Insurance-linked securities such as catastrophe (CAT) bonds are designed to transfer tail risk to capital markets, but this transfer can fail when insured losses become more closely aligned with financial-market stress, raising the compensation investors require in joint-loss states. Cyber risk and natural disasters are salient applications because severe losses can coincide with broader portfolio distress and erode the diversification logic supporting CAT bond issuance. Motivated by evidence of market-wide stress after major cyber incidents, we develop a two-regime continuous-time framework in which catastrophe-loss intensity is stochastic and dependence between catastrophe-loss claims and equity returns strengthens in disaster states. A joint reinsurance-portfolio problem yields closed-form insurer policies, economically meaningful welfare losses from misestimating disaster-state dependence, and the affordable spread for CAT issuance. Catastrophe losses realized in low-wealth states raise investors’ required spreads. Together, these imply a bid-ask feasibility condition under which standard CAT bonds can fail to clear even as market depth increases. We propose a catastrophe put option (CatPut), an insurer-funded, catastrophe-conditional equity-index put triggered only when both catastrophe and equity thresholds are breached. It restores market-clearing feasibility over a nontrivial, layer-specific design space, highlighting that targeted crisis-contingent redesign can be more effective than market deepening alone.

This study examines how firms respond to carbon regulations in their ESG management and whether ESG ratings are properly aligned with actual emissions. Using 592 publicly listed Korean firms from 2015 to 2024, we estimate Difference-in-Differences (DiD) models with firm and year fixed effects and address staggered adoption of carbon regulations. We find evidence that the earlier Target Management System (TMS), as a weaker regulation than the Emissions Trading Scheme (ETS), has no significant effect on firms’ ESG environmental scores, whereas entry into the stricter ETS is associated with a statistically significant and persistent increase in environmental scores. Yet these rating gains do not translate into real mitigation by showing no statistical evidence on a change in the rate of greenhouse gas (GHG) emissions reduction. Moreover, Granger causality tests indicate that increases in environmental scores do not precede subsequent reductions in emissions intensity. Our results suggest a decoupling between environmental scores and actual emissions outcomes after ETS entry. These findings highlight the need for regulatory designs that can induce firms to initiate measurable and verifiable reductions in greenhouse gas emissions.

This study investigates whether investors distinguish between contagious and noncontagious cyber risks in the cross-section of stock returns. Although recent literature establishes cyber risk as a priced factor in financial markets, asset pricing theory predicts that only its non-diversifiable component should command a risk premium, suggesting that the key dimension is not cyber risk broadly defined, but its potential for systemic propagation. We propose a novel text-based framework using a transformer-based language model to classify contagious cyber incidents and estimate firm-level ex-ante exposure via a machine learning algorithm. Our event-study analysis reveals that contagious incidents are associated with significantly larger and more persistent negative abnormal returns over extended event windows, with these effects particularly pronounced for highly leveraged firms. Furthermore, our results show that firms with higher exposure to contagious cyber risks earn a significant positive risk premium of approximately 33 basis points per month, whereas exposure to noncontagious risk is not priced. We conclude that investors selectively price cyber risk, identifying contagion as the primary channel through which cyber threats become a systematic, non-diversifiable risk factor.

We examine how managerial preferences, internal risk-control practices, and financial frictions jointly shape insurance demand among small and medium-sized enterprises (SMEs). Using a unique dataset of 693 Korean SMEs that links 2023 survey measures of managerial risk aversion and risk-control orientation with 2020-2022 financial statements, we estimate baseline OLS models, a nonlinear CEO-ownership threshold model, and finite mixture regressions to capture latent heterogeneity in insurance-demand behavior. The OLS results show that SMEs led by more risk-averse managers purchase significantly more discretionary insurance, and that internal risk-control activities complement rather than substitute for insurance. Insurance demand also rises with tax incentives and financial constraints but declines with firm size. Threshold estimation reveals a non-monotonic ownership effect: insurance demand decreases with CEO ownership below 37% but increases beyond it. Finite mixture models uncover three distinct behavioral regimes - reference-and-control, preference-only, and financial-driven -each characterized by different mechanisms linking preferences, controls, and financial conditions to insurance purchases. These findings demonstrate that SMEs do not follow a single representative demand rule and that pooled estimates obscure economically meaningful heterogeneity. The results offer implications for targeted insurance design, SME risk-management policy, and credit-risk assessment.

