How Fast Do Insider Trades Become Public? Form 4 Filing Lag, Measured

1. The rule: two business days, not two calendar days

Under 17 CFR 240.16a-3(g)(1), a Form 4 must be filed before the end of the second business day following the day on which the subject transaction was executed. The Form 4 General Instructions say the same thing in General Instruction 1(a), and the SEC's investor bulletin on Forms 3, 4, and 5 repeats it in plain language. Two consequences follow. First, weekends and federal holidays do not count: a Friday trade has until Tuesday, not Sunday. Second, the filing date is the date the public learns about the trade, while the transaction date is when the trade happened — and any price analysis that uses the filing date as the trade date is misdating the trade by exactly the lag measured below.

2. What we measured

The measurement uses this site's own parsed Form 4 corpus: 316,796 parsed filings dated 2025-01-02 to 2026-10-02. The window is the same 90-day window as the site's insider screens — filings dated 2026-07-31 to 2026-10-29, effectively through the corpus cutoff of 2026-10-02 — giving 25,512 filings and 59,214 transaction rows. Two cuts are reported: code P (purchase) and code S (sale) rows only, which are the rows that describe an insider's own buying or selling decision; and all transaction codes, for comparison.

Business-day lag is defined as the number of business days strictly after the transaction date through the filing date, so a filing on the same day as the trade scores 0, the next business day scores 1, and a filing on the second business day after the trade scores 2 — the deadline itself. Business days are Monday to Friday excluding U.S. federal holidays (OPM calendar, 2025–2026). Rows without a transaction date (43 rows, 0.07% of the window) are excluded from the lag calculation and reported here, not silently dropped from the counts.

3. Results: the deadline is the median

All figures in the table above are as of corpus cutoff 2026-10-02.

Three readings follow. First, the typical Form 4 is not early: the median is exactly 2 business days in both cuts, and the 90th percentile is also 2. Nearly half of P/S rows (11,779 of 24,191) land precisely on the deadline day. Second, same-day and next-day filings are common but not the norm: 9.5% of P/S rows are filed the same day and 37.6% the next business day. Third, the tail is long but thin: 1,006 P/S rows (4.16%) show a lag over 2 business days, and 449 of those (1.86% of P/S rows) stretch beyond 10 business days. The all-codes tail is longer (99th percentile 68 business days), which is expected: it includes compensation, gift, and administrative rows whose reporting path differs from a plain market purchase or sale.

By transaction code

Code meanings follow the Form 4 code guide. The table covers the 59,158 rows that have both a transaction code and a parseable transaction date. Code V (voluntarily reported early) had no rows in this window, so it has no row. Codes with fewer than 30 rows are pooled in the last row, as planned, instead of being shown one by one.

Three patterns stand out, and none of them is a violation finding. First, sales are the fastest large code: 97.35% of code S rows are filed within 2 business days, against 88.98% of code P rows. Purchases look slower on average (mean 10.13 business days, against 2.23 for sales) but faster at the median (1 business day, against 2 for sales). That split is the signature of a long tail: a small number of purchase rows carry transaction dates years before their filing date, and those rows pull the mean up while leaving the median almost untouched. The median, not the mean, is the right summary for what a typical purchase row looks like. Second, code J ("other") is the slowest large code, at 20.28% over 2 business days with a mean of 12.50. That is expected for a residual category: "other" collects transactions that do not fit the standard codes, so its reporting path is the least uniform. Third, code L looks extreme — median 5, mean 34.51, 73.47% over 2 business days — but it has only 49 rows, so a handful of very old transactions move every figure. Small-sample codes are pooled in the last row for the same reason: a percentage computed on 1 or 11 rows says almost nothing, and publishing it on its own would mislead more than inform.

Two codes sit at the fast end for a different reason. Codes D and E both have a median of 0, meaning at least half of their rows were filed the same business day as the transaction. Dispositions back to the issuer and expirations of short derivative positions are typically arranged by the issuer or by the terms of the instrument, so the filing can be prepared before the transaction date arrives. That is a workflow difference, not a compliance signal.

By reporter type

Each row is assigned to exactly one reporter bucket, in this priority order: 10% owner first, then officer, then director, then other. A reporter who is both an officer and a director therefore counts as an officer, and a 10% owner who is also an officer or director counts as a 10% owner. The buckets use the filing's own flags (isTenPercent, isOfficer, isDirector); "other" means none of those three flags is set. This table covers the same 59,158 coded, scored rows as the code table, so the two tables reconcile row for row.

The reporter split is narrower than the code split. Officers, the largest group at 35,133 rows, file within 2 business days 94.69% of the time; 10% owners are close at 94.41%. Directors trail both, at 91.53% within 2 business days and a mean of 7.20, but their median is still 2 — again, a long tail rather than a typical delay. The "other" bucket is the outlier at 75.00% within 2 business days, but it holds only 536 rows and, by construction, the filings whose reporter does not fit the three standard flags, so it should not be read as a group of people behaving differently. The practical takeaway for tracking is modest: reporter type changes the tail of the lag distribution more than its centre, and the centre is 2 business days for every group.

