More than words: quantifying language to measure firms' fundamentals

Document Information:
Title:	More than words: quantifying language to measure firms' fundamentals
Author(s):	Tetlock P C, Saar-Tsechansky M, Macskassy S
Journal:	Journal of Finance, Jun 2008, Volume: 63 Issue: 3 pp.1437-1467 (31 pages)
Issn:	0022-1082

Keywords:	Financial Forecasting, Financial Reporting, Language, Stock Markets, Transaction Costs, Usa
Article type:	Research paper
DOI:	10.1111/j.1540-6261.2008.01362.x
Reference:	37AP327 (Permanent URL)
Abstract:	Purpose - To assess the forecasting value of linguistic indicators in financial reporting. Design/methodology/approach - Reviews prior studies of qualitative information research. Takes approximately 350,000 news stories from Wall Street Journal and Dow Jones News Service between 1980 and 2004, covering 1,063 firms while they were in the S&P 500 Index. Measures the proportion of negative words in each, as well as a stationary measure of negative media content. Runs regressions of standardized unexpected quarterly earnings (SUE) and standardized analysts' forecast error for quarterly earnings (SAFE) against several variables including returns and transaction costs. Findings - Finds that reporting is clustered around earnings announcements, implying information about fundamentals is disseminated. Shows that negative words in media reports forecast low earnings, after a short lag, and returns are also predictable. Indicates that the market underreacts in the first day after announcement. Suggests investors react less to non-earnings statements with negative words. Practical implications - Notes that transaction costs exceed any profit from a trading strategy based on such forecasts, and market prices underreact to negative words. Originality/value - Presents a new predictor of firm share prices that appears to indicate market inefficiency.