Three essays on asset management: Index tracking, conditional volatility, and switching return momentum.

摘要

This dissertation studies three aspects of asset management: Dynamic index tracking, conditional volatility, and return momentum under regime switching. The first study is motivated by the increasing popularity of cointegration-based statistical methods in financial applications. We explore the practical use of cointegration-based index tracking using sector ETFs. Methodologically, this analysis adds a new dimension to current cointegration-based tracking strategy: sector asset class. The empirical results show this strategy can make sector over- and under-bets to generate consistent positive return in excess of the underlying index, thus can be used as a simple statistical sector rotation model. In the second paper, we measure the conditional covariance in four broadly defined US equity style markets. Estimating the covariance or correlation matrices of multiple sets of financial data has been an active research issue. There exist a variety of models to generate conditional covariance matrix. Determining what is the best model is indeed an empirical and an implementation issue. We adopt five commonly used forecasting methods: the simple moving average; the exponentially weighted moving average; the constant conditional correlation multivariate GARCH; the dynamic conditional correlation multivariate GARCH; and the orthogonal GARCH model. To best evaluate the performance of the different forecasting models out-of-sample, we apply both standard statistical criteria and economic measures such as Value at Risk at the portfolio level. We find that empirically there does not exist a single model which dominates irrespective of the forecasting horizon. The third study seeks the multifactor explanation of momentum profit in the presence of Markov regime switching. We present a regime-dependent version of Fama-French three-factor model and find evidence that different regimes do exist. Particularly, in one of the two regimes, short-term return continuation is supported by size and value risk factors. The estimated regime probability is further related to macroeconomic explanatory variables in order to find the driving force behind regime switch. Both NBER recession and default spread are found to strongly correlated to regime probability.