A Box, and Eighty Percentage Points

In Austria, over 99% of people are registered organ donors. In Germany, next door, culturally similar and with similar survey attitudes to donation, the figure is around 12%.

The difference is a box. In Austria you are a donor unless you opt out; in Germany you are not unless you opt in.

Johnson and Goldstein's comparison of European countries found the same pattern everywhere: opt-out countries cluster above 85%, opt-in countries mostly below 30%, and the gap is far larger than any difference in stated attitudes. A default that takes thirty seconds to change is doing more work than everything anybody believes about donation.

Three mechanisms compound to produce this. Inertia: changing anything costs effort, however small, and small costs stop people who are not strongly motivated. Implied endorsement: a default signals what the designer thinks is normal or advisable, and people reasonably read it that way. And the reference point: once something is the status quo, moving away from it is coded as a loss, which loss aversion then amplifies.

Thaler and Sunstein built a policy programme on this and called it nudging, with a deliberately narrow definition. A nudge changes the choice architecture without removing any option or significantly changing incentives, and it must be easy and cheap to avoid. A default qualifies. A tax does not, because it changes incentives. A ban does not, because it removes an option.

That narrowness matters, because the political defence (libertarian paternalism, steering people toward their own goals while leaving them free) depends entirely on the escape being genuinely easy. A "nudge" that is hard to avoid is just a mandate with better public relations.

The most successful application on record is Save More Tomorrow, designed by Thaler and Benartzi. Employees are enrolled in a pension by default, and contributions rise automatically with future pay rises rather than immediately. The second half is the clever part: because the increase comes out of a raise not yet received, it is never experienced as a cut in take-home pay, so present bias is routed around rather than argued with. Participation and savings rates rose sharply, and the design has since been written into national policy in several countries.

Then the field got ahead of itself, and the correction is worth being honest about.

Nudging became enormously fashionable. Behavioural insights teams were set up in dozens of governments. Hundreds of studies reported effects. And in 2022 an analysis of the published literature found that once publication bias is corrected, the average nudge effect across the field shrinks close to zero. A separate comparison of academic studies with large-scale trials run by government behavioural units found the government trials producing effects around one sixth the size of the published academic ones.

Defaults survive this, and they are the exception rather than the rule. The honest summary is that a small number of specific mechanisms (defaults, friction removal, well-timed reminders) work reliably, and the general category was oversold by roughly the amount you would expect from a field that got a bestselling book before it got a replication crisis.

The objections to nudging as policy are worth taking seriously rather than dismissing, because the standard reply is correct and insufficient.

The standard reply is that there is always some choice architecture. The form has to be laid out somehow; the default has to be something. This is true. It does not answer the question, which is who sets it, toward what end, and whether the person being steered can see it happening. A defence of nudging that rests on inevitability licenses any architecture at all.

The second objection is substitution. Cheap behavioural interventions can crowd out structural ones that would do more, precisely because they are cheap and politically easy. If a benefit is not being taken up because the application is nine pages long, simplifying the form is excellent. If it is not being taken up because the benefit is too small, a reminder letter is a way of appearing to act.

Which brings up the intervention with the best evidence and the least glamour: removing friction.

Across take-up trials (benefits, college financial aid, vaccination, tax credits) reducing the number of steps and pre-filling what the state already knows consistently outperforms telling people about the thing. The effects are frequently large, in double-digit percentage points, where information campaigns produce a few.

Mullainathan and Shafir's framing is the useful one. Administrative burden is a tax, paid in time, attention and anxiety, and it falls most heavily on exactly the people with the least of those to spare. A nine-page form is not neutral between a comfortable applicant and a stressed one. Removing it is not a clever psychological trick; it is the removal of a charge that nobody decided to levy.