摘要：《Journal of Risk and Insurance》为季刊，每年4期，每期发表文章8-10篇左右。2024年影响因子为2.1，是风险管理与保险领域的顶级权威学术期刊。该期刊主要发表保险经济学和风险管理主题的理论和实证方面的学术论文，可以为保险市场的实
声明：本系列文章基于原期刊目录和摘要内容整理而得，仅限于读者交流学习。如有侵权，请联系删除。
期刊介绍：
《Journal of Risk and Insurance》为季刊，每年4期，每期发表文章8-10篇左右。2024年影响因子为2.1，是风险管理与保险领域的顶级权威学术期刊。该期刊主要发表保险经济学和风险管理主题的理论和实证方面的学术论文，可以为保险市场的实践、决策和监管以及企业和家庭风险管理提供重要的信息。
本期看点：
●灾害风险最优公私共担机制：最优的政府干预策略依赖于个体损失之间的相关程度，当损失呈现中等程度正相关时，政府通过事后救助计划来补充私人保险是最优方案。而当损失相关性较高时，政府再保险成为更优选择，尽管在政府效率不及私营企业的情况下，并不会完全替代私人保险。
●首次经历洪水的周边居民会上调对洪水风险的认知，并开始主动购买洪水保险。
●洪灾发生后，社区在参与率和行动强度两个方面的风险缓解活动均显著增加。社区具备一定的适应能力，但也暴露出以灾害驱动为主的响应机制存在效率低下的问题，并引发了社区间响应能力差异带来的公平性担忧。
●一种新型神经网络架构可以提升对已报告但尚未结案赔案的损失金额预测能力，机器学习在提升精算预测能力方面具有很大的潜力，保险公司应逐步转向更具粒度的数据应用方式。
●基于大型语言模型的通用性文本差异分析框架，拓展了其在保险科技与风险管理领域中的应用场景。
●利用机器学习规范保险行业的风险文化发现，保险公司的风险文化受到不确定风险战略、在风险定义、执行与报告方面的限制、诉讼决策以及风险管理实践等因素的显著影响。
●科罗拉多州“Peak健康联盟”的设立提升了保险公司的市场议价能力，并促使平均保费下降，保费下降主要归因于医疗服务价格的降低，而非覆盖范围或风险结构的变化。
※ 本期目录
●Catastrophe risk sharing among individuals, private insurance, and government
●Learning from experience: Flooding and insurance take-up in the flood zone and its periphery
●Community responses to flooding in risk mitigation actions: Evidence from the community rating system
●Advancing loss reserving: A hybrid neural network approach for individual claim development prediction
●Cyber risk assessment for capital management
●Banding together to lower the cost of health care? An empirical study of the Peak Health Alliance in Colorado
●Textual analysis of insurance claims with large language models
●Regulating risk culture in the insurance industry using machine learning
Catastrophe risk sharing among individuals, private insurance, and government
个体、私人保险与政府之间的灾害风险共担机制研究
作者
Ruo Jia（北京大学经济学院），Jieyu Lin（岭南大学商学院），Michael R. Powers（清华大学经济管理学院/苏世民书院），Hanyang Wang（印第安纳大学凯利商学院）
摘要：Limited research has been conducted on the optimal public–private risk-sharing for catastrophe risks. This paper develops a theoretical framework to study the risk-sharing decisions and interactions of three types of catastrophe-market participants: a large number of individuals, a large number of private insurers in a competitive market, and a government that can choose between alternatives of re/insurance or ex post relief. Our analysis shows that the optimal government intervention varies depending on the correlation levels among individual losses. For moderately positive levels of loss correlation, it is optimal for the government to offer an ex post relief program to supplement private insurance. However, for higher levels of loss correlation, government reinsurance becomes optimal, although not to the extent of replacing private insurance if the government is less efficient than private firms. In sum, as catastrophe-loss correlations increase, that is, as the risk becomes more catastrophic, more risk-sharing tools and funding are needed to maximize social welfare.
