Abstract The rise in online transactions has made credit card fraud a significant global concern, necessitating detection strategies that are both highly accurate and practically viable. While existing literature extensively explores machine learning techniques to address class imbalance, most studies optimize for traditional statistical metrics, overlooking the asymmetric financial costs and strict operational constraints of real-world fraud detection. This study bridges this gap by proposing a comprehensive, cost-sensitive ensemble framework evaluated on a real-world European cardholder dataset. We move beyond the traditional F 1 score by adopting the cost-sensitive F β metric to reflect real financial impact. Through exhaustive benchmarking, we show that while eXtreme Gradient Boosting (XGBoost) combined with Borderline SMOTE achieves the highest single-model performance, our proposed soft-voting ensemble integrating Logistic Regression and Random Forest with SMOTE delivers the best overall performance ( F β = 0.8287). To ensure practical viability, we introduce a Top- K operational constraint evaluation reflecting limited human investigation bandwidths. Additionally, an ablation study demonstrates that there is no universal remedy for class imbalance; optimal interventions are highly model-dependent. Finally, by validating our framework on a feature-transparent simulated dataset, model interpretability analysis reveals the ensemble’s capacity to capture the critical importance of environmental risk factors, shifting the focus beyond solely customer-centric anomalies.
Abstract Agricultural insurance is yet to achieve widespread adoption due to ongoing challenges in risk assessment. This study proposes a climate-responsive insurance framework that integrates meteorological and remote sensing data into statistical modeling to enhance risk assessment. We estimate the predictive density of yields by analyzing historical yield data and weather information across three crop life cycle phases: sowing, growing, and harvesting. A nonparametric Bayesian framework is used to derive the conditional yield distribution through a copula-based joint dependence structure between yield and climate variables. Bayesian inference is used to quantify the relative influence of climate factors at each crop phase on yield variability. These phase-specific contributions are used as weights to combine the conditional yield distributions, resulting in a more weather-informative predictive density. This approach shows improved predictive accuracy in cross-validation and outperforms conventional univariate models in out-of-sample rating performance, supporting more reliable premium rate estimation. These features make it promising for large-scale insurance programs compared to prevalent univariate models. Our findings under various climate stress scenarios underscore the need for such a framework to inform sustainable, climate-responsive insurance policies and to enhance both farmer participation and resilience to climate risks.
Abstract A captive – short for a captive insurance company – is an alternative technique for businesses to finance risk. Captives are formed by their parent companies which can retain and mutualize risks within corporate structures. Theoretically, captives can enable parent companies to access efficient risk financing, streamline risk management, and optimize cash flows with recaptured premiums. This study aims to assess the effects of captive formation on cash flow among Nikkei 225 companies from 2000 to 2023. Our empirical evidence shows that captives provide non-financial Japanese companies with potential opportunities to boost cash flow from operating activities rather than from investing activities. In particular, a positive relationship between captive utilization and operating cash flow is more likely to materialize when firms have a short public trading history and possess a higher proportion of intangible assets. Overall, we find that captives can improve operating cash flow, which may have positive value implications for shareholders.
Several empirical studies have attempted to clarify what factors alleviate or exacerbate insurance fraud. However, no one has attempted to clarify whether ethics alleviates insurance fraud because there is a lack of sufficient data. Therefore, we use the responses of Japanese consumers to a questionnaire to analyze the characteristics of those who received insurance payments for COVID-19 infections. Our findings are as follows: First, those with lower ethical standards received COVID-19 insurance payments shortly after purchase that raises suspicions of insurance fraud. Second, those who obtained information from a typical messaging app (X(Twitter) or LINE) tended to receive suspicious insurance payments. Third, those with greater insurance knowledge tended not to receive suspicious insurance payments. Collectively, these results indicate what factors encourage and discourage insurance fraud.
This study examines how feelings of regret and rejoicing affect participation in risk-pooling arrangements. The main finding reveals that whether feelings of regret and rejoicing have positive or negative effects on participation in a risk-pooling arrangement depends on the forms of utility, regret, and rejoicing functions and the magnitude of accident probability. In a common situation in which the utility and rejoicing functions are strictly concave, the regret function is strictly convex, the magnitude of the accident probability is small, feeling regret promotes participation in a risk-pooling arrangement, whereas rejoicing might hinder participation. Hence, individuals who experience rejoicing do not participate in a risk-pooling arrangement when the degree of concavity of the rejoicing function is large, whereas individuals who do not experience rejoicing always choose to participate.
