Stochastic Volatility Models: Heston and SABR in Practice

The Black-Scholes-Merton model, with its assumption of constant volatility, provides a foundational framework for option pricing. However, the empirical reality of financial markets reveals that implied volatility is not constant across strike prices and maturities, giving rise to the well-known "volatility smile" or "skew." To address this, quantitative analysts have developed a class of models known as stochastic volatility models, which treat volatility itself as a random process. Among the most prominent of these are the Heston model and the SABR model.

The Heston Model: A Mean-Reverting Approach

The Heston model, introduced by Steven Heston in 1993, is a two-factor model that describes the evolution of an asset price and its variance. It is particularly popular because it allows for a semi-analytical solution for the price of a European option. The model is defined by the following system of stochastic differential equations (SDEs):

[ dS_t = \mu S_t dt + \sqrt{v_t} S_t dW_t^S ] [ dv_t = \kappa (\theta - v_t) dt + \xi \sqrt{v_t} dW_t^v ]

where:

• (S_t): The asset price at time (t).
• (v_t): The variance of the asset price at time (t).
• (\mu): The risk-neutral drift of the asset price.
• (\kappa): The rate of mean reversion of the variance.
• (\theta): The long-run mean of the variance.
• (\xi): The volatility of the variance (often called the "vol of vol").
• (dW_t^S) and (dW_t^v): Wiener processes with correlation (\rho).

The correlation parameter (\rho) is important for generating the volatility skew. A negative correlation between the asset price and its volatility ((\rho < 0)) means that as the asset price falls, its volatility tends to rise, which is consistent with the leverage effect observed in equity markets.

The SABR Model: A Popular Tool for Interest Rate Derivatives

The SABR model, which stands for "Stochastic Alpha, Beta, Rho," was developed by Hagan, Kumar, Lesniewski, and Woodward (2002). It has become an industry standard, particularly in the interest rate derivatives market, for its ability to capture the volatility smile. The SABR model describes the evolution of a forward price and its volatility:

[ dF_t = \alpha_t F_t^\beta dW_t^F ] [ d\alpha_t = \nu \alpha_t dW_t^\alpha ]

where:

• (F_t): The forward price at time (t).
• (\alpha_t): The stochastic volatility at time (t).
• (\beta): The exponent, which determines the backbone of the volatility smile. (\beta=1) corresponds to a lognormal model, while (\beta=0) corresponds to a normal model.
• (\nu): The volatility of the stochastic volatility.
• (dW_t^F) and (dW_t^\alpha): Wiener processes with correlation (\rho).

The SABR model does not have a closed-form solution for option prices. However, Hagan et al. derived a highly accurate asymptotic approximation for the implied volatility, which has made the model extremely practical for real-time applications.

Calibration and Implementation Challenges

Both the Heston and SABR models require calibration to market data, which involves finding the set of model parameters that best reproduces the observed prices of options. This is typically done by minimizing the sum of squared differences between the model-implied volatilities and the market-implied volatilities.

• Heston Model Calibration: Calibrating the Heston model can be computationally intensive, as it involves a multi-dimensional optimization problem. The choice of the objective function and the optimization algorithm can significantly impact the results. Furthermore, the calibration can be unstable, with different parameter sets producing similar fits to the market data.

• SABR Model Calibration: The SABR model is generally easier to calibrate, thanks to the analytical approximation for the implied volatility. It is often calibrated on a slice-by-slice basis, meaning that a separate set of parameters is found for each option expiry.

A Numerical Example: Implied Volatility Smile

To illustrate how these models can generate a volatility smile, consider the following table of market-implied volatilities for options on a particular stock with a 3-month expiry:

This U-shaped pattern is the classic volatility smile. A calibrated Heston or SABR model would be able to reproduce this smile by adjusting its parameters. For example, in the Heston model, a negative correlation (\rho) would contribute to the downward-sloping part of the smile (the skew), while the volatility of volatility (\xi) would contribute to the curvature of the smile.

Suppose we calibrate a SABR model to this data and obtain the following parameters:

Using these parameters in the SABR implied volatility formula, we could then price options at any strike, not just the ones observed in the market, in a way that is consistent with the observed smile.

Conclusion

Stochastic volatility models like Heston and SABR represent a significant advancement over the constant-volatility framework of Black-Scholes. They provide a more realistic description of market behavior and are indispensable tools for pricing and hedging derivatives in markets with a pronounced volatility smile. While their implementation and calibration can be challenging, the rewards in terms of improved accuracy and risk management are substantial. For the quantitative professional, a deep understanding of these models is not just an asset, but a necessity.