Events Library

This work aims at providing a possible solution for the reduction of the gap between supply and demand of cyber insurance. We set up a model for the dynamics of data breaches trough an Integer Valued GARCH model to consider time-dependent effects. In order to better capture the dynamics of data breaches we refer to INGARCH models that assume data breaches have a zero-inflated Negative Binomial distribution with the purpose of considering the fact that, in some cases, the victims of the breach do not report the attack. We refine loss estimates and risk assessment by looking for lead-lag relationships between cyber risks and economic variables (as in De Giovanni et al. (2021)) . We define a parametric insurance product as a possible way to overcome some of the issues to be faced in analyzing and managing cyber risks described in details in Eling et al. (2021), Eling et al. (2022), Malavasi et al. (2022) and discuss the implications through numerical applications.

Actuaries are used to perform complex computations for a long time. Be it for reserving, pricing, mortality analysis or ALM among others, several skills are inherent to the role of actuaries (as data management, predictive modelling, coding, operationalization, reporting,...) in their goal to support decision-making processes. With the emergence of data science and new technologies, actuaries must extent their tool box in order to improve the efficiency, robustness and overall performance of their analysis. In this presentation, we introduce new trends and techniques that actuarial engineers should master (at least some of them): (i) creative data sourcing and management: use of external (open) data, Natural Languages Processing and features engineering are key skills to complete databases and avoid the ""shit in, shit out"" effect. (ii) use of advanced predictive modelling techniques: machine learning offers new opportunities but must be handled with care to avoid overfitting or black-box effect. Mastering machine learning techniques, cross-validation and model interpretability tools is essential. (iii) upgraded coding and operationalization skills: open sources software like R or Python allow actuarial engineers to implement their own algorithms and develop robust tools/application supporting their computation. Mastering these software together with e.g. GitHub Copilot to accelerate coding with AI, container tools (e.g. Docker) to operationalize computation engines or Sphinx for automatic documentation will bring actuaries to an absolute other level. (iv) innovative reporting tools: Excel remains essential but other tools like PowerBI, R/Shiny or Python Dash could be quick wins for actuaries to improve their reporting and make it available to the management. We strongly believe that there is a sweet spot for actuarial engineers between business and IT to develop robust actuarial and financial computation engines improving the decision-making process. Thanks to their business knowledge, actuaries will be better place than any profession to helps insurance companies thrive should they improve their technological skills

At our independent consultant firm, we work for Insurance Companies (ICs), they hire us because they need agility, an independent point of view, an adequate speech for different audiences (actuary, finances, commercial, product development and even underwriting, legal, IT and auditory) or a combination among them. Being an external party allows us to have the perspective that the day-to-day duties sometimes do not allow the IC to answer questions that go beyond product or reserve compliance, maintenance, or development. Using data from the “Claims File Book” (CFB) for “Health High Cost” (HHC), I developed a Predictive Model that answers the question: How much will I pay for a received claim? The answer to this question would allow the IC to estimate optimal opening reserves, support finance planning among other business objectives. In other words, we may use modern approaches to answer business questions that seem simple, but fundamental. In this paper, the predictive model is described and some of the implications of using data science approaches for ICs in Colombia and undeveloped markets are discussed. Final remarks: The paper that I am submitting was written for the “Captstone” course of the “Data Science Professional Certificate” that I finished last year (https://www.edx.org/certificates/professional-certificate/harvardx-data-science). I am adapting the document for this event, in the next few weeks, I will submit the final document.

There have been several studies on the determinants of insurance consumption in developing countries and particularly in sub-Saharan African countries. Most of them focus on insurance consumption in rural areas and reach the conclusion that financial illiteracy and lack of liquidity constrain insurance consumption in rural area and mostly farmers. Given the proven growing African middle class, educated and with growing revenue, we focus in this study on the middle class with the aim to identify the determinants of insurance consumption. Using data from Harmonized Survey on Households Living Standards 2018-2019 in Côte d’ivoire, the study identifed that, the main determinants are the marital status, the literacy, the Socio-professional category and the employer, the revenue, the living area and also the holding of another insurance. Furthermore we highlight the under-consumption paradox of some type of insurance (life and household) given the identified determinant which need further analysis.

This paper seeks to examine the application and effectiveness of Generalized Linear Models (GLM) in estimating Incurred But Not Reported (IBNR) reserves within diverse insurance portfolio contexts. Differing from prior studies, our focus extends to evaluating GLMs in demanding scenarios, including portfolios characterized by escalating loss ratios, the integration of new portfolios, and those featuring outlier claims. Drawing upon the foundational work of some previously published articles which emphasized the efficacy of GLMs in modelling IBNR, our research delves into the robustness of these models under more intricate conditions. Through a comprehensive comparative analysis and sensitivity test, we aim to unravel how GLMs respond to variations in portfolio characteristics. The expected results might reveal crucial insights into both the adaptability and constraints of Generalized Linear Models (GLMs). These findings will provide valuable perspectives for actuaries and researchers seeking to enhance the precision of IBNR estimates within complex and dynamic environments. This research contributes to the ongoing discourse on refining actuarial methodologies for more accurate and reliable reserve estimations in the ever-evolving insurance portfolios.

