Mortality Studies GPV Workshop

Actuarial Experience Studies:
Mortality

By
Luc St-Amour
July 19, 2012

Mortality Experience Studies - Content






Use of Mortality Experience
Source of Mortality Studies
Best Estimate Assumptions
Margin for Adverse Deviation
Experience Studies






Exposure Measurement

Credibility
Additional / Special Considerations

2

Use of Mortality Experience




Valuation
Pricing
Products without mortality Risk?






Still needed as mortality could be a contingency with respect to persistency

Life insurance
Life annuity

3

Source of Mortality Studies


Industry Studies / Mortality Tables



Representative of your Company’s experience?
Indonesia vs Other Countries (e.g. US Society of Actuaries (SOA) mortality tables)





Population Mortality





Not representative of insured mortality (re: underwriting)

Reinsurers’ experience
Own experience





May not be representative due to differences in underwriting standards, socio economic factors, etc


Most Relevant
Credibility

Important to understand the experience underlying the mortality table






Insured mortality
Basis of the table (e.g. SOA tables are valuation tables and include “built in” margins)
Period of observation – experience may be dated and mortality improvements / deterioration may be
required to make the table current
Care needed to ensure that mortality adjustments are appropriate (for example, when using population
mortality as the source for mortality improvements)
Impact of distribution system (e.g. agency vs brokerage)

4


Best Estimate Assumption – Data Differentiation








Life insured’s/Annuitant’s age, sex, smoking habit, health, lifestyle
Duration since issue of policy
Plan of insurance (e.g. term, whole life) and its benefit provided
Company’s underwriting practice
Size of policy
Company’s distribution system (e.g. agency, brokerage, etc.) and marketing
practice
Each “cell” should be as homogeneous as possible

5


Margin for Adverse Deviation(MfAD)





MfAD should increase the liability – requires testing of “direction” of MfAD (i.e. increasing or
reducing the best estimate assumption)
Size of MfAD reflects the degree of uncertainty of the best estimate assumption
Uncertainty relates to misestimation of and deterioration from the best estimate assumption
Canadian Standard for life insurance: K / ex per 1,000 lives





Canadian Standard for annuities







High Margin: K = 15
Low Margin: K = 3.75 (i.e. 25% of High Margin)

High Margin: -20%
Low Margin: -5%

Selection of High or Low margin depends on “Significant Considerations”
In the presence of a Significant Consideration, the MfAD must be at least average of low and
high margin

6

Margin for Adverse Deviation(MfAD) - continued


Significant Considerations – Misestimation of best estimate assumption:








Significant Consideration – Deterioration of best estimate assumption:







Low credibility of the Company’s own experience
Lack of homogeneity
Unrefined method used to determine best estimate assumption (e.g. using single equivalent age for joint policies instead of
each individual age)
Change in underwriting practice in the Company


Anti-selection is present (e.g. re-entry products; underwriting, sales force)
Unfavourable mortality developments have emerged (e.g. AIDS)
Persistency rate of product is low
Premium structure does not recognize mortality differentials as precisely as the rest of the market (e.g. unisex rates; no
distinction between smokers/non-smokers)

Similar considerations apply to annuity business


But favourable mortality developments would be important consideration

7

Experience Studies – Best Estimate Assumption



qx = Deaths / Exposure
Exposure to risk





Deaths refer to Actual Claims






Critical Element of experience studies
Reconcile to Claims registry/P&L
Use annuity payments for annuitant mortality experience

Compute based on count (i.e. number of deaths) and amount
Annual Studies, but use several years’ experience

8

Exposure - Measurement






Study Period
Include all policies exposed to risk during the period
Need to determine entry point and exit point
Entry Point:





Exit Point:





End of exposure period for policies in-force at the end of the study period
Date of termination for other policies

Cause of Termination is an important consideration






Start of exposure period for policies in-force at the beginning of the study period
Issue date of policy for new business

Death claims exposed to the end of the year of death
Lapses/surrenders – exposed to the date of termination
This would be reversed if doing lapse study: lapsed policies exposed for the full period but terminations due to death/surrender
exposed only date of termination

Treatment of late reported claims?



