Journal of Risk and Insurance最新文献

In the underwriting and pricing of nonlife insurance products, it is essential for the insurer to utilize both policyholder information and claim history to ensure profitability and proper risk management. In this paper, we apply a flexible regression model with random effects, called the Mixed Logit-weighted Reduced Mixture-of-Experts, which leverages both policyholder information and their claim history, to categorize policyholders into groups with similar risk profiles, and to determine a premium that accurately captures the unobserved risks. Estimates of model parameters and the posterior distribution of random effects can be obtained by a stochastic variational algorithm, which is numerically efficient and scalable to large insurance portfolios. Our proposed framework is shown to outperform the classical benchmark models (Logistic and Lognormal GL(M)M) in terms of goodness-of-fit to data, while offering intuitive and interpretable characterization of policyholders' risk profiles to adequately reflect their claim history.

About a quarter of long-term care insurance (LTCI) policy holders aged 65 let their policies lapse before death, forfeiting all benefits. We find that lapse rates are substantially higher among the cognitively impaired in the Health and Retirement Study. This generates a pernicious form of dynamic advantageous selection, as the cognitively impaired are more likely to use care. Simulations show that an inappropriately optimistic asset drawdown path further increases the individual welfare cost of unanticipated lapses. Meanwhile, we find evidence of a significant but very small role for either strategic or financial motives for lapsing.

On-demand insurance is an innovative business model from the InsurTech space, which provides coverage for episodic risks. It makes use of a simple fact in a practical way: People differ in their frequency of exposure as well as the probability of loss. The extra dimension of heterogeneity can be used to screen the insured and shifts the utility-possibility frontier outward. We provide a sufficient condition under which type-specific full insurance at the actuarially fair price is incentive compatible. We also show that our results hold for various real-world implementations of on-demand insurance.

Technological progress has improved insurers' ability to monitor policyholders and has led to usage-based insurance (UBI) contracts that incorporate behavioral risk factors in pricing. Economic theory predicts that any informative monitoring signal is adopted in equilibrium. In practice, the demand for UBI is still low to date with market shares in the single digits. We modify the standard moral-hazard model in insurance economics by trading off a simpler effort model for a richer strategy space, and by focusing on the use of monitoring for premium differentiation. In our model, an informative monitoring technology is in use if it is sufficiently accurate. Otherwise, the premium incentive from monitoring is not large enough to alleviate the incentive-compatibility constraint to an extent that would make policyholders better off. Our results help explain the slow adoption of UBI contracts in practice and provide an avenue to increase their appeal to policyholders.

Ignoring the effects of inflation in retirement planning can have severe consequences for an individual's future financial well-being. Yet, many pension funds do not communicate inflation-related information, presumably for the fear of reduced contributions once the members understand how low the “real” return on saving for retirement is. As an alternative prediction, the provision of inflation information could increase pension contributions, because it reveals possible pension shortfalls. In cooperation with a major German pension fund, we conduct a field experiment, in which we vary the inflation information provided to the fund members, to explore this important issue. Among all participants, we find mostly positive but insignificant effects of the inflation information on pension contributions. Among those participants who voluntarily changed their pension contributions after the experimental intervention, the provision of inflation information significantly raises the likelihood of increasing pension contributions.

We present a model of a longevity risk transfer market with different market players (primary insurers, reinsurers, and capital market investors) and investigate how market dynamics and the market players' roles evolve with progressing market saturation. We find that reinsurers' appetite for longevity risk is the key driver in the early stage of market development. Since diversification benefits with other businesses decrease with every transaction, the reinsurance market is intrinsically antimonopolistic. With the increasing saturation of the reinsurance sector as a whole, its competitiveness shrinks leading to rising expected risk-adjusted returns for capital market investors. We show that in a saturated market, reinsurers should assume the entire longevity risk from primary insurers, diversify it within their business mix, and subsequently pass on only specific (nondiversifiable) components of the longevity risk to the capital markets. Our findings provide valuable suggestions on how to make the best use of the market's limited risk absorption capacity.

We analyze insurance demand when insurable losses come with an uninsurable zero-mean background risk that increases in the loss size. If the individual is risk vulnerable, loss-dependent background risk triggers a precautionary insurance motive and increases optimal insurance demand. Prudence alone is sufficient for insurance demand to increase in two cases: the case of fair insurance and the case where the smallest possible loss exceeds a certain threshold value (referred to as the large loss case). We derive conditions under which insurance demand increases or decreases in initial wealth. In the large loss case, prudence determines whether changes in the background risk lead to more insurance demand. We generalize this result to arbitrary loss distributions and find conditions based on decreasing third-degree Ross risk aversion, Arrow–Pratt risk aversion, and Arrow–Pratt temperance.

Fraud is a significant issue for insurance companies, generating much interest in machine learning solutions. Although supervised learning for insurance fraud detection has long been a research focus, unsupervised learning has rarely been studied in this context, and there remains insufficient evidence to guide the choice between these branches of machine learning for insurance fraud detection. Accordingly, this study evaluates supervised and unsupervised learning using proprietary insurance claim data. Furthermore, we conduct a field experiment in cooperation with an insurance company to investigate the performance of each approach in terms of identifying new fraudulent claims. We derive several important findings. Unsupervised learning, especially isolation forests, can successfully detect insurance fraud. Supervised learning also performs strongly, despite few labeled fraud cases. Interestingly, unsupervised and supervised learning detect new fraudulent claims based on different input information. Therefore, for implementation, we suggest understanding supervised and unsupervised methods as complements rather than substitutes.

Accurate pricing of crop insurance policies relies on forecasts of probability densities of crop yields. Yield densities are dynamic, time series data on yields are often limited, and yield data are spatially correlated. We examine linear pooling of potentially related, but almost surely misspecified, crop yield density forecasts. The pooled forecasts combine densities from other spatial units based on out-of-sample forecast performance. The pooled densities result in more accurate premium rates which can reduce incentives for adverse selection. The approach is applicable to any insurance setting where the statistical model for the loss variable is likely to be misspecified and the underlying data-generating processes are potentially related.