Narratives and past crashes
“History of past events, whether personally experienced or not, plays a crucial role in the assessment of extreme crash probabilities,” argue Goetzmann et al. in this paper. Investors use past crash events to understand why things are happening, deduce causalities between different states of the world and predict the future. These narratives are being propagated by the press which serves as a mechanism of aggregate market feedback. How important are these crash narratives for predicting market volatility? The authors answer this question by using computational linguistics to extract narratives that appear in the news in the days following the 1987 and 1929 crashes and estimate their prevalence in news articles each day for over 30 years. They conclude that this measure captures “crash narratives” in the media:
We show that the measure corresponds with major market events throughout the sample period, including the LTCM collapse, 9/11, the Great Financial Crisis, the European Debt Crisis, and the COVID-19 pandemic. This suggests that journalists broadly employ narratives used following the 1987 crash in spite of differences in contexts. It has a significant relationship with lagged, market-based crash indicators in the post-1987 period. Manual inspection suggests that these patterns are unlikely to be driven by the prevalence of specific words alone, but rather by the sequencing and distribution of words representing narrative structure.
Crash Narratives
Authors: William N. Goetzmann, Dasol Kim, Robert J. Shiller
From: Yale University, US Department of the Treasury
