Inside the Nudge Unit

even those of us involved knew enough about cognitive biases to be wary of overconfidence. Thus one of the team’s first decisions was to plan its own demise if things did not work out. Well aware that many new government initiatives fail but yet linger indefinitely, the team built in a sunset clause. After two years, the BIT would be evaluated by the Cabinet Office and unless it could present solid evidence that it could produce results and save the British taxpayers money, it would be dissolved.

We tested whether adding a single sentence such as ‘most people pay their tax on time’ would boost repayment rates. It did –and by several percentage points, enough to bring forward tens of millions of pounds.

The team’s objectives read like a mission impossible: to transform the approach of at least two major departments; to inject a new and more realistic understanding of human behaviour across UK government; and to deliver at least a tenfold return on its cost. If it failed, it was to be shut down on its second anniversary –leaving enough time for voters to forget the whole embarrassing episode before the next election.

The team found that adding a simple (and true) statement on tax reminders that ‘most people pay their tax on time’ encouraged far more people to do so.

Getting the unemployed to think about what they could do in the next two weeks, instead of asking them what they had done in the previous two weeks, significantly increased the numbers off benefits at three months, getting tens of thousands back to work faster and trimming millions of days off benefits.

The US Office of War Information, created by Franklin D. Roosevelt, was reputed to have produced more than 200,000 different posters during the Second World War. These campaigns pursued a wide range of objectives, from encouraging people to buy war bonds; eating different, and previously unpalatable, foods; planting ‘victory gardens’; and bolstering the preparedness to fight and support the Allies. These campaigns were often highly successful. For example, billions of dollars were raised from US war bonds from the civilian population, including more than a billion from children alone. Similarly, vegetable production from the 50 million civilian ‘victory gardens’ is estimated to have exceeded that of commercial vegetable production.

while laws and punishments have often proved reasonably effective at getting people to stop doing something, they are often much less effective at getting people to start doing something, and certainly to persist with it.

A ‘nudge’ is essentially a means of encouraging or guiding behaviour, but without mandating or instructing, and ideally without the need for heavy financial incentives or sanctions.

Robert Cialdini, the author of Influence: The Psychology of Persuasion,

there’s a basic behavioural phenomena that applies to both people and organisations around ‘commitment’: people and organisations are much more likely to do something if they were previously engaged in even a small commitment or helped build it themselves (the IKEA effect). Cialdini gives a neat example, in his early popularisation of the psychology of influence, of a restaurant that was struggling with large numbers of customers failing to turn up for bookings. Staff taking phone bookings were instructed to make a seemingly tiny change: to pause after asking customers ‘Would you let us know if you can’t make it?’ The pause –imagine it in your own head –leads to customers to fill it with a response, such as ‘sure’, or just ‘uh-huh’. And the effect? The number of customers failing to turn up without calling more than halved.

The decision to set up BIT with a default that it be shut after two years was based on my own experience of government, plus the desire to create a sense of urgency and momentum. I had seen too many units set up by previous governments that had failed to deliver, but limped on because no one got around to shutting them down. This way we would have a clear, time-limited target. It also enabled us to live by our own principles: we had effectively switched around the normal default of government teams so that the Prime Minister and Cabinet Secretary would have to make an active choice to keep the team going

embrace empirical methods. You’ll need to demonstrate to sceptics that your new approach works, and to quantify the impact. But, more fundamentally, you should follow a logic of test, learn, adapt –behavioural science is well suited to experimental approaches, and human responses are complicated and hard to perfectly predict.

the head of a tax office in a mediumsized country will have 50–100,000 staff, and be responsible for collecting hundreds of billions of pounds in revenue that every other department will rely on. To those who have spent years oiling and tuning it, what to you look like a ‘few’ ingenious changes will to them look like a dangerous risk of throwing grit into the cogs of their great machine. They know perfectly well that if the changes you make lead to things going wrong, tax collection not happening, or mistakes being made, support from your political masters will evaporate. And chances are they’ll have to pick up the pieces.

Just like MINDSPACE, EAST is a mnemonic. If you want to encourage a behaviour, you should think about making it:  Easy.  Attract.  Social.  Timely.

If you want someone to do something –pay their taxes, recycle, or take on an extra employee –a pretty good start is to ‘make it easy’.

