Loading...
Publication
Análise do sentimento do investidor no mercado brasileiro: evidências empíricas por meio de notícias técnicas
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 1.37 MB | Adobe PDF |
Authors
Abstract(s)
O crescimento das notícias veiculadas por meios eletrónicos tem transformado o modo
como os investidores tomam decisões. O conteúdo e o tom dessas notícias impactam
diretamente o fluxo de caixa previsional das empresas, refletindo-se tanto no valor das ações
quanto no Ibovespa. Em consonância com os princípios das Finanças Comportamentais, tem se destacado uma área de estudos dedicada a analisar o sentimento do investidor e a
compreender o modo como os fatores comportamentais influenciam o mercado.
Nos últimos anos, diversos estudos têm sido realizados com o objetivo de mensurar e
compreender esse efeito, utilizando técnicas de machine learning ou abordagem baseada em
dicionários léxicos. Na literatura brasileira, é pouco analisado o poder preditivo dos notáveis
dicionários Harvard-IV, Loughran e McDonald e VADER na análise de sentimento do
investidor, aplicado a notícias dos media, principalmente avaliando o impacto de cada
dicionário com base em 3 (três) variáveis independentes.
Perante esta lacuna, o objetivo geral desta investigação foi analisar a relação entre o
sentimento do investidor, mensurado por meio da aplicação de 3 (três) dicionários léxicos a um
conjunto alargado de notícias financeiras, com o comportamento do mercado acionista
brasileiro. Foram analisadas 9.515 notícias do período de novembro de 2021 a agosto de 2024,
agrupadas em 697 observações diárias. A base das notícias, do tópico “Finanças”, do serviço
“Valor +News”, foi adquirida ao jornal Valor Econômico, com licenciamento específico para
fins académicos, tendo sido processadas na plataforma Google Colab por meio de um código
Python desenvolvido no âmbito do trabalho. Os dados foram modelados por meio da regressão
linear múltipla aplicada a séries temporais, e analisados usando o software Stata. A validade e
robustez do modelo foi confirmada com os testes referentes à estacionariedade,
heterocedasticidade, autocorrelação dos resíduos e da multicolinearidade das variáveis.
A extração do sentimento do investidor através dos dicionários Loughran e McDonald
e VADER revelam significância estatística positiva na explicação do comportamento do
Ibovespa, reforçando o valor informacional das notícias na formação do sentimento dos
investidores e, consequentemente, no mercado acionista brasileiro. As variáveis analisadas taxa
de câmbio, Bitcoin e risco Brasil mostraram-se estatisticamente significativas, enquanto o ouro
e a frequência diária de notícias não apresentaram qualquer relevância estatística.
Relativamente ao sentimento mensurado pelo dicionário Harvard-IV tornou-se significativo
apenas após o cálculo da primeira diferença, embora a variável risco Brasil tenha perdido a sua relevância. Em síntese, os resultados destacam a influência de fatores subjetivos na análise do comportamento dos investidores, em conformidade com os princípios das Finanças
Comportamentais.
The growth of electronic news has transformed the way investors make decisions. The content and tone of this news has a direct impact on the forecast cash flow of companies, reflected in both the value of shares and the Ibovespa. In line with the principles of Behavioral Finance, an area of study has emerged dedicated to analyzing investor sentiment and understanding how behavioral factors influence the market. In recent years, several studies have been carried out with the aim of measuring and understanding this effect, using machine learning techniques or approaches based on lexical dictionaries. In the Brazilian literature, there is little analysis of the predictive power of the notable Harvard-IV, Loughran and McDonald and VADER dictionaries in investor sentiment analysis, applied to news from the media, mainly evaluating the impact of each dictionary based on 3 (three) independent variables. Given this gap, the general objective of this research was to analyze the relationship between investor sentiment, measured by applying three (3) lexical dictionaries to a wide range of financial news, and the behavior of the Brazilian stock market. We analyzed 9,515 news items from November 2021 to August 2024, grouped into 697 daily observations. The news base, from the “Finance” topic of the “Valor +News” service, was acquired from the Valor Econômico newspaper, with specific licensing for academic purposes, and was processed on the Google Colab platform using Python code developed within the scope of the work. The data was modeled using multiple linear regression applied to time series, and analyzed using Stata software. The validity and robustness of the model was confirmed with tests for stationarity, heteroscedasticity, autocorrelation of the residuals and multicollinearity of the variables. The extraction of investor sentiment using the Loughran and McDonald and VADER dictionaries revealed positive statistical significance in explaining the behavior of the Ibovespa, reinforcing the informational value of news in shaping investor sentiment and, consequently, the Brazilian stock market. The analyzed variables exchange rate, Bitcoin and Brazil risk were statistically significant, while gold and the daily frequency of news were not statistically significant. The sentiment measured by the Harvard-IV dictionary became significant only after calculating the first difference, although the Brazil risk variable lost its relevance. In summary, the results highlight the influence of subjective factors in the analysis of investor behavior, in line with the principles of Behavioral Finance.
The growth of electronic news has transformed the way investors make decisions. The content and tone of this news has a direct impact on the forecast cash flow of companies, reflected in both the value of shares and the Ibovespa. In line with the principles of Behavioral Finance, an area of study has emerged dedicated to analyzing investor sentiment and understanding how behavioral factors influence the market. In recent years, several studies have been carried out with the aim of measuring and understanding this effect, using machine learning techniques or approaches based on lexical dictionaries. In the Brazilian literature, there is little analysis of the predictive power of the notable Harvard-IV, Loughran and McDonald and VADER dictionaries in investor sentiment analysis, applied to news from the media, mainly evaluating the impact of each dictionary based on 3 (three) independent variables. Given this gap, the general objective of this research was to analyze the relationship between investor sentiment, measured by applying three (3) lexical dictionaries to a wide range of financial news, and the behavior of the Brazilian stock market. We analyzed 9,515 news items from November 2021 to August 2024, grouped into 697 daily observations. The news base, from the “Finance” topic of the “Valor +News” service, was acquired from the Valor Econômico newspaper, with specific licensing for academic purposes, and was processed on the Google Colab platform using Python code developed within the scope of the work. The data was modeled using multiple linear regression applied to time series, and analyzed using Stata software. The validity and robustness of the model was confirmed with tests for stationarity, heteroscedasticity, autocorrelation of the residuals and multicollinearity of the variables. The extraction of investor sentiment using the Loughran and McDonald and VADER dictionaries revealed positive statistical significance in explaining the behavior of the Ibovespa, reinforcing the informational value of news in shaping investor sentiment and, consequently, the Brazilian stock market. The analyzed variables exchange rate, Bitcoin and Brazil risk were statistically significant, while gold and the daily frequency of news were not statistically significant. The sentiment measured by the Harvard-IV dictionary became significant only after calculating the first difference, although the Brazil risk variable lost its relevance. In summary, the results highlight the influence of subjective factors in the analysis of investor behavior, in line with the principles of Behavioral Finance.
Description
Keywords
Sentimento do investidor Abordagem léxica Notícias Investor sentiment lexical approach news