Are Sell Recommendations More Accurate? Evidence from Consensus Analyst Target Prices

Sammanfattning

This study examines whether analyst sell recommendations are more accurate than buy recommendations, and whether the accuracy difference varies between growth and value firms. Using consensus target prices for US-listed firms from Capital IQ over Q1 2004 to Q1 2025, forecast accuracy is evaluated using mean absolute forecast error, the Diebold-Mariano test, and the Mincer-Zarnowitz regression. The results indicate that sell recommendations are less inaccurate than buy recommendations, exhibiting a mean absolute error of 32 percentage points compared to 41 for buy recommendations. However, both recommendation types display noticeable forecast errors, and the sell recommendation sample is considerably smaller than the buy recommendation sample, with 74 positive signals for every negative signal. The accuracy gap declines with firm size but remains present even for the largest firms. Evidence of a value or growth effect is weaker overall, but the gap widens for small growth firms, driven primarily by the buy-side accuracy deteriorating while the sell-side accuracy remains stable. Overall, the findings suggest that sell recommendations may contain more accurate forecasting information than buy recommendations. However, the results should be interpreted with caution, as both recommendation types exhibit substantial forecast errors.