Nudge theory: what survives after a decade of policy experiments

In 2008, Richard Thaler and Cass Sunstein published Nudge. The premise: people don't make decisions like the rational agents of textbook economics; small changes in how choices are presented can produce large changes in what people choose. The policy implication: governments should use defaults, framings, and "choice architecture" to nudge citizens toward better outcomes.

The book became influential enough to spawn behavioral-insights units in dozens of governments. Thaler won a Nobel Prize. The framework has had fifteen years of practical testing. The evidence is less unambiguously supportive than the early enthusiasm suggested.

1. The most-cited successes

A handful of nudges have produced large, robust effects across multiple replications:

Organ donation defaults. Countries with opt-out organ donation systems have much higher consent rates than opt-in countries. The effect is enormous — often 90%+ vs. 15% — and has been replicated repeatedly (Johnson & Goldstein, 2003).

Retirement savings auto-enrollment. Automatic enrollment in 401(k) plans dramatically increases participation. Effects are large and durable (Madrian & Shea, 2001).

Tax-form deadline reminders. Sending taxpayers behaviorally framed reminders increases compliance. Effects are smaller but reliable (Behavioural Insights Team, 2014).

These are the showcase cases. They have driven much of the popular framing of nudge theory.

2. The 2022 meta-analysis

In 2022, Maximilian Maier and colleagues published a meta-analysis of 200+ nudge interventions, accounting for publication bias. The headline finding: the true effect size of nudges, after correcting for publication bias, was much smaller than the published literature suggested. The average effect was on the order of d = 0.05 — quite small (Maier et al., 2022).

This doesn't mean nudges don't work. It means the average nudge produces a small effect, and the well-known successes are unusually robust cases that have come to represent a literature dominated by much smaller and less reliable findings.

A 2022 response by DellaVigna and Linos, looking specifically at nudges deployed by government behavioral-insights units in the real world (where the original studies tended to have been done in carefully controlled lab conditions), found effects approximately one-third the size of comparable academic studies (DellaVigna & Linos, 2022).

3. The conditions

Nudges work best when:

• The desired behavior is one people already weakly want
• The cost of the new behavior is low
• The choice has a default that can be shifted

They are weakest or counterproductive when:

• The desired behavior conflicts with strongly-held preferences
• The choice involves substantial real costs (financial, social)
• Targeted users feel manipulated rather than helped

4. The political economy

A criticism that has aged better than the nudge enthusiasm: small behavioral interventions tend to replace larger structural changes in the policy discussion. When a country could mandate cleaner air or could redesign a form, the political pressure tends to favor the form redesign. The form redesign is cheaper, more politically palatable, and produces measurable but small effects. The mandate is harder and produces larger effects.

The "nudge instead of regulate" tendency is real, and it has limited what behavioral economics has accomplished at scale.

5. The honest position

Nudges work — modestly, in specific conditions, for specific outcomes. The framework remains useful for low-stakes, low-effort policy adjustments where people's existing preferences point in a roughly good direction. It's much weaker as a tool for changing behavior that runs against strong existing preferences or incentives.

The post-2022 consensus among behavioral economists is more humble than the 2010s public-policy enthusiasm. Most of the interesting policy problems are not nudge-shaped problems.