Eliciting Social Norms

In their seminal paper, Krupka and Weber (2013) develop an incentivized elicitation method
for identifying social norms that uses simple coordination games. Participants are asked to rate the social appropriateness of various allocation decisions and are incentivized to match with the modal answer in their ratings.

In this study, we want to test whether the elicited social norms will be different if we use different elicitation methods or if we ask participants to state the social appropriateness of various controversial statements, instead of simple allocation decisions.

We will consider 5 different elicitation methods:
The first method consists of Krupka and Weber's method. Participants rate the social appropriateness of various allocation decisions and statements, but are incentivized to guess the modal response in their ratings. It therefore requires them to anticipate others' guesses.

The second method is non incentivized and is a variant of Krupka and Weber (2013)'s method. It removes the financial incentive associated with the coordination game. We ask participants to guess the social appropriateness rating most frequently given in the study for each of the allocation decisions and statements.

The third, fourth and fifth methods are designed to elicit social norms whithout having participants try to anticipate others' guesses.

The third method is non-incentivized: it consists of directly asking participants to state the social appropriateness of various actions and statements. We will refer to participants in this treatment as the "evaluators".

The fourth method is incentivized: it consists of asking participants to guess the social appropriateness ratings reported by the evaluators. They receive an additional $2 if their guess corresponds to the answer most frequently given by the evaluators.

Lastly, the fifth method is non-incentivized and mimics our third method: it consists of asking participants to guess the appropriateness ratings that an hypothetical group of evaluators would have reported.

At the end of the experiment, we will collect information on demographic characteristics, social preferences, cognitive ability and k-level thinking. We will explore potential correlations between these traits and differences in norm reporting within and across treatments.

2024-01-03

2025-01-03

We will collect social appropriateness ratings from participants.
We will also collect information on demographic characteristics, social preferences, cognitive ability and k-level thinking.

We use a between- subject design in an individual decision-making study. Each participant is randomly assigned to one treatment and is paid a participation fee of $4. The experiment is conducted on prolific and comprises 2 stages.

In the first stage, each participant will read the descriptions of a situation in which one person, "Individual A", must make an allocation decision. For each situation, participants will be given a description of the allocation decision faced by Individual A. This description will include several possible choices available to Individual A. After they read the description of the decision, participants will be asked to report appropriateness ratings for each possible allocation choice on a scale. The method used to report these ratings will differ across treatments. In this stage, we will therefore test whether the elicited social norms differ across elicitation methods.

In the second stage, participants will be presented 15 statements and will report the appropriateness rating associated with each of these statements. These statements include 10 controversial statements and 5 non-controversial statements, and will be presented to participants in a random order. We follow Braghieri (2021) and define controversial statements as statements for which there is a strong perceived social-acceptability direction. It is seen as more socially acceptable to disagree or agree with the statement than the opposite. On the contrary, non-controversial statements consist of statements for which both agreeing and disagreeing with the statement is perceived as socially acceptable.
In this stage, we will test whether the elicited social norms differ across elicitation methods and across the nature of the statement.

Each participant is randomly assigned to one of 5 elicitation methods:

Participants in treatment 1 are assigned to our first elicitation method. Following Krupka and Weber (2013), participants state the social appropriateness of the different allocation choices and statements.
At the end of the session, we will randomly select one of the allocation choices and one of the statements. For the choices selected, participants will receive an additional $2 if their rating matches the modal rating.

Participants in treatment 2 are assigned to our second elicitation method. They will guess for each of the allocation choices and statements the social appropriateness rating most frequently given by participants in the study.

Participants in treatment 3 are assigned to our third elicitation method. They are asked to report their ratings as truthfully as possible, based on their opinions of what constitutes socially appropriate or socially inappropriate behavior. We will refer to participants in this treatment as the "evaluators".

Participants in treatment 4 are assigned to our fourth elicitation method. They are asked to guess the social appropriateness rating most frequently given by the evaluators for each of the allocation choices and statements. At the end of the session, we will randomly select one of the allocation choices and one of the statements. For the choices selected, participants will receive an additional $2 if their rating matches the answer most frequently given by the evaluators.

Lastly, participant in treatment 5 are assigned to our fifth elicitation method. They are asked to guess the social appropriateness rating that would be most frequently given by an hypothetical group of evaluators.


At the end of the session, participants will be asked to complete a post-experiment survey including two non-incentivized k-level thinking games as well as questions on demographic characteristics, social preferences and cognitive ability.

The first k-level thinking game is a standard beauty contest game. Participants have to choose an integer between 0 and 100. The winner is the person whose number is closest to 2/3 times the average of all numbers chosen by participants in the study.

The second k-level thinking game consists of asking participants to guess the winning number in another beauty contest game.

Not available

computer program

session

No

750

750

150 participants per treatment.

We use the effect sizes suggested by Krupka and Weber (2013) and by a pilot study run at the Pittsburgh Experimental Economics Laboratory in April 2023 ( RCT ID: AEARCTR-0011132) to test whether these 5 elicitation methods yield similar or different elicited norms and whether it depends on the nature of the statement.