This study investigates how the interaction of political accountability and local officials’ career incentives shapes the market for Municipal Corporate Bonds (MCBs) in China, taking the 2017 local government debt personal responsibility rule as a quasinatural experiment. We develop a stylized incomplete-information bargaining model to analyze how the rule reshapes the bargaining equilibrium by rendering officials’ observable characteristics credible signals of bailout incentives. Using a dataset of prefecture-level MCBs from 2008 to 2020, we empirically test the model’s predictions and focus on separating officials’ incentive effects from their inherent ability. Our core findings show that post-announcement of the rule, each additional year of a local party secretary’s remaining time to retirement, a proxy for bailout incentives, reduces MCB spreads by approximately 2.5 basis points and increases issuance volume by about 2.0%. These effects are significantly amplified in fiscally stressed cities. Notably, under the 2017 rule, cities led by party secretaries with stronger bailout incentives can expand MCB issuance, which is contrary to the rule’s original intent to rein in local borrowing.

This paper constructs a call pressure index (CP) from China's SSE 50 ETF option market and finds a robust U-shaped nonlinear predictability for directional option returns as measured by log returns. The effect reflects that extreme call pressures—whether unusually low (reversal) or high (momentum)—contain information, while moderate levels are dominated by noise trading. Robustness checks using delta-hedged returns confirm that predictability stems primarily from directional exposure rather than volatility dynamics. The predictability is stronger in high-volatility and down-market states and survives controlling for implied skewness, variance risk premium, and other common predictors. A simple timing strategy based on rolling-window forecasts achieves a Sharpe ratio of 0.97, which further increases to 2.43 after applying a prediction threshold. A parsimonious volume-based indicator captures unique predictive information beyond complex proxies, offering a feasible path for emerging markets lacking proprietary order flow data.

Why do fertility subsidies consistently fail in China? Why do consumptionand fertility collapse globally despite intact household book wealth? This paper proposes the Family Long Cycle Hypothesis (FLCH), extending thedecision-making unit of the life cycle hypothesis from an individual to an intergenerational family network, and expanding the budget constraint from a singlelifetime resource constraint to a dual constraint of “total resources + liquiditystructure”. The core mechanism is: intergenerational “blood-transfusion” homepurchase locks in network liquidity without changing household net assets, completing a balance sheet morph of “book wealth unharmed, decision-making paralyzed” at the moment of purchase. This liquidity depletion spikes effective riskaversion, forming a multiplier effect with income uncertainty, causing fertility—the irreversible commitment with the longest duration—to enter the corner solution region first. Within this region, the elasticity of fertility decisions to costsubsidies is strictly zero, but they remain highly sensitive to liquidity repairand uncertainty reduction. Consequently, this paper proves the fundamentalmechanism difference between consumption subsidies (cash rewards, childcarefee waivers) and capital transfers (mortgage principal write-down, unconditionalcash transfers), and proposes three effective policy directions: reducing incomeuncertainty, repairing family liquidity, and raising the reference income of thebottom 90% of the population. The model nests the standard life cycle hypothesis as a special case and is distinguishable from the competitive saving hypothesis on six pairs of opposing predictions. The theory also explains theasymmetric “fast-falling, slow-rising” adjustment of housing prices: the drop isdriven by defensive behavior (business cycle scale), while the recovery is constrained by intergenerational liquidity reconstruction (intergenerational scale of10–20 years).

We investigate whether overnight news sentiment adds predictive value for overnight returns. We focus on the CSI300 Index, whose ETFs are widely held by Chinese retail investors. Sentiment indi-cators are constructed from minute-level overnight news using a fine-tuned RoBERTa model. These indicators are combined with market-based variables to predict overnight returns via regression and machine learning. Results show that while the sentiment alone has predictive value, its incremental contribution disappears once the A50 overnight return is included.

This paper studies missing firm characteristics in the Chinese stock market and their implications for empirical asset pricing. Relative to the U.S. market, missing firm characteristics in China remain underexplored despite substantial differences in data availability and disclosure environments. Using a dataset of 106 firm characteristics from 1992 to 2021, we document a pronounced cliff-shaped pattern in missingness, with missing rates falling sharply after 2000. We then compare expectation-maximization (EM) and mean imputation (MN) in both univariate characteristic-sorted portfolios and machine-learning applications that combine many predictors. Results indicate that, in univariate analysis, the two methods produce very similar return spreads because they assign largely the same stocks to the extreme deciles. In machine-learning applications, however, EM-imputed data generally produce better-performing prediction-sorted portfolios than mean-imputed data. These findings provide new evidence on missing firm characteristics in a major emerging market and highlight the importance of imputation choices in machine-learning asset-pricing applications.

