Paternalism in the Welfare State

Last registered on July 23, 2023

View Trial History

Pre-Trial

Trial Information

General Information

Title

Paternalism in the Welfare State

RCT ID

AEARCTR-0009144

Initial registration date

April 12, 2022

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published

April 14, 2022, 11:53 AM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated

July 23, 2023, 1:25 AM EDT

Last updated is the most recent time when changes to the trial's registration were published.

Locations

Country

United States of America

Region

Primary Investigator

Name

Zachary Freitas-Groff

Affiliation

Stanford University

Contact Primary Investigator

Other Primary Investigator(s)

PI Name

Sandro Ambuehl

PI Affiliation

University of Zurich

Contact Investigator

PI Name

Tony Fan

PI Affiliation

Stanford University

Contact Investigator

PI Name

B. Douglas Bernheim

PI Affiliation

Stanford University

Contact Investigator

Additional Trial Information

Status

In development

Start date

2022-08-02

End date

2024-06-30

Keywords

Behavior, Lab, Welfare

Additional Keywords

paternalism, online, survey

JEL code(s)

D6, D63, D64, D9, D91, D3, D31, H53, I38

Secondary IDs

Stanford IRB-53441

Prior work

This trial does not extend or rely on any prior RCTs.

Abstract

Assistance towards the poor in Western economies is often delivered in the form of in-kind transfers. This form is puzzling from the vantage point of standard microeconomic theory, according to which cash transfers are the most efficient means of raising recipients' welfare. To understand the motivations for such policies, we conduct a lab-in-the-field experiments in the US. General population respondents have the opportunity to place restrictions on a transfer we deliver to a welfare recipient who participates in the study. We examine the causes underlying the restrictions individuals impose. We quantify the importance of various motives including respect for the recipient's autonomy. We characterize how these motivations differ depending on whether the welfare recipient is black or white, male or female, and whether the recipient has children. In counterfactual simulations, we then examine whether and how approval of paternalistic constraints on welfare payments would change if individuals had accurate beliefs about the recipients' use of the funds as well as about the demographic composition of the recipient pool.

External Link(s)

Registration Citation

Citation

Ambuehl, Sandro et al. 2023. "Paternalism in the Welfare State." AEA RCT Registry. July 23. https://doi.org/10.1257/rct.9144-3.0

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

Intervention (Hidden)

We vary the demographic group of a Supplemental Nutrition Assistance Program (food stamps) recipient that respondents make decisions about. Respondents decide whether the person receives food, gift cards, or a choice between the two.

Intervention Start Date

2022-08-02

Intervention End Date

2024-06-30

Primary Outcomes

Primary Outcomes (end points)

• meanBelief: The respondent’s incentivized belief about the average total gift card value the welfare recipient would trade for the food box, elicited using a multiple price list with eight dollar amounts ($0, $25, $45, $60, $70, $85, $105, $130)
• patFoodSw: The range out of the nine defined by the eight dollar amounts ($0, $25, $45, $60, $70, $85, $105, $130) in which the respondent switches from sending the welfare recipient the food box to letting them choose between the food box and gift cards worth that amount.
• surrSw: The range out of the nine defined by the eight dollar amounts ($0, $25, $45, $60, $70, $85, $105, $130) in which the respondent switches from sending the welfare recipient the food box to sending them gift cards worth that amount.
• We again elicit the first and the last of the above three outcomes after informing the respondent whether the welfare recipient prefers food or gift cards worth $70.
We had originally planned to use the dollar amounts ($0, $20, $35, $45, $55, $65, $80, $100) but edited the questionnaire to reflect the new price grid after piloting. We also originally planned to provide information about the welfare recipient’s preferences above or below $50 but edited the questionnaire to use $70 as the new benchmark after piloting.

Primary Outcomes (explanation)

We assign the midpoint of each interval as the value of the outcome for respondents who switch in that interval. For example, someone who lets respondents choose between food and gift cards at $45 and above but who forces food at $35 and below would have patFoodSw = $40.

Secondary Outcomes

Secondary Outcomes (end points)

• sdBelief: The respondent’s incentivized belief about the standard deviation of the total gift card value the welfare recipient would trade for the food box, elicited using a multiple price list with eight dollar amounts ($0, $25, $45, $60, $70, $85, $105, $130)
• sdBelief, elicited after informing the respondent whether the welfare recipient prefers food or gift cards worth $50.
• altruism: The dollar amount out of a list of six dollar amounts ($0, $10,…,$50) that the respondent would give to the welfare recipient if the amount remaining out of $50 goes to a random U.S. taxpayer.
• recFood: Whether the respondent recommends choosing food if they give the welfare recipient the choice between food and gift cards (rather than the food box).
• Miscellaneous questions regarding how respondents use the food and gift cards, how they spend the marginal dollar from each, the price of the food box, welfare recipients’ demographics, the acceptability of different choices, political views, and demographics.

