Superstar Billionaires: Public Image and Demand For Taxation

Each participant is presented with a randomly-selected individual from the list of 5 of the wealthiest billionaires compiled by Forbes (Elon Musk, Jeff Bezos, Bill Gates, Mark Zuckerberg, and Michael Bloomberg). Each participant is asked a few questions about the billionaire (prior and posterior beliefs) and is randomly allocated to an information treatment. After the information provision stage, we include a battery of questions about the individual's attitudes towards the taxation of billionaires and support for specific policies.

2024-01-24

2024-03-31

The full set of questions is included in the attached sample questionnaire. We ask individuals for their desired income tax rates for billionaires as well as the desired corporate tax rates for the companies they founded. We include a few questions about the personal opinions of the billionaires and the companies they founded (e.g., whether they agree with the statement "Elon Musk deserves the wealth he has."). We elicit support for four real policy proposals to tax the ultra-wealthy. Last, we have three behavioral outcomes: i) the amount people are willing to donate to an organization supporting a fair tax system; (ii) whether the individuals sign a petition in support of a policy proposal; (iii) the willingness to pay for a backpack with the logo of the company founded by the billionaire.

To improve the precision of the estimates, in addition to estimating the effects on specific outcomes, we'll also construct an index for each of the categories of outcomes (e.g., support for policy proposals).

The experiment is designed to provide a horse-race between different reasons why the public may demand taxes for the ultra-wealthy. In each treatment arm, we provide information related to one aspect (e.g., the wealth of the billionaire, the tax rates that they pay) and then we measure the effects of the information on the support for taxation.

Each subject is randomly assigned to one of three treatment arms, each exploring a different mechanism:

1) Perception of wealth.
2) Perception of tax rates.
3) Public Image (luxury lifestyle and hard work).

We attach PDF documents with screenshots of the questions used in each treatment arm (using Elon Musk as an example).

Within each treatment arm, we collect prior and posterior beliefs to be able to distinguish the effects between the individuals who are updating their beliefs upwards versus downwards vs not updating at all. We elicit prior and posteriors beliefs with predefined scales, except in the wealth treatment and the luxury treatment, where we we elicit prior and posterior beliefs through open-ended questions. In that case, we will apply a log-transformation of the priors and posteriors for the analyses.

We will also invite participants to respond to a follow-up survey 2 weeks after they responded to the baseline survey, to be able to measure the persistence of the effects on beliefs and attitudes.

Not available

Randomization done by a computer (Qualtrics software)

Individual respondent

No

We aim to recruit around 6,000 participants for the baseline survey through the panel provider Prolific. Two weeks later, we will invite all the participants to complete a follow-up survey. Based on our prior experience with the platform, we expect that around 65% of the participants will respond to the follow-up survey.

Same as cluster. We exclude data points based on the following criteria: - Attention check: We include an attention check in our survey. Any participant who does not pass the attention check will be excluded from the analysis and will not count towards the number of completes. - Survey completion time: Participants whose response time falls below the minimum threshold will be excluded from the analysis. - Previous participation: We exclude participants who have already participated in one of the pilot surveys. - Outliers: In the wealth treatment and the luxury, we elicit participants' prior and posterior beliefs through an open-ended question. To minimize the influence of outliers, we will take the following steps: (i) We will drop respondents with outlier prior beliefs; (ii) We will winsorize posterior beliefs.

The probabilities of assignment to each treatment arm are described in the section Experimental Design.

There is no simple answer to this question as: (i) There are multiple treatment arms (e.g., information on wealth or tax rates); (ii) there are multiple outcomes (e.g., desired tax rate or the probability of signing a petition); (iii) and we expects heterogeneous treatment effects due to belief updating. However, as an illustration, we can provide some calculations based on the results from a small pilot. We expect that the wealth information (relative to no wealth information) would have a minimum detectable effect on the desired income tax rate of 3.5 percentage points.