Fifty Years of Testing Without Proof of Results

Discrimination testing has been part of the 401(k) system for more than 50 years and was designed to ensure that plans benefit rank and file employees, not just owners and highly compensated employees. It was introduced to address real abuses in earlier plan designs, but it is worth asking whether it continues to function as intended.

There is no shortage of research on participant behavior, including auto-enrollment, escalation, and plan design, and entire bodies of literature have been built around what drives participation and savings outcomes. Despite decades of use, there is no widely cited body of empirical evidence demonstrating that discrimination testing itself drives improved outcomes for non-highly compensated employees.

If testing were effective, you would expect to see measurable improvements tied to it. Instead, the factors that consistently move the needle are design based, especially auto-enrollment and escalation. Many of the behavioral tools that actually improve participation were not available when ERISA’s testing framework was created.

What testing does reliably produce is cost and complexity, as plans hire TPAs to run tests, correct failures, and in many cases use cross-testing to achieve contribution flexibility that the system otherwise restricts. Cross-testing exposes the core contradiction: the rules restrict outcomes, then permit more complex methods to achieve similar results anyway, which suggests the restriction itself is not doing what it claims to do.

Safe harbor already provides a way to eliminate testing, but it requires employer contributions with immediate vesting, which many employers are reluctant to adopt. In many plans, non-highly compensated employees bear a portion of the administrative costs associated with testing and cross-testing, even though those structures are frequently used to support contribution flexibility for highly compensated employees.

There is also an assumption built into the system that large differences in contributions between highly compensated and non-highly compensated employees are inherently unfair. In many cases, those differences are driven by income rather than plan design, and some employees simply do not have the financial capacity to defer a meaningful portion of their pay, regardless of how the plan is structured or how much education is provided. In those cases, limiting contributions for those who can save more does not necessarily improve outcomes for those who cannot, and in some instances it simply reduces overall savings without meaningfully increasing participation or balances among non-highly compensated employees.

Auto-enrollment with escalation directly addresses participation and savings behavior without requiring annual recalculation or correction cycles, and it allows participants to opt out, which matters given that many employees may not be in a position to contribute consistently. Top-heavy rules and other layers of compliance are designed to address participation and concentration after the fact, while design based approaches address those issues more directly by building participation and balances over time.

Plans could be given a clear choice: continue with testing, or adopt a design based structure such as auto-enrollment with escalation that eliminates the need for both testing and cross-testing.

The goal should be improved participation and outcomes, not maintaining layers of compliance that lack a demonstrated connection to them.