Social tipping points and forecasting norm change

James Andreoni, Nikos Nikiforakis, Simon Siegenthaler 30 April 2021

a

A

Social norms – informal rules about what actions should be rewarded or sanctioned – are ubiquitous in human societies. Some social norms have been around for so long that it is difficult to imagine life without them. And yet, these norms are sometimes replaced in favour of new rules that better reflect the preferences of the members of a society. Two prominent examples involve the abandonment of norms against homosexuality, and norms supporting gender discrimination in the workplace. As can be seen in Figure 1, social change can sometimes be rapid and nearly impossible to anticipate.

Figure 1 This is how fast the US changes its mind

Source: Tribou and Collins (2015)

For norm change to occur, there needs to be a social wave of individuals that moves against the status quo; for instance, against the ban on same-sex marriage and for the normalisation of being gay. If the group that mobilizes is large and visible, the social cost for embracing a new behaviour decreases and eventually reverses. A social tipping point occurs (e.g. Schelling 1978, Bicchieri 2016, Centola et al. 2018). An illustration of such a process is the rapidly increasing number of people in 2013 who signalled their support for same-sex marriage using red equal signs, and later rainbow flags, on their social-media profiles.

Figure 2 As the US Supreme Court took up arguments in key marriage rights cases in 2013, the red equal sign released by the Human Rights Campaign replaced more than 15 million profile pictures on social media platforms in support of same-sex marriage

Historical data clearly documents instances of sudden social change (e.g. Kuran 1991, Jones 2009, Amato et al. 2018). These data, however, do not permit us to identify models that can predict social tipping. This is problematic, as it is difficult to know if social norms accurately reflect societal attitudes at any given point in time, or whether there is a hysteresis which could call for an intervention to stimulate change. Recently, researchers started using experiments in controlled laboratories to study the process of social change (e.g., Centola et al. 2017, Smerdon et al. 2019, Bicchieri et al. 2020). In a new study (Andreon, Nikiforakis and Siegenthaler 2021), we present evidence from a large-scale lab experiment designed to test and empirically validate a widely used theoretical class of model for social tipping: threshold models. 

Threshold models

Threshold models assume that an individual’s willingness to deviate from a social norm depends on the proportion of others in the society that previously deviated from it. A society reaches a ‘tipping threshold’ when the proportion of people deviating from the norm becomes large enough that even individuals who are risk averse, conformist, or have pessimistic expectations about the prospects of change have an incentive to follow suit (e.g. Granovetter 1978, Schelling 1978, Efferson et al. 2020).

Intuitively, the size of the tipping threshold measures the entrenchment of an established norm. The higher the threshold, the more difficult it becomes for a society to abandon a norm. This relation, however, had not been tested, nor is there evidence to identify at what threshold a social norm can persist even at the detriment of most of a society’s members. 

Using numerical simulations, we show that in such a model there is a critical tipping threshold of 35% of the population, for plausible distributions of risk/conformity preferences and expectations. That is, when the benefits and costs of norm change are such that the tipping threshold falls below the critical threshold of 35%, then the probability of norm abandonment is predicted to be close to 100%, because enough individuals are willing to persist in leading change. In environments where incentives are such that the tipping threshold exceeds 35%, however, the probability of social tipping quickly drops to 0%, because the critical mass for triggering change is unlikely to accrue.

Experiment design

In Andreoni et al. (2021), we use a threshold model to predict social tipping in 54 experimental societies, each consisting of 20 individuals, and test different policy interventions that could promote norm change. The individuals interact over multiple periods in a social-tipping game with the following incentives: 

  • failing to choose the same behaviour as most others results in a penalty – depending on the treatment, either exogenously fixed or endogenously chosen by the participants – which creates a need for coordination and conformity, and 
  • preferences over alternatives gradually change over time, which creates a need for change.

That is, in the first round everyone prefers alternative A (equilibrium A), but preferences change over time until after a certain point, when nearly all members of a society would prefer change to alternative B (equilibrium B). The process by which preferences change is public knowledge, thus ruling out incorrect beliefs about others’ preferences as a reason for detrimental norm persistence (e.g. Granovetter 1978, Smerdon et al. 2019). Despite this, reaching a tipping point can be difficult given the pressure to conform and the history of adherence to the old norm, which affects individual expectations.

Persistence of detrimental norms

The data from the experiment demonstrates that societies can easily get caught in conformity traps. In particular, the baseline condition TT-43 implements a tipping threshold of 43% for which the model predicts no social tipping. The first panel in Figure 3 depicts behaviour for this condition in the six societies that participated in TT-43. The increasing solid line shows that the proportion of individuals who would prefer to abandon the old norm increased over time (the dashed curve shows the expected proportion). However, the green line with circled markers shows that the proportion of individuals who choose to abandon the norm never exceeded 20%. Hence, societies never reached the tipping threshold (given by the horizontal line at 43%), and norm abandonment did not occur.

Figure 3 Time series of norm abandonment in TT-43 with a tipping threshold of 43% (horizontal line) and for the condition where subjects chose endogenously the pressure to conform and thus the tipping threshold (TT-Endo)

Notes: Norm abandonment is shown as the line with circled markers. The dashed concave curve indicates the theoretically expected fraction of subjects preferring to abandon the norm; the solid increasing line the corresponding realised fraction.

