Revealed Social Norms

Abstract

When someone decides whether to take a prosocial action, they might be influenced by what others around them do. This could happen for two reasons. The first is standard externalities: for instance, if one person’s prosocial decisions lowers the cost of doing the same for another person. The second are social norms – people may want to conform to the decisions of others. A well known non-identifiability result shows that standard peer effect estimations cannot separate these two models (Brock & Durlauf 2001, Blume et al 2015, Boucher & Fortin 2016), despite their fundamentally different motives and often divergent policy implications (Young 2015, Bhattacharya et al 2024). We develop a method to overcome this problem and identify social norms in observational data. Using a theoretical model of choices in groups, we show that potential multiplicity of social norms can create incentives for behaviour that are very different from those under externalities. These differences, although not identifiable in peer effect regressions, can be identified by looking at the full distribution of choices across groups. We develop this method and identify social norms in the context of students’ voluntary participation in Cambridge’s large-scale asymptomatic Covid-19 testing programme. We show that students develop opposing social norms in different households. This heterogeneity is concealed by both peer effect estimates and aggregate testing patterns, whilst it can reverse some of the policy conclusions that flow from them.