Stigma and Social Safety Net Program Participation

There are three intervention groups and one control group. The control group will receive no intervention during the survey.

Two text-based interventions intend to reduce stigma and stigmatizing views by using vignettes describing SNAP as a tool for upward mobility. In one vignette, SNAP is described as a tool to help someone to invest in their career (Treatment 1). In the other vignette, SNAP is described as a tool to help someone invest in their kids so that their kids can become upwardly mobile in the future (Treatment 2).

Within each Treatment 1 and Treatment 2, we cross-randomize the indicated gender of the SNAP recipient: male or female.

One text-based intervention provides information on how SNAP is funded to correct a commonly held misperception (Treatment 3). In surveys and qualitative interviews, people often report that they avoid receiving SNAP benefits so as not to take away benefits from others who may need them more. However, SNAP is an entitlement program, so benefits are not rationed and anyone who is determined eligible for SNAP can receive benefits. The idea of the intervention is to provide this information to survey participants. The information is also customized based on whether the respondent answered the question correctly or incorrectly earlier in the survey.

Individuals assigned to this treatment arm receive no information indicating gender.

2023-09-18

2024-01-18

Stigmatizing attitudes index, overall perceived stigma index, demonstrated interest in enrolling in SNAP, and second-order beliefs questions.

Stigmatizing attitudes index: The mean of answers to the following questions: Out of 100 individuals receiving SNAP, how many… (a) do you think should avoid telling other people that they receive SNAP; (b) would you judge negatively if you noticed them using SNAP in the grocery store; (c) would you look down on because they receive SNAP; (d) do you think are less motivated to work because they receive SNAP; and (e) do you think would be taking the place of someone who needs SNAP more than they do.

Overall perceived stigma index: The mean of answers to the questions in the self stigma and perceived external stigma indices. (See below.)

Demonstrated interest in enrolling in SNAP is measured by a dummy variable which equals 1 if the survey respondent clicked a link to take a quiz to check their household’s SNAP eligibility status, and 0 otherwise. This is meant to capture a real-world outcome: a first step towards applying for SNAP.

Second-order beliefs questions: Respondents are asked to estimate the average response of other survey respondents to the following two questions: (1) Out of 100 individuals receiving SNAP, how many do you think are less motivated to work because they receive SNAP? and (2) Out of 100 individuals receiving SNAP, how many would you judge negatively if you noticed them using SNAP in the grocery store?

Self stigma index, perceived external stigma index. Perceived overall stigma views and stigmatizing attitude individual questions. Also, measures for the first stage impacts of the interventions. Pending data access, outcomes of the eligibility screener. Support for SNAP spending (wave 2 only).

Self stigma index: The mean of answers to the following questions (with slight variations on question wordings based on whether the respondent received SNAP before): (a) I would rather support myself than use SNAP. (b) Applying for SNAP made me think less of myself. (c) When I participate in SNAP, I avoid telling other people about it.

Perceived external stigma index: The mean of answers to the following questions (with slight variations on question wordings based on past receipt of SNAP): (a) Most people look down on me because I applied for SNAP. (b) If someone found out I applied for SNAP, they would think I lack work ethic. (c) Because I enrolled in SNAP, other people think I am taking the place of someone who needs benefits more than me.

Individual questions in the perceived overall stigma and stigmatizing attitudes indices listed above. We will use methods to correct for multiple hypothesis testing.

First-stage outcomes for misperception intervention: (1) Enrolling in SNAP makes me feel like I am taking the place of someone who needs benefits more than I do. (2) Because I enrolled in SNAP, other people think I am taking the place of someone who needs benefits more than me. (3) Out of 100 individuals receiving SNAP, how many… do you think would be taking the place of someone who needs SNAP more than they do.

First-stage outcomes for vignette interventions: Out of 100 individuals receiving SNAP, how many…(1) do you think are adult men; (2) adult women; and (3) children. Summary measures of text responses to the question: "Please briefly describe what you think of when you think of a typical household participating in SNAP."

First-stage outcomes for work vignette interventions: (1) Participating in SNAP makes me less motivated to work. (2) If someone found out I applied for SNAP, they would think I lack work ethic. (3) Out of 100 individuals receiving SNAP, how many…do you think are less motivated to work because they receive SNAP?

Outcomes of the eligibility screener may include: completed the screener, found eligible or found ineligible, started an application and submitted an application.

Support for SNAP spending: Asks respondent whether she prefers current levels of spending on SNAP, more, or less spending.

Our sample includes survey respondents from Prolific. The survey respondents will answer a series of questions at baseline. Demographic questions asked at baseline include household size, state of residence, and prior SNAP participation status. We also collect respondents’ general preferences for redistribution and a survey module designed to capture an individual's propensity to give socially desirable answers. At baseline, we also collect respondents’ perceptions about features of the SNAP program and recipient population. Key factual questions are incentivized to encourage respondents to report their best guess.

The respondents will then be randomly assigned into one of the four main treatment arms, stratified by prior SNAP participation status. There are four blocks or strata: current SNAP recipient, past SNAP recipient, never a SNAP recipient, or missing SNAP status. Conditional on the prior SNAP participation group, respondents are randomly assigned into one of the four treatment arms with equal probability. Respondents in each treatment group are shown only one piece of information described in the interventions section above; those in the control group are shown no information.

Then, survey respondents will answer the remaining survey questions. Questions provided after the intervention include the outcome questions described above and additional demographic questions.

The main experimental design is randomization to the four treatment arms. The following describes additional, secondary experimental design features.

For the question which measures the initial prevalence of the misperception, ALL survey respondents are randomly assigned to either a TRUE or FALSE framing. The individuals randomly assigned to the misperception intervention will then be told the truth and so will receive slightly different versions of the intervention based on whether they initially answered the TRUE or FALSE question framing.

For each of the interventions which provide vignettes about work or family, we randomly assign the indicated gender for the person in the vignette. For each vignette type, the “man” or “woman” indications are assigned with equal probability.

To test for social desirability bias, all survey respondents will be randomly assigned to one of four conditions during the measurement of stigmatizing attitudes questions: (a) no additional prompt, (b) a prompt reminding respondents their answers are “confidential and anonymous”, (c) a prompt indicating we are “interested in your personal opinions”, and (d) both prompts.

At baseline, all survey respondents are given a selection of questions from a measure developed by social psychologists to measure their propensity to give socially desirable answers (Crowne & Marlowe 1960; Reynolds 1982). For example, one question asks respondents whether they “never hesitate to go out of [their] way to help someone in trouble”, where the socially desirable answer is true. For each respondent, we will calculate a social desirability score based on the number of socially desirable answers. We will test the robustness of our results to controlling for this social desirability score.

Endline demographic questions include SNAP receipt in one’s social network, age, gender, income, race, ethnicity, political affiliation, education, work status, and zipcode.

In the second wave, participants are randomized to have the additional demographic questions at the very end of the survey (as before) or the very beginning of the survey.

In the second wave, we add the following demographic questions: length of time ever participated in SNAP and number of children in the household.

Randomization will be done automatically by the Qualtrics online survey platform.

Individual survey respondents

No

1200 individual survey respondents

1200 individual survey respondents

300 control, 300 work vignette (150 work male, 150 work female), 300 family vignette (150 family male, 150 family female), 300 misperception correction