Exam LTAM

You have what it takes to pass updated 08/26/19

SURVIVAL DISTRIBUTIONS
SURVIVAL DISTRIBUTIONS Force of Mortality Curtate Expectation
𝑑𝑑 𝑑𝑑 • First Moment
𝑓𝑓# (𝑡𝑡) d𝑡𝑡 7𝑝𝑝# d𝑡𝑡
𝑙𝑙#L7
b b
Probability Functions 𝜇𝜇#L7 = =− =−
𝑆𝑆# (𝑡𝑡) 7𝑝𝑝#
𝑙𝑙#L7 𝑒𝑒# = 𝐸𝐸[𝐾𝐾# ] = t 𝑘𝑘 ⋅ R|𝑞𝑞# = t R𝑝𝑝#
Survival Function
𝑓𝑓# (𝑡𝑡) = 𝑆𝑆# (𝑡𝑡) ⋅ 𝜇𝜇#L7 = 7𝑝𝑝# ⋅ 𝜇𝜇#L7 Ru0 Ruv
𝑇𝑇# : future lifetime or time-to-death of (𝑥𝑥)
• Second Moment
𝑆𝑆# (𝑡𝑡): Probability that (𝑥𝑥) survives 𝑡𝑡 years Finding X𝑝𝑝# Using Force of Mortality b b
= Pr[𝑇𝑇# > 𝑡𝑡] X 𝐸𝐸[𝐾𝐾#m ] = t 𝑘𝑘 m ⋅ R|𝑞𝑞# = t(2𝑘𝑘 − 1) R𝑝𝑝#
= Pr[𝑇𝑇0 > 𝑥𝑥 + 𝑡𝑡|𝑇𝑇0 > 𝑥𝑥] X𝑝𝑝# = exp [− \ 𝜇𝜇#L7 d𝑡𝑡] Ru0 Ruv
0
𝑆𝑆0 (𝑥𝑥 + 𝑡𝑡) #LX • Variance
=
𝑆𝑆0 (𝑥𝑥) = exp [− \ 𝜇𝜇^ d𝑦𝑦] 𝑉𝑉𝑉𝑉𝑉𝑉[𝐾𝐾# ] = 𝐸𝐸[𝐾𝐾#m] − (𝑒𝑒# )m
#
𝑆𝑆# (𝑡𝑡) must satisfy:
Temporary Expectation
• 𝑆𝑆# (0) = 1 Properties of Force of Mortality X
∘
• 𝑆𝑆# (∞) = 0 • 𝜇𝜇#L7 ≥ 0 𝑒𝑒#:X| = \ 𝑡𝑡 ⋅ 7𝑝𝑝# 𝜇𝜇#L7 d𝑡𝑡 + 𝑛𝑛 X𝑝𝑝#
b 0
• 𝑆𝑆# (𝑡𝑡) is a non-increasing function of t • ∫0 𝜇𝜇#L7 d𝑡𝑡 = ∞ X

=\ 7𝑝𝑝# d𝑡𝑡
Actuarial Notations Adding/Multiplying a Constant 0
Xfv X
7𝑝𝑝# : Probability that (𝑥𝑥) survives 𝑡𝑡 years ∗
• 𝜇𝜇#L7 = 𝜇𝜇#L7 + 𝑘𝑘 ⇒ X𝑝𝑝#∗ = X𝑝𝑝# ∙ 𝑒𝑒 fRX 𝑒𝑒#:X| = t 𝑘𝑘 ⋅ R|𝑞𝑞# + 𝑛𝑛 ⋅ X𝑝𝑝# = t R𝑝𝑝#
= Pr(𝑇𝑇# > 𝑡𝑡) ∗ R
• 𝜇𝜇#L7 = 𝑘𝑘 ∙ 𝜇𝜇#L7 ⇒ X𝑝𝑝#∗ = g X𝑝𝑝# h Ru0 Ruv
= 𝑆𝑆# (𝑡𝑡)

∘
7𝑞𝑞# : Probability that (𝑥𝑥) dies within 𝑡𝑡 years Express 𝑝𝑝’s or 𝑞𝑞’s in terms of 𝜇𝜇 Relationship between 𝑒𝑒# and 𝑒𝑒#
b ∘
= Pr(𝑇𝑇# ≤ 𝑡𝑡) . 𝑒𝑒# ≈ 𝑒𝑒# + 0.5
7𝑝𝑝# =\ i𝑝𝑝# ⋅ 𝜇𝜇#Li d𝑠𝑠
= 𝐹𝐹# (𝑡𝑡) 7

