Increasing employer compliance with state auto-IRA plan

We will randomly assign employers to one of the three conditions with equal probabilities:

Control - Status quo: the current compliance letter used by the state.
Message A - Benefits: a simplified, formal letter highlighting benefits of program participation for employees and the workforce.
Message B - Process: a simplified, formal letter highlighting how easy and costless the process is for the employer.

Each letter asks employers to either register for the state auto-IRA program, or report that they are exempt from the program because they offer a qualifying retirement plan. Both actions are taken via a state website.

2026-02-19

2026-04-03

Initiation of registration or exemption process
Registration or exemption submission

Initiation of registration or exemption process: an indicator for beginning the registration or exemption process via the state website in the four weeks following the mail date. Note: this outcome depends on data availability; if data on process initiation are not available, our primary outcome will be registration or exemption, as detailed below.

Registration or exemption submission: an indicator for submitting the state auto-IRA registration for or the exemption form in the four weeks following the mail date.

Employer Upload of Employee Information

Employer Upload of Employee Information: among employers who register for the state auto-IRA program, we will construct an indicator for uploading the employee information required for program implementation – the next step in the registration process – in the four weeks following the mail date.

In a stratified randomization, the 8,174 employers that are part of the experimental universe will be randomly assigned to one of the three conditions with equal probabilities:

Status quo: the current compliance letter used by the state.
Message A - Benefits: a simplified, formal letter highlighting benefits of program participation for employees and the workforce.
Message B - Process: a simplified, formal letter highlighting how easy and costless the process is for the employer.

Each letter asks employers to either register for the state auto-IRA program, or report that they are exempt from the program because they offer a qualifying retirement plan. Both actions are taken via a state website.

The randomization will be stratified by (a) employer location in a metropolitan vs non-metropolitan geographic area, based on zip code areas and the RUCA codes; (b) a categorical measure of employer size; and (c) a binary variable for whether a uniquely identified employer shares an address with another employer in our sample.

We will check randomization balance across the stratification variables using a multinomial regression model (for differences across all three treatment conditions). If a joint significance test of all covariates yields p < .05, we will re-randomize.

Not available

Randomization done in office by a computer

Employers

No

Status quo: the current compliance letter used by the state (N = 2,728 employers)
Message A - Benefits: a simplified, formal letter highlighting benefits of program participation for employees and the workforce (N = 2,725 employers)
Message B - Process: a simplified, formal letter highlighting how easy and costless the process is for the employer (N = 2,721 employers)

8,174 employers

Status quo: the current compliance letter used by the state (N = 2,728 employers)
Message A - Benefits: a simplified, formal letter highlighting benefits of program participation for employees and the workforce (N = 2,725 employers)
Message B - Process: a simplified, formal letter highlighting how easy and costless the process is for the employer (N = 2,721 employers)

Based on the experimental universe of 8,174 employers, and assuming that, at baseline, 2% of employers will initiate the process after being contacted, power of 80%, and a standard alpha of 0.05, the minimum detectable effect (MDE) is 0.9 pp for the pairwise comparison of simplified outreach vs. status quo outreach, and 1 pp for the pairwise comparison between employers assigned to receive Message A - Benefits vs. Message B - Process.

Analysis plan