Included in the year of study
Best to perform study a few months after close of observation period to reduce the number of late reported claims

9

Exposure – Measurement
0

r

s

New Entrants

Withdrawals

n lives, aged x+r

w lives, aged x+s

1

Beginning

Ending

A lives, age x

B lives, aged x+1
D deaths during the period

B=A+n–w-D

10

Exposure – Measurement
0

r

s

New Entrants

Withdrawals

n lives, aged x+r

w lives, aged x+s

1

Beginning

Ending

A lives, age x

B lives, aged x+1
D deaths during the period

D = A * qx + n * 1-r qx+r – w * 1-s qx+s

11

Exposure – Mortality for periods less than 1 year




Balducci formula:
0≤t≤1
1-t q x+t = (1-t) * q x,
Simple assumption to compute exposure
Alternate Formulas:
• Uniform distribution of deaths
0≤t≤1
• tq x = t * q x,
• 1-t q x+t = (1-t) * q x, / (1-t*qx)




Constant force of mortality
t

tq x = 1- (1-qx) ,
(1-t) ,

1-tq x+t = 1- (1-qx)

0≤t≤1

0≤t≤1
0≤t≤1

Referring to general formula “D = A * qx + n * 1-r qx+r – w * 1-s qx+s “, Balducci is simplest to use in
exposure formula

12

Exposure – Measurement
0

r

s

New Entrants

Withdrawals

n lives, aged x+r

w lives, aged x+s

1

Beginning

Ending

A lives, age x

B lives, aged x+1
D deaths during the period

D = A * qx + n * 1-r qx+r – w * 1-s qx+s
Using Balducci,
D = A * qx + n * (1-r) * qx – w * (1-s) qx
D = qx * [ A + n * (1-r) – w * (1-s) ]
With qx = D / E, we have
E = A + n * (1-r) – w * (1-s)

13

Exposure – Measurement – Example 1
0

r

s

New Entrants

Withdrawals

n lives, aged x+r

w lives, aged x+s

1

Beginning

Ending

A lives, age x

B lives, aged x+1
D deaths during the period

In the above observation period, we have:
A = 1,000
n = 40
w = 30
r = 1/4
s = 2/3
B = 990
Assuming Balducci, Solve for qx,

14

Exposure – Measurement – Example 1 - Solution
0

r

s

New Entrants

Withdrawals

n lives, aged x+r

w lives, aged x+s

1

Beginning

Ending

A lives, age x

B lives, aged x+1
D deaths during the period

Solution:
First Step: Determine Number of Deaths D
B=A+n–w–D
D=A+n–w-B
D = 1,000 + 40 – 30 - 990
D = 20
15

Exposure – Measurement – Example 1 - Solution
0

r

s

New Entrants

Withdrawals

n lives, aged x+r

w lives, aged x+s

1

Beginning

Ending

A lives, age x

B lives, aged x+1
D deaths during the period

Solution:
Second Step: Determine Exposure assuming Balducci
E = A + (1-r) * n – (1-s) * w
E = 1,000 + (1-1/4) * 40 – (1-2/3) * 30
E = 1,020

16

Exposure – Measurement – Example 1 - Solution
0

r

s

New Entrants

Withdrawals

n lives, aged x+r

w lives, aged x+s

1

Beginning

Ending

A lives, age x

B lives, aged x+1
D deaths during the period

Solution:
Third Step: Calculate qx
qx = D / E
qx = 20 / 1,020
qx ≈ 0.0196
17

Example




Calculate Exposure contributed for each of the following lives.
Observation period begins March 1, 2000.
Balducci hypothesis is assumed.

Case

Birth Date

Other Facts

1

Oct. 1, 1980

Withdrew Feb. 1, 2002

2

Dec. 1, 1981

Died Apr. 1, 2003

3

Apr. 1, 1979

Died May 1, 2004

4

Mar. 1, 1980

Died Apr. 1, 2000

5

Nov. 1, 1980

Withdrew Aug. 1, 2004

18

Example 2 - Solution
Case

Birth Date

Other Facts

1

Oct. 1, 1980

Withdrew Feb. 1, 2002

Exposure Period - Calendar Year
Number of
Months Description
Year
2000
10 Mar 1 - Dec 31
2001
12 Jan 1 - Dec 31
2002
1 Jan 1 - Jan 31
Total
Age at Entry

Exit
Feb 1, 2002

Entry
Mar 1, 2000

23
19 5/12
Oct 1, 1999

Oct 1, 2000

Oct 1, 2001

Oct 1, 2002

Exposure Period - By Age
Number of
Age
Months
Description
19
7 Mar 1, 2000 - Sept 30, 2000
20
12 Oct 1, 2000 - Sept 30, 2001
21
4 Oct 1, 2001 - Jan 31, 2002
Total

23

19

Example 2 - Solution
Case

Birth Date

Other Facts

2

Dec 1, 1981

Died April 1, 2003

Exposure Period - Calendar Years
Number of
Year
Months Description
2000
10 Mar 1 - Dec 31
2001
12 Jan 1 - Dec 31
2002
12 Jan 1 - Dec 31
2003
11 Jan 1 - Nov 30
Total
Age at Entry

