Financial market volatility and cross-asset correlations vary substantially over the business cycle, yet widely used resampling methods for tail risk forecasting remain backward-looking and fail to incorporate forward-looking economic conditions. This is a critical limitation because economic conditions fundamentally shape market risk profiles. To address this gap, we propose Forward-looking Economic State Probability-Weighted Simulation (FEWS), a novel and scalable approach that generates the joint return distribution of multiple assets conditional on predicted economic state probabilities. FEWS retains the strengths of resampling methods and adapts simulations to forward-looking macroeconomic conditions, thereby enhancing forecasting accuracy and mitigating the bias–variance trade-off inherent in the choice of sample window length. FEWS also captures key stylized facts of asset returns—such as time-varying volatility and correlations, the leverage effect, and asymmetric tails and dependencies—and allows these features to vary across economic states. Out-of-sample tests on equity portfolios show that FEWS outperforms benchmark methods, delivering more accurate tail risk forecasts and reducing sensitivity to the choice of sample window length. • We propose FEWS, a resampling framework conditioned on future economic conditions. • FEWS simulates forward-looking joint return distributions for tail risk forecasting. • FEWS captures economic-state-contingent volatility, leverage, and correlation dynamics. • Out-of-sample tests on equity portfolios show FEWS outperforms benchmark methods. • FEWS alleviates the bias–variance trade-off in forecasting.

his study investigates the application of \emph{risk-sensitive reinforcement learning} on heavy-tailed return series by comparing two primary algorithms: REINFORCE with baseline (REINFORCE-BL) and episodic batched actor--critic (A2C-B). Initial exploratory analysis reveals an asymmetric return distribution with numerous extreme \emph{outliers}, rendering variance-based risk measures inadequate and motivating the integration of tail-based risk measures—specifically Value at Risk (VaR), Conditional Value at Risk (CVaR), and Entropic Value at Risk (EVaR)—into the RL objective function. This study constructs a simple portfolio environment with discrete actions (market entry, market exit, and \emph{hold}) and trains both algorithms under four scenarios: risk-neutral, VaR, CVaR, and EVaR. Experimental results demonstrate that A2C-B consistently outperforms REINFORCE-BL across all scenarios, exhibiting higher average long-term rewards, faster convergence rates, and more stable \emph{learning curves}. While VaR and CVaR penalties significantly reduce rewards and increase learning volatility for REINFORCE-BL, A2C-B experiences only moderate reward reductions while maintaining stability. In the EVaR scenario, both algorithms yield high rewards, yet A2C-B retains a slight advantage in terms of stability. These findings indicate that in environments with heavy-tailed returns, employing coherent risk measures (particularly CVaR and EVaR) within an actor--critic framework offers a more compelling trade-off between tail risk control and average performance, serving as a viable \emph{baseline} for the development of risk-sensitive RL in finance and actuarial science.

Differentiated impacts of climate physical risks on the Indian power sector.

Non-determinism in deep learning algorithm design and implementation leads to performance variation, meaning model performance is not a single value, but rather a distribution. These model performance distributions are underexplored despite their impact on robustness. We investigate the robustness of deep learning performance to sources of non-determinism, specifically focusing on how performance distributions differ across various architectures and tasks. We conducted 186 experiments on state-of-the-art image classification (ResNet, ViT) and time series forecasting (Autoformer, iTransformer, NLinear, TSMixer) architectures. Each experiment was run 100 times with different random seeds to generate performance distributions, resulting in 18,600 runs. Robustness was quantified using metrics for spread, symmetry, and tail risk. Performance distributions are frequently non-Gaussian, particularly in time series forecasting. Model size does not systematically affect robustness - larger image classification models show fewer outliers but not lower spread, while smaller time series models show lower spread but more extreme underperformers. Training duration does not scale linearly; early stopping effectively balances performance and robustness. Mean performance does not predict robustness - time series forecasting shows moderate correlation while image classification shows none. Time series models produce nearly three times more underperforming outliers than image classification models, indicating substantially higher tail risk. Tail risk poses serious concerns for Trustworthy AI in high-stakes applications. Models performing well on average may exhibit long tails and extreme outliers revealed only through distributional analysis. Mean performance alone should not guide model selection; assessment of spread, symmetry, and tail risk is essential for reliable model assessment where consistent performance is critical.

