Information Matrix

Last year Iain wrote about a smooth model to identify mortality shocks, using Swedish population data to illustrate the impact of the 1918 influenza pandemic. 

One of the most written-about models for stochastic mortality projections is that from Lee & Carter (1992). 

A particular leitmotif of 2010 is productivity — getting more work done with the time and resources available.  Often this is about controlling costs, but in the insurance sector in the European Union it is also about adapting to scarcity of resources: with Solvency II looming, there is strong Europe-wide demand for actuarial expertise.

Stephen has written about the challenges in using population cause-of-death data for mortality analysis and forecasting. Another potential source of data is computerised patient records such as the General Practice Research Database (GPRD).

On 18th May 2010 we ran a seminar on mortality projections for clients, including hands-on use of the Projections Toolkit.

We have often written about how modelling the force of mortality, μx, is superior to using the rate of mortality, qx.

Trends in cause of death can be an instructive way of looking at past mortality, although we have previously seen that we have to be very careful that an apparent "trend" is not due to changes in recording.  Leaving aside the problems of shifting classification over time, what of the categories themselves?

There are two fundamentally different ways of thinking about how mortality evolves over time: (a) think of mortality as a time series (the approach of the Lee-Carter model and its generalizations in the Cairns-Blake-Dowd family); (b) think of mortality as a smooth surface (the approach of the 2D P-spline models of Currie, Durban and Eilers and the smooth versions of the Lee-Carter model).

When projecting mortality rates it is common for people to ask what sort of changes in causes of death might be required to achieve a particular scenario.  Often one is asked to posit what causes of death have to be "eliminated", and the results can lead to the conclusion that a particular projection is unlikely and therefore too prudent.

We have written before about how survival models make better use of available data.