## Putting the TAS into Longevitas

In the UK the Board for Actuarial Standards (BAS) has published a series of Technical Actuarial Standards (TAS).  At the time of writing there are three standards which are either in force or shortly will be: one for data (TAS-D), one for reports (TAS-R) and one for actuarial models (TAS-M).

TAS-D concerns the quality of data used in actuarial reports and analysis.  The emphasis is on documenting the data definitions, together with the validation checks which have been performed and the purpose to which the data will be used.  If data are incomplete or inaccurate, then these limitations are to be clearly documented to enable the reader to know how much reliance can be placed on them.  TAS-D applies to all data used in actuarial reports produced since 1st July 2010.

TAS-R concerns the content of actuarial reports, with particular emphasis on transparency.  The aim is that all assumptions should be disclosed, that any uncertainties or limitations of the data should be clearly stated, and that the report is comprehensible.  TAS-R is interesting for its specific content on an area which has greatly raised its profile in actuarial work recently: uncertainty.  Actuaries are now required to state what uncertainty lies over their assumptions, and what the impact of this uncertainty could be.  Although TAS-R does not mandate the use of stochastic models, judicious use of such models makes it easier for an actuary to ensure that uncertainty is thoroughly explored.  TAS-R applies to all actuarial reports produced since 1st April 2010.

TAS-M concerns the models built by actuaries.  The aim is to ensure that such models are relevant and fit for the purpose to which they are put.  Besides the usual requirements for careful documentation of assumptions and limitations, TAS-M also makes the distinction between "neutral" and "non-neutral" estimates.  As with TAS-R, TAS-M does not mandate the use of statistical or stochastic models.  However, demonstrating that an estimate is "neutral" in terms of TAS-M is relatively straightforward with a stochastic or statistical model.  TAS-M will apply to all actuarial models on or after 1st April 2011.

The requirements for the various TASs are sensible: nobody can argue against careful documentation of model assumptions or data limitations, for example.  The only drawback of the TASs is the extra costs they will bring: ensuring compliance for each actuarial report will take some amount of time and will therefore add to costs.  However, this might be an opportunity to take stock of how things are done and consider whether computerisation could ease the burden.  Rather than just adding extra manual work to an existing process, it might be possible to use software to improve quality at the same time.

Assume we have a random variable, $$X$$, with expected value ... Read more