CEO Buys Look Like the Strongest Insider Signal. The Data Says Otherwise.
Published 2026-09-04. By Theodor Nielsen, founder of Form4API.
Answer
Across 61,884 open-market insider purchases, CEO buys had the highest average 3-month return of any officer role at +10.51% — but the weakest median at just +0.26% and the lowest hit rate at 50.4%, statistically a coin flip. CFO buys averaged less (+9.39%) but had the best combination of median (+5.00%) and hit rate (63.1%) of any officer category. The common framing — a CEO buying is the strongest insider signal — inverts once you look past the average.
CEO buys vs. CFO, director, and every other officer role
We classified every open-market insider purchase (transaction code P) by the buyer's officer/relationship role, excluding derivatives, superseded rows, and 10b5-1-planned trades, then measured the matured 3-month return for each role.
| Role | n | Mean 3m | Median 3m | Hit rate |
|---|---|---|---|---|
| All buys (baseline) | 61,884 | +7.67% | +2.46% | 56.8% |
| Director (non-officer) | 22,638 | +7.55% | +3.21% | 59.1% |
| 10% owner only | 8,587 | +5.06% | +1.18% | 54.9% |
| CEO | 7,673 | +10.51% | +0.26% | 50.4% |
| CFO | 2,174 | +9.39% | +5.00% | 63.1% |
| Other officer | 2,121 | +8.45% | +4.73% | 65.4% |
| President | 1,088 | +5.86% | +2.83% | 61.7% |
| COO | 744 | +34.4% | +3.53% | 58.2% |
| Chairman(n too small — aside only) | 349 | −5.74% | −3.21% | 43.3% |
Absolute, split-adjusted returns anchored at the first close on or after the filing date. Open-market purchases only (code P), excluding derivatives, 10b5-1 plan trades, and superseded rows. 2015-2026. Role is current title/relationship flag — see Methodology.
CEO sits at the top of the mean column and the bottom of both the median and hit-rate columns among major roles — the only role where those rankings point in opposite directions this sharply. CFO and director both land above the pooled baseline on every metric that matters for a single trade.
Why the CEO mean is misleading
The CEO mean is pulled up by a small number of enormous winners, not by CEO buys generally performing well. 3.28% of CEO buys returned more than +100% at 3 months — roughly 1.7x the rate for CFO buys and nearly 2.5x the rate for director buys. A handful of trades, plausibly concentrated in micro-cap and biotech names where a single catalyst can multiply a stock overnight, post triple-digit gains and drag the average well above what a typical CEO buy actually returns.
CEO buys > 100% return
3.28%
CFO buys > 100% return
1.98%
Director buys > 100% return
1.33%
COO shows the same shape even more extremely: a +34.4% mean sitting on a +3.53% median, with 10.75% of COO buys returning over +100%. The mean alone tells the wrong story for both roles — it answers "what happens if you took every buy in the dataset at once," not "what is one buy more likely to do," which is how this signal actually gets used.
Is this gap statistically real?
We ran a 10,000-resample bootstrap on the median 3-month return for each role to check whether the CEO underperformance could be noise.
- CEO median — 95% CI[−0.05%, +0.88%]
- CFO median — 95% CI[+4.48%, +5.54%]
- Director median — 95% CI[+2.99%, +3.51%]
- CFO − CEO gap — 95% CI[+3.98pp, +5.37pp]
- Director − CEO gap — 95% CI[+2.28pp, +3.43pp]
The CEO interval crosses zero at every horizon tested (1m, 3m, 6m) — we can't rule out a true CEO edge of nothing. CFO and director stay clearly positive at every horizon, and neither gap interval crosses zero. This is a measurable difference, not sampling noise.
Consistent with prior academic research
This isn't a pattern unique to Form4API's data. A 2014 study in Managerial Finance — Knewtson & Nofsinger, “Why are CFO insider trades more informative?”, Vol. 40 No. 2, pp. 157-175 — found the same ranking: CFO purchase portfolios outperformed CEO purchase portfolios. The authors tested two explanations and the data favored the scrutiny hypothesis: CEOs face more public visibility and oversight, so they trade more conservatively and less aggressively on their own information. The alternative — that CFOs simply have superior financial judgment — was not supported. That's a plausible mechanism behind what shows up in our data as a fat right tail on CEO returns rather than a shifted center.
What this data can and can't tell you
Role is a snapshot of today, not of the trade date
We classify role from the insider's current title, not a point-in-time record at the transaction. Someone promoted from CFO to CEO after their trade shows up under "CEO." We don't have a title-history table to correct this.
24.5% of the clean population had no role flag, and was excluded
That unclassified bucket (15,159 rows) concentrates heavily in 2023-2026 filings — likely an ingestion gap for recent filings, not a real "no role" population. Disclosed rather than dropped silently.
