Investor sentiment in the Chinese stock market

Abstract

The thesis encompasses three essays in empirical asset pricing and focuses on investor sentiment in the Chinese stock markets. Traditional models of asset pricing assume that information is promptly processed and integrated into the prices of asset. In reality, pervasive investor sentiment may distort the assumption. Proponents of the efficient market hypothesis suggest that investor sentiment should not be considered as a pricing factor, since the mispricing caused by sentiment can be eliminated through trades made by rational speculators and arbitraging. However, behavioral finance theory argues that the impact of investor sentiment cannot be disregarded if it leads to uninformed demand shock and the cost of arbitrage is high. The three essays specifically explore whether the pricing effect of sentiment exists in Chinese stock markets.

In Chapter 2, we focus on market-, survey-, text- and search-based investor sentiment proxies and the impact of aggregate sentiment extracted from them in the Chinese stock market. Using data from 2008 to 2019, we find that individual sentiment proxies have limited return predictability, while the aggregate sentiment measures extracted from the four types of sentiment proxies show significant positive predictability both in- and out-of-sample. Moreover, the aggregate sentiment measures can deliver sizable economic gains to a mean-variance utility investor in an asset allocation exercise. This study advances our understanding of investor sentiment and its asset pricing and prediction implications in China.

In Chapter 3, we construct a text-based measure of investor sentiment by extracting the comments from individual investors on stock message boards in China. Using data from 2008 to 2020, we provide extensive evidence that the investor sentiment captured by our measure positively predicts cross-sectional stock returns in the following ten months. The text-based sentiment reveals that investor trading behaviors are influenced by online messages posted by individual investors. By longing the stocks with high sentiment and shorting the stocks with low sentiment, this long-short strategy based on textual sentiment measure produces significant economic value. In addition, we perform a range of robustness tests and confirm that the return predictability is not due to firm characteristics, common risk factors, investor attention, or alternative sentiment indicators.

In Chapter 4, we introduce disaster-induced sentiment measures derived from disaster-related search terms based on the Baidu Index and Google Trends to characterize investors' responses to disaster events. In particular, we study how disaster-induced sentiment affects stock returns in the Chinese and the US stock markets. Using data from 2007 to 2021, we find that disaster-induced sentiment measures negatively predict country-level market returns in the short term. The sentiment based on Baidu index increases the explanatory power of the return variation in predicting the Chinese stock market in comparison with Googling sentiment used in the literature. In China, the coastal provinces and provinces with high GDP are more heavily influenced by disaster-induced sentiment. Based on disaster-related internet search data, this study promotes our understanding of the impact of disaster-induced sentiment on the performance of stock markets.

Date of Award	Nov 2023
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Xiaoquan Liu (Supervisor) & Ying Jiang (Supervisor)