UK – Academics in London have come up with a new way of predicting how long groups of people will live, which they claim could transform the way pension funds, annuity providers and governments forecast mortality rates.
The new blueprint for building mortality models has been devised by academics at the Pensions Institute at Cass Business School, which is part of City University London.
David Blake, co-author of the research and director of the Pensions Institute, said: "Rather than propose yet another new mortality model, we outline and implement a general procedure for building a mortality model from scratch."
This 'general procedure' can be applied to any set of data to construct a mortality model to fit all ages across the population, the institute said.
The procedure is presented as addressing problems with the many new mortality models that have been developed in recent years.
These models often involve ad-hoc extensions to existing models that have "questionable demographic significance", the institute said.
On top of this, the new models have trouble giving realistic forecasts of specific mortality rates, it said.
"The general procedure is a way of constructing mortality models that are tailored to specific datasets," said Blake.
"This means it is able to identify the idiosyncratic features of different populations, which conventional off-the-peg models are unable to do."
The procedure works by sequentially extending a simple mortality model, said Andrew Hunt, also co-author of the study.
First the model is extended with freely varying age effects, which take whatever shape fits the data best, and these are then replaced with a simpler, parametric age function that does the same job, he said.
"It uses a combination of expert judgement and a toolkit of functional forms," said Hunt.
"This then achieves a good fit to the data with a relatively parsimonious model whose age effects can then be interpreted in light of the underlying socio-economic and demographic drivers of changing mortality rates."
After testing on UK data, the general procedure easily did better than simpler models such as the Lee-Carter model, the institute said.
In addition to producing more economical models than those generated using a mechanical algorithm, the procedure also gave well-specified cohort effects, which the institute said are essential for reliable forecasting of mortality rates.