The Nudge

Someone must decide what happens when the chooser does nothing. A pension form can leave the employee outside the plan until she signs, or place her inside until she declines. A menu has a first item; a notice has a heading; a form has boxes already filled or waiting to be filled. These arrangements precede the decision they appear merely to receive. Where the arrangement changes what people do, designing it becomes a form of influence, exercised even by the designer who thinks he has left everyone alone. The nudge begins from this observation and gives it a purpose: arrange the choosing so that people are more likely to reach an outcome that serves them, while preserving their ability to choose otherwise. The difficulty lies inside that last promise. An open alternative establishes one kind of freedom. Whether the path toward it has been made visible, manageable and worth taking requires a closer look.¹

Libertarian paternalism joins freedom of exit to a judgment about the chooser's good. Thaler and Sunstein defended steering that leaves options available and aims to improve the person's position, judged with attention to what fuller information, deliberation and self-command might change. Their defense depends on the familiar fact that choosing can be costly: the form is postponed, the instructions confuse, the preset investment looks like advice. Giving the designer no guidance leaves those influences in place. Yet the justification also demands a distinction between helping the chooser and recruiting her for someone else's end. A seller who arranges a subscription to increase revenue may use the same inertia as an employer who arranges a savings plan to assist employees. The shared instrument tells us how influence works. The purpose, the burdens of refusal and the consequences determine the case for using it.

The evidence on automatic enrollment shows how strongly a default can enter the decision while leaving the welfare question open. Madrian and Shea studied a large American corporation that changed its retirement-plan rules in 1998. New hires were enrolled unless they opted out; the selected contribution rate and investment allocation were also preset. Comparing employee cohorts, the authors found substantially higher participation and considerable persistence at those default settings. They separately examined the accompanying change in eligibility and described the limits of their comparisons. The historical result concerns behavior in that company under those rules. It gives no general instruction that every person should save at the preset rate or hold the preset portfolio. Indeed, the authors stressed that a default which brought employees into the plan could also hold them at a low contribution rate and a conservative allocation. Participation, the amount saved and its investment are different outcomes, each capable of responding to the same starting point.²

A default can increase enrollment and still harm some of the people it moves. Consider an illustrative hundred people: eighty would gain ten from joining a program, and twenty would lose thirty, measured against staying out by a stipulated informed evaluation. Suppose a fifth of each group follows the default without attending to the decision, while everyone else selects the outcome that serves them. With staying out as the default, sixty-four join and the total gain is six hundred and forty. With joining as the default, all eighty beneficiaries join, along with four people who lose thirty each: eighty-four join, and the total gain is six hundred and eighty. The change produces a net gain of forty while imposing losses on four people. If their loss were fifty apiece, the same enrollment figures would accompany a net deterioration of forty. The menu and the observed participation increase would be identical; the judgment would reverse.³

The evaluation therefore needs evidence that does more than count the choices the design has caused. In the illustration the gains and losses were given; in an actual institution they must be investigated. A person's informed explanation, later experience and repeated decisions under clearer conditions can bear on the judgment, without any one response becoming an infallible measure of her interest. The original defense of libertarian paternalism proposed comparisons of consequences and, where direct welfare inquiry proved difficult, proxies such as explicit decisions or the frequency of opting out. It also acknowledged that these proxies can inherit the very problems they are supposed to resolve. A majority's answer may depend on the starting point, and remaining enrolled may reflect the same inertia that prevented joining. Treating either response as decisive would let the architecture certify its own success.⁴

Preserving alternatives makes the power to steer more answerable, but it does not remove the power. The person who can decline a preset option must be able to recognize it, understand its consequence and act on the refusal. A default buried in an unreadable form and a plainly presented recommendation occupy different positions in that account. So do a design that saves the chooser work and one that deliberately makes departure wearisome. An institution can compare simpler instructions, active choice and different starting points, while counting the effort each requires. These comparisons preserve the practical force of the nudge's defense: some arrangement will be used, and an examined one can serve people better. They also preserve the chooser's claim against the architect: the convenience of steering cannot by itself establish the right destination.⁵

The nudge returns freedom to the circumstances in which a person exercises it. An option may remain formally open while its presentation makes it easy to overlook; an intervention may assist many while burdening a few; a deliberate arrangement may outperform an accidental one without acquiring authority over every end. The question is what the arrangement does, for whom, and on what understanding of their good. Keeping the exit open is a meaningful restraint on the designer. Making the benefit answerable to evidence is another. Together they leave the final question of this volume exactly where the act of choosing alone cannot settle it: when a person has selected what was offered, what have we learned about what she wants?