Do You Still Have to Take an RMD If You’re Still Working?

There is a real exception here, and it is narrower than almost everything written about it suggests.

If you are past the age at which required distributions normally begin and you are still working, one account may be allowed to wait. One. Not your IRAs, not the plans you left behind at former employers, and not your plan at all if you own enough of the business — and “enough” turns out to be a much lower bar than most owners assume, because the tax code counts your family’s stock as yours before it measures.

So this page is mostly about the edges. The rule itself takes a sentence. Everything that matters is in who it excludes.

Do you still have to take an RMD if you’re still working?

From an IRA, yes. From the plan at the employer you are still working for, possibly not.

The rule sits in the definition of the required beginning date. 26 U.S.C. §401(a)(9)(C)(i) defines that date as April 1 of the calendar year following the later of “the calendar year in which the employee attains the applicable age” and “the calendar year in which the employee retires.” Treas. Reg. §1.401(a)(9)-2(b)(1) says the same thing in the same shape, with one clarification worth holding onto: the second limb is “the calendar year in which the employee retires from employment with the employer maintaining the plan.”

Read those two limbs as a race between two dates. Ordinarily the age limb wins, because most people stop working before required distributions start, and the year they retire is long past by the time the age arrives. Keep working past that age and the retirement limb wins instead, and the clock does not start until you stop.

That is the whole exception. Everything below is a limit on it.

A word on the age limb first, because it is not one number and a lot of published material still treats it as one. The applicable age is 73 for people born in 1951 through 1958, and 75 for people born in 1960 or later. Treas. Reg. §1.401(a)(9)-2(b)(2)(iv) gives age 73 for an employee “born on or after January 1, 1951, but before January 1, 1959,” and paragraph (vi) gives age 75 for an employee “born on or after January 1, 1960.” Birth year 1959 falls between them, in a paragraph the regulation prints as “[Reserved]” at §1.401(a)(9)-2(b)(2)(v). Two older cohorts sit above both: paragraph (ii) gives 70½ for an employee born before July 1, 1949, and paragraph (iii) gives 72 for one born between then and the end of 1950. Anyone in those rows is well past the question by now, but the table does not start at 73 and it is worth not saying that it does. If that is your birth year, the published regulation genuinely does not answer you, and no article can. I have written up when required distributions actually start at more length, and the mechanics of the division itself live in how an RMD is calculated.

Who counts as a five-percent owner?

Far more people than the phrase suggests, because your family’s ownership is counted as yours.

This is the exclusion that does the most damage, and it is where nearly every competing page stops after one sentence. 26 U.S.C. §401(a)(9)(C)(ii)(I) removes the retirement limb “in the case of an employee who is a 5-percent owner (as defined in section 416) with respect to the plan year ending in the calendar year in which the employee attains the applicable age.” Treas. Reg. §1.401(a)(9)-2(b)(3)(i) states the consequence directly: for such an employee, “the employee’s required beginning date is April 1 of the calendar year following the calendar year in which the employee attains the applicable age.” No retirement limb, no waiting, no exception.

So everything turns on what a five-percent owner is, and the statute does not say — it cross-references. Follow the reference and you land on 26 U.S.C. §416(i)(1)(B)(i), which defines the term as, for a corporation, “any person who owns (or is considered as owning within the meaning of section 318) more than 5 percent of the outstanding stock of the corporation or stock possessing more than 5 percent of the total combined voting power of all stock of the corporation.” For an employer that is not a corporation, the same clause reads on “any person who owns more than 5 percent of the capital or profits interest in the employer.”

Three things in that definition are worth slowing down for.

One thing before those three. §401(a)(9)(C)(ii)(I) opens “except as provided in section 409(d)” — a narrow carve-out about employer securities in a tax-credit employee stock ownership plan. It will not apply to most readers, but the removal of the retirement limb is not quite unconditional, and this page quotes the rest of that subclause, so the opening words belong here too.

“More than.” The line is not five percent. It is more than five percent, both in §416(i)(1)(B)(i) and in the parallel voting-power test. An interest of exactly five percent is not over the line. That is a thin margin to plan a retirement date around, but it is what the words say.

“Or is considered as owning within the meaning of section 318.” This is the phrase that catches people, and it is the reason the term reaches so far past the people who think of themselves as owners. 26 U.S.C. §318(a)(1)(A) says an individual “shall be considered as owning the stock owned, directly or indirectly, by or for” their spouse — excluding a spouse legally separated under a decree of divorce or separate maintenance — “and his children, grandchildren, and parents.” A legally adopted child counts as a child by blood under §318(a)(1)(B).

