This paper investigates the predictive power of news topics for stock returns in the Chinese equity market. The Quantile Auto-Encode (QAE) model is innovatively employed to extract latent factors embedded in media news, addressing challenges such as heavy-tailed distributions, conditional heteroskedasticity, and quantile heterogeneity in asset pricing. Using monthly Chinese stock return data from April 2005 to December 2022, the QAE model significantly outperforms both the IPCA and AE models in out-of-sample evaluations, achieving higher total and predictive R-squared values as well as improved annualized Sharpe ratios. The narrative-based factors estimated by these models exhibit smaller pricing errors than those from traditional asset pricing models, indicating superior accuracy in capturing the systematic risk structure. Moreover, topics concerning firms' business activities, operations, strategies, and profitability exhibit stronger pricing power.

We study whether machine learning (ML) forecasts can enhance option portfolio performance by relaxing strict delta neutrality. We propose a confidence-scaled hedging framework that dynamically adjusts hedge ratios according to the classification results of ML models. Using option and underlying ETF data, we find that moderate confidence scaling improves Sharpe ratios relative to a benchmark, while aggressive scaling increases volatility and weakens long-term returns. The results highlight that ML forecasts can be translated into economically meaningful improvements in derivatives trading and risk management.

We use neural networks to model portfolio weights as nonlinear functions of firm characteristics. Applying our method to the equity market in the United States excluding micro-cap stocks, we find the long-only portfolio from the network with four hidden layers performs best and tends to have investments in stocks exhibiting low risk, high value, and high profitability, the characteristics that contribute significantly to the out-of-sample performance. While more complex networks are associated with increased turnovers and elevated trading costs, we find that autoregressive smoothing policies are effective in reducing portfolio turnover and improving model performance.

We introduce a novel decomposition of stock market returns into a fundamental component (FC), capturing long-term growth, and an unexpected capital gain component (UC), reflecting short-term fluctuations. The decomposition is explicit, straightforward to estimate, and designed to correct model misspecification in traditional predictive regressions. The FC aligns with highly persistent valuation ratios, while the UC is predictable via standard regressions, further improved by shrinkage. Using 41 predictors from Goyal et al. (2024), we find 33 significant predictors of market returns, compared with only 5 in their study. Combining forecasts across predictors reaffirm equity risk premium predictability out-of-sample.

Major events can have a lasting impact on the financial markets and affect the temporal aggregation of tail risk. We capture the dynamics of jump intensity using a generalized autoregressive conditional heteroskedasticity with autoregressive jump intensity (GARCH-ARJI) model, derive analytical formulas for the first four moments of cumulative returns, and utilize them to calculate VaR based on the Johnson distribution method. Our numerical experiments reveal that skewness decreases sharply while excess kurtosis rises in the short term, particularly when initial jump intensity is high. In the long term, the time diversification effect causes skewness and excess kurtosis to converge slowly to zero. Our out-of-sample backtesting analysis on S&P 500, FTSE 100, and DAX 30 total return indexes shows that it is important to incorporate the time-varying jump intensity when forecasting tail risk. (c) 2025 International Institute of Forecasters. Published by Elsevier B.V. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

This paper examines the relationship between news content and stock price movement in the Chinese stock market, proving that higher textual similarity of firm news to peers is accompanied with less idiosyncratic information and stronger stock return synchronicity. Our findings remain robust after applying the firm fixed effects, using the PSM method, expanding sample windows, and introducing instrumental variables. Additionally, the effect is pronounced for firms with poor information environments and high investor attention, in bull markets and under conditions of lower uncertainty. The effect varies depending on the authority of the news publishers and the themes of the news narratives.

This paper examines the effect of news topic concentration (TCR) on Chinese stocks. Our empirical findings reveal a U-shaped relationship between TCR and market reaction, which is mediated by investor attention. Information ambiguity complements attention’s role at low TCR. Moreover, TCR has a greater effect on stock prices when firms have lower institutional ownership, higher information uncertainty, and when the news originates from state-controlled media. Additionally, we observe that the U-shaped patterns differ across dominant meta-themes. Our results highlight the importance of the structure of the information present in its integration into asset prices and emphasize the role of journalists as key information intermediaries.

This paper investigates the impact of smartphone trading on investor attention allocation. We find a positive connection between smartphone trading ratio and return comovement. Smartphone traders relatively pay less attention to firm-specific information and more to information at the market level than computer-based traders. The impact of smartphone trading is positively correlated to the Internet searching level but negatively correlated to social media discussion intensity and analyst following. We also show that market boom and crush would increase the effect of smartphone trading on attention allocation.

This paper investigates the effect of retail investor attention on market reaction following analysts' recommendation revisions in the Chinese stock market. The results suggest that the impact of retail attention on post-recommendation revision drift is asymmetric. Specifically, retail attention mitigates post-upgrade announcement drift, whereas it aggravates post-downgrade drift. After a series of arrangements to address the potential endogeneity concerns and ensure robustness, the results still hold. Moreover, we reveal retail attention facilitates the absorption of individual firms' information and increases liquidity to attenuate post-upgrade drift, whereas retail attention induces underreaction due to short-sale restrictions and disposition effects regarding stronger post-downgrade drifts. The results of the additional heterogeneity test provide further evidence of channel tests and reveal that prior positive sentiment can accelerate the integration of good news into stock prices. These findings enrich the existing literature on retail investors' role in the price discovery process following the release of fundamental information.

