Are nudges worth the hype?

Have you ever heard about those little flies drawn in urinals? A few years ago, I hadn’t. Not until my grandfather gave me Misbehaving, a book by Richard Thaler, Nobel Prize winner and one of the founders of behavioural economics. I started reading it out of curiosity, looked up a bit more about the author, and discovered that he is also known for popularising the Nudge in another book of the same name, co-authored with Cass Sunstein.

A fly painted at the bottom of a urinal (Image generated with ChatGPT)

That is where I came across a story that has since become a classic: the tiny flies painted at the bottom of urinals in Schiphol airport in Amsterdam. To reduce splashes, someone had the idea of printing a little fly near the drain. Users started aiming at the fly. Splashes went down. Thaler calls this his “favourite illustration of a nudge”: you don’t ban anything, you don’t punish anyone, you simply make it easier to do the “right” thing.

I read this, I smile: it’s clever, slightly ridiculous. At the time, I would never have bet that I’d end up publishing a paper in Communications Psychology on behavioural public policies (not really about urinals, I’ll get to that…).

When behavioural public policies misfire

Behavioural public policies came back onto my radar properly in 2022. That year, a meta-analysis by Mertens and colleagues was published. It compiled dozens of studies on behavioural interventions and estimated their average effect. Using this method, you find that many behavioural interventions do have a positive effect. So far, so good. The problem is that several papers published in response showed that, as soon as you correct for publication bias (the fact that studies “that don’t work” are less likely to be published), that effect becomes much more uncertain.

Is all this just one big con dressed up as science? I don’t think so. In many contexts, behavioural policies have shown that they can change important behaviours, sometimes at relatively low cost. The real problem lies in how we use them: as a universal toolbox that can be almost mechanically applied to very different problems, on the grounds that “they work on average”. Yet we have known for a long time that the effects of these interventions are highly heterogeneous: some people respond a lot, others not at all.

In essence, this means we still do not really know when and for whom behavioural interventions truly work. Is that such a big deal? One might argue that we should “give them a chance”: as long as there is a positive average effect, isn’t that already good enough? That is more or less what I thought… until I came across a Swiss study that made me deeply uncomfortable.

In this study, Claus Ghesla and his co-authors look at green electricity default options. The idea is simple: if you set a “green” electricity contract as the default – one that is more expensive but emits less CO₂ – will people stick with it? The answer is yes: the default option works, emissions go down, the policy is effective on average.

But the authors do not stop there. They compare the contracts actually chosen under the default regime with the contracts the same people would choose if you removed that default and asked them to actively select an option. And at that point, the picture changes. The poorest households are particularly likely to stay in the green default option, even when that is not what they would have chosen spontaneously. They therefore pay more for their electricity than they would wish to. By contrast, wealthier households are more likely to opt out of the default and select a different contract. The very same intervention that reduces emissions therefore leads the most modest households to pay a higher bill than they would like, while leaving part of the potential contribution of the better-off untapped.

From there, a question becomes hard to avoid: when we celebrate the average effectiveness of a behavioural policy, who is picking up the bill? More broadly, instead of only asking whether behavioural interventions “work”, shouldn’t we be asking for whom these interventions are actually appropriate? To answer that question, we needed to change our level of analysis.

The psychology of poverty

To try and resolve this puzzle, I turned to a growing scientific literature: the psychology of poverty.

Living in poverty is not “just” about having less money. It means living in a context of economic, social and sometimes legal scarcity: less access to your rights, less power, fewer opportunities. It means living with urgent needs to meet and considerable uncertainty about the future.

Under those conditions, it is hardly surprising that the preferences and attitudes of individuals are systematically different from those of people living in more affluent environments. When you bring the studies together, robust patterns appear: people in poverty are, on average, more conformist, more present-oriented, feel less in control of their lives, and are more socially vigilant. What this literature emphasises is that these characteristics are not “irrational biases”: they are appropriate responses to risky, constrained and unpredictable contexts.

Once you see this, the question of behavioural interventions looks very different. All these motivations, which vary with living conditions, can determine whether a policy has a substantial effect or not on a given individual. This is no longer a technical detail: it means that some interventions, depending on the psychological profile of their target group, may reduce inequalities, and others, on the contrary, may worsen them, all while having a positive effect on average.

At that point, my co-authors and I decided to turn this line of questioning into a larger project. Rather than simply highlighting the ethical implications of behavioural interventions, we set out to propose a general framework: what does the psychology of poverty teach us about behavioural public policies?

The psychology behind behavioural interventions

In practice, we began by mapping out two things: (a) on one side, the main types of behavioural interventions (default options, self-regulation tools, information campaigns, social belonging interventions, social norms); (b) on the other, the psychological dimensions associated with poverty (conformism, present orientation, sense of agency, social vigilance). We formulated predictions about the heterogeneous effects we should expect, then confronted them with the existing literature. In many cases, the map and the territory matched fairly well. For default options, for example, the link with poverty became clearer.

