Os impactos de notícias na performance das ações de empresas de capital aberto no Brasil

Abstract

The financial market is known for its dynamism, with newly published information being transmitted almost instantly to the prices of assets traded on the stock exchange. With the growth of digital publications, monitoring news that may influence the performance of publicly traded stocks has become a highly relevant field of study within financial markets. Thus, the aim of this study is to examine the existence of a causal relationship between the current sentiment expressed by financial news portals and the future performance of stocks in the market, through the application of a BERT-based model for sentiment classification of online publications. The findings indicate an absence of causality between the two variables investigated, as five out of seven stock performance predictions did not yield positive returns. However, due to the structure of the data-processing algorithm, the correlation procedures adopted, and the generation of investment allocation recommendations, this research contributes to future academic work in the field by suggesting several avenues for achieving improved results and for sugesting other potential causal relationships among the variables involved in the problem. Furthermore, the validation of the FinBERT model for sentiment analysis, as well as the identification of words most associated with positive and negative sentiments, may support subsequent studies and benefit professionals seeking applications in corporate communication.