New to these reports? Start here
- Dotted-underlined words have a plain-English definition — hover or tap them. Every term is also on the glossary page.
- "Null" means we found nothing, not that something broke. Most reports here are negative results, on purpose — knowing an idea doesn't work is the point.
- Two questions get asked separately. First, is the effect real? Second, is it already priced into the betting odds? An effect can be completely real and still useless to bet on.
- A "calibration" row is a self-check. It runs the same method on something already known to be true. If that fails, the whole report is unreliable — so it's reported alongside the findings.
- If a confidence interval includes zero, the real effect might be nothing at all, so no claim gets made.
Strategy 009: Insider Cluster Buying
Source
Cohen, Malloy & Pomorski, Decoding Inside Information (Journal of Finance, 2012). Also Lakonishok & Lee (2001), Jeng et al. (2003).
Thesis
Insiders (executives, directors) trading their own stock have an information advantage. Open-market PURCHASES are the strongest signal because insiders don't get paid in cash they then have to buy stock with — they're affirmatively committing capital.
A single insider buy can be idiosyncratic (small portfolio diversification, etc.). A cluster of 3+ insiders buying within a short window (30-60 days) is much more informative: independent informed parties acting on similar information.
Formula
For each week: * For each stock, look at past 30 days of Form 4 insider transactions * Count distinct insiders (by name) who executed an open-market PURCHASE * If ≥ 3 distinct insiders, the stock is a "cluster" signal * Hold all current cluster signals, equal-weight, max 30 names * Hold each entry for 90 days
Why it might work
- Documented in 20+ academic studies across 30+ years
- Mechanism is fundamental (insiders genuinely know more)
- Cluster filter cuts noise; single-insider signal has been arbed away by Form 4 aggregators since 2010s
Why it might NOT work
- Form 4 filings are public within 2 days — heavily monitored by hedge funds and retail trackers (OpenInsider, Finviz Insider, etc.)
- Open-market buying is rarer than insider selling; very few clusters fire in any given week in modern markets
- Cluster might trigger on a small/illiquid name where retail can't execute at backtest price
Implementation (smell test version)
- Data source: yfinance (
Ticker.insider_transactions) — gives only ~18 months of history. Will use this as a smell test only. - Universe:
sp500 - Cluster definition: ≥ 3 distinct insider PURCHASE events within past 30 days (we look for transaction text containing "Purchase", "Acquisition", or "Open Market")
- Hold: equal-weight cluster names, rebalance weekly, hold 90d
Caveats unique to the smell test: * Backtest window: Aug 2024 → present (~18 months) * yfinance transaction parsing is fuzzy (text-based) * No grant/award/gift events should count, but parsing may miss edge cases — manual spot-check required
If smell test shows positive signal, invest in proper SEC EDGAR Form 4 scraper for full historical backtest (10+ years).
Expected outcome
- Academic literature: +5-10%/yr alpha on cluster signals
- Our 18-month window: too short for statistical confidence; need to see if alpha is directionally positive (worth investing more) or nullnull resultA test that found nothing. "Null" is the starting assumption that there is no real effect; a "null result" means the data gave us no reason to abandon that assumption. It does not mean the data was missing or the test failed to run./negative (reject)
Pass/fail criteria (smell test)
- Promising: alpha > 0pp on whatever signal volume fires; worth building proper EDGAR pipeline
- Reject: alpha ≤ 0pp, OR fewer than 5 cluster signals fire in the 18 months (universe too thin even on full S&P 500)