Nudge, nudge. Think, think. Say no more …

I hear the Nudge unit is in the news again …

I am waiting for the government to establish a Dig in the Ribs unit. Maybe even a Slap and Tickle unit, who knows?

Don’t be silly. Remind me what Nudge is again?

It started as a concept, “libertarian paternalism”, advanced by two American academics, Richard Thaler and Cass Sunstein. The idea was that the government could help people to help themselves without violating their liberty – for instance, by assuming they would like to make pension contributions unless otherwise stated. Then it became a book and the concept got a bit broader and a bit vaguer and more generally about the use of psychology and behavioural economics in policymaking. Then “Nudge” became a fashionable label to be slapped on any policy in search of a headline. Finally, David Cameron set up the Behavioural Insight Team – aka the Nudge unit – to do more research on the subject. The Cabinet Office published some of their findings this week.

This sounds an unpromising pedigree …

Well, the idea of using a popular economics book as a branding exercise certainly smacks of superficiality. But the idea of doing actual experiments to improve policy is a good one and, perhaps surprisingly, that’s what the Behavioural Insight team has been doing.

Why are you surprised?

I am surprised – pleasantly surprised – because these experiments risk failure and take time, neither of which are qualities calculated to endear them to politicians. The Nudge unit has managed to get around this by conducting some experiments on a large scale and examining short-term questions, such as whether people will respond to a letter about tax. This makes it possible to produce results very quickly.

For example?

Let’s say somebody has been fined in court but has not paid. You could send in the bailiffs. Or you could send a text message explaining that if the fine isn’t paid quickly, the bailiffs will be on their way. The Behavioural Insight team and the courts service ran a randomised trial, sending no text message to some people and a variety of text messages to others to see which approach works best. It turns out that text messages are highly effective and even more effective is a text message that mentions the miscreant’s name. The difference between no message and a personalised message is that instead of one in 20 people immediately paying up, one in three people do. That adds up to 150,000 occasions on which the bailiffs need not be called in.

This doesn’t sound like rocket science …

No, and it’s not brain surgery either. But it does appear to work. Sometimes these effects are mind-numbingly obvious. For instance, a letter sent by HM Revenue and Customs to chase up tax from doctors was vastly more effective after being written in a straightforward way with the key messages and request for action at the top of the letter. It was just as effective as an alternative that shoehorned in many fancy behavioural insights.

Why do we need these experiments at all, then?

There are a couple of reasons. The first is that human psychology is full of surprises. The report from the Cabinet Office contains no jaw-dropping discoveries but it does show that some sensible-sounding interventions have no effect, while others have very large effects. The second reason is that credible experiments tell a powerful story. I am told that the phone is ringing off the hook at the Behavioural Insight team office – government departments are queueing up to get some of that Nudgey goodness.

Nice for them, but I thought Nudge was all about persuading us to go to the gym

That has been its reputation, but so far the focus has been persuading people to pay taxes and fines and get a driving licence. A lot of the details are, frankly, pedestrian. But if core functions of government can be conducted more effectively the stakes are far from trivial. These experiments suggest that they can.

Do the experiments really stack up?

Most of them are work in progress. But while the Cabinet Office has obviously rushed to publish early results, they don’t look like a botched job. In an ideal world we’d get peer review, a registry of experiments and eventually systematic reviews. For now, I think we should be grateful that somebody is bothering to ask what works.

Also published at ft.com.