Topic modeling in news articles: an example from Brazil
Published in UCL Repository, 2022
Even though economic uncertainty impacts a country’s economic activity, it remains a challenge to measure it. The effects of uncertainty can be even more profound and prolonged on emerging market economies. This is the case of Brazil, an emerging market economy that expe-rienced a recession between 2014- 2016, political instability, and corruption scandals in the last decade. To identify the topics related to economic uncertainty and how these topics inform the country’s Economic Uncertainty Index (IIE-Br), we used Latent Dirichlet Allocation (LDA), a topic modeling approach. This technique allows us to identify the hidden topics from an extensive collection of texts, in our case, newspapers. For this purpose, we used the newspaper article database provided by Funda ̧c ̃ao Getulio Vargas (FGV). Our database consists of 8263 articles related to economic uncertainty between 2009-2020 from Estad ̃ao, one of Brazil’s most influential newspapers. We extracted fifteen topics related to economic uncertainty, such as elections, politics, and international relations, between others, that gave us more information on the topics that drive uncertainty the most in Brazil and how these topics have changed over time.