Racial and Ethnic Disparities in Retirement Outreach among Low-to-Middle Income Workers

This project’s primary goal is to design and pilot a digital outreach program/intervention intended to improve retirement knowledge and preparedness among low to moderate income (LMI) Black and Hispanic workers. The precise details of the intervention and experimental design will depend on findings from a screening survey and an extensive literature review (e.g., Viceisza et al., NBER Working Paper 30456, 2022). Those details will be filed as an update to this initial submission.

*** ADDED May 24, 2023 ***

The screening survey closed in mid-April, leading to three main findings.

First, per question 2 in the survey N=796 individuals do not have an employer-sponsored retirement account themselves, that is, they selected categories 2, 3, or 4. This is the sample that will be exposed to our intervention. It is distributed as follows: 384 White, 110 Black, and 302 Hispanic people.

Second, per question 3 in the survey a significant proportion of individuals (~72 percent) selected categories 1, 2, 3, or 4. This is the sample that we consider at risk of not saving for retirement. We return to how this impacts the experimental design in the relevant section further below.

Third, per question 4 in the survey most individuals identified two primary barriers to saving for retirement: (A) not having enough money to put in a retirement account (category 2) and (B) not knowing enough about retirement savings account (category 3). Recall that the categories were presented in random order. This result underscored our priors, and thus previously proposed treatments, as discussed in the experimental design section further below.

The screening survey is included under Docs and Materials.

*** ADDED May 24, 2023 ***

The baseline survey and experiment protocol are included under Docs and Materials.

2023-04-01

2023-07-31

The primary outcomes as of now are likely to be (1) retirement knowledge as measured by the Retirement Knowledge Scale/RKS (Blanco et al. 2020) and (2) willingness to open/invest in a retirement account. However, this could change based on a findings from the screening survey.

*** ADDED May 24, 2023 ***

The baseline survey that is being included with this update comprises the following outcomes:
1. Retirement account ownership, whether employer-sponsored or not (self-reported)
2. Current retirement contributions and frequency (self-reported)
3. Future retirement contributions/plans based on hypothetical and real (experiment windfall/lottery) income changes
4. Hypothetical allocation of windfall spending
5. Retirement knowledge as measured by RKS
6. Financial literacy

The RKS scale measures three important areas related to retirement knowledge and preparedness: 1. Reflection about retirement plans (questions 1 and 2); 2. Engagement with information related to financial planning for retirement (questions 3, 4, 5, 6, and 7); and 3. Evaluation of retirement preparedness (questions 8, 9 and 10). RKS items are coded lower (higher) for those answers that represent low (high) levels of reflection, engagement, and evaluation of retirement preparedness. We transform these responses linearly to a 0-100 possible range, assigning 0 for 1, 33.3333 for 2, 66.6666 for 3 and 100 for 4. We will use the average of the item responses such that the RKS scales also have a 0-100 possible range. Further details are discussed in https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4305160.

The secondary outcomes as of now are likely to be self-reported (1) retirement account ownership, (2) balances in such accounts, and (3) familiarity with and usefulness of information provided about retirement preparedness. The outcomes could also change based on findings from the screening survey.

*** ADDED May 24, 2023 ***

These secondary outcomes are included in the previous list (see primary outcomes). We might add some questions on usefulness of information provided in the intervention as part of subsequent surveys. The primary deciding factor will be length of such surveys.

Subjects will be randomly assigned to retirement outreach treatments relative to a control group. The exact nature of the treatments and experimental design will depend on the screening survey and feasibility of execution by the implementing partner.

Tentatively, the experiment will proceed as follows. The conditions will be revised/finalized in consultation with the implementing partner and based on a screening survey.

First, a screening survey will be fielded to identify (A) strata of eligible subjects (for example, those who have not started to plan for retirement versus those who have), (B) their reasons for falling within these strata (for example, their reason for being in that stage of retirement planning), and (C) other relevant types of heterogeneity.

Second, subjects will randomly be assigned to one of the following conditions:

(i) a treatment group that is exposed to an outreach intervention in which they are presented with retirement preparedness materials such as the mySocialSecurity statement and/or a retirement calculator; *** ADDED May 24, 2023 *** We have titled this “information only” condition Group B.

(ii) a treatment group that is exposed to the above plus (possibly hypothetical) “house money” funds (i.e., funds provided by the researchers, with varying stakes) which can be partitioned across (1) a retirement account, (2) a standard savings account, or (3) “cash” (a gift card). The allocation can be however they want, for example, they can choose to place all the funds in only one of these modes. The stakes of the house money might also be randomly varied; *** ADDED May 24, 2023 *** We have titled this “monetary” condition Group C.

(iii) a control group that will only complete pre- and post-surveys. At most, this control group will receive a placebo treatment. *** ADDED May 24, 2023 *** We have titled this “control” condition Group A.

Third, a random subset of subjects will receive a follow-up mailing or email ("nudge") with additional information about retirement planning.

*** ADDED May 24, 2023 ***

The experiment design has not changed significantly. The three conditions are still intact, with the following modifications:

First, Group A (control) will not receive a placebo treatment. They will only participate in surveys.

Second, Group B will receive four weeks of educational materials as well as some follow-up nudges.

Third, as before Group C will receive the same information as Group B PLUS a monetary windfall, that is, a random subset of participants will be chosen to receive an additional payout based on a decision they made during the study. This information will be communicated to them in the consent form (attached) and reiterated in the surveys (attached).

The randomization will be done by the implementing partner's algorithm.

Individual. *** ADDED May 24, 2023 *** Also see “sample size by treatment arms” further below.

No

Not clustered.

300-900 individuals. *** ADDED May 24, 2023 *** Based on a response rate of 80 percent, we are expecting a sample size of N=634 individuals across all three conditions.

100-300 individuals in control, 100-300 individuals in the first treatment, 100-300 individuals in the second treatment.

*** ADDED May 24, 2023 ***

Based on the screening survey question 3:

Participants who answered categories 1 (one) through 4 (four) have an equal chance of being randomized to Groups A, B, or C.

Participants who answered category 5 (Yes, I have been for six months or more) have an equal chance of being randomized to Groups A or B. In other words, this subset of participants will not be randomized to the monetary group.

This will be conducted once we settle on the final set of main outcomes based on the screening survey. *** ADDED May 24, 2023 *** Due to the limited number of racial and ethnic minorities in the UAS, the sample size is based on recruitment/participation constraints. Ex-post power calculations (e.g., a la Maniadis et al. 2014) will be conducted as necessary.