Volatility long memory is a stylized fact that has been documented for a long time. Existing literature have two ways to model volatility long memory: component volatility models and fractionally integrated volatility models. This paper develops a new fractionally integrated GARCH model, and investigates its performance by using the Standard and Poor’s 500 index returns and cross-sectional European option data. The fractionally integrated GARCH model signi?cantly outperforms the simple GARCH(1, 1) model by generating 37% less option pricing errors. With stronger volatility persistence, it also dominates a component volatility model, who has enjoyed a reputation for its outstanding option pricing performance, by generating 15% less option pricing errors. We also con?rm the fractionally integrated GARCH model’s robustness with the latest option prices. This paper indicates that capturing volatility persistence represents a very promising direction for future study.

In this paper, we have examined the long memory property of Indian stock market by analyzing the trading volume series. Given the absence of trading volume index data, we have constructed trading volume series for the Indian stock market. We used maximum likelihood method to analyze the constructed trading volume index. The estimation of ARFIMA model, obtained a signicant parameter for the order of fractional integration, and this could be consistent with the long autocorrelations observed in the trading volume series. The ndings that stock trading volume is a long memory process is robust, given dierent estimating methods, different subsamples, temporal aggregation and tests on individual stocks. Because of the conditional heteroscedasticity in the series, we have also carried out ARFIMAGARCH procedures to check whether long persistence were robust in the presence of conditional heteroscedasticity.