Global and Extreme Dependence Between Investor Sentiment and Stock Returns in European Markets

The authors investigate the global and extreme dependence structure between investor sentiment and stock returns in 7 European stock markets (Belgium, France, Germany, Greece, the Netherlands, Portugal, and the UK), over the period 1985–2015. Global dependence refers to the correlation of changes in sentiment and stock returns over the whole range of these 2 variables, and extreme dependence refers to the local correlation of high (i.e. asymptotic) changes in sentiment and high stock returns. Using copula models and a bootstrap procedure, 6 statistical tests are performed for this purpose. Among the results of the tests, the authors highlight those that provide evidence of contemporaneous lower extreme dependence and contemporaneous upper extreme independence between sentiment and returns. As policy implications, these results suggest that financial stability can be promoted if regulators consider the impact of their decisions on investor sentiment. Also, the results seem to support the arguments in favor of short selling ban during turmoil periods. Finally, overall, the results are relevant for both investors and regulators and reinforce the importance of considering investor sentiment to better understand the behavior of financial markets.     

Comparing automated text classification methods

Online social media drive the growth of unstructured text data. Many marketing applications require structuring this data at scales non-accessible to human coding, e.g., to detect communication shifts in sentiment or other researcher-defined content categories. Several methods have been proposed to automatically classify unstructured text. This paper compares the performance of ten such approaches (five lexicon-based, five machine learning algorithms) across 41 social media datasets covering major social media platforms, various sample sizes, and languages. So far, marketing research relies predominantly on support vector machines (SVM) and Linguistic Inquiry and Word Count (LIWC). Across all tasks we study, either random forest (RF) or naive Bayes (NB) performs best in terms of correctly uncovering human intuition. In particular, RF exhibits consistently high performance for three-class sentiment, NB for small samples sizes. SVM never outperform the remaining methods. All lexicon-based approaches, LIWC in particular, perform poorly compared with machine learning. In some applications, accuracies only slightly exceed chance. Since additional considerations of text classification choice are also in favor of NB and RF, our results suggest that marketing research can benefit from considering these alternatives.     

Internal pyramid structure,contract enforcement,minority investor protection,and firms’ performance: Evidence from emerging economies

This study investigates the relationship between internal pyramid structure and performance of Chinese, Pakistani, Malaysian pyramidal firms, the effect of judicial efficiency and minority investor protection on this relationship. The results show that the pyramid structure of Pakistani firms is more complicated than Chinese and Malaysian firms, both vertically and horizontally. The study finds that the impact of control layers on performance is negative and stronger than control chains. Moreover, the results illustrate that the effect of control layers on performance at Chinese firms is negative but lower than at Pakistani and Malaysian firms. However, control chains have insignificant association with performance at Chinese pyramid firms. We find that efficient judiciary abates the negative impact of control layers and chains on performance. Our results reveal that in the absence of efficient courts the minority investors’ protection have insignificant impact on the association between internal pyramid structure and firms’ performance.     

The impact of shareholders and creditors rights on IPO performance: An international study

This paper examines the impact of cross-country variation in shareholders' and debt holders' rights on post-IPO performance and survival of newly listed stocks across the globe. Using a sample of 10,490 initial public offerings (IPOs) in 40 countries between 2000 and 2013, we find that post-IPO performance and survival is better in countries with stronger shareholder protection, but the impact of creditor protection is negative i.e. stronger creditor protection leads to poor post-IPO performance and survival. This effect is driven by rules requiring creditors’ consent for company reorganization and the mandatory replacement of incumbent managers. Reputable IPO advisors exacerbate the positive impact of shareholder rights and the negative impact of creditor rights.     

Islamic and conventional portfolios optimization under investor sentiment states: Bayesian vs Markowitz portfolio analysis

This paper investigates the portfolio optimization under investor’s sentiment states of Hidden Markov model and over a different time horizon during the period 2004–2016. To compare the efficient portfolios of the Islamic and the conventional stock indexes, we have employed two approaches: the Bayesian and Markowitz mean-variance. Our findings reveal that the Bayesian efficient frontier of Islamic and conventional stock portfolios is affected by the investor’s sentiment state and the time horizon. Our findings also indicate that the investor’s sentiment regimes change the Islamic and the conventional optimal diversified portfolios.Moreover, the results show that the potential diversification benefits seem to be more important when using the Bayesian approach than when applying the Markowitz approach. This finding is valid for the bearish, depressed, bullish and calm states in Islamic stock markets. However, the diversification of potential portfolios is significant only for the bullish and the bubble states in the conventional financial markets.The findings of the study provided additional evidence for investors to exploit googling investor sentiment states to evaluate the portfolio performance and make an optimal portfolio allocation.     

