Modelling and valuing multivariate interdependencies in financial time series

This thesis investigates implications of interdependence between stock market prices inthe context of several financial applications including: portfolio selection, tests of marketefficiency and measuring the extent of integration among national stock markets.In Chapter 2, I note that volatility spillovers (transmissions of risk) have been found innumerous empirical studies but that no one, to my knowledge, has evaluated their effectsin the general portfolio framework. I dynamically forecast two multivariate GARCHmodels, one that accounts for volatility spillovers and one that does not, and constructoptimal mean-variance portfolios using these two alternative models. I show thataccounting for volatility spillovers lowers portfolio risk with statistical significance andthat risk-averse investors would prefer realised returns from portfolios based on thevolatility spillover model.In Chapter 3, I develop a structural MGARCH model that parsimoniously specifies theconditional covariance matrix and provides an identification framework. Using the modelto investigate interdependencies between size-sorted portfolios from the Australian StockExchange, I gain new insights into the issue of asymmetric dependence. My findings notonly confirm the observation that small stocks partially adjust to market-wide newsembedded in the returns to large firms but also present evidence that suggests that smallfirms in Australia fail to even partially adjust (with statistical significance) to large firms?shocks contemporaneously. All adjustments in small capitalisation stocks occur with alag.Chapter 4 uses intra-daily data and develops a new method for measuring the extent ofstock market integration that takes into account non-instantaneous adjustments toovernight news. This approach establishes the amounts of time that the New York, Tokyoand London stock markets take to fully adjust to overnight news and then uses thisThis thesis investigates implications of interdependence between stock market prices inthe context of several financial applications including: portfolio selection, tests of marketefficiency and measuring the extent of integration among national stock markets.In Chapter 2, I note that volatility spillovers (transmissions of risk) have been found innumerous empirical studies but that no one, to my knowledge, has evaluated their effectsin the general portfolio framework. I dynamically forecast two multivariate GARCHmodels, one that accounts for volatility spillovers and one that does not, and constructoptimal mean-variance portfolios using these two alternative models. I show thataccounting for volatility spillovers lowers portfolio risk with statistical significance andthat risk-averse investors would prefer realised returns from portfolios based on thevolatility spillover model.In Chapter 3, I develop a structural MGARCH model that parsimoniously specifies theconditional covariance matrix and provides an identification framework. Using the modelto investigate interdependencies between size-sorted portfolios from the Australian StockExchange, I gain new insights into the issue of asymmetric dependence. My findings notonly confirm the observation that small stocks partially adjust to market-wide newsembedded in the returns to large firms but also present evidence that suggests that smallfirms in Australia fail to even partially adjust (with statistical significance) to large firms?shocks contemporaneously. All adjustments in small capitalisation stocks occur with alag.Chapter 4 uses intra-daily data and develops a new method for measuring the extent ofstock market integration that takes into account non-instantaneous adjustments toovernight news. This approach establishes the amounts of time that the New York, Tokyoand London stock markets take to fully adjust to overnight news and then uses this