But in the real world of things, of people and of bureaucracy, friction matters a great deal. Just as a real weight pushed across a real table will soon grind to a halt as a result of friction, a human impulse to do something soon grinds to a halt when it becomes a hassle. Hence John, the character at the start of this chapter, really did mean to start saving, and to get that ‘money on the table’. He just didn’t get around to it because it was an effort, involved tedious paperwork and was less attractive and urgent than all the other things he could be doing in the next hour or day. Frictional costs are not a peripheral issue. Rather, they often make all the difference between something happening or not, be it a stone rolling down a slope, or a policy succeeding or failing.

In 1980, the Federal Republic of Germany (West Germany) introduced spot fines on motorcyclists not wearing helmets. The primary motivation was to reduce head injuries, but it had an unexpected and dramatic impact in a totally different area: thefts. In the wake of the change, motorcycle thefts fell by 60 per cent, and stayed down.

Laboratory studies show that an easy-to-read message is not only more likely to be understood, it is also more likely to be believed. By way of a simple illustration, subjects are more likely to believe a statement as true when it is written in bold rather than standard text.

We found that tax letters written in plain English, with a clear, simple request at the beginning, could often be 200–300 per cent more effective than the originals we compared them with. And, generally speaking, we found that ‘less is more’. For example, we found that click-throughs by businesses to get further information in response to emails from a major government agency could be increased by 40 to 60 per cent by reducing the volume of text in the email. Similarly, we found that sign-ups –from stopping smoking to youth skills and employment programmes –were significantly increased when the website landing pages were simplified and de-cluttered.

Low-income families, with children approaching college age, were offered a free service to pre-fill the college application forms using their tax and income information. The researchers found that pre-filling the forms, which took no more than a few minutes and cost just a few dollars, increased university enrolment from these lower income families by around a quarter (34 to 42 per cent), and increased the proportion receiving a scholarship by around a third. 9 The researchers also found that giving information, the standard government response, made no difference at all. Clearly the intervention was extremely cost-effective: a great many forms can be pre-filled for the equivalent cost of an expensive new scholarship or outreach programme.

in one series of letters from our tax department, the HMRC, taxpayers were given a web address where they could find the online tax forms to complete. If the recipient typed in the address, it took them to the HMRC’s webpage with the tax form clearly on it. All the person had to do then was click on the form and, of course, fill it in. It turned out that just under one in five people who received these letters did indeed go on to fill in the form and pay their tax as a result. We ran a trial testing a tiny change. The trial compared the effect of the original letter to one almost identical except that the web-link took people directly to the form, instead of to the page with the form on it. In other words, it saved people a single click, albeit at the cost of a couple of extra characters of text. The result? A 22 per cent increase in the proportion of people completing their tax forms in response to the letter.

certainly give ‘making it easier’ your best shot before spending an extra billion or two on tax subsidies.

doctors who argue that alcoholic drinks should contain calorie labelling, rather than information about how many ‘units’ of alcohol they contain, the idea being that some individuals (unaware of what an unhealthy intake of alcohol actually constitutes) might cut down on their drinking when they realise a glass of wine is equivalent to a slice of cake.

if you ever are in such a crowded situation in need of help, a pretty good idea is to point or look at someone in particular and say ‘I need help!’

Cialdini et al.’ s work also gave clues about messages that might work even better. His work suggests that people are more influenced by the behaviour of those they see as being more like themselves than by people in general. For example, guests are more likely to reuse hotel towels, thereby saving energy, if told that previous guests had done so, rather than people in general. On this basis we tested the line that ‘most people in your local area pay their tax on time’, and found this raised the payment rate by more than 2 per cent over control. Similarly, a little known exception to the rule that the more litter on the ground, the more likely a person is to discard a leaflet is that you are even less likely to drop litter in an otherwise perfectly clean environment with a single piece of litter than one with no litter at all. The single piece of litter seems to remind us that littering is the exception, and that disposing of it properly is the norm. This finding helped to shape the line that ‘most people with a debt like yours’ had already paid (‘ debt norm’). This was even more effective than the local social norm. Combining the two approaches into one, explaining that most people in your area had already paid, and that you were one of the few yet to do so raised the repayment rate by 5.4 percentage points, or an impressive 16 per cent increase in payment rate relative to the control letter.

Several studies have indicated that eyes or faces looking at us tend to make us behave more virtuously. Hence, posters with faces looking at us have been shown to increase substantially the likelihood that people will, for instance, clear away their trays in a canteen or use an honesty box in an unsupervised environment, while other images and messages had little effect.

a clear example of the ‘big mistake’. The normal centrepiece of campaigns to get more women on boards is a statistic along the lines ‘isn’t it shocking that only 25 per cent of board members are women?’ (less in some countries). It is shocking, but it’s also likely to be a message that inadvertently normalises the situation. On the other hand, if such campaigns made the equally valid point that ‘90 per cent of companies have women on their boards’, then the signalling is very different.

