The construction of lapse stress scenarios is commercially important for insurers, especially for writers of guaranteed products and unit-linked business. Yet for such an important tool, relying on judgmental 'mass lapse' events comes with its own risks. Fortunately, new stress test techniques hold out the possibility of more robust data driven approaches.

As any Actuary reading this will know, lapse modelling is challenging from a technical point of view, because:

  • A lack of data limits the power of statistical tests; this implies that a wide range of 'wrong' models are not rejected by the models.
  • Causal explanations for lapses cover a wide spectrum of theories so expert judgement does little to reduce the range of possible models.
  • The requirement to extrapolate to extreme percentiles implies that the choice of model has a large impact on the output.

Where model risk is such an important consideration, it cannot be right to fit a single 'best estimate' model to the data and use that to construct percentiles. Other plausible models, including some producing a higher frequency of extreme events, must also be taken into account. Otherwise there is a concern that firms give undue focus to mitigating reported risk according to one arbitrary model while potentially exacerbating risk under other models. This is why firms often augment their fitted lapse risk distribution with subjectively assessed 'mass lapse' events.

New stress test techniques, based on the statistical concept of prediction intervals, provide a rigorous way to address capital requirements when model and parameter uncertainty are an important part of the risk. This holds out the possibility of more robust data-driven approaches in place of ad-hoc judgemental mass lapse stresses.

I'll be exploring the construction of lapse stress scenarios in more depth at the Life Conference and Exhibition 2013, 10-12 November in Edinburgh.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.