How to model the implied volatility surface accurately and obtain efficient predictions? Evidence from index options.

In this paper we show applicability of the 7-factor implied volatility surface specification, proposed by Chalamandaris and Tsekrekos (2011), on index options. Moreover, we claim that any previously proposed parametric specification contains heteroskedasticity and autocorrelation in the fitting residuals, which results in inefficient volatility estimations. We propose a Residual Correction Model that enables us to reduce the heteroskedasticity and autocorrelation and is superior in both in- and out-of-sample performances over alternative implied volatility surface specifications. We find statistical and economical evidence of the relevance of both the 7-factor specification and Residual Correction Model. Moreover, we show that a combination of the two captures valuable information regarding describing and predicting index implied volatility surfaces, as it produces accurate volatility estimates and efficient surface predictions, and enables us to support profitable trading strategies in the absence of transaction costs. Furthermore, we show that making use of error-correction models and restricted dynamic factor State Space Models help us to obtain more accurate option volatility predictions than alternative time series models.