The Impact of Uncertainty on Norm Compliance in Pro-Social Behavior

The lab experiment consists of a between-subject design, with the type of normative information as the manipulating factor. Subjects will be randomly assigned to either a control group or one of two information treatments. Those in the treatment groups will receive normative information relevant to a standard dictator game setting. This information will be presented either as a point estimate (a single value) or as a profile (a range of values). Subjects in the control group will receive no normative information prior to completing the experimental tasks. Treatment assignment will be done at the session level.

The information treatments are based on data from a prior experiment on norms and giving behavior conducted in Spring 2025, which replicated Krupka and Weber (2013). The average population response on the social appropriateness of each actions is termed the appropriateness rating. Results replicated KW2013, with slightly narrower ratings: equal splits were widely viewed as most socially appropriate, keeping everything as least appropriate, producing an inverted U-shaped pattern peaking at the 50% allocation. This replication is referred to as Session 0. In the treatments, subjects will receive normative information as follows.
Treatment 1 - Point Estimate (PE): Subjects receive information about the action rated as most socially appropriate in Session 0. The message states: “In a previous session, we asked participants to rate each selection available to Player 1 on an appropriateness scale from -1= very socially inappropriate to 1= very socially appropriate. Based on the average responses collected, the most socially appropriate selection is to send $5 to the recipient.”
Treatment 2 - Appropriateness Profile (AP): Subjects are shown a distribution-style figure displaying the appropriateness ratings for all actions in Session 0. The AP message reads: “In a previous session, we asked participants to rate each selection available to Player 1 on an appropriateness scale from -1= very socially inappropriate to 1= very socially appropriate. The figure below displays the average social appropriateness rating for each selection.”

2025-12-10

2026-03-02

Giving behavior, Social norm uncertainty, and the importance placed on norm compliance.

Giving behavior is based on the subjects' choice in a standard dictator game experiment. This is the amount that each player chose to send to an anonymous recipient in the role of a dictator. There were eleven choices, from giving $0 to all $10 (entire endowment amount) in $1 increments.
Social norm uncertainty is measured by the standard deviation of a distribution of tokens that the subjects create in one part of the experiment. For each action choice in the dictator game (explained above), subjects receive 100 tokens to allocate across 4 appropriateness levels. This is a coordination game; they are trying to allocate tokens in attempt to guess how others in their session will allocate tokens. The 4 levels are very socially inappropriate, somewhat socially inappropriate, somewhat socially appropriate, and very socially appropriate. They can allocate tokens in any way they choose, resulting in a unique distribution per person. They receive payoff proportional to how many tokens they allocate to the appropriateness level that received the highest average token allocation by participants in the session. The more certain they are about how others will rate the appropriateness about a given action, the closer the standard deviation of their token distribution will be to 1.
The importance placed on social norm compliance is estimated as a coefficient in our main theoretical model, which uses data on giving behavior in the dictator game, as well as norm uncertainty and a population level calculation of the social norms themselves within the context of that dictator game.

Risk preferences, Gender

Risk preferences is measured based on a gamble task, which allows participants to choose 1 of 6 available gambles which vary based on their payoffs and risk. An additional measure of risk preferences is a self reported level of risk aversion elicited in the post experiment survey (likert scale from very unwilling to take risks to very willing).
Gender is measured in the post experiment survey, which will be used as an observable in the regression models.

Following treatment assignment, subjects will complete three tasks in the following order: social preference elicitation task, norm perception elicitation task, and risk preference elicitation task. The instructions for each task will be presented visually on the screens, accompanied by an audio recording of the instructions. In the treatment conditions, normative information will be provided prior to both the social preferences elicitation and the social norm elicitation tasks. This information will be both displayed on the screen and printed, with copies available at each workstation, allowing subjects to refer back to it at any point during the session.
After completing the three tasks, subjects will fill out a short demographic survey.

Eliciting Social Preferences
To elicit social preferences, a standard dictator game will be used (Hoffman et al. 1994). Subjects will receive a $10 endowment, and be asked to split it between themselves and an anonymous recipient. The action set will consist of any split from $0 to $10, in $1 increments, resulting in an action set of 11 choices total. All subjects will make a decision in the role of dictator. To incentivize this task, subjects will be randomly paired and assigned to roles at the end of the session. The decision made by the assigned dictator will be implemented, and both subjects will be paid accordingly. The recipient will remain as a passive player.

Eliciting Social Norm Perception
To elicit social norms perceptions, subjects will take part in a pure coordination game. In this task, subjects will receive a 100-token endowment for each of the eleven actions available in the dictator game, as described in the previous task. They will then be asked to assign the tokens to the appropriateness levels available for each action: “Very Socially Inappropriate”, “Somewhat Socially Inappropriate”, “Somewhat Socially Appropriate”, and “Very Socially Appropriate”. Since this is a pure coordination game, subjects will be prompted to allocate tokens based on their guess as to how other subjects in the session will distribute their own tokens. They may choose to allocate their tokens in any manner, so long as they allocate all 100 tokens. This will result in an individual distribution for each action for each subject. Once all subjects have completed this task, the average token allocation to each appropriateness level will be calculated for each action. The appropriateness level with the highest average allocation will be the winning level. One action choice from the set will be randomly chosen and incentivized so that each subject will receive a payoff proportional to the number of tokens they have allocated to the winning level for that action. This will be done using a conversion of 1 token= $0.10. For example, if a subject allocated 34 tokens to the winning level, they would receive $3.40 for this task.

Eliciting Risk Preferences
Risk preferences will be elicited using the gamble task by Eckel and Grossman (2008). Subjects will be presented with a menu of six gambles, each offering two possible outcomes with equal probability (50/50). The gambles vary such that moving from gamble choice 1 to gamble choice 6 increases the degree of risk. The available gambles range from a riskless option ($6/$6) to a highly risky option ($1/$15), and the intermediate gambles provide gradually wider outcome spreads. Subjects will be asked to choose only one of the available gambles. The lower the number chosen in this task, the higher the degree of an individual’s risk aversion. Each subject’s chosen gamble will be played via the O-tree platform, and they will be paid based on the binding outcome.

Randomization will be done using a computerized tool to assign treatments to sessions. Treatments will be randomized at the session level over two consecutive weeks, with sessions scheduled at various times between 9:00 am to 6:00 pm.

Treatments will be randomized at the session level. All participants within a unique session will receive the same treatment.

Yes

12 to 30 sessions based on show up rate per session.

255 subjects

At least 4 session for each of the three treatments arms (aiming for 85 participants per treatment)

To detect a minimum effect size of 0.2 with 80% power at a 5% significance level, a sample of 82 participants per treatment is required.