When insider filings get corrected

Study #2 · originals 2026-07-01 → 2026-07-31, amendments observed through 2026-08-24 · every number below regenerates from one script on a free key

A Form 4 is not immutable. When a filing is wrong — a price, a share count, a missed transaction — the insider re-files it as a Form 4/A, and the amendment carries amendment_date: the date of the original submission it corrects, as filed. That makes correction behavior measurable, and it matters for one specific reason: any dataset snapshotted at time T still holds the superseded version of every filing whose correction had not yet arrived at T. This page measures how big that window is.

Method. All 21,479 Form 4 and 4/A filings with filed_date from 2026-07-01 through 2026-08-24 (55 days), fetched from GET /v1/insider-trades one filed-date at a time with since=/until= (a fixed window returns the same population forever). Correction lag per 4/A = filed_date minus amendment_date. Amendments with a missing or malformed original date (0) or an original date after their own filing date (0) are counted here, not silently dropped. The cohort view then asks: of the 11,052 originals filed in July 2026, how many had a 4/A naming their submission date by 2026-08-24?

342

amendments analyzed

1.6%

of filing traffic is 4/A

19d

median correction lag

≥0.9%

of July originals amended

How much filing traffic is corrections

Over the full 55-day fetch range: 21,137 original Form 4s and 342 Form 4/A amendments — 1.6% of everything filed. Inside July 2026 itself (the same fixed window as study #1): 11,241 filings, of which 11,052 originals and 189 amendments.

Correction lag: how old is the filing being corrected?

Days between the original submission date a 4/A names and the day the 4/A itself reached EDGAR. This is exactly how long the superseded version was the only version anyone had.

lag	amendments	share
same day	16	4.7%
1 day	32	9.4%
2-3 days	29	8.5%
4-7 days	46	13.5%
8-14 days	34	9.9%
15-30 days	34	9.9%
31-90 days	73	21.3%
91-365 days	66	19.3%
over a year	12	3.5%

Percentiles (nearest-rank): p50 = 19d · p75 = 81d · p90 = 148d · p95 = 307d · p99 = 728d · max = 1,649d. The oldest correction in this window reached back to a submission from 2021-12-31.

The July cohort: a lower bound on the amendment rate

Of the 11,052 original Form 4s filed in July 2026, amendments observed through 2026-08-24 named the submission dates of an estimated 95 distinct originals — 0.9%. That is a floor, not the final number: the last July original had only 24 days of observation, and the lag distribution above shows corrections arriving months later. 97 amendment filings named a July submission date in total (an original can be amended more than once); their median lag was 7d.

Of those July-targeting amendments: 33 (34.0%) within 2 days · 52 (53.6%) within 7 days · 82 (84.5%) within 30 days of the original submission.

Is the population complete?

An amendment study over an incomplete population is quietly wrong, so the script cross-checks every day's filing count against SEC EDGAR's own daily master index (distinct accessions, form types 4 and 4/A, counting only rows whose Date Filed matches the index day — EDGAR re-disseminates corrected filings under their original filed date). Result: 38 of 38 index days match exactly; 17 days had no published index (weekends/holidays) and zero filings in the API — consistent.

What this means for a data pipeline

Amendments are first-class events. A 4/A arrives as a new filing with its own accession number; it does not overwrite the original in EDGAR, and it should not silently overwrite it in your store either. Key on accession, keep both, and let form_type + amendment_date mark the supersession.
Snapshots age. The lag distribution above is the correction traffic your pipeline will see after any point-in-time export. A dataset frozen at month-end still holds the pre-correction version of whatever gets amended later.
Re-analysis needs the amendment, not just the original. The API serves 4/As in the same stream and shape as originals (form_type: "4/A", amendment_date set), and webhook subscriptions deliver them as they arrive.

Reproduce this

The whole study is one standard-library Python script that runs on the free tier (the default ~8-week range is ~160 requests; at the free tier's 10 req/min expect about twenty minutes). The --verify-coverage flag hits EDGAR directly and needs a contact identity in the User-Agent, per SEC fair-access policy — set your own.

# free key, no card: https://filingpulse.io/signup.html
export FILINGPULSE_API_KEY=fp_...
export EDGAR_CONTACT=you@example.com

curl -O https://filingpulse.io/examples/form4_amendments_study.py
python form4_amendments_study.py \
    --since 2026-07-01 --until 2026-07-31 \
    --observed-through 2026-08-24 \
    --out results.json --verify-coverage

The committed results this page renders from: form4_amendments_2026-07.json. All three date bounds are fixed, so the population never changes under you — that is what makes the study reproducible rather than a screenshot.

Limitations

The cohort rate is a lower bound by construction: corrections keep arriving after any observation horizon, and the last day of the cohort had only 24 days of observation.
amendment_date names a date, not an accession, so "distinct originals amended" is an estimate keyed on issuer CIK + reporting owners + original date — two same-day originals by the same insider at the same issuer would collapse into one.
A 4/A can itself be amended; EDGAR's field does not say which prior submission (original or earlier amendment) is being corrected, so chains are counted as separate amendments of the named date.
One window is one window. Correction behavior varies with filing waves; treat these numbers as a measurement of this range, not a law of nature. This is an observed re-filing record, not a compliance determination and not a list of violations.
Dates are strings exactly as filed; malformed ones are excluded and counted in the method box.

Sources

SEC Form 4 — the ownership form; amendments are re-filings of it as 4/A
EDGAR daily index — the per-day ground truth the coverage check reads
Study #1: How late is Form 4 data? — the filing-delay side of the same point-in-time question
FilingPulse API docs · query patterns · free screener