Entry
Mar 1, 2000

Died
Apr 1, 2003

Exit
Dec 1, 2003

45
18 3/12
Dec 1, 1999

Dec 1, 2000

Dec 1, 2001

Dec 1, 2002

Dec 1, 2003

Exposure Period - By Age
Number of
Age
Months Description
18
9 Mar 1, 2000 - Nov 30, 2000
19
12 Dec 1, 2000 - Nov 30, 2001
20
12 Dec 1, 2001 - Nov 30, 2002
21
12 Dec 1, 2002 - Nov 30, 2003
Total

45

Note: Since the termination is by death, we give a full year of exposure in the year of death
20

Example 2 - Solution
Case

Birth Date

Other Facts

3

Apr 1, 1979

Died May 1, 2004

Exposure Period - Calendar Years
Number of
Months Description
Year
2000
10 Mar 1 - Dec 31
2001
12 Jan 1 - Dec 31
2002
12 Jan 1 - Dec 31
2003
12 Jan 1 - Dec 31
2004
12 Jan 1 - Dec 31
2005
3 Jan 1 - Mar 31
Total
Age at Entry

Entry
Mar 1, 2000

Died
May 1, 2004

Exit
Apr 1, 2005

61
20 11/12
Apr 1, 1999

Apr 1, 2000

Apr 1, 2001

Apr 1, 2002

Apr 1, 2003

Apr 1, 2004

Apr 1, 2005

Exposure Period - By Age
Number of
Age
Months Description
20
1 Mar 1, 2000 - Mar 31, 2000
21
12 Apr 1, 2000 - Mar 31, 2001
22
12 Apr 1, 2001 - Mar 31, 2002
23
12 Apr 1, 2002 - Mar 31, 2003
24
12 Apr 1, 2003 - Mar 31, 2004
25
12 Apr 1, 2004 - Mar 31, 2005
Total

61

Note: Since the termination is by death, we give a full year of exposure in the year of death
21

Example 2 - Solution
Case

Birth Date

Other Facts

4

Mar 1, 1980

Died Apr 1, 2000

Exposure Period - Calendar Years
Number of
Months Description
Year
2000
10 Mar 1 - Dec 31
2001
2 Jan 1 - Feb 28
Total
Age at Entry

12
20

Exposure Period - By Age
Number of
Months Description
Age
20
12 Mar 1, 2000 - Feb 28, 2001
Total

Died
Apr 1, 2000

Entry
Mar 1, 2000

Exit
Mar 1, 2001

Mar 1, 2000

Mar 1, 2001

12

Note: Since the termination is by death, we give a full year of exposure in the year of death
22

Example 2 - Solution
Case

Birth Date

Other Facts

5

Nov 1, 1980

Withdrew Aug. 1, 2004

Exposure Period - Calendar Years
Number of
Months Description
Year
2000
10 Mar 1 - Dec 31
2001
12 Jan 1 - Dec 31
2002
12 Jan 1 - Dec 31
2003
12 Jan 1 - Dec 31
2004
7 Jan 1 - Jul 31
Total
Age at Entry

Entry
Mar 1, 2000

Exit
Aug 1, 2004

53
19 4/12
Nov 1, 1999

Nov 1, 2000

Nov 1, 2001

Nov 1, 2002

Nov 1, 2003

Nov 1, 2004

Exposure Period - By Age
Number of
Months Description
Age
19
8 Mar 1, 2000 - Oct 31, 2000
20
12 Nov 1, 2000 - Oct 31, 2001
21
12 Nov 1, 2001 - Oct 31, 2002
22
12 Nov 1, 2002 - Oct 31, 2003
23
9 Nov 1, 2003 - Jul 31, 2004
Total

53

23

Credibility


Function of number of claims




Use to blend own experience with industry experience






CIA Educational Note sets 100% credibility at 3,007 death claims (based on Poisson distribution; 90%
confidence interval with 3% margin of error)
Assumption that Industry Tables are 100% credible

Blended qx = Z * own experience + (1-Z) * Industry experience
Z is credibility factor = minimum { (Number of Claims / 3007) ^ (1/2) , 1 }
Examples of Credibility Factors

Number 30
of
Claims

120

271

481

752

1083

1473

1924

2436

3007

Z

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0.10

24

Example 3




Calculate Actual-to-Expected experience by number of lives and by amounts based on the
following data
Calculate Credibility at age 22, assuming 3,007 claims for full credibility
If e52 = 28.89, calculate the Margin for Adverse Deviation (MfAD) for High Margin situations and
total “valuation” mortality at age 52 based on Canadian Actuarial Standards of Practice

