eThesis - School of Business electronic theses

Time series momentum has been one of the most interesting topics in the recent asset pricing literature. This thesis contributes to this existing literature in three ways. First, we study single-asset time series predictability using a sample of 178 bond indices, commodity indices, equity indices, and foreign exchange rates from 1970 to 2015, and show that the previously documented single-asset time series predictability results are also present in this much larger data set. Second, we study cross-asset time series predictability using a broad, international sample of 57 bond indices, equity indices, and foreign exchange rates from 1980 to 2015, and find evidence for five distinct kinds of cross-asset predictability, three of which are new to the literature. Finally, we apply these results to the construction of new cross-asset time series momentum trading strategies, and show that these new strategies typically outperform the single-asset strategies previously considered in the literature.