Identification of High-Frequency Herding Behavior in the Chinese Stock Market: An Agent-Based Approach

Abstract

The existing literature often uses various aggregate measures to track the influence of sentiment on asset prices. Rather than using proxies like volume or cross-sectional dispersion, it might, however, be preferable to estimate a full-fletched model that allows for sentiment formation among investors along with fundamental innovations. This paper estimates one such model proposed by Alfarano et al. (J Econ Dyn Control 32(1):101–136, 2008). We apply the simulated method of moment estimator proposed by Chen and Lux (Comput Econ 2018, https://doi.org/10.1007/s10614-016-9638-4) to investigate the herding behavior in the Chinese stock market using high-frequency data. The asset pricing process is driven by fundamental factors and sentiment change due to idiosyncratic changes of expectations and herding effects. We find evidence of both autonomous switches of sentiment as well as of herding effects in the Chinese stock market. The autonomous switching tends to dominate the herding effect. Both autonomous and herding effects are stronger during the crisis year 2015 than in other periods.