H.S. Ng and K.P. Lam (PRC)
Financial Forecasting, Range-based Volatility
Range-based volatility is one of the major streams in the volatility study. Garman and Klass [1] proposed an efficient volatility estimator based on the extreme values, open and close. Conventionally, the estimator can be obtained at the market close or can be predicted with a limited accuracy by its lagged values. Nowaday, the latest and cumulative in traday data, which embeds a rich informational content, is ignored in the prediction model while it definitely enhances the accuracy. In this paper, we introduce an incremental in traday prediction problem to explore a unidirectional causal relationship between the volatility estimator and the incre mental information. Generally, the causal effect due to the incremental information can reflect through a proper selec tion of the incremental information content and should in crease as the time instant on the same trading day. Two dif ferent incremental information contents as exogenous input of a linear model are selected to demonstrate these features. In addition to the selection of exogenous input, we believe that the selection of model also affects the causal effect. We examine the relationship using a linear and a nonlinear neural networks. Empirical results show that the prediction accuracy of nonlinear model is better than that of linear model. It implies that the nonlinear model can reflect the causal effect in a better way.
Important Links:
Go Back