7 Recursive Formulas
7𝑝𝑝# + 7𝑞𝑞# = 1
7𝑞𝑞# = \ i.𝑝𝑝# ⋅ 𝜇𝜇#Li d𝑠𝑠 ∘ ∘
𝑒𝑒# = 𝑒𝑒#:X| + X𝑝𝑝# ⋅ 𝑒𝑒#LX
∘

;|7𝑞𝑞# : Probability that (𝑥𝑥) survives 𝑢𝑢 years 0

;L7
and dies within the following 𝑡𝑡 years . 𝑒𝑒# = 𝑒𝑒#:X| + X𝑝𝑝# ⋅ 𝑒𝑒#LX
;|7𝑞𝑞# =\ i𝑝𝑝# ⋅ 𝜇𝜇#Li d𝑠𝑠
∘ ∘ ∘
= ; 𝑝𝑝# ⋅ 7 𝑞𝑞#L; ; 𝑒𝑒#:zLX| = 𝑒𝑒#:z| + z𝑝𝑝# ⋅ 𝑒𝑒#Lz:X|
= ; 𝑝𝑝# − ;L7 𝑝𝑝#
𝑒𝑒#:zLX| = 𝑒𝑒#:z| + z𝑝𝑝# ⋅ 𝑒𝑒#Lz:X|
= ;L7 𝑞𝑞# − ; 𝑞𝑞# Moments
Complete Expectation 𝑒𝑒# = 𝑝𝑝# (1 + 𝑒𝑒#Lv )
Curtate Future Lifetime • First Moment
𝐾𝐾# : number of completed future years by ∘
b
𝑒𝑒# = 𝐸𝐸[𝑇𝑇# ] = \ 𝑡𝑡 ⋅ 7𝑝𝑝# 𝜇𝜇#L7 d𝑡𝑡
(𝑥𝑥) prior to death 0
𝐾𝐾# = ⌊𝑇𝑇# ⌋ b
=\ 7𝑝𝑝# d𝑡𝑡
Pr[𝐾𝐾# = 𝑘𝑘] = R 𝑝𝑝# ∙ 𝑞𝑞#LR = R|𝑞𝑞# 0

Life Table • Second Moment

b
𝑙𝑙#L7
7𝑝𝑝# = 𝐸𝐸[𝑇𝑇#m ] = \ 𝑡𝑡 m ⋅ 7𝑝𝑝# 𝜇𝜇#L7 d𝑡𝑡
𝑙𝑙# 0
b
7𝑑𝑑# 𝑙𝑙# − 𝑙𝑙#L7 = \ 2𝑡𝑡 ⋅ 7𝑝𝑝# d𝑡𝑡
7𝑞𝑞# = =
𝑙𝑙# 𝑙𝑙# 0
• Variance
7𝑑𝑑#L; 𝑙𝑙#L; − 𝑙𝑙#L;L7
;|7𝑞𝑞# = = ∘ m
𝑙𝑙# 𝑙𝑙# 𝑉𝑉𝑉𝑉𝑉𝑉[𝑇𝑇# ] = 𝐸𝐸[𝑇𝑇#m ] − r𝑒𝑒# s

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 1

Special Mortality Laws Fractional Ages INSURANCE INSURANCE
Constant Force of Mortality UDD (0 ≤ 𝑡𝑡 < 1)
𝜇𝜇# = 𝜇𝜇 Use linear interpolation: Type of
𝐄𝐄𝐄𝐄𝐄𝐄
7𝑝𝑝# = 𝑒𝑒
f{7
𝑙𝑙#L7 = (1 − 𝑡𝑡) ⋅ 𝑙𝑙# + 𝑡𝑡 ⋅ 𝑙𝑙#Lv Insurance
∘ 1 = 𝑡𝑡 ⋅ 𝑞𝑞#
7𝑞𝑞# Discrete
𝑒𝑒# = 𝑞𝑞#
𝜇𝜇 b
𝜇𝜇#L7 =

