Comunicação política no Facebook e previsão eleitoral - Análise de big data da eleição presidencial brasileira de 2018 no Brasil: Big data analysis of the 2018 Brazilian presidential election Brazil

• Autor(a) principal: Magalhães Firmino, Leonardo
• Data de Publicação: 2019
• Outros Autores: Murta, Felipe
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Revista Lumina
• Texto Completo: https://periodicos.ufjf.br/index.php/lumina/article/view/28589
• topic: comunicação política; previsão eleitoral; opinião pública; mídias sociais; big data

This article aims to understand how the quantitative analysis of political communication in social media, through indicators of engagement to the political speech, is able to predict the electoral outcome. This study is a big data analysis of the first round of the 2018 presidential election in Brazil. Specifically, around 10,000 posts were collected from official Facebook campaign pages between June 1 and October 7, 2018. Regarding voting intent, a series of daily representative of the Brazilian population opinion polls were conducted for the same period. In order to understand whether there is a causal relationship between candidate engagement on Facebook and voting intent, predictive analysis was performed by testing two empirical approaches: one with aggregated data and the other with 90 multiple linear regression prediction models. We conclude the analysis by comparing the actual election results with all predictive models. The results showed that both the aggregate and regressive approaches demonstrate that the candidate engagement rate on Facebook is a good electoral predictor. The results reinforce the theories that defend the relevance of data coming from political behaviour in social media as good electoral predictor.