ABSTRACT This study examines behavioural differences in risk aversion (variance risk), prudence (downside risk), and temperance (tail risk) between public and private sector employees, employing higher-order risk preference theory. Using two monetary-incentivized behavioural tasks (Multiple Price List and Deck-Schlesinger Lottery) with 1000 participants in Japan, we compare risk behaviours across sectors. Results demonstrate that public employees exhibit significantly higher temperance after accounting for individual-level covariates, indicating stronger aversion to extreme outcomes. Sectoral differences in risk aversion and prudence become insignificant once public service motivation is controlled. These findings highlight the importance of distinguishing multiple risk dimensions in sectoral comparisons.

Modern slavery as tail risk: an extreme value theory analysis of ESG, agrifood, and financial sectors

This paper attempts to apply the vine copulas methodology to assess the interdependence among the exchange rate market, equity indices, precious metals and energy resources within the selected BRICS economies. Using the ARFIMA-GJR-GARCH model, the residuals of the daily returns from foreign exchange rates, precious metals, equity indices, and energy prices of the BRICS economies for the period from January 1, 2003, to August 2023 were filtered. The empirical findings reveal a persistence of shocks and an asymmetric response to positive and negative news. Elevated volatility was observed across equity, precious metals, and energy markets, indicating substantial risks that necessitate robust risk management strategies. The results illustrate the heightened sensitivity of BRICS economies to external shocks, such as the Global Financial Crisis and the COVID-19 pandemic, which have triggered market volatility across currencies, stock market returns, and energy prices. This study emphasises the crucial importance of diversification, given the strong co-movement among asset classes, particularly during periods of extreme market volatility. Furthermore, the vine copulas analysis reveals intricate co-movements between assets, contributing to enhanced portfolio management strategies. Assets such as oil and gold serve as effective hedges. At the same time, foreign exchange rates play a significant role in investment decisions, underscoring the necessity for meticulous risk assessment and diversification strategies. These findings emphasize the vulnerability of BRICS economies to external shocks and highlight the imperative of effective risk management and diversification in navigating these dynamic markets. • ARFIMA-GJR-GARCH and C/D-vine copulas capture BRICS multi-asset dependence • C-vine fits Brazil, Russia, China; D-vine suits India and South Africa • Silver is central node; lower-tail dependence links silver with BRICS assets. • Oil and gold hedge; weak tau pairs provide diversification • Asymmetric tail risks stress need for volatility-aware BRICS risk management

Tail risk spillovers between Chinese USD-denominated bond market and Chinese stock market from a frequency-domain perspective

A queue-battery model for electrified traffic flow considering endogenous congestion propagation

Monitoring joint tail risks: An application to growth and inflation

Time-varying tails and the tail risk premium

ABSTRACT This study examines the dynamic tail risk spillovers between sustainability indices and traditional commodity markets in China using a CAViaR-TVP-VAR framework from 2014 to 2024. The results reveal significant time-varying and asymmetric risk transmission patterns, with ESG and low-carbon assets consistently serving as net risk transmitters, while traditional commodities especially energy-related indices act primarily as risk receivers. Methodologically, we innovate by integrating macroeconomic policy texts into a large language model (LLM)-based forecasting framework, demonstrating that policy augmentation substantially improves the predictive accuracy of tail risk connectedness, particularly during crisis periods. These findings underscore the growing systemic role of sustainable finance in China’s transition to a green economy and offer valuable insights for cross-market risk management and green policy design.

