Mortality Working Group

2 Mortality Working Group Chair: Paul Lewis (South Africa) Co-vice-chairs: Allen M. Klein (US) Brian Ridsdale (UK) 30 members + 31 interested persons Devoted to the world wide study of mortality Mortality Information Base lists areas of investigation by the group

3 Mortality Working Group Vision: The Mortality Working Group will be pre-eminent international actuarial body to provide insights and knowledge with respect to mortality and trends in mortality.

What s in this presentation? 4 1. The use of mortality tables 2. Measuring longevity 3. Projection modelling mortality 4. Understanding history mortality 5. International view 6. Asian Countries 7. Example of an annuity

The Use of Mortality Tables The tool an insurance company uses to analyse the probability of a claim in case of life insurance are the probability of dying (qx) or surviving (px=1-qx). The qx depends on gender, age, but also way of living, health etc. Therefore the mortality rates used should be based on observations in the insured population Observations over the whole country should be adjusted for own observations (level of mortality) In particularly the mortality in an insured population, or in the pension world are lower than the average whole population mortality. The difference depends for pensions on White or Blue collar It is observed that people with higher incomes have a lower mortality than people with lower incomes. So people with higher incomes live longer.

The Use of Mortality Tables The mortality tables can directly be derived from own observations, but in most cases not enough observations are available to get statistical reliable data. In most countries the tables to use are based on whole population mortality adjusted (e.g. with a factor) to get mortality for the insured population.

The Use of Mortality Tables Therefore it is very important for pension funds and insurance companies that reliable mortality observations over the whole population in a country are available, preferable every calendar year. In the internet site human mortality database detailed and yearly observations for many countries are published, sometimes even since 1800. www.mortality.org My advise for pension funds and insurance companies in countries where not yearly mortality data over the country are available to ask statistical offices to set up such statistics. Particularly Asian countries are less available in the HMD (only Japan and Taiwan are available)

Age\Calendar y 2015One Dimensional 0 0.0028226 1 0.0002678 2 0.0001103 3 0.0001103 4 0.0001213 5 0.0000842 6 0.0000707 7 0.0000458 8 0.0000362 9 0.0000304 10 0.0000368 11 0.0000431 12 0.0000545 13 0.0000647 14 0.0000983 15 0.0001357 16 0.0001402 17 0.0001870 18 0.0002430 19 0.0003094 20 0.0003414 21 0.0003512 22 0.0003319 23 0.0003132 24 0.0003001 25 0.0002790 26 0.0003160 27 0.0003750 28 0.0004035 29 0.0004755 30 0.0004929 31 0.0004875 32 0.0004811 33 0.0005171 34 0.0005190 35 0.0005728 36 0.0006134 37 0.0007096 38 0.0007813 39 0.0008383 40 0.0009008 Table Depending on one calendar year, age and gender

Measuring Longevity To measure longevity just a one dimensional table is not enough. We need also information about the future A projection for the future need to be made

