TL;DR
This paper introduces a novel semi-parametric approach to estimating hedge ratios that better accounts for the non-normal distribution of financial asset returns. By leveraging the Cornish-Fisher expansion, this method offers improved risk reduction in terms of Value at Risk (VaR) and Conditional Value at Risk (CVaR) compared to traditional minimum-variance hedging and a previously proposed non-parametric method.
What Problem Does This Paper Address?
Traditional financial hedging strategies often rely on minimizing portfolio variance. However, this approach assumes that asset returns are normally distributed or that investors have quadratic utility, assumptions that rarely hold in practice. When returns are non-normal, minimizing variance can paradoxically increase other risk measures like negative skewness and kurtosis. This means that a portfolio optimized for low variance might still be risky when evaluated by metrics like Value at Risk (VaR) and Conditional Value at Risk (CVaR), which are more sensitive to extreme events and the shape of the return distribution's tails. The paper addresses this by developing a hedging strategy that explicitly considers these higher moments of the return distribution, aiming to provide a more robust reduction in VaR and CVaR.
Methods
The core innovation of this paper is the development of a semi-parametric method for estimating minimum-VaR and minimum-CVaR hedge ratios. This method is built upon the Cornish-Fisher expansion.
Here's a breakdown:
- Cornish-Fisher Expansion: This is a mathematical tool used to approximate the quantiles (like those used in VaR calculations) of a probability distribution. It starts with the quantile of a standard normal distribution and then adjusts it based on the higher moments (skewness and kurtosis) of the actual distribution.
- Analytical Derivation: The authors use the Cornish-Fisher expansion to derive analytical expressions for the first-order conditions for minimizing VaR and CVaR.
- Numerical Solution: These derived conditions are then solved numerically using sample estimates of variance, skewness, and kurtosis from historical data.
This approach is "semi-parametric" because it uses a parametric tool (the Cornish-Fisher expansion) but estimates its parameters (moments) from non-parametric historical data. This contrasts with:
- Minimum-Variance Hedging: A standard approach that only considers the second moment (variance).
- Non-Parametric Historical Simulation (Harris and Shen, 2006): This method directly uses historical data to estimate VaR and CVaR without assuming a specific distribution shape, but it relies heavily on large samples to accurately capture tail events.
Datasets
The study uses daily data for the period 25/02/1994 to 19/08/2008. The datasets consist of:
- Four UK Equity Indices (Spot Prices): FTSE 100, FTSE 250, FTSE Small Cap, and FTSE All Share.
- Two Futures Contracts: FTSE 100 futures and FTSE 250 futures.
The data was obtained from DataStream. The authors use log returns for estimating hedge ratios and simple returns for calculating hedge portfolio returns.
Key Insights and Ideas
- Beyond Variance: The paper strongly reinforces the idea that minimizing portfolio variance is insufficient for robust risk management when returns are non-normal. Higher moments (skewness and kurtosis) significantly impact tail risk measures like VaR and CVaR.
- Cornish-Fisher as a Bridge: The Cornish-Fisher expansion proves to be an effective bridge between parametric assumptions (like normality) and non-parametric data-driven approaches. It allows for the incorporation of higher moments without requiring a full parametric model.
- Efficiency with Shorter Samples: A key advantage of the semi-parametric approach is its ability to achieve good results with shorter historical data samples compared to purely non-parametric methods. This is because it leverages the structure provided by the Cornish-Fisher expansion rather than relying solely on the rare events in the extreme tails of the data.
- Practical Implications: The method's ability to reduce both skewness and kurtosis, leading to lower VaR and CVaR, makes it a more attractive tool for practitioners seeking to manage downside risk effectively.
Key Results
The paper compares three hedging strategies: minimum-variance, non-parametric minimum-VaR/CVaR, and the new semi-parametric minimum-VaR/CVaR.
- Semi-Parametric Superiority: The semi-parametric approach consistently outperforms both the minimum-variance and the non-parametric methods in reducing negative skewness and excess kurtosis.
- VaR and CVaR Reduction: Consequently, the semi-parametric method generally leads to greater reductions in VaR and CVaR.
- For VaR, the average reduction was 45.58% with the semi-parametric approach, compared to 40.06% for minimum-variance and 42.69% for the non-parametric method.
- For CVaR, the average reduction was 45.31% with the semi-parametric approach, compared to 42.05% for minimum-variance and 41.79% for the non-parametric method.
- Efficiency: The semi-parametric approach works well even with relatively short historical data samples, unlike the non-parametric method which struggled with its small sample size.
- Minimum-Variance Drawbacks: While minimum-variance hedging effectively reduces standard deviation (average reduction of 45.00%), it can sometimes increase skewness and kurtosis, leading to less favorable VaR and CVaR reductions in certain cases.
Limitations and Future Work
- Time-Varying Volatility: The current semi-parametric approach assumes constant volatility. The authors suggest that a natural extension would be to incorporate time-varying volatility, perhaps by adapting Exponentially Weighted Moving Average (EWMA) or GARCH-type models, similar to how it's done for long-only portfolios. This could lead to further improvements in VaR and CVaR reduction.
- Data Dependence: While less dependent than purely non-parametric methods, the semi-parametric approach still relies on historical data to estimate moments. The accuracy of these estimates is crucial.
- Model Adaptability: The paper focuses on equity index futures. Further research could explore the applicability and performance of this method across different asset classes and hedging instruments.
Summary updated