An empirical application of a two-factor model of stochastic volatility

This contribution focuses on the modelling of volatility of returns in Czech and US stock markets using a two-factor stochastic volatility model, i.e. the volatility process is modeled as a superposition of two autoregressive processes. As the volatility is not observable, the logarithm of the daily range is employed as the proxy. The estimation of parameters and volatility extraction are performed using the Kalman filter. We have obtained a meaningful decomposition of the volatility process into one highly persistent factor and another quickly mean-reverting factor. Moreover, we have shown that although the overall level of the volatility of returns is roughly the same in both markets, the US market exhibits substantially lower volatility of the volatility process.