Read that list again, because the direction of it is asymmetric in a way that surprises people. Your children’s stock is attributed to you. Your grandchildren’s stock is attributed to you. Your parents’ stock is attributed to you. Siblings are not on the list. Neither, going the other way, are your children’s spouses.

There is more than family attribution in §318. Stock held by a partnership or an estate is attributed proportionately to the partners or beneficiaries under §318(a)(2)(A). Stock held by a trust is attributed to the beneficiaries in proportion to their actuarial interests under §318(a)(2)(B)(i), and stock held in a grantor trust is attributed to the grantor under §318(a)(2)(B)(ii). Stock you merely hold an option over is treated as owned under §318(a)(4). And a corporation’s stock is attributed down to a shareholder under §318(a)(2)(C) — a provision the top-heavy rules modify specifically for this purpose. 26 U.S.C. §416(i)(1)(B)(iii)(I) directs that §318(a)(2)(C) “shall be applied by substituting ‘5 percent’ for ‘50 percent’,” which is to say the corporate attribution rule is deliberately loosened here so it catches more, not less.

“If the employer is not a corporation.” Partnerships, LLCs and professional practices are not exempt from any of this. The capital-or-profits test in §416(i)(1)(B)(i)(II) reaches them directly, and §416(i)(1)(B)(iii)(II) provides that ownership in a non-corporate employer is determined under regulations “based on principles similar to the principles of section 318 (as modified by subclause (I)).”

One narrowing rule cuts the other way, and it is a genuine relief for anyone working inside a corporate group. 26 U.S.C. §416(i)(1)(C) provides that “the rules of subsections (b), (c), and (m) of section 414 shall not apply for purposes of determining ownership in the employer.” Those are the controlled-group and affiliated-service-group aggregation rules. Ownership is measured in the actual employer, not across a family of related companies.

The practical shape of all this is uncomfortable. A dentist with a small stake in the practice, a founder who transferred most of the company to a child years ago, a family business where the second generation runs it and the first generation still comes in three days a week — all of them can be five-percent owners for this purpose, and in the last two cases the person tested may own almost nothing in their own name. If there is any family ownership at all in the business you work for, this is a question to put to your plan administrator in writing rather than a conclusion to reach yourself.

When is that ownership actually tested?

Once, in a specific year, and not continuously.

Treas. Reg. §1.401(a)(9)-2(b)(3)(ii) is precise about it: “a 5-percent owner is an employee who is a 5-percent owner (as defined in section 416) with respect to the plan year ending in the calendar year in which the employee attains the applicable age.” The statute frames it identically at §401(a)(9)(C)(ii)(I).

That single sentence carries two consequences that almost nobody writes down.

Ownership before that year does not matter. Someone who founded the business, ran it for thirty years, and sold their entire interest well before reaching the applicable age is not a five-percent owner for this purpose, however the company was owned for the previous three decades.

And the test is looser inside that year than people assume. §416(i)(1)(A) defines a five-percent owner as an employee who is one “at any time during the plan year” — so a single day of qualifying ownership inside the tested year is enough, and a stake sold partway through it does not undo the year. The tested year itself is fixed by your birthday, not by your plans: it is the plan year ending in the calendar year you reach your applicable age — which is a year that arrives whether or not you were paying attention to it, and which for most plans ends on December 31. If a transfer of ownership is contemplated anyway, the year it lands in relative to that one is not a small detail.

There is one carve-out from the whole ownership test. Treas. Reg. §1.401(a)(9)-2(b)(3)(iii) provides that the five-percent-owner rule “does not apply in the case of a governmental plan (within the meaning of section 414(d)) or a church plan (within the meaning of § 1.401(a)(9)-6(g)(4)(i)),” and 26 U.S.C. §401(a)(9)(C)(iv) says the same at the statutory level for both the ownership exception and the actuarial-adjustment rule beside it.

Which of your accounts does the exception actually reach?

One. The plan at the employer you have not left.

It never reaches an IRA. 26 U.S.C. §401(a)(9)(C)(ii)(II) removes the retirement limb “for purposes of section 408(a)(6) or (b)(3)” — the required-distribution provisions for individual retirement accounts and annuities. The regulation implements that by pointing IRAs at the owner rule instead: Treas. Reg. §1.408-8(b)(1)(i) says “an IRA owner’s required beginning date is determined using the rules for employees who are 5-percent owners under §1.401(a)(9)-2(b)(3). Thus, the IRA owner’s required beginning date is April 1 of the calendar year following the calendar year in which the individual attains the applicable age.” There is no retirement limb in that sentence at all. Every IRA owner is treated the way an owner-employee is treated, which is to say the exception simply does not exist on that side of the wall.

