Trading off autonomy and efficiency in choice architectures : self-nudging versus social nudging

To overcome ethical objections to choice architecture interventions, Thaler and Sunstein (2008) suggest asking individuals to set their own nudge autonomously. Our online experiment (n=1080) faithfully implements this idea for social dilemmas where individual and collective interests often diverge and social nudges can conflict with autonomy. General-population subjects play a ten-round, ten-day public goods game. Non-participation triggers default contributions. We test three default nudges: An exogenous selfish nudge of zero contribution, an exogenous social nudge of full contribution, and an autonomous self-nudge where subjects select their own default contribution. Their performance is tested under four different information structures. We, first, document default choice under autonomy: Only between three and eight percent of subjects set their own default to either zero or full contribution. Second, autonomy and efficiency conflict: Group-level contributions under self-nudging are consistently lower than under the social nudge, which strictly dominates the selfish nudge. When committed to autonomy, the policy-maker - to maximize efficiency - best combines self-nudging with an information structure with public defaults.