Multiple Identities and Attentiveness

This is a two-arm between subject design. We vary whether there is 1) one dimension of identity or 2) two dimensions of identity made salient to subjects. When there is only one salient dimension of identity, we will randomly assign them one of the two available dimensions. Subjects then have to allocate some credits (i.e., experimental currency) between 2 random participants given information on their identities. More detail is provided in the attached Analysis Plan.

2021-05-21

2021-08-30

Identity strength on each dimension as measured by their decisions in the third-party allocation task

The effective allocation to a participant with same identity when the two random participants have opposite identities in the dimension.
(If they choose to implement their raw allocation decision, the effective allocation is just their "raw" allocation, otherwise, it will be a 50-50 allocation which is the (expected) allocation when choosing to split randomly or equally).

We have 2 secondary-measures which are sub-components of our main measure. 1)The raw allocation to a participant with the same identity when the two random participants have opposite identities in the dimension. 2) Whether participants choose to implement their raw allocation decisions, or decide to always split randomly or equally between the two random participants.
Beliefs of the distribution of identities on the two dimensions elicited in an incentivized way in the questionnaire

In the questionnaire after the allocation task, each participant has to answer a question of 4 different beliefs. Their answers are incentivized to be accurate. The 4 beliefs are as follows:
1) Among all participants who belong to the group that agrees with the statement on religion, what is the percentage that belong to the group that agrees with the statement on government;
2) Among all participants who belong to the group that disagrees with the statement on religion, what is the percentage that belong to the group that agrees with the statement on government;
3) Among all participants, what is the percentage that belong to the group that agrees with the statement on religion;
4) Among all participants, what is the percentage that belong to the group that agrees with the statement on government. One of the four questions is randomly selected for payment. If the selected answer is within 2% of the true value, the participant will get a bonus (with a higher bonus if the answer is within 1% or exactly correct).

This study uses an online experiment to examine the effect of the number of dimensions of salient identity on identity strength and ingroup biases in each dimension. At the beginning of the experiment, participants will have particular identities made salient to them. They subsequently play an incentivized third-party allocation task which is used to elicit strength of their identities. At the end of the experiment, there is a post-experiment questionnaire, which includes an incentivized elicitation of their beliefs of the distribution of identities on both dimensions. More detail is provided in the attached Analysis Plan.

There are 2 main stages in the experiment.

The first stage is the identity dimension priming stage: the priming is done through asking participants to consider and choose their standing on a general statement on religion or/and limited government. In the one-dimension treatment, they are only asked about one (random) statement. In the two-dimension treatment, they are asked about both statements in a random order, and the order will be kept throughout the experiment. Participants will then be shown the groups they belong to based on their choices, with graphic symbols for each group, which will further prime their identities.

The second stage is a third-party allocation task aiming to elicit participants’ strength of identities in an incentivized way. Participants will decide how to allocate a number of credits between 2 randomly selected other participants in the following manner. Firstly, they will be asked how they would like to allocate the credits in 3 different cases: 1) if one of the two selected participants has the same identity as theirs while the other has the opposite identity on the salient dimension; 2) both have the same identity and 3) both have the opposite identity. In the two-dimension treatment, in each case, a statement and graphic will highlight that on the other dimension of identity, the participant might have the same or opposite identity.

Then, they will be asked whether they would like to make the allocation based on their decisions, to equally divide the credits, or to randomly divide the credits between the two participants. For those in the two-dimension treatment, they will then be asked the same questions on the second dimension. Each participant’s decisions will be used to determine the payoffs of two randomly selected other participants. For those in the two-dimension treatments, one set of decisions will be randomly selected to be implemented.

Finally, after the second stage, we will have a post-experiment questionnaire. Those in the one-dimension treatment will be given and asked about the statement on the other dimension. The questionnaire will include questions on their strength of identities, (incentivized) beliefs about the distribution of and the correlation between identities on both dimensions, understanding and other standard demographic variables like birthyear, gender, education level, state currently living in, race, personal income, and percentage of personal income to household income.

At the end of the study, participants will be required to provide their Prolific ID. They will be credited with their payoff after the study is complete and payoffs are calculated given random matching.

More detail is provided in the attached Analysis Plan.

Randomization done by the experiment program when a session begins.
Stratified sampling is used to reduce sampling error.
There are six strata given by a participant's political affiliation and religion affiliation: Demographic/Republican/Independent X Religious/Non-religious. The information is given by the demographic screener on Prolific.
Simple random sampling is applied within each stratum: half of the participants in each stratum will be randomly assigned to the 1 dimension treatment, while the other half will be assigned to the 2 dimension treatment.

Randomization unit is at the individual level.

No

480-750 participants

480-750 participants

240-375 participants in control group with one salient dimension of identity (one of the two dimensions randomly selected)
240-375 participants in treatment group with two salient dimensions of identity (randomise the order of the two dimensions)
80-125 participants in each stratum

The software Optimal Design (Raudenbush et al., 2011) is used for power calculation. We only have one key hypothesis so no adjustment for multiple hypotheses testing is needed. A fixed-effect blocked trial model is used. In the pilot, we found the two stratification variables explained around 8.17% of the variance for the main outcome variable. Using this number, power = 0.8, alpha = 0.05, with equal SD and six strata, a sample size N=480 (80 in each stratum) is associated with a MDE of 0.247. In Chen and Li (2009), chatting with one’s ingroup results in an increase in ingroup bias with effect sizes ranging from 0.08 to 0.29 SD. If we take an intermediate value of around 0.2 SD as an estimate for our effect size, we will need a sample size of N=750 (125 in each stratum).Since we do not know the exact effect size and underlying variation, we will first target a sample size of 480. After gaining a better idea of the variance of the underlying population, we will potentially follow up with a larger sample size if needed.