The rough volatility theory is a relatively new and promising approach to model the volatility of financial markets. The roughness arises from the fractal-like characteristics observed in stock price dynamics that are hard to capture with traditional models. The main idea behind rough volatility is that the variations in the stock prices (or returns) exhibit a self-similar roughness over different time scales.
In mathematical terms, rough volatility refers to models where the instantaneous volatility is driven by a fractional Brownian motion (fBm) with Hurst exponent $H < \frac{1}{2}$. The path of the fBm is rough (as opposed to regular Brownian motion with $H = \frac{1}{2}$), which implies that it has more jagged trajectories and exhibits strong persistence or anti-persistence behavior.
Let’s describe the rough volatility framework more formally. Consider a stochastic process St representing the stock price at time t. In the geometric Brownian motion representation, the price dynamics can be written under the risk-neutral measure as:
dSt = μStdt + σtStdWt,
where μ is the drift, σt is the instantaneous volatility, and Wt is a standard Brownian motion. The rough volatility hypothesis suggests that the log-volatility process ln σt is a function of a fBm:
ln σt = f(BtH),
where BtH is a fractional Brownian motion with Hurst exponent H, and f is a deterministic function. The fBm is defined as:
$$B^H_t = c_H \int_{-\infty}^t \frac{dB_u}{(t-u)^{H-\frac{1}{2}}},$$
where $c_H = \sqrt{2H}\cdot\sqrt{\Gamma(\frac{3}{2}-H)\Gamma(H+\frac{1}{2})\Gamma(2-2H)}/\Gamma(2H-1)$, dBu stands for regular Brownian increments, and Γ is the gamma function.
One way to apply this framework is to consider a volatility process that is related to the q-variation of the fBm. In this case, the volatility can be defined as:
$$\sigma_t = \exp\left(\int\limits_{t-h}^t k_H(\cdot,u)dW(u)\right),$$
where h is the memory parameter, $k_H(\cdot,u) = \frac{G_H}{(1 + (u-t)^2)^\gamma}$ is the kernel function, GH is a scaling constant, and $\gamma=\frac{3}{2}-H$.
Application in the modeling of financial markets:
Rough volatility models provide a novel perspective on the modeling of asset prices and their volatilities. Some benefits of rough volatility models and their applications include:
1. Better empirical fit: Rough volatility models have been shown to capture the scaling and roughness properties observed in historical data more accurately than classical models like Black-Scholes, Heston, or GARCH.
2. More efficient option pricing: By accounting for the observed roughness and correlation structure in the volatility process, rough volatility models lead to more realistic option prices, improving the risk-neutral pricing of derivatives.
3. Improved parameter estimation: Rough volatility models provide a convenient framework for filtering and estimating volatilities from observed data, utilizing techniques like Maximum Likelihood Estimation, Fourier-based estimation, or dynamic programming algorithms like the Kalman filter.
4. Better calibration: Fast and efficient volatility estimation techniques in rough volatility models help improve the calibration of option pricing models to market data, leading to reduced pricing and hedging errors.
In conclusion, rough volatility offers an innovative way to model the fractal-like behavior of asset prices and their implied volatilities. Researchers are actively exploring different aspects of rough volatility, including estimation techniques, numerical methods for option pricing, and extensions to other asset classes. The incorporation of rough volatility into financial models can bring significant improvements to risk management, option pricing, and forecasting in financial markets.