Does a Form 144 Filing Mean the Stock Will Drop? Evidence From 75,654 Insider Sale Notices

Summary

No. Across 75,654 Form 144 notices filed from April 2023 to October 2025, the typical stock did not fall after an insider gave notice of a proposed sale. One year after the filing became public, the median stock was up 5.07% and had beaten the median listed stock by 2.23 percentage points, with 52.41% of events beating it.

The same stocks trailed the S&P 500 by 14.03 points over that year, which looks like a sell signal until you measure the same stocks one year later, away from the filing. They trailed the S&P 500 by 13.52 points then, against 12.29 points in the real windows on the same events. The paired difference is 1.16 points with a Newey-West t-statistic of 0.50. The gap belongs to the kind of company whose insiders file Form 144s, not to the filing.

The pattern that does hold is what comes before the filing. In the three months up to entry, the median stock gained 8.44% and beat the median listed stock by 8.16 points, and 67.32% of events beat it. Insiders file after a rally, not before a fall.

Two secondary results. Notices without a disclosed Rule 10b5-1 plan were followed by weaker returns than plan notices filed on the same day: 1.65 points weaker over three months, with a t-statistic of 5.83. They still did not trail the median listed stock. Large notices, measured against shares outstanding or in dollars, also failed to predict a drop once the same companies were measured away from the filing.

Abstract

We measure closing-price returns after 55,992 Form 144 filing events, built from 75,654 notices of proposed sale filed on EDGAR between 13 April 2023, the first day electronic filing was mandatory, and 16 October 2025. The panel covers 3,146 companies and 626 filing dates. An event is one company and one filing date. Entry is the first market close strictly after the filing date. Returns are measured over 21, 63, 126 and 252 S&P 500 market sessions and compared with two benchmarks over identical dates: the S&P 500 price index and the median listed US stock.

At one year, the median event returned 5.07%, against 18.37% for the S&P 500 and 2.56% for the median listed stock. Relative to the S&P 500 the median event return is -14.03% and 36.44% of events beat the index; relative to the median listed stock it is +2.23% and 52.41% beat it. Date-clustered Newey-West t-statistics are -28.70 and 4.84.

A shifted-window placebo measures the same events 252 sessions later. Against the S&P 500 the paired one-year difference is 1.16 points with a t-statistic of 0.50; at one, three and six months it is -0.14, -0.01 and -0.21 points. Against the median listed stock the real window is better than the shifted one by 3.99 points at one year, with a t-statistic of 4.75. Neither benchmark shows weaker returns after a filing than the same stock showed a year later. In the 63 sessions before entry, the median event beat the median listed stock by 8.16 points, with a t-statistic of 24.09.

What is a Form 144?

A Form 144 is a notice of proposed sale. Under Rule 144 of the Securities Act of 1933, an affiliate of a company, which in practice means an officer, a director, a holder of 10% or more, or someone controlling or controlled by them, must file one when they plan to sell more than 5,000 shares or more than $50,000 of the company's stock under the rule in any three-month period. The notice is filed at the same time the seller places the order with a broker or sells directly to a market maker.

Three features of the form matter for reading it:

• It announces a plan, not a completed sale. The notice gives the number of shares to be sold, their aggregate market value, the broker and an approximate sale date. The sale itself is later reported on Form 4 for officers, directors and 10% holders.
• The seller signs a representation. By signing, the seller represents that they do not know any material adverse information about the company that has not been publicly disclosed.
• It records any Rule 10b5-1 plan. A seller relying on a written trading plan or trading instruction under Rule 10b5-1 enters the date the plan was adopted. Since the SEC's 2022 amendments, officers and directors must wait a cooling-off period of at least 90 days after adopting a plan before it can trade.

Forms 144 for companies that report to the SEC must be filed electronically on EDGAR since 13 April 2023. Before that date most notices were paper filings, so earlier electronic records are a voluntary minority. Our panel starts on the mandate date for that reason.

The question we ask is simple and is the one investors type into search engines: when a Form 144 appears, should you expect the stock to drop?

Data and sample construction

From notices to events

Our source is the Form 144 notices in our EDGAR archive filed from 13 April 2023 to 20 October 2026: 114,979 notices covering 3,748 companies. Each notice is linked to its issuer, and each issuer to its primary US listing. Roughly 33,000 to 35,000 notices are filed each year, with peaks in February and March, May and June, August, and November and December, the months after earnings when company trading windows open.

The event is one company and one filing date. Several notices on one day, from one seller or several, become one event. That gives 84,265 events. A small number of companies file very often: Ginkgo Bioworks had 726 notices, Meta Platforms 584, Dell Technologies 508, CoreWeave 468, Airbnb 423 and Datadog 421. The ten most frequent filers account for 4.0% of notices, and the median company has 11 events, so no single company drives the results.

Attrition

The panel holds 75,654 notices from 13 April 2023 to 16 October 2025. The median event proposed a sale worth $469,195.