Deferred pension benefits provide effective hedging against longevity risk, yet very few people currently opt for them in Japan. This paper proposes a new public pension plan designed to incentivize individuals to delay their pension claim and hedge longevity risk. Specifically, we propose a Hybrid Lump Sum plan (“ Hybrid plan”) that combines elements of the Lump Sum plan and the current deferred plans, achieving both the incentivization of delayed claims and mitigation of longevity risk. First, we evaluate both individual and overall longevity risk under the Hybrid plan through simulation analysis to clarify its effect. Second, through an original questionnaire survey, we show that the Hybrid plan offers a stronger incentive to delay claiming age compared to the current deferred plan. Moreover, the characteristics of deferred beneficiaries are identified using a logistic regression model. Finally, using parameters obtained from the questionnaire survey, we illustrate how the Hybrid plan can reduce overall longevity risk in some instances.
In the context of a globally aging population, nursing care coverage has become a critical issue in insurance research. Japan, where population aging is advancing ahead of global trends, may serve as a reference for many other countries. The country’s public long-term care insurance system, introduced over two decades ago, is administered by 1,571 municipalities (as of 2020), representing a more localized governance structure compared to the 47 prefectures. Half of the system’s financial resources derive from long-term care insurance premiums, which vary across municipalities. Accurate projections of these premiums are essential for informing policies related to social security financing and public fiscal burdens. However, current estimates lack precision. This study projects the long-term care certification rate and insurance premiums at the municipal level for the next 20 years and analyzes regional disparities. Results indicate that nationwide premiums nationwide are expected to increase by 1.5–1.6 times over two decades, with disparities widening between municipalities. By 2040, half of the 20 municipalities with the highest premiums are projected to be in Osaka Prefecture, while substantial increases are also anticipated in commuter areas surrounding Tokyo. These findings underscore the need for targeted policy interventions to mitigate regional inequalities and ensure the financial sustainability of the long-term care system.
Payments to ameliorate disaster damages are typically provided through public assistance programs or insurance. For large-scale natural hazard disasters, governments have historically played a major role in providing this assistance, especially as comprehensive private disaster insurance coverage can rarely be sustained over time. In Switzerland, New Zealand, and Spain, publicly owned disaster insurance schemes were established as early as a hundred years ago. Currently, with an increasing risk of climate change–related disasters, private insurance retreat is happening more frequently in more places, and the need to devise sustainable public disaster insurance (PDI) systems is gaining policy attention. Useful lessons can be extracted from analysing the processes that have led to the establishment of existing PDIs. Using a narrative review, we describe the process of PDI establishment and identify recurrent themes associated with this process, including the incentives involved, role of risk knowledge, government deliberations, legislative proposals, and associated changes in disaster mitigation policies. We analyse the main implications of these past experiences for current and future transitions to PDI systems following private insurance retreat and explore what is still missing in our knowledge about PDIs’ performance.
Experience rating (called “claim experience rating (CER)” in this study), which provides insurance premium discounts without any claims, is widely known. In line with advances in IT, insurance firms can also introduce a new rating (called “effort experience rating (EER)” in this study), which provides insurance premium discounts in an effort to lower the accident probability. This study aims to investigate the rating that is in equilibrium in the competitive market. In particular, we discuss the possibility of asymmetric outcomes in insurance firms choosing different ratings and the critical factor in deciding the ratings of insurance firms. Our main result is that the maximum disutility level for making an effort is a critical factor to decide the ratings of insurance firms and asymmetric outcomes are achieved when it is not high.
This paper analyzes insurance claim count data using a Regime-Switching Integer-Valued Generalized Autoregressive Conditional Heteroscedasticity (RS-INGARCH) model. Weekly claim counts from a Ugandan insurance company, spanning 2020–2024 and covering Motor Private and Motor Commercial lines of business, are used. The study examines both the COVID-19 period (March 2020–January 2022) and the post-COVID-19 period, which saw major structural shifts in claims. The RS-INGARCH model outperforms benchmark models such as the Integer-Valued Autoregressive (INAR) and single-regime INGARCH models, based on Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Mean Squared Error (MSE), and out-of-sample forecasting accuracy. Performance gains were especially notable when the COVID-19 period was included, suggesting the RS-INGARCH(1,1) model better captures abrupt changes in claim dynamics. The Autocorrelation Function (ACF) plots and Ljung–Box test confirms reduced residual autocorrelation under RS-INGARCH(1,1), and the INGARCH(1,1) specification performed best among alternative lag choices. A two-regime specification was preferred over three regimes, as the latter introduced instability and overfitting due to limited data in infrequent regimes. The study recommends RS-INGARCH(1,1) models for forecasting and risk evaluation during periods of structural change, especially in emerging insurance markets.
