Inverse Volatility Weighting

The weight of asset i is calculated as 1 / σ_i divided by the sum of the inverses of all asset volatilities. This ensures that assets with lower historical volatility receive a higher capital allocation, and vice versa.

• Formula: w_i = (1/σ_i) / Σ (1/σ_j)
• A 10% vol asset gets twice the weight of a 20% vol asset
• The sum of all weights always equals 1 (fully invested)

The fundamental flaw of this heuristic is that it implicitly assumes all pairwise correlations are zero. It ignores the diversification benefits of negatively correlated assets and the risk concentration of highly correlated ones.

• Two assets with 10% vol and 0.99 correlation get the same weight as two with -0.99 correlation
• Fails to penalize cluster risk in similar asset classes
• True risk parity requires inverting the full covariance matrix

Unlike true risk parity which requires convex optimization or matrix inversion, inverse volatility weighting is a closed-form calculation. This makes it trivial to implement in spreadsheets or low-latency execution systems.

• No covariance matrix inversion required
• No convergence checks or iterative solvers
• Computation scales linearly with the number of assets: O(n)

The weights are entirely dependent on the lookback window used to estimate historical volatility. A short window makes weights jumpy; a long window makes them slow to react to regime changes.

• Using a 20-day window creates high turnover and transaction costs
• Using a 200-day window creates stale weights during volatility spikes
• Common practice: Exponentially Weighted Moving Average (EWMA) with a decay factor of 0.94

Inverse volatility weighting is a special case of Equal Risk Contribution only when all asset correlations are identical. In real markets with heterogeneous correlations, the two methods produce divergent portfolios.

• ERC equalizes marginal risk contributions, not just inverse vols
• Inverse vol weighting over-allocates to low-vol assets in correlated clusters
• The divergence grows with the dispersion of pairwise correlations

Despite its theoretical limitations, inverse volatility weighting is widely used as a baseline benchmark and in contexts where simplicity and transparency are paramount.

• Futures strategies: Scaling position sizes by contract volatility to equalize dollar risk
• Multi-asset ETFs: A transparent, rules-based alternative to black-box optimizers
• Risk budgeting floors: Setting minimum allocation constraints before running full optimization

The parent framework that Inverse Volatility Weighting approximates. True risk parity allocates capital so each asset contributes equally to total portfolio volatility, not just individual volatility. This requires accounting for cross-asset correlations using a covariance matrix. Inverse volatility is a special case of risk parity only when all correlations are zero.

The mathematical formalization of risk parity. ERC uses convex optimization to find the portfolio weights where the marginal risk contribution of every asset is identical. Unlike inverse volatility weighting, ERC solves for weights that satisfy:

• Equalizing the product of weight and marginal volatility
• Accounting for the full covariance structure
• Avoiding concentration in highly correlated assets

A machine learning approach that bypasses the need to invert the covariance matrix. HRP uses hierarchical clustering on the correlation matrix to build a tree of similar assets, then allocates capital from the leaves up using inverse variance weighting within each cluster. This avoids the estimation errors that plague inverse volatility when the covariance matrix is ill-conditioned.

A statistical remedy for the primary weakness of inverse volatility weighting: estimation error. Shrinkage blends the noisy sample covariance matrix with a structured target (like constant correlation) to produce a more robust estimator. Key techniques include:

• Ledoit-Wolf shrinkage toward a single-factor model
• Oracle approximating shrinkage for large dimensions
• Reduces the impact of extreme historical observations on weights

A dynamic mechanism that scales portfolio exposure to maintain a constant ex-ante volatility level. While inverse volatility weighting sets relative weights, volatility targeting determines absolute leverage. The combination creates a portfolio where:

• Relative risk is balanced across assets
• Absolute risk is stabilized through time
• Leverage adjusts automatically as market calm or turbulence shifts

A time-series model that allows correlations to evolve over time, addressing the static nature of inverse volatility weighting. DCC estimates time-varying correlation matrices that capture:

• Correlation breakdowns during crises
• Regime shifts in asset relationships
• Gradual decoupling or convergence trends Risk parity weights updated with DCC adapt to changing market structure rather than relying on long-term averages.