ABSTRACT
Herding among analysts emerges when analysts give priority to their peers’ opinions instead of their own beliefs or information. Some circumstances may enhance or restrain this type of behaviour. We postulate that market sentiment is one of them. This article analyses the effect that investor sentiment may have on analysts’ herding behaviour in the U.K. Our results suggest that ‘easy situations’ such as analysing easy-to-value securities and releasing optimistic information at times of high market sentiment clearly reduce herding practices, whereas herding clearly increases in difficult situations when analysts have to release negative information at moments of high investor sentiment.
Acknowledgements
We would like to thank the anonymous referees for their very useful comments and suggestions.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 As well as the major international firms that regularly send their recommendations to I/B/E/S, contributors to this database include some domestic analysts, which results in wider coverage in European countries. Nevertheless, like other forecast databases, FactSet is affected by potential survivorship bias, and also selection bias, because it collects recommendations and forecasts from brokerage houses that collaborate on a voluntary basis. Correction of these two biases is not possible.
2 Size is measured as the market capitalization and volatility is obtained as the SD of stock returns for the previous 12 months.
3 Previous studies include investor survey findings (Brown and Cliff Citation2005), dividend premium (Baker and Wurgler Citation2004) and turnover (Scheinkman and Xiong Citation2003), among others.
4 The reason for the choice of these measures is their relationship with the level of sentiment used by Baker and Wurgler (Citation2006), together with data availability. Constructional details of the turnover index can be found in the work by Baker and Stein (Citation2004), and those of the volatility premium in the work by Baker, Wurgler, and Yu (Citation2012). The consumer confidence index, which is available from the European Commission website, has been used in numerous studies, such as by Brown and Cliff (Citation2005), Lemmon and Portniaguina (Citation2006) and Schmeling (Citation2009), among others.
5 Following Baker and Wurgler (Citation2006) and Schmeling (Citation2009), the macroeconomic variables considered are the industrial output index, durable and non-durable goods consumption and the unemployment rate.
6 In order to reduce the EPS skewness effect, we consider the median consensus instead of the mean consensus.
7 The observations are winsorized at the 99% level.
8 For the sake of clarity, in the remainder of the article we report only the estimations performed on named analysts’ forecast data using the method employed by Hribar and McInnis (Citation2012). It is worth noting, however, that the results coincide with those obtained using the alternative sample.
9 The causality test was carried out through a VAR technique where both indices were included. According to the ADF test, all the variables are stationary. The criterion for the lag choice was the Schwarz criterion.
10 Henceforth, for the sake of clarity, we present only the monthly estimates. The quarterly estimates, which are consistent with the latter, are available from the authors upon request.
11 When the size variable is used to classify the assets, it is also observed that herding increases in hard-to-value portfolios.
12 When the volatility variable is used to classify the assets, it is also observed that herding increases in hard-to-value portfolios.
13 The influence on the S statistic is only significant for hard-to-value portfolios when the assets are ranked by size, for which mimetic behaviour intensifies when sentiment and portfolio are taken into account.
14 We appreciate this idea suggested by the anonymous reviewers.