Two Dimensional Generation Table Depending on gender, age and calendar year. Moves via diagonal. Age\Calendar y 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 0 0.0028226 0.0027110 0.0026051 0.0025046 0.0024091 0.0023185 0.0022323 0.0021504 0.0020725 0.0019984 0.0019279 1 0.0002678 0.0002528 0.0002388 0.0002257 0.0002135 0.0002020 0.0001912 0.0001812 0.0001718 0.0001629 0.0001546 2 0.0001103 0.0001026 0.0000955 0.0000890 0.0000830 0.0000775 0.0000724 0.0000678 0.0000635 0.0000596 0.0000559 3 0.0001103 0.0001042 0.0000986 0.0000934 0.0000885 0.0000839 0.0000797 0.0000758 0.0000721 0.0000687 0.0000655 4 0.0001213 0.0001173 0.0001134 0.0001096 0.0001060 0.0001025 0.0000990 0.0000957 0.0000925 0.0000894 0.0000864 5 0.0000842 0.0000803 0.0000767 0.0000732 0.0000700 0.0000669 0.0000640 0.0000613 0.0000587 0.0000563 0.0000539 6 0.0000707 0.0000675 0.0000644 0.0000615 0.0000587 0.0000561 0.0000536 0.0000512 0.0000490 0.0000469 0.0000449 7 0.0000458 0.0000428 0.0000400 0.0000375 0.0000352 0.0000331 0.0000311 0.0000293 0.0000277 0.0000262 0.0000248 8 0.0000362 0.0000334 0.0000309 0.0000286 0.0000265 0.0000247 0.0000230 0.0000214 0.0000201 0.0000188 0.0000177 9 0.0000304 0.0000276 0.0000251 0.0000229 0.0000210 0.0000193 0.0000177 0.0000164 0.0000152 0.0000141 0.0000131 10 0.0000368 0.0000337 0.0000310 0.0000286 0.0000264 0.0000244 0.0000226 0.0000210 0.0000196 0.0000183 0.0000172 11 0.0000431 0.0000398 0.0000368 0.0000341 0.0000317 0.0000295 0.0000276 0.0000258 0.0000242 0.0000227 0.0000214 12 0.0000545 0.0000506 0.0000471 0.0000440 0.0000411 0.0000385 0.0000362 0.0000340 0.0000321 0.0000303 0.0000287 13 0.0000647 0.0000603 0.0000562 0.0000526 0.0000493 0.0000463 0.0000436 0.0000411 0.0000388 0.0000368 0.0000349 14 0.0000983 0.0000932 0.0000886 0.0000842 0.0000802 0.0000764 0.0000729 0.0000696 0.0000666 0.0000637 0.0000611 15 0.0001357 0.0001297 0.0001240 0.0001187 0.0001137 0.0001090 0.0001046 0.0001004 0.0000965 0.0000928 0.0000893 16 0.0001402 0.0001322 0.0001247 0.0001179 0.0001117 0.0001060 0.0001007 0.0000959 0.0000914 0.0000873 0.0000836 17 0.0001870 0.0001767 0.0001671 0.0001583 0.0001503 0.0001429 0.0001360 0.0001297 0.0001239 0.0001186 0.0001136 18 0.0002430 0.0002303 0.0002187 0.0002079 0.0001980 0.0001888 0.0001803 0.0001724 0.0001652 0.0001584 0.0001522 19 0.0003094 0.0002963 0.0002840 0.0002725 0.0002617 0.0002516 0.0002421 0.0002332 0.0002249 0.0002170 0.0002097 20 0.0003414 0.0003278 0.0003150 0.0003029 0.0002916 0.0002810 0.0002710 0.0002615 0.0002527 0.0002443 0.0002364 21 0.0003512 0.0003374 0.0003244 0.0003122 0.0003007 0.0002898 0.0002796 0.0002700 0.0002610 0.0002525 0.0002444 22 0.0003319 0.0003186 0.0003060 0.0002942 0.0002831 0.0002727 0.0002629 0.0002536 0.0002449 0.0002368 0.0002290 23 0.0003132 0.0003000 0.0002878 0.0002762 0.0002654 0.0002553 0.0002458 0.0002369 0.0002285 0.0002207 0.0002133 24 0.0003001 0.0002871 0.0002750 0.0002637 0.0002531 0.0002432 0.0002339 0.0002252 0.0002171 0.0002094 0.0002023 25 0.0002790 0.0002656 0.0002532 0.0002417 0.0002311 0.0002213 0.0002122 0.0002037 0.0001959 0.0001886 0.0001819 26 0.0003160 0.0003030 0.0002908 0.0002795 0.0002689 0.0002589 0.0002497 0.0002410 0.0002329 0.0002253 0.0002182 27 0.0003750 0.0003632 0.0003519 0.0003413 0.0003312 0.0003217 0.0003126 0.0003040 0.0002959 0.0002881 0.0002808 28 0.0004035 0.0003917 0.0003805 0.0003699 0.0003599 0.0003503 0.0003413 0.0003328 0.0003247 0.0003170 0.0003097 29 0.0004755 0.0004647 0.0004544 0.0004444 0.0004349 0.0004257 0.0004170 0.0004085 0.0004004 0.0003927 0.0003852 30 0.0004929 0.0004818 0.0004712 0.0004611 0.0004514 0.0004421 0.0004332 0.0004247 0.0004165 0.0004087 0.0004012 31 0.0004875 0.0004749 0.0004628 0.0004514 0.0004405 0.0004302 0.0004204 0.0004110 0.0004021 0.0003936 0.0003856 32 0.0004811 0.0004668 0.0004533 0.0004405 0.0004284 0.0004169 0.0004061 0.0003958 0.0003861 0.0003770 0.0003683 33 0.0005171 0.0005028 0.0004891 0.0004761 0.0004637 0.0004519 0.0004406 0.0004299 0.0004196 0.0004098 0.0004005 34 0.0005190 0.0005036 0.0004890 0.0004751 0.0004618 0.0004493 0.0004373 0.0004260 0.0004152 0.0004049 0.0003951 35 0.0005728 0.0005567 0.0005414 0.0005268 0.0005129 0.0004995 0.0004868 0.0004747 0.0004631 0.0004520 0.0004414 36 0.0006134 0.0005956 0.0005786 0.0005625 0.0005472 0.0005326 0.0005187 0.0005055 0.0004930 0.0004810 0.0004697 37 0.0007096 0.0006922 0.0006754 0.0006593 0.0006438 0.0006290 0.0006148 0.0006011 0.0005880 0.0005754 0.0005633 38 0.0007813 0.0007634 0.0007461 0.0007295 0.0007135 0.0006982 0.0006834 0.0006693 0.0006556 0.0006425 0.0006299 39 0.0008383 0.0008185 0.0007995 0.0007812 0.0007638 0.0007470 0.0007309 0.0007155 0.0007008 0.0006866 0.0006730 40 0.0009008 0.0008784 0.0008570 0.0008364 0.0008167 0.0007978 0.0007797 0.0007623 0.0007456 0.0007297 0.0007144

