Redistributive Preferences with Unequal Life Expectancy

Participants will complete an online vignette survey about tax policy. In the survey, they are asked to judge which of two hypothetical individuals is more deserving of a tax break. The vignettes manipulate characteristics such as life expectancy, retirement costs, consumption orientation, and cost of living while holding income and taxes constant.

The survey consists of several randomized vignette pairs. Each vignette presents two individuals with identical incomes and taxes but different circumstances (e.g., long vs. short life expectancy, high vs. low retirement costs, future- vs. present-oriented consumption, high vs. low cost of living). Respondents indicate relative deservingness on a 7-point scale and provide a written explanation. Between-subject randomizations include vignette order, position of “Individual A” vs. “Individual B,” and baseline vs. higher income levels.

2025-09-29

2025-10-15

(1) Average relative deservingness rating
(2) Distribution of relative deservingness on 7-point scale
(3) Distribution of relative deservingness on aggregate 3-point scale
(4) Net relative deservingness share

For (1), we test the sample mean on the 7-point scale
For (2), we test sample distribution of responses on the 7-point scale
For (3), we convert the 7-point scale to a 3-point scale (-1 - A more deserving, 0 - Equally deserving, 1 - B more deserving)
For (4), we calculate the share that sees A as more deserving than B, minus the share that sees B as more deserving than A

Responses to follow-up questions on perceived value of money and preferred tax rates for people with different characteristics.
Response to a belief question on dependance of life expectancy on factors within / beyond one's control.

These serve as robustness and quality measures and for mechanism analysis.

This is a vignette-based survey experiment. Respondents are randomly assigned to different vignette orders and treatments (income levels, individual order). They complete a series of pairwise comparisons, rating who is more deserving of a tax break, followed by general evaluative questions.

Vignettes 1,2,3,4 are randomized, and vignettes 5,6 are randomized; half of participants see Individuals A/B reversed; a subsample receive high-income vignettes ($100,000 gross vs. baseline $50,000). Outcomes are collected via online survey.

Computerized randomization within the online survey platform.

Individual

No

-

Total planned N = 1000 participants.

Main within-subject sample: 600 participants (sufficient to detect d = 0.2 with Bonferroni correction across four main within-subject contrasts when conservatively assuming zero within-person correlation).

Between-subject robustness arm: 400 additional participants with the high income vignette version.

We power the study to detect Cohen’s d = 0.2 (small) at 80% power, two-sided α = 0.05. Primary one-sample tests (is mean ≠ 4 on the 7-point scale): n ≈ 196 without multiplicity adjustment; applying Bonferroni for 4 tests (α = 0.0125) yields n ≈ 279. Within-subject contrasts (paired t-tests across vignette cells): with conservative assumption of zero within-person correlation, the detectable effect d = 0.2 on the difference implies n ≈ 558 under Bonferroni for 4 tests; we target n = 600. Between-subject robustness test high income: for d = 0.2 at 80% power and α = 0.05 requires ~393 per arm (total 786) — we add 400 new participants and use an equal-sized control subset from the main sample. Mapping to the scale, d = 0.2 corresponds to ~0.3 scale points assuming SD ≈ 1.5 from pilot data.