Secondary Outcomes (explanation)

Experimental Design

We present respondents with a Supplemental Nutrition Assistance Program (SNAP, food stamps) recipient and one of two different food boxes (a healthy box and a representative box), each worth about $50. One in 20 respondents makes the choice about a real person. Respondents learn that the welfare recipient may or may not be real and do not learn whether they are. Respondents then make a series of choices regarding whether the welfare recipient receives the food box or a set of gift cards of their choice worth a given dollar amount.

In order to run these treatments, we first run a survey with SNAP recipients where we elicit their preferences for the food box compared to gift cards.

After running the initial survey, we recruit a new panel of 500 respondents. First, we altered the way we elicited respondents’ beliefs (meanBelief, sdBelief) to drag and drop tags into bins rather than using a grid. Second, we randomly provided information about SNAP recipients’ alcohol consumption to some respondents and not to others.

Experimental Design Details

We present respondents with a Supplemental Nutrition Assistance Program (food stamps) recipient and one of two different food boxes (a healthy box and a representative box), each worth about $50. We randomize respondents to view a welfare recipient who is a single male, a single female, or a single female with children and either black or white. One in 20 respondents makes the choice about a real person; the remaining respondents view each demographic group with equal probability. Respondents learn that the welfare recipient may or may not be real and do not learn whether they are. Respondents then make a series of choices regarding whether the welfare recipient receives the food box or a set of gift cards of their choice worth a given dollar amount.

In addition, we randomly inform respondents whether the welfare recipient chose the food box or the gift cards worth a given amount ($50), and we then have the respondents make the choices again. Which choice the welfare recipient made is again assigned randomly with equal probability to either choice for those respondents who do not pair with a real welfare recipient.

In order to run these treatments, we first run a survey with SNAP recipients where we elicit their preferences for the food box compared to gift cards.

After running the initial survey, we recruit a new panel of 500 respondents and 150 SNAP recipients with a few different features. First, we alter the way we elicit respondents’ beliefs (meanBelief, sdBelief) to drag and drop tags into bins rather than using a grid. Second, we randomly provide information about SNAP recipients’ alcohol consumption to some respondents and not to others. Third, we update the recipients’ price grid to match the grid used for respondents. (In the initial survey, SNAP recipients faced the price grid ($0, $20, $35, $45, $55, $65, $80, $100).)

Randomization Method

Random number generator on Qualtrics platform

Randomization Unit

Individual

Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters

Initial survey: 4,000 respondents (including 3,800 respondents who make choices about artificial welfare recipients with pre-specified demographics); 140-160 real welfare recipients for incentives.

Second round: 500 respondents; 150 real welfare recipients for incentives.

Sample size: planned number of observations

Initial survey: 4,000 respondents (including 3,800 respondents who make choices about artificial welfare recipients with pre-specified demographics); 140-160 real welfare recipients for incentives. Second round: 500 respondents; 150 real welfare recipients for incentives.

Sample size (or number of clusters) by treatment arms

Approximately 300 respondents in the young white male bin
Approximately 300 respondents in the young black male bin
Approximately 600 respondents in the young white female (without children) bin
Approximately 600 respondents in the young black female (without children) bin
Approximately 600 respondents in the young white female (with children) bin
Approximately 600 respondents in the young black female (with children) bin
Approximately 300 respondents in the old white male bin
Approximately 300 respondents in the old black male bin
Approximately 300 respondents in the old white female (without children) bin
Approximately 300 respondents in the old black female (without children) bin

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

With our current sample size, resampling our pilot data indicates that we have the following power to detect a $5 effect of the following demographics on the listed outcomes: Outcome Power for parental status Power for race patFoodSw 91.1% 99.4% surrSw 99.8% 100% surrSw, elicited again after observing the welfare recipient chose $50 in gift cards over the food box 99.1% 100% surrSw, elicited again after observing the welfare recipient chose the food box over $50 in gift cards 100% 100% The power to detect a $5 effect of gender on the listed outcomes will fall between parental status and race, and thus is always larger than 90%.

Supporting Documents and Materials

IRB