Backtested insider buying as an earnings predictor for stocks that have reported so far
Did some light weight backtesting on insider buys for stocks that have reported earnings so far in last 2 weeks.
Pulled every Form 4 filing from EDGAR for the last 3 months, cross-referenced against 80 stocks that moved >3% on earnings over the past 2 weeks (Jan 27 - Feb 13). Checked if insider buying before earnings predicted the direction.
| Signal | Count | Avg Earnings Move | % Correct Direction |
|:---|:---:|:---:|:---:|
| Insider bought before earnings | 19 | -11.5% | 21% |
| Insider sold before earnings | 50 | +1.4% | 50% |
| No insider activity | 22 | -12.0% | 27% |
Insider buying was **wrong 79% of the time**. Worse than having no signal at all. Insider selling was a literally a coin flip.
### The failures
These insiders bought with their own money in the 3 months before earnings. All of them got destroyed:
| Ticker | Insider Buy $ | Earnings Move | What happened |
|:---|---:|:---:|:---|
| PFSI | $200K | -33.2% | Director bought at $89, stock cratered |
| RAL | $512K | -31.8% | Three different insiders bought the week before. All wrong. |
| RBBN | $74K | -27.9% | Director bought at $2.06, still dropping |
| MOH | $100K | -25.5% | Director bought at $125. Molina missed by 739%. |
| AZTA | $190K | -22.8% | Board member bought at $27 |
| LUMN | $500K | -21.6% | VP bought 78K shares at $6.35. Beat estimates by 209% and still dropped 22%. |
| CVCO | $867K | -20.3% | CEO bought at $462-500. Most expensive wrong call in the sample. |
LUMN is the funniest one. Beat earnings by 209%, insider bought $500K the week before, stock still dropped 22%.
### The only wins
| Ticker | Insider Buy $ | Earnings Move | What happened |
|:---|---:|:---:|:---|
| UAA | $219M | +20.4% | Prem Watsa (Fairfax Financial) loaded 35M+ shares across multiple days in January |
| ENPH | $723K | +38.6% | CEO Kothandaraman bought repeatedly at $30 and $51 |
| MSTR | $3.3M | +26.1% | Multiple insiders bought. Also had $6.5M in selling. MSTR gonna MSTR. |
The only buys that worked were either **massive** ($219M from a known value investor) or **from the CEO specifically** buying repeatedly.
Every single director buy, VP buy, and board member buy under $1M failed.
### What the data actually says
Small insider buys are noise. Directors buy $100K-$500K for optics or because their governance guidelines require minimum holdings. It tells you nothing about next quarter.
The only insider signal were worth watching:
- CEO/CFO buying >$1M with their own cash (not options)
- Cluster buying (3+ insiders at the same time)
- Size relative to compensation matters. For example a CFO buying $2M when they make $500K is a different signal than a board member buying $200K when they're worth $50M
Also point to note insider buying is generally a conviction signal about the next 12 months, not the next earnings print.
### Methodology
- 80 stocks that moved >3% on earnings (Jan 27 - Feb 13, 2026)
- Insider data from EDGAR Form 4 filings, last 3 months
- Only counted open market purchases (code P), not option exercises or grants
- Price data from Polygon
- Filtered to market cap >$100M