Whole Life
∘ 1 1 − 𝑡𝑡 ⋅ 𝑞𝑞# 𝐴𝐴# = t 𝑣𝑣 RLv ⋅ R|𝑞𝑞#
𝑒𝑒#:X| = (1 − 𝑒𝑒 f{⋅X )
𝜇𝜇 𝑓𝑓# (𝑡𝑡) = 7𝑝𝑝# ∙ 𝜇𝜇#L7 = 𝑞𝑞# Ru0
Continuous
Uniform Distribution Constant Force of Mortality (0 ≤ 𝑡𝑡 < 1) b
𝑙𝑙# = 𝑘𝑘(𝜔𝜔 − 𝑥𝑥) Use exponential interpolation: 𝐴𝐴̅# = \ 𝑣𝑣 7 ⋅ 7𝑝𝑝# 𝜇𝜇#L7 d𝑡𝑡
0
1 1 𝑙𝑙#L7 = (𝑙𝑙# )vf7 ⋅ (𝑙𝑙#Lv )7
𝜇𝜇# = ⇒ 𝜇𝜇#L7 = Discrete
𝜔𝜔 − 𝑥𝑥 𝜔𝜔 − (𝑥𝑥 + 𝑡𝑡)

Term Life
7
7𝑝𝑝# = (𝑝𝑝# ) 𝐴𝐴v#:X| = 𝐴𝐴# − X𝐸𝐸# ⋅ 𝐴𝐴#LX
𝑙𝑙#L7 𝜔𝜔 − (𝑥𝑥 + 𝑡𝑡)
7𝑝𝑝# = = 𝜇𝜇#L7 = − ln(𝑝𝑝# )
𝑙𝑙# 𝜔𝜔 − 𝑥𝑥 Continuous
𝑓𝑓# (𝑡𝑡) = 7𝑝𝑝# ∙ 𝜇𝜇#L7 = 𝑒𝑒 f{⋅7 ⋅ 𝜇𝜇
𝐴𝐴̅ v = 𝐴𝐴̅# − X𝐸𝐸# ⋅ 𝐴𝐴̅#LX

𝑙𝑙# − 𝑙𝑙#L7 𝑡𝑡 #∶X|

7𝑞𝑞# = =
𝑙𝑙# 𝜔𝜔 − 𝑥𝑥 Select & Ultimate Mortality Discrete

Deferred Whole Life

𝑙𝑙#L; − 𝑙𝑙#L;L7 𝑡𝑡 The age at which a person is selected is = 𝐴𝐴# − 𝐴𝐴v#:X|
X|𝐴𝐴#
;|7𝑞𝑞# = =
𝑙𝑙# 𝜔𝜔 − 𝑥𝑥 denoted as [𝑥𝑥].
∘ 𝜔𝜔 − 𝑥𝑥
= X𝐸𝐸# ⋅ 𝐴𝐴#LX
𝑒𝑒# = Continuous
2 Select mortality is written as 𝑞𝑞[#]L7 where 𝑥𝑥
∘ 𝑛𝑛 ̅ ̅ ̅ v
𝑒𝑒#:X| = X𝑝𝑝# (𝑛𝑛) + X𝑞𝑞# r s is the selected age and 𝑡𝑡 is the number of X|𝐴𝐴# = 𝐴𝐴# − 𝐴𝐴 #∶X|
2
years after selection. = X𝐸𝐸# ⋅ 𝐴𝐴̅#LX
Beta Distribution

The mortality after the select period is called Discrete

Endowment
𝑙𝑙# = 𝑘𝑘(𝜔𝜔 − 𝑥𝑥)}
the ultimate mortality, where: 𝐴𝐴 v = X𝐸𝐸# = 𝑣𝑣 X X𝑝𝑝#

Pure
𝛼𝛼 𝛼𝛼 #:X|
𝜇𝜇# = ⇒ 𝜇𝜇#L7 = 𝑞𝑞[#]L7 = 𝑞𝑞#L7
𝜔𝜔 − 𝑥𝑥 𝜔𝜔 − (𝑥𝑥 + 𝑡𝑡) Continuous
}
𝑙𝑙#L7 𝜔𝜔 − (𝑥𝑥 + 𝑡𝑡) N/A
7𝑝𝑝# = = Ä Common Approach
𝑙𝑙# 𝜔𝜔 − 𝑥𝑥 Discrete

