A Randomized Controlled Trial Varying Unconditional Cash Transfer Amounts in the United States

In partnership with a non-profit organization, we will randomly assign approximately 5,000 low-income individuals in the US to one of three conditions: Control, Small Cash, or Large Cash. The Control group will not receive any payments beyond those given as compensation for their study participation. The Small Cash group will be given a one-time unconditional cash transfer (UCT) of $500. The Large Cash group will be given a UCT of $2,000.

In addition, each of these groups will be cross-randomized to either have access to the partner non-profit organization’s online platform or not. The purpose of the platform is to help program participants connect with one another, potentially fostering an environment of mutual help-seeking and -giving. However, our primary analyses will collapse across these groups, examining only the main effects of varying UCT amounts.

2020-08-12

2021-12-31

We have four primary classes of outcomes: financial well-being, psychological well-being, cognitive capacity, and physical health. Each of these four classes of outcomes will be composed of several measures and aggregated into an index. We list the outcomes we will include in our indices in the analysis plan. All survey measures are included in the “Docs & Materials” section. With one exception (described below), all outcomes will be measured at all time points.

Please see the analysis plan for a description of how the indices will be constructed.

In addition to the four indices, we will also measure several other outcomes for exploratory purposes. These outcomes will not be reported as our confirmatory hypotheses. All survey measures are included in the “Docs & Materials” section.

All participants will be adults living in the United States who have signed up to receive financial assistance from the non-profit organization. To be eligible, participants must be low-income (defined in varying ways depending on the region and funder restrictions), have been adversely affected by the coronavirus pandemic, and have not previously received funding from the partnering non-profit organization.

Participants consenting to participate in the study will be asked to complete a profile survey (which will ask about basic demographics) and a baseline survey (“t1 survey”). Participants who complete both the profile and t1 surveys will then be randomly assigned to either receive no UCT (“Control” group), a $500 UCT (“Small Cash” group), or a $2,000 UCT (“Large Cash” group). All UCTs will be paid approximately 3.5 weeks after the t1 survey is completed (the reasoning for this is explained in the next paragraph).

Approximately one week after the UCT payments are made, participants will be asked to complete the t2 survey. One month after the t2 survey, participants will be sent the t3 survey. Two months after the t3 survey, participants will be sent the t4 survey. Note that this timing ensures that all four surveys (t1, t2, t3, and t4 surveys) are administered at approximately the same time of month. The timing of the UCT payment (exactly 1 week before the t2 survey, or approximately 3.5 weeks after the t1 survey) allows us to capture short-term (1 week) effects of the payments. Conditional on us being able to secure funding, participants will also receive a fifth survey several months after t4 (at t5), which will allow us to capture even longer term effects.

Participants will have as much time as they want to complete the profile and t1 surveys. For all subsequent surveys, they are given up to 10 days to complete each survey. The surveys will vary in small ways across the time periods.

Participants will be paid $20 for each survey they complete, plus an additional $20 completion bonus if they complete all the surveys. In addition, participants will receive 2 lottery tickets for completing the profile and first main survey, and one lottery ticket for each survey they complete after that. They receive an additional 5 lottery tickets if they complete all the surveys. Each lottery ticket is worth a chance at winning an additional $1,000 prize. Payments for each completed survey will be made immediately after the survey is completed. The completion bonus and lottery payments (if any) will be made after the fourth survey. A third party organization will deposit both the UCTs and survey payments directly into participants’ bank accounts, the information for which must be provided at sign up.

In addition, all participants will also be randomly assigned to either have access to the organization’s online platform or not. This platform aims to connect households receiving help from the organization to support social capital creation and maintenance. As described further in our analysis plan, however, our primary specifications will collapse across these groups and only examine the main effects of the UCT amounts.

Randomization will be done in an office using a computer program.

The randomization will be done at the individual level.

No

We aim for a final sample of approximately 4,502 participants. To account for attrition (assumed to be 10%), we are asking the partner organization to enroll approximately 5,002 participants. However, the exact number of participants the partner organization ultimately recruits and we analyze will depend on potential participants’ interest in enrollment, retention, and budgetary constraints. The partner organization will recruit participants until we have run out of money (in our case, $2,500,000) or the participant pool has dried up, whichever comes first.

Same as the planned number of clusters.

Given that the Large Cash group is 2.5 times as expensive as the Small Cash group, and the Small Cash group is 5 times as expensive as the Control group, we employ unequal sample sizes to maximize power given our budget (Guo & Luh, 2009; Schouten, 1999; Torgerson & Campbell, 1997). Collapsing across the online platform randomizations, we aim for a final sample of 2,768 individuals in the Control group (initial sample of 3,076 individuals), 1,130 in the Small Cash group (initial sample of 1,256 individuals), and 604 in the Large Cash group (initial sample of 671). (Splitting across the online platform randomizations, each of these groups would be divided by two.) As we mention above, the exact number of participants will ultimately depend on participants’ interest, our ability to retain them in the study over time, and our budget.

We calculate our minimum detectable effect sizes using our final sample sizes (post-attrition), a 90% target power level, one baseline measure, and one endline measure. We conservatively assume a 0.10 correlation between baseline and endline measures and equal standard deviations across groups. We begin with a standard alpha level of 0.05 and adjust it using a Bonferroni multiple hypothesis testing correction for 4 outcomes (financial, psychological, cognitive capacity, and physical health). Using these parameters, we estimate a minimum detectable effect size of 0.14 SDs for testing the difference between Control and Small Cash for each of our 4 indices, 0.17 SDs for testing the difference between Control and Large Cash, and 0.19 SDs for testing the difference between Small Cash and Large Cash.

Pre Analysis Plan