Preferences for Rights: Targeting Experiment

There are two main modules, and each participant is randomly assigned to complete only one of the two modules. Unlike AEA #0012065, all tasks in this follow-up are hypothetical.

In the targeting module, participants choose between concentrating benefits on fewer, poorer applicants and providing smaller benefits to a broader group. Participants are asked to choose between giving an in-kind good plus $100 in cash to a subset of the poorest of 10 applicants or giving the in-kind good alone to all 10 applicants. Using an adaptive multiple-price list that varies the number of applicants receiving the targeted benefit, we identify each participant’s indifference point between targeted and universal provision. We randomize the good across participants from a set of 13 goods (lawyer for an eviction proceeding, bus passes, health care vouchers, YMCA membership, community college tuition, food, child care, rental assistance, contraception, job training, pre-K tuition, internet access, and cash as a benchmark).

In the second module (rights), participants are randomly assigned to consider six of the 13 goods and indicate whether it should be considered a right. The module then includes several qualitative questions about one of the six goods (randomly selected) intended to understand the reasons for their response.

2026-09-04

2026-09-05

We focus on three main outcomes from the targeting (first) module. We will measure targeting via a continuous targeting choice (using the adaptive MPL to identify the smallest number of the poorest applicants for whom the participant prefers targeted provision), a binary targeting choice (whether the participant chooses to give to all 10 vs. give 5 applicants the good and cash), and a binary indicator for whether the participant always provides the good universally. Rights choice (second module): whether the participant indicates the good should be considered a right

We focus on:
1. The levels of these outcomes across different goods
2. The between-good correlation between targeting and rights choices.

We also intend to use the first outcome (targeting) as an input into a calibrated model of how transfer programs that provide different goods should be targeted to recipients at different incomes, when participants express good-specific welfare weights.

The rights outcome is constructed from a 3-category item (should be a right, desirable but not a right, or neither desirable nor a right). We code it as a binary indicator equal to 1 if the participant selects "should be a right."

From the rights module: we will look at the relationship between our primary outcomes and the following outcomes (all elicited on a 1-7 scale)
1. Perceptions of how valuable the good is
2. Whether recipients still have a right to a good even if they would not benefit from receiving it
3. Preference for following standard procedure
4. How morally worthy distributing the good is
5. How important it is to provide the good even when individually could afford to buy it themself (not included when the good is cash)
6. Whether it matters if the government or philanthropic groups provide the good
7. How important it is to guarantee the good even when people do not value it (not included when the good is cash)

Our experiment proceeds in several parts. First, we randomize participants to complete one of the two modules.They then complete either the targeting module (where one of 13 goods is selected as the focus) or the rights module (where six of 13 goods are selected for our primary question, and one of those six is selected for additional questions). Finally, we ask demographic and political questions. We recruit participants on Prolific.

Randomization done in Qualtrics via Javascript and Survey Flow Randomizer

Individual

No

5,101 people

5,101 people

Targeting module: Targeting module: 510 per good for each good from AEA #0012065 (4 total); 255 people per all other goods (9 total). That is 4,335 people in the targeting module.
Rights module: 766 people (15% of total sample), each responds for six goods, approximately 354 responses per good.