Momentos superiores e correlações dinâmicas: uma otimização de portfólio aplicada ao mercado brasileiro de ações

Abstract

Portfolio theory became established in the financial literature following the formulation of the Mean-Variance Model proposed by Harry Markowitz in 1952. Despite its relevance, the approach presents limitations by restricting itself to the parameters of mean and variance and by assuming static correlations among assets. The present study analyzed the application of an Mean-Variance-Skewness-Kurtosis model calibrated with dynamic correlations via Dynamic Conditional Correlation GARCH and examined its effectiveness as a methodological alternative to the classical framework. The dataset consisted of simple daily returns of stocks included in the IBrX 50 Index from January 2020 to May 2025, filtered by the Novo Mercado listing segment. A walk-forward approach was used to assess the temporal consistency of the methodology, in addition to regular rebalancing to evaluate the model’s adaptability to new data. The results indicate competitive advantages in adopting the proposed methodology compared to classical optimization: the portfolio generated a higher cumulative return (104.50% versus 62.36%), superior risk-adjusted performance measured by the information ratio (5.78% versus 1.58%), and lower asset concentration according to the HHI (0.0649 versus 0.2239), as well as improved adaptability in response to the inclusion of new information. For future research, it is recommended that financial analysis of the companies be incorporated into the asset-filtering process, along with an assessment of the implementation costs associated with the modeling.