We introduce a deep visualization (DV) framework that turns conventional commodity data into images and extracts predictive signals via convolutional feature learning. Specifically, we encode futures price trajectories and the futures surface as images, then derive four deep‑visualization (DV) predictors, carry ($bs_{DV}$), basis momentum ($bm_{DV}$), momentum ($mom_{DV}$), and skewness ($sk_{DV}$), each of which consistently outperforms its traditional formula‑based counterpart in return predictability. By forming long–short portfolios in the top (bottom) quartile of each DV predictor, we build an image‑based four‑factor model that delivers significant alpha and better explains the cross‑section of commodity returns than existing benchmarks. Further evidence shows that the explanatory power of these image‑based factors is strongly linked to macroeconomic uncertainty and geopolitical risk. Our findings reveal that transforming conventional financial data into images and relying solely on image-derived features suffices to construct a sophisticated asset pricing model at least in commodity markets, pioneering the paradigm of image‑based asset pricing.

We examine how mutual funds form stock market expectations and the implications of these beliefs for asset returns, using a novel text-based measure extracted from Chinese fund reports. Funds extrapolate from recent stock market and fund returns when forming expectations, with more recent returns receiving greater weight. This recency tendency is weaker among more experienced managers. At the aggregate level, consensus expectations positively predict short-term future market returns, both in and out of sample. At the fund level, expectations are positively related to subsequent fund performance in the time series. In the cross-section, however, superior performance arises only when funds accurately forecast market direction and adjust their portfolios accordingly. This effect is stronger for optimistic forecasts and among funds with greater exposure to liquid stocks. Our findings highlight the conditional nature of belief-driven performance, shaped jointly by forecasting skill and the ability to implement views in the presence of execution frictions such as short-selling and liquidity constraints.

We construct a novel risk-based similarity peer network by applying machine learning techniques to extract a comprehensive set of disclosed risk factors from firms' annual reports. We find that a firm's future returns can be significantly predicted by the past returns of its risk-similar peers, even after excluding firms within the same industry. A long-short portfolio, formed based on the returns of these risk-similar peers, generates an alpha of 84 basis points per month. This return predictability is particularly pronounced for negative-return stocks and those with limited investor attention, suggesting that the effect is driven by slow information diffusion across firms with similar risk exposures. Our findings highlight that the risk factors disclosed in 10-K filings contain valuable information that is often overlooked by investors.

How do Held-to-Maturity (HTM) securities that limit the impacts of banks’ unrealized capital loss on the regulatory capital measures affect banks’ exposure to deposit run risks when policy rates increase? And how should regulators design policies on classifying securities as HTM jointly with bank capital regulation? To answer these questions, we develop a model of bank runs in which banks classify long-term assets as HTM or Asset-for-Sale (AFS). Banks trade off the current cost of issuing equity to meet the capital requirement when the interest rate increases against increasing future run risks when the interest rate increases further in the future. When banks underestimate interest rate risks or have limited liability to depositors in the event of default, capping held-to-maturity long-term assets and mandating more equity capital issuance may reduce the run risks of moderately capitalized banks. Using bank-quarter-level data from Call Reports, we provide empirical support for the model’s testable implications.

This paper investigates the contribution of futures market information to enhancing the predictive accuracy of macroeconomic forecasts, using data from China. We employ three cat-egories of predictors: monthly macroeconomic factors, daily commodity futures factors, and daily financial futures variables. Principal component analysis is applied to extract key fac-tors from large data sets of monthly macroeconomic indicators and daily commodity futures contracts. To address the challenge of mixed sampling frequencies, these predictors are incor-porated into factor-MIDAS models for both nowcasting and long-term forecasting of critical macroeconomic variables. The empirical results indicate that financial futures data provide modest improvements in forecasting secondary and tertiary GDP, whereas commodity futures factors significantly improve the accuracy of PPI forecasts. Interestingly, for PMI forecast-ing, models relying exclusively on futures market data, without incorporating macroeconomic factors, achieve superior predictive performance. Our findings underscore the significance of futures market information as a valuable input to macroeconomic forecasting.

This paper investigates the price overreaction during the initial continuous trading period of the Chinese corn futures market. Using a dynamic modeling algorithm, we identify the overreaction behavior of intraday high-frequency (1 min and 3 min) prices during the first session of daytime trading. The results indicate that the overreaction hypothesis is confirmed for the daytime prices of the Chinese corn futures market. We also find a noticeable reduction in overreaction following the introduction of night trading and this decline appears to diminish over time. Furthermore, this paper conducts an overreaction trading strategy to assess traders’ returns, revealing a slight decline in average return after the introduction of night trading. This study provides valuable insights and recommendations for exchanges and regulators in monitoring overreaction and formulating effective policies to address it.

We examine the pricing ofsystematic information asymmetry, induced by Chinese gov-ernment intervention, in the cross-section of stock returns. Using market-wide order im-balance as a proxy for systematic information, we observe a strong correlation betweenthe standard deviation of market-wide order imbalance and economic policy uncertainty.Furthermore, we find a significant positive relationship between the sensitivity of stocks tosystematic information asymmetry (OIBeta) and their future returns. The average monthlyreturn spread between high- and low-OIBeta portfolios ranges from 1.30% to 1.77%, andthis result remains robust after controlling for traditional risk factors. Our results providesubstantial evidence that the pricing of OIBeta is driven by systematic information asym-metry rather than alternative explanatory channels.

This paper employs a proprietary data set on commodity producers’ profit margins (PPMG) and establishes a robust positive relationship between commodity producers’ profitability growth and future returns of commodity futures. The spread portfolio that longs top-PPMG futures contracts and shorts bottom-PPMG futures contracts delivers a statistically significant average weekly return of 36 basis points. We further demonstrate that profitability is a strong SDF factor in commodity futures market. We theoretically justify our empirical findings by developing an investment-based pricing model, in which producers optimally adjust their production process by maximizing profits subject to aggregate profitability shocks. The model reproduces key empirical results through calibration and simulation.

This study reconsiders the timing of the equity risk factors by using the flexible neural networks specified for image recognition to determine the timing weights. The performance of each factor is visualized to be standardized price and volatility charts and `learned' by flexible image recognition methods with timing weights as outputs. The performance of all groups of factors can be significantly improved by using these ``deep learning--based'' timing weights. In addition, visualizing the volatility of factors and using deep learning methods to predict volatility can significantly improve the performance of the volatility-managed portfolio for most categories of factors. Our further investigation reveals that the timing success of our method hinges on its ability in identifying ex ante regime switches such as jumps and crashes of the factors and its predictability on future macroeconomic risk.

We construct firm-level indicators to capture the engagement in the Belt and Road Initiative (BRI, henceforth) via textual analysis. We find that higher firm engagement in BRI predicts higher stock returns in the subsequent 12 months. The top 10% high-BRI firms have 12.42% higher annual returns than bottom 10% low-BRI firms in China A-Share market. Additionally, two fundamental channels of increased earnings and reduced liabilities explain the higher expected returns of high-BRI firms. Furthermore, we reveal that the phenomenon is more pronounced among non-state-owned enterprises. For large-cap firms, BR Report is a more effective indicator for predicting future stock returns, while BR Beta performs better for small-cap firms. These findings contribute to the measurement of firm engagement in BRI and its impact on the stock market.