Stochastic conditonal range, a latent variable model for financial volatility

In this paper we introduce a parameter driven model for the dynamics of range, the stochastic conditional range (SCR). We propose to estimate its parameters by Kalman filter, importance sampling and simulated maximum likelihood depending on the hypotheses on the distributional form of the innovations. The model is applied to a large subset of the S&P 500 components. A comparison with of its fitting and forecasting abilities with the CARR model shows that the new approach can provide an interesting alternative.