• Geoff Mulgan

Nudges, behaviour change and complex systems

20 years ago I co-wrote a survey of behavioural policy options, that was then published by the UK Cabinet Office. It's now disappeared from the government websites and the Internet so here is the version of it I did a few years later in my book 'The Art of Public Strategy'. It tried to take a fairly comprehensive and systemic view. It should be read alongside the very useful accounts of work by BIT and more recent books like Damon Centola's work on behaviiour change.

Much of government is about changing peoples’ minds. Even more of it is about changing their behaviour – cutting water use or drink driving, or cultivating a sense of national identity. The UK health regulator NICE determines which treatments are cost effective and measures them in terms of costs per QALY (‘quality adjusted life year’). The available research shows that smoking cessation programmes cost around £500-1000 per QALY, compared to over £20,000 for Tamoxifen for breast cancer prevention and anything up to £800,000 for beta interferon for Multiple Sclerosis.[i] Clearly the right actions to change behaviour can achieve remarkable results and at relatively low cost (for NICE any treatment needs to cost under £20-30,000 per QALY to justify introduction within the NHS).

Yet changing behaviour is a delicate subject. No-one wants to be told that there’s something wrong with their behaviour. When I worked in the UK government I asked each department what it knew about behaviour change and commissioned a programme of research on what was known. It proved hard to track down who knew about the field, and what lessons had been learned. But even when the work was completed I felt wary about putting any of our conclusions into the public domain for fear of the scope for misunderstanding – deliberate or not. On the day that we eventually did publish a background paper the London Times ran a front page story misquoting a passage from the paper as proof that government was about to introduce a ‘fat tax’ to penalize the obese. Mild hysteria descended on the spin doctors and the health department (fortunately Tony Blair was always much more calm about the silliness of the media than many around him).

These sensitivities have not gone away. But strategies to change behaviour are less controversial. Many of us have driven too fast, drunk too much alcohol or dropped litter. But we want to live in societies without too much speeding or drunkenness and where streets are clean. Governments have many ways of changing behaviour, from the soft to the brutal and hard. They can introduce prison sentences or heavy penalties for drunk driving. They can catch more people who break the law with speed cameras for cars, or CCTV for litterers. They can raise the taxes on drinks, or wrapping paper, or highly polluting cars. They can make speeches or run advertising campaigns to frighten us, or persuade us to change.

There are few fields where governments aren’t trying to shape behaviour. Crime and antisocial behaviour are strongly influenced by people’s willingness to restrain themselves and to restrain others around them. Education can never simply be done to passive pupils (though some teaching methods have appeared to try): it also depends on students’ willingness to work and learn. Cutting levels of waste depends on householders’ willingness to separate their rubbish; cutting levels of traffic congestion depends on persuading people to use trains, buses, bicycles or trams rather than cars.

An ideal society might be one where people automatically policed themselves – saving governments the bother of fines and enforcers. Yet that is easier said than done, and it might make for a sadly anxious society. Instead the only alternative options lie in different mixes of sticks, carrots and persuasion to encourage people to take greater responsibility and do the right thing.[ii] In the field of welfare, successive governments have tried to persuade people to do more to find jobs and to acquire new skills rather than becoming dependent on benefits. In health it has become increasingly apparent that once the threat of infectious diseases is contained better health depends more on the lifestyle and behaviour of citizens (diet, exercise, smoking, drinking) than it does on the quality of health care. Banning smoking in public places has done more for life expectancy than many billions spent on pharmaceuticals.

Economic considerations have forced policy-makers to take seriously what might have been considered soft or marginal issues. The UK government’s review of long-term health funding in the early 2000s concluded that the public’s willingness to take responsibility for their own health would dramatically alter future funding pressures, and concluded that much more effort should be made to influence it.[iii] But strategies to change behaviour enter delicate philosophical territory. The appropriate division of responsibility between the individual, community and state has always been contested. In past decades there were furious arguments over whether the state should provide fluoride in the water, enforce the wearing of seat belts or force children to stay in school up to the age of 16. For many traditions of social and political thought greater personal responsibility is a good in itself because it enables society to function with a less coercive state and judicial system and allows public goods to be provided with a lower tax burden. The exercise of responsibility strengthens individual character and moral capacity and greater personal responsibility – in terms of restraint and support for others - enhances the quality of life of the whole community. But for others any suggestion of using state power to enhance personal responsibility smacks either of a conspiracy to pare back essential public services or of an excessively paternalistic desire to control peoples’ lives.

