Analysts’ Qualitative Statements and the Profitability of Favorable Investment Recommendations
Marcus Caylor, Mark Cecchini and Jennifer Winchel
Accounting, Organizations and Society, Vol. 57, (February) 2017, pp. 33-51
We use a novel machine learning text methodology that allows us to create a numerical text signal from the narrative in financial analysts’ reports with buy ratings. This text signal is the result of a text methodology which distinguishes words and phrases that are important for classifying profitable and unprofitable buy ratings in analysts’ reports. Using our text signal, we find that what analysts write actually yields important information about buy ratings’ profitability, controlling for information conveyed in the quantitative summary measures identified by prior research, such as target price and earnings forecasts. By taking long (short) positions in the top (bottom) quintile of the text signal, we developed trading strategies for twelve-month buy-and-hold portfolios that generated economically significant returns. When we further disaggregated the text signal into discrete information categories, we found that discussions of historical financial and nonfinancial performance measures have significant predictive power. Overview
8 | Journal Publications - Financial Times Top 50
Made with FlippingBook - Online catalogs