Chairman (n=349) is the smallest, most dramatic-looking row
More than 60x smaller than the CEO bucket. Included for completeness, not leading with it — and neither should you.
Association, not a backtest
Every return here is a historical price move after a real filing — no transaction costs, no execution model, no causal claim. Not investment advice.
Doesn't contradict our earlier buy-side study
The pooled baseline here (+7.67%, 56.8%) differs slightly from the insider buying returns study (+8.6%, 58%) because of a wider date window (2015-2026 vs Oct 2023-Feb 2026), not a change in method.
Methodology
- Population. Open-market purchases (code P) with a matured 3-month return, 2015-2026. n=61,884 after exclusions.
- Exclusions. Derivatives, 10b5-1 plan executions, and superseded rows removed before role classification.
- Role classification. Current title/relationship flag. Unclassified rows (24.5%) excluded from this comparison.
- Return definition. Absolute, split-adjusted, anchored at first close on/after filing date, measured at 63 trading days.
- Statistical testing. 10,000-resample bootstrap for 95% confidence intervals, run at 1/3/6-month horizons.
Reproduce it
The same role flags and return fields are exposed on the API. There is no query parameter for a specific title like CEO or CFO — role filters only at the coarser officer / director / ten_percent_owner level. Pull officer-flagged open-market buys with matured returns, then group by the per-row insiderTitle field client-side to reproduce the table above:
curl "https://api.form4api.com/v1/transactions?code=P&exclude_10b5=true&role=officer&has_returns=true" \ -H "X-Api-Key: $FORM4API_KEY"
Each row carries insiderTitle alongside the isOfficer / isDirector / isTenPercentOwner flags and the return fields (return1d/1w/1m/3m/6m, stored as fractions — 0.109 = +10.9%), so bucketing by CEO vs CFO vs president is a filter on that string once you have the officer rows in hand. The screener guide covers the full filter grammar, and /docs has the complete field reference.
Frequently asked questions
Is a CEO buying stock a strong signal?
Less strong than the mean alone suggests. Across 61,884 open-market insider purchases, the median CEO buy returned just +0.26% at 3 months, and only 50.4% were positive at all — a coin flip. The average (+10.51%) is inflated by a thin slice of outsized winners, not by CEO buys performing well as a rule.
Which officer role has the most reliable signal?
CFO, by median (+5.00%) and hit rate (63.1%) — the best combination of any officer category, with a bootstrap confidence interval that stays positive at every horizon tested. Director (non-officer) buys are close behind, with a median of +3.21% and a 59.1% hit rate.
Why does the CEO mean look so much better than the median?
Right-skew. 3.28% of CEO buys return more than +100% at 3 months — roughly 1.7x the rate for CFO buys and nearly 2.5x the rate for director buys — and that thin tail of enormous winners drags the average well above what a typical CEO buy actually returns. COO shows the same shape even more extremely: a +34.4% mean sitting on a +3.53% median.
Is this a real difference, or noise?
We bootstrapped the median 3-month return for each role 10,000 times. The CEO median's 95% confidence interval crosses zero at every horizon tested (1, 3, and 6 months) — we cannot rule out a true CEO edge of nothing. The CFO and director intervals stay clearly positive at every horizon, and the CFO-minus-CEO and director-minus-CEO gap intervals don't cross zero either. This is a measurable difference, not sampling noise.
Has anyone else found this?
Yes. A 2014 peer-reviewed study in Managerial Finance (Knewtson & Nofsinger, "Why are CFO insider trades more informative?", Vol. 40 No. 2, pp. 157-175) found the same ranking — CFO purchase portfolios outperformed CEO purchase portfolios — and attributed it to the scrutiny hypothesis: CEOs face more public visibility and oversight, so they trade more conservatively on their own information. The alternative explanation, that CFOs simply have superior financial judgment, was not supported by their data.
Does this contradict Form4API's other insider-buying study?
No. The pooled baseline here (+7.67% average, 56.8% hit rate) differs slightly from the earlier buy-vs-sell study (+8.6%, 58%) because of a wider date window (2015-2026 vs October 2023-February 2026) applied to a role-level cut of a similar population — not a change in method or a correction.
Bottom line
Financial media treats a CEO buying their own stock as the loudest, most bullish signal a company can send. In 61,884 open-market insider purchases since 2015, it's the officer role with the weakest median return and lowest hit rate of any major category — propped up in the headlines only by a thin, fat-tailed slice of outsized winners. CFO and director buys, quieter and less covered, are the more consistent signal by every measure except the one the headlines use.
Population statistics from publicly filed SEC Form 4 data, for research and education. Not investment advice; past performance does not predict future returns.