Many of the issues in this territory are strongly shaped by perceptions of fairness. It seems fair to impose the costs of a harm onto the person most responsible for producing it – so that smokers should pay more for the health needs of passive smokers; big companies should pay more for the pollution they cause. These judgements depend in part on perceptions of how much moral responsibility the perpetrator really has: a child is viewed differently from an adult, someone suffering from a genetic disorder will be seen in a more forgiving light than someone who has no excuse for their behaviour.

Public perceptions of what’s fair, and of the right balance between paternalism and responsibility, change over time. For example, the vast majority of the British public believe that it is the State’s responsibility to provide healthcare for the sick, but less than a third think that it is the State’s responsibility to provide a job for everyone who wants one. Many more now believe that the state has a responsibility to protect children or to punish domestic violence than in the past. Conversely more people expect to be able to buy or own their own house, or to choose their own pension.

Most governments have traditionally relied on two sets of tools to shape behaviours. One set are the tools of law, prohibition and coercion: forcing people to behave in a particular way because of the pain or harm that would result from penalties. Coercion works much of the time: it stops people from stealing, killing and raping. But prohibitions can notoriously fail, as happened to the US’ prohibition of alcohol use in the 1920s and the prohibitions of hard drug use that did little to stem rising demand from the 1960s to the 2000s.

The other set of tools are economic: incentives to reward one kind of behaviour over another. These are effective in some circumstances – for example in encouraging people to shift to more ecologically benign cars. The World Bank claims that in developed countries a 10% increase in cigarette prices reduces smoking by about 4%.[iv] Price signals have been used successfully to shift people onto unleaded petrol, to cut alcohol consumption. In the USA, high sales taxes on snack foods have also been shown to cut consumption. [v]

But incentives are often ineffective. Better incentives to save should encourage people to save more. Yet the available evidence suggests that while they encourage shifts between savings vehicles, from less to more tax efficient, they do little if anything to affect aggregate savings rates. Sometimes incentives can have perverse effects. For example, in order to reduce the number of parents dropping their children off late at an Israeli nursery, fines were introduced for late arrival. But rather than reduce lateness, the fines led to a marked increase in the number of children dropped off late. The fines led parents to feel that they were entitled to drop their children off late, since they were now paying for it.

An alternative approach lets government set the defaults for behaviour – making it as easy as possible to make choices that suit the individual’s interests and the communities’ interests, while not removing the freedom to make another choice. This has been particularly favoured in fields like pensions – with default state-guaranteed options that citizens can choose to opt out of – or health, where a doctors prescriptions come along with some freedom to choose differently (the phrase ‘libertarian paternalism’ - setting default options in the interests of the public but enabling them to opt for alternatives has been used to describe these options).[vi] These work in part by creating new social norms. A small scale but impressive example is the Irwell Valley Housing Association’s ‘Gold Service scheme’ which rewards tenants who pay their rent on time (and abide by various other commitments to good behaviour) with quicker emergency repairs, discounts on home insurance and faster access to improvements. The scheme has brought down costs, evictions and voids, and has worked by rewarding good behaviour and making this the norm, rather than through penalties.[vii]

There are many more rounded views of how human behaviour really works, and a rich literature that’s full of insights [viii] One school of thought emphasises conditioning – how people are shaped by rewards or punishments that follow closely on a particular behaviour. So speed cameras that reliably catch speeding drivers and make them immediately aware that they have been caught literally condition changed behaviour. Experimental evidence shows that the most difficult behaviour to change is that which has been learnt through a programme of increasingly intermittent rewards. This helps to explain puzzling phenomena, such as why people stay with abusive partners – the relationships were originally rewarding, but the intervals between the positive experiences gradually became larger (and may even follow episodes of abuse).

The very different ecological view looks at how individual choices are affected by interpersonal relationships (of people close at hand) as well as by broader social pressures and encouragements .[ix] Young people are more likely to take up smoking if their peers do – one study found a 1000% increase in smoking if two peers smoke, compared to a 26% increase if they have a parent who does .[x] Robert Cialdini has shown that people are twice as likely to litter if their environment is dirty because of the powerful impact of perceived group norms. Similarly if you want to persuade people to cut their air travel, don’t warn them that if everyone carries on flying the environment will collapse: instead try to persuade them that everyone else if taking fewer flights. Social marketing approaches draw on these ecological theories to change either the individual or the environment around the individual, or both. The changed behaviour of individuals and changed environment then, hopefully interact, gradually establishing new social norms.