This paper examines how mobile device usage affect retail investors’ lottery behaviour. Using data from the Chinese stock market, which is dominated by retail investors, we observe a structural change in mobile investing trends in 2015. By employing a structural shift model, we find that higher mobile device usage in the market disproportionately increases the trading volume of lottery stocks and amplifies the lottery-related pricing anomaly. Our results suggest that the use of mobile devices increases retail investors’ demand for lottery stocks by attention-induced trading, but fail to find evidence that mobile devices influence retail investors’ inherent skewness preferences.

Satellites can “see” oil inventory in oil tanks, but they are sensitive to cloud cover. Cloud cover introduces a new uncertainty related to information quality. We measure such information uncertainty by assessing cloud cover over floating roof oil tanks. Using a cloud cover index, we demonstrate that higher information uncertainty leads to lower future returns (mean effect) and a stronger momentum anomaly (interaction effect). These two effects can be explained by investor overconfidence and arbitrage costs, respectively. An investor with a mean–variance preference obtains sizable gains in terms of certainty equivalent return, which accounts for the mean effect.

This study investigates the impact of remote meetings on the participation of female analysts in corporate visits. We find that online visits significantly increase female analysts' participation, especially under greater security risks or work–family conflicts. Additionally, our analysis shows that an increase in the number of female analysts enhances communication between analysts and management and improves forecast performance following corporate visits. These findings suggest that online visits can effectively boost female analysts' participation and enhance their work quality.

Excessive supply chain concentration (SCC) introduces various risks, and the factors contributing to SCC are complex and have been rarely explored. This study examines the economic impact of digital transformation on supply chains by analysing data from China's A-share listed companies (2010-2021) using regression models. The findings demonstrate that digital transformation significantly reduces supply chain concentration and increases the firms' influence within the supply chain. The mechanism analysis shows that digital transformation makes firms more attractive to suppliers and customers, eases financing constraints, boosts firm reputation, and reduces SCC. The impact is more pronounced in firms facing higher environmental uncertainty or those positioned upstream in the supply chain. Except for blockchain, all five dimensions of digital transformation help to reduce SCC, thereby improving total factor productivity and economic growth. Digital transformation also creates a spillover effect, raising expectations for digital transformation and productivity among suppliers and customers. These findings have significant implications for understanding digital transformation practices and advancing supply chain management. The conclusions remain robust after various tests, including changes to the measures of variables and controls for reverse causality using lagged treatment and instrumental variables.

This paper investigates how the trading behavior of investors using mobile devices differs from that of investors using PCs in the context of limited attention. We hypothesize that mobile device-based trading is associated with a stronger “ranking effect” that investors disproportionally trade top-ranked stocks. By exploiting the exogeneity of the return-based ranking of price limit events, we examine the ranking effect on retail investors' buying behavior and find supportive evidence that (1) based on the absolute returns of price limit events, top-ranked stocks are more heavily bought than the lower ranks by investors using mobile devices, while there is no significant ranking effect for investors using PCs; and (2) the ranking effect is enhanced with the increase of the number of contemporaneous price limit events.

The promising empirical results presented using high-frequency data show that the log-volatility behaves essentially as a fractional Brownian motion (fBm) with a Hurst exponent smaller than 0.5. Motivated by these findings, we propose the autoregressive rough volatility (ARRV) model, which combines the fractional Gaussian noise (fGn) process and time series models to forecast volatility. We apply this model to the VIX index by adopting the fBm approximation technique, and our results indicate that the ARRV model can significantly improve VIX out-of-sample forecast accuracy, particularly during turbulent times.

This study examines the association between analysts’ site visits and stock price crash risk using a dataset of Chinese firms listed on the Shenzhen Stock Exchange (SZSE) from 2012 to 2019. We find that analysts’ site visit frequency is positively associated with future stock price crash risk, and the positive association is primarily driven by the firms that analysts keep silent on. Furthermore, we confirm that analysts’ silence is associated with negative news about firms’ fundamentals. We also show that the effect of analysts’ silence on crash risk is more pronounced when analysts face conflicts of interest, when firms have poorer information transparency, and when firms have fewer alternative information channels. Overall, our findings shed light on the dark side of analysts’ site visits.

There is increasing attention on information transfers along supply chain partners for firm (extreme) events. This growing literature finds spillover effects following certain types of firm events. Using data from credit rating actions of Chinese-listed firms over the period between March 2007 and May 2020, we examine the spillover effects of supply chains by focusing on the market reactions of event firms to the action announcements. We find strong evidence of spillover effects driven by the market reactions of event firms, which are enhanced through information diffusion channels as supply chain partners receive more investor attention. Moreover, the effects are stronger when event firms' market reactions are negative, event firms are non-stated-owned, the industry concentration of event firms is higher, or the supplier-customer business relationship is closer. Overall, these findings highlight the role of investor attention and network characteristics in supply chain spillovers.

With the data of listed companies in China’s A-share market,this paper examines the relationship between smartphone trading and herding behavior when market price experiences large fluctuate.Smartphone trading enables investors to respond to information rapidly,but its information processing is not efficient as computer trading due to the device and function limit.Besides,smartphone users are more susceptible to the interference of information and emotions from mobile social media and mobile social platforms.It is found that smartphone trading enhances herding behavior under market pressure in both bullish and bearish sentiment,and the effect is more pronounced in bullish sentiment;the effect is more significant when investors have more informationsearching activities.With further exploration,it is found that smartphone trading and intra-day market volatility have a compounding effect on herding behavior.