Can fully agentic AI nowcast stock returns? We deploy a state-of-the-art Large Language Model to evaluate the attractiveness of each Russell 1000 stock each trading day, starting in April 2025 when AI web interfaces enabled real-time search. Our data contribution is unique along three dimensions. First, the nowcasting framework is completely out-of-sample and free of look-ahead bias by construction: predictions are collected at the current edge of time, ensuring the AI has no knowledge of future outcomes. Second, this temporal design is irreproducible once the information environment passes. Third, our framework is fully agentic: we do not feed the model curated news or disclosures; it autonomously searches the web, filters sources, and synthesises information into quantitative predictions. We find that AI possesses genuine stock-selection ability, but that its predictive power is concentrated in identifying future winners. A daily value-weighted portfolio of the 20 highestranked stocks earns a Fama-French five-factor plus momentum alpha of 19.4 basis points and an annualised Sharpe ratio of 2.68 over April 2025–March 2026. The same portfolio accumulates roughly 49.0% cumulative return, versus 21.2% for the Russell 1000 benchmark. The strategy is economically implementable: the average bid-ask spread of the daily Top-20 portfolio is 1.79 basis points, less than 10% of gross daily alpha. However, the signal remains asymmetric. Bottom-ranked portfolios generally exhibit alphas close to zero, while the strongest predictive content sits in the extreme top ranks. Delayed-entry tests further show that predictability does not vanish after a single day; rather, the signal remains positive over a broad window of subsequent entry dates, consistent with slow information diffusion rather than a fleeting overnight anomaly.

We document a novel "news extrapolation" behavior among investors, which distorts the market reaction to corporate news. Specifically, investors tend to extrapolate the value of past news in the immediate reaction to the newly arrived news. News extrapolation generates a biased price reaction to news, which is completely reversed afterwards. Furthermore, the tendency of news extrapolation is related to the recency, consistency, and value uncertainty of news. Investors extrapolate not only from news of the same category but also from news of different categories. By analyzing the trading behavior and sentiment of different investor groups, we find that retail investors tend to be news extrapolators, while institutional investors trade against the news extrapolators.

Periods of market crisis are often accompanied by heightened fear and information overload, which can induce information avoidance behaviors such as the ostrich effect. While prior research has documented investors’ tendency to avoid unfavorable information, little is known about how different information sources—and trust in those sources—jointly shape such behavior under extreme uncertainty. Drawing on Granular Interaction Thinking Theory (GITT) and employing Bayesian Mindsponge Framework (BMF) analytics, this study examines how investors’ regular securities-related information sources is associated with the ostrich effect during the 2022 market downturn in China, and how these associations are conditioned by trust. Using survey data from 1,451 Chinese individual stock investors, we model investors’ recalled frequency of temporarily disengaging from stock investing as an indicator of information avoidance. The results show that regularly consulting professional sources, financial newspapers, and online forums is associated with information avoidance, whereas reliance on personal relationships and company disclosures is not. Importantly, trust moderates these relationships in distinct ways. Higher trust in professional sources is associated with reduced information avoidance, while higher trust in financial newspapers and online forums amplifies avoidance behavior. Among all sources, the interaction between trust and information referral is strongest for financial newspapers. These findings suggest that trust does not uniformly mitigate fear-driven avoidance. Instead, when combined with high-entropy information sources, trust can exacerbate cognitive and emotional strain, increasing investors’ propensity to disengage. By highlighting the joint roles of informational entropy and trust, this study advances behavioral finance research and offers practical insights for investors, policymakers, and regulators seeking to improve decision-making resilience during periods of market crisis.

We examine how social networks shaped by high-speed rail connections influence investor attention and market reactions to earnings announcements in China. Firms in high-centrality cities exhibit stronger immediate and subsequent responses in investor attention, stock price, and trading volume to earnings news. Further analysis shows that earnings-induced local attention predicts future attention spillovers to intercity investors, amplifying both price and volume reactions after announcements. Overall, these findings indicate that high-speed rail networks foster investor social networks that facilitate the dissemination of firm news and help explain predictable patterns in investor behavior and market pricing.