Other approaches exploit people’s desire for cognitive consistency. Where there is a clash between their actions and their values, people often resolve the discrepancy by changing their values or attitudes rather than their behaviour. For example, if someone agrees to take on a boring task for a very limited reward, there is a dissonance between their behaviour (doing the task) and their reasoning (they would only do a boring task if there’s a decent reward). One way out of this dissonance is to stop doing the task by changing their behaviour. Another is to. convince themselves that the task is actually quite interesting .[xi]

Policy tools that exploit this aspect of human nature encourage people to make public commitments and then put pressure on them to live up to their commitments. These may be commitments to learn, to help their children with home work (in Parent-School contracts for example), to diet or to help others. Studies of the level of activity in ‘staged’ crime scenes show that individuals who agree to ‘watch over’ someone else’s property become over 400% more likely to attempt to prevent a theft than those who are aware that something is being stolen but have no such prior commitment to protecting it.

Another set of methods go with the grain of the heuristics, or mental shortcuts, that people use to make decisions [xii] People assume that events that they can easily call to mind, or that are easy to imagine, are more frequent and therefore likely to happen. Hence they tend to be more nervous about flying than driving because airplane crashes are easy to recall (even though hundreds of airlines have never had a fatality). Similarly, the larger the jackpot in a lottery, the more tickets are bought, because the consequences of a large prize attract more attention and are easy to imagine. Policies that emphasise individual success stories (such as disabled people who have triumphed in their careers) or that highlight dramatic risks (such as the violent effects of driving while drunk) are exploiting these heuristics. There is good evidence that people value things differently depending on whether they are gaining or losing them. Loss tends to be felt more keenly than gain. Messages stressing the potentially negative consequences of ill health tend to be more effective than those that phrase the benefits in terms of potential gains. Warning people that they stand to lose money if they don’t get more educational qualifications works better than promising that they could earn more.

In making judgements of this kind Daniel Kahneman has shown that everyone discounts the future but in very different ways .[xiii] So people living chaotic or impoverished lives apply especially high discount rates as a result of their immediate circumstances – making it less likely that they will make longer-term investments in their health, welfare, security or education. Hence policies that try to influence individuals’ investment in their future (e.g. promoting personal pensions or adult education) tend to widen inequalities as those with high discount rates fail to take up new opportunities.

Another set of approaches take advantage of the social context of behaviours – and try to influence those around the individual. To persuade sexually-active teenagers to use condoms they need to know what type of condoms work best and how to use them properly; to believe that potential sex partners won't reject them because they want to use condoms; and to have the confidence in themselves to state their wishes clearly before or during sex. Self-efficacy refers to people’s confidence in their ability to take action and to persist with that action .[xiv] Bandura’s research shows that self-efficacy can be increased by setting and rewarding small incremental goals, along with very overt monitoring and feedback. This works well for things like training patients to manage their own care. Surprisingly simple methods can have a big impact: recovering depressives who record on a daily basis the good things that have happened to them and what they owe to others get better much quicker than others.

This last example is a reminder that humans are social creatures. Reciprocity is a powerful social force. We respond to being placed in some sort of debt, even if unwillingly. This is the technique used in direct mail when a pen or pre-stamped envelope is supplied with a request for money. Wine tasting at vineyards works on a similar principle – a little is given free, but a lot is realised in return – though without a formal contract. It has been argued that the strongly generational pattern of civic behaviour in the USA resulted from the GI bill – the public funding of free college education for World War II veterans - which triggered a cycle of reciprocity that lasted a lifetime. There may be ways in which similar effects can be achieved through ‘social gifts’ such as educational bursaries, or publicly subsidized children’s trust funds, rather than couching such public expenditure in terms of ‘rights’ to services.

