Modeling daily realized futures volatility with singular spectrum analysis

Using singular spectrum analysis (SSA), we model the realized volatility and logarithmic standard deviations of two important futures return series. The realized volatility and logarithmic standard deviations are constructed following the methodology of Andersen et al. [J. Am. Stat. Ass. 96 (2001) 42–55] using intra-day transaction data. We find that SSA decomposes the volatility series quite well and effectively captures both the market trend (accounting for about 34–38% of the total variance in the series) and, more importantly, a number of underlying market periodicities. Reliable identification of any periodicities is extremely important for options pricing and risk management and we believe that SSA can be a useful addition to the financial practitioners’ toolbox.