Summary of our predictions

Why would default options tend to have a stronger average effect on poorer people? We know that people living in precarious conditions are, on average, more conformist. When you do not have much room for error, following what others do, or what is presented as “normal”, is one way of limiting risks. Opting out of a default option requires time, information, and a sense of legitimacy in choosing something else. All of that is often easier when you live in a more stable and affluent context.

So, a default option such as a green electricity contract is quite likely to have stronger effects in disadvantaged populations. The empirical literature seems to point in that direction. This is excellent news if the default is protective, subsidised and aligned with their priorities. It is much less encouraging if the default is costly, even in the name of a cause that policymakers or researchers consider noble.

We can apply similar reasoning to other tools. Present orientation, for instance, can make some self-regulation devices particularly helpful: reminders not to miss an important deadline, support in setting goals, public commitments to carry out an action. These tools can assist those for whom it is difficult to project themselves into the future, which is more common in contexts of poverty. Conversely, personal agency (feeling able to influence one’s life) largely determines the impact of large-scale information campaigns (on smoking, nutrition and so on). A policy that relies on reading, understanding, organising and changing one’s routines often benefits more those who already have the most resources at their disposal.

Everything seems to line up… until we turn to one last category of interventions: social norms.

Social norms

A descriptive social norm involves telling someone what other people around them do: ‘your neighbours use less water than you’, ‘most residents in your area recycle’, and so forth. This type of message can already have a direct effect: when we are doing worse than others on a behaviour that is morally valued, we risk feeling left behind and wanting to move closer to the norm.

We might think the story ends there: social norms mainly affect those who deviate most from the norm. But when you look at studies that also measure socio-economic status, the picture is more complicated. Some interventions appear to change the behaviour of the better-off more, others that of the worse-off. To understand what is going on, we have to distinguish between two types of social norms: private social norms and public social norms.

Private social norms allow individuals to compare their behaviour to that of others in a confidential way. For example, a letter sent to a household informing them of their water consumption relative to that of their neighbours. For that sort of intervention to work, people need to be willing to cooperate even if no one will ever see the effort they make. That often relies on a form of relatively unconditional cooperation, which we find more frequently in more affluent circles.

In a U.S. study by Ferraro and Miranda, the authors sent messages of this kind about water consumption. They observed that richer households reduced their consumption more after receiving these reports. In that case, it is of course hard to know whether this reduction is simply due to the fact that these households used more water to begin with, and therefore had more room to cut back. But the pattern is clear: this kind of tool draws on motivations that are more prevalent among better-off groups.

Public social norms, by contrast, make behaviour visible to others. For example, by requiring transparent rubbish bags that make it obvious whether people are sorting their waste. Akbulut-Yuksel and Boulatoff studied such a policy in Canada, the Clear Bag Policy. Their conclusion: the scheme increases recycling and reduces total waste, with stronger effects in lower socio-economic neighbourhoods. How can we interpret this? When reputation is explicitly at stake, people who are more socially vigilant, which is more common in contexts of poverty, may be particularly sensitive to such schemes. The very same vigilance that makes trust more costly can also make public behaviours more responsive to the gaze of others.

Taking people’s motivations seriously

What interests me in these examples is not to say ‘this intervention is good, that one is bad’. It is to show that as long as we do not look at how poverty shapes our psychology, we cannot really tell for whom a policy is a genuine nudge and for whom it is a burden

More deeply, what seems crucial to me is to take seriously the motivations of the people we want to help. Truly wanting to help someone in a precarious situation means accepting a more unpredictable environment, more urgent needs, and therefore a stronger preference for immediate and certain forms of support. It also means accepting that what appears “reasonable” from a more comfortable position is not necessarily reasonable from another vantage point.

Ultimately, these differences in motivation across life contexts are both a boon and a bane. A boon, because we can design schemes that build on these preferences to reduce inequalities: reminders that genuinely help those most at risk of missing a deadline, or interventions that help people set goals in ways that narrow achievement gaps. A bane, because the very same levers can be used, intentionally or not, to impose even more constraints on those with the least room for manoeuvre.

So, what are we supposed to do with all this, concretely? Behind these programmes, there are policymakers, researchers, consultants, but also all of us, as citizens who can vote, campaign, support or criticise projects. For me, this work is an invitation to go beyond our personal preferences. Beyond the reflex ‘this programme will work, it feels like it is going in the right direction’. Beyond our intuitions, towards examining who interventions affect, how, and at what cost.

What I have learnt from this project is that behavioural interventions truly succeed when they take seriously the motivations of the people they target and the context they come from.

And if we take that conclusion seriously, another question follows quite naturally: if what really matters are the motivations and the needs of the people we are trying to help… why not start by asking them what they need? Perhaps we would learn a great deal by talking more with them, by putting them more in the driving seat. Perhaps that will be the topic of a future research project.