The importance of social interactions explains why face to face actions can be so much more powerful than impersonal ones delivered by letter or email. For example, individual face to face approaches have encouraged people to make greater use of alternatives to the car. In Perth in Australia door to door visits combined with peer pressure encouraged people to use their cars less. Some schemes ask volunteers to keep a travel diary to help them think about which journeys could be made by other means. Similar results have been found in terms of civic and political behaviour. A randomised experiment with 30,000 voters in the USA to see how voter turnout might be increased compared the effectiveness of leaflets, telephone campaigns and face-to-face reminders of a forthcoming election, all using a non-party political message highlighting the importance of voting .[xv] Leaflets had a modest effect, boosting turnout by around 2.5%. Telephone calls had, if anything, a negative effect. But face-to-face contact – someone turning up at your doorstep to remind you in advance – boosted turnout by around 10 to 15% .[xvi] Despite its relatively high cost, face-to-face contact was ultimately more cost-effective. Other methods take advantage of the extent to which people look for social proof (ie that others changing their behaviour in a similar way) or to the influence of an authority figure, the doctor, judge or policeman.

Some recent policy directions have used these insights and theories. The welfare to work programmes pioneered in Scandinavia introduced much greater conditionality into welfare and took advantage of aversion to loss (the threat that benefits would be removed), psychological discount rates (giving extra emphasis to immediate consequences), the authority of personal advisers and a sense of societal reciprocity. In health there has been much discussion over the years about making personal responsibility for health and the responsible use of health services more explicit by adding conditions, penalties or rewards alongside patients’ rights. Even modest compacts can take advantage of a similar sense of reciprocity, especially if they encourage patients to make visible commitments about future behaviour and reinforce the sense of new social norms taking shape.

These devices can balance the powerful psychological forces that stop us taking care of our own health. Discounting makes us disinclined to change our behaviour now for a long-term gain in health or longevity (rather the drink or burger today than the extra year tomorrow). Asymmetry of losses versus gains make us disinclined to give up our current satisfaction (smoking) for a potential gain (feeling fitter). And our psychological defences and attributions make us feel that early death and morbidity are things that happen to others, not us.

A good example of how psychology can be exploited is Thailand’s success in reducing the transmission of AIDS and other sexually transmitted diseases. This was achieved by a sustained, multilevel attempt to change social norms concerning condom use. The campaign combined consultation with national information campaigns, active engagement of at-risk groups (commitment and consistency), severe penalties for brothels not following safe practices (economic sanctions), and practices that empowered prostitutes to insist on condom use (self-efficacy). But perhaps the most important aspect of the programme was that together all the elements created a sense that habits were changing (social proof) and fostered the emergence of new social norms.

There are many other examples. Graduation incentives that pay a young person to stay on in school (such as Educational Maintenance Allowances in the UK) have shown that a positive conditionality can be more effective than the conventional negative forms. Restorative justice models in which offenders are held directly accountable to victims, and sometimes have to restore the losses they suffered, have also proven effective. They make the human harm and suffering caused more real and salient for the offender (availability and simulation), and they empower the victim by giving them an opportunity to express their feelings (self-efficacy and closure). Restorative justice creates opportunities for ‘reintegrative shaming’ – the offender is given an opportunity to apologise and make-up to the community the harm they have done, while the community can in turn forgive them; together these help offenders re-establish a more positive image of themselves both in their own mind and for others. Parenting classes have been found to be extremely effective both at changing parent’s behaviour and impacting on child behaviour – effects that are rapid and sustained. Anti-social behaviour has been reduced by 30% where parents of offending adolescents have attended parenting classes.

Changes in the physical environment can also influence the norms that affect crime. Even minor changes to street lighting can have an impact.[xvii] People are also influenced by how others are behaving. The term ‘broken windows’ was coined in New York for a style of policing which aimed to change social norms of acceptable behaviour by tackling visual signs of disorder, such as broken windows and burnt-out cars. These methods are given credit for cutting crimes as diverse as ‘threatening behaviour’ and ‘drug-dealing’. Finally, neighbourhood characteristics and community behaviour also influence how individuals behave. Work in the USA has shown how neighbourhoods with higher levels of ‘collective efficacy’ – where more neighbours know each other and are more likely to intervene in minor incivilities (such as children playing truant or teenagers hanging around) – suffer significantly lower levels of crime.[xviii] These positive effects are found even after controlling for socio-economic factors and prior levels of crime, suggesting the effect is causal.

Behaviour change isn’t yet a settled science. But the best behaviour change strategies draw on careful analysis of what is shaping behaviour to shape detailed interventions tailored to the very different motivations that people have. Crucial to this is detailed market research, quantitative analysis, and qualitative research using psychological profiling techniques to segment the population. To influence obesity, for example, populations have to be segmented not only according to their health status and current behaviour but also according to their capacity and willingness to change. Then, targeted strategies can be introduced for each group, combining the right mix of incentives and peer pressure, individual encouragements and group encouragements, alongside reinforcing messages from opinion leaders.

