Predictability of Analyst Stock Recommendation Revisions

Abstract

On average, analysts’ stock recommendation revisions have immediate effects on stock prices. However, recent research indicates that only a small subset of recommendations are influential in the sense that they are associated with significant returns. Given the relative infrequency of recommendation revisions compared to other analyst output, we expect that analysts predictably signal future recommendation revisions through changes in the tone of sequential research reports and quantitative forecasts. Consistent with our expectation, we find that future recommendation revisions are positively associated with the signed change in analyst report tone, expected returns implied from target prices, and target price revisions. To test whether the predictability of recommendation revisions preempts market reactions when revisions are announced, we construct a predictability measure using changes in research report tone, analysts’ quantitative output, and firm characteristics. We find that our predictability measure is associated with attenuated market reactions to recommendation upgrades and a lower likelihood that the upgrades are classified as influential. We also find that our recommendation revision prediction model is better at predicting upgrades than downgrades and has similar performance to prediction models in other settings in the literature predicting rare events, such as fraud prediction models.