Individual stock sentiment beta and stock returns

Using the data in Chinese stock market, we measure the individual stock sentiment beta, which is defined as the sensitivity of individual stock returns to the individual stock sentiment changes. We demonstrate that stocks in the highest individual stock sentiment beta portfolio have significantly higher excess returns, CAPM alpha, Fama-French three-factor alpha and Fama-French five-factor alpha. Besides, we find that the high individual stock sentiment beta stocks are smaller, younger, more volatile stocks with higher price and higher market beta. After controlling for firm characteristic, the returns of High-Low individual stock sentiment beta portfolios are still significantly positive. Moreover, we show the effect of the individual stock sentiment beta on stock returns is positive and significant in different stock markets, in different sample periods, and in bull and bear market. Besides, the results of the Bayes-Stein individual stock sentiment beta are still stable.     

央行货币政策报告文本信息、宏观经济与股票市场

本文利用金融情感词典和文本分析技术,分析中国人民银行货币政策执行报告的文本情绪、文本相似度和文本可读性等多维文本信息,刻画央行货币政策执行报告的文本特征,探究货币政策报告的文本信息与宏观经济和股票市场的关系。实证研究发现,货币政策报告的文本情绪的改善会引起显著为正的股票市场价格反应,报告文本相似度的增加会引起股票市场波动性的显著降低,报告可读性对公布后股票市场的波动性影响不显著。货币政策报告文本情绪还与诸多宏观经济指标显著相关。进一步研究发现,引起股票市场显著反应的是报告文本情绪中反映货币政策指引的部分,而反映宏观经济历史状态的部分对股票市场的影响不显著。本文从文本大数据分析角度证明了我国央行沟通的有效性,对国内央行沟通相关研究形成了有益补充。     

The impact of investor structure on stock price crash sensitivity: Evidence from China's stock market

Stock price crash sensitivity refers to the conditional probability of a stock crash when the market collapses. It focuses on individual stocks' sensitivity to the market crash and can affect stock pricing significantly. Although the crash sensitivity of China's stock market is very high as a whole (Weigert, 2016), different individual stocks show varying degrees of crash sensitivity. This paper, adopting the perspective of institutional investors, explores the reasons for the difference in crash sensitivity in China's stock market, and finds that: First, institutional investors' shareholdings is positively related to firms' stock price crash sensitivity. However, after dividing institutional investors into professional (represented by financial institutions) and non-professional institutional investors (represented by general legal persons), we find that only professional institutional investors' shareholdings is negatively related to firms' stock price crash sensitivity. Second, the impact of professional institutional investors on the crash sensitivity is influenced by stock liquidity and media sentiment: when the stock liquidity of listed companies is good or the media sentiment is strong, the negative impact of professional institutional investors on the crash sensitivity is accordingly high. This paper, by highlighting the investor structure, attempts a pioneering exploration of the influencing factors of the difference in stock price crash sensitivity in China. Our empirical results enrich research on stock price crash sensitivity and the heterogeneity of institutional investors. They can also serve to guide regulatory authorities' development of institutional investors and efforts to maintain market stability.     

Comparative study of deep learning models for analyzing online restaurant reviews in the era of the COVID-19 pandemic

Online reviews remain important during the COVID-19 pandemic as they help customers make safe dining decisions. To help restaurants better understand customers’ needs and sustain their business under current circumstance, this study extracts restaurant features that are cared for by customers in current circumstance. This study also introduces deep learning methods to examine customers’ opinions about restaurant features and to detect reviews with mismatched ratings. By analyzing 112,412 restaurant reviews posted during January-June 2020 on Yelp.com, four frequently mentioned restaurant features (e.g., service, food, place, and experience) along with their associated sentiment scores were identified. Findings also show that deep learning algorithms (i.e., Bidirectional LSTM and Simple Embedding + Average Pooling) outperform traditional machine learning algorithms in sentiment classification and review rating prediction. This study strengthens the extant literature by empirically analyzing restaurant reviews posted during the COVID-19 pandemic and discovering suitable deep learning algorithms for different text mining tasks.     

The impact of COVID-19 on the stock market crash risk in China

This study investigates the impact of the COVID-19 pandemic on the stock market crash risk in China. For this purpose, we first estimated the conditional skewness of the return distribution from a GARCH with skewness (GARCH-S) model as the proxy for the equity market crash risk of the Shanghai Stock Exchange. We then constructed a fear index for COVID-19 using data from the Baidu Index. Based on the findings, conditional skewness reacts negatively to daily growth in total confirmed cases, indicating that the pandemic increases stock market crash risk. Moreover, the fear sentiment exacerbates such risk, especially with regard to the impact of COVID-19. In other words, when the fear sentiment is high, the stock market crash risk is more strongly affected by the pandemic. Our evidence is robust for the number of daily deaths and global cases.