In the wake of increasing global climate concerns, understanding how financial markets respond to climate-related uncertainties has become crucial for investors and policymakers. This study investigates the directional predictability of climate policy uncertainty (CPU) and sustainability uncertainty index (ESGUI) on four major African stock markets, namely: Nigeria, Egypt, Morocco and South Africa. Using cross-quantilogram method, we analyze the magnitude, direction, and persistence of predictability across different quantiles of both uncertainty measures and market returns from 2000 to 2024. Our results reveal significant asymmetric patterns across these markets. Lower (higher) uncertainty quantiles exhibit positive (negative) predictability during bearish (bullish) market conditions. The Nigerian and Moroccan markets demonstrate stronger reactions to fluctuations of both CPU and ESGUI, compared to South African and Egyptian markets, with Nigerian market being particularly sensitive to sustainability uncertainty. Time-varying analysis reveals that predictability strengthens during global financial stress periods and major climate policy events (2008–2009, 2015, 2020–2022), with sustainability uncertainty effects showing greater persistence than climate policy uncertainty across all markets. These findings underscore the importance of incorporating climate and sustainability factors in investment decisions and policy development across African economies, with implications for portfolio diversification, risk management, and regional financial stability.
Data pricing plays a key role in promoting the efficient circulation of data elements in the market. However, existing methods still have limitations in terms of targeting, dynamics, and comprehensiveness. This study proposes a data price prediction model based on Beluga Whale optimized XGBoost to provide scientific support for data market pricing decisions. First, we collect transaction data from the API zone of the Youyi Data platform and perform data cleaning and sample screening. Second, we preprocess the data using methods such as outlier treatment, one-hot encoding, and logarithmic transformation. Lasso regression is used for feature selection to construct a data price prediction model containing 19 features. Finally, the BWO algorithm is used to optimize the hyperparameters of XGBoost to improve the model’s predictive performance. Experimental results show that the BWO-XGBoost model improves the R2 metric by 2.3
In the context of rising geopolitical uncertainty in recent years, the trend of systemic risk spreading across industries has become increasingly prominent. This paper takes the Chinese A-share market from January 3, 2017, to May 16, 2025, as the research sample, focusing on the association between the Sino-US tariff conflict and market structure. Employing complex network methods, we construct undirected networks of highly correlated stocks using the correlation threshold approach and analyze the dynamic evolution of key network topological indicators under both fixed and rolling time windows. The results show that during the intensive implementation of tariff measures between early 2018 and the end of 2019, the network structure deviated significantly from its long-term equilibrium, characterized by higher overall connectivity and stronger local clustering. In contrast, during the regulatory shock of 2021, the network exhibited structural fragmentation, with rising average path length, declining average degree and clustering, and increasing assortativity, indicating a concentration of influence within a limited number of core communities. These two patterns are not contradictory but represent two distinct paradigms of systemic vulnerability: synchronized fragility, characterized by high interconnectedness and rapid market-wide co-movement, and decoupled fragility, characterized by network fragmentation and the concentration of influence within core communities. Furthermore, community structures identified through the Louvain algorithm reveal that many stocks formed cross-industry, non-traditional coupling relationships during the crisis, reflecting correlation-based risk co-movement patterns beyond conventional industry classifications. The findings of this study provide investors with a structured perspective for identifying risks under external shocks and offer valuable insights for regulators in building early-warning systems.
For VaR and other risk measures calculation, the assumption of i.i.d. normal log returns is typically used. For a time series that significantly violate independence assumptions, people usually consider fitting a time series model and apply VaR computation on the fitted residuals, which are assumed to be independent. However, as the fitted model is a good fit in regions where most of the data fall instead of a good fit in the tails. It causes the fitted residuals are actually not independent in practice and hence the risk measures computed may not be reliable. This article introduces a new econometric approach to the estimation of VaR and convex risk measures in financial risk management. Using decoupling inequalities, we show that methods in EVT can apply to the dynamically scaled residuals. This approach, therefore, addresses the pitfalls of EVT techniques while enabling them to realize their widely recognized potential for estimating extreme quantiles and probabilities.
Applying Reinforcement Learning (RL) to economic modeling reveals a fundamental tension: standard RL agents do not behave as competitive price-takers. While macroeconomic models typically assume agents take aggregate conditions as given, a naive single-agent RL algorithm learns to internalize its impact on the environment, effectively exerting market power. We illustrate this divergence in a search-and-matching model with concave production, showing that a standard RL agent converges to a monopsonistic allocation rather than the competitive equilibrium. Furthermore, we show that RL’s approach to time discounting distorts standard intertemporal economic trade-offs. To resolve these issues, we propose a Calibrated Mean-Field Reinforcement Learning framework that restricts the agent to optimize against an exogenous aggregate state to restore price-taking behavior, and recalibrates the RL reward structure to reflect true economic opportunity costs. By iteratively updating the aggregate state, our algorithm converges to a stationary fixed point that recovers the competitive equilibrium. This approach provides a method for embedding learning algorithms into dynamic general equilibrium models.
