We propose a time-varying framework for tail risk contagion based on conditional higher co-moments (Co-HCM), derived from a DCC-GARCH-MGH model that provides closed-form expressions for dynamic co-moments. Applying this CoHCM approach, we construct tail contagion networks across Belt and Road Initiative (BRI) stock markets. Our ffndings indicate that covariance-based metrics underestimate the ex-tent of epidemic transmission, while the CoHCM metrics reveal China’s pivotal role in spreading outbreaks and identify a distinct cluster of core transmission hubs, particularly during the 2015 Chinese stock market crisis. Dynamic contagion further exhibits cross-country heterogeneity that the Southeast Asian markets synchronize tightly with China during crises, while smaller and resource-driven markets display more inter-mittent contagion patterns. These ffndings highlight the importance of higher co-moment dependence for monitoring systemic risk in interconnected emerging markets.

We examine how investors who experienced paper gains during a bubble-crash episode, deemed as investors “escaping from death”, adjusted their future risk-taking. Using detailed transaction-level data and a quasi-experiment based on an unanticipated government intervention in the 2007–08 Chinese stock market, we find that investors who “escaped from death” reduce risk-taking behaviors over the next five years. The evidence shows that the change in risk taking is likely at-tributable to reference-dependent preferences. However, the effect diminishes over time and investors “escaping from death” do not exhibit a diminished tendency toward risk-taking when confronted with a stock market bubble crash again.

This study applies large language models (LLMs) to measure technological links and examines its predictive power in the Chinese stock market. Using the BAAI General Embedding (BGE) model, we extract semantic information from patent textual data to construct the technological momentum measure. As a comparison, the measure based on traditional International Patent Classification (IPC) is also considered. Empirical analysis shows that both measures significantly predict stock returns and they capture complementary dimensions of technological links. Further investigation through stratified analysis reveals the critical role of investor inattention in explaining their differential performance: in stocks with low investor inattention, IPC-based measure loses its predictive power while BGE-based measure remains significant, indicating that straightforward information is fully priced in while complex semantic relationships require greater cognitive processing; in stocks with high investor inattention, both measures exhibit predictability, with BGE-based measure showing stronger effects. These findings support behavioral finance theories suggesting that complex information diffuses more slowly in markets, especially under significant cognitive constraints, and demonstrate LLMs’ advantage in uncovering subtle technological connections that traditional methods overlook.

While many studies document cross-stock predictability where returns of some stocks predict returns of other similar stocks, most evidence comes from US markets. Following Chen et al. (2019), we identify peer firms based on historical return similarity and construct a Peer Return Gap (PRG) measure, defined as the difference between a stock’s lagged return and its peers’ returns. Our empirical evidence from Chinese markets shows that past-return-linked peers strongly predict focal firm returns. A long-short portfolio sorted on PRG generates an equal-weighted monthly return of 1.26% (t = 3.81) and a Fama-French five-factor alpha of 1.10% (t = 2.86). These abnormal returns remain unexplained by several alternative factor models.

Contrary to the US, Chinese stock markets exhibit negative overnight returns that appear to be highly affected by the extent of information asymmetry. China's "T+1" trading rule, which prohibits same-day selling, exacerbates adverse selection for uninformed buyers by limiting them to react to post-trade information. An information asymmetry-driven price discount thus emerges at market open, generating negative overnight returns, which further decrease with information asymmetry. Consistent with adverse selection, empirical evidence reveals lower overnight returns during market declines and high-volatility periods, with robust negative relationship between overnight returns and information asymmetry proxied by firm size, analyst coverage, and earnings announcement proximity. A model is introduced to rationalize our findings. This framework also sheds light on China's "opening return puzzle", the phenomenon that prices rise rapidly in the initial 30 minutes of trading, by showing how reduced adverse selection enables rapid price recovery during opening session.

Based on 5-minute high-frequency trading data, we examine the time-varying causal relationship between herding behavior and multiscale spillovers (return, volatility, skewness, and kurtosis) in the Chinese stock market. We employ the novel time-varying Granger causality test proposed by Shi et al. (2018), which is based on the recursive evolving algorithm developed by Phillips et al. (2015a, 2015b), to identify real-time causal relationships and capture possible changes in the causal direction. Our findings reveal a strong relationship between herding and spillover effects, particularly with odd-moment (return and skewness) spillovers. For most of the study period, a bidirectional causal relationship was found between herding and odd-moment spillovers. These results imply that herding behavior is a key driver of spillover effects, especially return and skewness spillovers, which are primarily transmitted through the information channel. By contrast, volatility and kurtosis spillovers are more strongly driven by real and financial linkages. Furthermore, spillover effects also affect herding behavior, highlighting the intricate feedback loop between investor behavior and risk transmission.

We investigate the relationship between the intensity of intraday return reversals and future stock returns in the Chinese stock market. We find that a high frequency of positive overnight returns followed by negative daytime returns predicts one-month ahead returns positively. The analysis shows that daytime retail investors tend to overly sell their own rising stocks at market open, accepting lower stock prices in exchange for liquidity. As the price pressure attenuates, these stocks experience subsequent price increases, implying a positive relationship between return reversals and future returns.

We examine whether large language models (LLMs) can extract contextualized representation of Chinese news articles and predict stock returns. The LLMs we examine include BERT, RoBERTa, FinBERT, Baichuan, ChatGLM and their ensemble model. We find that tones and return forecasts extracted by LLMs from news significantly predict future returns. The equal- and value-weighted long minus short portfolios yield annualized returns of 90% and 69% on average for the ensemble model. Given that these news articles are public information, the predictive power lasts about two days. More interestingly, the signals extracted by LLMs contain information about firm fundamentals, and can predict the aggressiveness of future trades. The predictive power is noticeably stronger for firms with less efficient information environment, such as firms with lower market cap, shorting volume, institutional and state ownership. These results suggest that LLMs are helpful in capturing under-processed information in public news, for firms with less efficient information environment, and thus contribute to overall market efficiency.

We utilize the Bayesian model scan approach to examine the best performing models in a set of 15 factors discovered in the literature, plus principal components (PCs) of anomalies unexplained by the initial factors in the Chinese A-share market. The Bayesian comparison of approximately eight million models shows that HML, MOM, IA, EG, PEAD, SMB, VMG,PMO, plus the four PCs, PC1, PC6, PC7, PC8 are the best supported specification in terms of marginal likelihoods and posterior model probabilities. We also find that the best model outperforms existing factor models in terms of pricing tests and out-of-sample Sharpe ratio.

This study highlights asymmetries in liquidity risk pricing from the perspective of heterogeneously informed traders facing changing levels of preference uncertainty. We hypothesize that higher illiquidity premium and liquidity risk betas may arise simultaneously in circumstances where investors are asymmetrically informed about their trading counterparts’ preferences and their financial firms’ timely valuations of assets . We first test the time-varying state transition patterns of IML, a traded liquidity factor of the return premium on illiquid-minus-liquid stocks, using a Markov regime-switching framework. We then investigate how the conditional price of the systematic risk of the IML fluctuate over time subject to changing levels of preference uncertainty. Empirical results from the Chinese stock market support our hypotheses that investors’ sensitivity to the IML systematic risk conditionally increase in times of higher preference uncertainty as proxied by the stock turnover and order imbalance. Further policy impact analyses suggest that China’s market liberalization efforts, contingent upon its recent stock connect and margin trading programs, reduce the conditional price of liquidity risk for affected stocks by helping the incorporation of information into stock prices more efficiently. Tighter macroeconomic funding conditions, on the contrary, conditionally increase the price of liquidity that investors require.