I'm giving away rather too much information about my age when I say I started work in 1990 right after graduating from university.   Not long into my first job at a UK insurer, I was called to a meeting of the actuarial department.  At this meeting was a man with the job title of Research Actuary.  During the meeting he suggested that our employer's investment strategy should be set such as to target a "fixed probability of ruin".   The idea was to accept that insolvency was always a possibility, but to recognise this and pick the assets — and the business strategy — to match the liabilities to make sure this possibility was suitably unlikely.  As I was fresh out of university, I couldn't imagine running a risk business any other way.

However, I vividly remember the reaction to this suggestion  a wordless stare from the senior actuary and much looking at feet from the junior ones.  Although the suggestion sounded almost obvious to my naïve ears, it was clear from the body language around the room that the proposal was regarded as impractical and other-worldly by everyone else.   However, modern readers will recognise the suggestion as being the very essence of the current ICA regime in the UK.  It is also the foundation of the forthcoming Solvency II regime in the EU — managing an insurer's affairs such that there is only a 0.5% probability of insolvency over a one-year period.   The Research Actuary's proposal even anticipated the use test, namely where an office uses this approach to manage its affairs on a day-to-day basis and not treat liability valuation as a once-a-year exercise.   In fact, the only problem with the proposal was that it was around 15 years ahead of its time.

Fast forward to 2012 and I still see occasional parallels with 1990.  Nowadays there is little disagreement that stochastic asset models are useful  swings in asset values can be large and unpredictable, so it makes sense to acknowledge this unpredictability.   However, when it comes to projecting future mortality, there is still in some quarters a reluctance to recognise that annuitant mortality rates in twenty years' time are almost as unknowable as the level of the FTSE-100 Index.  The need to use stochastic projection models for mortality projections is little different from the need to use stochastic models for asset returns: both are important basis elements with substantial amounts of uncertainty over the future. The sensible approach is the same in both cases: acknowledge the uncertainty, pick one or more stochastic models and calibrate them to some data.  This applies as much to mortality projections as it does to models of asset returns and yield curves.

I suppose the Research Actuary could console himself in one respect: at least he wasn't as badly treated as others with ideas ahead of their time, like Galileo. A hard and disapproving stare from the senior actuary wasn't quite as scary as a visit from the Inquisition.  It wasn't too far off, though!

### RECENT POSTS

The upcoming EU General Data Protection Regulation places focus on ... Read more
Assume we have a random variable, $$X$$, with expected value ... Read more
Our new book, Modelling Mortality with Actuarial Applications , describes ... Read more
Stephen Richards is the Managing Director of Longevitas
##### Models in the Projections Toolkit
All models in the Projections Toolkit are statistical, i.e. they are extrapolative models where central projections come with a statement of uncertainty.