Liability insurance is a vital source of financial protection and risk management for individuals and organizations. However, its widespread adoption carries certain unintended consequences, including the amplification of liability risks for uninsured and underinsured populations. This may result in a liability insurance consumption spiral, where purchases by some incentivize others to follow suit. The current study provides the first empirical tests of the consumption spiral hypothesis. Using data from 280 Chinese cities over an eight-year period (2011–2018), we find evidence that liability insurance purchases influence those in nearby geographic units and subsequent time periods. These knock-on effects are substantial and support the notion that purchase decisions are positively correlated. Additionally, our results reveal a key externality of liability insurance markets and the ways in which consumption diffuses across space and time.
India introduced a new healthcare scheme, the Pradhan Mantri Jan Arogya Yojana (PMJAY), in 2018, which has been in effect since 2019. This scheme aims to achieve universal health coverage and supports the catastrophic Out-of-Pocket Expenditure (OOPE) of below poverty line population in Indian states. This study assesses the impact of PMJAY in two selected districts of the Indian state of Bihar considering a sample of 324 respondents. Propensity score matching (PSM) and entropy balancing (EB) are two widely used methods to ensure the robustness of impact assessment exercises. The EB method is considered relatively robust and therefore, we use and compare both. The average effect of PMJAY on the beneficiaries using EB shows a significant rise in healthcare utilization; the values are 2.51, 0.79 and 0.36 for OPD visits, hospitalizations, and surgeries, respectively. Our sample consists of 22 percent PMJAY-beneficiaries. The findings highlight the scheme’s potential to improve healthcare access and outcomes for the vulnerable population. Beneficiaries have reported better post-hospitalization quality of life and are more likely to return to work regularly after treatment. These insights may assist policymakers in further improving their effectiveness by raising awareness among the eligible households, as there are still barriers to enrollment and information asymmetries with regard to utilization.
Human resources (HR) are a driver and a risk factor for achieving business objectives. HR risks therefore should form part of risk management to derive the realistic risk exposure of a company. As companies have to inform the public of their relevant risks, HR risks should also be disclosed if they are relevant. However, evidence from existing studies shows that HR risks are not gaining the same attention as other risks, e.g. financial risks. This also stems from the fact that as of today there are no widely accepted conceptualizations for HR risk management and only little empirical evidence on the prevalence of HR risks. We therefore analyzed the annual report of listed companies from the EURO STOXX 50 from the European Union, Nifty 50 from India and Dow Jones Industrial Average from the US. We found that 87 companies mention risks in the area of HR, mainly from recruiting, retention and development. One interesting finding of our study is that companies, which apply the COSO enterprise risk management standard, are far more likely to disclose HR risks. Our study also provides a basis for categorizing HR risks for further conceptualizations.
A captive insurance subsidiary (or captive for short) has been increasingly accepted as a valid risk management solution, formed primarily to insure the loss exposures of its parent company. This study assesses the effect of captive formation on cash flow from an international perspective based on the 2020 S&P Global 100 index constituents. As a wholly-owned insurance company that underwrites the risks faced by its parent company, a captive performs a dual purpose of self-insurance and mutualization that theoretically helps improve cash flow because recaptured premiums can be retained within the corporate structure and efficiently invested. However, our empirical results fall short of evidence to endorse the recent assertion that captives produce better cash flow, implying that recaptured premiums do increase cash hoard under captives at a given point in time, but do not necessarily lead to improved cash flow on an annual basis at the corporate level. In all, our analysis suggests that the financial incentives to form captives may lie not in increasing shareholder value, but in serving corporate risk-financing needs, addressing perceptions of high commercial premiums and maximizing managerial preferences.
