Data for “Legitimacy and Digital Engagement with the Federal Police: Audience Sentiments on Social Media During the Government Transition”

Abstract: This research systematically investigated how the institutional communication strategies of the Brazilian Federal Police, through its official Instagram account, influence the perception of institutional legitimacy and public engagement, especially during the governmental transition. To this end, a database was built with 1,112 posts and more than 138,000 comments, collected between December 2021 and October 2024. The methodology combined analysis of engagement metrics, thematic classification of the publications, human coding of the content of comments, and automated sentiment analysis using a BERT model fine-tuned for Portuguese. The results indicated that content about professional training and career entry generated higher proportions of positive interactions (positive comments in relation to negative ones) and engagement (likes), while operational and institutional posts showed a greater propensity for ambivalent and polarized reactions. It was also observed that, after the elections, there was a reduction in average engagement and a relative increase in critical manifestations. The study culminated in the development of an updated Social Media Manual with data-based guidelines to strengthen the institutional legitimacy of the Federal Police in the digital environment. The evidence reinforces that legitimacy is a dynamic process built at the intersection between institutional discourse, political context, and public perceptions expressed in online interactions.
Keyword: digital communication, institutional legitimacy, Federal Police, social media, sentiment analysis.

README
Code used to extract Instagram post data: python baixar_instagram.py
Code used to train BERTimbau (fine-tune): fine_tune_bertimbau.py
Code used to perform sentiment analysis: analise_sentimento_BERTimbau_finetuned.py
Code used to consolidate the sentiment analysis results: consolidar_dados_e_sentimento.py
Comments and Sentiment Analysis: Comments_and_Sentiment_Analysis.xlsx
Code used for Theme Analysis in comments with KHCoder: KHCoder_Text_Codes.txt
Tabulated data for Regression Models: Data Posts.dta
STATA code (do-file): Code Data Posts.do