A SEP or a SIMPLE IRA is an IRA for this purpose, even though both are employer arrangements and both feel like workplace plans. Treas. Reg. §1.408-8(a)(4) is explicit: IRAs receiving employer contributions “under a SEP arrangement (within the meaning of section 408(k)) or a SIMPLE IRA plan (within the meaning of section 408(p)) are treated as IRAs, rather than employer plans, for purposes of section 401(a)(9).” A self-employed person still working at the age their birth year points to, holding a SEP IRA, has no exception to use. That is worth saying plainly, because the population most likely to keep working past the applicable age overlaps heavily with the population whose retirement savings sit in a SEP.

It never reaches an old plan you left behind. The retirement limb in Treas. Reg. §1.401(a)(9)-2(b)(1)(ii) is the year you retire “from employment with the employer maintaining the plan.” A 401(k) sitting at a company you left is a plan whose employer you retired from at the time you left. Its clock started on the ordinary schedule. Working somewhere else now does nothing for it, and this is the single most common way somebody relying on the exception ends up short.

There is one useful gloss on that. Treas. Reg. §1.401(a)(9)-2(b)(5) provides that where a plan is maintained by more than one employer, an employee “who retires from employment with any of those employers but continues to be employed by another employer that maintains the plan is not treated as having retired for purposes of paragraph (b)(1)(ii) of this section.” Multiemployer and multiple-employer arrangements — common in the trades and in some professional associations — follow the plan, not the individual employer.

And a plan is allowed to switch it off. This is the limit that gets omitted most often. Treas. Reg. §1.401(a)(9)-2(b)(4) allows a plan to “provide that the required beginning date for purposes of section 401(a)(9) for all employees is April 1 of the calendar year following the calendar year described in paragraph (b)(1)(i) of this section, without regard to whether the employee is a 5-percent owner” — that is, a plan is free to start everybody on the age schedule and offer the exception to nobody. Plenty of plans do exactly that, for the ordinary administrative reason that one rule is cheaper to run than two.

So the answer for your own plan is not in the tax code. It is in the plan document, or in the summary plan description, or in a written answer from the administrator. Get it in writing before you rely on it, because the cost of being wrong is a missed distribution rather than an argument.

If it helps to have the numbers in one place, I keep them all in the Ultimate Retirement Guide.

What happens in the year you actually retire?

The clock starts, and the year you stop working is the first distribution year — not the year after.

The required beginning date runs from the later of the two years, so once you retire, that retirement year becomes the reference year. Your first required distribution is for the year you retire, and it may be deferred to April 1 of the following year on the same terms as anyone else’s first one.

Which sets up the trap that the whole grace period is famous for, in a sharper form than usual. Take that deferral and the first distribution lands in the same calendar year as your second one, because the second is due at year end regardless. Two required distributions, one tax return. I have written that mechanism up in full alongside the table and the deadlines, and it applies here with an extra edge: the year you retire often still contains most of a year’s salary. Stacking a first and second required distribution on top of a nearly full working year is a specific and avoidable way to have a very expensive twelve months.

Note also what retiring mid-year does not do. It does not prorate anything. There is no partial-year required distribution. The amount is the ordinary calculation — the balance as of the last valuation date in the preceding calendar year, over your denominator — whether you retired in January or in December. For an IRA that date is fixed at December 31 by Treas. Reg. §1.408-8(b)(2); for an employer plan it is whatever the plan’s own valuation date is, which is usually but not always the same day.

One honest gap: the regulation does not define what “retires” means. There is no hours test, no severance definition, no rule about consulting for the same employer afterwards. Treas. Reg. §1.401(a)(9)-2(b)(5) is the only gloss, and it addresses a narrow multiple-employer case. Someone winding down to two days a week, or leaving as an employee and returning as a contractor, is in territory the published text does not resolve, and the plan administrator’s reading of its own document is what will actually govern the reporting.

Can rolling an old plan into your current one bring it under the exception?

Possibly, and the mechanics of it are worth understanding before anyone treats it as a plan.