This study first examines the dimensions of organizational risk culture as a latent factor, measures its robustness, and assesses its impact on risk management performance through empirical validation in the insurance industry context. Building upon existing literature, a conceptual model is developed with five exogenous factors – “leadership commitment and support,” risk governance,” “risk awareness and communication,” “competency and resources,” and “performance review and rewards’ as antecedents of organizational risk culture and risk management performance as an endogenous factor. Using a descriptive research design and a comprehensive thirty-one-item scale, we collected primary data from Indian general insurance companies to validate the causal model of risk culture using structural equation modeling (SEM). The results indicate that ‘leadership commitment and support’ is the foremost determinant of risk culture, followed by competency and resources, risk governance, and risk awareness and communication. The analysis further reveals that risk culture strongly impacts the risk management performance in the Indian insurance sector. In addition to demonstrating empirical evidence, the study also provides the risk culture assessment tool to risk management practitioners from the insurance industry.
Despite the inherent benefits of usage-based auto insurance (UBI), such as loss prevention and reduced premiums, the UBI market has encountered limitations hindering its expansion in practical terms. This study aims to explore the attitudes of consumers towards UBI and the pivotal factors influencing their decision to adopt UBI. According to the survey, a better understanding of the rationale behind pay-as-you-drive auto insurance (PAYD) and pay-how-you-drive auto insurance (PHYD) plays an important role in policyholders’ acceptance. In addition, consumers with higher auto insurance premiums and individuals willing to share their data are more likely to buy UBI. Our research suggests that the promotion of UBI should be aligned with increasing individuals’ willingness to share personal information and reducing privacy concerns. This could not only increase the willingness to purchase UBI but also help balance discount expectations between policyholder and insurer.
Consistent crop information is vital for the survival of the crop insurance sector, which relies on historical crop data, weather records, meteorological information, and farmers’ details. In India, fragmented, low-quality, and costly data have led to adverse claims ratios, forcing insurance companies to exit the market. This paper proposes an integrated framework that assimilates crop details, quality satellite data, and an actuarial model for crop yield estimation. We use kernel density estimation for risk assessment and emphasize the critical role of bandwidth calculation. Our research indicates that traditional heuristics for bandwidth selection can be misleading. A visualization of the fitted distribution with a frequency histogram can often provide tell-tale signs of an erroneous conclusion from the heuristics. We emphasize the role of the modeller’s judgment in determining the optimal bandwidth that is free from overfitting or over-smoothing. The framework bridges the gap between data and the insurer. The proposed model is of regulatory importance as it solves the issue of missing data and improves risk assessment, which will improve crop insurance market penetration and farmers’ participation and thereby promote stability in the crop insurance sector.
Reserve calculation is crucial for insurance companies. Due to the long-term nature of life insurance products, stochastic interest rate models are more suitable when calculating the premium and reserve of a life insurance product. In this article, we use several popular mean-reverting stochastic interest rate models to study the impact of the model and its parameters on the values of reserves for life insurance products. We employ linear regression, moment estimation, and error optimization methods to calibrate the Vasicek, Cox-Ingersoll-Ross (CIR), CIR#, and Chen models. Our analysis reveals that when applying mean-reverting stochastic interest rate models to a whole life insurance policy, the initial interest rate and the long-term mean have a significant influence on the net reserve value. However, the speed of reversion and volatility only marginally impact the net reserve. Additionally, we observe that the numerical result of the net reserve is sensitive to the time period of the interest rate data and the term structures of the yield rates used in the analysis.
Abstract Previous studies have shown that a single social insurance project may improve subjective well-being. Yet, the effect of the overall social insurance system and the underlying mechanism are unknown. This study investigates the impact of the social insurance system on individuals’ subjective well-being and the impact mechanisms using the China Family Panel Studies (CFPS) conducted in four consecutive waves (2014, 2016, 2018, 2020). By applying the structural equation modeling approach, we construct a focused longitudinal path model and find that the social insurance system has a direct positive effect on individuals’ subjective well-being, measured by life satisfaction and job satisfaction. We explored three longitudinal mediation patterns including self-rated health, household income, and trust in the government, however, none of them are significant mediators. Through subgroup analysis, it is found that the male group and the age over 60 group benefit more from the social insurance system regarding life satisfaction improvement.
Abstract This study employs a Poisson model with Generalized Estimating Equations (GEE) to identify the key factors in the Moroccan auto insurance rating system. Crucial factors identified include gender, tariff-determining usage codes, vehicle usage, and driver experience. A comparison of this study model with Morocco’s current rating system reveals the presence of pricing suboptimality, which can be improved by enhancing the incorporation of driving history and adjusting the Bonus–Malus system to better reflect actual claim patterns and risk profiles.
