Yes — Your Designated Roth Contributions Are Counted in 401(k) Nondiscrimination Testing

The test counts the deferral. It has no interest in the tax label.

A designated Roth contribution is not a different kind of contribution. It is the same elective deferral, with a different tax character attached at the moment it goes in. You gave up the deduction; the plan still recorded a deferral. The ADP test — the actual deferral percentage test — looks at deferrals as a share of pay, then compares the average for the highly compensated group against the average for everyone else. Roth dollars sit in that average exactly like pre-tax dollars.

This is the part people get backwards. There is an intuition that Roth money is somehow already settled, already taxed, therefore out of scope. Testing is not about your taxes. It is about whether the plan is being used disproportionately by better-paid people. A Roth election does nothing to change the answer to that question.

Two other buckets are tested elsewhere, and confusing them costs money. Employer match and after-tax non-Roth contributions run through the ACP test, not the ADP test. So if you are contributing after-tax dollars with the intention of converting them inside the plan later, that money is exposed to a different test with a different failure mode — one that can be triggered by exactly the same problem, a workforce where the lower-paid half is not participating much. And an in-plan Roth conversion is not a contribution at all, so it is not tested again; the money was tested in the year it was originally deferred.

One limit governs your pre-tax and Roth deferrals together, which is a separate constraint from testing and worth keeping straight. We cover that split elsewhere.

The failure does not reach you as a memo. It reaches you as money you did not ask for.

When a plan fails the ADP test, the correction usually runs downhill from the highest deferrers in the highly compensated group. The plan hands part of your deferral back. You receive a distribution and a 1099-R. The plan sponsor has a deadline for making that correction cleanly, and missing it exposes the employer to an excise tax — confirm the current deadline and correction options with the plan administrator, because the mechanics have been revised more than once.

Here is what makes a Roth refund different from a pre-tax refund. On the pre-tax side, the returned deferral becomes taxable income, which is annoying but symmetrical — you never got the deduction to begin with, so you are back where you started. On the Roth side, you already paid tax on that dollar. Returning it does not undo the tax you paid. What comes back is basis plus the earnings attributable to it, and the earnings portion is generally taxable. So you paid tax going in, you pay tax on the growth coming out, and the dollar is now sitting in a taxable account instead of inside a tax-free wrapper.

That last part is the real cost and it is quiet. The refund does not restore your contribution room. The year's shelter is gone. If you were deliberately loading Roth space during your final high-earning years — because you expect a stretch of low-income years afterward and wanted a pool of money that never gets taxed again — a corrective distribution removes exactly the asset you were trying to build, and there is no second attempt at that year.

Which source gets refunded is not always fixed. Some plan documents specify pro-rata across pre-tax and Roth. Some let the administrator or the participant direct it. The answer lives in the plan document, and asking for the language before December is a very different conversation from asking after the 1099-R arrives.

Whether any of this can touch you was decided by the plan's design, not by your election.

Some plans are built so the ADP test never applies. A safe harbor design trades a required employer contribution for relief from that test. If your plan uses one, the whole question is largely academic for you — although the ACP test can still matter if the plan accepts after-tax non-Roth money.

If your plan is not safe harbor, your deferral capacity is functionally rented from your coworkers. It depends on how much the non-highly-compensated group defers, and you have no lever on that. This is the structural point: your Roth election looks like a personal decision made in a benefits portal, and it is not. It is a personal decision inside a collective result.

Design changes are also calendar-bound. A sponsor generally cannot decide in October that this year will be a safe harbor year. Elections that reshape a plan year tend to have to be made before that year starts. So an employee who discovers the problem when the refund lands has already missed the window to influence the following year — unless they raise it early enough for the sponsor to act on the next plan year.

Ask the plan administrator three factual questions: is this plan safe harbor, has it failed ADP or ACP testing in any of the last few plan years, and are interim testing results run mid-year. A plan with a history of failing is not a surprise. It is a pattern nobody bothered to tell participants about.

If you own the business you are about to leave, the test is your problem twice.

An owner-employee is almost always in the highly compensated group, regardless of what the payroll says, because ownership can determine that status on its own. So the person with the most to gain from maxing Roth deferrals in the final years before a sale or a wind-down is also the person first in line for the refund.

