Talk Title: Hidden-Factor Modeling of Climate-Driven Losses: An Illustration Using U.S. Crop Indemnities 

Abstract: Climate risk induces complex dependence across losses at geographically distant locations. Accurately forecasting the joint distribution of losses is therefore important for portfolio-level risk management, yet remains challenging and relatively underexplored. Using U.S. crop insurance indemnities as a representative setting, this study argues that dependence in climate-driven losses arises from two distinct sources: local spatial spillovers and latent common climate factors. To jointly capture these sources of dependence, we adopt a spatial panel regression model that incorporates both a spatial autoregressive component and a latent factor component. Using data from multiple representative counties across five Farm Resource Regions, we generate one-year-ahead predictive distributions at the county level and subsequently the predictive distribution of aggregate crop indemnities. We compare the proposed model with baseline models that exclude either the latent factor component or both sources of dependence. The results show that the model ignoring both components substantially underestimates aggregate indemnities, while the model relying solely on spatial spillovers misattributes the effects of latent factors to local spillovers. In contrast, the main model, by allowing for heterogeneous responses to common climate regimes, hence, the possibility of capturing negative dependence, achieves markedly improved in-sample fit, more accurate out-of-sample median forecasts that closely track realized aggregate indemnities, and substantially tighter prediction intervals without loss of coverage. These improvements translate into a more economically grounded assessment of diversification potential, more balanced capital allocations, and more credible stress-testing outcomes. 

Talk Title: Lambda Expected Shortfall

Abstract: The Lambda Value-at-Risk (Λ-VaR) is a generalization of the Value-atRisk (VaR), which has been actively studied in quantitative finance. Over the past two decades, the Expected Shortfall (ES) has become one of the most important risk measures alongside VaR because of its various desirable properties in the practice of optimization, risk management, and financial regulation. Analogously to the intimate relation between ES and VaR, we introduce the Lambda Expected Shortfall (Λ-ES), as a generalization of ES and a counterpart to Λ-VaR. Our definition of Λ-ES has an explicit formula and many convenient properties, and we show that it is the smallest quasi-convex and law-invariant risk measure dominating Λ-VaR under mild assumptions. We examine further properties of Λ-ES, its dual representation, and related optimization problems.

Talk Title: Financial Valuation of Variable Annuities with Pandemic Risk

Abstract: In this talk, we consider how an insurance company manages insurance contracts that relate to financial markets, such as equity-linked insurances or variable annuities. We introduce stochastic mortality, which is not necessarily independent from the financial market. This is a key feature of the recent Corona crisis: increasing mortality together with falling stock prices. We propose a model that is able to capture uncertain circumstances and compute the associated impact of the pandemic risk.

Talk Title: Valuation of Insurance Liabilities: Merging Market- and Model-Consistency. 

Abstract: In this talk, we consider the fair valuation of liabilities related to an insurance policy or portfolio in a single period framework. We define a fair valuation as a valuation which is both market-consistent (mark-to-market for any hedgeable part of a claim) and model-consistent (mark-to-model for any claim that is independent of financial market evolutions).

We introduce the class of hedge-based valuations, where in a first step of the valuation process, a best hedger for the liability is set up, based on the traded assets in the market, while in a second step, the remaining part of the claim is valuated via a model-consistent valuation. We also investigate the class of two-step valuations, the elements of which are closely related to the two-step valuations which were introduced in Pelsser and Stadje (2014). We show that the three introduced classes of valuations (fair, hedge-based and two-step) are identical.

Several simple examples are given to illustrate the concepts introduced.

Keywords: Fair valuation of insurance liabilities, market-consistent valuation, model-consistent valuation, mean-variance hedging

Talk Title: Bayesian Mortality Forecasting with a Conway-Maxwell-Poisson Specification

Abstract: This work presents a novel approach to stochastic mortality modelling by using the Conway--Maxwell--Poisson (CMP) distribution to model death counts. Unlike standard Poisson or negative binomial distributions, the CMP is a more adaptable choice because it can account for different levels of variability in the data, a feature known as dispersion. Specifically, it can handle data that are underdispersed (less variable than expected), equidispersed (as variable as expected), and overdispersed (more variable than expected). We develop a Bayesian formulation that treats the dispersion level as an unknown parameter, using a Gamma prior to enable a robust and coherent integration of the parameter, process, and distributional uncertainty. The model is calibrated using Markov chain Monte Carlo (MCMC) methods, with model performance evaluated using standard statistical criteria such as residual analysis and scoring rules. An empirical study using England and Wales male mortality data shows that our CMP-based models provide a better fit for both existing data and future predictions compared to traditional Poisson and negative binomial models, particularly when the data exhibit overdispersion. Finally, we conduct a sensitivity analysis with respect to prior specification to assess robustness. This is joint work with J.S.T. Wong, University of Essex. 