Prices, benchmarks and statistics

We use raw closes on the company's primary US listing and require positive volume at entry and exit. Any one-year window that crosses a recorded split is excluded, and the same panel is used at every horizon. Dividends are excluded from both the stock and the benchmarks, so every comparison is between price returns.

Entry is the first S&P 500 session strictly after the filing date, so the filing is public before the position opens. Horizons are 21, 63, 126 and 252 sessions on the S&P 500 calendar, reported as one month, three months, six months and one year.

We use two benchmarks over identical dates:

• The S&P 500 price index. It answers the practical question of whether the stock did better than an index fund.
• The median listed stock. For every window we take the median buy-and-hold return across the primary US listings of every company in our coverage that traded at both ends and had no recorded split in the window, between 5,082 and 6,816 stocks depending on the date. It answers whether the stock did better than a typical company.

The two benchmarks diverged sharply in this period. Over the panel's one-year windows the S&P 500 returned 18.37% and the median listed stock 2.56%, because a few very large companies led the index. Any result reported against only one of them should be read with that in mind.

"Relative return" means the stock's return minus the benchmark's return over the same dates. It is not a factor-model alpha. The pooled median is the primary statistic. For inference we collapse each group to one median per filing date and compute a Newey-West t-statistic with a lag equal to the fourth root of the number of dates, the same procedure as our study of congressional trades.

The placebo moves each event's entry forward by 252 sessions and repeats the measurement with the same horizons, benchmarks and exclusions. Both the real and the shifted window must be complete, traded and free of splits, which limits the one-year placebo to events filed by 11 October 2024.

What happens to a stock after a Form 144 filing?

Stocks rose after Form 144 filings in absolute terms, did slightly better than the typical listed company, and did much worse than the S&P 500. All three statements are true over the same windows. Which one sounds like "the stock drops after a Form 144" depends only on the benchmark.

Is the gap caused by the filing or by the company?

A large S&P 500 gap with a t-statistic of -28.70 looks like strong evidence. It is strong evidence about these companies. It is not evidence about the filing until the same companies are measured at a time when no filing is at stake.

Against the S&P 500 the two windows are almost identical at every horizon. The companies whose insiders file Form 144s trailed the index whether or not a filing had just happened. The median event was a company worth $4.80 billion at entry, and in 2023 to 2026 the index was led by a handful of companies hundreds of times that size.

Against the median listed stock the one-year window after the filing was better than the year after it, by 3.99 points. That is the opposite of a sell signal. It fits the run-up described next: stocks that have just risen tend to keep a little of that strength for a while.

Do insiders sell after the stock has gone up?

Yes, and this is the clearest pattern in the data. We measured the 63 sessions up to entry for the 81,808 events with complete prices over that window.

Two in three filings came after the stock had beaten the typical listed company over the previous three months. The pattern is almost the same with and without a disclosed 10b5-1 plan, at 8.21 and 7.97 points against the median listed stock. This window includes the filing day itself, so part of the gain may come from the filing date.

There are ordinary reasons for this. Company trading windows open after earnings, restricted stock vests on a schedule, and 10b5-1 plans often sell at preset prices. An insider who sells more after a rise is taking money off the table, which tells you about their finances and their plan, not about the next quarter.

Are sales without a 10b5-1 plan more bearish?

Slightly weaker, not bearish. We split events into those where every notice gave a 10b5-1 plan adoption date (25,074 events) and those where none did (29,774). We excluded 1,144 mixed days. A missing plan date means no plan was disclosed on the notice, not that the sale was proven discretionary.

On the 625 filing dates that carried both kinds of notice, the median gap between plan and no-plan events was 0.81 points at one month, 1.65 at three months, 1.87 at six months and 2.23 at one year, with t-statistics of 4.35, 5.83, 3.89 and 2.23. Because both groups share the same entry and exit dates, this comparison does not depend on the benchmark.

The gap also survives the placebo. Against the S&P 500, plan events beat their own shifted windows by 2.64 points at one year (t of 2.33), while events without a plan matched theirs (0.39 points, t of -0.81). Netting each group's shifted window, plan events led by 1.60 points at three months (t of 3.20) and 3.22 points at one year (t of 2.85).

So the useful reading is narrow. A notice without a plan date is followed by somewhat weaker returns than a planned sale on the same day. It is still not followed by a drop against the typical listed stock.

Do bigger sales predict worse returns?

Not once the companies are measured away from the filing. We measured size two ways: the shares in the day's notices as a share of the shares outstanding stated on the form, and the aggregate market value of the proposed sale.

The two measures point in opposite directions, which is the first sign that they describe companies rather than intent. Sales that are large against shares outstanding happen at small companies: the median company value at entry was $0.75 billion for the 1% or more group, against $11.91 billion for sales under 0.01%. Sales that are large in dollars happen at large companies after large gains: the $10 million or more group had a median company value of $26.56 billion and had beaten the median listed stock by 13.41 points in the three months before entry, against 5.95 points for sales under $1 million. In no row was the year after the filing reliably worse than the same stock a year later.