We document “foreign bias” in AI financial predictions, reversing the classic home bias. U.S.-based ChatGPT is systematically more optimistic than China-based DeepSeek about Chinese firms—in price predictions and directional forecasts—yet significantly less accurate. Evidence supports an information-availability mechanism: bias is strongest when U.S. media coverage of Chinese firms is limited and attenuates for cross-listed firms. Crucially, injecting Chinese news eliminates the prediction gap. Both models produce similar forecasts for U.S. firms, consistent with broader worldwide coverage. LLMs trained in different information environments can create divergent signals, with implications for investors and policymakers as AI increasingly intermediates global markets.

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.

The Hurst exponent is a key parameter for characterizing the long memory of high-frequency time series. However, traditional estimators often exhibit systematic biases due to the influence of high-frequency endogenous noise and low-frequency trend shifts. Theoretical derivations show that endogenous noise contemporaneously correlated with the latent signal possesses a spectral density in the first-differenced series that is asymptotically equivalent to a squared sine functional form. Accordingly, the proposed estimator incorporates a corresponding spectral density component to fit the high-frequency error. Simultaneously, the model introduces a SCAD penalty term to control the low-frequency spectral divergence caused by structural breaks, thereby mitigating spurious long memory in parameter estimation. Monte Carlo simulations demonstrate that the Penalized Mixture Whittle estimator yields smaller finite-sample biases and root mean square errors in scenarios involving both trend disturbances and endogenous noise. Empirical analysis shows that the estimates obtained using this method are robust to changes in sampling frequency. In further volatility forecasting experiments on commodity futures, the linear forecasting model constructed based on the parameter set achieves higher prediction accuracy than benchmark models such as HAR, as confirmed by the Diebold-Mariano test. This paper provides an effective econometric tool for high-frequency data inference in the presence of composite statistical disturbances.

This study develops a novel deep learning framework for bank distress prediction, designed to overcome the limitations of static network analysis and to enhance model interpretability. We propose a Spatio-Temporal Attention Network that uniquely captures the time-varying nature of systemic risk. Methodologically, it introduces two key innovations: (1) a dynamic interbank network whose connection weights are adjusted by the volatility of the Shanghai Interbank Offered Rate (SHIBOR), reflecting real-time market liquidity changes; and (2) a dual spatio-temporal attention mechanism that identifies critical time steps and pivotal contagion pathways leading to a distress event. Empirical results demonstrate that the model significantly outperforms traditional benchmarks across key metrics including accuracy and F1-score. Most critically, the architecture proves exceptionally effective at reducing Type II errors, substantially minimizing the failure to identify at-risk banks. The model also offers high interpretability, with attention weights visualizing intuitive risk evolution patterns. We conclude that incorporating dynamic, liquidity-adjusted networks is crucial for superior predictive performance in systemic risk modeling.

This study proposes an innovative CPO-VMD-PConv-Informer framework to forecast the KBW Nasdaq Financial Technology Index (KFTX). The framework comprehensively incorporates the effects of eight representative uncertainty indicators on KFTX price predictions, including the Economic Policy Uncertainty Index (EPU) and the Geopolitical Risk Index (GPR). The empirical findings are as follows: (1) The proposed CPO-VMD-PConv-Informer framework demonstrates superior predictive performance across the entire sample period, achieving R² values of 0.9681 and 0.9757, significantly outperforming other commonly used traditional machine learning and deep learning models. (2) By integrating VMD decomposition and CPO optimization, the model effectively enhances its adaptability to extreme market volatility, maintaining stable predictive accuracy even under structural shocks such as the COVID-19 outbreak in 2020. (3) Robustness tests show that the proposed model consistently delivers strong predictive performance across different training-testing data splits (9:1, 8:2, and 6:4), with the MAPE remaining below 2%. These findings provide methodological advancements for forecasting in the KFTX market, offering both theoretical value and practical significance.