The second panel in Figure 3 shows behaviour in condition TT-Endo, which is identical to TT-43 except that subjects could set the tipping threshold themselves by choosing how much others are penalised when failing to conform to the norm. In this condition, we did not impose strong incentives for coordination. Choosing a low penalty would correspond to a high tolerance for norm digressions. Strikingly, subjects set the tipping threshold too high for tipping to occur and five out of six societies failed to achieve a norm change. The reason is threefold: 

  • those who prefer the status quo (e.g. a ban on same-sex marriage) initially choose high penalties for norm deviators 
  • to prevent the costs associated with attempts at transitioning to the new norm, individuals who would prefer change also start to choose high penalties; hence 
  • change fails to occur even when nearly everyone would prefer the new alternative. That is, creating a hostile environment between those seeking change and those favouring conformity can perpetuate the status quo.1 

How to escape the conformity trap

Can policy promote beneficial norm change? The model suggests that effective interventions lower the tipping threshold. We examine this possibility in two conditions. Condition TT-30 implements a tipping threshold of 30% by increasing the benefit of change. In the field, this could correspond to an information campaign about the economic gains, health benefits, or scientific advances associated with a norm change. Condition TT-23 implements a tipping threshold of 23% by exogenously lowering the penalty for norm transgressions. In the field, this could correspond to campaigns that promote tolerance towards alternative beliefs and behaviours, or that illuminate the moral certitude of a particular stance. The first two panels in Figure 4 show that these conditions were indeed successful. In eleven out of twelve societies, the fraction of individuals deviating from the established norm increased over time until the tipping threshold was reached, and rapid norm change followed.

We also explored the efficacy of interventions that could affect expectations about change. In condition ‘public awareness’, we highlighted that all members of a society would prefer change by providing precise information about the predominant preferences in any given period. In ‘poll’, individuals could express their preferred social alternative via a poll that took place when a clear majority preferred abandoning the old norm. The lower two panels in Figure 4 show that, despite a high tipping threshold of 43%, these interventions led to change in ten out of twelve societies. Social tipping can therefore critically depend on a common understanding of and perspective on the benefits from change. For example, choosing marriage equality as the angle for gay rights helped to create more sympathy for gay people, and generated enough momentum in a sufficiently broad demographic to bring about the normalisation of being gay.

Figure 4 Time series of norm abandonment for the conditions with high benefit from change with a tipping threshold of 30% (TT-30), low cost of miscoordination with a tipping threshold of 23% (TT-23), and two conditions designed to affect collective expectations (Public Awareness and Poll)

Conclusion

Our research shows that when social change requires mass coordination, there can be a disjunction between what society really wants (e.g. granting marriage equality) and the outcome (e.g. great social costs for supporting gay rights). The change in social and institutional arrangements will not occur if only relatively few people mobilize. This may happen if people worry about the social costs due to their mobilization – which we show can be high even if a large majority would be in favour of change – or believe that the protests are unlikely to garner broad support.

Social learning is often not enough to promote change. Choices can mask true preferences. Indeed, in one condition, we show that increasing the speed of social feedback (e.g. through new technologies) can prevent rather than promote change.

Two factors consistently helped hasten beneficial change in our study. The first is a common understanding of the benefits from change, which can result from events that attract public attention, opinion polls that aggregate information, or by finding an angle on an issue that appeals to a broad demographic (e.g. marriage equality has proven to be fertile ground for the acceptance of gay rights). The second factor is perseverance. Social tipping in our setting critically depended on a group of leaders who persisted in moving against the established equilibrium, even at great personal cost.    

References

Schelling, T (1978), Micromotives and macrobehavior, New York: WW Norton & Company.

Bicchieri, C (2016), Norms in the wild: how to diagnose, measure, and change social norms, Oxford University Press.

Centola D et al. (2018), “Experimental evidence for tipping points in social convention”, Science 360, 1116–1119.

Kuran, T (1991), “The East European revolution of 1989: is it surprising that we were surprised?”, American Economic Review 81: 121–125.

Jones, S (2009), “Dynamic social norms and the unexpected transformation of women’s higher education, 1965–1975”, Social Science History 33: 247–291.

Amato, R, L Lacasa, A Diaz-Guileara and A Baronchelli (2018), “The dynamics of norm change in the cultural evolution of language”, Proceedings of the National Academy of Sciences 115: 8260–8265.

Smerdon, D, T Offerman and U Gneezy (2020), “‘Everybody’s doing it’: on the persistence of bad social norms”, Experimental Economics 23: 392–420.

Bicchieri C, E Dimant, S Gächter and D Nosenzo (2020), “Observability, social proximity, and the erosion of norm compliance”, CESifo Working Paper No. 8212.

Andreoni, J, N Nikiforakis and S Siegenthaler (2021), “Predicting social tipping and norm change in controlled experiments”, Proceedings of the National Academy of Sciences 118: e2014893118.

Granovetter, M (1978), “Threshold models of collective behavior”, American Journal of Sociology 83: 1420–1443.

Efferson, C,  S Vogt and E Fehr (2020), “The promise and the peril of using social influence to reverse harmful traditions”, Nature Human Behaviour 4: 55–68.

Tribou, A and K Collins (2015), “This is how fast America changes its mind”, Bloomberg.

Endnotes

[1] We tested the robustness of the occurrence of conformity traps in different environments that varied the speed of feedback, the size of society, and introduced extra incentives for instigators of change. The persistence of detrimental norms was observed in all settings.

a

A

Topics:  Frontiers of economic research

Tags:  social norms, norm change, social tipping points

Professor of Economics, University of California, San Diego

Professor of Economics, New York University in Abu Dhabi

Assistant Professor, School of Management, University of Texas at Dallas

Events

CEPR Policy Research