Correlation Risk: Why 10 "Diversified" Stocks Can Still Be One Bet
You hold 10 stocks. AAPL, MSFT, NVDA, GOOGL, AMZN, META, TSLA, AMD, AVGO, NFLX. Congratulations — you own one position with 10 ticker symbols.
When the NASDAQ drops 3%, your "diversified" portfolio drops 3.1%. That's not diversification. That's concentration with extra steps.
Effective Number of Bets
The mathematically correct way to measure diversification is the effective number of bets:
Where wᵢ is the weight of position i in your portfolio.
Examples:
• 10 positions, all equal weight (10% each): N_eff = 1/Σ(0.1²) = 1/0.1 = 10 ✓
• 10 positions, but 5 are highly correlated: effective N drops to ~4–5
• 10 positions, all in the same sector: effective N ≈ 1–2
The key insight: N_eff accounts for correlation. Two 50/50 positions in AAPL and MSFT (correlation ~0.85) give you N_eff ≈ 1.3. You're essentially holding one bet with slightly less idiosyncratic risk.
How to Calculate the Correlation Matrix
For a portfolio of N stocks, the correlation matrix is an N×N matrix where each element is:
Where R is the return series (daily or weekly) and Cov is the sample covariance over a lookback window (60–120 days typical).
In practice, you don't need the full matrix. You need two numbers:
- Average pairwise correlation — if this is above 0.5, your portfolio is effectively 2–3 bets, not N bets.
- Maximum cluster size — group stocks with ρ > 0.7. If one cluster has 6+ members, you have a concentration problem.
Sector vs Factor Correlation
Obvious correlation is easy to spot (all tech stocks move together). Factor correlation is subtler:
- High-beta cluster: High-growth stocks correlate regardless of sector. A "diversified" portfolio of high-beta energy + tech + consumer discretionary is one high-beta bet.
- Rate sensitivity: In a rising-rate environment, long-duration assets (growth, unprofitable tech, REITs) all correlate ~0.8+ regardless of sector.
- Liquidity correlation: In a crisis, everything correlates to 1.0. Diversification fails exactly when you need it most. This is why position sizing matters more than stock selection.
Practical Diversification Rules
- Max 3 positions per "factor cluster." If you define clusters by sector + beta + rate sensitivity, no cluster should have more than 3 of your positions.
- Target N_eff ≥ 5 for a 10-position portfolio. If your calculation shows N_eff < 4, you're more concentrated than you think.
- Stress-test the correlation matrix. Calculate correlations in both calm and crisis windows. If calm-window ρ = 0.3 but crisis-window ρ = 0.9, your "diversification" is an illusion.
- Keep 20–30% in uncorrelated assets. Cash, short-duration bonds, or gold have near-zero correlation to equities in most regimes. This is your real diversification.
What GemStox Does About This
When GemStox generates multiple signals for the same user, the system checks pairwise correlation between the suggested positions. If two signals are for stocks with ρ > 0.7, the second signal's position size is automatically reduced. You get the informational value of both signals without the hidden concentration.
Diversification that's actually measured, not assumed.
GemStox signals include correlation-aware position suggestions so your portfolio doesn't secretly become one big bet. Try it for $3.
See Correlation-Aware Signals →