Diversification
Diversification is the portfolio construction principle of spreading investments across multiple assets, sectors, geographies, or strategies such that the imperfect correlation between holdings reduces the portfolio's total risk below the weighted average of its individual component risks. It is the primary mechanism through which investors can reduce idiosyncratic risk without sacrificing expected return.
Key takeaways
- Diversification eliminates idiosyncratic (company-specific) risk but cannot eliminate systematic (market-wide) risk.
- The risk-reduction benefit of adding assets diminishes as portfolio size increases; most idiosyncratic risk is eliminated by 20–30 well-diversified holdings.
- Diversification benefits depend on correlation: assets with correlation near 1.0 provide little benefit, while negative or zero correlations provide maximum benefit.
- In crisis periods, correlations across asset classes frequently converge toward 1.0, reducing the effectiveness of cross-asset diversification precisely when it is most needed.
- True diversification requires uncorrelated return sources, not merely different labels—many apparently diverse assets have latent common factor exposures.
Explanation
Diversification is rooted in the mathematical properties of portfolio variance. For a two-asset portfolio, variance is: σ²_p = w₁²σ₁² + w₂²σ₂² + 2w₁w₂σ₁σ₂ρ₁₂, where w₁ and w₂ are portfolio weights, σ₁ and σ₂ are individual asset volatilities, and ρ₁₂ is the correlation between assets. When ρ₁₂ < 1, portfolio variance is less than the weighted average of individual variances—the defining mathematical condition for diversification benefit. When ρ₁₂ = -1 (perfect negative correlation), complete risk elimination is theoretically possible through optimal weighting.
Extending to N assets, portfolio variance becomes: σ²_p = Σᵢ Σⱼ wᵢwⱼσᵢσⱼρᵢⱼ. As N grows large, the contribution of individual variances (diagonal terms) shrinks, and the portfolio variance approaches the average covariance between pairs of assets. This demonstrates that diversification cannot reduce risk below the average pairwise correlation level of the portfolio—systematic risk, captured by common factor exposures (market beta, credit beta, etc.), persists regardless of how many assets are held.
In modern portfolio theory (Markowitz, 1952), diversification is optimized through mean-variance optimization, which identifies portfolio weights that maximize expected return for a given level of portfolio variance. The set of optimal portfolios traces the efficient frontier—the uppermost boundary of achievable return/risk combinations. The optimal portfolio for a given investor lies on the efficient frontier at the point where their indifference curves (reflecting risk aversion) are tangent to the frontier.
For hedge fund portfolios, diversification is pursued across several dimensions: strategy (long/short, macro, event-driven, arbitrage), time horizon (short-term momentum versus long-term value), geography, and factor exposure. A hedge fund of funds explicitly manages diversification across managers, aiming to combine strategy specialists whose returns are driven by genuinely different underlying mechanisms. The challenge is that diversification benefits are estimated using historical correlations, which can be unstable and particularly misleading during stress regimes.
Practitioners have documented a phenomenon known as correlation breakdown or correlation contagion during market crises: assets that appeared uncorrelated under normal conditions exhibit sharp correlation increases during drawdowns. This occurs because common investors with leveraged positions across asset classes are forced to liquidate broadly, creating artificial co-movement. The 2008 financial crisis provided a stark example, as asset classes from equities to credit to commodities to emerging markets all fell simultaneously—a 'correlation goes to one' event that devastated many multi-strategy hedge funds.
Formula
σ²_p = w₁²σ₁² + w₂²σ₂² + 2·w₁·w₂·σ₁·σ₂·ρ₁₂
Example
A portfolio manager holds two stocks, each with 30% annual volatility. If the correlation between them is 0.8, the two-stock portfolio has volatility of: σ_p = √(0.5² × 0.30² + 0.5² × 0.30² + 2 × 0.5 × 0.5 × 0.30 × 0.30 × 0.8) = √(0.0225 + 0.0225 + 0.018) = √0.063 ≈ 25.1%. If the manager instead selects a second stock with correlation of 0.2 to the first, portfolio volatility drops to: σ_p = √(0.0225 + 0.0225 + 2 × 0.5 × 0.5 × 0.30 × 0.30 × 0.2) = √(0.0225 + 0.0225 + 0.0045) = √0.0495 ≈ 22.2%. Selecting truly uncorrelated assets (ρ = 0) reduces volatility further to √0.045 ≈ 21.2%—equal to each stock's volatility divided by √2. This illustrates how lower correlations provide increasingly significant diversification benefits.
Related terms
Arbitrage Beta Breakdown Calmar Ratio Contagion Correlation Covariance Efficient Frontier Emerging Markets Event Driven Financial Crisis Five Factor Model