Abstract
This paper elicits and quantifies narratives from open-ended surveys sent daily to U.S. stockholders during the first wave of the COVID-19 pandemic. Using textual analysis, we extract thirteen narratives and measure their prevalence over time. A validation analysis confirms the behavioral and economic relevance of the retrieved narratives. Moreover, we find that the narratives contain predictive information for future excess stock and bond returns, and this predictability remains when controlling for contemporaneous information stemming from news and social media. Finally, we find evidence that political identity is reflected in the narrative tone.
Original language | English |
---|---|
Journal | Journal of Applied Econometrics |
Volume | 38 |
Issue | 4 |
Pages (from-to) | 512-532 |
Number of pages | 21 |
ISSN | 0883-7252 |
DOIs | |
Publication status | Published - 1 Jun 2023 |
Keywords
- COVID-19
- textual analysis
- latent Dirichlet allocation
- narrative economics