Risk-factor diversification has attracted a great deal of attention since the 2008 Global Financial Crisis. The academic research in this field, however, has centred on analysing investment performance with little research in other areas. Motivated by industry interest and a paucity of academic research in the field, this thesis examines the behaviour of risk-factor diversification. Specifically, it examines performance, return correlation, and return predictability. This thesis makes a number of original contributions to academia and the practical management of investment portfolios. The first empirical chapter examines the performance of risk-factor diversification across a number of dimensions. Even when limited to U.S. stock and bond risk factors, the evidence indicates that risk-factor diversification offers the potential to outperform asset-class diversification. However, unlike previous studies, this thesis does not find that risk-factor diversification is universally superior to asset-class diversification across all market environments and time periods. The second empirical chapter explores the exceedance correlation between risk-factor returns and reveals three distinguishing empirical facts: exceedance correlations are generally symmetric; where asymmetries exist, positive asymmetries are as common as negative; and normality is rejected for both positive and negative exceedance correlations.