Whether it is trying to increase the welfare of citizens through public policy, or adapting organizational culture to a new business environment, changing behavior is one of the biggest challenges faced by managers and policymakers. Theoretically, harnessing the power of social influence is a highly cost-effective and relatively simple mechanism for achieving widespread behavioral change. In practice, however, the ability to use targeted interventions in this way to trigger broader behavioral change is far more complex than is often assumed.
In a policy setting, one relatively popular approach is to use social influence as a mechanism for behavioral change. If people are open to or motivated to behave like others around them, this interest in conforming, driven by incentives to coordinate their behavior with others in that population, can be harnessed to amplify the impact of an isolated intervention. The initial intervention changes the behavior of a group of people within a population, and this in turn, through the mechanism of social influence, changes the behavior of others in what is known as a spillover effect. In simple terms the policymaker is helping some people within a population to help other people in the population change behavior.
This approach offers several benefits. As the majority of behavioral change results from social influence, rather than external intervention, it is cost-effective. Change from within may confer greater legitimacy possibly than external intervention. And there may be less likelihood that people will believe that their traditions and culture are under attack, leading to a backlash.
Social influence has been shown to be an important element in a range of behaviors and practices such as domestic violence, smoking, alcohol consumption, energy conservation, child marriage and female genital cutting that policymakers might want to affect, for example. Recently it has also been seen in so-called “nudge” approaches to policy-making, where individuals are “nudged” towards a particular behavior, and theoretically social influence mechanisms may magnify that effect.
However, new research by Charles Efferson (HEC Lausanne, University of Lausanne) and his academic colleagues Sonja Vogt (University of Bern) and Ernst Fehr (University of Zurich) reveals that the ability to effect change through social influence is not as straightforward as it intuitively seems, or as policymakers might like. After extensive research in the field on a range of cultural practices that harm girls and women, including female genital mutilation, Efferson, and his colleagues were surprised to see that the result on the ground in terms of behavioral change did not reflect the success that policymakers might have expected.
Just because people at individual level appear open to conforming and coordinating with respect to a specific behavior does not mean policymakers can rely on the spillover effect to change behaviors across the population. Instead, the research suggests that the presence and size of a spillover effect depends on a number of characteristics about the population. For example, even if people are susceptible to conformity and coordination, they may have different trigger points or levels of sensitivity to social influence and behavior change. So it cannot be assumed, that a particular proportion of the population changing its behavior in a positive way will cause the rest to follow.
As a result preliminary research and analysis on the population to evaluate the potential for deploying social influence and the spillover effect is essential. Otherwise, there is a considerable risk that valuable time and expense will be wasted on designing and deploying an intervention that is intended to harness the power of spillovers when the spillover effect is unlikely to work.
The researchers note that the distribution of behaviors within a population can provide a number of clues about social influence and spillovers.
For example, in a situation where there is good behavior G and bad behavior B, if the local target population exhibits a mix of behaviors, with a relatively even distribution between those that choose G and those that choose B, this is unlikely to be a good target for a spillover effect. Because, assuming that people in the group routinely mix with others who think and behave differently, and social influence and the conformity and coordination mechanism was going to tip people towards one or another type of behavior, it should already have done so. The population would already be more polarized in respect of behavior G or B.
Secondly, when applying an intervention to a population with a range of thresholds among individuals at which they are likely to change behavior, the policymaker has to decide which subset of the population to target initially. The subset targeted can make a significant difference to the outcome. Intuitively it might seem sensible to target a subset of people most open to persuasion about changing behaviors. However, this would leave the spillover effect with the difficult task of converting the people more resistant to persuasion.
Alternatively, policymakers might choose to target a population subset more resistant to change, reasoning that the spillover effect would then be easier to obtain in the rest of the population. However it would already be difficult to gauge whether or not, and the extent to which, any initial intervention would be successful in changing the behavior of the tougher-to-persuade group. Or the size and scope of any intervention that would be required to change behavior.
A random approach
The best approach in this situation, conclude the researchers, is to target a truly random sample. This is on the basis that it is balanced in a way that moderates the problems of both an amenable target (it does not leave the hardest task for spillovers) and a resistant target (there is less risk that the effect of the initial intervention will be limited, thus restricting the spillover effect). This means that spillover effect has fewer people who are resistant to change to convert in the rest of the population. If, however, the people within the random sample who are resistant to change do not respond to the intervention, but instead prove resistant to various degrees, the spillover effect may be small or non-existent, regardless of the size of the intervention.
Other findings reveal that where a particular behavior or practice confers a particular cultural identity, this creates an additional barrier to a spillover effect. It provides additional intrinsic value to the practice and works against spillovers in nearly all situations. People carry out the practice as part of belonging to a particular ingroup, and distinguishing themselves from a particular outgroup. Policymakers can try to reduce the link between practice and cultural identity, possibly by transferring an association with cultural identity to another activity. Although this is particularly difficult if the attempt to transfer is made by someone from an outgroup.
Overall, the main message from the research is that using social influence as means of changing behaviors within a population may seem feasible from an initial cursory assessment of attitudes. However, just because people seem open to changing their behavior through social influence at an individual level, that is no guarantee of success at a wider population level. Instead, a cost-effective and sensible approach is to conduct a thorough analysis of how attitudes and behavior vary within versus between groups and then consider the findings in the context of this research.
Related research paper: Efferson, C., Vogt, S. & Fehr, E. The promise and the peril of using social influence to reverse harmful traditions. Nat Hum Behav 4, 55–68 (2020).