Essays on the Quantification of Newspaper Articles

This dissertation examines how large-scale newspaper and institutional text can be transformed into quantitative indicators for economic and policy analysis. It develops supervised text-classification approaches tailored to three empirical applications. Chapter 1 studies whether newspaper text improves nowcasts of German GDP. Using a corpus of 12.4 million German news articles, it combines business-cycle sentiment with topic modeling and shows that text-based indicators add predictive information beyond standard hard data. Chapter 2 analyzes how central bank communication affects media reporting and, through the media, household inflation expectations. Based on more than 2.1 million German newspaper articles and ECB press conferences from 2002 to 2023, it shows that households react asymmetrically to media signals, responding more strongly to news about rising inflation and dovish ECB quotes than to disinflationary or hawkish signals. Chapter 3 studies how German newspapers cover and frame the German Baltic fishery. It shows that political events are the main catalysts of media reporting, that fishery representatives receive more personalized and favorable coverage than environmental NGOs, and that the media portray the debate as a conflict between socio-economic and ecological coalitions. Overall, the dissertation shows that text-based indicators are most informative when they are designed to match the corpus and the research question at hand.

Level thesis:degree_level

thesis.doctoral

Grantor dc:publisher

Christian-Albrechts-Universität zu Kiel

Year

2026

Author dc:creator

• Bär, Jasper

Contributors dc:contributor

• Carstensen, Kai
• Boysen-Hogrefe, Jens

• Text as data
• Macroeconomic forecasting
• Central bank communication
• Media discourse analysis
• Sentiment analysis
• Monetary policy

Repository record source_url

https://macau.uni-kiel.de/receive/macau_mods_00008183

OAI identifier oai:identifier

oai:macau.uni-kiel.de:macau_mods_00008183