Bubbles Talk : Narrative Augmented Bubble Prediction

Financial bubble theories emphasize the importance of behavioral mechanisms cen- tered around investor beliefs, which can be potentially gleaned from prevailing nar- ratives, that reflect investors’ psychological states and link them to economic events. By summarizing market narratives into meaningful and economically relevant features, guided by bubble theories, we offer a novel approach to bubble prediction. We then test whether the variation of narratives and bubble measures are related on a predictive basis, as bubble theories imply. Our findings reveal that most of our narrative fea- tures exhibit statistically significant predictive power for bubble measures, and that the narrative-augmented models outperform non-augmented benchmarks in out-of-sample tests. These results offer new insights into the understanding of bubbles and lay the foundation for using narratives to develop early warning systems (EWS) for bubble for- mation and deflation, and for investigating the causal relationship between narratives and economic events