Mortality at Old Ages...Expon. (2NT) Expon. (3NT) Exponential curves capture mortality growth by age well. These curves show no signs of 2NT and 3NT mortality converging, but data

Mortality at Old Ages

Ben Blakeslee

111/11/2020© 2020 Munich American Reassurance Company. All Rights Reserved. Copyrighted presentation; do not forward.

Old Age Mortality – Why don’t we know already?

Industry data is fairly credible through age 99 and fits the VBT15 industry table reasonably wellSource: https://www.soa.org/resources/research-reports/2019/2009-2015-individual-life-mortality/

211/11/2020

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

30.0%

-

1,000

2,000

3,000

4,000

5,000

6,000

7,000

8,000

50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98

ILEC Mortality Data by Attained Age

Claim Count Actual Qx VBT15 Expected Qx

Credibility

Based on the number of

deaths observed in

each category

Several approaches

taken to measure credibility Generally,

100% credibility

requires >1000 deaths

© 2020 Munich American Reassurance Company. All Rights Reserved. Copyrighted presentation; do not forward.

Old Age Mortality – Why don’t we know already?

311/11/2020

0.0%

20.0%

40.0%

60.0%

80.0%

100.0%

120.0%

0

1000

2000

3000

4000

5000

6000

1980 1983 1986 1989 1992 1995 1998 2001 2004 2007

Age 80+ Duration 6+ A/E by Issue Year

Claim Count A/E VBT15

More recent data shows a downward trend away from VBT15 at older ages This is challenging because data is not

available for recent issue years yet


Old Age Mortality – Why has it trended downward?

411/11/2020

Significant underwriting changes have been introduced over time:

Smoker / Non-Smoker

1980 1990 2000 2010

Blood Testing(HIV/AIDS)

Preferred Class Underwriting

Multiple Preferred Classes

Mature Age Underwriting

Product design changes and interest rate changes are another a significant factor


Underwriting Changes Over Time

511/11/2020

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

80.0%

90.0%

100.0%

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

Issue Year

Business Mix by Number of Preferred Classes

1 2 3 4

One way to measure the underwriting changes in

industry experience is via the number of preferred classes

Number of preferred classes increased by issue year

• Only shows NT Classes • Excludes SI Business



611/11/2020

0.0%

20.0%

40.0%

60.0%

80.0%

100.0%

120.0%

-

10,000

20,000

30,000

40,000

50,000

60,000

70,000

1 2 3 4Number of Preferred Classes

Age 80+ Duration 6+ A/E

Number Of Deaths A/E ($) 2015VBT

Structures with more preferred classes have lower mortality in industry data



711/11/2020

Structures with more preferred classes attract more healthy lives

This is expected as additional segmentation yields discounts for healthy lives and charges for unhealthy lives

Preferred classes in the newer structures also have lower mortality than corresponding classes in older structures

A/E is based on VBT 15

Business Mix 1 NT 2 NT 3NT 4NT1

100%64.1%

44.1% 51.7%2 30.2% 25.6%3

35.9% 25.7%13.6%

4 9.0%

A/E VBT15 1NT 2 NT 3NT 4NT1

93.9%81.6% 65.1% 64.7%

2 74.9% 77.8%3

112.3% 98.2% 90.8%4 110.9%

Total 93.9% 95.0% 81.8% 77.8%

74.3% 77.3%


Preferred benefit by duration

811/11/2020

0.0%

20.0%

40.0%

60.0%

80.0%

100.0%

120.0%

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

3NT Structure A/E by Duration VBT15 Ultimate

Best Pref Std Total

Even while underwriting selection effect is wearing off, preferred differentiation shows no

separation



911/11/2020

0.0%

20.0%

40.0%

60.0%

80.0%

100.0%

120.0%

140.0%

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

Total A/E by Duration VBT15 Ultimate

2NT 3NT

More modern 3NT structures have shown signs of leveling off to ultimate mortality rates

sooner than 2NT



1011/11/2020

0.0%

20.0%

40.0%

60.0%

80.0%

100.0%

120.0%

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

Preferred/Standard Ratio

2NT 3NT

Both 2NT and 3NT structures show little sign of preferred wear-off even many years after

underwriting


Preferred structures by age

1111/11/2020

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

30.0%

35.0%

80 82 84 86 88 90 92 94 96 98

Female Age 80+ Duration 6+ Qx

2NT 3NT HMD VBT15 Expon. (2NT) Expon. (3NT)

Exponential curves capture mortality growth by age well

These curves show no signs of 2NT and 3NT mortality converging, but data is limited

Extrapolating only a few years implies significantly different qx at age 99

The extrapolation does give the counterintuitive result that 2NT structure has higher mortality than the general population at older ages

Population Source: https://www.mortality.org/cgi-bin/hmd/country.php?cntr=USA&level=1


Preferred mix at older ages

1211/11/2020

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97

Business Mix by AA

% Preferred 2NT % Preferred 3NT

Industry data shows declining concentration in preferred

classes at older ages, likely different than a model projection may show

2NT structures are ~64% preferred class at younger ages and 3NT structures show ~74% in preferred

classes


Mortality Improvement at Older Ages

1311/11/2020

0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%

14.0%

16.0%

18.0%

78 80 82 84 86 88 90 92 94 96

2NT Experience by AA and Study Year

2009-12

2013-15

Expon. (2009-12)

Expon. (2013-15)

0.0%

1.0%

2.0%

3.0%

4.0%

5.0%

6.0%

7.0%

8.0%

9.0%

78 80 82 84 86 88 90 92 94

3NT Experience by AA and Study Year

2009-12

2013-15

Expon. (2009-12)

Expon. (2013-15)

Splitting experience by study year limits credibility,

but does not show clear signs of mortality

improvement in later years of the study


Beyond Insured Data

1411/11/2020

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

80.0%

90.0%

100.0%

95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121

Female Unismoke Qx

VBT15 Exponential to .5 at 112 Exponential to 1 at 121

At the point that data is no longer available, extrapolation is needed

Extrapolation can be impactful even given the same start and

end points

Two common approaches:• Exponential• Deceleration (Gamma-

Gompertz, Cubic)


Old Age Extrapolations

Source: https://pubmed.ncbi.nlm.nih.gov/31295741/1511/11/2020

Old age mortality data often has issues such as age

exaggeration and heaping

Some have argued that as data quality has improved it

has increasingly pointed to an exponential slope at high ages

Others have found that even after cleaning data, there is a

deceleration pattern


When should the table end?

• Jeanne Calment, the oldest documented person, died at age 122 in 1997.

• Her record can’t be passed for at least 5 years

Source: https://en.wikipedia.org/wiki/Jeanne_Calment

1611/11/2020

1875-1997


1711/11/2020

Questions?

Ben [email protected]


Mortality at Old AgesOld Age Mortality – Why don’t we know already?Old Age Mortality – Why don’t we know already?Old Age Mortality – Why has it trended downward?Underwriting Changes Over TimeUnderwriting Changes Over TimeUnderwriting Changes Over TimePreferred benefit by durationPreferred benefit by durationPreferred benefit by durationPreferred structures by agePreferred mix at older agesMortality Improvement at Older AgesBeyond Insured DataOld Age ExtrapolationsWhen should the table end? Questions?

Documents

Mortality at Old Ages...Expon. (2NT) Expon. (3NT) Exponential curves capture mortality growth by age well. These curves show no signs of 2NT and 3NT mortality converging, but data