Milliman has just published a new report summarizing the results of a recent survey of current practice in the modelling of dynamic policyholder behaviour (DPB) and management actions (MA) for life insurance business.
There are 56 companies represented in the survey, across Europe, the United States, and Japan.
The survey revealed some interesting results. For instance:
• For variable annuities/unit-linked with guarantees, only around 50% of respondents model at least one type of DPB. This increases to 85% for other types of products (what we have termed “traditional” products).
• Of respondents offering guaranteed annuity options (GAOs) on traditional products, only around 16% of respondents were modelling them with dynamic take-up rates.
• Around 60% of companies have monitored DPB experience against that predicted by their models; of these, almost half say their models predicted experience well.
• Most companies in the United States and Europe model assets and liabilities, the interactions between them, and some type of future investment strategies (a form of MA) for certain classes of business. However, future investment strategies modelled are often oversimplistic, for instance being invariant to economic conditions.
• Only a minority of companies hold a formal documented plan for management actions. Most companies also don’t monitor actual management actions against those predicted by their models.
• Actuaries were the most prominent group involved in setting modelled MA rules, followed by investment and risk management professionals. Other professionals were more rarely involved.
DPB and MA are becoming increasingly important aspects of modelling as more focus is placed on stochastic calculations and the tails of distributions. In particular, Solvency II in Europe specifies requirements for both DPB and MA, so we expect significant work will be required of companies in these areas, particularly as they should form a key component of a company’s risk management.
Market turmoil in recent years across various regions highlights the importance of working to understand and model how management may react to such scenarios. However, the survey results show that many companies are failing to model DPB for some key options. Modelling of MA is also underdeveloped in many cases, with some key actions not being modelled at all, or in an oversimplistic way that doesn’t appropriately reflect reality.
DPB and MA predicted by models should also be monitored against actual experience as it emerges, with models being refined over time.