Negativity Bias of Analyst Forecasts

Abstract

In contrast to the conventional view that analysts forecast optimistically, we provide evidence of the Negativity Bias. Analysts show negative forecast bias associated with their relative local income growth, whether the growth is positive or negative. The bias is stronger for negative growth than for positive growth. The negative bias also directly affects the bias of the next analysts of the same and peer earnings being forecast. Our results suggest non-fundamental factors at work.

Keywords:

• Negativity bias
• Mood effect
• Contrast effect
• Forecast bias

JEL CLASSIFICATIONS:

• G24
• G41

Acknowledgment

We thank Hamish Anderson, Ling Cen, Jie Gan, Gilles Hilary, David Hirshleifer and Md Humanyun Kabir for helpful comments and Andrea Bennett and Chris Moore for proofreading our manuscript.

Notes

1 Analysts’ income plausibly consists of an industry component, a performance-based component and a local component. The industry component has mainly variation over time, but not cross-sectional variation. As for the performance-based compensation, it should capture significant cross-sectional variation. However, it should not be associated with behavioural bias because analysts consider it as what they deserve for their effort, skills or ability. What remaining will be geographical variation. Although we measure an analyst’s local income by the income of the state in which the analyst is located, its variation across states plausibly captures the cross-sectional local income variation of analysts for the following reasons. First, the finance sector of the state is generally a significant sector in the state. In fact, relative income for the finance sector of the state is highly correlated with the state overall relative income (between 0.84 and 0.95). Second, analyst income is highly correlated with capital market activity, as shown in Groysberg, Healy, and Maber (2011). Therefore, our relative local income measure may be associated with cross-sectional variation in analysts’ behavioural bias, as we explain our hypotheses below. Moreover, as Owyang, Piger, and Wall (2005) report that contemporaneous economic growth varies significantly across states, there is significant cross-sectional variation of the relative local income growth. In addition, the relative local income measure of a particular state varies over time. Hence, it may also be associated with time-series variation in behavioural bias. We provide statistics concerning the variability of the relative local income in Section 1 below.

2 Oshio, Nozaki, and Kobayashi (2011) extend the literature of Western countries to Asian countries.

3 Conceptually, true GDP growth and true income growth are considered interchangeably as economic growth (e.g., Henderson, Storeygard, and Weil 2011)

4 We obtain qualitatively the same results when we exclude those observations for which analysts are located in those states that are more than 5% relevant for the businesses of the close competitors (García and Norli 2012).

5 We acknowledge Kee-Hong Bae and Hongping Tan for providing us with location data for the period 1995–2010. The procedure used to identify analysts’ locations is the same as that used in Bae, Stulz and Tan (2008) and Bae, Tan and Welker (2008). Using Nelson’s Directories, we manually check cases with the same analyst, the same research firm and multiple locations. We exclude observations for which there is insufficient information to clearly identify the location of the analyst.

6 The results that are not tabulated are available from the authors upon request.

7 The magnitude of the coefficients is comparable to those in Table 4. The coefficients of positive LOC-TO-US_GROWTH is between −0.044 and −0.048 whereas the coefficients of negative LOC-TO-US_GROWTH is between 0.073 and 0.076. The statistical significance is slightly weaker than that in Table 4.

8 We further conduct a logistic transformation of the dependent variable.

9 In our regressions, we have two explanatory variables that control for the effect that there is different amount of information available for forecasting at different points of time. They are the horizon between the forecast release date and the data date of the earnings being forecast and the cumulative monthly stock returns between the data date of the last annual earnings and the most recent month prior to the forecast release day.

10 Welch (2000) refers herding to mutual imitation.