A critical barrier to labor mobility arises from institutional constraints that im-pose discriminatory costs on migrants. Using China’s hukou system as a case study,we construct a novel, outcome-based measure of mobility frictions that infers thesediscriminatory costs. We document a systematic relationship between our frictionmeasure, migrants’ decisions to leave behind families (“partial migration”), remit-tances, and expenditure patterns. Our estimated spatial general equilibrium modelencompasses these features and examines how mobility frictions interact with tradeliberalization to shape migration, inequality, and welfare. Trade-exposed regionsbenefft from attracting migrants, while high-friction regions experience muted laborreallocation and smaller welfare gains.

本文重点探讨与人工智能和资本市场相关的一个问题,即如果AI统治算法，证券市场会变怎样？ 本文以哲学上的 “二律背反” 和香农熵理论为方法，分析金融人工智能对证券市场信息效率的影响。 熵作为信息论中的一个重要概念， 被用来衡量信息的混乱程度， 即信息的不确定度；“二律背反” 则强调了一种辩证视角。 经实证研究发现： 人工 智能对于证券市场效率的影响并非二元对立，AI技术降低信息 “噪音”， 提高投资者认知概率（降熵）;但AI算法的同质化、 市场过度反应或其他风险因素会削弱市场效率， 增加市场的信息不对称性 （升熵）。AI对以中美为代表的新兴市场和成熟市场的影响因市场而异， 我们认为人工智能的出现不是一般性技术冲击 而是属于技术 “奇点” 或约瑟夫·熊彼特的 “创造性破坏” 的范畴， 是真正长期将对资本市场产生持续性影响的重大因素， 这也是本文的一个主要研究价值。

Do investors respond to what fund managers say, or how they look saying it? Using 2,000 live-streamed sessions by Chinese ETF managers and multimodal machine learning, we show that managers’ facial expressions, not their words, drive fund flows. A one-standard-deviation increase in positive facial affect raises next-day flows by 0.17pp (260% of mean). Vocal tone shows weak effects; textual sentiment shows none. Critically, facial expressions predict flows but not returns, indicating pure persuasion rather than information transmission. Effects strengthen when investors are emotionally vulnerable (down markets, retail-heavy funds) and persist 2-3 weeks before dissipating. Our findings challenge the emphasis on textual disclosure in finance and raise questions about investor protection as video communication proliferates.

This paper develops an autonomous framework for systematic factor investing via agentic AI. Rather than relying on sequential manual prompts, our approach operationalizes the model as a self-directed engine that endogenously formulates interpretable trading signals. To mitigate data snooping biases, this closed-loop system imposes strict empirical discipline through out-of-sample validation and economic rationale requirements. Applying this methodology to the U.S. equity market, we document that long-short portfolios formed on the simple linear combination of signals deliver an annualized Sharpe ratio of 2.75 and a return of 54.81%. Finally, our empirics demonstrate that self-evolving AI offers a scalable and interpretable paradigm.

This paper examines the impact of Exchange-Traded Fund (ETF) ownership on corporate earnings management. We find that ETF ownership is associated with a significant reduction in earnings management, and this result remains robust across a wide range of endogeneity tests and robustness checks. Further analyses reveal that ETFs exert a pronounced mitigating effect on sales manipulation, production manipulation, and expense manipulation. Mechanism tests indicate that ETFs curb earnings management by improving stock liquidity and strengthening external monitoring. We also find that the influence of ETFs is stronger in private firms, in firms with lower information transparency, and in firms with CEO duality, suggesting that ETFs serve as a more prominent external governance force when internal governance mechanisms are relatively weak. Overall, this study enriches the literature on the economic consequences of ETFs and provides new empirical evidence that financial innovation in emerging markets can help alleviate the information risk faced by investors.

This study investigates the role of contextual cues in memory-based decision-making within high-stakes trading environments. Using trade records from a large Chinese brokerage firm, we provide evidence that both extreme events (COVID-19 quarantines) and everyday contexts (geographic locations) trigger the recall of previously traded stocks, increasing the likelihood of subsequent orders for those stocks. The observed patterns align more closely with similarity-based recall than with alternative channels. Welfare analysis reveals that these memory-induced trades lead to substantial losses for the representative investor's portfolio. We also find evidence at the market level: when the geographical distribution of quarantine risks is recalled, the probability of recalling the cross-sectional stock return-volume distribution from the same day increases by 1.6 percentage points. This study provides evidence from a real-world setting for memory-based theories, particularly similarity-based recall, and highlights a novel channel through which contextual cues affect financial markets.

We examine the information content of the call-put implied volatility spread (IVS) of Shanghai Stock Exchange 50 ETF options. Empirically, the IVS significantly and negatively predicts future SSE50 ETF returns at both weekly and monthly horizons. This predictability is robust both in-sample and out-of-sample, which stands in contrast to prior evidence from the U.S. options market. We explore several potential explanations and show that the IVS is closely linked to the option-cash basis. Its predictability is consistent with the model of Hazelkorn, Moskowitz, and Vasudevan (2023), where the option-cash basis reflects liquidity demand common to both options and underlying equity markets.

随着汽车金融下沉市场的拓展与多源异构数据的爆发，传统信用评分模型在兼顾预测精度与特定场景泛化能力时遭遇瓶颈。本文提出一种基于多模态混合专家模型（Multimodal Mixture of Experts, MMoE）的深度风控框架。该框架依托企业级AI中台，通过动态门控网络（Gating Network）将借款人的结构化征信、非结构化文本语义及动态行为特征智能路由至专属专家网络。基于 LendingClub 公开数据集的实证研究（有效映射汽车金融多模态场景）表明，MMoE 模型在 AUC 与 KS 指标上显著优于 LightGBM 等主流基准模型，且其期望校准误差（ECE）降至 0.015。研究证实，门控路由机制不仅提升了长尾人群的逾期预测准度，更为深度学习在金融领域的应用提供了宏观可解释性视角。本研究为金融机构构建高并发、易扩展的下一代智能风控底座提供了系统性的工程路径与理论支撑。

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.

This paper studies how mutual fund managers update their beliefs as macroeconomic conditions change. Using regulator-mandated reports from Chinese mutual funds, we measure the intensity of belief updating from year-over-year changes in stated outlooks and decompose those updates into macro and micro themes. We show that belief updating is state-contingent: funds with more intensive belief updating shift their narratives toward macro (micro) topics during recessions (expansions) and concurrently reduce (increase) procyclical stock exposures and on-site company visits. This state-contingent belief updating predicts superior performance when matched to prevailing economic conditions, with macro-oriented updates paying off mainly for high-updating funds in recessions and micro-oriented updates paying off more broadly in expansions. Investors recognize this signal of skill, allocating greater flows to these funds, especially when past returns are less informative. Finally, belief updating is stronger for younger managers and for funds from newer, smaller families, consistent with signaling under career and competitive pressures.