Giving Content to Investor Sentiment: The Role of Media in the Stock Market

Source: Tetlock, P. C. (2007) · Journal of Finance 62(3), 1139–1168 · doi:10.1111/j.1540-6261.2007.01232.x

TL;DR

One of the first papers to turn news text into a quantitative market signal. Tetlock measures the

pessimism of the Wall Street Journal's daily "Abreast of the Market" column with a content-analysis

program and shows that **high media pessimism predicts downward pressure on prices followed by reversion

to fundamentals, and that unusually high or low pessimism predicts high trading volume**. The

patterns fit noise/liquidity-trader theories, not media-as-new-information.

Financial news content could induce, amplify, or merely reflect investor sentiment. Using daily text

from the WSJ "Abreast of the Market" column over 1984–1999 (16 years), Tetlock builds a media

pessimism measure and estimates its intertemporal links with returns and volume in vector

autoregressions (VARs), testing the sentiment view against the new-information and "sideshow" views.

Construction: count words in the 77 General Inquirer (Harvard-IV psychosocial) categories per

day; take the first principal component, which loads on pessimistic words — the "pessimism factor."

Returns: high pessimism predicts downward price pressure that reverts over the following days

— temporary sentiment-driven mispricing, not fundamental news.

Volume: both unusually high and unusually low pessimism predict high trading volume.

Robustness: results hold when a time gap is allowed between media release and the return window,

and under alternative GI-based pessimism measures. Statistical tests reject the new-information and

no-relation hypotheses.

The template for NLP in finance: systematically scored text predicts markets, paving the way for

dictionary methods (and the finance-specific Loughran–McDonald lexicon) and later embedding- and

LLM-based approaches.

General-purpose dictionaries misclassify finance language (motivating Loughran–McDonald 2011).

A single column over one era; effects are short-horizon and small relative to costs.

Timestamp/leakage care is essential when aligning news with returns.

Tetlock, P. (2007) — Giving Content to Investor Sentiment — Journal of Finance

Loughran, T. & McDonald, B. (2011) — When Is a Liability Not a Liability? Textual Analysis, Dictionaries, and 10-Ks — Journal of Finance

Tetlock, P., Saar-Tsechansky, M. & Macskassy, S. (2008) — More Than Words — Journal of Finance

Provenance: verified/generated from the paper's full text.