Rethinking safe harbor for employer matching rules

Introduction

Close to 100 million Americans have access to an employer-sponsored defined contribution retirement account, and every year, private employers contribute over $250 billion to their employees’ retirement plans, with most plans offering matching contributions.¹ These matching incentives tie employee compensation to voluntary saving behavior and vary widely across employers, shaping both how much workers save and how employer contributions are distributed. The regulatory framework governing matching formulas has seen little substantive change over the past 20 years. Most recent legislative efforts (e.g., SECURE and SECURE 2.0) have instead emphasized behavioral interventions such as automatic enrollment and automatic escalation. With recent evidence finding more modest effects of these “nudges” than previously thought (Choukhmane 2025; Choi et al. 2024), revisiting the design of matching formulas is pressing. Since the Revenue Act of 1942, federal policy toward employer-sponsored plans has pursued two goals: expanding retirement saving and preventing tax-preferred benefits from accruing disproportionately to highly compensated workers. Matching formulas are ostensibly firm choices, yet these employer decisions are heavily shaped by federal regulation. In practice, as we show, safe harbor designs, which offer employers simplified compliance pathways, function as a powerful policy lever, channeling an increasing number of plans toward a specific set of matching formulas. This brief discusses the limits of current safe harbor matching formulas and proposes a reform. We argue that the existing safe harbor formulas are poorly designed: they feature high match rates applied over narrow contribution ranges, which heavily subsidize saving that would have occurred anyway while failing to extend incentives to a broader set of workers. Drawing on new evidence from Carranza, Choukhmane, Greig, O’Dea & Schmidt (2026),² we propose that matching safe harbor formulas should be reformed to incorporate a non-elective contribution (an employer contribution made to all eligible employees regardless of whether they contribute themselves) paired with a less steep, more extended match. This design has the potential to increase total saving while guaranteeing that every eligible employee receives an employer contribution. It can do so at any chosen level of employer cost.

Challenge: Matching, nondiscrimination testing, and safe harbor rules

Institutional background

Defined contribution (DC) plans are the dominant vehicle for private retirement saving in the United States. Around two-thirds of U.S. workers have access to an employer-sponsored retirement account and most of them work for employers that match their saving. Americans held over $14 trillion of assets in employer-sponsored DC accounts at the end of 2025.³ Federal law requires that the benefits provided by 401(k) plans do not disproportionately favor highly compensated employees.⁴ Israel Preminger & O’Dea (2026) provide a detailed description of the legislative history of nondiscrimination testing and safe harbor. The current section aims to provide a (non-comprehensive) summary. The requirement not to disproportionately favor highly compensated employees is enforced through two ‘nondiscrimination’ tests:

• The Actual Deferral Percentage (ADP) test, which compares the average elective deferral rates (i.e., the contribution employees choose to make) of highly compensated and non-highly compensated employees, and
• The Actual Contribution Percentage (ACP) test, which performs the same comparison for employer matching contributions and employee after-tax contributions.

Plans that fail either of these tests must take (often costly) corrective action or risk losing their tax-qualified status. Amid concerns that the administrative burden and uncertainty of annual nondiscrimination testing were acting as a disincentive to plan formation, Congress created safe harbor provisions that allow employers to satisfy the ADP and ACP tests automatically by adopting specified plan designs. There are two basic safe harbor matching formulas: i) the basic match formula under the ‘traditional’ safe harbor, which requires employers to match 100% of employee contributions up to 3% of compensation and 50% for the next 2% and ii) the basic match formula under the ‘QACA’ (Qualified Automatic Contribution Arrangement) safe harbor, which involves a 100% match on the first 1% of compensation contributed, plus 50% on the next 5%. Employers may alternatively qualify for the safe harbor with an “enhanced” formula that is at least as generous at every deferral rate as the corresponding basic formula, subject to additional constraints (including that the match rate does not increase with the deferral rate). Table 1 summarizes the existing matching safe harbors. Separately, employers may also be treated as satisfying the ADP test through a non-elective contribution of at least 3% of each eligible non-highly compensated employee’s compensation, paid regardless of whether the employee makes any deferrals.⁵ A plan using this non-elective safe harbor may also provide matching contributions. Although such a plan is deemed to satisfy the ADP test, those matching contributions must satisfy the ACP test unless they meet the separate ACP safe-harbor requirements.

Origins and evolution of nondiscrimination testing and safe harbor rules

The requirement that retirement plan benefits be distributed equitably has roots in provisions of the tax code that are almost a century old. The Revenue Act of 1921 first extended tax preferences to employer retirement plans, but it did not regulate the distribution of the benefits across employees. The Revenue Act of 1942 introduced the first nondiscrimination requirements, and the Revenue Act of 1978 established Section 401(k) plans along with the ADP test as a formal mechanism for enforcement. The Tax Reform Act of 1986 created the ACP test and modified the ADP test into the form that remains in place today.

