arXiv study of 6.28M headlines finds media-market link is firm-specific

A new arXiv paper from Shivansh Verma, Soham Tulsyan, Sashwat Dhanuka, and Anirban Sen asks a classic question with unprecedented data: do financial news headlines move stock prices, or just reflect information already priced in? The researchers built a corpus of 6.28 million headlines covering 26 major US firms from 2015 to 2025, then filtered for firm-specific, materially relevant coverage. They constructed daily stance measures and ran panel regressions plus vector autoregressions with data-driven structural breaks, focusing on how the relationship shifted around the COVID-19 pandemic's 2020 shock.

The headline result: there is little evidence of a persistent market-wide change in media stance or stock returns following the pandemic. Instead, dynamic media-market relationships appear only for a subset of firms around their own structural breaks. This nuance challenges prior studies that relied on aggregate sentiment indices, which can mask firm-level heterogeneity. The authors argue that understanding how news and prices interact under changing conditions requires firm-specific analysis, and they offer a replicable framework for doing so. The paper is posted on arXiv (2608.05899) under computational finance, and—as always with preprints—has not been peer-reviewed.