When does copying insider buying actually pay?
We took every meaningful open-market insider purchase since 2006 — 69,355 of them — and asked one simple question about each: if you had bought the same stock the day the filing became public and held it for three months, would you have beaten a plain index fund? This page is the answer, split by the kind of market it happened in and by industry — because it turns out those two things decide almost everything.
In a normal market, you would likely have done about as well just buying an index fund. The edge from copying insider buys in calm conditions is one to two points over three months, before trading costs — thin enough that costs could eat it. Where the edge has been real is stressed markets: buys made while the market was being sold off went on to beat the index by three to five points, across every strategy variant we tested. Insider buying is a tool you take out when things are frightening, not an everyday replacement for an index fund.
1. It works best when the market is frightened. The biggest edge, by far, came from buys made during crashes and panics — up to about +5% over 3 months versus the index.
2. In calm markets the edge is small — one to two percentage points, before trading costs. Real, but thin. And in the current market we can't measure any edge at all.
3. The industry matters enormously. Some industries carry a real edge; banks (a third of all insider buying!) carry almost none; and a few — energy, chip makers — only work when their own sector is being dumped.
Every percentage below answers the same question: how much better (or worse) did these buys do than a small-company index fund (the Russell 2000, via IWM) over the following 3 months?
Worked example: +4.8% means that if the index fund returned 0% over those 3 months, the insider-buy stocks returned +4.8% on average. If the fund fell 20%, they fell about 15.2%. The number is the gap, not the raw return.
Why compare to a small-company fund and not the S&P 500? Because insider buying happens mostly at small companies. Compare those to the S&P and you are really measuring company size, not stock-picking. (Since late 2022 the S&P has beaten small-caps by ~36 points — that alone would make any small-cap list look terrible.)
"Average" vs "typical": the average can be dragged up by a few huge winners. The typical (median) figure is what the middle trade did — when the two disagree a lot, most trades did worse than the average suggests, and we say so. Confidence is shown in words: Solid = very unlikely to be luck, Weak = could be luck, Unproven = treat as chance. (The raw statistics are in the fold at the bottom.)
1 · Eleven kinds of market, one test each
A single 20-year average would hide the whole story, because it would blend crashes into booms. So we split 2006–today into eleven periods by what the market was actually doing, and ran the identical test inside each. Amber rows are the stress periods. Bars to the right of the line = the buys beat the index fund; left = they lagged it.
The pattern: the four best periods are all panics or the recoveries right after them. In ordinary bull markets the edge shrinks to a point or two, and in the current market (bottom row) it is statistically indistinguishable from zero. That is why we describe insider buying as a drawdown signal — a tool for frightened markets, not an everyday money machine. Anyone who quotes you one blended average for "insider buying beats the market" is averaging 2008 into 2024 and hoping you don't ask.
The picture over time — signal vs the market, same windows
Green line: put the average result of each month's insider buys (held 3 months) end to end, from the start of the window you pick. Grey line: the small-cap index over the same entry dates and holding windows. Dashed: the S&P 500. Where green pulls away from grey, the buys were beating the market you could have bought instead. Pick a period — the lines restart at zero so every window is a fair race from its own starting line.
The rule: when an insider files a genuine open-market purchase — not a scheduled 10b5-1 plan, not a routine same-month-every-year habit — buy the same stock at the close of the next trading day after the filing appears, in an equal dollar amount, and sell after 63 trading days (~3 months). No discretion, no exceptions. Every number on this page is that one rule, applied 69,355 times.
Where you would see these trades: the live signals feed is exactly this filter running in real time — each row is a qualifying filing, already screened the same way the study screened. What the study adds is the context: when (frightened markets) and where (which industries) that rule has actually paid, and when it has not.
Every strategy, in every kind of market
These are the same strategies the live scorecard tracks — run through the full 20 years instead of 13 months, one column per market period. Each cell: how much that strategy's buys beat (green) or lagged (red) the small-cap index over the next 3 months. Hover a cell for the trade count. The pattern to notice: every row lights up in the same columns. The strategy you pick matters less than the market you are in — no variant we tested escapes the weather.
