Dependent Care FSA Nondiscrimination Testing: 55% and 25% Tests

A Dependent Care FSA must pass three nondiscrimination tests under Internal Revenue Code Section 129 every plan year, and if it fails, the highly compensated employees and larger owners in the plan lose the tax-free treatment on the amount that caused the failure. That reclassified benefit becomes taxable wages subject to income tax, Social Security, and Medicare tax. With the exclusion capped at $7,500 per employee ($3,750 if married filing separately), a failed test can hand the company’s leadership a surprise tax bill on money they thought was already sheltered. Dependent Care FSA nondiscrimination testing exists to prevent that outcome from being planned around.¹

The three tests under Section 129(d) each ask a different question. The eligibility test asks whether the right people can join the plan. The 55-percent average benefits test asks whether the dollars flowing through the plan are spread reasonably between rank-and-file employees and highly paid ones. The concentration test asks whether more than a quarter of the plan is going to the largest owners and their families. A plan has to pass all three.

Who the Tests Are Actually About

Two employee groups drive the results. They overlap but are not the same, and both need to be identified before any testing runs.

A highly compensated employee under IRC Section 414(q) is either a 5-percent owner at any point during the current or prior plan year, or an employee whose compensation exceeded a set threshold in the prior year. For the 2026 plan year, the compensation threshold is $160,000 in 2025 pay.² An employer can narrow the compensation-based group by electing the top-paid group rule, which limits HCE status to the top 20 percent of earners.³

The concentration test looks at a smaller circle: individuals who own more than 5 percent of the company’s stock, capital, or profits interest, together with their spouses and dependents. For a corporation, that is more than 5 percent of the outstanding stock or voting power; for a partnership or LLC, more than 5 percent of capital or profits.¹ Every more-than-5-percent owner is also automatically an HCE, so these individuals are subject to all three tests.

Family Attribution Can Change Who Is an Owner

Ownership for testing isn’t limited to shares held in a person’s own name. Under IRC Section 318, stock owned by a spouse, children, grandchildren, or parents is attributed to the individual.⁴ An executive who personally holds 2 percent while her spouse holds another 4 percent is a more-than-5-percent owner for testing purposes. A founder who transferred all his shares to his adult children still counts as owning them. Attribution runs through spouses, children, grandchildren, and parents, and it does not chain: stock attributed from a child to a parent cannot then be re-attributed sideways to a sibling.

The Eligibility Test

Section 129(d)(3) requires that the plan benefit employees under a classification the IRS would not consider discriminatory in favor of HCEs.¹ The evaluation borrows principles from the coverage rules under Section 410(b): the classification has to be reasonable, based on objective business criteria, and the coverage rate for non-HCEs should be at least 70 percent of the coverage rate for HCEs.⁵

Opening the plan to all employees is the cleanest way through this test. Plans that limit eligibility to salaried staff, management, or specific locations have to show the resulting group doesn’t tilt toward HCEs.

The 55-Percent Average Benefits Test

Even a plan open to everyone can fail on the dollars. Section 129(d)(8) requires that the average benefit received by non-HCEs be at least 55 percent of the average benefit received by HCEs.¹

The calculation divides total dependent care benefits used by each group by the total number of eligible employees in that group, including employees who chose not to participate. That denominator is where most failures come from. If 200 non-HCEs are eligible but only 15 enroll, the non-HCE average is diluted across all 200, and the ratio collapses.

Section 129(d)(8)(B) offers one relief valve. When benefits are provided through salary reduction, the plan may exclude employees earning less than $25,000 from the average benefits calculation.⁶ Low earners are the least likely to enroll, and pulling them out of the denominator often rescues the ratio. Model results both ways before deciding.

The 25-Percent Concentration Test

Section 129(d)(4) sets a hard cap: no more than 25 percent of all dependent care assistance the employer pays or incurs during the year can go to more-than-5-percent owners and their spouses and dependents.¹ Both direct employer contributions and salary-reduction amounts count toward the total.

This is the test that trips up small companies. If two equal owners of a 12-person shop both elect the maximum, they can easily consume more than a quarter of the plan. Say total plan reimbursements for the year come in at $40,000 and the two owners account for $15,000 of that. The owners are at 37.5 percent, and the test fails. Employers with concentrated ownership need to watch claim volume during the year rather than discovering the problem after the plan year closes.

