Mortality improvements outlook 2019+

1. Mortality improvements outlook 2019+ Discussion hosted by SIAS and the CMI Mortality Projections Committee 2 April 2019 Staple Inn Hall, London The views expressed in this presentation are those of the presenters and not necessarily those of their employers, the CMI or the Staple Inn Actuarial Society.

2. IntroductionTim GordonChair, CMI Mortality Projections Committee SIAS/CMI meeting – Mortality improvements outlook 2019+

3. CMI CMI • Wholly owned by Institute and Faculty of Actuaries • Independent executive and management Funded by subscription but free for academics and non-commercial research Mission To produce high-quality impartial analysis, standard tables and models of mortality and morbidity for long-term insurance products and pension scheme liabilities on behalf of subscribers and, in doing so, to further actuarial understanding. Our vision is to be regarded across the world as setting the benchmark for the quality, depth and breadth of analysis of industry-wide insurance company and pension scheme experience studies SIAS/CMI meeting – Mortality improvements outlook 2019+

4. Agenda SIAS/CMI meeting – Mortality improvements outlook 2019+

5. England & Wales quarterly and annual centred average SMRs Source: CMI Working Paper 119. NB Straight lines are indicative only. SIAS/CMI meeting – Mortality improvements outlook 2019+

6. CMI_2018 • Using the Core Model for DB pension schemes requires justification: • The CMI Model is calibrated to lives-weighted national data • But DB pension schemes are (a) a SE-biased subset and (b) liability-weighting increases the SE bias • Evidence suggests higher improvements for higher SE groups for past 15 years – using the Core Model implies this ceases immediately • New addition to initial improvements parameter A • A is more intuitive – prefer to Sκ (which may migrate to Advanced) • You can change both – but why would you? • Please download version 2 (updated today) – version 1 uses (and documents in the output) the Advanced parameter constant addition to mortality improvements by age even if you revert it to Core SIAS/CMI meeting – Mortality improvements outlook 2019+

7. England & Wales cumulative standardised mortality rate vs 2009-2018 average Source: Provisional CMI calculations based on ONS E&W data to week 11. NB CMI quarterly monitor will be based on data to week 13. SIAS/CMI meeting – Mortality improvements outlook 2019+

8. CMI_2018 and initial improvementsCobusDaneelCMI Mortality Projections Committee SIAS/CMI meeting – Mortality improvements outlook 2019+

9. CMI Model versions • CMI_2009 to CMI_2015 • Annual updates to data; and some tweaks to method • CMI_2016 • Larger changes in method including: • New “period smoothing parameter” to control responsiveness • Change in shape of long-term rate • CMI_2017 • “Business as usual” annual update • CMI_2018 (published 7 March 2019) • Period smoothing parameter reduced from 7.5 to 7 • New “initial addition to mortality improvements” parameter • Adjustment to population data at ages 85+ SIAS/CMI meeting – Mortality improvements outlook 2019+

10. Progression of cohort life expectancy Age 65, CMI Model, S3PxA, illustrative long-term rate of 1.5% p.a. SIAS/CMI meeting – Mortality improvements outlook 2019+

11. Initial mortality improvements • Model is calibrated to data for England & Wales • Evidence of higher recent improvements for certain populations • Higher socio-economic groups • CMI SAPS dataset • We encourage users of the Model to adjust the Core parameters to suit the population they use it for. Average mortality improvements 2008-2015 with 95% observational confidence intervals SIAS/CMI meeting – Mortality improvements outlook 2019+

12. How to adjust initial improvements? • Period smoothing parameter SK • Intended to reflect how quickly we recognise new data • Some users have used it to modify initial improvements • Impact on improvements of a given change in SK varies over time • Recommend use of the “initial addition to mortality improvements” (A) • Introduced in CMI_2018 • Specifies additional initial age-period improvements at ages 20-85 • Tapers to nil at age 110 (like the Core shape of the long-term rate) • Affects historical improvements also; relevant since the base table date • More complex changes possible using Advanced parameters • Software to be updated to allow “A” to be used with Advanced parameters SIAS/CMI meeting – Mortality improvements outlook 2019+

13. How much to adjust initial improvements? • Historical analyses of specific datasets can be a guide to past differences in mortality improvements. • For projections, also need to consider other factors, including: • Credibility of analyses – statistical uncertainty, and any artefacts of the data • Which factors have caused past differences, and whether they are likely to persist • The extent to which mortality rates for different groups may converge or diverge over the convergence period SIAS/CMI meeting – Mortality improvements outlook 2019+

14. Super-healthy:How good can population longevity get?Dr Tim CrayfordGroup Medical DirectorJust plcWith thanks to Matthew Edwards of Willis Towers Watson SIAS/CMI meeting – Mortality improvements outlook 2019+