目前关于灾害风险最优公私共担机制的研究仍较为有限。本文构建了一个理论框架，以分析三类灾害市场参与者——大量个体、大量处于竞争市场中的私人保险公司，以及可以在再保险或事后救助之间进行选择的政府——在风险分担中的决策与相互作用。研究发现，最优的政府干预策略依赖于个体损失之间的相关程度。当损失呈现中等程度正相关时，政府通过事后救助计划来补充私人保险是最优方案。而当损失相关性较高时，政府再保险成为更优选择，尽管在政府效率不及私营企业的情况下，并不会完全替代私人保险。总体而言，随着灾害损失相关性的上升，即风险变得更加系统性和灾难性，社会福利最大化所需的风险共担工具和资金支持也随之增加。
Learning from experience: Flooding and insurance take-up in the flood zone and its periphery
经验驱动下的学习：洪水事件对洪水区及其周边地区保险购买行为的影响
作者
Ivan Petkov（纽约城市大学皇后学院），Francesc Ortega（纽约城市大学皇后学院）
摘要：Flood insurance take-up remains low outside of the 100-year flood zone (SFHA), where purchasing insurance is entirely voluntary, despite the availability of affordable policies. Merging building footprints and inundation data for a large-scale flooding episode, we document substantial flood risk in the periphery of the SFHA and show that the storm led to large increases in take-up. But, while in the SFHA, the increase vanished after 3 years, it was highly persistent in the periphery. The extent of flooding and the type of policies purchased indicate that periphery residents who experienced flooding for the first time revised upwardly their beliefs about flood risk and began purchasing flood insurance. We also argue that increased granularity in flood risk communication could increase take-up before catastrophic flooding occurs.
尽管可获得价格合理的保单，在100年一遇洪水区（SFHA）以外的地区，洪水保险的购买率仍然较低，因为在这些地区购买保险完全属于自愿行为。本文将建筑物轮廓数据与一次大规模洪灾的淹没数据相结合，记录了SFHA周边地区存在显著的洪水风险，并发现该次风暴导致洪水保险购买率大幅上升。然而，在SFHA内部，这一上升趋势在三年后消失，而在周边地区却表现出高度的持续性。洪灾的严重程度以及所购买保单的类型表明，那些首次经历洪水的周边居民上调了对洪水风险的认知，并开始主动购买洪水保险。我们进一步指出，如果洪水风险沟通更加细化，有望在灾难性洪水发生之前提高保险覆盖率。
Community responses to flooding in risk mitigation actions: Evidence from the community rating system
社区应对洪灾的风险缓解行为：来自社区评级系统的证据
作者
Yanjun（Penny）Liao（未来资源研究所Resources For the Future），Simon Sølvsten（欧洲风险与韧性研究中心），Zachary Whitlock（未来资源研究所Resources For the Future）
摘要：This paper studies the impact of disaster experiences on communities' engagement in risk mitigation actions, focusing on flooding in the United States. We measure risk mitigation actions using communities' scores in the Community Rating System, an incentive program that scores flood preparedness and mitigation activities and rewards communities with flood insurance premium discounts. Leveraging a panel of communities from 1998 to 2019, we find a significant increase in risk mitigation activities following flood events, in both participation rates and intensity of actions. The effects continue to increase up to 10 years. Communities with greater capacity, particularly those in urban areas, exhibit a much stronger response. The findings highlight the adaptive capacity of communities but also raise several concerns regarding the inefficiency of disaster-driven responses and inequitable outcomes across communities.
本文研究灾害经历对社区参与风险缓解行为的影响，重点关注美国的洪水风险应对。我们通过“社区评级系统”（Community Rating System，CRS）中的评分来衡量社区的风险缓解行为。CRS是一项激励计划，用于评估社区在洪水防范和缓解方面的措施，并通过降低洪水保险保费来奖励得分较高的社区。基于1998年至2019年的社区面板数据，我们发现洪灾发生后，社区在参与率和行动强度两个方面的风险缓解活动均显著增加，且这种效应可持续上升长达十年。具备更强治理能力的社区，尤其是城市地区，表现出更强的应对反应。研究结果表明，社区具备一定的适应能力，但也暴露出以灾害驱动为主的响应机制存在效率低下的问题，并引发了社区间响应能力差异带来的公平性担忧。
Advancing loss reserving: A hybrid neural network approach for individual claim development prediction
推进损失准备金估计：一种用于个别赔案发展预测的混合神经网络方法
作者
Judith C. Schneider（汉诺威莱布尼茨大学风险与保险研究所），Brandon Schwab（汉诺威莱布尼茨大学风险与保险研究所）
摘要：Accurately estimating loss reserves is critical for the financial health of insurance companies and informs numerous operational decisions. We propose a novel neural network architecture that enhances the prediction of incurred loss amounts for reported but not settled claims. Moreover, differing from other studies, we test our model on proprietary datasets from a large industrial insurer. In addition, we use bootstrapping to evaluate the stability and reliability of the predictions and Shapley additive explanation values to provide transparency and explainability by quantifying the contribution of each feature to the predictions. Our model shows superiority in estimating reserves more accurately than benchmark models, like the chain ladder approach. Particularly, our model exhibits nuanced performance at the branch level, reflecting its capacity to effectively integrate individual claim characteristics. Our findings emphasize the potential of using machine learning in enhancing actuarial forecasting and suggest a shift towards more granular data applications.
