The Effects of Herding on Betas and Idiosyncratic Risk

Abstract

This paper investigates the consequences of herding on systematic and idiosyncratic risk for stocks traded on S&P 500. Herding behavior is measured through a state-space model. Using monthly data from 1999 to 2017, different periods of herding and adverse herding are present. Evidence shows that the state space model identifies the significant herding effects on both risk measures for specific portfolios. Our findings validate the expected implications of herding on betas but not of adverse herding. In addition, the low-beta anomaly is not confirmed on our beta-based portfolios. On the other hand, we confirm the risk-return relationship. We attribute this evidence to overpriced values of high beta assets as well as to the effects of adverse herding on the systematic and idiosyncratic risk. Finally, we also show that the herding level could serve as a systematic driver of returns improving the portfolio performance of traditional ‘anomaly’ based strategies.

Keywords:

• Herd behavior
• Systematic risk
• Idiosyncratic risk
• Herding strategies

Acknowledgements

The authors would like to thank the anonymous referees for their constructive comments that helped to substantially improve the final version of this paper.

Notes

1 See for example, Levy (1974), Fabozzi and Francis (1977), Ferson and Harvey (1991, 1993), Clinebell et al. (1993), Woodward and Anderson (2009)). More recently, Cederburg and O’Doherty (2016) reconsider the abnormal performance of beta-sorted portfolios and find a substantial variation of betas over time.

2 Beta instability has also been attributed to earnings surprises (Ball et al. (1993)), investment opportunities (Berk et al. (1999), Gomes et al. (2003)), leverage (O’ Doherty (2012), Garlappi and Yan (2011)) or even to changes in regulatory frameworks (Pham et al. (2018)).

3 We performed cross sectional regressions on a monthly basis to confirm the significant role of idiosyncratic risk. Our results indicate a positive (0.189) and statistically significant (t-statistic of 2.4) tradeoff between idiosyncratic risk and expected return. The detailed results are available from the authors upon request.

4 Following Messis and Zapranis 2014b) we adopt the state space model for detecting herding behavior. The method considers herding as a measure of the behaviour of investors who follow the performance of specific factors, styles or macroeconomic signals, and thus proceed on buying or selling individual assets at the same time disregarding their underlying risk-return relationship. Moreover, state space models provide a more detailed analysis over time, Demirer, Kutan, and Chen (2010). Finally, the model is free from the influence of idiosyncratic components, Hwang and Salmon (2004).

5 Cederburg and O’Doherty (2016) find the beta anomaly to be resolved through the conditional CAPM.

6 For example, Friend and Blume (1970), Fama and French (1992, 1996) show that strategies based on the performance of portfolios formed on lagged betas do not confirm the spread between high-low beta-based portfolios.

7 The results are similar regarding the systematic and the idiosyncratic risk when the CAPM or the 4FM are employed.

8 We find similar differences in alphas when portfolios are formed based on betas estimated from the CAPM and 4FM.

9 Arbitrageurs will try to correct the mispricing of high beta assets. However, their limited risk-bearing capacity prevent them for taking the right number of short positions, leading to equilibrium overpricing (Hong and Sraer (2016)).

10 We have also considered significant positive and negative changes in market returns to be one standard deviation away from the mean. The general conclusion remains the same.

11 For robustness check we also used slightly different values from $\left|0.02\right|$ and found no significant differences in the main conclusions.