15. Main concepts Objectives • Can we define a super-healthy group from the general population? • What is their mortality, and how different is it from population mortality? • Does super-healthy mortality tell us anything about when longevity improvements might stop? For debate • How credible is it that any whole population’s mortality might fall to the level of people who are super-healthy today? SIAS/CMI meeting – Mortality improvements outlook 2019+

16. Industry context Medically underwritten annuities • Sold since the 1990s • Developed into a sophisticated market place • Most companies & reinsurers now have a basis to allow them to rate the lives of people who have medical risk factors • That basis is underpinned by a floor, which is the healthiest conceivable mortality: these are generally annuitants who have never had anything wrong with them at the time of underwriting. • No previous illnesses • No medications • Not overweight or underweight • No significant risk factors – smoking, obesity etc • Good postcode • It is hard to conceive of a healthier sub-group of the population. SIAS/CMI meeting – Mortality improvements outlook 2019+

17. Improvements Base mortality • What can a firm underwriting the healthiest lives safely assume? Improvements • Almost all (re)insurers and pension funds assume a constant long-term future improvement rate in projecting longevity improvements (typically 1.25-1.75%) • What does this imply for our understanding of average life expectancy? • Is it reasonable for an average person now to become ‘as if’ super healthy today at some future point? SIAS/CMI meeting – Mortality improvements outlook 2019+

18. Objectives of the improvement analysis • Construct a life table of annual mortality by age for males with no significant current recorded conditions, no significant medical history, and no significant medication history from ages 40+ • Explore the rate at which population mortality would have to improve before all individuals in the general population experienced the mortality currently experienced by this group • Compare this rate to the assumptions typically used by insurers (etc) SIAS/CMI meeting – Mortality improvements outlook 2019+

19. Clinical Practice Research Database • Comprises medical treatment information on 21 million anonymised patient lives followed for up to 25 years • 1.8bn consultations • 1:12 people in UK • 600 currently contributing GP practices • Medical history mostly captured, risk-factors less so • Data quality varies by time (some lack of consistency pre-2010) SIAS/CMI meeting – Mortality improvements outlook 2019+

20. Selection criteria for our super-healthy group • Absence of • Heart disease, stroke • Diabetes • Cancer • Respiratory disease • Neurological disease • Other major conditions (eg IBD, kidney disease) • NB we also mean absence of any history of these conditions • Presence of • Good HbA1c • Good BMI • Good socio-economic bracket (via ONS IMD) • Good smoking habit (ie never-smoker!) Data volumes: 676,000 man-years exposure from disease criteria, pre-risk factor stratification. Risk factor effects computed from larger data set using GLMs rather than cutting data to create individual ‘cells’. SIAS/CMI meeting – Mortality improvements outlook 2019+

21. Healthy mortality We can compare the mortality of the healthy group against population, and also against typical enhanced annuity profile. The relative ‘health discount’ varies from -75% initially to circa -20% from age 75. Also as a lower bound we have a ‘permanently healthy’ life – defined as not having any major conditions at the start of each year SIAS/CMI meeting – Mortality improvements outlook 2019+

22. Jim Vaupel: Broken limits to life expectancy 2002 Oppen & Vaupel Science, 2002 SIAS/CMI meeting – Mortality improvements outlook 2019+

23. In Search of Methuselah: Estimating the Upper Limits to Human Longevity Olshansky et al Science 1990 SIAS/CMI meeting – Mortality improvements outlook 2019+

24. SIAS/CMI meeting – Mortality improvements outlook 2019+

25. ELT17, the super-healthy table and convergence: How Low Can You Go? SIAS/CMI meeting – Mortality improvements outlook 2019+

26. Aim of this section • Take the CPRD-derived life-table of annual mortality by age for males with no significant current recorded conditions, no significant medical history, and good risk factors (non-smokers etc. as per previous slides) from ages 40+ • Explore the rate at which population mortality (ELT17) would have to improve before all individuals in the general population experienced the mortality currently experienced by this group • Compare this rate with typical insurer improvement assumptions SIAS/CMI meeting – Mortality improvements outlook 2019+

27. SIAS/CMI meeting – Mortality improvements outlook 2019+

28. CMI Convergence to synthetic healthiest @ 1.5% SIAS/CMI meeting – Mortality improvements outlook 2019+

29. Cohort ALE from 65 for always healthy from 2067 under various long-term improvement assumptions SIAS/CMI meeting – Mortality improvements outlook 2019+

30. Cohort ALE from 65 for always healthy from 2067 under various long-term improvement assumptions 2017 healthiest SIAS/CMI meeting – Mortality improvements outlook 2019+

31. Cohort ALE from 65 for always healthy from 2117 under various long-term improvement assumptions 2017 healthiest SIAS/CMI meeting – Mortality improvements outlook 2019+

