Equal Risk Contribution (ERC)

ERC relies on the Euler decomposition theorem for homogeneous risk functions. Total portfolio volatility is perfectly decomposed into additive components: the product of each asset's weight and its Marginal Risk Contribution (MRC). The ERC objective forces these components to be equal across all assets.

• Objective: Minimize the dispersion of risk contributions.
• Constraint: The sum of risk contributions equals total portfolio volatility.
• Result: No single asset dominates the risk profile.

Finding ERC weights is a convex optimization problem when constrained to long-only positions. The objective minimizes the variance of risk contributions. Because the problem is convex, a global minimum is guaranteed, avoiding the local minima traps common in non-convex portfolio optimizations.

• Solver: Sequential quadratic programming (SQP) or interior-point methods.
• Alternative: Analytical solution exists for the two-asset case.
• Stability: More robust than mean-variance optimization to input estimation errors.

A critical distinction from Inverse Volatility Weighting is ERC's explicit use of the correlation matrix. Assets with high correlation to the rest of the portfolio receive lower weights because their diversification benefit is limited. Conversely, assets with low or negative correlation receive higher weights.

• Mechanism: The covariance matrix, not just the diagonal variances, drives allocation.
• Outcome: True diversification across independent risk sources.
• Metric: Maximizes the Effective Number of Bets (ENB).

ERC weights are highly sensitive to the quality of the input covariance matrix. Using a naive sample covariance matrix leads to unstable, noisy weights. Production implementations rely on robust estimators like Covariance Shrinkage (Ledoit-Wolf) or Exponentially Weighted Moving Average (EWMA) to improve out-of-sample performance.

• Shrinkage Target: Constant correlation or single-factor model.
• EWMA Lambda: Typically 0.94 (RiskMetrics standard) for monthly data.
• Rebalancing Frequency: Directly tied to the covariance estimation window.

While both target diversification, ERC and the Maximum Diversification Ratio (MDR) portfolio differ fundamentally. MDR maximizes the ratio of weighted-average asset volatility to portfolio volatility. ERC equalizes risk contributions. MDR often results in highly concentrated portfolios, whereas ERC guarantees a minimum weight for all assets.

• ERC: Balances risk contributions; no asset is dropped.
• MDR: Maximizes the diversification ratio; can assign zero weights.
• Use Case: ERC is preferred when exposure to all assets is mandated.

A tail-risk extension of ERC shifts the risk measure from volatility to Expected Shortfall (CVaR) or maximum drawdown. Drawdown Parity equalizes the contribution of each asset to the portfolio's peak-to-trough decline. This addresses the criticism that volatility treats upside and downside movements symmetrically.

• Risk Measure: Conditional Value-at-Risk (CVaR) at a 95% or 99% confidence level.
• Optimization: Non-convex for drawdown; requires heuristic or scenario-based approaches.
• Benefit: Directly controls the pain point for loss-averse investors.