Abstract
This article investigates volatility transmission process between the US, the UK and Japanese stock index futures markets. Most importantly, we examine that whether structural changes have effect on volatility transmission process. We use Iterated Cumulative Sums of Squares (ICSS) algorithm proposed by Inclan and Tiao (Citation1994) to identify time points of structural changes exiting in the financial time series. Our results show that there is no common structural change in variances for three futures returns. This implies that diversification across stock index futures markets is possible. We find that volatility in three stock index futures markets are directly affected by its own lagged volatility. There are asymmetric volatility transmission effects between Japan and the UK and Japan and the US. In addition, there are bidirectional cross market volatility transmission between the UK and the US. However, this relation does not hold after controlling for structural changes in the bivariate Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model. We find that the measure of volatility transmission differs in intensity from that otherwise estimated. These findings support that structural changes in variance and GARCH model misspecification influence information flow and hence the scheme of transmission.
Notes
1 Kawaller et al. (Citation1987) point out that the lead from stock index futures to cash prices extends for 20 to 45 min. Further, Stoll and Whaley (Citation1990) note that stock index futures returns tend to lead stock market returns by about 5 min on average, after controlling for infrequent trading and bid/ask spreads of the component stocks. Chan (Citation1992) finds that futures market is the dominant source of market-wide information, therefore futures markets usually lead their corresponding cash markets.
2 Many studies also apply the ICSS algorithm to investigate volatility changes. For example, Huang and Yang (Citation2001) use the algorithm to examine the impact of settlement time changes on the volatility in the Shanghai and Shenzen Stock Exchange. Chelley-Steeley and Tsorakidis (Citation2009) use the algorithm to identify dates on which there were significant changes to the volatility of Greek drachma.
3 We identify the best-fitting specification of conditional-mean equation by Box–Jenkins techniques for S&P 500 and FTSE 100 futures. The partial autocorrelation function suggests that the Autoregressive Moving Average (ARMA) (||1, 7||, 0) model would be appropriate for S&P 500 futures return series, and ARMA(||1||, 0) model would be appropriate for FTSE 100 futures return series. The statistics are shown in .