Do Nudges Really Work?

by Fabio Tollon

I have always been deeply impressed by the way behavioural nudges can promote socially desirable outcomes. From “opt-out” retirement plans, flies in urinals, and speed camera lotteries, nudges big and small can be a force for good. But not all nudges are created equal. Nudge theory has taken the world by storm (with organizations and governments using these techniques), and so you might be forgiven for thinking that these behavioural interventions get it right most of the time. Well, as is often the case, things do in fact go wrong in the world of nudging. Thankfully, however, things going wrong is a wonderful opportunity for us to learn and improve upon our theories.

Failed behavioural interventions are actually quite widespread, but as with nudges themselves, not all failures are created equal. While it is far beyond the scope of this article for me to go into each way behavioural interventions may fail, I will highlight a few. My reason for doing so is that better appreciation of how things go wrong can be used to better understand the relevant causal features of behavioural interventions, leading to the systematic improvement of such interventions.

The first kind of example I’d like to engage with is that of backfiring. As you might imagine, backfiring is when a specific behavioural intervention gets things exactly backwards. That is, the actual behavioural change is in the opposite direction than the desired one. As an example, many educational campaigns that aim at improving dietary choices quite naturally choose to provide information on the potentially dangerous effects of unhealthy eating. However, in practice, this strategy often results in people who already have strong concerns regarding their health (“dieters”) to actually display an increase in their consumption of unhealthy foods after receiving such information. A candidate hypothesis for this behavioural outcome is reactance, which (in psychology) is “an unpleasant motivational arousal (reaction) to offers, persons, rules, or regulations that threaten or eliminate specific behavioural freedoms”. In other words, “dieters”, even though they are concerned with their health, seem to not like their dietary options being restricted. Their response to such restriction is to indulge in the very behaviour that they ostensibly want to avoid. Reactance is also at play when people make use of reverse psychology, as such usage is an attempt to influence a person to choose the opposite of what is requested. Evidence like this means that we should take careful note of the way in which the desired behaviour is described. It seems that individuals have a negative psychological reaction when encountering overly prescriptive language, and so a suggestion might be to use more descriptive language. The latter would focus on the behaviour of the desired target group, as opposed to simply signaling which kinds of actions are to be encouraged or discouraged. Indeed, this was found to be the case regarding tax compliance:

“Informing taxpayers that more than 90% of taxpayers had already complied in full with their obligations under the tax law increased the likelihood of tax compliance fourfold, whereas threatening taxpayers with information about the risks of punishment for noncompliance had no effect.”

The study, which was conducted on more than 200 000 individuals in the UK, shows just how much the specific linguistic construction of messages can matter.

The second example concerns the ways in which the positive effects of a seemingly successful nudge in the short-term are offset by later undesirable behaviour. Evidence of this comes from charitable organizations sending reminders to potential donors in an attempt to increase donations. In the short-term, it was found that such interventions are indeed effective at increasing donations. However, this effect was offset by the fact that “avoidance behaviour” also increased, as people who had been contacted also had a higher chance of unsubscribing from the mailing list. Further evidence for this kind of error comes from the literature on dietary supplements. Here the authors found that participants who took a placebo (which they believed was a dietary supplement) were less inclined to participate in multiple forms of health-related behaviour.

“They expressed less desire to engage in exercise and more desire to engage in hedonic activities (Experiment 1), expressed greater preference for a buffet over an organic meal (Experiment 1), and walked less to benefit their health (Experiment 2) compared with participants who were told the pills were a placebo.”

Dietary supplements are explicitly marketed for their health benefits, and those who take them are therefore already predisposed to want to take care of their health. However, after taking the supplements these individuals appear to develop an inflated sense of their body’s resilience, which seems to hinder their ability or willingness to not engage in unhealthy behaviour. By attempting to “nudge” consumers by playing up the health benefit of certain products (in this case dietary supplements) there is a chance that this leads to the opposite kind of behaviour than was originally intended. While it is of course an open question as to what the exact psychological mechanisms are that give rise to these results (as I have only presented behavioural evidence here), it in any case seems that there is a loss of motivational control by the subject. Both of the aforementioned cases (donations and dietary supplements) point to the need for us to take the actual mechanisms that lead to behavioural change more seriously. More specifically, they call on researchers to take temporal aspects of change into account, and to reflect on the long-term implications of such change, as opposed to simply being concerned with whether behaviour can be successfully changed in the moment.

The third and final example concerns nudges that result in negative side effects. This is different from the example above in that negative side effects are not necessarily the inverse of what is desired. Evidence of this kind of mistake comes from environmental campaigns that aim to reduce household consumption of water. In one large-scale study it was found that such campaigns, while reducing water consumption (and therefore achieving the desired behavioural change), result in an accompanying increase in electricity consumption. Specifically,

“The results show that residents who received weekly feedback on their water consumption lowered their water use (6.0% on average), but at the same time increased their electricity consumption by 5.6% compared with control subjects.”

This suggests that we ought to have multiple evaluative criteria that span not just the range of behaviours associated with the intervention, but also consider behaviours that might be tangential or susceptible to influence from the targeted intervention. It also highlights how it is at times unproductive to frame calls for environmental concern in such simplistic terms. We might be better off slowly “greening” human subjectivities, so that instead of individuals believing that just by saving water they are doing enough, we cultivate a kind of psychological maturity in which agents understand that environmental concern is a necessary part of our lives. Such concern is not simply about “saving water” or “reducing consumption” (although it of course incorporates these things), but rather about wide-spread, structural changes and issues. Such changes, however, can only be accomplished by a “greening” of our psychology, which would allow us to better appreciate the interconnectedness of our actions, and not see such concern as a kind of cost-benefit analysis where my using energy efficient lightbulbs somehow justifies a 10 minute shower.

From what I have described above one might begin to feel disheartened regarding the efficacy of behavioural interventions. While it is true that I have spent my time here detailing failed behavioural interventions, my purpose has been to illuminate these failings so that we might learn from them. Even the smallest interventions can have a large impact, and it is therefore imperative that we leverage this fact in order to promote socially desirable outcomes. It is therefore important that we keep in mind the specific ways in which nudges might fail, as each failure will have its own unique solution.