32. Caveat: generalisability of population improvements to healthy people? • CMI suggestive evidence of increased improvements for normal health vs ill-health retirees • L&G / Barnet Waddingham / Hymans analysis showing higher improvement rates amongst people in higher SEGs • Suggests that the healthiest people have experienced the highest improvement rates – some assume up to 5% in recent years • Can this be sustained? Will we see a widening of the longevity inequality? SIAS/CMI meeting – Mortality improvements outlook 2019+

33. Lu et al – improvements by SEG SIAS/CMI meeting – Mortality improvements outlook 2019+

34. So what is the healthiest ALE in CPRD? • Highly synthetic group • 92.86 at 40 years • 93.81 at 65 years • These people have no observable illness at the start of the year, they experience a certain amount of incident disease, and some of them die within this calendar year. This implies the elimination of chronic diseases and the removal of existing medical history from all people at 65 years • Their rates of death are typically around 20% of the ELT17 table, increasing from the mid-eighties. SIAS/CMI meeting – Mortality improvements outlook 2019+

35. Average annual improvement required for population ALE to reach 93.8 years Long-term rate as low as 1.25% is sufficient to reduce mortality of everyone to current healthiest within 100 years Assumes improvements tapering from 85-110 years SIAS/CMI meeting – Mortality improvements outlook 2019+

36. By how much would mortality need to be reduced to take average longevity to… SIAS/CMI meeting – Mortality improvements outlook 2019+

37. What could change for the very healthiest? • Future incidence of disease? • Obesity, Diabetes? • Early uptake of longevity-promoting drugs by the highest social classes? • Statins, LDL Vaccine? • Metformin? • New pharmaceuticals? SIAS/CMI meeting – Mortality improvements outlook 2019+

38. Conclusions • Maximum ALE from 65 years in identifiable current UK populations is 93.8 years (male) • The population as a whole could reach this level within 100 years at typical industry-standard assumptions • There has been a significant down-turn in improvements in the UK. It is uncertain both as to the cause and as to whether this will be sustained • Is it reasonable for projections not to assert a research-based assumption of maximum longevity? • In other words: should we move from the idea of a constant LTR to a gradually decreasing LTR – especially for deferred pension liabilities?(NB different from tapering > age 85) SIAS/CMI meeting – Mortality improvements outlook 2019+

39. Through the lens of a mortality reduction framework: Recent slow-down in mortality improvement versus long-term viewJoseph LuDirector, Longevity ScienceLegal & General SIAS/CMI meeting – Mortality improvements outlook 2019+

40. Reduction in mortality rates – a framework Historical data back up to 2013 Suppose a projection of reduction in death rate, considering: We are mortals – there’s a limit to reduction Historical differentiation – gender, age groups, causes of death. Historical trends associated with: improvement in risk factors – to consider prevention of diseases in the future. treatments – to consider medical breakthroughs in the future. To consider barrier to technology, access of health & social care, uptake by patients. SIAS/CMI meeting – Mortality improvements outlook 2019+

41. Maximum reduction in mortality SIAS/CMI meeting – Mortality improvements outlook 2019+

42. A projection from 2013 onwards 1. Maximum reduction 2. How to get to maximum SIAS/CMI meeting – Mortality improvements outlook 2019+

43. Inconveniently, reality did not follow projection • Now, to consider what have happened. • Will the events affect long term? • Framework of model • Maximum reduction • Projection of reaching maximum reduction SIAS/CMI meeting – Mortality improvements outlook 2019+

44. What happened? High winter deaths.Will this continue or revert to ‘normal’? 2014/15 2015/16 2016/17 2017/18 2018/19 (UK under 4 z-score) EuroMOMO SIAS/CMI meeting – Mortality improvements outlook 2019+

45. Potential new solutions to improve living conditions • Examples: • Financial facilities to liquidate housing wealth to improve living conditions • Specialist housing for retirement • New-built suitable for retirement and other ages • Renovation of existing homes. SIAS/CMI meeting – Mortality improvements outlook 2019+

46. Rich-poor gap in death rates suggests there is potential for improvement through existing knowledge Longevity Science Panel (2018) Most Affluent = Least Deprived IMD etc. SIAS/CMI meeting – Mortality improvements outlook 2019+

47. Widening mortality gap SIAS/CMI meeting – Mortality improvements outlook 2019+

48. Death rates fell more for the more advantaged SIAS/CMI meeting – Mortality improvements outlook 2019+

49. Medical development Examples from Longevity Forum – Buck Institute and AgeX SIAS/CMI meeting – Mortality improvements outlook 2019+

50. Potential in medical advancement.E.g. Increase in venture capital investment http://www.evaluate.com/vantage/articles/data-insights/venture-financing/rise-and-rise-corporate-venture-arms SIAS/CMI meeting – Mortality improvements outlook 2019+