ECB researchers use LLMs to measure geoeconomic tension
A new ECB study shows how AI improves the measurement of geopolitical and geoeconomic developments.
The European Central Bank (ECB) researchers have published a working paper introducing a Large Language Model-based method for measuring geopolitical and geoeconomic tensions in the euro area.
The paper develops the LLM Geoeconomic and Geopolitical Tension index, or LGPT, using a large dataset of European newspaper articles in local languages.
Researchers analysed almost 20 million articles from newspapers in France, Germany, Italy and Spain, covering the period from 1999 to 2025.
The methodology combines a fine-tuned multilingual BERT model with GPT-4o in a two-stage classification process.
BERT is used to filter articles likely to relate to geopolitical or geoeconomic tension, while GPT-4o classifies relevant articles and extracts structured information.
The index distinguishes narrower geopolitical tensions from geoeconomic tensions, including economic policy or the use of resources for geopolitical purposes.
It also breaks geoeconomic tension into four sources: trade, energy, finance and technology.
They also show how the index can be integrated into macroeconomic modelling to assess the effects of geoeconomic tensions on output and inflation in the euro area.
Why does it matter?
The paper shows how LLMs can be used as analytical tools for economic policymaking, not only as chatbots or productivity software. Measuring geoeconomic tension more precisely matters because trade conflict, energy security, financial fragmentation and technology restrictions can affect inflation, output and financial stability in different ways. A multilingual approach is especially relevant for the euro area because it captures local-language reporting from major member states rather than relying only on English-language media or keyword lists.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!