1
The Optimal Product Mix for Hedging
Longevity Risk in Life Insurance Companies
H.C. Huang, Sharon S. Yang, Jennifer L. Wang, Jeff T. Tsai
National Cheng-chi U. /Soochow U./ National Cheng-chi U./National Taiwan U.
2
Outline
Introduction
Literature review
Immunization Strategy
Modeling Longevity Risk
Numerical Illustration
3
Introduction
Hedging longevity risks has taken on an increasingly important role for life insurance companies.
According to the concept of natural hedging, life
insurance can serve as a dynamic hedge vehicle against unexpected mortality risk.
To help life insurers achieve a better natural hedging
effect, we propose an immunization model that incorporates a stochastic mortality to calculate the optimal level of a product mix to effectively reduces longevity risks for insurance companies.
4
Introduction-Con’t
Research Purpose
Using US mortality experience, we
demonstrates that our proposed model can
lead to calculate the optimal product mix
and thus effectively reduce longevity risks
for insurance companies.
5
Literature Review
Mortality risk and pricing issues for annuity
products
Friedman and Warshawsky (1990)
Frees, Carriere, and Valdez (1996)
Mortality derivatives and survival bonds
Blake and Burrows (2001)
Charupat and Milevsky (2001) Lin and Cox (2005 )
Dowd, Blake, Cairns, Dawson (2006) Denuit, Devolder, Godernaiaux (2007)
6
Literature Review-Con’t
Stochastic mortality
Lee and carter (1992)
Marceau and Gaillardetz (1999)
Lee (2000) and Yang (2000)
Milevsky and Promislow (2001, 2002)
Renshaw and Haberman (2003)
Pitacco (2004)
Cairns et. al. (2006)
Schrager (2006)
Natural hedging
7
No discussion in the insurance literature so
far addresses product strategies for natural
hedging.
This paper attempts to fill this gap.
8
The total liability V of the insurer equals
the sum of the liabilities for different
business
To achieve that the effect of changing
mortality on total liability is immunized.
Immunization Strategy
life annuityV
=
V
+
V
0 dV dμ =9
Under the assumption of constant force of mortality,
we define mortality duration for insurance and annuity as follows
Immunization Strategy-Con’t
1
life life lifedV
D
d
V
μ= −
μ
⋅
1
annuity annuity annuitydV
D
d
V
μ= −
μ
⋅
10
Immunization Strategy-Con’t
Therefore, we can achieve by setting
the mortality duration of total liability equal to 0.
where . 0 dV dμ = life annuity life annuity
D
μ=
D
μ⋅
ω
+
D
μ⋅
ω
0
D
μ=
11
The optimal product mix of liability proportions:
Immunization Strategy-Con’t
.
a n n u ity
life a n n u ity life
u
D
D
D
μω
=
−
12
Immunization Strategy-Con’t
The concept of duration employs assumptions of
constant cash flows, flat yield curves, and parallel shifts in interest rates. However, these assumptions may not be realistic in practice.
Kalotay, Williams, and Fabozzi (1993), David,
Merrill, and Panning (1997), and Gajek, Ostaszewski, and Zwiesler (2005) propose effective duration as an alternative risk measure, which also applies to
13
Immunization Strategy-Con’t
Effective Mortality Duration
2
life life life eu lifeV
V
D
V
μ
−−
+=
×
× Δ
2 annuity annuity annuity eu annuity V V D V μ − − + = × × Δ14
15
Stochastic Mortality Model
Lee Carter Model
: the central death rate for age x in year t
exp(αx):the general shape of the mortality schedule
βx : rates decline rapidly and which slowly over time in response to change in κt. κt : is a stochastic process , , s.t 1 0
ln
x t x t x t x x t x t andm
β κα
β κ
ε
= ==
+
+
∑
∑
, x t m16
Practical Issues
Mortality experience for life insurance product
is different to that for annuity product.
Mortality experience is different to countries.
Model risk and parameter risk are important in
dealing with natural hedging.(Melnikov and
Romaniuk 2006; Koissi, Shapiro and Hognas
2006)
17
Mortality Investigation
US mortality experience obtained from HMD data base.
