| Abstract: |
Abstract This thesis examines the use of differential evolution in a real-world portfolio op- timization task based on US stock data. We empirically test the capability of the algorithm to find an inter-sector allocation that outperforms a broad-market stock index. Two constrained sector ETF portfolios are constructed to simulate realis- tic agent-based settings and performance of the competing portfolios is analyzed in terms of both return and risk. The results are further extended to include Markowitz’ global minimum variance portfolio and a naive 1/N portfolio. We show that the con- structed portfolios are indeed capable of outperforming the market whilst simultane- ously maintaining lower tail risk, however, the performance significantly deteriorates if the portfolios are rebalanced based on rolling data windows. Overall the algorithm delivers satisfying results while providing the user with a relative freedom when choosing portfolio constraints. JEL Classification: C61, G11, G17, G19 Keywords: portfolio optimization, exchange-traded funds, differen- tial evolution, empirical analysis Title: Sector ETF Portfolio Optimization Using Differential Evolution Author’s e-mail: rene.holesinsky@gmail.com Supervisor’s e-mail: frantisek.cech@fsv.cuni.cz 1 |