| Strategy | Late credit boom | 2008 crisis | Post-crisis recovery | Euro crisis | Mid-cycle bull | Oil bust | Late-cycle bull | COVID crash | Stimulus melt-up | Rate shock | Now (AI era) |
|---|---|---|---|---|---|---|---|---|---|---|---|
| All qualifying buys | +0.4% | +4.8% | +4.0% | +3.6% | +1.4% | +2.8% | +1.5% | +0.7% | +1.8% | -0.4% | -0.2% |
| Cluster buys (>=3 insiders) | +1.0% | +5.4% | +4.7% | +5.4% | +1.0% | +3.7% | +0.9% | -0.1% | +0.1% | +1.5% | -0.2% |
| Brand-new position | -0.7% | +5.3% | +2.6% | +13.1% | -0.1% | +1.8% | +0.8% | +1.0% | -0.1% | -0.5% | -0.9% |
| Grew stake 30-99% | +2.2% | +8.7% | +3.7% | +3.2% | +1.7% | +2.4% | +1.3% | +7.0% | +2.3% | +0.1% | -0.6% |
| Grew stake 100%+ | -1.4% | +9.4% | +5.5% | +3.3% | +0.8% | +0.6% | +1.9% | -1.8% | +3.1% | +3.1% | -0.8% |
| CFO / COO tier | +0.5% | +4.1% | +1.7% | +3.2% | +1.3% | +7.7% | +2.1% | +1.4% | +2.7% | +2.2% | -1.4% |
| Directors | -1.0% | +5.6% | +2.0% | +3.8% | +1.4% | +2.7% | +1.0% | -2.0% | +0.5% | -0.1% | +0.5% |
| Officers | +0.9% | +5.0% | +4.3% | +2.0% | +1.8% | +3.6% | +1.9% | +3.8% | +3.5% | +0.1% | -0.8% |
| $100K+ buys | +0.2% | +4.4% | +3.6% | +4.0% | +0.3% | +1.6% | +1.1% | +0.4% | +2.0% | -1.3% | -0.9% |
Where are the Overreaction and below-insider-cost strategies? They cannot be reconstructed over 20 years, and we would rather leave a gap than fake one. Overreaction requires knowing that every seller was on a pre-scheduled 10b5-1 plan — the disclosure checkbox for that only exists since April 2023. Below-insider-cost requires the insider's own fill price compared to the market at scan time, which the historical record does not retain. Both remain on the live scorecard, where they are measured forward in real time — their strong 13-month numbers are promising and unproven at 20-year scale, and that is exactly how we label them.
2 · Where it works — by industry
"In market stress" = the same test, restricted to buys made during the five dated windows below. "In normal markets" = every other time between 2006 and today.
- Global financial crisis (Oct 2007 – Mar 2009) — S&P 500 fell ~57%; Lehman and Washington Mutual failed
- Euro crisis & US downgrade (Apr – Oct 2011) — US lost its AAA rating; a fast ~19% drop
- Oil bust (May 2015 – Feb 2016) — crude fell from ~$60 to $26; energy earnings collapsed
- COVID crash (Feb – Mar 2020) — 34% fall in 23 trading days
- Inflation & rate shock (Jan – Oct 2022) — fastest rate-hiking cycle since 1981
"This sector today" is live, not historical: we track each industry through a well-known sector fund (named in the table) and call it "being sold off" when that fund is at least 10% below its own 1-year high. Why it matters: for some industries the entire edge only exists in that condition.
| Industry | Edge vs index fund | Average | Typical trade | Confidence | In market stress | In normal markets | This sector today |
|---|---|---|---|---|---|---|---|
| Medical devices
35 companies · 1,662 buys
|
+7.5% | +1.2% | Weak | +14.5% | +5.7% |
21% below its 1-yr high — being sold off now
tracked via IHI |
|
| Transport & logistics
23 companies · 775 buys
|
+6.8% | +1.8% | Solid | +16.2% | +3.4% |
9% below its 1-yr high — normal
tracked via IYT |
|
| Biotech & pharma
54 companies · 2,378 buys
|
+5.3% | -1.1% | Weak | +0.1% | +7.6% |
8% below its 1-yr high — normal
tracked via XBI |
|
| Other
62 companies · 2,662 buys
|
+5.0% | +0.6% | Solid | +2.9% | +6.1% | not tracked | |
| Industrials & manufacturing
229 companies · 8,303 buys
|
+3.5% | +1.2% | Solid | +3.4% | +3.5% |
8% below its 1-yr high — normal
tracked via XLI |
|
| Retail & consumer
63 companies · 2,810 buys
|
+3.2% | +1.5% | Solid | +4.5% | +2.7% |
9% below its 1-yr high — normal
tracked via XLY |
|
| Food, drink & tobacco
16 companies · 418 buys
|
+2.7% | +0.5% | Unproven | +7.1% | +1.6% |
7% below its 1-yr high — normal
tracked via XLP |
|
| Services
91 companies · 3,642 buys
|
+2.2% | -0.8% | Weak | -0.4% | +3.2% | not tracked | |
| REITs
57 companies · 3,462 buys
|
+2.1% | 0.0% | Unproven | 0.0% | +2.9% |
6% below its 1-yr high — normal
tracked via VNQ |
|
| Software & IT services
42 companies · 2,006 buys
