Comparing Links between Topic Trends and Economic Indicators in the German and Polish Academic Literature

Abstract

The popularity of econometric analyses that include variables obtained from text mining is growing rapidly. A frequently applied approach is to identify topics from large corpora, which makes it possible to determine trends that reflect the changing relevance of topics over time. We address the question of whether such topic trends are linked to quantitative economic indicators typically used for analysing the objects described by a topic. The analysis is based on academic economic articles from Poland and Germany from 1984 to 2020. There is a specific focus on whether relationships between topic trends and indicators are similar across national economies. The connection between topic trends and indicators is analysed using vector autoregressive models and Granger causality tests.

Keywords

• topic modelling
• text analysis
• latent dirichlet allocation
• granger causality
• topic trends