The actuarial valuation of insurable risks in a changing risk landscape
This joint PhD project will be based at KU Leuven with a minimum 12 month stay at The University of Melbourne.
Project title: VALERIA: Valuation and Advanced Learning methods for Emerging, global Risks In Actuarial science
The insurance industry faces fundamental changes that will not be tackled by incremental improvements of existing techniques, but call for entirely new insurance pricing paradigms. The dynamics of emerging risks such as cyber and weather related risks need to be handled with little or no past data. At the same time, for more traditional covers the wealth of data that is collected now presents new challenges (e.g., computational or ethical) and opportunities (e.g., statistical power). Bringing together the Leuven-based expertise on machine learning practice for insurance data with the knowledge on stochastic processes, behavioral data and dependencies from the Melbourne team, this PhD project will focus on:
- formulas for discrimination-free insurance pricing;
- predictive modeling tools for the actuarial valuation of emerging risks; and
- the creation of data analytic tools for a customer-centric, usage based insurance paradigm that bundles selected products and even services.
Differential pricing is a foundation of modern-day insurance and deals with the actuarial valuation of risk by calculating a fair price for a new policy sold to a given risk profile. The so-called best estimate price is calculated as the ‘expected frequency times expected impact’ of the insured event resulting from standard regression models (the generalized linear models, or GLMs) for claims data. These predictive models are key in a business that is highly regulated, strongly valuing the explainability of the algorithms driving decisions with impact on customers. After selling a contract to a client, the insurer is liable for the claims arising from this contract. Capital must be held to meet these future liabilities. Calculating the necessary amount of capital is the job of a reserving actuary. Even though these key actuarial tasks are treated in silos in current insurance practice and literature, reserving is the mere continuation of pricing. Whereas pricing happens at the onset of the insurance policy – before any coverage has been provided – reserving is, in some way, an updated pricing of the insurance policy. The pricing actuary values the total loss on a policy from ground-up, while the reserving actuary assesses the total loss in the presence of some (though incomplete) information on the development of occurred claims.
As such, we will design dynamic, responsive and resilient pricing and reserving techniques for traditional but also emerging risk types, including machine learning methods that balance predictive value and acceptability by major stakeholders (e.g. explainable to management, discrimination-free pricing). Access to real data and strong links with practice will ensure applicability and relevance of our developments.
Bringing together the available expertise (in Leuven) on machine learning practice with the knowledge on stochastic processes and dependencies (from Melbourne) the first PhD project will focus on (1) a probabilistic framework for discrimination-free pricing in tariff plans, (2) predictive modeling tools for the actuarial valuation of emerging risks and (3) the creation of data analytic tools for a new insurance paradigm: customer-centric, usage based, bundling selected products and even services.
The project will be complemented by the project on Combined actuarial and financial valuation of hybrid insurance liabilities and the collaboration will ensure a successful completion of the project
Principal Investigators (PIs)
Co-Principal Investigators (co-PIs)