The 80–120 rule is an election, not a threshold

Here is the conversation, more or less verbatim, that happens every spring somewhere between a plan sponsor, a TPA and an audit partner.

“We are at 85 participants with account balances. We filed as a large plan last year. So we are stuck carrying large and we need another audit, right?”

Not from participant count alone. At 85 beginning-of-year participants with account balances, the administrator is not compelled to file another large-plan return merely because the prior return was large. The election in 29 CFR 2520.103-1(d) is permissive. But a small-plan filing category does not itself eliminate the audit: a small pension plan that cannot satisfy a separate exception or the conditions of 29 CFR 2520.104-46 must still attach an IQPA report.

Two rules, stacked

Start with what the 2023 change did and did not touch. For plan years beginning on or after 1 January 2023, a defined-contribution plan counts participants with account balances at the beginning of the plan year. That changed the input. It did not change the threshold the input is measured against, and it did not change the structure of the filing-category rules.

What you have, in order, is:

The default. 100 or more participants with account balances means large-plan filing. Fewer than 100 means small-plan filing. This is the ordinary rule and it applies across the whole range.
The election. Where the count is between 80 and 120 inclusive, the administrator may elect to complete the return in the same filing category as the prior year’s return.

The operative word in the second rule is may. It confers an option on the administrator. It does not remove the default that would otherwise apply, and it does not bind the plan to whatever it did last year.

The shape of the rule

Outside 80–120, there is exactly one available answer and no choice. Inside 80–120, there are at most two available answers: the one the 100-participant rule produces, and the one the prior filing produces. Where those coincide, the election is a no-op.

So run the four cells

Inside the band, only the count and the prior category matter. There are four combinations, and only two of them are genuinely elective.

Participants with balances at beginning of year	Prior return	Default under the 100 rule	Actually available
80–99	Small	Small	Small only. Election is a no-op.
80–99	Large	Small	Small or large. Genuinely elective.
100–120	Large	Large	Large only. Election is a no-op.
100–120	Small	Large	Large or small. Genuinely elective.

The second row is the one from the conversation at the top, and it is the row that gets answered wrongly most often. A plan at 85 beginning-of-year participants with account balances that filed large last year is not compelled to carry large. Because 85 is below 100, the default is small-plan filing. The prior large filing gives the administrator the option to elect large. Declining that option is the default rule applying.

The fourth row is its mirror and it is the one that surprises people in the other direction: a plan at 108 balances that filed small last year may elect to stay small, even though the count is over 100.

Notice also what the first and third rows mean in practice. When the prior filing already matches the default, nothing about the band is doing any work. Plenty of the anxiety about the 80–120 range is spent on plans where there was never a decision to make.

Why this is the row that matters right now

The 80–99-with-a-prior-large-filing cell is not a corner case. It is the exact population the 2023 counting change created. Consider what happened to a plan with 60 funded accounts and 55 eligible employees who had never contributed:

Under the old method it counted 115 participants and filed large.
Under the current method it counts 60 and files small.

Plans that landed further up the distribution — say 130 eligible, 85 funded — came out of the transition sitting inside the band with a large filing history behind them. That is a lot of plans, all of them carrying an audit relationship, all of them arriving at the question at the same time. It is precisely why the question feels ambient right now and did not five years ago.

Getting the count right is harder than getting the rule right

Everything above assumes the count going in is correct, and in practice that is where more answers go wrong than in the rule itself. Two errors dominate.

Using the wrong population

“Participants with account balances” is not active headcount, not eligible headcount, and not the number the payroll system reports. An eligible employee who has never deferred and holds no balance is not counted. This is the whole point of the change, and it is still routinely undone by pulling the wrong column off a census.

Dropping terminated participants who still hold a balance

This one runs the other way and understates the count. A terminated employee whose balance has not been distributed is a participant with an account balance. The test is the balance, not employment. Plans with high turnover and small balances that sit undistributed for years can carry a surprising number of these, and leaving them out is the fastest way to talk yourself under a threshold you are not actually under.

And the date

The measurement is at the beginning of the plan year for a continuing plan. A first-year plan is different: there is no prior return to carry, so no election exists, and the plan uses its end-of-year count against the 100-participant threshold.

Small-plan filing is not the finish line

Landing in the small-plan category answers the filing-category question and only that question. It makes the audit waiver at 29 CFR 2520.104-46 available; it does not apply it. The waiver is conditional and every condition has to hold:

At least 95% of plan assets were qualifying plan assets at the end of the preceding plan year — or every person handling the non-qualifying assets is bonded for at least their full value, which is an enhanced bond on top of the ordinary ERISA section 412 bond, not the usual 10% one.
The Summary Annual Report carries the specific disclosures: institutions holding qualifying assets and amounts, the surety where enhanced bonding is used, and notice that participants may examine or obtain the supporting evidence without charge.
Those statements and the bond evidence are actually furnished, without charge, on request.

Both halves of the first condition get missed. The threshold is 95%, not a comfortable majority. And the measurement date is the prior year end, which means a plan that cleaned up its non-qualifying holdings during the year has not fixed the test for the year it is filing. If a limited-partnership interest is a non-qualifying plan asset, a small plan holding 12% in that interest will not satisfy the 95% route; using the alternative requires bonding the person handling the non-qualifying asset for at least its full value, not merely the ordinary section 412 amount.

The practical version

If you want one sentence to carry into the next conversation:

Inside 80 to 120 the 100-participant rule still decides the default, and the prior filing only ever adds a second option on top of it.

Which means the useful questions, in order, are: what is the beginning-of-year count of participants with account balances, what did the prior return actually file as, do those two point at the same category — and if the answer is small, do all three waiver conditions hold?

The first two are facts about a filing. Both of them are public.

A correction we made to our own tool

An earlier version of the screen on this site returned “IQPA audit generally required” for the 80–99-with-prior-large case, the exact row this post is about. It treated the election as binding rather than permissive. That is fixed, and the fix is in the changelog. We would rather say so than quietly correct it.

The 80 to 120 band is an election, and most of the argument is about the wrong rule