What the Nudge Unit proved.

In February 2010, shortly before the UK general election, a small group of civil servants and academics began meeting in the Cabinet Office to plan something unusual.

The plan was to apply the findings of behavioral economics — the field that studies how people actually make decisions rather than how economists assumed they did — to the practical problems of government. Organ donation. Tax compliance. Job seeking. Energy consumption. Public health messaging. The hypothesis was that small, low-cost changes to the way choices were presented could produce large improvements in outcomes without regulation, without financial incentives, and without telling anyone what to do.

The team had a name — the Behavioural Insights Team — and a mandate to prove its value or be disbanded within two years.

The first experiments

The BIT's early work was deliberately modest. Its members knew they needed wins, and they chose experiments where the evidence from academic research was strongest and the barriers to implementation were lowest.

The first major success came from tax compliance. HM Revenue & Customs was struggling to collect tax from people who had received letters about overdue payments and simply not responded. A small randomized experiment tested whether different letter variants could change response rates.

The winning variant — which told recipients that most people in their local area had already paid their taxes — increased response rates by five percentage points. At the scale of HMRC's operations, this translated to hundreds of millions of pounds in additional tax collected annually. The intervention cost essentially nothing.

A second early experiment concerned organ donation. Britain had an opt-in system: people had to actively register to donate organs after death. BIT tested different messages on the organ donation registration website, asking whether telling people that thousands of patients needed donors, or that many of their neighbors were already registered, changed registration rates. The social norm message outperformed all others.

These were small experiments. They were also, from a government perspective, revolutionary — not because the interventions were clever, but because they were *tested*. Government had been writing letters and designing websites for a century without systematically measuring which versions worked better.

The EAST framework

As BIT grew — it was spun out as a social purpose company in 2014, with the UK government retaining a stake — it developed a practical framework for applying behavioral science to policy problems.

EAST: Easy, Attractive, Social, Timely.

**Easy** — if you want someone to do something, make it the path of least resistance. Reduce friction, simplify forms, use defaults in favor of the desired behavior.

**Attractive** — draw attention to the choice point through personalization, imagery, or salience. Generic messages are processed as noise.

**Social** — show that others are doing the desired thing. Social norms are among the most consistent behavioral levers in the literature.

**Timely** — intervene at the moment when people are most receptive. A message about pension savings is more effective when someone has just started a new job than after they've been in a role for a decade.

EAST is not a theory. It is a diagnostic checklist — a way of asking, systematically, what barriers might be preventing a behavior that people would generally support if those barriers were removed.

The scope of what was learned

By 2020, BIT had run more than 750 randomized trials across governments in the UK, the United States, Australia, Singapore, and more than 30 other countries. Its staff had grown from 7 to more than 200. The findings were neither uniformly positive nor uniformly predictable.

Some patterns were extremely robust — simplification and friction reduction consistently improved take-up of beneficial programs. Social norm messages consistently improved tax compliance and energy use. Personalization consistently outperformed generic messaging.

Other hypotheses failed. Financial literacy programs consistently showed no effect on financial behavior. Generic job training consistently showed no effect on earnings. Workplace wellness programs, despite their ubiquity, showed no consistent health effects. In each case, the failure was informative: the absence of an effect was a finding, not a disappointment.

The most important thing BIT proved was not any specific behavioral insight. It was that a government agency could operate as a learning institution — that randomized trials could be embedded in day-to-day operations, that negative findings could be published without institutional embarrassment, and that systematic experimentation could become routine rather than exceptional.

The institutional design

BIT's organizational structure mattered as much as its findings.

The team was deliberately small and protected from the normal incentives of civil service — where programs are rarely evaluated and never discontinued. It had a mandate to publish results, including null results, which removed the usual disincentive to run experiments that might fail.

It was positioned close to political power — initially in the Cabinet Office, with access to ministers — which gave it the ability to implement findings quickly when they were positive. And it had a two-year prove-or-disband mandate, which created urgency without creating permanence.

The model has been replicated imperfectly. Most government innovation units lack one or more of these elements: either they don't publish null results, or they don't have implementation authority, or they don't have access to senior decision-makers, or they have no accountability for outcomes. The full BIT model — rigorous evaluation, publication of all results, direct government access, and implementation authority — is rarer than its imitators suggest.

What it means for civic experimentation

The BIT story proves three things that are directly relevant to the ES model.

First, that behavioral science can be applied to government problems at scale. The academic findings from Kahneman, Thaler, Sunstein, Cialdini, and others were not laboratory artifacts. They replicated in real government operations, with real populations, affecting real outcomes.

Second, that governments can run randomized experiments as a matter of routine — not as exceptional research projects, but as ordinary operational practice. The barrier is not methodological. It is institutional: norms, incentives, and risk tolerance.

Third, that the learning institution is the outcome, not the insight. BIT's most valuable contribution is not any single finding — it is the demonstration that a government body can build a culture of experimentation, sustain it over time, and generate compounding knowledge. The unit's effectiveness grew over the decade not because behavioral science improved, but because the institution got better at asking questions, designing tests, and applying what it learned.

That is the model ES is trying to propagate — not at the national government level, but in the library systems, parks departments, school districts, and permit offices where most people interact with government most of the time.

Next issue: Elinor Ostrom and what she showed about governance that economists missed.