Can investor sentiment be a momentum time-series predictor? Evidence from China

Abstract

This paper challenges the prevailing view that investor sentiment is a contrarian predictor of market returns at nearly all horizons. As an important piece of "out-of-sample" evidence, we document that investor sentiment in China is a reliable momentum signal at monthly frequency. The strong momentum predictability is robust under both single- and multi-regressor settings, and is statistically and economically significant both in and out of sample, enhancing portfolio performance as shown by our numerical examples. More importantly, we find a striking term structure that local sentiment shifts from a short-term momentum predictor to a contrarian predictor in the long run. Cross-sectional analysis reveals that sentiment is more of a small-firm effect. Finally, we confirm that global sentiment spills over to the local Chinese market, as it predicts negatively future returns over the longer horizons and in the cross section

Does Investor Sentiment Predict the Near‐Term Returns of the Chinese Stock Market?

“…Panel C reports the results for the full sample (1997–2017). We find that the lagged sentiment positively relates to the subsequent market returns with a bias‐corrected slope coefficient of 1.79% ( t ‐stat = 4.14) that corroborates Han and Li (). However, we suggest that the bubble period in the sample had induced these results.…”

“…Our results remain similar when we use sentiment index data of Han and Li (). For example, we find that the OLS slope coefficient is 0.51% ( t ‐stat = 0.79) and a bias‐corrected slope coefficient is 0.64% ( t ‐stat = 1.12) once we exclude the 2006–2008 period from Han and Li () sample. We are thankful to Xing Han for providing their sentiment index data for replication purposes.…”

“…9 Finally, equation (3) tests whether our results remain robust when a dummy variable controls the effect of the bubble period. 10 We report the results in panel H. The results in panel H show that there is a strong positive association 8 Our results remain similar when we use sentiment index data of Han and Li (2017). For example, we find that the OLS slope coefficient is 0.51% (t-stat = 0.79) and a bias-corrected slope coefficient is 0.64% (t-stat = 1.12) once we exclude the 2006-2008 period from Han and Li (2017) sample.…”

“… Following Baker and Wurgler (2007),Chen et al (2014), andHan and Li (2017), we scale each individual sentiment proxy by its prior 6-month moving average.© 2018 International Review of Finance Ltd. 2018 …”

Overnight momentum, informational shocks, and late informed trading in China

“…The existence of the three alternative boards reflects China's multi-tier capital market system and offers a simple yet rigid classification of the firm size (see footnote 9). Admittedly, an alternative size-based classification would be to divide all listed firms into size-tercile groups based on the market value of equity as in Han and Li (2017). In fact, our findings remain unchanged under the alternative size classification method.…”