University of the Aegean
A new approach to dynamic stochastic mortality modelling for managing mortality-longevity risk
Abstract
dc:descriptionThe continuous increase in life expectancy observed in recent years, especially in developed countries, as well as the actuarial risks raised due to the ageing population have triggered scientists and actuaries to develop appropriate models in order to model and project mortality rates. A great enabler to achieve this is the stochastic mortality models which serve as a steppingstone for the development of plans for managing the mortality risk for the sake of safeguarding the solvency of insurance organisations. In this context, this doctoral thesis aims to further explore the domain of mortality modelling and identify the methodologies empowering the existing models, highlight their advantages and disadvantages and use them to draw a research line for further contributing to the domain. Based on this analysis, this thesis introduces and evaluates a new multiple-component stochastic mortality model, namely the Hatzopoulos-Sagianou (HS) model, which goes beyond the prominent solutions, aiming to address the identified limitations. HS offers a dynamic model structure and is based on a parameter estimation methodology, which aims to reveal significant and distinct age clusters by identifying the optimal number of incorporated period and cohort effects. The latter maximise the captured variance of the mortality data, enable the attribution of an identified mortality trend to a unique age cluster and contribute to the interpretability of the modelling results. Having HS as the basis, this thesis further extends the model through the use of various link functions, and diverse distributions in the model's estimation method and differentiates the HS approach by modelling the number of deaths using the Binomial model. In addition, the HS is reformed by introducing a new form of link functions with a particular focus on the use of heavy-tailed distributions leading to an enhanced estimation methodology, improving even more the model's goodness-of-fit, and producing more fine-grained age clusters. The above-mentioned developments lead to a software implementation of the HS in the form of a Matlab toolbox called: HS-Tool. The latter provides the means for replicating the full cycle of a mortality model, i.e., fitting the stochastic mortality model, assessing its goodness-of-fit and performing mortality projections. The HS-Tool introduces codebase-related improvements that contribute to the model's ability to provide a “plug 'n' play” user experience, while the integration of multi-criteria decision making methods in the model's workflow achieve increased tool automation and convergence to “optimal” values for critical model parameters.The model's qualities are supported by experimental results showing that the HS model is able to achieve high scores over diverse qualitative and quantitative evaluation metrics, while the identified mortality rates come into agreement with well--established findings of the mortality literature.
Degree
thesis:*- Grantor dc:publisher
- University of the Aegean
- Year dc:date
- 2022
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Sagianou, Aliki
- Σαγιάνου, Αλίκη
Subjects
dc:subject × 19- Στοχαστικά μοντέλα θνησιμότητας
- Στοχαστική μοντελοποίηση θνησιμότητας
- Διαχείριση κινδύνου μακροζωίας
- Γενικευμένα γραμμικά μοντέλα
- Πολυκριτηριακή λήψη αποφάσεων
- Stochastic mortality models
- Stochastic mortality modelling
- Longevity risk management
- Generalised linear models
- Multi creteria decision making
- Sparse principal component analysis
- Μαθηματικά
- Φυσικές Επιστήμες
- Μαθηματικά, άλλοι τομείς
- Μοντελοποίηση και Προσομοίωση
- Mathematics
- Natural Sciences
- Mathematics (miscellaneous)
- Modeling and Simulation
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/52824