What Is a Nudge? Changing Behaviour by Changing the Default

Richard Thaler and Cass Sunstein defined a nudge in 2008 as any aspect of the choice architecture that alters behaviour predictably without forbidding options or significantly changing economic incentives, and they insisted that the intervention must be cheap and easy to avoid. The example that made the case is organ donation. Countries where people must opt in to the donor register have consent rates that are frequently a fraction of those in countries where people are on the register unless they opt out, and the difference between neighbouring countries with similar cultures can be enormous. Nobody is forced and nobody is paid; only the default changed. Comparable effects appear in pension enrolment, where automatic enrolment with an opt-out raised participation dramatically in every country that adopted it, and the reason defaults work is a combination of inertia, the effort of deciding, and the implicit recommendation a default carries.

Behavioural units in governments and companies use a recognisable set of techniques:

• •Defaults, which set what happens if a person does nothing, and which are the single most powerful tool available
• •Simplification, since complexity causes people to postpone decisions indefinitely, so shortening a form or pre-filling it raises completion substantially
• •Social norms, telling people what others like them actually do, which has produced measurable effects on tax payment, energy use and antibiotic prescribing
• •Salience and timing, placing information where the decision is made rather than in advance, and prompting at the moment when action is possible
• •Framing, since the same fact described as a gain or a loss produces different choices, with losses weighing more heavily
• •Commitment devices, letting people bind their future selves, which works because people know their intentions will weaken
• •Feedback, giving people information about their own behaviour that they would otherwise never see

The honest assessment has become more modest than the early enthusiasm. Individual well-designed nudges produce real effects, and the effects are usually small in percentage terms while being extremely cheap, which is what makes them attractive: a letter redesign costing almost nothing that raises tax payment by a percentage point is excellent value even if it is not transformative. Systematic assessment complicates the picture. A comparison of results from academic publications with results from the large-scale trials run by behavioural units found that published effects were several times larger, which is consistent with publication bias, and that the real-world average effect is small. Several prominent findings have failed to replicate, and the field has been damaged by fraud cases involving researchers whose work on dishonesty was itself found to rest on fabricated data. The reasonable summary is that defaults and simplification work reliably, social norm messages work sometimes, and many of the more elaborate interventions do not.

Two criticisms are substantial. The first is ethical: nudging works by exploiting predictable failures of reasoning rather than by persuading, which treats citizens as objects of management, and it is not transparent in the way an argument or a tax is, since a person cannot easily notice that a default was chosen for them. Thaler and Sunstein's response, libertarian paternalism, is that some arrangement of choices is unavoidable, so the only question is whether it is arranged thoughtfully or accidentally, and that preserving the freedom to opt out is what makes it acceptable. The second objection is political: nudging is frequently proposed as an alternative to regulation, and critics argue that it substitutes cheap interventions for structural change, letting governments appear to act on obesity, climate or poverty while avoiding measures that would work better and cost more politically. The same techniques are used commercially with no such constraint, where dark patterns deliberately exploit the same tendencies to extract consent, purchases and data, which demonstrates that the toolkit is neutral and the intent is not.

A nudge changes how options are presented without removing any or changing their cost, and defaults are the most powerful version, which is why opt-out organ donor registers and automatic pension enrolment produce far larger shifts than incentives do. Simplification, social norms, framing and timing make up the rest of the toolkit. Real-world effects are small and cheap, several times smaller than published studies suggested, and the objections are that nudging bypasses reasoning and substitutes for regulation.