The Cluster Buy Signal: Deconstructing the Excess Return Illusion
We analyze 12,853 synchronized executive purchases to determine if clustered insider buying generates true excess returns, or if the data is hiding the real story.
"There are many reasons an insider might sell, but there is only one reason they buy: they believe the stock will go up."
While this Wall Street adage makes for a great headline, quantitative finance requires more than just intuition. Prior academic research on insider trading has documented abnormal returns following insider purchases, particularly when purchases are clustered among multiple insiders. But how exploitable are these returns in the real world once you factor in market drift and execution slippage?
To find out, our quantitative team ran a historical backtest on 12,853 distinct Cluster Buy events using the InsiderAlpha Master Dataset. We wanted to move beyond the hype and evaluate this signal as a true quantitative feature.
Rigorous Backtest Parameters
Before looking at the numbers, it is critical to define the exact parameters of our backtest to prevent look-ahead bias and survivorship bias.
- Universe: US Equities (NYSE/Nasdaq). Excludes OTC markets, penny stocks (Micro-caps under $50M), and biotech companies (to prevent FDA binary events from skewing the distribution).
- Survivorship Bias: The dataset includes delisted companies and bankruptcies. If a company went to zero, it is reflected in the returns.
- Entry Execution Timing (T+1): You cannot buy the stock before the SEC filing is public. Our backtest assumes market-open execution on the trading day after the final Form 4 in the 72-hour cluster is officially filed.
- Benchmark: We use the Russell 2000 as our primary benchmark, as insider cluster buys overwhelmingly occur in small-to-mid-cap equities rather than mega-cap tech.
The Results: Absolute Return vs. Excess Return
The data reveals a non-linear relationship between the size of the cluster and the expected absolute return. However, absolute returns in a bull market mean very little. Here is the true Excess Return generated over a 180-day holding period:
| Cohort | 180-Day Absolute Return | Russell 2000 Benchmark | Excess Return |
|---|---|---|---|
| 3 Insiders | +12.55% | +8.10% | +4.45% |
| 4 Insiders | +11.10% | +8.10% | +3.00% |
| 5 Insiders | +5.65% | +8.10% | -2.45% |
| 6+ Insiders | +21.82% | +8.10% | +13.72% |
Note: The 3,784 transactions in the 6+ cohort represent individual Form 4 filings. The statistical sample for the 6+ cohort therefore consists of approximately 540 independent corporate events, rather than 3,784 individual trades.
Deconstructing the 6+ Cluster: The Distribution Reality
At first glance, the 6+ Insider cohort appears to be a Holy Grail, generating a +13.72% excess return in 180 days. However, as quants, we must look at the distribution of these returns. Averages can be dangerously misleading if the variance is massive.
When we break down the 6+ cluster cohort, the reality of the strategy emerges:
- Mean (Average): +21.82%
- Median: +6.40%
- Win Rate (Trades > 0%): 54.2%
- Maximum Drawdown: -42.1%
What does this mean? The +21.82% average is heavily skewed by a right-tail distribution. In approximately 10% of cases, the stock surged over +150% (often due to an unexpected acquisition or aggressive buyback). However, almost half of the time (45.8%), the trade lost money. Trading this strategy blindly would result in massive drawdowns.
The Missing Feature: Dollar Conviction
The raw count of insiders buying is only half the equation. The other half is the relative size of the purchase. Four directors buying $10,000 each in a $10 Billion company is statistical noise. One CFO buying $2 Million in a $100 Million company is a screaming anomaly.
A robust algorithmic model shouldn't trade solely on a cluster count. It requires a weighted feature we call Insider Conviction: (Total Cluster Dollar Volume / Market Cap)
Limitations
As with any quantitative backtest, this analysis has limitations that must be considered before deploying capital. This analysis does not account for:
- Transaction costs and bid-ask spreads.
- Portfolio concentration limits and sizing constraints.
- Liquidity constraints (slippage) in smaller micro-cap companies.
- Tax implications.
- Macroeconomic market regime changes.
The Verdict: Feature vs. Strategy
At InsiderAlpha, we don't just scrape Form 4s. We convert regulatory data into quantitative intelligence.
If you read a headline that says "Buy any stock where 6 insiders buy and you'll make 21%", run the other way. While Cluster Buying is not a standalone trading strategy, it is an exceptionally powerful quantitative feature.
A modern algorithmic model uses an equation similar to:
Cluster Score + Dollar Conviction + Earnings Trend + Valuation + Technical Momentum = Final Signal
When a coordinated, high-dollar cluster buy aligns with strong fundamentals and technical momentum, the probability of a breakout increases significantly.
Want to build your own model using these features? Download our premium CSV database featuring pre-calculated Cluster Sizes, Dollar Conviction, and ROI tracking for over 1.5 million transactions, allowing you to run your own rigorous out-of-sample tests.