Crying Wolf: Media Warnings and the Predictability of Financial Bubbles
Marco Grotteria,
Afonso A. Monteiro and
Constantine Yannelis
Cambridge Working Papers in Economics from Faculty of Economics, University of Cambridge
Abstract:
We develop a framework that applies large language models (LLMs) to 170 years of The Economist (1843–2013), extracting article-level indicators of speculative content and aggregating them into a monthly bubble index, which we evaluate against 22 documented episodes in ten countries. We document a phenomenon we term "crying wolf": the press warns of overvaluation in every month and assigns to an imminent crash a probability several times its realized frequency, so the level of its warnings carries little information. Their variation, however, is informative. In 74 percent of comparisons between a bubble month and a normal month, the index is higher in the bubble month, and for the Dot-com episode the share is 94 percent. For a policymaker who regards a missed bubble as several times more costly than a false alarm, the index reduces the expected loss by a fifth to a third relative to ignoring it; for one who weighs the two errors equally, it yields no gain. The index also forecasts lower future returns. Out of sample, the index on its own reduces the mean squared error of 12-month NASDAQ return forecasts by 11 percent relative to the historical mean, and its coefficient remains significant with valuation controls and after correcting for predictor persistence. Applied to September 2026, the index assigns a probability of up to 97 percent to an AI bubble.
Keywords: Bubble; Asset Prices; Artificial Intelligence (search for similar items in EconPapers)
JEL-codes: C55 G01 G12 G14 G17 N20 (search for similar items in EconPapers)
Date: 2026-11-15
References: Add references at CitEc
Citations:
Downloads: (external link)
https://www.econ.cam.ac.uk/sites/default/files/pub ... pe-pdfs/cwpe2692.pdf
Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.
Export reference: BibTeX
RIS (EndNote, ProCite, RefMan)
HTML/Text
Persistent link: https://EconPapers.repec.org/RePEc:cam:camdae:2692
Access Statistics for this paper
More papers in Cambridge Working Papers in Economics from Faculty of Economics, University of Cambridge
Bibliographic data for series maintained by Jake Dyer ().