Exposure
Age

Actual Claims

1000 qx

Number

Amount

Number

Amount

22

.67

471,797

17,119,826

196

6,685

37

.90

660,127

29,317,759

661

23,777

52

2.91

1,549,048

93,872,023

4,506

228,785

67

16.32

1,098,383

36,781,306

12,720

335,609

82

64.87

534,162

8,686,284

28,878

430,786
25

Example 3 - Solution



Step 1 – Calculate Expected Claims
Expected Claims = Exposure * q x

Exposure
Age
22
37
52
67
82
TOTAL

Expected Claims

1000 qx

Number

Amount

Number

Amount

0.67
0.90
2.91
16.32
64.87

471,797
660,127
1,549,048
1,098,383
534,162
4,313,517

17,119,826
29,317,759
93,872,023
36,781,306
8,686,284
185,777,198

316
594
4,508
17,926
34,651
57,995

11,470
26,386
273,168
600,271
563,479
1,474,774
26

Example 3 - Solution



Step 2 – Actual to Expected
A / E = Actual Claims divided by Expected Claims

Actual Claims
Age
22
37
52
67
82
TOTAL

Expected Claims

A/E

Number

Amount

Number

Amount

Number

Amount

196
661
4,506
12,720
28,878
46,961

6,685
23,777
228,785
335,609
430,786
1,025,642

316
594
4,508
17,926
34,651
57,995

11,470
26,386
273,168
600,271
563,479
1,474,774

62%
111%
100%
71%
83%
81%

58%
90%
84%
56%
76%
70%

27

Example 3 – Solution



Calculate Credibility at age 22, assuming 3,007 claims for full
credibility
Solution:


Credibility = minimum { (Actual Claims / 3007) ^ 0.5, 100%}



Actual Claims [22] = 196



Credibility = minimum { (196 / 3007) ^ 0.5 , 100%}



Credibility = minimum { 26% , 100%} = 26%

28

Example 3 - Solution




If e52 = 28.89, calculate the Margin for Adverse Deviation (MfAD) for
High Margin situations and total “valuation” mortality at age 52
based on Canadian Actuarial Standards of Practice
Solution:




MfAD [x] = K / ex , per 1,000, with k = 15 for High Margin situations
MfAD [52] = 15 / 28.89 per 1,000
MfAD [52] = 0.52 per 1,000



Total Valuation Assumption = Best Estimate Assumption + MfAD (*)




Total q 52 = 2.91 per 1,000 + 0.52 per 1,000
Total q 52 = 3.43 per 1,000

(*) This is applicable for life insurance under “normal condition”. For annuities and “death supported” business, the MfAD
would be deducted from the Best Estimate Assumption in order to increase the actuarial liability.
29

Best Estimate Assumption – Additional Considerations



Persistency




Underwriting Practices





Anti-selective lapses
Medical
Non-medical

Mortality Improvement





Project current mortality from dated studies
Required for annuity business
Accepted practice in Canada
Typical approach:

• qx,n = qx * (1 – improvement) n, where n is number of years since the study of the
underlying qx was performed





“Back-to-Back” contracts
Annuity Business
Death Supported Mortality


May happen with high quota share reinsurance arrangements

30

Example 4



You are given the following qx for males based on a mortality study centered in 2010
Assuming 2% annual mortality improvement, calculate the qx for a policy issued in 2012 for a
male aged 25 at issue

Age

1000 qx

25

.76

26

.78

27

.80

28

.81

29

.82

Adjusted 1000 qx

31

Example 4 - Solution


Step 1 – Determine number of elapsed years since the mortality study

Age

Attained Year

Elapsed Years

25

2012

2

26

2013

3

27

2014

4

28

2015

5

29

2016

6

32

Example 4 - Solution


Step 2 – Determine mortality improvement factor by elapsed year since study

Age

Elapsed
Years

Improvement

25

2

Improvement factor [2] = (1-2%) ^ 2 = 0.9604

26

3

Improvement factor [3] = (1-2%) ^ 3 = 0.9412

27

4

Improvement factor [4] = (1-2%) ^ 4 = 0.9224

28

5

Improvement factor [5] = (1-2%) ^ 5 = 0.9039

29

6

Improvement factor [6] = (1-2%) ^ 6 = 0.8858

33

Example - Solution



Step 3 – Calculate the adjusted q x
Adjusted qx,n = base qx * improvement factor [n], where n is the elapsed number of years since
the mortality study.

Age

Base
1000 qx

Improvement
Factor

Adjusted 1000 qx

25

.76

0.9604

.730

26

.78

0.9412

.734

27

.80

0.9224

.738

28

.81

0.9039

.732

29

.82

0.8858

.726

34

Sources of Information



Canadian Institute of Actuaries Standards of Practice
Canadian Institute of Actuaries Educational Notes:





Expected Mortality: Fully Underwritten Canadian Individual Life Insurance Policies
Margins for Adverse Deviation

Society of Actuaries Mortality Table Construction (Batten) , Prentice Hall, 1978

35