Insider buying is the most repeated idea in retail investing: when the people running a company buy its stock with their own money, follow them. We hold every Form 4 filed since 2021, so we tested it. Bought blindly, the idea does not work.
What we tested
Every open-market purchase (Form 4 transaction code P) of common stock filed with the SEC since January 2021, priced between $1 and $5,000, where we hold prices for the company. That covers 4,223 of the 5,516 companies with an insider purchase (76.6%). The entry is the first close after the filing date, because that is the first price anyone reading the filing could have traded at. The trade date is earlier, and using it would quietly assume you knew about the purchase before it was public.
Returns use split- and dividend-adjusted prices, and are measured against the Russell 2000 (IWM) over the same 90 days. We use the small-cap index as the fair benchmark because insider buying is concentrated in small companies; against the S&P 500 the same purchases look worse (-0.86%), but most of that gap is the size factor, not the signal.
What we found
On their own, insider-bought stocks did rise: +2.69% on average over 90 days. But the market rose too. Measured against the Russell 2000 the average purchase returned -0.11%, the median purchase -1.76%, and only 45.4% of purchases beat the index. That is not distinguishable from zero.
The famous filters do not rescue it. Big purchases, purchases by the top three officers, and cluster buying, where three or more insiders buy within 90 days, all trailed the index too:
| Strategy | Buys | Mean vs R2000 | Median | Beat index | t |
|---|---|---|---|---|---|
| Every purchaseopen-market buys, Form 4 code P | 40,427 | -0.11% | -1.76% | 45.4% | -0.8 |
| Big purchases$250K or more | 10,208 | -0.85% | -2.28% | 44.2% | -2.3 |
| C-suitechief executive, finance or operating officer | 2,970 | -1.67% | -3.63% | 42.2% | -2.4 |
| Clusterthree or more insiders within 90 days | 14,376 | -0.73% | -1.17% | 46.7% | -3.4 |
t is the t-statistic of the mean: beyond about 2 either way, the result is unlikely to be chance. Negative means the strategy trailed the index.
The one lead, and why it also failed
Cut by features a trader could know on the filing day, one group looked promising: purchases made while the stock was within 10% of its one-year high. Those beat other near-high stocks on the same days by a few percent at 90 days. On 26 September 2026 we matched each purchase to stocks of the same sector and size on the same day, also near their high, with no insider buying, and charged 1% for trading costs. The edge disappeared: the median insider purchase returned -0.09% after costs against +0.41% for its matched controls.
What insider filings are actually good for
This does not make Form 4 useless. It makes it context rather than a signal. A chief executive selling outside a pre-scheduled 10b5-1 plan, three directors selling in the same week, or a large holder filing to sell before a quiet quarter are facts worth knowing about a company you own. They are not, on their own, a reason to buy one.
How to check us
The full study, with every strategy, both benchmarks and the losing cuts, is on the backtest page. An earlier version of this study used unadjusted prices and overstated the result; that error and its fix are listed on our corrections page. Every figure on WaveMan comes from public filings, and a sample is re-checked against the original source every night (how). Not investment advice.