By the 1980s and 1990s, policymakers had raised concerns that nondiscrimination testing discouraged plan formation, particularly among small employers. The Small Business Job Protection Act of 1996 responded by creating traditional safe harbor plans that allowed employers to satisfy the tests automatically, with the goals of increasing employee participation and encouraging more employers to offer 401(k) plans. The specific numerical formulas in these plans do not appear to have been chosen based on evidence about their effects on saving behavior. The origins of the basic match (100% on the first 3%, 50% on the next 2%) are not entirely clear, but the Federal Thrift Savings Plan matching structure was referenced in congressional discussions and plausibly served as the model. The Pension Protection Act of 2006 subsequently added the QACA safe harbor, pairing a modified matching formula with auto-enrollment, but the structure of the safe harbor matching formulas has otherwise remained unchanged for nearly three decades.

Safe harbor formulas rise to dominance

Choukhmane, Dedyo, O’Dea, & Schmidt (2026) document that safe harbor formulas have come to dominate the landscape of retirement plan design. Using a newly constructed dataset that aims to cover the universe of U.S. retirement plans with more than 100 participants, they find that the share of matching plans whose formulas meet or exceed the safe harbor matching requirements has grown dramatically over the past two decades, from under a fifth of plans in 2003 to close to half by 2023.

Figure 1 illustrates these trends. Figure 1a shows the share of matching plans whose formulas meet or exceed the safe harbor matching requirements. Figure 1b repeats the exercise but counts only plans that also satisfy applicable auto-enrollment and vesting requirements associated with safe harbor status. The levels in Figure 1b are lower, but the upward trend is the same.

Safe harbor formulas are especially persistent. Plans that adopt a safe harbor matching formula are much less likely to change their formulas over any given horizon than plans using non-safe harbor formulas are to change theirs. The regulatory benefits of safe harbor status appear to create a “lock-in” effect: once a plan qualifies for the nondiscrimination testing exemption, firms are less likely to redesign. This stickiness implies that the dominance of safe harbor formulas is likely to persist and deepen in the years ahead.

Despite safe harbor rules effectively determining match design for millions of workers, there has been little formal evaluation of their effects or design, and very little policy innovation in their structure.

Limitations with the current safe harbor formulas

The current safe harbor matching formulas share a common feature: they all begin with a 100% match rate on the first percentage point of employee contributions. While this structure may appear generous, the high initial match rate creates two important limitations.

First, a high initial match rate subsidizes contributions that many employees would have likely made even without the match. Carranza et al. (2026) show that employee saving is inelastic with respect to the match rate: dollar-for-dollar matches are much more costly than lower match rates, and yet they generate only modestly more additional saving. Because the current safe harbor formulas concentrate employer dollars at a high match rate over a narrow contribution range, much of that spending goes to workers who would have contributed at that level regardless. Greig et al. (2024) find that, among Vanguard-administered plans, the majority of matching dollars flow to employees who contribute beyond the match cap, meaning the match provides no marginal incentive to save more. The match, for many individuals, functions more as a windfall than as an incentive.

Second, with around a third of eligible employees not participating in their employer plan, many receive no employer contribution at all under a pure matching formula. The current safe harbor matching design therefore concentrates employer spending among higher-saving employees, who tend to have higher personal, spousal, and parental incomes, exacerbating inequality in retirement wealth accumulation (Choukhmane et al. 2025).

The proposal: Reforming safe harbor matching formulas

We propose that Congress reform the safe harbor matching formulas around two principles: i) any new safe harbor matching plan should include a non-elective employer contribution, guaranteeing all eligible employees receive a minimum contribution regardless of their own saving, and ii) the match component should use a lower match rate applied over a wider range of employee contributions.

To illustrate, consider a formula of a 2% non-elective contribution plus a 25% match on contributions up to 8% of compensation, which produces a maximum employer cost of 4% for an employee contributing 8% or more. This is the same maximum employer cost as under the current basic match, which reaches 4% of compensation for a worker contributing 5% or more in a traditional safe harbor plan.

The important difference is how employer dollars are allocated. Under the current traditional safe harbor match, employer contributions are concentrated among employees who contribute enough to receive the full match. Under this illustrative formula, all eligible employees would receive a 2% contribution, while the match would be spread more gradually over a broader range of employee saving. The result is a formula that preserves a saving incentive while establishing a contribution floor for non-participants and low savers.

