(Follow Up) Can a Shared National Identity Trump Ethnic Identity? Evidence from a Field Experiment in Singapore (with RCT ID AEARCTR-0007559)

The new intervention will be an online survey on 1,300 participants, adults residing in Singapore. The survey will ask a bunch of attitudinal and socioeconomic questions. Beyond that, we will carry out a conjoint experiment, a supermarket voucher experiment, and a donations experiment. All details in the box below.

The individual characteristics we will collect are are ethnicity, marital status, language spoken at home, monthly income, occupation, place of birth, legal status (born Singaporean/became Singaporean/permanent resident), year start residing in Singapore, apartment size, apartment tenure, number of people living in the apartment, education, religion, religiosity, strength of ethnic and national identification, views on national identity, views on inter-ethnic relations, perceptions on ethnic stereotypes (heard of them or not & agreeable or not), experience of ethnic discrimination, experience of stereotyping, political interest, political preferences, sources for political information, level of degree of satisfaction with different public services (public schools, government hospitals, access to electricity, access to water/sanitation, overall safety and security), views on what the state should do to provide for the needy,

2024-03-20

2024-04-10

See all details below.

For intervention 1: Characteristics of recipient associated to higher levels of altruism. These are five characteristics: ethnicity, household size, socioeconomic status, occupation status, nationality.

For intervention 2: Rates of donations to the different NGOs

For intervention 3: Rates of keeping the voucher for the different supermarkets.

All details provided in the word document attached under "analysis plan"

All details provided in the word document attached under "analysis plan"

There will be three experiments:

1. Conjoint experiment: The primary goal is to allow us to tease out whether people are more likely to exhibit altruism towards those they think deserve their help.

2. Donations experiment: To examine whether people are more likely to exhibit altruism to their co-ethnics, we will ask participants to choose a local NGO to make a $10 donation to. We will randomly select 100 participants and make (with our funds) the donations according to their preferences. Participants will be reminded adequately that all donations will be made, that this is not a hypothetical scenario but a real one. The goal of this question is to examine whether on average participants prefer to donate to coethnic organizations, or organizations that cut across ethnic lines (like, for example, support for the disabled or the arts). The exact wording of this question in the questionnaire is given below.

3. Supermarket voucher experiment: The primary goal of this is to examine whether residents are more likely to disregard free vouchers from a particular supermarket than from other ones.

Not available

All randomizations will be done algorithmically by the survey company provider: Qualtrics.

Only one level of randomization for both interventions: Individual

No

1,300 individuals

1,300 individuals

For conjoint experiment, all randomizations will be given equal probability.

For the donations experiment, all individuals whose choices are effectively donated will be selected with equal probability (although this will be done ex post, so it should not alter the results in any way).

For the voucher experiment, all participants will be randomly allocated with equal probabilities (one third) to each of the groups: Sheng Siong, NTUC, Fair Price

Check **Power Analysis for Conjoint Experiments*** https://osf.io/preprints/socarxiv/9yuhp The attached document included in "analysis Plan" includes the relevant computations *********************************** * 2. Donations Experiment * *********************************** *********************************** * 3. Voucher Experiment * *********************************** These are our power computations using STATA/MP16. power twoproportions average1 average2, n(#) twosided For an effect size of 5% (return rates differing by at least 5%): power twoproportions 0.1 0.15, n(866) power twoproportions 0.5 0.55, n(866) power twoproportions 0.9 0.95, n(866) Comparison groups will be of size 433. This means that for an effect size of 5%, for two-sided comparisons, we have a statistical power ranging from 0.6046 (if keeping averages are 0.1 & 0.15) to 0.7987 (if return averages are 0.9 & 0.95). Power drops to 0.3133 at the 0.5 vs 0.55 keeping rates. For an effect size of 10% (return rates differing by at least 10%): power twoproportions 0.1 0.2, n(866) power twoproportions 0.5 0.6, n(866) power twoproportions 0.9 0.99, n(866) For an effect size of 10%, for two-sided comparisons, we have a statistical power ranging from 0.9855 (if keeping averages are 0.1 & 0.2) to 1 (if return averages are 0.9 & 0.99999). Power drops to 0.8420 at the 0.5 vs 0.6 keeping rates.