It turns out that one of the most powerful, but underused tools in the policymakers’ armoury is simply to provide a more accurate echo chamber of what others are doing. Since, in general, policy is aligned behind what the majority of people are already doing –such as paying tax and not hurting others –it is a tool that can be applied widely.

The weight curves given to US parents to plot their child’s development, for some years, have not been the actual weight curves of US children. Rather, they are distributions of ‘healthy weight’, driven by the concern that if parents used the actual weight curves of US children it would exacerbate the problem of obesity even further

many teachers and leaders soon learn that they had better follow a ‘tight-loose’ rule: enforce the good behaviour you want to see early on and you will generally find that later you can be much more relaxed.

why put the signature at the end? What if forms asked you to sign an honesty declaration before, rather than after, filling them in? Dan Ariely, Max Bazerman and colleagues tested this idea in the USA in a series of lab studies and found that people were indeed less likely to cheat if they signed a declaration of honesty before rather than after the opportunity to cheat. 3 They also tested the idea in an elegant real-world study on car insurance, where drivers were required to estimate how many miles they were likely to drive in the coming, or typical, year. The more miles you drive, the higher your insurance is likely to be, so drivers do have an incentive to ‘underestimate’. When the signature was brought to the top of the form, drivers declared on average an extra 2,428 miles, or a little over 10 per cent more. This in turn cost the average driver around $ 97 extra, a significant amount. Signing the honesty declaration before they filled in the number made them significantly more honest.

BIT ran a trial, led by our environment lead Marcos Pelenur, with a large retailer in the UK expressing the energy efficiency of appliances in terms of how much they are likely to cost to run over the products’ typical lifespan (typically around nine years). 5 Sure enough, when labels had this extra information consumers tended to buy slightly more expensive but more efficient products, at least in a product class with high usage costs, such as washer-dryers. Clearly, the effectiveness of the intervention rests on supplying that information while the consumer is still deciding what to buy.

‘hyperbolic discounting’: the further into the future a cost or benefit, the disproportionately smaller it becomes relative to immediate costs and benefits. We also know that this discounting curve is not smooth, but drops away more sharply with the boundaries that our minds use to divide up the future. Viewed from Monday, three or four days away is pretty close, but viewed from a Friday, three or four days away is another world away –it’s next week.

once we’ve mentally labelled something as ours, we value it much more highly than we did before. That’s one of the reasons so many of us find our homes full of bits and pieces that we really should have thrown out long ago, and if we saw the same type of thing in other people’s homes we’d surely think they should throw them out.

self-control is like a precious good that we use up. Lab studies show that people are much more likely to choose chocolate cake rather than a healthy snack after taking part in a task that requires self-control. 12 Similar effects appear to impact on professional judgements, too. In a now famous study, Danziger et al. found that judges moved from around 65 per cent favourable parole judgements at the beginning of the day to close to zero by late morning. After lunch, positive judgements surged back up to 65 per cent again, before dropping away by the end of the day. 13 Recent studies have shown similar results with other professional groups. The incidence of hand-washing has been found to fall across care workers’ 12-hour shifts, with the drop accelerating the more stressful the shift was. Furthermore, the shorter the break before the shift started, the faster the rate of decline. 14 Similarly, doctors have been found to prescribe more antibiotics as the day wears on. The rate is partly reset after lunch before it drifts up further through the afternoon, ending at around 20 per cent higher than when the day began. 15 It’s an effort to say no to patients who you feel probably don’t need antibiotics, and an effort to keep washing your hands. We’ve seen similar effects in BIT analysis, too, such as the changing probability that social workers will bring a child into care through the week. Decision fatigue, as psychologists call it, has very wide effects and is often far more powerful and pervasive than we realise.

Effective interventions often bring together all three elements of timeliness. The intervention is targeted before the behaviour has become entrenched; the intervention is aligned to a moment when it is likely to be most salient or when the existing behaviour is disrupted; and its design will help the person overcome their own time inconsistency –helping them to do what their future self would have wished.

