Repricing individual annuities after a mortality improvement release
Halcombe Assurance writes level and RPI-linked immediate annuities in the UK. In-force business is £2.6bn across 31,000 policies; new business runs at about £400m a year, of which 55% is medically underwritten through a third-party triage service. Pricing uses the S3PMA tables with 1.25% a year mortality improvements tapering to 0.75% in the long term. The latest CMI release shows improvements over 2011–2020 were materially lower than the previous model implied, while the 1960s cohort effect remains strong and recent weekly data are volatile. The firm's solvency coverage ratio is 155% of SCR, with SCR of £1.4bn. A competitor has cut its rates by about 8bps and the sales team wants a matching cut. The enhanced book is 30% reinsured under a quota share with recapture terms, and the reinsurer has indicated it will reprice at the next renewal. The board's risk appetite limits any single pricing decision to a 15 percentage point fall in cover. The actuarial function asks whether the improvement basis, cash-flow matching and underwriting loadings should be reviewed together or separately.
1. Recommend how the mortality improvement basis should be updated, and explain what evidence you would require before signing off the change.
Read the worked answer
Frame the question properly. The issue is not what the latest release says but how much weight one release deserves, and what the cost of being wrong in each direction is. On a £2.6bn book, 0.1% a year of improvement is worth tens of basis points of price - far more than the 8bps competitive question.
Evidence I would require before signing off
- The model itself. Central, 0.5% and 1.5% variants; sensitivity of the smoothing parameter; whether the lower improvements are broad-based across ages or driven by a handful of high-age years.
- Distortions. COVID-period deaths and the resulting effects on the denominator and on 2020–2023 data, the 2022–23 winter, and whether weekly volatility is signal or noise.
- Appropriateness of the reference data. Population tables used as the reference may not represent an annuitant book that skews socio-economically advantaged; check own-experience A/E by age, sex, pension size and postcode, and assess the credibility of the firm's own data.
- Cohort effects. The 1960s cohort remains strong. A model that lowers the long-term rate on the back of 2011–2020 may be removing a cohort effect through the back door.
- Consistency with the reinsurer's basis and with the recapture terms, since the treaty will be repriced.
Recommended approach
- Update the base table to the latest available.
- Do not move the long-term rate to the model central on one release; hold 0.75% until a trend is confirmed.
- Reduce the initial rate modestly and judgementally - for example from 1.25% to around 1.0% - and document the reasons rather than mechanically adopting the central projection.
- Test the impact on solvency, on the value of in-force and on new business margins, and check the book can absorb a repeat of the last decade.
- Commit to a mid-year review trigger, and write a formal opinion setting out the weight given to the population model versus own experience - that is the document a review will challenge.
The improvement basis should be reviewed separately from and before the pricing response to the competitor, because the pricing decision depends on the answer, not the other way round.
2. Advise the board on whether to match the competitor's 8bps reduction, quantifying the cost and identifying what really drives the decision.
Read the worked answer
Quantify the direct cost
- 8bps on £400m of new business is about £320k of annual premium income foregone.
- Capitalised over an annuity duration of roughly 15–20 years, at a factor of about 15, that is around £4.8m of economic value.
- Against SCR of £1.4bn and coverage of 155%, the coverage impact is a fraction of one percentage point - comfortably inside the 15 point risk appetite limit.
So the affordability test is passed easily. That is the trap: the decision should not be driven by the risk appetite limit at all.
What actually drives the decision
- Relative to the basis risk, 8bps is noise. A 0.1% error in the improvement assumption is worth far more than the price cut. Spending management attention on the competitor while the longevity basis is uncertain is a misallocation.
- Why is the competitor cutting? A different asset strategy, a reinsurance quote, a desire for volume, or a mispriced loss-leader. Matching transfers their mistake onto Halcombe's balance sheet.
- Volume versus margin. Losing volume spreads fixed costs over less business, but writing unprofitable volume is worse. There is no virtue in a 100% market share of a loss-making product.
- Adverse selection cuts both ways. Raising rates sharply pushes the best risks to competitors; cutting rates attracts the business no one else wants.
Recommendation
- Do not make a blanket 8bps cut.
- Reprice selected segments where the revised improvement basis genuinely supports a lower premium - typically the older ages and the level (non-RPI) product where the longevity sensitivity is different.
- Compete on underwriting and service in the enhanced segment, where the triage capability and the 55% medically underwritten mix are a real advantage.
- Monitor competitor pricing, sales volumes and lapses monthly against a pre-agreed review trigger.
- Ensure the repricing and the reinsurance renewal are negotiated together - the reinsurer's new rate will determine whether any segment can be priced down at all.
The answer to give the board is: we are not following an 8bps cut we cannot justify, we are repricing where our own evidence supports it, and we will revisit if the sales data show we are losing the business we want.