For the 1% or more group the median and the date-level t-statistic disagree in sign, so we treat that row as uninformative rather than as evidence of a drop.

Does it matter who is selling?

Little. We grouped events by the relationship written on the notices: any officer on the day (39,170 events), directors with no officer (10,594), and 10% holders or affiliates (5,504).

Officers and directors look alike. The 10% holder and affiliate group is the noisiest: its full-panel one-year return is slightly below the median stock, but in the earlier events that have a placebo, its real window beat the shifted window by 12.96 points. Its stocks tended to weaken later, not straight after the sale.

Several insiders filing on the same day did not change the answer. Days with three or more sellers (3,678 events) returned 0.31 points more than the median listed stock over a year, against 2.25 points for days with one seller, and the placebo difference was not significant.

Does the result hold every year?

The absence of a drop holds in every year. The small lead over the median listed stock does not.

Filings in 2025 were followed by returns indistinguishable from the median listed stock. Against the S&P 500, every year shows a similar gap, which is the composition effect the placebo identifies.

Limitations

Notices, not trades. A Form 144 states an intention to sell. Some proposed sales are smaller than stated or never happen, and the notice can follow the order by hours. We measure from the first close after the filing date, which is the earliest an outside investor could act on the notice.

Short period. Electronic filing became mandatory in April 2023, so the panel covers two and a half years of filings and the one-year placebo covers eighteen months. It is one market regime, led by a few very large companies. We would not assume it generalizes to other periods.

Benchmarks disagree on size. The S&P 500 and the median listed stock differed by about 16 points over the average one-year window. We report both throughout because neither is the obvious comparison for a group that mixes small growth companies with the largest companies in the market.

Stale quotes. Some listed stocks do not trade enough to move over a month, which pins the median listed stock at exactly zero on several dates. This affects the benchmark level, not the paired placebo or the same-date plan comparison.

Fields as filed. Shares outstanding, relationship and plan adoption date come from the notices as the sellers filed them. We excluded 60 events whose shares outstanding were missing or smaller than the shares to be sold from the size cut and did not correct any other values.

Unrecorded corporate actions. We excluded windows crossing a recorded split. A split we have not recorded would distort a few returns, which is one reason medians rather than means are the primary statistic.

Reproducible specification

Source: Form 144 notices on EDGAR filed from 13 April 2023, linked to the issuer's primary US listing. Price data through 20 October 2026.

Event: one issuer and one filing date; all notices that day are combined.

Entry: the first S&P 500 session strictly after the filing date, at the raw close, with positive volume.

Horizons: 21, 63, 126 and 252 S&P 500 sessions; run-up window from 63 sessions before entry to entry.

Exclusions: incomplete one-year window, missing traded close at entry or exit, any recorded split inside the window. The same panel serves every horizon.

Benchmarks: S&P 500 price index; median buy-and-hold return of all primary US listings traded at both ends of the identical window with no recorded split in it.

Statistics: pooled medians; Newey-West t-statistics on per-filing-date medians with a lag of the fourth root of the date count.

Placebo: entry shifted forward 252 sessions, same horizons, benchmarks and exclusions, compared as a paired per-event difference on events where both windows exist.

Groups: 10b5-1 plan adoption date present on every notice that day, on none, or mixed (excluded from that cut); shares to be sold over shares outstanding; aggregate market value as filed (about 4% of the events in the $10 million or more group include a notice whose stated value implies a price more than five times the prior close, most likely a filing typo); relationship as filed; number of distinct sellers that day. Company value is the largest shares outstanding stated on that day's notices times the entry close.

Conclusion

A Form 144 is a notice that an insider plans to sell, filed because the sale is large enough to require one. Over two and a half years of mandatory electronic filings, it did not mark the start of a decline. The stocks behind these notices did trail the S&P 500 badly, but they trailed it just as badly a year later, and they slightly beat the typical listed company over the same windows.

What a Form 144 reliably tells you is that the stock has usually just gone up. Insiders sell after rallies, when trading windows open and stock grants vest. A notice without a 10b5-1 plan date is a little weaker than a planned sale on the same day, which is the only cut where the reason for selling seems to matter, and even then the stock did not fall behind the typical company.

If you want to check the notices for a stock yourself, every stock page on Equibles has a Proposed Sales tab, and the free Equibles MCP server lets an AI agent read the same filings, along with the Form 4 trades that follow them. Our earlier study of what happens after an insider buys covers the other side of the trade.

References

U.S. Securities and Exchange Commission, Rule 144: Selling Restricted and Control Securities.

U.S. Securities and Exchange Commission, Form 144: Notice of Proposed Sale of Securities Pursuant to Rule 144.

U.S. Securities and Exchange Commission, Updating EDGAR Filing Requirements and Form 144 Filings, Release No. 33-11070, compliance date 13 April 2023.

Investor.gov, Form 144.

Federal Reserve Bank of St. Louis, S&P 500 (SP500), FRED.