准确估算损失准备金对保险公司的财务稳健至关重要，并影响众多运营决策。本文提出了一种新型神经网络架构，用于提升对已报告但尚未结案赔案的损失金额预测能力。与其他研究不同，我们在某大型工业保险公司提供的专有数据集上对模型进行了测试。此外，我们运用自助法（bootstrapping）评估模型预测的稳定性与可靠性，并通过Shapley加性解释值（SHAP）量化各特征对预测结果的贡献，以增强模型的可解释性和透明度。研究结果显示，该模型在损失准备金估计方面显著优于传统基准模型，如链梯法（Chain Ladder）。特别是，我们的模型在分支机构层面表现出细微的性能差异，反映了其有效整合单个索赔特征的能力。研究强调了机器学习在提升精算预测能力方面的潜力，并指出保险公司应逐步转向更具粒度的数据应用方式。
Cyber risk assessment for capital management
用于资本管理的网络风险评估框架研究
作者
Wing Fung Chong（赫瑞–瓦特大学数学科学麦克斯韦研究所；精算数学与统计系），Runhuan Feng（清华大学金融系），Hins Hu（康奈尔大学系统工程系），Linfeng Zhang（俄亥俄州立大学数学系）
摘要：This paper introduces a two-pillar cyber risk management framework to address the pervasive challenges in managing cyber risk. The first pillar, cyber risk assessment, combines insurance frequency-severity models with cybersecurity cascade models to capture the unique nature of cyber risk. The second pillar, cyber capital management, facilitates informed allocation of capital for a balanced cyber risk management strategy, including cybersecurity investments, insurance coverage, and reserves. A case study, based on historical cyber incident data and realistic assumptions, demonstrates the necessity of comprehensive cost–benefit analysis for budget-constrained companies with competing objectives in cyber risk management. In addition, sensitivity analysis highlights the dependence of the optimal strategy on factors such as the price of cybersecurity controls and their effectiveness. The framework's implementation across a diverse range of companies yields general insights on cyber risk management.
本文提出了一个“双支柱”的网络风险管理框架，以应对当前网络风险管理中普遍存在的挑战。第一支柱是网络风险评估，融合了保险中的频率–严重性模型与网络安全事件级联模型，从而更好地刻画网络风险的独特特性。第二支柱是网络资本管理，旨在辅助企业在网络安全投入、保险覆盖与资本储备之间实现科学的资金配置，推动更为均衡的网络风险管理策略。基于历史网络事件数据与现实假设的案例研究表明，对于预算有限且面临多重管理目标的公司而言，开展全面的成本效益分析至关重要。此外，敏感性分析指出，最优管理策略对网络安全控制措施的价格与效果高度敏感。该框架在不同类型企业中的应用，揭示了网络风险管理的一般性规律与策略启示。
Banding together to lower the cost of health care? An empirical study of the Peak Health Alliance in Colorado
通过联合采购降低医疗成本？——对科罗拉多州Peak健康联盟的实证研究
作者
Mark K. Meiselbach（约翰斯·霍普金斯大学彭博公共卫生学院），Matthew D. Eisenberg（约翰斯·霍普金斯大学彭博公共卫生学院；凯里商学院）
摘要：This paper evaluates the effectiveness of Peak Health Alliance, a public–private initiative in Colorado aimed at lowering health care costs for employers and enrollees by increased bargaining power through the formation of a health care purchasing alliance. Using 2017–2021 plan data provided by the Colorado Department of Regulatory Affairs: Division of Insurance, we use difference-in-differences, event study, and synthetic control methods to compare changes in premiums in counties where Peak operated to other counties in Colorado before and after its implementation. The results suggest that Peak was associated with an increase in insurer market power and led to a 13%–17% decrease in average premiums, depending on the empirical specification. We further assess mechanisms underlying these effects and find evidence that lower prices were the most likely mechanism behind the estimated effect of Peak. Study results provide insights about the future of such public–private partnerships and their potential effectiveness.
