The general principles for determining the financial performance of a company is that revenue is earned as goods are delivered or services provided, and that expenses in the period are made up of the costs associated with this earned revenue. In the insurance industry, premium payments are typically made upfront, and can provide coverage for several years, or be paid many years before the coverage period starts. The associated costs are often not fully known until many years later. Hence, complexity arises both in determining how a premium paid should be earned over time, and in valuing the costs associated with this earned premium. IFRS 17 attempts to align the insurance industry with these general accounting principles. We bring this new accounting standard into the realm of actuarial science, through a mathematical interpretation of the regulatory texts, and by defining the algorithm for profit or loss in accordance with the new standard. Furthermore, we suggest a computationally efficient risk-based method of valuing a portfolio of insurance contracts and an allocation of this value to subportfolios. Finally, we demonstrate the practicability of these methods and the algorithm for profit or loss in a large-scale numerical example.

No abstract

We consider a simple long short-term memory (LSTM) neural network extension of the Poisson Lee-Carter model, with a particular focus on different procedures for how to use training data efficiently, combined with ensembling to stabilise the predictive performance. We compare the standard approach of withholding the last fraction of observations for validation, with two other approaches: sampling a fraction of observations randomly in time; and splitting the population into two parts by sampling individual life histories. We provide empirical and theoretical support for using these alternative approaches. Furthermore, to improve the stability of long-term predictions, we consider boosted versions of the Poisson Lee-Carter LSTM. In the numerical illustrations it is seen that even in situations where mortality rates are essentially log-linear as a function of calendar time, the boosted model does not perform significantly worse than a simple random walk with drift, and when non-linearities are present the predictive performance is improved. Moreover, boosting allows us to obtain reasonable model calibrations based on as few data points as 20 years.

No abstract