This collides with the exit in a specific way. Workforce changes in the run-up to a sale — layoffs, a hiring freeze, turnover among lower-paid staff, part-timers dropping out of the plan — move the non-highly-compensated average. The plan can start failing tests it used to pass, for reasons that have nothing to do with the plan and everything to do with the transaction. Meanwhile the buyer's diligence will look at plan compliance history, and an uncorrected testing failure is a real item to resolve, not a footnote.

There is a second timing trap. Plan termination or a final short plan year still requires testing, and the correction has to happen while there is still an administrator and a payroll system to do it. Sequencing the plan's end against the sale's close is one of those small operational decisions that gets left to whoever is least busy, and it is the kind of gap that only becomes visible afterward. The sale itself carries enough tax structure of its own that the retirement plan tends to fall off the agenda.

A refund that lands in your first retirement year hits a different tax picture than the one you planned around.

Corrective distributions typically arrive in the calendar year after the deferral. For someone who retires at the end of a plan year, that means the refund shows up in the first year of retirement — the year they were most likely engineering to be low-income. Which tax year the amount is reportable in has been changed by legislation and guidance, so confirm the current treatment with the plan administrator and whoever prepares your return rather than assuming the rule you last read still holds.

The knock-on effects are where this stops being a small number. Income in a given year can feed into Medicare premium determinations on a lookback basis, into the affordability calculation for Marketplace coverage if you are retiring before Medicare eligibility, and into how much of a Roth conversion you can do that year before crossing into a higher bracket. An unexpected distribution is not large in isolation. It is large when it sits on top of a conversion you already executed, or when it pushes a coverage calculation past a cliff you were steering around by a few hundred dollars.

This is the classic gap between advisors. The recordkeeper knows the test failed. The third-party administrator processes the correction. Payroll issues the money. The CPA sees the 1099-R next spring. The person doing your conversion plan in November knew nothing about any of it. Each one did their piece correctly and the outcome was still wrong.

What you can still change is narrow but real: how much you defer and how early in the year you defer it, whether you ask for the plan document's refund ordering language, whether the sponsor considers a design change for the next plan year, and whether the people planning your retirement-year income know a refund is possible. What closes is the year itself. Once the correction is processed, that year's Roth space is gone and cannot be rebuilt.

What to actually do

• Ask the plan administrator, in writing, whether the plan is a safe harbor design for the current plan year, and whether ADP and ACP testing apply to you.
• Ask whether highly compensated employees have received corrective distributions in any of the last three plan years, and roughly how large they were — a repeat pattern is a planning input, not a surprise.
• Request the plan document language on which source corrective distributions come from: pro-rata across pre-tax and Roth, a fixed order, or participant direction.
• Ask whether interim testing is run before year-end, and get on the list to be told the result while you can still adjust your remaining deferrals.
• If your plan accepts after-tax non-Roth contributions and you were planning to convert them inside the plan, confirm how those amounts affect the ACP test before you rely on that route.
• Write down which calendar year a refund would land in, and check it against the other income decisions already scheduled for that year — a conversion, a coverage application, a lookback year for Medicare premiums. Confirm the current rules for each against the primary source.
• Tell one person — the one who sees your whole tax picture — that a refund is possible, so the 1099-R is not the first anyone hears of it.

How this shows up

A senior engineer plans to retire at the end of the year and deliberately loads Roth deferrals in the final months, wanting a pot of money that will never be taxed again during the low-income years before Social Security begins. The plan fails the ADP test. In March, part of the deferral comes back with attributable earnings, reported on a 1099-R, in the exact year the household was holding income down to qualify for Marketplace coverage at a manageable cost. Nothing was done incorrectly. The refund was simply not in anyone's model.

The owner of a fourteen-person firm expects to sell within two years and has been maxing Roth deferrals to use up the last high-bracket years. A hiring freeze and turnover among lower-paid staff shift the non-highly-compensated average downward. The plan begins failing a test it passed for years, the owner's own deferrals are the first refunded, and the correction history surfaces in the buyer's diligence — where it becomes a line item to fix rather than a private annoyance.

Frequently Asked Questions