A influência do investimento estrangeiro no mercado acionário brasileiro: uma análise com técnicas estatísticas e de machine learning

Abstract

This study investigates the relationship between foreign investment flow on B3, the Ibovespa index (IBOV), and its constituent stocks, aiming to determine whether foreign flows can serve as leading indicators of movements in the Brazilian stock market. To this end, statistical techniques such as Pearson correlation, linear regression, and Granger causality are applied, along with machine learning methods including the K-Means algorithm and principal component analysis (PCA), using normalized time series. The data analyzed covers the period from August 19, 2024, to June 14, 2025, including daily foreign investment flow values provided by B3 and the closing prices of the IBOV and selected stocks retrieved from Yahoo Finance. Different time windows are considered to assess the robustness of the relationships and the presence of dynamic patterns. The results show that the larger the dataset, the lower the Pearson correlation between foreign flow and the IBOV, but the greater the evidence of Granger causality. The PCA and K-Means analysis revealed clusters with high foreign participation associated with high IBOV levels, suggesting a positive relationship between these variables. Regarding the IBOV constituent stocks, independence was observed in short-term windows, but stronger correlation emerged in the medium and long term, indicating structural cohesion of the index. However, the correlation between foreign flow and individual stocks was low, suggesting that the direct impact of foreign investors on specific stocks is limited.