本文评估了“Peak健康联盟”（Peak Health Alliance）在控制医疗支出方面的政策效果。该联盟是科罗拉多州的一项公私合作项目，旨在通过组建医疗服务采购联盟提升议价能力，从而为雇主及参保人降低医疗成本。基于科罗拉多州监管事务部保险司提供的2017年至2021年保险计划数据，本文采用双重差分法（Difference-in-Differences）、事件研究（Event Study）以及合成控制法（Synthetic Control）等计量方法，对Peak联盟实施前后、其所在县与其他县的保费变动进行对比分析。研究结果显示，Peak联盟的设立提升了保险公司的市场议价能力，并在不同实证设定下促使平均保费下降了13%至17%。进一步机制分析表明，保费下降主要归因于医疗服务价格的降低，而非覆盖范围或风险结构的变化。本研究为探索公私合作机制在降低医疗费用方面的作用提供了经验证据，并对相关政策的推广与优化具有现实启示意义。
Textual analysis of insurance claims with large language models
利用大型语言模型对保险理赔文本的分析研究
作者
Dongchen Li（华东师范大学统计学院，教育部统计与数据科学前沿理论与应用重点实验室），Zhuo Jin（澳大利亚麦考瑞大学，精算与商业分析系），Linyi Qian（华东师范大学统计学院，教育部统计与数据科学前沿理论与应用重点实验室；华东师范大学中国普惠养老金融研究中心），Hailiang Yang（西交利物浦大学数学与物理学院，金融与精算数学系）
摘要：This study proposes a comprehensive and general framework for examining discrepancies in textual content using large language models (LLMs), broadening application scenarios in the insurtech and risk management fields, and conducting empirical research based on actual needs and real-world data. Our framework integrates OpenAI's interface to embed texts and project them into external categories while utilizing distance metrics to evaluate discrepancies. To identify significant disparities, we design prompts to analyze three types of relationships: identical information, logical relationships and potential relationships. Our empirical analysis shows that 22.1% of samples exhibit substantial semantic discrepancies, and 38.1% of the samples with significant differences contain at least one of the identified relationships. The average processing time for each sample does not exceed 4 s, and all processes can be adjusted based on actual needs. Backtesting results and comparisons with traditional NLP methods further demonstrate that our proposed method is both effective and robust.
本研究提出了一个基于大型语言模型（LLMs）的通用性文本差异分析框架，拓展了其在保险科技与风险管理领域中的应用场景，并结合实际需求与真实数据开展了实证研究。该框架通过集成OpenAI接口对文本进行嵌入表示，并投射至外部分类体系中，同时利用距离度量方法评估文本差异。为识别关键性语义差异，本文设计提示词以分析三类文本关系：信息一致性、逻辑关系与潜在关联关系。实证结果表明，22.1%的样本存在明显语义差异，其中38.1%的显著差异样本包含至少一种识别出的文本关系类型。每个样本的平均处理时长不超过4秒，整体流程可根据实际需求灵活调整。回测分析与与传统自然语言处理（NLP）方法的对比结果进一步验证了所提出方法的有效性与稳健性。
Regulating risk culture in the insurance industry using machine learning
利用机器学习规范保险行业的风险文化
作者
Aparna Gupta（伦斯勒理工学院，Lally管理学院），Abena Owusu（蒙特克莱尔州立大学，Feliciano商学院）
摘要：We investigate whether the price paid for insurance explains dishonesty in reporting an insurance claim. In our laboratory experiment, participants earn money in a real-effort task but risk losing some of this income through one of four randomly assigned, privately observed loss amounts. Before observing their loss, participants indicate their reservation price for insurance that pays an indemnity equal to their stated loss. Participants are insured if their randomly assigned premium is less than their stated reservation price. This mechanism provides data on each participant's consumer surplus from insurance. After receiving their cash earnings minus their assigned loss in private, participants report their loss. We find that the insured report modestly but statistically insignificant larger losses than the uninsured. Among the insured, we find no clear evidence that their reporting of excess losses increases in the randomly assigned price of insurance or decreases in the consumer surplus from insurance.
本研究运用文本分析和机器学习方法，探讨保险行业中风险文化与监管之间的关系。通过分析公司10-K年度报告，我们将企业划分为不同的风险文化集群，并发现保险公司的风险文化受到不确定风险战略、在风险定义、执行与报告方面的限制、诉讼决策以及风险管理实践等因素的显著影响。时间序列预测分析显示，相较于风险文化正在改善的公司，维持较差风险文化趋势的大型保险公司更难实现逆转。此外，Dodd–Frank法案实施后，保险公司的风险文化有所提升。我们的研究结果强调了监管措施在监督和管理保险公司风险实践方面可能带来的积极作用。
更多推荐文章
【保险学术前沿】期刊Journal of Health Economics 2025年101卷目录及摘要
【保险学术前沿】The Geneva Papers on Risk and Insurance 2025年50卷第2期目录与摘要
来源：13个精算师
