The actuarial profession is concerned with expected rates of death (and implicitly, survival) in order to cost financial promises made to (say) policyholders of life insurance companies and members of defined benefit pension plans.
To this end, longevity risk is simply defined as the risk (from a financial perspective) that a group of people, on average, live longer than anticipated.
Organisations exposed to longevity risk are thus required to make assumptions about the expected mortality rates of relevant group(s) of people.
The approach of the Actuarial Profession in this regard has undergone considerable scrutiny and has been the subject of much debate. In very broad terms, current best practice involves decomposing this assumption into two elements
current expected rates of death
an allowance for expected future improvements in mortality
Each of these is typically derived by reference to historical mortality experience data. The first is likely to result from an examination of observations from the recent past. The second typically results from a statistical extrapolation of historical trends in mortality rates.
For precisely this reason, assumptions relating to longevity are then not reactive to real world events. It is not difficult to conceive of real world events that would be widely acknowledged to have a very significant impact on actual realised rates of survival (over and above those implied by current assumptions). Yet under the status quo, the occurrence of such events is not coupled with a corresponding change to longevity assumptions. Indeed, the passing of such events will have little or no immediate impact upon longevity assumptions adopted.
Furthermore, there is typically a long delay between the occurrence of such events and their effects being revealed in mortality data as illustrated by the Mortality Improvement Framework with an explicit example provided here.
A related point is that practitioners do not necessarily have a clear understanding as to which future real world events are already embedded in their "base case" improvement assumptions. The lack of such a framework makes it difficult to justify a structured response as and when such events / "information" / other indicators come to pass.
This might be viewed as surprising especially as one element of the assumptions set explicitly relates to views on future improvements in mortality. Some external observers might expect this aspect to be more dynamic based upon factors such as continually emerging developments and our continually changing environment.
In fact, under current practice it is difficult to allow for any aspect of longevity risk which is not already captured in past data to some degree.
In other words, our future longevity estimates are largely based upon information gathered by looking into the rear view mirror – and this is not without foundation. But can we do better?
This initiative seeks to form the building blocks of a concept which may help to lead to a supplementary approach which
is more forward-looking in nature and
seeks to capture elements that are widely agreed to influence future improvements without yet revealing themselves in past empirical mortality data