Projection Models Mortality To estimate the future mortality development is one of the biggest challenges. Should is be pure statistical, based on historical figures? What data should be used? Should expert judgement play an important role in the model? Particularly expert judgement from the medical world? How far can we look into the future?

Projection Models Mortality This is my opinion: We can not use a pure statistical/mathematical model. Expert Judgement from the medical world should be added. We should understand the historical data.

13 We should understand the history Life expectancy at age zero, Netherlands 90 female smoking reduced the increase of life expectancy after 1990 80 70 (2) 60 (1) Male 50 Spanish Flu in 1918 Hunger winter end of WW2 Female 40 30 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000

Explaining the History 14 (1) The Hump In the early 50 ts for the ages 45-75 for male the mortality rates went up. This was caused by: Smoking of cigarettes Traffic accidents Heart failure All these impacts are the result of behaviour Smoking Eating habits in combination with less healthy exercise habits More driving in cars This flat period of development make the insurers not aware of the potential longevity risk in their portfolio

Explaining the History 15 (1) The Hump (cont.) During the seventies all three causes changed Less smoking for male Traffic get safer (in 1969 yearly more than 3000 traffic deaths in NL, nowadays around 600) Medical developments regarding heart attacks in combination with a healthier way of living (healthier food, more exercise) and the mortality rates went down again

Explaining the History 16 (2) Trend change in 2001 The increase of the life expectancy suddenly went up to more than 0.3 years per year (before that between 0.15 and 0.2), both for male and female Happened in almost the whole Western World Reasons Continuation of less smoking (particularly male) Angioplasty as a treatment in case of an heart attack. This increased the survival chance dramatically

Mortality Development 17 The development of life expectancy depends on: Medical development And is it available? Behaviour Drinking, smoking, eating habits,... Environment Drinking water, one of the most important reasons of the increase of the life expectancy in the developed countries Pollution Water, air Climate And so climate change New diseases Resistance against medicines (antibiotics)

We can split development in 3 parts: 18 90 Life expectancy at age zero, Netherlands 80 70 2 60 50 3 40 1 30 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Male Female

We can split development in 3 parts 19 (1) No development of e(0) High volatility People less protected against extreme weather, flu epidemics Tuberculosis High mortality for young children In 1850: e(0) male: 38.3; e(5) male: 50.8! Comparable with the underdeveloped countries

We can split development in 3 parts 20 (2) After the industrial revolution Steep increase of life expectancy Medical developments Cleaner drinking water Seen as THE most important reason for improvement Environment Better protection: heating in houses, toilets etc. Comparable with emerging countries

We can split development in 3 parts 21 (3) Typical for developed countries Developments like the quality of drinking water are reaching the limits Change in life expectancy depends more of: Behaviour Medical developments Both can have positive or negative effects. Particularly behaviour can cause more independency in development between male and female.