Psychologists do not yet have a prominent place in policy design. But looking to the future, these methods are likely to loom ever larger in the world of policy makers and strategists. Behind all of them lies the simple insight that dreams rest on habits: the freedoms that people aspire to depend on the habits that they and millions of others follow day by day.


[i] These are the fuller figures: smoking cessation programmes - counselling and nicotine replacement, £503-692/QALY (Source: NE Derbyshire PCT), £817-1,040/QALY (Source: 2005 Netherlands study). Tamoxifen for breast cancer prevention in high-risk groups: £21,7367/QALY (Source: Cybert, S; Obstetrics & Gynecology; Sept 2004). Beta Interferon for Multiple Sclerosis £39,972 to £810,481 (Source: NICE/ MS Research Trust) [ii] This chapter draws on several sources. My ESRC lecture in 1996 on ‘Coproduction and personal responsibility’; the Demos collection ‘Missionary Government’ which drew attention to the rising importance of culture and behaviour change in public policy; and the Strategy Unit paper ‘Personal responsibility and changing behaviour’ published in 2003, which was followed up by another paper on culture change in late 2007. [iii] Wanless D. (2002) Securing Our Future Health: Taking a Long-Term View. HM Treasury. [iv] TCurbing the epidemic. Governments and the economics of tobacco control.T Washington: The World Bank, 1999 [v] TPEffect of a snack tax on household soft drink expenditure, University of Wisconsin working paper .Tefft (2006) [vi] Sunstein, CS and Thaler, RH (2003) Libertarian paternalism is not an oxymoron. AEI-Brookings Working paper 03-2. [vii] For more information see [viii] One of the most useful recent books is Cialdini, R.B. (2001) Influence: science and practice. London: Allyn & Bacon. Other sources include Glanz, K., Rimer, B., Lewis, F.M. eds. (2002) Health Behaviour and Health Education: Theory Research and Practice. Fourth ed. San Francisco, CA: Jossey-Bass. Tversky, A. and Kahneman, D. (1974) ‘Judgment under uncertainty: Heuristics and biases.’ Science, 185: 1124--1131. Pavlov, I.P. (1927). Conditioned Reflexes. Oxford: Oxford University Press; Skinner, B.F. (1953). Science and Human Behavior. New York, NY: Macmillan. [ix] McLeroy, K.R., Bibeau, D., Steckler, A., Glanz, K., (1988) "An Ecological Perspective on Health Promotion Programs," Health Education Quarterly 14(4), 351-377; Reed, E. S. (1996). Encountering the World: Toward an Ecological Psychology. Oxford: Oxford University Press; Barker, R.G. (1968) Ecological psychology. Concepts and methods for studying the environment of human behavior. Stanford, CA: University of California Press. [x] Dudley, Haller and Portes (1978) ‘Peer influences on aspirations: a reinterpretation’ American Journal of Sociology 74, 2. [xi] Festinger, L. (1957) A Theory of Cognitive Dissonance. Stanford, CA: Stanford University Press. [xii] Tversky, A., & Kahneman. D. (1974) ‘Judgment under uncertainty: Heuristics and biases’. Science 185: 1124--1131. 185 Bettman, J. R. (1979) An Information Processing Theory of Consumer Choice. Reading, MA: Addison-Wesley. [xiii] Kahneman, D., Knetsch, L. and Thaler, R. H. (1990) Experimental Tests of the Endowment Effect and the Coase Theorem, Journal of Political Economy, Vol. 98 (6): p. 1325-48. 17 Rothman & Salovey (1997) and (1999). [xiv] Bandura, A. (1986) Social Foundations of Thought and Action: A Social Cognitive Theory. Englewood Cliffs, NJ: Prentice Hall. [xv] Gerber, A.S. and Green, D.P. (2000) ‘The effects of canvassing, telephone calls, and direct mail on voter turnout: a field experiment’ American Political Science Review, 94 (3): 653-663. [xvi] Gerber, A.S. and Green, D.P. (2000) cited above [xvii] Effects of improved street lighting on crime: a systematic review, Home Office Research Study 251, David P Farrington & Brandon C Welsh (2002) [xviii] See the work of Robert Sampson et al.