Many of the world’s most pressing societal, environmental, and health challenges are characterized by two sources of uncertainty: the ecological uncertainty of a potential disaster of unknown impact and the social uncertainty of not knowing how others will react to prevent this disaster. The investments to minimize these collective risks are costly, creating individual (short-term) incentives to delay or reduce investments in risk prevention, especially if it is uncertain how severe the risk would be (in terms of probability and/or impact). On the other hand, passing the safe bandwidth between sustainable and exploitative behavior can potentially have catastrophic results. Social norms have often been proposed as solutions to these challenges. Providing simple behavioral instructions about how to behave, people might look for social norms particularly in situations characterized by environmental uncertainty. However, since social norms are formed endogenously and in interaction with collective behavior, they tend to be stronger in situations with less uncertainty, where beliefs about the right course of action are more strongly embedded. This creates a paradox where social norms are weak or malfunctioning precisely when they are needed most.

We experimentally test this claim in two experiments in which we present subjects with various collective decision settings (a threshold public good setting and common pool resource setting) that vary in terms of environmental uncertainty. In the first experiment (the baseline), subjects have no information about the behavior of their group members. They decide how much to behave and are paid according to the collective returns after a random draw of the computer determines whether a disaster takes place. After making their decision, but before knowing the outcome of the interaction, they are also asked their personal normative beliefs, empirical expectations, and normative expectations. In the second experiment, people can use part of their earnings of the experiment to buy social information. If they do, we present them with social information from Experiment 1 stressing either the number of sustainable decisions (cooperative treatment) or the amount of miscoordination (uncooperative treatment). We expect that people more often pay to receive social information for intermediate risk levels (i.e., highest environmental uncertainty). Moreover, we expect that receiving positively framed social information increases the level of sustainable behavior compared to the baseline (Experiment 1), while negatively framed social information decreases it. All hypotheses, the experimental design, and the planned analyses are preregistered using the AsPredicted template. The experiment is programmed in oTree and will be conducted with a representative sample of Prolific users. Experiment 1 will be conducted in March and Experiment 2 will be conducted in April.