How old can a human become? 22 The oldest confirmed human became 122 years and 164 days Jeanne Calment Born: 25 February 1875 Secret? On all her food olive oil Port wine diet 1 kg of chocolate a week

How old can a human become? 23 It cannot be proven that the max age is increasing Medical experts mention that a real life span exist per person, depending on the genetic passport, but will be limited to around 125-130 years. Mortality rates seem to be almost constant above age 105 at a level 0.5-0.6.

International View 24 Following the ideas countries that are in situation 3 should have comparable trend developments (same level of development) Also following several studies the uncertainty should be comparable over the countries (I would like to add: under the same circumstances ) and can be used in case a lack of data exist in a country Li Lee CBS

Countries in (3) 25

Countries in (3) 26

What can we see in those graphs? It is less accurate to model mortality completely independent from other countries It looks like countries coming from (2) are converging in level and development of mortality once they reach (3) This should be taken into account in the models.

Asian Countries 90 Development life expectancy (age 0) in some Asian countries 85 80 75 70 Korea Male Korea Female Japan Male Japan Female Taiwan Male Taiwan Female 65 60 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Don t look only to your own country A conclusion of this graph is that when modelling Korea independent from for example Japan, the life expectancy of Korea will grow fast above Japan More reasonable is that it grows to the same level (goal table approach) I did something like that for Korea, let the development grow to Japanese projections 29

Projection + Uncertainty for Japan 100 95 results projection Japan life expectancy age 0 90 85 80 MALE BE MALE 95% FEMALE BE FEMALE 95% 75 70 2012201420162018202020222024202620282030203220342036203820402042204420462048205020522054205620582060

Comparison Japan with other developed countries (male) e0 e0 e0 male 2012 BE 2060 2060 incl. 95% unc. 95% uncertainty AUSTRIA 78.24 88.13 92.66 4.53 BELGIUM 77.78 86.28 90.80 4.52 DENMARK 77.82 87.17 92.34 5.17 FRANCE 78.62 88.46 92.50 4.04 ITALY 79.96 88.66 93.78 5.13 NETHERLANDS 79.28 87.85 92.82 4.97 NORWAY 79.36 87.77 92.77 5.00 UK 78.99 89.39 93.19 3.80 USA 76.84 85.91 89.95 4.04 SWISS 80.57 89.27 92.80 3.53 SWEDEN 79.97 87.24 90.53 3.29 FINLAND 77.38 87.49 92.78 5.29 CANADA 79.65 88.85 92.33 3.47 JAPAN 79.82 87.43 91.89 4.46 Based on new stochastic model

Comparison Japan with other developed countries (female) e0 e0 e0 female 2012 BE 2060 2060 incl. 95% unc. 95% uncertainty AUSTRIA 83.53 90.89 94.51 3.62 BELGIUM 82.92 88.78 92.53 3.75 DENMARK 81.91 89.31 94.43 5.12 FRANCE 85.03 91.75 94.92 3.18 ITALY 84.84 91.58 95.80 4.22 NETHERLANDS 82.94 88.37 92.74 4.37 NORWAY 83.60 89.83 93.99 4.16 UK 82.83 90.26 93.83 3.56 USA 81.44 86.48 90.07 3.59 SWISS 84.79 90.69 93.59 2.90 SWEDEN 83.77 88.51 91.32 2.81 FINLAND 83.81 91.49 95.80 4.31 CANADA 83.74 89.49 92.56 3.07 JAPAN 86.29 93.89 97.82 3.93 Based on new stochastic model

33 Example of an Annuity Male Discount age rate 2% 65. Annuity age 65 male Japan % 2012Without trend 15.292 2012-2060 Including BE Trend 16.394 7.2% including 95% trend uncertainty 17.164 IM solvency 95% 0.77 4.7% (multi year) Discussion: trend uncertainty is a typically multiyear risk A shock happens not just in one year. This is different from for example the 1 in 200 one year approach of Solvency II.

34 Conclusions/Advice Important for countries to have reliable, yearly day available to analyse possible developments Look in modelling not only to your own country, but also to other (nearby) countries. Impact longevity on liabilities is high when discount rates are low and impact will be low when discount rates are high. This doesn t mean that when discount rates are high the modelling of longevity is less important: it is still there but only more hidden.

THANK YOU! www.actuaries.org Moving the profession forward internationally