Social Media Monitoring to Predict Bank Runs: Signals and Limitations

ABSTRACT This article examines whether real‐time social media monitoring could have provided early warning signals for recent bank runs by analysing sentiments and emotions in 160,000 tweets regarding Silicon Valley Bank (SVB), Signature Bank (SBNY) and a control group of banks that did not experience runs. The results show statistically significant negative deviations in sentiment and emotions prior to the bank run for both SVB‐ and SBNY‐related tweets, with sharp increases being visually evident for SVB. The control group indicates that similar strong negative co‐movements across multiple sentiments and emotions do not usually occur in other banks. However, the analysis also shows that potential social media signals are prone to false positives within expectable fluctuations, making them unfavourable for practical applications as a standalone warning system. Furthermore, the findings highlight significant heterogeneity between social media reactions to different banks, emphasising that reliable thresholds for social media monitoring require bank‐specific calibration.