The growing emphasis on Environmental, Social, and Governance (ESG) principles, particularly in corporate sectors, shapes investment trends and operational strategies, whose shift is supported by the increasing role of media in monitoring and influencing corporate ESG performance, thereby driving the energy innovation. Therefore, based on reported events from Baidu News and patent text information of Chinese A-share listed companies from 2012 to 2022, this study innovatively applied machine learning and text analysis to measure ESG news sentiment and corporate energy innovation indicators. Combing with reputation, stakeholder, and agency theories, we find that a good reputation conveyed by positive ESG textual sentiments in the media significantly promotes corporate energy innovation, and the effect is mainly realized through alleviating financing constraints and agency problems and promoting green investment. Further analysis shows that ESG news sentiment promotes corporate energy innovation mainly among private firms, non-growth-stage firms, high-energy-consuming firms, and regions with better green finance development and higher ESG governance intensity. From the perspective of ESG news content and information content, greater ESG news attention can also exert an energy innovation incentive effect, in which the incentive effect exerted by positive media sentiment in the environmental (E) and social (S) dimensions, as well as excellent attention, is more robust. This study provides new insights for promoting green and low-carbon development and understanding the external governance role of media in corporate ESG development.

This study introduces an innovative approach for constructing multimodal investor sentiment indices and explores their varying impacts on stock market returns. We employ the RoBERTa model to quantify text-based sentiment, the Google Inception(v3) model for image-based sentiment measurement, and a multimodal semantic correlation fusion model to comprehensively consider the interplay between textual and visual sentiment features. These sentiment indices are further categorised into industry-specific investor sentiment and market-wide investor sentiment, enabling separate analyses of their effects on stock markets. Furthermore, we leverage these indices to build a multifactor stock selection model and timing strategies. Our research findings demonstrate that multimodal sentiment analysis yields superior predictive accuracy. Industry-specific investor sentiment exerts bidirectional positive influences on stock market returns, whereas market-wide investor sentiment indices exhibit unidirectional impacts. Integrating industry-specific investor sentiment into our multifactor stock selection model effectively enhances portfolio returns. Furthermore, combining market-wide investor sentiment with timing strategy optimisation further augments this advantage.

We introduce a hypergraph-based approach to analyze information flow and investor attention transfers through news outlets in financial markets. Extending traditional graph models that focus on pairwise interactions, our hypergraph framework captures higher order relationships between firms that are simultaneously mentioned in the same news article. We develop a random walk based centrality framework that considers both the properties of the hyperedges (news articles) and the nodes (firms). This framework allows us to more accurately simulate investor attention flows and to incorporate different theories of investor behavior, such as category learning and investor attention theory. To demonstrate the effectiveness of our attention centrality, we apply it to the Chinese CSI500 market index from 2016 to 2021, where our centrality measures improve the prediction of future returns, with improvements ranging from 6.3% to 14.0% compared to traditional graph-based models. This improvement implies that our centrality measure can better capture investor attention transfers on the news hypergraph. In particular, we find that investors pay more attention to news that covers both a greater number of firms and firms on which the sentiments are more negative. Although we focus on financial markets in this research, our hypergraph framework holds potential for broader applications in information systems — for example, in understanding social or collaboration networks.

Using a comprehensive dataset for Chinese listed companies from 2009 to 2021, we find that online media is negatively associated with cash dividend level, and the proportion of positive news has a negative moderating effect on this relationship. Our results support the "information intermediary" effect and exclude the "external governance" and "market pressure" effects. We further propose that online media weakens the positive relationship between cash dividends and past earnings (rather than the future), indicating that cash dividends contain signals of improvement in past earnings and are replaced by online news. We also find that only firms with more positive news pay dividends that have signaling effects, and there is a synergistic effect between positive news and dividend signal. Additional results show that the effect of online media on dividend policy is more pronounced than traditional media, which has almost no influence. Our main conclusions remain valid after addressing potential endogeneity issues and conducting various robustness tests.

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 study how financial markets integrate news about pollution abatement costs into firm values. Using China’s automation of pollution monitoring, we find that firms with factories in bad-news cities---cities that used to report much lower pollution than the automated reading---see significant declines in stock prices. This is consistent with the view that investors expect firms in high-pollution cities to pay significant adjustment and abatement costs to become “greener.” However, the efficiency with which such information is incorporated into prices varies widely---while the market reaction is quick in the Hong Kong stock market, it is considerably delayed in the mainland ones, resulting in a drift. The equity markets expect most of these abatement costs to be paid by private firms and not by state-owned enterprises, and by brown firms and not by green firms.