Understanding Redistribution Across Countries

In this project we examine the effect of different redistribution policies and of the prospect of upward mobility on anti- and pro- social behaviour.

2022-09-26

2022-10-10

We collect the following primary outcomes:
• An indicator variable equal to one if an individual decreased another subject’s income in part 1 and zero otherwise
• An indicator variable equal to one if an individual increased another subject’s income in part 1 and zero otherwise
• The amount a subject decreased their partner’s income by in part 1
• The amount a subject increased their partner’s income by in part 1
• Rankings of policies across different treatments and types

See attached The amount a subject decreased (increased) their partner’s income by in Part 1: this is limited to the values 0,2,4,6,8 and 10 ECUs.

Rankings of policies: at the end of the experiment, each respondent must answer a short survey. One of the questions asks:
Please rank the following from redistribution policies from most (1) to least (5) preferred. Firstly, for yourself, and secondly how you think others in your community generally prefer them.
The second order beliefs will be incentivized - participants who correctly identify the preferred rankings of the majority of participants in their country will be paid.

Our design aims to measure the impact of redistribution policies on an individual’s acceptance of advantageous and disadvantageous inequality. We will conduct an online experiment in multiple countries. We will match participants and create a level of inequality in endowment between the pair. We will further have different treatment arms representing different redistribution (tax-transfer) mechanisms. Each member of the pair will then have the opportunity to pay to increase or decrease the partner’s income. We are interested in inequality aversion across treatment arms. Our experimental design will allow us to examine whether preferences for inequality are influenced by the form of public redistribution (e.g. universal basic income, transfers funded by progressive taxation) and whether preferences for inequality are influenced by the prospect of upward mobility.

See attached

Randomization will be carried out by a computer (Javascript).
For each condition, we will stratify by gender and self-reported income level.

The randomization unit will be the individual for all treatments.

No

The number of clusters is the same as the number of individuals.

We will run the experiment in 4 countries, namely the USA, India, Indonesia and Germany. For each country we will collect 1,000 observations. 4,000 in total. The experiment has been designed on the Qualtrics platform. It will be distributed via a link to 1,000 participants in each country. We have signed a contract with Expilab Research who will deal directly with the panel providers. We will provide the link to Expilab Research who will send the link to the panel providers. The panel providers then distribute the link to their lists of respondents. We have stratified the sample by income and by gender. On clicking the link, panel respondents land on the welcome page where they enter their income level and gender. If we already have sufficient responses from this gender/income strata, they will not be allowed to enter the experiment. Dropping observations We will drop all observations that meet two of the following criteria: • Are >2 standard deviations faster than the average for their country; • Fail comprehension questions at a rate >2 higher than the average for their country; • Show little variation in column chosen in the survey questions

200 per treatment arm, per country. Randomly allocated into treatment arm.

Our power calculations, based on our pilots (N=110), show that with 200 participants per treatment arm and per country we can obtain a minimum detectable effect size for the following key outcomes: 0.11 (pay to decrease).