Talk Title: Insurance Experience Rating with Endogenous Deductible Choices

Abstract: Insurers adjust premiums for renewing policyholders based on past claims, a practice known as experience rating. This paper examines experience rating in insurance contracts with deductibles, a key instrument for mitigating inefficiencies arising from information asymmetry between policyholders and insurers. We propose a copula-based panel data model that accounts for the endogeneity of deductible choice, particularly when the assumption of contemporaneous exogeneity is violated—consistent with theories of adverse selection and moral hazard. We also introduce a computationally efficient algorithm for inference and prediction. The framework accommodates various types of non-Gaussian outcomes and is well-suited for predictive applications. We apply the method to a government property insurance program, using historical claims data to develop an experience rating scheme. Results reveal a negative relationship between deductible choice and underlying risk, providing empirical support for endogenous selection behavior. Compared to standard approaches that treat deductibles as exogenous, our model enables more refined risk segmentation and improved identification of profitable business.

Talk Title: Constrained Portfolio Optimization with Estimation Risk: Theory and ESG Applications

Abstract: This study investigates constrained mean-variance portfolio optimization while accounting for estimation risk in the first two moments of asset returns. We begin by characterizing the optimal constrained portfolio in the absence of estimation risk and derive its analytical structure. When the mean vector and covariance matrix need to be estimated from historical data, the standard plug-in approach may lead to substantial out-of-sample performance deterioration. To mitigate this issue, we propose a combined three-fund portfolio construction tailored to the constrained setting. The optimal combination coefficients are obtained by maximizing the expected out-of-sample utility subject to the constraint. We adopt the Environmental, Social, and Governance (ESG) constraint as an empirical illustration. Through extensive empirical analysis, we show that the proposed combined portfolio consistently outperforms the plug-in portfolio in terms of certainty equivalent return, demonstrating its effectiveness in addressing estimation risk under the constraint settings.

Talk Title: Representation Learning for Ratemaking and Reserving: Using Auxiliary Data to Enhance Loss Models

Abstract: Actuarial loss models for ratemaking and reserving are often constrained by sparse data, short histories, and limited covariates, even when rich auxiliary information exists outside the loss dataset. This talk presents a unified, probabilistic view of representation learning in which each unit of analysis (county, policy, or claim) has an underlying latent risk state that generates observable auxiliary signals (weather events, policy-transaction histories, or claim narratives) and also drives losses. By estimating compact embeddings from these auxiliary signals and incorporating them into familiar actuarial models, we can improve predictive accuracy, risk classification and stability in low-credibility settings. In a case study, we illustrate how pretrained language representations can convert claim narratives into predictors within a regression-based frequency-severity loss reserving model, improving reserve forecasts. We conclude with practical considerations for transparency, interpretability, and model governance as increasingly central themes in the AI era of actuarial science.

Talk Title: CADA-Flow: Capturing Complex Dependence in Cyber Breach Risk via Deep Learning

Abstract: Large-scale cyber data breach incidents have occurred frequently in recent years, posing substantial challenges to global cybersecurity and risk management systems. There is strong demand for systematic and effective approaches to modeling and pricing cyber data breach risk. Reliable prediction is mainly hindered by data sparsity,  and complex dependence patterns across time, geography, and industry sectors. We develop a two-part deep learning framework that jointly models (i) the probability of observing a breach event and (ii) the conditional distribution of loss severity given an event. The framework adopts a unified encoder--decoder architecture: an attention-based recurrent encoder learns dynamic temporal--spatial representations from historical signals, while two specialized decoders handle the zero-event component and the nonzero severity component, respectively. For severity, we employ a conditional normalizing-flow decoder to produce flexible full-distribution forecasts with improved tail behavior and calibration. In an empirical study, the proposed framework consistently outperforms classical statistical approaches and competitive deep learning baselines in distributional predictive accuracy. We further show how the resulting predictive distributions translate into insurance-relevant outputs, including portfolio-level risk measures and group- and firm-level premium indications.