Recomputation note: what changed in the headline table, and why

This expansion recomputed every figure from the same parsed corpus and window as the original table (316,796 parsed filings; filings dated 2026-07-31 to 2026-10-29, effectively through the corpus cutoff of 2026-10-02). The recomputation reproduces the original table almost exactly: the P/S and all-codes medians, 90th and 99th percentiles, within-2 and over-2 counts, and the P/S mean all match. Two values were corrected in this expansion. The all-codes mean is 4.94 business days, the mean of the 59,165 scored all-codes rows, so 4.94 is used here. The all-codes within-2 share is 93.64% (55,404 of 59,171), a rounding correction. Row accounting for the headline table: the all-codes count of 59,171 is 59,165 scored rows plus 6 rows filed before their transaction date, which are counted in the total but in neither the within-2 nor the over-2 bucket; the P/S count of 24,191 is 24,190 scored rows plus 1 such negative-lag row. The two breakdown tables above cover 59,158 rows: the 59,165 scored all-codes rows minus 7 rows that carry a transaction date but no transaction code. A further 43 rows in the window carry a transaction date written with a timezone suffix (for example "2026-09-30-05:00") that does not parse as a plain date; they are excluded from all lag figures here and are listed in the recomputation file published with this article. Method and definitions are on the Methodology page.

4. What "over 2 business days" does — and does not — mean

Rows with a lag over 2 business days are described here only as "rows that appear late." That wording is deliberate, because a raw lag over 2 does not prove a late filing. For trades under a Rule 10b5-1 plan or a discretionary transaction where the reporting person does not select the execution date, the rule deems the notification date — not the trade date — to be the execution date, with a cap at the third business day after the trade date (17 CFR 240.16a-3(g)(2)–(4)). A plan trade can therefore look late on raw dates while being timely under the rule. Separately, the corpus does not distinguish original filings from amendments, so a long lag can reflect an amended filing rather than a late original. Six rows in the all-codes cut (one in the P/S cut) even show a negative lag — a filing dated before the transaction date — which is a data quirk, not a trading pattern, and is small enough (0.01%) not to move any headline number.

5. Why this matters for tracking insider buying

A tracker built on Form 4 data is always looking at trades that happened days ago, and the median trade in this window became public a full two business days after it happened. Aggregating individual Form 4s into a market-wide "insider buying" indicator adds a second problem: most Form 4 rows are not market buys at all. In the same 90-day window, compensation rows (codes A, M, F) outnumbered market rows (P + S), and sales outnumbered purchases by more than four to one. A market-wide insider indicator that counts every Form 4 row is mostly counting pay and routine selling, not conviction buying. The screens on this site avoid both traps by using transaction dates for price context, filing dates for "when the public learned it," and code P rows only for buying.

See also: Form 4 Transaction Codes Explained · How the STOCK Act 45-Day Disclosure Works · Insider cluster buying · Methodology.

Frequently asked questions

Can insider trading be tracked in real time? Not from Form 4. The form is due by the end of the second business day after execution, and the median filing in this window landed exactly on that deadline. Anything labelled "real-time insider trading" is tracking something other than completed Form 4 trades.

How long after a trade does a Form 4 usually appear? Two business days is the typical case: the median and the 90th percentile are both 2 business days in this window, for P/S rows and for all codes. Same-day filings are the exception (9.5% of P/S rows), not the rule.

Does a filing more than two business days after the trade mean the insider filed late? Not necessarily. For plan trades, the rule can deem the notification date to be the execution date, and the corpus does not separate amendments from originals. Such rows are described here only as appearing late.

Why is the 99th percentile so much longer than the median? Because the tail mixes different reporting paths: amended filings, plan trades with deemed execution dates, and (in the all-codes cut) compensation and administrative rows. The median describes the ordinary filing; the 99th percentile describes the exceptions.

Why do some codes show a much higher average lag than others? Because the mean is pulled by a long tail while the median is not. Code P purchases average 10.13 business days but have a median of 1; a small number of purchase rows carry transaction dates years before filing. Code L averages 34.51 on only 49 rows. Compare medians and within-2 shares first, and treat any mean computed on a small or heterogeneous group — code J ("other") and code L in particular — as a description of the tail, not of a typical filing. None of these figures, on their own, shows a late filing: the next section explains why.

Sources

• SEC Office of Investor Education and Advocacy, Updated Investor Bulletin: Insider Transactions and Forms 3, 4, and 5 (Form 4 within two business days). https://www.investor.gov/news-alerts/investor-bulletins/investor-bulletin-insider-transactions-forms-3-4-5 (checked 2026-10-29)
• 17 CFR 240.16a-3(g)(1) (Form 4 before the end of the second business day after execution) and (g)(2)–(4) (deemed execution dates for plan and discretionary transactions). https://www.ecfr.gov/current/title-17/chapter-II/part-240/subject-group-ECFR451036c2752e741/section-240.16a-3 (checked 2026-10-29)
• SEC, Form 4 General Instructions, General Instruction 1(a) (second-business-day deadline). https://www.sec.gov/files/form4.pdf (checked 2026-10-29; form confirmed via the 2026-10-29 supplement check of the 13:10 reply)
• GetCoattail parsed Form 4 corpus, as of corpus cutoff 2026-10-02 (316,796 filings; 90-day window filings dated 2026-07-31 to 2026-10-29, effectively through 2026-10-02; aggregation script and result JSON in the bundle).

Last updated: 2026-11-01. This is general information about a disclosure form, not tax, investment, or legal advice.

For education only — not investment advice. How every number on this site is measured: Methodology.