We examine how the sudden drop in public information search capability caused by Google’s withdrawal from China affects Chinese analysts’ earnings forecasts. We observe, after Google’s withdrawal, a decline in analysts’ forecast accuracy for firms with foreign trade relative to those without it. This decline suggests that the withdrawal hinders analysts’ acquisition of foreign information about firms, which decreases the quality of their earnings forecasts. We also find that the effect of the withdrawal on forecast accuracy is stronger for firms with higher business complexity and more opaque financial reporting and for analysts with weaker information processing capacity and more attention constraints. Additionally, we find that corporate site visits serve as an alternative information source that can compensate for the information loss caused by the Google withdrawal. Last, we document that the withdrawal reduces analysts’ forecast timeliness and increases their forecast dispersion.

We conceptually identify and empirically verify the features distinguishing FinTech platforms from non-financial platforms using marketplace lending data. Specifically, we highlight three key features: (i) Long-term contracts introducing default risk at both the individual and platform levels; (ii) Lenders’ investment diversification to mitigate individual default risk; (iii) Platform-level default risk leading to greater asymmetric user stickiness and rendering platform-level cross-side network effects (p-CNEs), a novel metric we introduce, crucial for adoption and market dynamics. We incorporate these features into a model of two-sided FinTech platform with potential failures and endogenous participation and fee structures. Our model predicts lenders’ single-homing, occasional lower fees for borrowers, asymmetric p-CNEs, and the predictive power of lenders’ p-CNEs in forecasting platform failures. Empirical evidence from China’s marketplace lending industry, characterized by frequent market entries, exits, and strong network externalities, corroborates our theoretical predictions. We find that lenders’ p-CNEs are systematically lower on declining or well-established platforms compared to those on emerging or rapidly growing platforms. Furthermore, lenders’ p-CNEs serve as an early indicator of platform survival likelihood, even at the initial stages of market development. Our findings provide novel economic insights into the functioning of multi-sided FinTech platforms, offering valuable implications for both industry practitioners and financial regulators.

This study investigates the impact of analysts' information source diversity on forecast bias and investment returns. We combine the GPT-4o model and text similarity, to extract the names of information sources from the text of analyst in-depth reports. Using 349,200 sources, we calculate information diversity scores based on the variety of data sources to measure analysts’ ability of selecting relevant information. The findings reveal that higher information diversity significantly reduces forecast bias and enhances portfolio returns. The effect is particularly pronounced for large companies, state-owned enterprises, those with low analyst coverage, low firm-specific experience, and reports with positive forecast revisions. Institutional investors recognize the value of this skill, while retail investors remain largely unaware, which contributes to financial inequality. This study highlights the critical role of information diversity in analyst performance.

We develop an information choice model where information costs are sticky and co-move with firm-level intrinsic uncertainty as opposed to temporal variations in uncertainty. Incorporating analysts' forecasts, we predict a negative relationship between information costs and information acquisition, as proxied by the predictability of analysts' forecast biases. Finally, the model shows a contrasting pattern between information acquisition and intrinsic and temporal uncertainty, where intrinsic uncertainty strengthens return predictability of analysts' biases through the information cost channel, while temporal uncertainty weakens it through the information benefit channel. We empirically confirm these opposing relationships that existing theories struggle to explain.

Developing an early warning model for company financial crises holds critical significance in robust risk management and ensuring the enduring stability of the capital market. Although the existing research has achieved rich results, the disadvantages of insufficient text information mining and poor model performance still exist. To alleviate the problem of insufficient text information mining, we collect related financial and annual report data from 820 listed companies in mainland China from 2018 to 2023 by using sophisticated web crawlers and advanced text sentiment analysis technologies and using missing value interpolation, standardization, and data balancing to build multi-source datasets of companies. Ranking the feature importance of multi-source data promotes understanding the formation of financial crises for companies. In the meantime, a novel Attention-based Fuzzy Neural Network (AFNN) was proposed to parse multi-source data to forecast financial crises among listed companies. Experimental results indicate that AFNN exhibits significantly improved performance compared to other advanced methods.

We show that the personal traits of analysts, as revealed by their political donations, influence their forecasting behavior and stock prices. Analysts who contribute primarily to the Republican Party adopt a more conservative fore- casting style. Their earnings forecast revisions are less likely to deviate from the forecasts of other analysts and are less likely to be bold. Their stock recommen- dations also contain more modest upgrades and downgrades. Overall, these analysts produce better quality research, which is recognized and rewarded by their employers, institutional investors, and the media. Stock market participants, how- ever, do not fully recognize their superior ability as the market reaction following revisions by these analysts is weaker.

Behavioral theories predict that firm valuation dispersion in the cross-section (‘‘dispersion’’) measures aggregate overpricing caused by investor overconfidence and should be negatively related to expected aggregate returns. This paper develops and tests these hypotheses. Consistent with the model predic- tions, I find that measures of dispersion are positively related to aggregate valuations, trading volume, idiosyncratic volatility, past market returns, and current and future investor sentiment indexes. Disper- sion is a strong negative predictor of subsequent short- and long-term market excess returns. Market beta is positively related to stock returns when the beginning-of-period dispersion is low and this rela- tionship reverses when initial dispersion is high. A simple forecast model based on dispersion signifi- cantly outperforms a naive model based on historical equity premium in out-of-sample tests and the predictability is stronger in economic downturns.

We propose a novel measure, DOT, to capture belief divergence on extreme tail events in stock returns. Defined as the standard deviation of expected probability forecasts generated by distinct information processing functions and neural network models, DOT exhibits significant predictive power for future stock returns. A value-weighted (equal-weighted) long-short portfolio based on DOT yields an average return of -1.07% (-0.98%) per month. Furthermore, we document novel evidence supporting a risk-sharing channel underlying the negative relation between DOT and the equity premium following extreme negative shocks. Finally, our findings are also in line with a mispricing channel in normal periods.

This paper investigates the high-frequency dynamics of macroeconomic expectations and disagreement among professional forecasters. We propose a novel mixed-frequency estimation approach that integrates daily asset returns with quarterly expectation data from the Survey of Professional Forecasters. Our findings indicate that consensus forecasts are updated efficiently according to Bayes' rule, independent of prior forecasts. By employing "representative forecasters" as proxies for real-world agents, we derive a simple yet intuitive evolution equation for disagreement, revealing that changes in disagreement are primarily driven by different interpretations of new information. Furthermore, we reconstruct daily series of expectations and disagreement concerning macroeconomic growth, achieving impressive R2 values of 93.3% and 84.5% against the true quarterly series.

We examine whether large language models (LLMs) can extract contextualized representation of Chinese public news articles to predict stock returns. Based on representativeness and influences, we consider seven LLMs: BERT, RoBERTa, FinBERT, Baichuan, ChatGLM, InternLM, and their ensemble model. We show that news tones and return forecasts extracted by LLMs from Chinese news significantly predict future returns. The value-weighted long-minus-short portfolios yield annualized returns between 35% and 67%, depending on the model. Building on the return predictive power of LLM signals, we further investigate its implications for information efficiency. The LLM signals contain firm fundamental information, and it takes two days for LLM signals to be incorporated into stock prices. The predictive power of the LLM signals is stronger for firms with more information frictions, more retail holdings and for more complex news. Interestingly, many investors trade in opposite directions of LLM signals upon news releases, and can benefit from the LLM signals. These findings suggest LLMs can be helpful in processing public news, and thus contribute to overall market efficiency.