Data period: US aged 25–100 from 1959 to 2002
Trend of Probabilities of Death for 10-Year Age Cohorts, (left: male; right: female)
1960 1965 1970 1975 1980 1985 1990 1995 2000 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 Calander year D eat h r a te age group 50-59 age group 60-69 age group 70-79 age group 80-90 1960 1965 1970 1975 1980 1985 1990 1995 2000 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 Calander year De a th ra te age group 50-59 age group 60-69 age group 70-79 age group 80-90
18
Model Fitting
SVD method (Estimation method has been discussed in Parameter Estimates of αx and βx in Lee-Carter model
Female Male age αx βx age αx βx 25 -7.35118 0.00938 25 -6.36892 0.00712 26 -7.31421 0.00910 26 -6.36916 0.00613 27 -7.27519 0.00881 27 -6.35958 0.00576 28 -7.20874 0.00883 28 -6.32864 0.00532 … … … … … … 98 -1.25084 0.00338 98 -1.14248 0.00188 99 -1.18752 0.00306 99 -1.09254 0.00163 100 -1.12742 0.00276 100 -1.04501 0.00141 101 -1.07054 0.00247 101 -0.99989 0.00120
19
Model Fitting-Con’t
Parameter Estimates of κt in Lee-Carter model
Female Male Year t κt κt 1959 26.64288 19.19884 1960 27.19024 19.59866 1961 23.80692 16.70521 … … … 1999 -23.11967 -29.11703 2000 -24.34012 -30.69297 2001 -24.70004 -32.10778 2002 -25.53304 -32.83798
20
Model Fitting-Con’t
Female Male MAPE 0.0682 0.0893 t t 1 ˆ X : forecasted value X : observed value ˆ 1 n t t t t where X X MAPE n = X − =∑
21
Forecasting Survival Probability
Estimated Confidence Interval of Simulated Survival Rate (left: male, right: female)
30 40 50 60 70 80 90 100 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Ages (x -25)P 2 5
HMD historical survival rate Stochastic mortality model 95% confidence interval 30 40 50 60 70 80 90 100 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Ages (x -25)P 2 5
HMD historical survival rate Stochastic mortality model 95% confidence interval
22
23
Assumptions of Numerical Settings
Product Deferred Annuity Life insurance
Age of insured 25 25
Gender Female Female
Coverage/payout benefit US$10,000 (per year) US$1,000,000
Coverage /payout benefit
period (years) Whole life Whole life Method of paying premium Single Single
Interest rate 4% 4%
Deferred period 30 None
Pricing mortality basis HMD, 2002 HMD, 2002
Forecasted mortality basis Stochastic mortality
model
Stochastic mortality model
24
Insurance Premiums of illustrated insurance
products with different mortality estimate
Pricing Mortality Basis Forecasted Mortality Basis by Lee-Carter (expected) Forecasted Mortality with 10% shock (unexpected) Coverage /payout benefit period 30-year Deferred Annuity Life 30-year Deferred Annuity Life 30-year Deferred Annuity Life 20-year term 37,026 12,836 37,043 11,959 37,495 10,772 30-year term 44,149 24,027 44,212 22,848 45,001 20,606 Whole life 46,749 129,328 46936 128,121 48,105 122,667
25
Optimal Product Mix Ratio
The K-ratio implies that if an insurance company sells one unit of an annuity policy, it should sell K units of life insurance policy to achieve the hedging effect and immunize itself against longevity risk. annuity life annuity e annuity annuity life life e life P D P w K w P D P μ μ = ⋅ = − ⋅
26
Product Mix Proportion and K-Ratio:
(
Women, single premium)
life ω Coverage /payout benefit period 20-Year Term Life Whole Life 11.6% 22.9% (0.380) (0.085) 16.0% 30.1% (0.660) (0.150) 20.7% 37.1% (0.950) (0.210) Whole life annuity
30-year term annuity 20-year term annuity
27
Product Mix Proportion and K-Ratio:
(
Men, single premium)
(
* In parentheses represent the K-ratios.
Coverage /payout benefit period 20-Year Term Life Whole Life 16.0% 31.0% (0.260) (0.096) 20.7% 38.1% (0.420) (0.150) 24.1% 42.8% (0.530) (0.190) Whole life annuity
30-year term annuity 20-year term annuity
28
Product Mix Proportion and K-Ratio:
(
Deferred Period)Product Mix Males Females
31.8% 26.8% 0.220 0.230 42.8% 37.1% 0.190 0.210 55.5% 49.5% 0.150 0.180 Whole life Whole life annuity
(deferred 40 years) Whole life Whole life annuity
(deferred 30 years) Whole life Whole life annuity
(deferred 20 years)
29
Product Mix Proportion and K-Ratio:
(
Different issued Age)* In parentheses represent the K-ratios.
Product Mix Males Females
42.8% 37.1% (0.190) (0.210) 56.8% 50.2% (0.170) (0.190) 71.3% 64.7% (0.130) (0.150) Whole life Whole life annuity
(issued at age of 45) Whole life Whole life annuity
(issued at age of 35) Whole life Whole life annuity
30
Product Mix Proportion and K-Ratio:
(
Between Gender and Age)Product Mix Males Females
56.8% 49.8% (0.170) (0.160) 55.6% 48.9% (0.250) (0.230) 44.5% 34.9% (0.528) (0.467) 57.3% 50.2% (0.210) (0.190) 56.0% 49.3% (0.300) (0.280) 44.5% 34.9% (0.532) (0.472) Whole life (Female, 35)
Whole life annuity
(issued at age of 55, Attend age 65) Whole life
(Female 35)
Deferred life annuity
(issued at age of 45, Attend age 65) Whole life
(Female, 35)
Deferred life annuity
(issued at age of 35, Attend age 65) Whole life
(Male, 35)
Whole life annuity
(issued at age of 55, Attend age 65) Whole life
(Male, 35)
Deferred life annuity
(issued at age of 45, Attend age 65) Whole life
(Male, 35)
Deferred life annuity
31
Product Mix Proportion and K-Ratio:
(with mortality shift 25%)
* The proportions have little difference to 10%-shift, because of the convexity to the productions are not significant. We have a good approximation in linear hedging.
20-Year Term Life Difference in Mortality Curve Shift Whole Life Difference in Mortality Curve Shift 11.6% 0% 22.7% 0% (0.380) 0.076 (0.084) -0.018 16.0% 0% 29.9% 0% (0.660) 0.185 (0.150) -0.014 20.9% 0% 37.1% 0% (0.960) 0.395 (0.210) 0.018 Whole life annuity 30-year term annuity 20-year term annuity
32
Conclusion and Further Research
This paper investigate the optimal strategy for
hedging the longevity risk of an annuity by
using life insurance products.
The proposed immunization model
incorporates stochastic mortality dynamics to
calculate an optimal product mix.
The results strongly demonstrate that the
proposed model can lead to an optimal
product mix and effectively reduce longevity
risks for life insurance companies.
33