This study examines shock transmission between U.S. and Asian REIT markets using a TVP-VAR time-frequency connectedness framework incorporating higher-order moments (volatility, skewness, kurtosis) and uncertainty factors—geopolitical risk (GPR), economic policy uncertainty (EPU), and climate policy uncertainty (CCPU). Analysis reveals a core-periphery structure where the U.S. REIT index dominates Asian real estate dynamics. Major events—the 2018 Sino-U.S. trade war, COVID-19 pandemic, 2022 geopolitical tensions, and 2023 Israel-Hamas conflict—trigger pronounced connectivity spikes. Uncertainty factors exhibit bidirectional relationships with REIT markets: GPR acts as an asymmetric risk source during geopolitical crises, EPU transmits volatility shocks during the Chinese REIT crisis, whereas CCPU primarily functions as a net recipient. Frequency analysis shows returns exhibit short-term dominance, volatility demonstrates long-term persistence, and kurtosis displays the highest long-term connectedness, indicating tail risks propagate more systematically than asymmetric shocks. Japanese and Chinese REIT indices emerge as consistently vulnerable to external shocks due to structural factors including demographic pressures, capital controls, and carry trade dynamics. Practical applications include minimum-connectedness portfolio strategies reducing investment risk, with Chinese indices serving as optimal hedging instruments. These findings inform policy coordination mechanisms and dynamic risk management strategies for investors and financial stability authorities monitoring systemic risk transmission across global REIT networks.

ABSTRACT We analyze the impact of biodiversity loss on sector profits and losses, as well as financial system losses, using a CoVaR approach based on quantile regression. We introduce a world biodiversity index and a measure called CoBiodiversity to capture the dependence of extreme sector losses and profits on changes in biodiversity, and vice versa. Furthermore, ΔCoBiodiversity and ExposureCoBiodiversity measures assess the response of the sector tail risk to biodiversity degradation and the vulnerability of biodiversity loss to deteriorating or improving sector returns, respectively. Our results show a biodiversity loss risk premium, particularly for losses in the Energy sector, and profits in the Financials and Information Technology sectors. We also find a high system's market risk conditional on biodiversity loss, particularly during periods of distress such as the 2008 global financial crisis and COVID-19. Finally, the Energy, Materials, and Industrials sectors contribute the most to biodiversity loss, while the Consumer Staples sector contributes the least. These findings are important for academics, regulators, and investors to understand the relationship between biodiversity and the financial system.

As obvious for Value-at-Risk (VaR), even Expected Shortfall (ES) may not detect tail risk adequately. A possible solution proposed in the literature is the adjusted ES. It is defined as the supremum of ES values over different confidence levels, adjusted with a deterministic function, the so-called target risk profile. By using a family of general monetary risk measures instead of a family of ESs we unify this concept. This leads to a new class of risk measures, called adjusted risk measures. As a main finding we present equivalent assumptions for an adjusted risk measure to be positively homogeneous, subadditive, convex and consistent with second-order stochastic dominance. Furthermore, we show that these conditions hold for several adjusted risk measures beyond the adjusted ES and we derive their dual representations. Finally, a case study based on the S[Formula: see text]P 500 demonstrates similarities and differences between the adjusted ES and several new adjusted risk measures. Numerical aspects for the calculation of these risk measures are discussed.

The integration of Environmental, Social, and Governance (ESG) considerations into corporate finance has generated extensive yet fragmented research. Existing findings are heterogeneous, and causal claims are often undermined by measurement noise, selection bias, and pronounced ESG rating divergence. This lack of consensus obscures the mechanisms through which ESG affects firm value. To address this gap, this paper reviews recent developments by organizing evidence around five transmission channels: information, risk management, stakeholder relations, innovation, and capital allocation. Focusing on studies with credible identification, the review synthesizes findings on financial performance, cost of capital, climate risk, disclosure, governance, and engagement. The analysis reveals that ESG value creation is conditional, depending systematically on industry materiality, governance quality, and institutional context. The most robust evidence supports ESG's role in reducing financing costs and mitigating tail risks, while evidence of persistent alpha remains mixed. By clarifying boundary conditions and mechanisms, this review provides an integrated framework that reconciles prior inconsistencies. The paper concludes with implications for managers, investors, and policymakers and outlines promising frontiers in climate finance and machine-learning-based measurement. This synthesis thus offers a structured foundation for future theoretical and empirical inquiry.