|
+1.9% | -1.8% | Unproven | +2.7% | +1.6% |
14% below its 1-yr high — being sold off now
tracked via IGV |
|
| Finance & real estate
46 companies · 1,676 buys
|
+1.4% | +0.7% | Weak | +1.4% | +1.4% |
2% below its 1-yr high — normal
tracked via XLF |
|
| Insurance
48 companies · 2,136 buys
ONLY WORKS IN A SELLOFF |
+1.3% | -0.2% | Weak | +3.6% | +0.3% |
6% below its 1-yr high — normal
tracked via KIE |
|
| Energy
51 companies · 2,352 buys
ONLY WORKS IN A SELLOFF |
+1.0% | -2.1% | Unproven | +6.8% | -1.3% |
0% below its 1-yr high — normal
tracked via XLE |
|
| Banks
184 companies · 21,815 buys
ONLY WORKS IN A SELLOFF |
+0.9% | +0.2% | Solid | +3.6% | +0.1% |
5% below its 1-yr high — normal
tracked via KRE |
|
| Utilities
36 companies · 2,510 buys
|
+0.7% | -0.2% | Unproven | +2.5% | +0.1% |
11% below its 1-yr high — being sold off now
tracked via XLU |
|
| Tech hardware & semis
70 companies · 1,774 buys
ONLY WORKS IN A SELLOFF |
+0.5% | -2.5% | Unproven | +4.9% | -1.2% |
15% below its 1-yr high — being sold off now
tracked via SMH |
|
| Construction
20 companies · 542 buys
|
-2.6% | -2.3% | Unproven | -1.8% | -2.9% |
24% below its 1-yr high — being sold off now
tracked via ITB |
What to actually take from that table
Banks are a third of all insider buying and carry almost no edge. Community-bank directors buy small amounts constantly — it is closer to a savings habit than a view on the business. They used to crowd the top of our live feed on sheer volume; we now rank them down, and every demoted signal on the feed says so and why.
Energy and chip makers flip sign. Insider buying in energy returned about +7% when the sector was being dumped and lost to the index when the sector was calm. Same for tech hardware. So our feed only surfaces those near the top while their sector is actually under pressure — the "This sector today" column above is that switch, live.
The most trustworthy row is not the biggest number. Industrials & manufacturing: a moderate edge, but across 229 companies with solid confidence — the strongest evidence in the table that something real is happening. Medical devices shows a bigger average from far fewer companies, with weak confidence: could be real, could be a handful of lucky trades.
3 · What this page cannot tell you
No trading costs. These are small, often illiquid stocks; spreads and slippage could eat much of a 1–2 point edge.
Only companies that survived. Stocks that were bought out or went bankrupt have no price history left to test. We measured
that gap instead of ignoring it: buyouts outnumbered bankruptcies about 8-to-1 among the ones we could trace, and since those push the
results in opposite directions, the net distortion is under one percentage point.
History, not prophecy. This describes 2006–today. The next crisis is under no obligation to look like the last five.
An average is not you. These are averages over thousands of trades. Any single trade can and does do anything.
For the statistically inclined — method & the raw numbers
| Period | vs S&P 500 | vs small-caps | % that beat the index | per-company mean | t (clustered) | n |
|---|---|---|---|---|---|---|
| Late credit boom | -1.10% | +0.41% | 48% | +0.46% | 0.6 | 6,600 |
| Global financial crisis | +5.89% | +4.83% | 54% | +6.26% | 4.4 | 10,119 |
| Post-crisis recovery | +5.63% | +4.03% | 51% | +3.99% | 3.6 | 7,907 |
| Euro crisis & US downgrade | +1.84% | +3.59% | 62% | +2.37% | 2.4 | 2,236 |
| Mid-cycle bull | +1.58% | +1.40% | 49% | +1.27% | 2.3 | 11,334 |
| Oil bust & industrial recession | +0.22% | +2.84% | 59% | +2.32% | 2.1 | 2,401 |
| Late-cycle bull | +0.71% | +1.51% | 51% | +1.57% | 2.6 | 10,570 |
| COVID crash | +2.31% | +0.67% | 39% | +0.40% | 0.3 | 1,235 |
| Stimulus melt-up | +3.33% | +1.81% | 51% | +1.77% | 2.2 | 4,844 |
| Inflation & rate shock | +0.29% | -0.38% | 47% | +0.86% | 0.7 | 2,533 |
| AI-led recovery (now) | -1.28% | -0.24% | 45% | +0.48% | -0.5 | 9,559 |
Study generated 2026-08-20 from SEC Form 3/4/5 data · this same table drives the ranking of the live signals feed, so the two can never disagree. Educational, not investment advice. Disclosures.