Guaranteed Minimum Accumulation Benefits (GMAB) are retirement savings vehicles that protect the policyholder against downside market risk. This article proposes a valuation method for these contracts based on physics-inspired neural networks (PINNs), in the presence of multiple financial and biometric risk factors. A PINN integrates principles from physics into its learning process to enhance its efficiency in solving complex problems. In this article, the driving principle is the Feynman-Kac (FK) equation, which is a partial differential equation (PDE) governing the GMAB price in an arbitrage-free market. In our context, the FK PDE depends on multiple variables and is difficult to solve using classical finite difference approximations. In comparison, PINNs constitute an efficient alternative that can evaluate GMABs with various specifications without the need for retraining. To illustrate this, we consider a market with four risk factors. We first derive a closed-form expression for the GMAB that serves as a benchmark for the PINN. Next, we propose a scaled version of the FK equation that we solve using a PINN. Pricing errors are analyzed in a numerical illustration.

The supreme body of the pension fund (board of trustees) is responsible for the overall management of the pension fund. Due to the fact, that the membership in the board of trustees is not a full-time job, the scope of their duties is enormous: the fulfilment of the mentioned legal requirements needs a lot of time and expertise. Pension fund accredited actuaries, investment consultants, auditors as well as pension fund management teams should fully support the board of trustees to make proper decisions. The reliable liability forecasting is very important for setting of the pension fund financing system and controlling it (with risk budgeting). Due to the fact, that many liability parameters are contingent on the development of yield curves and inflation, it is worth preparing the analysis of their historical data, visualizing them, and additionally reliably forecasting. It is very important to prepare the yield curve and inflation forecasting with the deep learning approach, visualization of such results and based on them the liability forecasting. Such kind of analysis helps the board of trustees to make their decisions and better understand the forecasting results (compared to affine models). Such approaches are useful as well for the forecasting of the international accounting results (IFRS, US GAAP, IPSAS) and for preparing the asset allocation to be the strong third contributor.

Modelling policyholder lapse behaviour directly impacts profitability and reserving for health insurers. Traditional experience studies methods, focusing on aggregate lapse rates, often fail to account for individual lapse behaviour. This limitation hampers the effective implementation of targeted retention strategies. In this study, we apply a gradient boosting tree-based modelling framework to forecast individual lapse rates. This framework allows us to extract information from a wide range of company data from Taiwan, including policy structure and individual records that are often overlooked by traditional models. The model ranks all in-force health policies by their relative probability of withdrawal in the next year. The model achieves a cross-validated AUC of 0.87, indicating good model performance. Among the top 5% of policies predicted to have the highest lapse rates, the average lapse rate was found to be 8.5 times greater than the overall average. This result underscores the model's effectiveness in pinpointing customers with a high likelihood of policy lapse. By adopting our approach, health insurers can more accurately identify targets for retention management and enhance profitability.

The insurance market is subject to increasing challenges over the last decade requesting a wise capital allocation. To respond to those challenges, solutions are available on the market, including tailor-made finality solution. The regulation in Europe has been highly demanding. Governance around Solvency II has strengthen the insurance industry to many aspects including costs and operational requirements. The market players had to genuinely integrate the numerical era and the exploitation of data to drive their management. The economic conditions have been fluctuating after several crisis leaving the market at an inflationary turning point in 2022. Those have been instrumental to the hardening of the capital source including the onerous access to reinsurance. The outlook for the future requires the companies to pay full attention on the operational efficiency and the insurability of the market, including in the natural catastrophe dimension. Today companies have to manage the future as well as the past as a trade of capital allocation. This paper will explore the opportunity of redeploying capital efficiently by isolating some legacy business and outline the key attention points in such transaction.

In the frame of solvency regulation, Reinsurance is usually bought to reduce the Solvency Capital Required, as part of risk mitigation techniques associated with the underwriting risk. Reinsurance may bring an effective risk transfer as far as the risk sharing mechanism is known in advance. When the ultimate net result depends on the claim experience, the risk sharing mechanism reduces. We have seen in the past many reinsurance contracts purchased by insurance undertakings with limited risk transfer, which are wrongly used to reduce the capital requirement. We provide several examples of reinsurance contract with limited or no risk transfer, we provide a measure of the risk transferred to the Reinsurers, considering the reinsurance conditions. With respect to excess of loss reinsurance, contracts including paid reinstatements or premium based on the claim experience (sliding scale premium) reduce the transfer of risk. In proportional reinsurance, the commissions based on the loss ratio/burning cost (sliding scale commission) may also reduce the transfer of risk. Our objective is to draw the attention to the Actuaries/Risk Managers of possible misuse of reinsurance as a risk mitigation technique and to provide a measure of the risk transferred to Reinsurers.