Michael Luca’s work illustrated how the ratings changed the nature of the market in more subtle ways. It showed how, as the coverage level of yelp.com increased in given areas, it disproportionately drove the growth of small, independent restaurants, whereas large chains were largely unaffected. In retrospect, we can see why. Imagine you are choosing a restaurant for lunch tomorrow. Chances are you have a good idea about the kind of food and service you will get in a familiar chain like Pizza Hut, but you know much less about the small independent pizza places around the corner. Is it worth taking a chance on one of the small independents? If you don’t know anything much about them, maybe not. But if you could find out how others had found them –and that they liked it –you might try one of them out.

Philip Tetlock has documented how, very often, political and policy experts aren’t very good at getting it right. 3 In particular, experts who have strong views of the world, with clear-cut but rigid theories, tend to make predictions that turn out to be wrong. In contrast, experts who make more accurate predictions tend to have much messier views of the world. They change their minds when new evidence comes along. They’re often full of doubt, so they don’t make great pundits on TV. But they’re more likely to get it right, because the world is, after all, a complicated place.

by some estimates as many of 80 per cent of jobs are filled by word of mouth.

A growing body of work suggests that if you prompt people to think ahead about how, when and where they will do something they are much more likely actually to do it. For example, if you want to encourage someone to vote, it’s much more effective to prompt them to make a plan about when and how they will get to the polling booth than simply appealing to their civic duty

We sought to give jobseekers a stronger sense of progress with a short checklist of all the steps between that stage and getting a job, with many of the tasks being ticked off even in the first session.

many as half of all published results are subsequently shown to have been ‘false’.

We know that extended unemployment is deeply damaging, leading to economic ‘scarring’ in the form of lower earnings even when people do get back to work, a scarring that can mark entire generations. We also know that unemployment has big impacts on well-being that far exceed those that result from the loss of earnings

Keynes’s ‘animal spirits’. Even if your business seems all right, with money in the bank and enough orders coming in, if you sense that other businesses around you are in trouble and there is no help at hand, then you will hold back. You won’t take on that extra employee, or invest in new equipment, if you think other businesses are in trouble.

businesses’ decisions are based on mental processes that are far from perfect, and in times of recession and doubt, much of policy is really about moving this sentiment. It may seem strange to think about it this way, but often the billions of pounds spent on schemes from tax breaks to quantitative easing (the printing of money) is more about sentiment than direct effect. If we think it’s getting better, or that someone has a plausible programme that might work to boost growth, then this itself can become a self-fulfilling prophecy, or at least one that will amplify the impact of the programme itself.

someone needs to make a final decision about which nudges are acceptable and which are not, and this should be someone other than the ‘nudgers’ themselves.

The clean-up of major oil spills counts as GDP growth, yet the depletion of the oil and the carbon it releases does not feature.

Psychologists have shown that most of us have a strong tendency to interpret the world through self-serving biases –i.e. we tend to take credit for things that go well, and blame others and the situation when things don’t. It seems that, to some extent, these biases help to keep us happy.

It is not for government to make these choices for people, or even to produce the citizens’ ‘trip adviser for life’. But governments –and potentially others –can help to generate such data, as the UK’s ONS has shown. They can, in the words of the 2010 Coalition Government, help find ‘… intelligent ways to encourage, support and enable people to make better choices for themselves’.

Better-informed consumers don’t just mean better functioning markets in the classical sense of lower prices or choosing lower fat desserts (see Chapter 7). It also implies the reshaping of consumer markets more deeply.

Those on the right choose to highlight the powerful role played by personal relationships, volunteering, freedom and control, and even organised religion (people who go to church are happier, though it’s the attendance rather than the religious belief that makes the difference). 43 In contrast, those on the left may choose to highlight the higher well-being returns to boosting income at lower income levels, the negative impacts of unemployment, and the role the state can play in reducing risk.

If you really want to achieve impact on a large scale, as psychologists have studied, it’s conversion not compliance that you’re after. For conversion, you need to persuade and convince, not force and insist.

‘Radical incrementalism’ is the idea that dramatic improvements can be achieved, and are more likely to be achieved, by systematically testing small variations in everything we do, rather than through dramatic leaps in the dark.

I hope that one of the things we’ll see emerging in the next couple of years is a more effective cross-national clearing house or platform that enables policymakers and practitioners from across the world to be better able to access and build the evidence on What Works. There is no shortage of opinions and even research, but patching together evidence is not a democratic process in the sense that evidence varies greatly in its quality, and that based on better and more robust methods should be accorded much more weight. Central to this platform will be a set of toolkits like that of the EEF we saw above (see Figure 44), but with an important extra column that shows the range of countries and places where the intervention was found to be effective (or not). A policymaker –or public service provider –can be much more confident about importing an intervention that has been replicated in five or six countries than one that has only been shown to work in one place alone.