Who Can Be Excluded From the Testing Pool

Section 129(d)(9) lets an employer disregard several categories of employees when running the eligibility and average benefits calculations, which can meaningfully change the outcome.¹

Employees who have not turned 21 and completed one year of service.
Employees whose benefits were the subject of good-faith collective bargaining, when testing a plan that covers only non-union employees.
Nonresident aliens with no U.S.-source earned income from the employer.

Whatever exclusions the employer applies have to be consistent from year to year. Picking and choosing to manipulate results is not permitted.

Controlled Groups Have To Be Combined

Companies under common ownership cannot test in isolation. Under IRC Sections 414(b) and 414(c), entities within a controlled group or affiliated service group are treated as a single employer for Section 129 purposes.⁷ Parent companies, subsidiaries, and brother-sister entities all get pulled into the same testing population. A professional practice that separates its owners into one entity and its staff into another still has to combine them for testing. The same applies to franchise structures and holding-company arrangements above the statutory ownership thresholds.

Running the Tests

Most employers run the tests as of the last day of the plan year, when participation, elections, and reimbursements are all final. Three categories of data are needed across every employee in the controlled group.

Compensation and ownership data: prior-year compensation to identify HCEs above the applicable threshold, current and prior-year ownership percentages for every owner and their family members, and officer status.
Eligibility data: which employees were eligible to participate during the plan year, whether or not they enrolled.
Utilization data: election amounts and actual reimbursements paid to each participant during the plan year.

The math itself is mechanical. For the eligibility test, compare the percentage of non-HCEs eligible against the percentage of HCEs eligible. For the 55-percent test, calculate average benefits for each group using all eligible employees as the denominator. For the concentration test, add up reimbursements to more-than-5-percent owners and their families and confirm the total stays at or below 25 percent of plan-wide reimbursements. Complete testing soon after the plan year closes, before W-2s go out, so any taxable excess can be captured on the correct year’s forms.

When a Test Fails

If preliminary testing signals a likely failure, the most common in-year correction is reducing HCE elections proportionally to bring the plan back into range. Cutting an HCE’s $7,500 election to something lower is not a welcome conversation, and it works best when the plan document reserves the right to adjust elections mid-year.

The longer-term fix is bringing more non-HCEs into the plan. Enrollment campaigns, education about the tax savings, and auto-enrollment features where the plan permits them can lift participation and improve the average benefits ratio. This does nothing for a plan year that has already closed, though.

If the plan fails and no correction is made, the discriminatory excess is reclassified as taxable wages for the affected HCEs and owners. That amount is subject to federal income tax, Social Security tax at 6.2 percent, and Medicare tax at 1.45 percent, and the employer owes its matching share of payroll taxes. Non-HCE participants keep their tax-free benefit regardless of the failure.¹ The reclassified amount goes on the affected employee’s Form W-2 for the year the benefit was provided, and getting the W-2 reporting wrong carries its own information return penalties.⁸

A Note on Form 5500

Nondiscrimination testing sits separately from Form 5500 filing, but the two often come up together. Welfare plans covered by ERISA generally have to file Form 5500 annually, with an exemption for plans that cover fewer than 100 participants, are unfunded or fully insured, and are not subject to Form M-1 requirements.⁹ Most DCFSAs are funded entirely through salary reduction and stay under 100 participants, so many employers qualify for the exemption. Those that do have to file are due by the last day of the seventh month after the plan year ends — July 31 for calendar-year plans.¹⁰

1
Office of the Law Revision Counsel. 26 USC 129 Dependent Care Assistance Programs
2
Internal Revenue Service. Notice 2025-67 – 2026 Amounts Relating to Retirement Plans and IRAs
3
Internal Revenue Service. Identifying Highly Compensated Employees in an Initial or Short Plan Year
4
Office of the Law Revision Counsel. 26 US Code 318 – Constructive Ownership of Stock
5
Office of the Law Revision Counsel. 26 USC 410 Minimum Participation Standards
6
Office of the Law Revision Counsel. 26 US Code 129 – Dependent Care Assistance Programs
7
Internal Revenue Service. Controlled and Affiliated Service Groups – Related Employers Phone Forum Presentation
8
Internal Revenue Service. Information Return Penalties
9
U.S. Department of Labor. Instructions for Form 5500
10
Internal Revenue Service. Form 5500 Corner