Do insiders
beat the market?

Backlog #18: “do corporate insiders beat the market?” — the open-market version of insider trading, not the congressional kind this desk has already checked three times over (004, 007, 017). Officers, directors and 10%+ owners have to disclose every open-market purchase of their own company's stock on SEC Form 4 within two business days. The folk theory is straightforward: these are the people who actually know how the quarter is going, so when they buy with their own money, it's a signal worth following. This run pulls the SEC's own structured Form 3/4/5 data set for 2024 Q2 — every non-derivative transaction coded P (open-market purchase, shares acquired) — and matches each one to its ticker's forward return against SPY over the ~126 trading days (roughly six months) that follow.

The headline says yes. Across 5,111 purchases matched to a full forward-return window (of 6,463 collected; the rest were tickers yfinance couldn't price, mostly thinly-traded micro-caps), the mean excess return — stock forward return minus SPY's forward return over the identical window — is +7.8 percentage points, 95% CI [+6.0, +9.6], p<0.001. That's not a fragile result: clip the top and bottom 1% of the distribution and it barely moves, +5.8 pts, CI [+4.5, +7.2]. Insiders, on average, bought stocks that went on to beat the market.

The typical purchase says no. The median excess return is -1.4 points — essentially flat, on the losing side of flat. The win rate — the share of purchases that actually beat SPY over the same window — is 48.0%, 95% CI [46.7%, 49.4%] — an interval that sits entirely at or below a coin flip. Those two paragraphs describe the same 5,111 purchases. The mean and the median disagree about the sign because the distribution has a long right tail: a purchase in Power Solutions International (PSIX) ran +903 points against SPY; one in TSS, Inc. (TSSI) ran +970. A handful of insiders bought right before a moonshot. Most didn't buy anything the market didn't already know.

Checked plainly: does a bigger purchase mean more conviction, and a bigger edge? The claim implies it should — an insider putting in $2 million is telling you something a $2,000 token buy isn't. It runs backward. Regressing excess return on the log₁₀ of the purchase's dollar value: slope -6.19 points per 10× increase in purchase size, 95% CI [-7.85, -4.53], p<0.001 — excludes zero, and it's negative. R²=0.010 is small, so dollar size is explaining almost none of the variance, but the direction is consistent and a rank-based check agrees: Spearman ρ=-0.186, p<0.001. Bigger buys, in this data, do not signal more; if anything they do slightly worse. The moonshots above were both comparatively small purchases.

Read plainly. There is a real, repeatable-looking average edge here, and it survives the obvious outlier check. But "insiders beat the market" as usually meant — follow the filing, buy what they buy — describes at best a coin flip for any single purchase, because the average is earned entirely in the tail: a small number of insider buys that preceded an enormous run-up, concentrated in exactly the kind of illiquid micro-cap stock where a Form 4 doesn't require much money to file. This is one quarter of purchases carried forward through mid-2024 to early 2025, a period when small caps broadly rallied; the SPY benchmark nets out the overall market's move but not a small-cap risk premium these purchases likely carry as a group, which this run does not attempt to isolate.