Getting Going on Savings: A Randomized Trial of Roth IRA Seeding for Young Workers

The intervention is designed to reduce upfront financial and informational barriers to Roth IRA saving among young adults in summer employment programs. All study enrollees will be offered an initial $50 contribution if they open a Roth IRA and successfully deposit the initial $50 contribution. Randomization will occur among participants who open accounts and successfully deposit the initial $50. Randomized participants will then be assigned to one of two contribution conditions. Participants assigned to the treatment condition will receive an additional $950 Roth IRA contribution, for a total program contribution of $1,000. Participants assigned to the comparison condition will receive an additional $50 Roth IRA contribution, for a total program contribution of $100. Both groups may receive baseline financial education, Roth IRA information, account-opening support, reminders, and guidance on funding and investing the account as specified in the final implementation protocol. If implemented, an independent messaging cross-randomization will assign participants to standard messages or additional messaging related to habit formation, reminders, Roth IRA funding, saving behavior, compound growth, account flexibility, or continued account engagement.

2026-07-01

2027-09-01

Roth IRA balance at the primary follow-up date (approximately 12 months after initial contribution)

Roth IRA balance will be measured using administrative brokerage account records at the primary follow-up date. Because randomization will occur among participants who have already opened a Roth IRA and successfully deposited the initial $50 contribution, Roth IRA account opening is part of the randomized-sample definition rather than a post-randomization outcome. Participants with no active observed Roth IRA balance at the primary follow-up date, including participants who close the account before follow-up, will be assigned a balance of zero where appropriate. The primary follow-up horizon will be specified before outcome data are analyzed and is expected to cover approximately 12 months after randomization.

• Roth IRA engagement: any Roth IRA funding, total contributions, participant-initiated contributions, net contributions, any withdrawal, total withdrawals, account closure, portfolio allocation, automatic reinvestment, and external-account link or recurring contributions.
• Financial knowledge and self-efficacy: financial literacy, comfort managing finances, confidence meeting financial goals, and trust in financial institutions.
• Savings behavior and financial management: any saving, amount saved, budgeting, and savings or investment accounts outside the study Roth IRA.
• Income, spending, and hardship: employment/enrollment, monthly income and spending, financial hardship, and use of alternative financial products.
• Credit report domain, if available: credit access, balances and borrowing, delinquency, collections, default, and bankruptcy.

Secondary outcomes will be measured using administrative brokerage records, participant surveys, credit report data, and program/administrative records where available. Roth IRA engagement outcomes will be constructed from account-level and transaction-level brokerage records, including balances, contributions, withdrawals, account status, portfolio allocation, reinvestment, and linked account or recurring contribution indicators. Survey outcomes will be constructed from participant responses on financial literacy, financial self-efficacy, savings behavior, budgeting, income, spending, hardship, and financial product use. Credit report outcomes, if available, will measure broader financial well-being, including credit scores or score presence, credit limits, balances, delinquency, collections, default, bankruptcy, and types of credit accounts.

The study is a randomized controlled trial among young adults participating in city-run summer employment programs, including the Boston Summer Youth Employment Program and potentially similar programs in other cities. Eligible participants are expected to be ages 18 to 24 and to have sufficient earned income during the study period to contribute to a Roth IRA. All study enrollees will be offered an initial $50 contribution if they open a Roth IRA and successfully deposit the initial $50 contribution. Randomization will occur among the group that opens accounts and successfully deposits the initial $50. Randomized participants will be assigned to one of two contribution conditions: a treatment condition receiving an additional $950 contribution after the common initial $50, for a total program contribution of $1,000, or a comparison condition receiving an additional $50 contribution after the common initial $50, for a total program contribution of $100. The current working design assumes approximately 800 randomized participants, with about 400 participants assigned to each contribution condition. The primary analysis will estimate the intent-to-treat effect of assignment to the higher total contribution relative to the comparison contribution among participants who opened an account and successfully deposited the initial $50 before randomization. Because account opening and the initial $50 deposit occur before randomization, the randomized comparison isolates the effect of the additional contribution amount, assuming all other post-randomization implementation features are held fixed across arms. The study may also include an independent messaging-based cross-randomization in which participants in both contribution conditions are randomized to receive either standard messages or additional messaging. If implemented as a balanced 2 x 2 factorial design, the working allocation would be approximately 200 participants per factorial cell. Final sample size, allocation, treatment descriptions, and stratification variables should be confirmed before submission.

Not available

Randomization will be conducted by computer after participants open a Roth IRA and successfully deposit the initial $50 contribution. There are no plans for pre-specified randomization strata such as city/program site, cohort, workshop, provider, outreach batch, or enrollment wave. If the messaging cross-randomization is implemented, contribution assignment and messaging assignment will be randomized as separate dimensions.

Individual participant. If the messaging cross-randomization is implemented, the messaging assignment is also randomized at the individual participant level.

No

N/A - individual-level randomization

Approximately 800 randomized participants, defined as participants who open a Roth IRA and successfully deposit the initial $50 contribution before randomization.

Approximately 400 randomized participants assigned to the treatment condition and 400 randomized participants assigned to the comparison condition. All randomized participants will have opened an account and deposited the common initial $50 before randomization. Treatment participants will receive an additional $950 contribution, for a total program contribution of $1,000; comparison participants will receive an additional $50 contribution, for a total program contribution of $100. If a balanced messaging cross-randomization is implemented, this implies approximately 200 participants per factorial cell: treatment/standard messages, treatment/additional messaging, comparison/standard messages, comparison/additional messaging.

Power calculations are based on approximately 800 randomized participants, with 400 participants in the treatment contribution condition and 400 participants in the comparison condition. These calculations apply to the randomized sample of participants who open a Roth IRA and successfully deposit the common initial $50 before randomization. Calculations assume two-sided tests with 80 percent power and a 5 percent significance level, individual-level randomization, no covariate adjustment, and no take-up adjustment. Under these assumptions, the standardized minimum detectable effect for continuous outcomes is approximately 0.198 standard deviations. For dollar-valued Vanguard administrative outcomes, the corresponding minimum detectable effect depends on the assumed outcome standard deviation. Illustrative calculations in the PAP imply an MDE of about $19.83 for outcomes with a standard deviation of $100 and about $9.92 for outcomes with a standard deviation of $50. For binary administrative outcomes, illustrative MDEs are approximately 3.84 percentage points when the control-group rate is 2 percent, 5.25 percentage points when the control-group rate is 5 percent, and 6.73 percentage points when the control-group rate is 10 percent. If the final implementation uses clustered randomization or a materially different randomized sample size, power calculations should be updated accordingly.