the position is an ethical frame, the later disputes over the average effect of nudges, including the 2022 reanalyses of the nudge literature by Maximilian Maier and colleagues and by Barnabas Szaszi and colleagues, bear on whether nudging works and leave the question of what is permissible where it was.

It is not the same as society's view of the right choice. The authors' standard is the chooser's own judgement. Restating it as nudging people towards what society prefers turns it into ordinary paternalism, which is a different and more easily rejected claim.

---