Project staff and support

Pascal Ah-Kine (Intern, University of Southampton)
Neil Robjohns (Company Supervisor, Barnett Waddingham)
Jon Forster (Academic Mentor, University of Southampton)
Melvin Brown (Technology Translator, Industrial Mathematics KTN)

This Internship project was hosted by Barnett Waddingham on behalf of The Actuarial Profession, in conjunction with the University of Southampton. It is part of the KTN's Industrial Mathematics Internships Programme, co-funded by EPSRC. Start date: October 2009; duration: 4 months.

Barnett Waddingham is an independent firm offering actuarial and consultancy services. The Intern worked with the firm’s Insurance Team, which provides actuarial advice and services in relation to life, health and general insurance.

The Actuarial Profession’s Continuous Mortality Investigation (CMI) carries out research, on behalf of the UK actuarial profession, into the mortality and morbidity experience of various types of long-term insurance and pensions business. Barnett Waddingham provides secretariat services to the CMI. This internship project relates specifically to analysis of CMI data.

Barnett Waddingham sought to account for the nature of the CMI’s pooled data and the heterogeneity within it, dividing explanatory factors into publishable (eg: year, age, gender, duration) and non-publishable (eg: office). The aim was to develop a model to allow analysis and publication of mortality / morbidity experience adjusted for changes in the mix of data, by non-publishable variables, over other explanatory variables (particularly time). Model parameters were to be computed with estimates of uncertainty and the viability of approach were to be examined.

The intern developed Bayesian hierarchical statistical models to account for the variability in insurance data due to office mix. This required developing computational algorithms to obtain the required estimates (adjusted for office mix).

Project description

The CMI collects data from a large number of contributing insurance companies and pension schemes, and analyses and reports on the experience of the whole pool of data. The investigations are independent and continuous, based on annual collections of data. Confidentiality of each contributor’s own data and results is paramount. The pooled data provides a view of market-level experience and gives greater credibility of results for analyses at a more granular level than any contributors could obtain from their own data. However, as experience varies significantly between insurers, such analyses are vulnerable to distortions arising from variations in the mix of insurers. These distortions may particularly affect the apparent patterns of experience over time and across age.

This project has led to the development of a new methodology in modelling and analysing the data available to the CMI. The use of Bayesian methods in producing reports has led to more effective interpretation of mortality and morbidity experience results produced from pooled data from a diverse range of insurance companies.

“The internship provided Pascal with a great opportunity to develop his statistical skills and to bring the advantages of modern statistical methodology to bear on a important practical problem. I expect that the work he has done will bring real benefits to the CMI.” said academic mentor Jon Forster, University of Southampton.