The idea is straightforward. The exception attaches to the plan of the employer you still work for. An old 401(k) at a former employer does not qualify — but if that balance were moved into your current employer’s plan, it would be sitting inside a plan that does. Treas. Reg. §1.401(a)(9)-7(b) says what happens to the balance: “the benefit of the employee under the receiving plan is increased by the amount rolled over for purposes of determining the required minimum distribution for the calendar year following the calendar year in which the amount rolled over was distributed.” Be careful about what that paragraph does and does not settle. It is a balance-computation rule. It says the money is counted in the receiving plan; it does not say, in terms, which required beginning date then governs it. That the rolled-in amount rides the receiving plan’s date follows from the required beginning date being determined plan by plan under §1.401(a)(9)-2(b)(1) — which is a sound reading and the one I would expect an administrator to take, but it is a reading rather than a sentence you can point at. On a move this consequential, that distinction is worth carrying into the conversation rather than leaving behind.

Three conditions have to hold before that is worth anything, and each is a question rather than an assumption.

Your current plan has to accept roll-ins at all. Many do; it is not universal, and a plan is under no obligation to.

Your current plan must not have adopted the uniform required beginning date. If it has, under §1.401(a)(9)-2(b)(4), moving money into it buys nothing — you have consolidated accounts, which may be worth doing anyway, but you have not deferred anything.

And you must not already be in a distribution year for the old plan. This is the timing constraint that ruins the move if it is left too late. Treas. Reg. §1.401(a)(9)-7(a) is unambiguous that “an amount that is a required minimum distribution under section 401(a)(9) is not eligible to be rolled over,” and that the distributed amount “is still taken into account by the distributing plan.” Once the old plan owes a required distribution for a year, that amount has to come out and be taxed; only the remainder can move.

None of that is advice to do it. It is a real planning move with real mechanics, and whether it suits a particular household depends on the investment menu in the current plan, the creditor protections that differ between a plan and an IRA, and whether deferring the income helps at all — which is the question the last section is about. What it is not is a decision to make from an article. Put both plan documents in front of the person who administers them.

What about a Roth account inside the plan?

A designated Roth account inside a workplace plan is already outside the calculation, so the exception has nothing to do for it.

Treas. Reg. §1.401(a)(9)-5(b)(3) provides that “for distribution calendar years up to and including the calendar year that includes the employee’s date of death, the account balance does not include amounts held in a designated Roth account (as described in section 402A(b)(2)).” That is a change from the older regime, and material published before it will tell you a Roth 401(k) carries a lifetime required distribution. It does not.

For a Roth IRA the answer has always been the same, and Treas. Reg. §1.408-8(b)(1)(ii) states it in one line: “No minimum distributions are required to be made from a Roth IRA while the owner is alive.”

So if you are working past your applicable age and your workplace savings are in the Roth side of the plan, none of the analysis above applies to that balance. It is the traditional balance — in the plan, in old plans, and in every IRA — that the required beginning date reaches.

What do people get wrong about this?

Four things, and the fourth is the one that costs the most.

They think it covers their IRAs. It never has. §401(a)(9)(C)(ii)(II) rules it out by name, and a SEP or SIMPLE IRA is an IRA for this purpose despite being an employer arrangement.

They think five-percent owner means what it sounds like. It means five percent counting your spouse’s, your children’s, your grandchildren’s and your parents’ interests as your own, plus attribution through partnerships, estates, trusts and options. Somebody who owns nothing in their own name can fail this test.

They think their plan cannot have taken it away. The regulation’s default is the later-of rule, but §1.401(a)(9)-2(b)(4) expressly permits a plan to put everybody on one uniform required beginning date instead — and a plan that has done so owes you a distribution on the age schedule no matter how long you keep working.

And they overestimate what deferral is worth. This is the argument I would most like you to leave with, and it runs against the way the exception is usually sold.

Deferring a required distribution does not remove the income. It moves it later, into a year where it will land on top of whatever else is happening then — and it leaves a larger balance behind, which produces a larger required distribution when the clock finally starts, against a denominator that is smaller because you are older. Meanwhile the years you spent deferring were years you were still earning a salary, which is to say years when your marginal rate was likely at or near its lifetime peak, and years when you were also not doing the kind of low-income conversion work that the window between retiring and required distributions is for.

Put those together and the exception is not a tax cut. It is a timing tool, and its value depends entirely on whether the year you are deferring out of is worse than the year you are deferring into. That is genuinely a matter of arithmetic on real figures — your salary, your other income, the size of the plan balance, the thresholds you are near — and it can come out either way. Sometimes it is clearly worth using. What it is not is free, and treating it as free is how somebody ends up with a larger balance, a shorter runway, and a first required distribution stacked on top of a final year of full pay.

If you want to see the shape of it on your own numbers, the RMD calculator will show you what the accounts that are already on the clock owe this year. The one that is waiting is the one to run twice: once as if you retired this year, and once as if you retired three years from now.