@quantpad: Insider buying backtest on 122,678 SEC Form 4 open-market purchases filed since 2018, joined to daily US equity prices. We tested whether following company insiders into their own stock actually beats doing nothing. Buying at the close of the first trading day after each filing returned +5.8% over six months, net of a 20bp round trip. That number holds up: it survives a doubled cost assumption (+5.6%) and standard errors clustered by calendar date (t = 2.43). On its own it looks like an edge. Then we measured what those same stocks did on ordinary days when no insider filed anything: +33.8% over the same six-month horizon. A randomly drawn basket of companies nobody's insiders were buying returned +29.5%. Following the insiders underperformed simply owning the names by about 28 percentage points. The obvious objection is that you are supposed to filter for conviction. Several insiders buying at once, six figures each. Every filter we tested made it worse. Cluster buys were 0.5pp worse than taking every purchase. Purchases of $100,000 or more were 5.6pp worse. The combined filter was 7.3pp worse. Purchases over $500,000 returned -7.7% at three months across 465 events (t = -4.92), and over $1m, -7.9%. Split by year, 2020 (+25.6%) and 2022 (+24.8%) carry the entire average. Both rebound years. Insiders buy the dip, so the trade is really a bet on the bounce, and four of the eight complete years lost money. Sample: 3,130 priced events across 88 companies, a fixed-seed random draw from the 4,506 issuers with at least one qualifying purchase. The filing census is complete; the price join is a sample, and we say so on screen. The CEO/CFO filter was not computed because the structured extract keeps the officer flag but not the officer's title, and faking it would have made a false test. The filings are free from the SEC. The eight years of daily prices for thousands of tickers needed to check them is the part that normally costs money, and it is included with QuantPad. #QuantPad #insidertrading #backtesting #stockmarket #quant

QuantPad

Open In TikTok:

Region: US

Monday 31 August 2026 19:28:40 GMT