Although the maximum employer contribution under the formula described above is the same as that offered under the current traditional safe harbor match, it might cost more for many employers because the non-elective contribution is paid to all eligible employees, including those who do not contribute. The size of the difference would depend on the plan’s workforce, especially the share of eligible employees who do not participate. But the non-elective and match parameters of the formula can be adjusted to meet different cost targets, and the Appendix presents alternative designs with the same structure at different cost levels, including designs with lower maximum employer costs than the current traditional safe harbor match.

In the spirit of the current safe harbor rules, which permit enhanced matching formulas provided they are at least as generous as the basic formula at every deferral level, the new safe harbor could similarly allow employers to exceed the baseline while retaining safe harbor status. Two enhanced formulas that satisfy this requirement would be:

• A 2.5% non-elective contribution plus a 25% match on contributions up to 8% of compensation (max employer cost: 4.5%). This formula directs more to the universal baseline, increasing the contribution floor for non-participants and low savers.
• A 2% non-elective contribution plus a 50% match on contributions up to 6% of compensation (max employer cost: 5%). This formula provides a stronger saving incentive at lower deferral rates, at the cost of a higher maximum employer outlay.

Table 2 illustrates the example proposal alongside the existing safe harbors.

Broadening coverage. The non-elective contribution ensures that all eligible employees accumulate some retirement wealth through their employer, even if they do not contribute to the plan themselves. This addresses a central limitation of pure matching formulas: that non-participants, who are disproportionately lower-income, receive nothing. Evidence from Carranza et al. (2026) shows that non-elective contributions do not crowd out employee saving. If anything, the evidence suggests that workers would contribute more. This means that adding a non-elective contribution will likely lead to a genuine addition to retirement wealth for recipients.

Incentivizing saving. The stretched match component extends the incentive to contribute over a wider range of employee contribution rates. Under the current traditional basic formula, the match incentive is exhausted once an employee contributes 5% of compensation; under the illustrative formula, the match continues to reward contributions over a broader range. Because saving is inelastic to the match rate, reducing the match rate on initial contributions has a modest effect on employee saving, while the extended range provides incentives to workers who might otherwise have stopped contributing at a lower threshold.

The proposal is deliberately framed in terms of principles rather than a single prescribed formula. We do not argue that the specific numbers in the illustrative example above are optimal. Rather, we propose that any reformed safe harbor matching formula should incorporate a non-elective contribution and should stretch the match across a wider contribution range at a lower rate. The precise calibration, including the size of the non-elective contribution, the match rate, and the cap, should be informed by further analysis and balanced against employer cost considerations.

Implementation

We propose that Congress introduce a safe harbor matching formula that combines a non-elective contribution with a stretched match and that satisfies the ACP and ADP tests.⁶ The new formula would require: i) a minimum non-elective contribution of at least a specified percentage of compensation for all eligible employees, and ii) a matching contribution at a rate that does not increase as the deferral rate increases and that extends over a contribution range wider than the current basic formula.

This reform could take one of three forms.

1. One approach would add the new formula as an additional safe harbor option alongside the existing traditional and QACA formulas, expanding employer choice without restricting it. Employers currently using the traditional or QACA safe harbor could transition to the new design if they find it better suited to their workforces, while employers satisfied with their current formula would face no mandatory change.
2. A second approach would replace the existing matching safe harbor formulas with the new design. This would have the advantage of directing all safe harbor plans toward a formula structure that could generate more saving and less inequality in employer contributions, and would not add further complexity to the code, but would require employers currently using the existing formulas to adjust their plan designs if they wanted to avoid losing safe harbor status.
3. A third approach, intermediate between the first two, could grandfather existing safe harbor plans, but require any new adoptions to be of the new proposed form.

The choice between these approaches involves a tradeoff between offering flexibility for employers and maximizing the reach of the reform. Under the first approach, the reform is unambiguously attractive to employers, since it expands the menu of qualifying safe harbor designs without removing any existing choices. However, precisely because it imposes no requirement to change, voluntary adoption may be slow, particularly given the stickiness of existing safe harbor formulas documented above. The second approach would remove options that employers currently use, while the third would restrict those options only for new adopters.

Discussion

Why safe harbor design matters more than it might appear

One might ask why reforming safe harbor formulas would matter if employers are free to choose any matching formula they wish. A response is that the rapid growth in safe harbor adoption, combined with the persistence of safe harbor formulas once adopted, means that the specific formulas written into the safe harbor provisions effectively determine matching designs for a large and growing share of U.S. retirement plans.

