Using Behavioural Insights More Effectively in Policy Interventions

ETHOS Digital Issue 05, Nov 2019

The last twenty years have produced a large body of research, many of them randomised controlled trials, showing that insights from behavioural economics can fruitfully be applied to public policy. So-called “behavioural insights” teams have mushroomed in many countries. Against an initial (and healthy) scepticism from economists, they demonstrated in many field experiments and policy pilots that behavioural insights (BI) can offer powerful policy tools. BI interventions play on motives that standard economic analysis overlooks or assumes away. BI interventions also often seem to have a convenience advantage over standard interventions. For example, they often only require making some behaviour visible, or adding messages, rather than active measures such as collecting fees or taxes. However, effectiveness and convenience are not always reliable indicators for quality, and deeper reflection is necessary to understand when BI should be used and when one should rely on policy tools suggested by standard economics.

Nudging Can Enable Standard Policy Interventions

Perhaps the most important contribution of BI interventions is to enable standard policy instruments, or to make them more effective. Consider the example of household electricity or water consumption. Study after study shows that households have very inelastic demand for electricity and water. This does not bode well for environmental policy, as taxes on electricity and water would have to be extraordinarily high to change behaviour in meaningful ways.

However, one needs to ask why the price sensitivity is so low. Is it because electricity and water are genuinely hard to substitute, or is there something special about electricity and water use? Our research suggests the latter: water and electricity use are difficult to perceive as one uses them, for instance while cooking, showering, or using the air conditioning. This lack of visibility leads individuals to give it too little weight in their decisions. Could this be the source of the low sensitivity to prices? In several studies, we found that simply providing individuals with a technology that makes their water use visible in real time has strong effects. For instance, in field experiments in Switzerland and Singapore, we found that individuals reduce water consumption by almost 20 percent if given real-time feedback on their water consumption in showers through a smart shower meter.¹ Showing individuals in real time how their behaviour translates into resource use had a very strong effect on behaviour, while price incentives (in the absence of such feedback) did not seem to have a strong effect.

Other evidence shows that households’ resource use becomes far more sensitive to prices and other interventions if they are provided with technology that helps them monitor their consumption. A study in the US gave households financial incentives to shift their electricity consumption away from peak periods to other times. There were two treatment conditions: in one, households were simply provided with the incentive. In the other, households additionally received an in-home display that allowed them to see their electricity use in real time.² The effect of price incentives were very small for the incentives-only group, in line with the previous evidence on low price sensitivity. However, households with an in-home display responded three times more strongly to the same incentives. Providing feedback, it seems, makes resource demand more price sensitive. This also applies to other types of interventions, as demonstrated by a field experiment we conducted in student residences in Germany. We provided one group with information on the carbon footprint of their showering. Another group received real-time feedback on their water use during shower, again using a smart shower meter. A third group received both real-time feedback on their water use and information on its carbon footprint. We found that providing information on the carbon footprint alone had no effect on behaviour. As in earlier studies, real-time feedback alone strongly reduced water use. Most interestingly, however, the combination of the two was most effective, leading to a reduction that was nearly twice as large as providing feedback alone. Again, feedback made behaviour more responsive: while information on the carbon footprint is ineffective by itself, it has a strong effect when combined with feedback, above and beyond what alone can do.

The general lesson here is that BI interventions, for instance through smart technologies, can give individuals better control over their behaviour, thereby making behaviour more responsive to standard policies. They help them make better decisions. For instance, it is highly unlikely that sustainability goals with respect to CO2 reduction can be achieved through BI interventions alone. Carbon pricing, in one form or another, will be a part. These incentives will lead to a greater behaviour change if individuals have effective tools that make their own behaviour easier to track. Thus, initiatives such as Singapore’s Smart Nation programme are likely delivering larger benefits than expected.

Nudging Isn’t Always Free

A widespread BI intervention is to play on individuals’ sense of pride or guilt by making their behaviour visible or by showing them how they fare relative to others. Such interventions are often quite effective, especially on individuals who are below average in desirable behaviours—such as going to the gym—or above average in undesirable behaviours. However, this effectiveness itself suggests that there may be a hidden cost to using BI in these circumstances. Take, for example, the case of gym visits: we could either make behaviour visible, thus achieving a large increase from individuals who feel shame because their gym visits are below average. Those above average will feel a bit better as well. However, because of loss aversion, the former effect outweighs the latter. For the overall population, this generates a psychological net cost, as losers lose more than winners gain. We could achieve the same number of gym visits by providing a financial incentive for each gym visit akin to many cashback schemes, thus enticing individuals to go more often. In contrast to the BI intervention, the cashback scheme could be financed by slightly increasing the membership price. Thus, no money gets wasted, while the incentivising effect is the same.

Economists from the University of Chicago illustrated this in a clever field experiment.³ They measured how much more often individuals went to the gym if their behaviour is made visible, through a published ranking to all gym members. As usual, BI interventions worked: people went to the gym more often. However, they also highlighted the psychological cost imposed by the intervention: below-average gym goers were shown to be willing to pay far more to stay off the ranking than above-average gym goers were willing to pay to appear on the ranking. Thus, introducing the BI intervention led to a net psychological cost. If cashback incentives were an option—and they clearly are in this case, they would be strictly preferable as they avoid the psychological cost.

BI policy tools are neither generally better, nor worse, than standard economic policy tools. In many cases, they complement each other, as the examples discussed above show. In other cases, BI interventions, albeit effective, may not be the preferred option if they generate a net psychological cost that standard interventions don’t produce. Most importantly, there is little guidance either way a priori.

Perhaps the most important contribution from the research in BI is the insistence on testing interventions in clean, randomised field trials before jumping to conclusions (as economists sometimes like to do). These trials have become increasingly sophisticated and valuable. Since there are no clear rules of thumb as to whether BI interventions work better in a particular setting, all new policies should be subjected to the same rigorous testing, whether BI is used or not. As prior examples—in Singapore and elsewhere—show, they make policy better by helping to pick BI interventions where appropriate and avoiding them where they are harmful.