•
Tetlock’s classic result: pessimism in a major WSJ column lines up with temporary downward pressure and higher trading volume—foundational alt-data intuition before modern transformers.
Paul C. Tetlock (Journal of Finance, 2007) gives content to investor sentiment using media text. High pessimism in a prominent Wall Street Journal column is associated with temporary downward pressure on market prices and with unusually high trading volume; prices tend to rebound afterward in the paper’s results. For after-work learners this is the classic proof that words in the business press can be measured and studied next to returns — long before FinBERT. This Hub note is educational reading for after-work Python practice. It is not investment advice, not a trading signal, and not an offer to buy or sell any security.
Before you fine-tune transformers, you need a reason to believe media tone is worth encoding at all. Vague claims like “the market feels bad” are not data. Tetlock’s problem was operational: turn a newspaper column into a numeric pessimism measure, then test whether that measure lines up with market returns and volume in a disciplined way.
The study focuses on the WSJ “Abreast of the Market” column and builds sentiment from word tallies (Harvard psychosocial dictionaries / pessimism counts in the paper’s design). High media pessimism predicts pressure on prices that partially reverses, consistent with temporary sentiment-driven effects, and coincides with elevated volume — investors trade more when the column is gloomy.
Modern pipelines replace dictionary counts with embeddings and FinBERT scores, but the research shape remains: define a text measure → align calendar dates → study return and volume associations with clear caveats. Read Tetlock to see how a careful alt-data claim is written.
Author: Paul C. Tetlock (2007). Open PDF host used in our brief: UTS-hosted copy of the JoF article.
One evening exercise (educational only):
count_neg / (count_pos + count_neg + 1e-6) in plain Python.This is literacy for alt-data papers, not a backtest.
When you want structured evening lessons on Python itself — variables, loops, functions, files, and first libraries — start with Introduction to Python Programming. Quant and macro courses are not the CTA here; build the language first.