Traditional indemnity-based cyber insurance struggles to keep pace with the escalating frequency and complexity of cyber threats, resulting in a widening protection gap for industry. To address this, we propose a novel cyber parametric insurance framework based on firm-specific cyber vulnerability that is estimated using a dynamic factor model incorporating geographic location, firm size, and industry sector. Our framework helps cyber insurers reduce basis risk by accounting for temporal dynamics and firm heterogeneity into cyber risk assessment. It can also facilitate to tailor a risk mitigation solution by considering substantial variations in corporate vulnerability that is integrated into the parametric index with the duration of cyberattacks. We provide supporting evidence on this benefit of the proposed framework by demonstrating a reduction in basis risk of our parametric insurance without increasing insolvency, compared to a benchmark. This study can contribute to the development of sustainable and scalable cyber parametric insurance markets.

Climate change risk emerges as a critical concern of Property & Casualty (P&C) insurers due to its causal effect on increasing natural disasters across the globe. However, a challenge to quantify ongoing climate change can make it difficult to assess how this progress may affect their underwriting performance. To address this concern, we employ the Actuaries Climate Index (ACI) to examine the impact of climate change risk on P&C insurers’ performance. We propose a local-level ACI by using meteorological information of regions in South Korea and employ a copula simulation to implement a stress test to determine the impact of climate change on insurers’ exposures. Our findings indicate that insurers with high geographic concentration in the capital region are more vulnerable to climate risk exposure.

Systemic cyber risk can cascade across interconnected networks, generating substantial Incurred But Not Reported (IBNR) losses that threaten insurers' capital adequacy. We present a framework that embeds IBNR into Solvency Capital Requirement (SCR calculations by coupling an SIR epidemic model parameterized by firms' operational resilience and sectoral network structure—with Monte Carlo simulations of loss propagation. Our analysis shows that highly resilient firms recognize losses more quickly, leading to shorter reporting periods and consequently higher IBNR in the early stages, whereas low resilient firms need longer reporting periods, causing their losses to increase over time. Sectoral comparison reveals more severe systemic impacts in finance than in information and demonstrates that large insurers maintain adequate solvency buffers while smaller insurers face solvency pressures. These findings highlight the need to incorporate IBNR, resilience, and network topology in cyber-SCR modeling.

This study investigates whether corporate environmental, social, and governance (ESG) activities enhance or hinder operational efficiency in the insurance industry, a sector widely regarded as inherently aligned with sustainability objectives. Contrary to the common expectation that ESG activities improve operational performance, our results show that insurers with higher ESG scores tend to exhibit lower business efficiency, particularly in terms of pure technical efficiency. This negative association persists when one-year lagged ESG measures are considered. We also provide evidence that while ESG activities are found to contribute to risk management and, in some cases, financial performance,these benefits do not translate into improvements in operational efficiency in the short term. Our findings suggest that ESG initiatives in the insurance industry may involve substantial operational and organizational costs that outweigh their efficiency-enhancing effects, at least in the near term.

We propose a matrix-based factor analysis model for predicting the probability of insurance claims. The model employs projected principal component analysis (PPCA), which enhances the estimation of unobserved latent factors by projecting a data matrix onto a linear space spanned by insured-specific features. This approach addresses the curse of dimensionality when the number of insured-specific features and insurance coverages is large, enabling more accurate estimation of claim probability than conventional methods. Using a health insurance dataset from a leading life insurer in South Korea, we demonstrate that the proposed model outperforms conventional and machine-learning benchmarks, such as logistic regression and XGBoost, in predicting claim probabilities. We further determine that our model can reduce computational time by approximately 86% and 98% compared to logistic regression and XGBoost, respectively. The proposed model can enhance claims management efficiency and strengthen financial stability through accurate claims prediction.