Evidence base

The equity case for reform draws on Choukhmane et al. (2025), which uses employer-employee linked data covering millions of Americans to show that, because workers with the same income make different contribution choices, matching incentives are unequally distributed even among similar-income coworkers. Carranza et al. (2026) build on this by linking survey responses to hypothetical matching scenarios with administrative 401(k) records from Vanguard to study how alternative match formulas could improve outcomes. They find that employee contributions are relatively insensitive to the match rate, and that non-elective contributions do not crowd out employee saving. These results suggest that plans pairing lower match rates with non-elective contributions can generate greater total saving and a more equitable allocation of employer contributions, and that many existing plans, including the current safe harbor formulas, are dominated along both dimensions.

The finding that saving responds modestly to the match rate is consistent with the literature on the causal effects of saving incentives. Studies exploiting cross-sectional variation in matches (Engelhardt and Kumar 2007; Mitchell et al. 2007), firm-level changes in match rates (Papke 1995; Kusko et al. 1994; Choi et al. 2002), variation in tax incentives (Ramnath 2013; Chetty et al. 2014), and experiments (Duflo et al. 2006) generally find that matching has a small positive effect on participation and limited effects on contribution rates. Most of these studies estimate the overall response to employer matching, jointly reflecting the match rate and the range over which it applies, and so do not isolate the effect of each margin separately. The limited evidence that does attempt to distinguish between them points to modest responses to both; Engelhardt and Kumar (2007), for instance, estimate an elasticity of total contributions with respect to the match rate of 0.12 in a specification that accounts for the nonlinear incentive created by the cap.⁷

Employer costs and wage responses

A natural concern is whether the proposed reform would increase employer costs. The proposal is best understood as changing the distribution of employer contributions, not necessarily their overall level. As discussed in the proposal section above, a formula with a non-elective contribution and stretched match can be designed so that employer costs per participating employee are comparable to the current traditional safe harbor. The overall change in costs at each employer, however, will depend on the size of the non-elective contribution and the share of eligible employees who do not participate.

A related question is whether employers would adjust wages or other benefits in response. A version of the proposal calibrated to be cost-neutral would not require broad wage cuts to finance higher benefit costs. It would, however, change how retirement contributions are distributed within firms, shifting resources from workers who would have received larger matches under current safe harbor formulas to workers who save less and currently receive little.

If wages adjusted in response, this would matter most if the goal were simply to redistribute total compensation across workers. But the reform’s main goal is to improve how compensation is delivered. Even if the total compensation of each worker remained unchanged, shifting compensation into retirement contributions rather than wages could increase retirement saving among workers who currently save little.

Vesting and auto-features

While this proposal focuses on matching, vesting schedules and auto-features are related plan-design choices that shape the equity effects of employer contributions. Carranza & Goodman (2025) find that vesting schedules increase the regressivity of employer contributions, while Choukhmane (2025) finds auto-enrollment reduces inequality in the distribution of employer contributions.

Under QACA (but not traditional) safe harbor rules, employers may impose a two-year cliff vesting schedule on matching contributions. If the non-elective contribution proposed here is subject to vesting, its equity benefits would be diluted for workers with short tenures.

The proposed structure could also be paired with auto-enrollment and auto-escalation. Auto-enrollment could help new hires capture the match, while auto-escalation might help more workers move along the stretched match schedule over time.

Distributional impacts

The precise distributional consequences of the proposed reform are difficult to evaluate, because they depend on which implementation approach is adopted, what existing formulas the new safe harbor would replace in practice, and how employer and employee behavior adjusts in response. That said, the clearest beneficiaries of the reform would be eligible employees who do not currently contribute to their employer’s plan. Under existing matching safe harbors, these workers receive nothing; under the proposed formula, they would receive the non-elective contribution. More broadly, the evidence from Carranza et al. (2026) suggests that combining a non-elective contribution with a stretched match would raise total saving rates on average, and would better target additional contributions toward lower-income workers and those who would otherwise save little.

Conclusion

Safe harbor rules, which were designed to simplify compliance with nondiscrimination testing, have gradually become a dominant force shaping how millions of workers save for retirement and how hundreds of billions of dollars in employer retirement contributions are distributed. Today, close to half of all matching plans use formulas that meet or exceed the safe harbor matching requirements. New evidence shows that these formulas are poorly targeted: they direct employer dollars toward high match rates over narrow contribution ranges, heavily subsidizing saving that many workers would have done anyway, while excluding non-participants, who are arguably most at risk of having inadequate saving.

We propose a new safe harbor formula that combines a non-elective contribution with a stretched match. The evidence is clear that safe harbor design matters: it shapes employer choices, can lock in plan features for years, and determines both how much workers save and who receives employer contributions. It is time to update these formulas to reflect what we have learned about how matching incentives work.