This article evaluates how alternative models of large-dimensional conditional covariance translate into the performance of out-of-sample portfolios in the Brazilian stock market. Using daily data from B3 for the period 2000–2024 and a rolling window structure, we estimated conditional covariance matrices and mapped them into weights of minimum and mean-variance portfolio strategies, as well as specifications with turnover and positivity constraints on the weights. Performance was evaluated using annualized return, standard deviation, information ratio and turnover. The results highlight the impact of transaction costs on valuation metrics, reinforcing the importance of strategies that favor a combination of stable covariance between the assets with low portfolio turnover.
Comissão Organizadora
Comissão Científica
Ferramenta completa de submissão, avaliação e publicação de anais para eventos científicos.
