Cultural Evolution / Human Development / Global Synthesis

Helping Social Change Spread

How social influence spreads change, and when it fails

0+adult participants in the team’s two published studies in Gezira State, Sudan

0considerations for planners, from Efferson, Vogt & Fehr (2020)

Executive Summary

Charles Efferson and Sonja Vogt study when an intervention’s effects spill over from the people it reaches to others who follow their lead, and what stops that from happening. They combine field, lab and online experiments with mathematical models. Much of their fieldwork has been on female genital cutting (FGC) in Sudan and on son bias, a preference for sons over daughters, in Armenia. People may not state their views on either topic openly, so Efferson and Vogt developed new ways to measure attitudes.

Who's behind this

Key terms

Two limits of the tipping-point approach

Cutting in Gezira did not match the coordination account
An influential approach in development treats harmful practices like FGC as social coordination norms. Because families want to match their neighbors, the argument goes, a community has to be pushed past a single tipping point, after which change spreads by itself. The approach implies that families within a community mostly do the same thing. In data that Efferson, Vogt and colleagues collected in 45 communities in Gezira State, Sudan, most communities had both families who cut and families who did not, and views varied widely. Models by Efferson, Vogt and Ernst Fehr show that ordinary differences between people, such as how strongly they respond to what others do, can severely limit spillovers. As people who respond weakly to their neighbors become more common, the scope for spillovers shrinks and eventually disappears, so a program that counts on spillovers may get none.
Backlash and polarization
Change that spreads from within a community might lower the risk of backlash, but Efferson and Vogt know of no study that tests this. They and Fehr argue that an intervention that makes people feel their culture is under attack can lead some to reject the new behavior to assert their cultural identity, which works against spillovers. In an online experiment by Sönke Ehret, Sara Constantino, Elke Weber, Efferson and Vogt, described below, partisan labels stopped change from spreading, and more polarized participants were less likely to switch.

How people differ, and whom to target

Son bias within households

In Armenia, the 2011 census counted more than 114 boys for every 100 girls among children under 9, against a natural ratio at birth of about 105. Matthias Schief, Vogt and Efferson measured son bias separately in husbands and wives, using questionnaires and implicit association tests with about 2,700 participants, including husbands’ mothers. On average, men were considerably more son-biased than women. The authors ask how reliably an intervention aimed at men would reach the more son-biased parent.

Whom to target

Programs to end cutting have often worked with the people and communities most open to change. In models by Efferson, Vogt and Fehr, a random, representative sample was “a more robust way to achieve spillovers” than a sample tilted toward people already open to change. The authors call targeting the most receptive people probably the worst strategy for spillovers, because it gives the intervention the easiest task and leaves the hardest one to social influence. Spillovers also depended on what people already thought. They were reliably large when most people were open to change, usually absent when most opposed it, and most sensitive to targeting in between.

Case study: FGC in Gezira, Sudan

Efferson and Vogt’s field studies of FGC in Gezira State, Sudan, were run with Ernst Fehr and Sudan-based colleagues Amy Elhadi, Nadia Ahmed Mohmmed Zaid and Hilal El Fadil Ahmed, together with UNICEF Sudan and Sudan’s National and Gezira State Councils for Child Welfare. Governments and international agencies had promoted abandonment of FGC for decades, and international agencies often assumed the practice was pervasive and entrenched within communities.

The team’s two published studies had more than 10,000 adult participants combined: 2,260 in 45 communities for a 2015 paper in Science, and nearly 8,000 in 127 communities for a 2016 paper in Nature. The Swiss National Committee of UNICEF was the main funder of the Nature study, which was also part of a European Research Council grant. Its funders had no role in design, data collection, analysis or writing.

The team dramatized a debate already going on in these communities. They produced four 90-minute, telenovela-style movies about the same extended Sudanese family, with screenplays by W. O. B. Alalfi. In three, family members argue about whether to keep cutting their daughters. The fourth, without that storyline, was the control. In a first experiment with 189 adults, 67–73% of those who watched a cutting movie had positive implicit associations with uncut girls immediately afterward, against 36% of those who watched the control. A second experiment, across 122 communities, measured attitudes about a week after the screenings. Only the movie combining arguments about personal values and marriage prospects had a significant effect by then, a small one of about a tenth of a standard deviation in attitude scores, mainly among people already less negative toward uncut girls. The movie experiments measured attitudes, not whether families stopped cutting.

Social tipping in politics and the environment

Since the Sudan studies, Efferson and Vogt have studied social tipping, in which change by a committed minority becomes self-reinforcing and sets a new norm, in politics and in environmental behavior.

Identity can stop tipping
In the online experiment by Ehret and colleagues, run around the 2020 US election, all-Democrat and all-Republican groups first settled on a shared choice. Then about half the players in each group were given incentives to switch. With neutral labels, the switch spread through the group. In another version, Democrat groups had settled on an image of a triumphant Biden and Republican groups on one of a triumphant Trump, so switching meant choosing the other side’s image. Those groups remained in costly disagreement. The authors describe tipping as “powerful but unreliable.”
The same intervention can tip, have little effect, or backfire
For environmental behavior, a model by Joël Berger, Efferson and Vogt suggests the outcome depends on how costly the new behavior is and on people’s social preferences, such as how much they dislike being better or worse off than others. When the behavior is cheap, a small start can be enough for it to spread. When it is costly, it may not spread, and telling people the behavior is less common than they thought can lead some of them to stop.
Tipping isn’t always the best outcome
In models by Efferson, Sönke Ehret, Lukas von Flüe and Vogt, when some people strongly prefer the old way, an intervention that leaves a community split between old and new behaviors can produce a higher average payoff than one that moves everyone to the new behavior. This assumes that people’s own payoffs capture all the costs of the old behavior. If the old behavior also harms people whose interests those payoffs leave out, as human-rights arguments hold for the girls who are cut, the comparison can change.
Solar panels spread in clusters
A review of social norms and climate action, led by Sara Constantino and co-authored by Vogt, describes rooftop solar spreading in clusters in parts of Europe. In the cases it describes, one or two early adopters on a street were not enough, but three or four homes close together could get neighbors adopting too.

Helping people help themselves

Efferson, Vogt and Fehr write that spillovers could let a planner “help people to help themselves.” A limited intervention could start change that continues without a long campaign. Their models also show that, even where people conform to those around them, spillovers “can range from irrelevant to indispensable.” They call for “a concentrated effort to measure and evaluate how social influence and everyday heterogeneity combine.”

“Endogenous change limits the need for the policy-maker to meddle directly in the culture of the target population. This could reduce the cost of reversing a harmful tradition, increase the perceived legitimacy of change, reduce the tendency for people to conclude that their culture is under attack, and reduce intercultural strife and the associated risk of backlash.”

Efferson, Vogt & Fehr (2020)

Nature Human Behaviour

Three considerations from their models

1. How attitudes vary
In Efferson, Vogt and Fehr’s models, how preferences are distributed within a community largely decides whether any spillover is possible. They call understanding that variation “essential.”
2. A random, representative target
Where like-minded people cluster, a random target is likely to seed each cluster with some people who change. In their models this made it “the most robust approach.” It also needs no survey of preferences beforehand, only care not to over-recruit particular kinds of people.
3. The link between the practice and group identity
Where a practice marks who belongs to a group, people may keep it to stay unlike outsiders. In the models, this identity link worked against spillovers under nearly all conditions. It was hardest to overcome when the people promoting change were seen as outsiders.

Published by Center for Human Progress (US) Inc., a 501(c)(3) organization