Who Deserves a Tax Break and Why? Evidence on Preferences for Taxing Personal Characteristics

2024-10-07

2025-10-07

For the policy proposals:

We elicit participants’ support for ten policy proposals using a 7-point scale, from “-3: Strongly oppose” to “+3: Strongly support.” We plan to analyze the following outcome variables:

1. Support_ip: Continuous measure of support by individual i for policy p (at the individual x policy level), ranging from -3 to +3.
2. BinSupport_ip: Binary measure of support by individual i for policy p (at the individual x policy level). The binary measure takes a value 1 if individual i displayed positive support for policy p (i.e., if support_ip >0), and zero otherwise.
3. Average support across all policies at the individual level
4. Share of policies supported at the individual level (share of policies for which BinSupport_ip=1)

For the vignette experiment:

We elicit participants’ support for tagging using a 7-point scale, from “-3: Strongly oppose” to “+3: Strongly support.” We plan to analyze the following outcome variables:

1. Continuous measure of support for tagging ranging from -3 to +3
2. Binary measure of support for tagging that takes a value of 1 if support for tagging in the vignette experiment is greater than 0 and 0 otherwise

Sample selection:

We recruit participants from the provider Bilendi. Participants who fail the attention check or any of the three comprehension checks cannot proceed with the experiment. We exclude participants from the study when the demographic quotas have been met. Among those who completed the survey, we will exclude those who took the survey multiple times.

Support for policies:

We elicit participants’ support for 10 different policy proposals. Each policy involves two groups of individuals in society distinguished by a single personal characteristic (e.g., women vs. men). The average income of the two groups differs. In each policy, individuals in the low-income group would owe slightly lower taxes compared to those in the high-income group. Participants’ support for these policies is measured using a 7-point scale, from “-3: Strongly oppose” to “+3: Strongly support.”

Participants are presented with ten different policies. The low-income groups in these policies are as follows: (1) Individuals who are NOT married, (2) Individuals aged 18 to 30, (3) Women, (4) Ethnic/racial minorities, (5) Individuals with disabilities, (6) Individuals NOT working in the service sector, (7) Individuals whose highest education is a high school diploma or less, (8) Individuals with visual impairments, (9) Individuals living in the poorest region, and (10) Individuals who financially support their children or other relatives.

Vignette Experiment:

In the vignette experiment, participants are presented with a scenario involving two groups of individuals distinguished by a single personal characteristic. One group has a lower average income than the other. We elicit participants’ support for a policy in which individuals in the low-income group would owe slightly lower taxes compared to those in the high-income group. Their support can range, on a 7-point scale, from “-3: Strongly oppose” to “+3: Strongly support.” Participants support “tagging” if they support this policy.

Participants are randomly assigned (between-subjects) to one of eight treatments in a 2 x 2 x 2 design with variations of the above scenario. The first treatment dimension varies whether the characteristic distinguishing the two groups is immutable or mutable (Treatments “Immutable” vs “Mutable). The second treatment dimension varies the strength of the correlation of the tag with ability (treatments “Low Corr” and “High Corr”). The third treatment dimension varies whether the tag is correlated with individuals' consumption needs (Treatments “Uncorr Needs” vs. “Corr Needs”).

Explanatory variables:

To explore the predictors of participants’ support for policy proposals and support for tagging in the vignette experiment, we ask participants several questions, such as their beliefs that tagging increases administrative costs or complicates the tax filing process.

Software-based randomization (Qualtrics)

Individual

No

1

3000 participants.

375 participants in each of the eight treatment cells.