Seleção e otimização de carteiras de ações do mercado brasileiro: Momentum, Markowitz e algoritmo genético.

Abstract

In this study, stocks from the Brazilian stock exchange were selected, using the momentum strategy initially developed by Jegadeesh and Titman, which consists of selecting winning stocks and selling losing stocks, from which four portfolios were built, an optimized portfolio using Algorithm Genetic (AG), one of them naive and the other two built on the theory of Markowitz portfolios, one of them with maximum Sharpe and the other with minimum variance. The objective of this study is to verify whether the performance of a momentum stock portfolio optimized through the Genetic Algorithm (GA) outperforms the market index (the IBOVESPA), a portfolio of momentum stocks with naive diversification, and a portfolio of momentum stocks optimized by the Markowitz model. Finally, it will be verified whether the GA optimization technique can generate excess risk-adjusted return superior to naive diversification and superior to the portfolio formed from the Markowitz theory, in addition to verifying if this portfolio outperforms the market. To verify if the results are satisfactory, some comparative metrics will be used, such as annualized return, Sharpe ratio, and to verify if these results are statistically significant, the statistical test of Ledoit and Wolf (2008) will be used. The results found revealed that the moment portfolios perform above the market in the first three quarters after the formation of the portfolio, after this period no statistical significance was found in the results. Regarding the methodologies of AG and Markowitz for calculating the weights of the portfolios, the results show that in the first two quarters the best portfolio was that of AG, however, after two quarters there is an inversion in the performance of the portfolios, in which the Markowitz portfolios start to outperform the portfolio created using genetic algorithm and the naive portfolio.