Demystifying diversification strategies by using portfolio optimization techniques

Abstract

In this dissertation, we study a series of diversification strategies in order to improve the understanding of the diversification of investments that was mathematically established by Harry Markowitz. Even though Modern Portfolio Theory considers a trade-off between generating high returns and lowering risks, investment processes inspired by the concept of diversification are generally only addressed with their diversification benefits. Therefore, we analyze the diversification strategies by using portfolio optimization techniques that primarily originated from the mean-variance framework, which considers returns as well as risks with equal importance, in order to more fully understand the quantifiable consequences of various diversification strategies. First, we investigate passive investing and performance benchmarking through analyzing the two most popular equity benchmark portfolios: the cap-weighted portfolio, and the equally-weighted portfolio. As conventional portfolio performance evaluations occur relative to benchmarks, the performance evaluation of the benchmark itself has never been a trivial issue. Thus, an alternative methodology for portfolio performance evaluation that can be conducted without peer information is proposed, and we find little or no evidence of either benchmark portfolio performing better than the average portfolio. In terms of performance benchmarking, however, equally-weighted portfolios exhibit more desirable properties than cap-weighted portfolios. Second, we examine the quantitative properties of asset allocation and asset classification. We derive and compare the closed form expressions for the portfolio performances of asset allocation and direct security selection, and we find that the majority of investors can benefit from employing asset allocation. Furthermore, our analysis indicates that the design of asset classes is a critical factor in determining the portfolio performance of employing asset allocation. Hence, we further test the two most widely used within-stock asset classification schemes, i.e. style and industry classifications, and find that the asset designs should not focus on diversification benefits only. Third, we discuss the viability of robo-advising, which was recently developed during the on-going expansion of financial technology (FinTech). Robo-advising attempts to lower the entry barrier to financial advising through utilizing automated but personalized algorithms, in order to attract investors with smaller accounts, who are ineligible to receive traditional financial advising services. We investigate the relationship between portfolio size and risk in order to examine the viability of robo-advisers in providing diversification benefits with limited portfolio sizes. The results indicate that a substantial investment is not necessary to gain diversification benefits.