Abstract: Stock price bubbles seriously affect the healthy operation of capital markets. Recently, with the development of behavioral finance and other financial theories, many studies have explored the causes of stock price bubbles from the perspective of investor behavior and market microstructure. Behavioral finance theory believes that investor sentiment plays an essential role in affecting the price of financial assets. Investors tend to be influenced by emotions to change their investment behavior, which to a certain extent, drives the speed and extent of stock price fluctuations and thus generates stock market bubbles. Moreover, market microstructure theory holds that stock liquidity, as a critical measure of market quality, will directly affect the speed and cost of market transactions, eventually acting on stock price fluctuations and leading to stock price bubbles. Therefore, investigating the relationship between investor sentiment, stock liquidity, and stock price bubbles can help maintain the stable development of our financial markets.
To explore the mechanisms of both investor sentiment and liquidity on stock price bubbles, we select nine Shanghai Stock Exchange(SSE) industry indices, including Energy, Materials, Industrials, Discretionary Consumption, Staples Consumption, Pharmaceutical, Information Technology, Telecom, and Utilities. Then based on dynamic factor analysis, this study constructs investor sentiment indicators by combining online social media data, such as the Baidu index, with traditional sentiment proxy variables, such as turnover rate. Subsequently, the bubbles in different industries are detected based on the Generalized Sup-ADF (GSADF) test. The GSDAF test results show that bubbles are detected in all nine industry indices, and the original hypothesis of the existence of unit root is rejected at the 1% significance level except for the Energy industry. Specifically, the highest number of bubbles is four in the Staples consumption sector, the second highest is three in the Industrials industry, and two bubbles exist in the Pharmaceutical, Information Technology, and Utilities sectors. In addition, the first bubble mostly lasted from late 2014 to mid-2015, with an average duration of 5 months, consistent with the occurrence of the stock market crash in June 2015. The second or third bubbles lasted for a relatively short period, and the timing of the bubbles varied widely across industries. Furthermore, the relationship between investor sentiment, liquidity, and stock price bubbles is further analyzed using the panel Logit model and mediating effects test. Since liquidity refers to the ability of a larger transaction to materialize quickly and with weaker impacts on the market’s price, this paper uses two widely used liquidity measures: The Amihud illiquidity ratio and the quoted spread (QS). That is, we present regression results based on both the Amihud and QS liquidity indicators, respectively. Furthermore, this paper also conducts robustness checks by replacing investor sentiment and liquidity indicators to exclude the bias due to the selection of investor sentiment and liquidity measures.
The results of both empirical analysis and robustness tests indicate that investor sentiment and liquidity have significant positive effects on the existence of stock price bubbles, and investor sentiment can lead to a further increase in the probability of stock price bubbles’ occurrence by promoting market liquidity. Specifically, investor sentiment and Amihud (QS) liquidity measure both positively impact stock price bubbles at the 5% significance level. When investor sentiment is high, investors’ trading behavior of mindlessly following will keep pushing up the stock price, causing the stock price to deviate from its intrinsic value. Then the elevated sentiment will accelerate the transmission of good news, enhance market liquidity, and increase market depth, leading to a rapid pull-up of the stock price and the creation of a bubble. Conversely, when investors’ sentiment is negative, the conservative investment strategy leads to declining stock prices, and the conservative investment strategy caused by negative sentiment reduces market depth. Then market liquidity is weakened with increasing transaction costs, resulting in a continuous decline in stock prices and the bursting of bubbles.
Based on the empirical results, this study can have important implications for preventing stock price bubble risk and provide the relevant theoretical basis for regulators to strengthen the regulation of market sentiment. In future research, we can further apply various deep-learning methods to predict stock price bubbles based on investor sentiment and liquidity proxies. This study has been supported by the National Natural Science Foundation of China under Grant 72171005.

Key words: investor sentiment, stock price bubbles, liquidity, GASDF test

摘要： 股价泡沫严重影响资本市场健康运行,而投资者情绪和市场流动性是影响泡沫膨胀和破灭的重要因素。为探究二者对股价泡沫的作用机制,本文选取上证行业指数,通过动态因子分析法,将百度指数等网络社交媒体数据与换手率等传统情绪代理变量相结合,构建多个行业的投资者情绪指标。随后基于广义右尾单位根(Generalized Sup-ADF, GSADF)检验法,检测了不同行业的泡沫,并运用面板Logit模型和中介效应检验方法,进一步分析了投资者情绪、流动性与行业股价泡沫之间的关系。实证分析及稳健性检验结果均表明,投资情绪与流动性对股价泡沫的存在性具有显著的正向作用,投资者情绪可以通过促进流动性的提升,导致股价泡沫产生的可能性的进一步提高。研究结论对于防范股价泡沫风险具有重要启示,为监管者对加强市场情绪监管提供相关理论依据。

关键词: 投资者情绪, 股价泡沫, 流动性, GSADF检验