Tail Risk Measures in Truncated Distributions and their Relationship with Inequality Indices

In August 2023, the launch of Shanghai Containerized Freight Index (SCFI) futures provides a suitable tool for risk management in the container shipping market, as well as new options for risk management of other financial assets. However, limited research exists on the influencing factors behind container freight rate fluctuations. This paper explores the nonlinear dynamic interdependence between the SCFI and 12 factors from the stock, commodity, carbon, and other markets using a data decomposition-reconstruction-based time-varying copula method, which can assist the stakeholders in hedging risk at different timescales. The findings reveal that most factors show no or limited upper tail dependence with SCFI in the short term. Medium- and long-term dependence is significantly stronger, indicating structural connections over longer horizons. Moreover, the dependence intensifies during extreme risk events. Generally, downside tail risks exert a greater influence on SCFI in the medium to long term, while upside tail risks are found to affect SCFI at any time horizon. This paper focuses on the tail risk interdependence analysis between SCFI and other assets, because the launch of SCFI futures makes the stakeholders to use this future to build risk management portfolios with other assets inevitably. The result provides useful implications to stakeholders with varying financial or investment attributes associated with shipping industry, aiding them in clarifying the different tail risk associations between SCFI futures and other assets at different timescales.

The transition toward sustainable and low-carbon power systems increasingly depends on the efficient accommodation of high shares of renewable energy while maintaining secure and reliable grid operation. In interconnected power systems, this challenge is intensified by strong cross-regional coupling, tie-line flow violation risks, and the high computational burden of fully coupled probabilistic assessments. To support the sustainable operation of renewable-rich interconnected systems, this paper proposes a probabilistic production simulation method that incorporates risk-aware tie-line exchange optimization. Sequential random sample paths are constructed by considering load fluctuations, renewable energy output uncertainty, and random outages of conventional units. Using cross-regional exchange power as coupling variables, a conditional value-at-risk (CVaR)-based pre-scheduling model is established to control tie-line and interface flow tail risks. Given the scheduled exchange power, cross-regional exchanges are transformed into regional boundary power injections, enabling decoupled sequential probabilistic production simulation for each region. The exchange schedule is then iteratively updated through marginal-value feedback. A four-region interconnected system is used for case-study validation. Results show that the proposed method improves renewable energy accommodation, reduces renewable curtailment, suppresses tie-line flow violation risk, and maintains high reliability assessment accuracy. Compared with the region-decoupled benchmark with fixed exchange power, the proposed method increases the renewable energy accommodation rate from 93.82% to 95.41% and reduces renewable curtailment from 312,162 MWh to 231,284 MWh, while also lowering expected energy not served and loss of load expectation. In addition, under the reported case-study setting, the proposed RC-IEF-PPS reduces the computation time from 5216.24 s for Full-PPS to 4074.63 s, i.e., by 21.9%, while maintaining comparable reliability assessment accuracy. These results indicate that the proposed framework can support the sustainable integration of high-penetration renewable energy by improving clean-energy utilization, operational reliability, and computational tractability in interconnected power systems.

Abstract Constructing an index-tracking portfolio involves closely replicating the performance of a benchmark index while minimizing deviations from it. In this paper, we propose a novel enhanced index tracking model that combines a quantile-regression-based deviation measure with linear second-order stochastic dominance (SSD) constraints. The objective is to control the tail risk of the tracking error while guaranteeing an enhancement of the portfolio return distribution relative to the benchmark. The proposed formulation leads to a linear optimization problem that remains computationally tractable under realistic portfolio constraints. The model is empirically evaluated using real-world data to assess the contribution of SSD constraints and to compare its performance with that of classical quantile regression. The empirical results show that both models outperform the benchmark index. However, when the investable universe is restricted through preselection, the proposed model delivers significant improvements in risk-return performance and tail-risk control.