Study finds financial journalism tracks bubble risks despite frequent crash warnings

Study finds financial journalism tracks bubble risks despite frequent crash warnings
Press signals bubble risks

Long-running market commentary often overstates the odds of an imminent crash, yet new research suggests that persistent financial anxiety in the press still contains useful signals about asset bubbles. A 170-year analysis of The Economist's archives finds that while journalists imply a much higher probability of collapse than history supports, the strongest surges in alarm often align with major market manias.

Highlights

  • Researchers analyzed nearly 800,000 articles from The Economist over 170 years using language models to estimate bubble risk indices for 22 historic episodes.
  • Articles imply a 20% average probability of market collapse within six months—four times higher than actual frequency—while bubble warnings are present every month, with no period below 12%.
  • Model signals peaked at 97% in 2007 (global financial crisis), 92% in 2000 (dotcom bubble), and 89% in 1889 (Baring crisis), generally aligning with real market excesses.

Research method and historical signals

As reported by Financial Times, a new paper by Marco Grotteria of London Business School and Afonso Monteiro and Constantine Yannelis of the University of Cambridge examines about 170 years of financial coverage to test whether media warnings help identify bubbles.

The researchers use large language models to analyse nearly 800,000 articles from The Economist, measuring tone, sentiment, relevance and references to overvaluation or policy concern. They then build three indices designed to estimate the probability that coverage is describing a bubble, using 22 major historical episodes across 10 countries as benchmarks.

The paper finds that articles imply an average 20% probability of a major market collapse within six months, about four times the historical frequency. Bubble warnings appear in every month of the full sample, and the implied probability never falls below 12%, while in the median month 36% of market-relevant articles express concern about overvaluation and 24% cite a regulatory or policy warning.

Even so, the strongest signals appear to coincide with recognised periods of market excess. The authors say their model gives a 97% bubble signal in 2007 as the global financial crisis begins, 92% at the peak of the dotcom bubble in 2000, and 89% during the 1889 Baring crisis tied to Argentina, while readings above 50% have usually matched episodes that later unravel.

In a previous article, we explained Weiss Ratings forecasting service aimed at helping U.S. investors navigate volatile market cycles.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.