Who actually carries the billions in late 401(k) deposits
Aggregate numbers about late 401(k) deposits invite two opposite wrong conclusions. A billion-dollar total suggests a handful of giant offenders and lets everyone else relax. A count in the tens of thousands suggests everyone does it and it must not matter. The distribution settles which reading is true, so this page measures it, on public filings, both ways it can honestly be measured.
How many plans, each year
From the same EFAST2 pipeline that builds our cohort (run of 2026-08-24): among plans with 100 or more participants, the number answering “Yes” on Schedule H line 4a was 18,492 for plan year 2022, 16,744 for 2023, and 15,570 for 2024. The subset flagged in all three years, the chronic cohort this site studies, is 6,769 plans. So each year, roughly fifteen to eighteen thousand large plans report the failure, and about two in five of a given year’s reporters are carrying an item across all three years rather than a one-off.
Concentration, measured on two bases that are not the same number
The chronic cohort reports $8,471,929,961 across the three years (computed on 2026-08-25, method below). Who carries it depends on what you count as “who”, and the two honest answers differ:
| basis | population | top 10 share | top 1% share |
|---|---|---|---|
| per plan (one row per EIN and plan number) | 6,769 plans | 20.6% | 38.3% (68 plans) |
| per sponsor EIN (a sponsor’s plans summed) | 6,668 EINs | 21.5% | 39.7% (67 EINs) |
The bases differ because an EIN is not unique per plan: 86 sponsors in this cohort carry more than one flagged plan, so summing per EIN moves their combined weight up the ranking. Any concentration figure quoted from this dataset must say which basis it uses; these were computed on 2026-08-25 and the per-EIN figures are the ones to use when the question is about employers rather than plans.
Either way the answer is the same shape. Computed on 2026-08-25 over the same cohort file, the top ten sponsors alone carry $1,821,991,876, about a fifth of everything, and one percent of the population carries about two fifths. The billion-dollar headline is a short tail.
And the rest of the population is mid-size
Strip the tail and the typical chronic plan is not a giant. On the same cohort, computed on 2026-08-25: median three-year amount $220,442, with 5,423 plans (80.1%) under $1 million total; median 452 participants, 4,917 plans (72.6%) under 1,000 participants, 1,731 plans in the 100-249 band, and only 448 at 5,000 or more. Geographically it tracks employer population, not a regional quirk: 53 states and territories, led by California (1,012 plans), Texas (559), New York (539), Illinois (343) and Pennsylvania (264), with 10 rows carrying no US state in the sponsor address.
So both wrong conclusions fail. The dollars are concentrated, so the total is not evidence of a universal problem. The plans are not: thousands of mid-size employers, a median under a quarter million dollars, in every state. If you administer a plan in that band and want to know whether your own filings are among the 6,769, the lookup at planremit.com/check reads the same public data by EIN.
Method
Cohort: chronic-cohort.csv, plans of 100+ participants answering “Yes” on Schedule H line 4a in 2022, 2023 and 2024, built by pipeline.py from the public EFAST2 Form 5500 datasets (read 2026-08-21). Per-year populations are that pipeline’s per-year counts from its 2026-08-24 run. Concentration, medians and state counts: blog_stats.py, computed on 2026-08-25. Amounts are sponsor-reported line 4a values, summed with no adjustment.
What these figures do not say
They do not say the top sponsors owe the most tax: the section 4975 excise tax is computed on interest and timing invisible in these filings, not on the reported amount. They do not de-duplicate corporate families beyond the EIN: related employers filing under separate EINs count separately. And a sponsor-reported amount can itself be wrong in either direction; this page measures what was filed, which is the only thing public data can measure.
Published by Neige AI, Inc., last reviewed August 25, 2026. See the method and sources.
This is independent research, not legal or tax advice. It quotes primary sources with citations. Verify anything load-bearing against the primary source itself before acting on it.