I apologize in advance if I've missed something in my cursory research of this, but here goes. Also, for simplicity I am assuming the employer in question only has DC plans, but I don't think that makes a difference.
Clearly, to determine whether a plan is top-heavy for a "plan year" you use account balances as of the "determination date," which also quite clearly is the last day of the plan year preceding the plan year for which the determination of top-heaviness, or not, is being made. E.g., to determine whether a plan is top-heavy for a calendar year 2019 plan year, you use balances as of 12/31/2018. There are potentially also certain addbacks to determination date account balances of keys and nonkeys for distributions that were made during the plan year preceding the year for which the top-heavy determination is made (i.e., in my example, during 2018 for the 2019 top-heaviness
determination), or potentially during a 5-year period in the case of some distributions. See IRC secs. 416(g)(1), (3), and (4)(C). And of course you have aggregation rules.
So the above gives you your key and non-key participant balances for purposes of determining whether more than 60% belong to keys. But then you have to figure out whether those balances belong to keys or non-keys. IRC sec. 416(I)(1) [Note: the "I" in 416(I)(1) should be lower-case, but I can't make that happen; sorry] tells you that the keys are the folks who meet certain requirements, e.g. percentage of ownership of the employer, "at any time during the plan year." Just looking at 416(I)(1) [see previous note], it would seem that the "plan year" being referred to for identifying keys is the current plan year, i.e. the year for which you are making the top-heavy determination, i.e. 2019 in my example. At least,
that's what I think, because 416(g)(1) tells you that a plan is top-heavy "with respect to any plan year," and to me, when 416(I)(1) says "during the plan year," they are talking about the same plan year. So it seems to me that based on the statutory language you would use ownership and compensation in 2019 to determine who are your keys and non-keys, and then you would go back to 12/31/2018 to see what those folks' determined to be keys based on their 2019 facts had in the plan for purposes of the "more than 60%" test.
But Treas. reg. 1.416-1, T-12 seems to say pretty clearly that the "plan year" being referred to in 416(I)(1) is not the plan year for which you are making the determination (i.e., 2019 in my example), but rather 2018. It does this by adding "containing the determination date" to 416(I)(1)'s simpler "plan year." I guess when, before EGTRRA, you looked back 5 years to
determine who were your keys (i.e., the keys were participants who at any time during "the plan year or any of the 4 preceding plan years" met one of the status tests), and under pre-EGTRRA Section 416(g) you were also dragging back in all distributions made during the 5 plan years ending on the determination date, it may have made sense for the IRS to want the same 5 year period for both purposes, i.e., the IRS may have been trying to simplify. But if that was the reason for departing from what otherwise seems the very plain statutory language of 416(I)(1), it no longer seems valid, since using balances as of the end of the previous year, and status as of any time in the current year, seems just as easy to do now.
Has anyone else had an issue with this? To anyone's knowledge, has the IRS ever commented on this, formally or informally?