We need to recognise our dangerous tendency to overconfidence and our presumption that what we do know is ‘right’. We need to follow in the footsteps of Archie Cochrane and Richard Feyman. We need to embrace doubt. We need to test, learn and adapt.

conference-goers fill up 68 per cent of their plates with the first three items they come across, regardless of whether the items are healthy fruit or a rich cooked breakfast.

a new ‘behavioural equilibrium’ is established that is reinforced by other people. A well-documented example of this phenomenon comes from interventions to encourage higher voter turnouts in the USA. Such interventions, such as prompting people to think about how, and at what time, they will go to the polling station, have been found not only to boost turnout in a coming election, but also in subsequent elections even without further intervention. Todd Rogers, at the Harvard Kennedy School of Government, has likened the persistence of the voting effect to moving the person into a ‘behavioural rip-tide’. 9 The intervention not only changes the person’s behaviour directly, it also moves it into the flow of other influences. In this example, once you become an active voter in the US, you come on to the ‘radar’ of local party activists who are then likely to reach out to you in subsequent elections. But it is also that you start to think of yourself as a voter; that you now know where the polling station is and how to get there; and perhaps that you start to become slightly more interested in politics and elections.

let me put it more strongly: I think it is unethical for governments not to do trials. 14 Governments, public bodies and businesses regularly make changes to what they do. Sometimes these changes are very extensive, such as when welfare systems are reformed, school curricula are overhauled, or professional guidelines are changed. No doubt those behind the changes think they are for the best. But without systematic testing, this is often little more than an educated guess. To me, this preparedness to make a change affecting millions of people without testing it is potentially far more unacceptable than the alternative of running trials that affect only a small number of people before imposing the change on everyone.

Once a month or so, when our Houses of Parliament are otherwise empty, we should ask 100 to 300 randomly chosen members of the public to come and help us decide on some new aspect or detail of lifestyle policy, just like the questions listed above. They should have the argument presented, hear from those against, and be asked to give a view. At the same time, government departments wrestling with behavioural and lifestyle issues would ask to have their issue presented at one of the ‘people’s parliaments’. The conclusion would be advisory, not binding, but governments would be expected to publicly explain how they followed or ignored the results.

Heads of government and business were traditionally not much interested in the details of forms and procedures that their citizens and customers had to deal with. Turns out that’s a big mistake. However good your strategy or product, if the person the product or intervention is aimed at has to wade through a poorly designed form or process, they will probably give up and do something else. Your strategists will be left back at HQ saying, ‘What went wrong?’ Getting these design details right has proved fertile ground for behavioural scientists. Much of what such scientists do is familiar to the design professions and ethnographers who study how real people interact with products and services. By understanding how people actually use services and products –what they like and what they get frustrated with –designers can reshape the products and services until they feel easier and more intuitive to use. But behavioural scientists have added two extra elements into the mix. First, they bring in the systematic study of how people think about the world and how they decide, including the mental shortcuts that we all rely on to get through everyday life. This gives the behavioural scientist a more informed idea about what might work better, including sometimes identifying seemingly counter-intuitive alternatives to the current set-up. Second, behavioural science brings in the practice of systematic testing and trialling. The combination, mixed with a little design flair, can prove highly effective (as we saw in Chapters 3 to 6). Addressing these design challenges is about more than ‘nudges’. It is about applying behavioural insights to everything that we do, from how information is presented to how an incentive or regulation is structured.

children who are steered towards believing that a test result is a measure of their inherent ability (‘ good result: you’re a smart kid’) show less persistence and lower subsequent performance on a difficult task than children who are steered towards believing that a test result is a measure of their effort (good result: good effort). In Dweck’s words, the latter type of feedback creates a ‘growth mindset’, or a theory of mind that personal achievements come from effort, 8 leading the child to try harder and not give up in the face of personal challenge.

Having money worries seems to preoccupy a chunk of our minds, even if we are not aware of it. The effect size on IQ is roughly equivalent to having had no sleep the night before, but on an ongoing basis. This gives useful clues to how welfare systems might be adapted, not just by throwing more money at them, but by avoiding systems and processes that factor in this unseen cognitive load.

EAST. If you want to encourage a behaviour –in yourself or in others –make it easy, attractive, harness social influence, and choose a time when most receptive.