Plotter for Hermite selection effect

Duration Selection effect added to log(mortality) © www.longevitas.co.uk

Many insurance risks exhibit selection effects at the start of a new policy. The simplest selection model is just a pair of parameters:

  1. SelectionInitial, the initial selection effect, and
  2. SelectionTerm, the period in years after which selection is presumed to have worn off and is zero at all longer durations.

For the Hermite splines we calculate the normalised duration, u, from the duration since outset, r, as follows:

u = r / SelectionTerm

and we can model selection with just h00(u), since selection is zero when r > SelectionTerm.

The two-parameter approach to selection is robust, but for datasets with strong and unusual selection patterns an optional third parameter describes the shape and direction of the selection effect immediately after the business is written. This optional third parameter uses the h10 spline. The h01 and h11 splines are unused for modelling selection, i.e. they have implict coefficients of zero.

The full, three-parameter adjustment to log(mortality) in respect of selection is therefore:

h00(uSelectionInitial + h10(uSelectionInitialGradient

when r < SelectionTerm. The selection adjustment is zero when rSelectionTerm.

Mandatory parameters:
SelectionInitial Initial effect at duration zero.
SelectionTerm Selection duration in years.

For portfolios with a long SelectionTerm or where, SelectionInitial is fairly strong, there is an optional parameter for the initial gradient (direction) of the selection effect:

Optional parameters:
SelectionGradient
The selection effect shown above is added to the basic age pattern for log(mortality). One can also add cyclic seasonal variaton and an age-related time trend. For full details of the model and its implementation, see Richards (2019).