Geopolitical risk tweets predicted some market changes

What the study found: The study found that a multilingual X/Twitter-based geopolitical risk sentiment series contained predictive information for several financial assets and market changes at different lag times.
Why the authors say this matters: The authors suggest that tracking geopolitical risk sentiment across languages may help relate public reaction to movements in financial assets and markets. They conclude this relationship was observable during the start of the Ukraine War.
What the researchers tested: The researchers collected more than 3.6 million tweets in seven languages from December 1, 2021 to April 30, 2022. They used Goldstein 1992 positive and negative geopolitical risk bigrams, applied sentiment analysis methods, built a daily time series of geopolitical risk sentiment, and tested its relationship with 39 financial assets and markets using Granger causality.
What worked and what didn't: Granger causality showed that the geopolitical risk time series contained predictive information for several assets and market changes at different lag times. The abstract does not specify which assets or markets were affected, or which sentiment methods worked best.
What to keep in mind: The abstract gives a broad summary and does not list the specific financial assets, the exact lag times, or the details of the sentiment-analysis methods. It also does not report limitations beyond the study period and data sources.

Key points

• More than 3.6 million tweets were analyzed in seven languages.
• The study built a daily geopolitical risk sentiment time series from tweet data.
• Granger causality indicated predictive information for several financial assets and market changes.
• The abstract does not identify which assets or exact lag times were involved.