Synthetic Diversification, Smart Randomization, and Commodity Indexing

Abstract

This thesis investigates the use of randomization in asset allocation, and introduces a dynamic commodity index. Randomizing asset holdings can lead to extra rebalancing gains, and lower inter-asset correlations. However, the gains are insignificant in practice. Momentum- and correlation-based smart randomization strategies can improve the performance and provide a promising basis for future research.

The second part of this thesis introduces a regime-based dynamically weighted commodity index. In this index, the commodity weights are determined by an optimization model that employs an underlying hidden Markov model. Including regimes in the allocation decision leads to vast improvements in index performance. Several extensions of the regime-based allocation model are introduced in the last chapter of this thesis.