Endowment
Read from the left to the right and then
𝜔𝜔 − 𝑥𝑥

Insurance
∘
𝑒𝑒# = continue downwards. 𝐴𝐴
#:X|
= 𝐴𝐴v#:X| + X𝐸𝐸#
𝛼𝛼 + 1
Continuous
Gompertz’s Law 𝑥𝑥 𝑞𝑞[#] 𝑞𝑞[#]Lv 𝑞𝑞[#]Lm 𝑞𝑞#Là 𝐴𝐴̅ ̅
= 𝐴𝐴 v + X𝐸𝐸#
𝜇𝜇# = 𝐵𝐵𝑐𝑐 # 𝑐𝑐 > 1, 𝐵𝐵 > 0 #:X| #:X|
30
𝐵𝐵𝑐𝑐 # 7
7𝑝𝑝# = exp Ñ− (𝑐𝑐 − 1)Ö 31 Varying Insurance
ln 𝑐𝑐 b
32
(𝐼𝐼𝐼𝐼)# = t(𝑘𝑘 + 1)𝑣𝑣 RLv ⋅ R|𝑞𝑞#
Makeham’s Law 33
Ru0
𝜇𝜇# = 𝐴𝐴 + 𝐵𝐵𝑐𝑐 # 𝑐𝑐 > 1, 𝐵𝐵 > 0, 𝐴𝐴 ≥ −𝐵𝐵 b
𝐵𝐵𝑐𝑐 # 7 ̅ ̅)# = \ 𝑡𝑡𝑡𝑡 7 ⋅ 7𝑝𝑝# 𝜇𝜇#L7 d𝑡𝑡
(𝐼𝐼𝐴𝐴
(𝑐𝑐 − 1)Ö ⋅ exp(−𝐴𝐴𝐴𝐴)
7𝑝𝑝# = exp Ñ− 0
ln 𝑐𝑐 X
̅ ̅) v êêê = \ 𝑡𝑡𝑡𝑡 7 ⋅ 7𝑝𝑝# 𝜇𝜇#L7 d𝑡𝑡
(𝐼𝐼𝐴𝐴 #:X|
0
X
í 𝐴𝐴̅) v êêê
(𝐷𝐷 = \ (𝑛𝑛 − 𝑡𝑡)𝑣𝑣 7 ⋅ 7𝑝𝑝# 𝜇𝜇#L7 d𝑡𝑡
#:X|
0
(𝐼𝐼𝐼𝐼) v v
êêê + (𝐷𝐷𝐷𝐷) #:X|
#:X|
v
êêê = (𝑛𝑛 + 1)𝐴𝐴#:X|

̅ ̅) v êêê + (𝐷𝐷
(𝐼𝐼𝐴𝐴 í 𝐴𝐴̅) v êêê = 𝑛𝑛 ⋅ 𝐴𝐴̅ #:X|
v
êêê
#:X| #:X|

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 2

Calculating 𝒏𝒏𝑬𝑬𝒙𝒙 and 𝒏𝒏𝟐𝟐𝑬𝑬𝒙𝒙 from the SULT Recursive Formulas ANNUITIES ANNUITIES

𝐴𝐴# = 𝑣𝑣𝑞𝑞# + 𝑣𝑣𝑝𝑝# ⋅ 𝐴𝐴#Lv
The SULT provides ó𝐸𝐸# , v0𝐸𝐸# , and m0𝐸𝐸# . 𝐴𝐴v#:X| v
Type of
êêê = 𝑣𝑣𝑞𝑞# + 𝑣𝑣𝑝𝑝# ⋅ 𝐴𝐴#Lv:Xfv|
êêêêêêê
𝐄𝐄𝐄𝐄𝐄𝐄
(𝐼𝐼𝐼𝐼)# = 𝑣𝑣𝑞𝑞# + 𝑣𝑣𝑝𝑝# (𝐴𝐴#Lv + (𝐼𝐼𝐼𝐼)#Lv ) Annuities
7𝐸𝐸# can be calculated as:
𝑙𝑙#L7
Due; Discrete
7 7 1/mthly Insurance
7𝐸𝐸# = 𝑣𝑣 ⋅ 7𝑝𝑝# = 𝑣𝑣 ⋅ b
𝑙𝑙#
(z) 1 𝑎𝑎̈ # = t 𝑣𝑣 R ⋅ R𝑝𝑝#
7Lò𝐸𝐸# = 7𝐸𝐸# ⋅ ò𝐸𝐸#L7 𝐾𝐾# = ⌊𝑚𝑚𝑇𝑇# ⌋
𝑚𝑚 Ru0

(z) Immediate; Discrete
m Pr °𝐾𝐾# = 𝑟𝑟¢ = £𝑝𝑝# ⋅ v 𝑞𝑞#L£ = v 𝑞𝑞# Whole Life
7𝐸𝐸# can be calculated as: z £ | z
m 7 m 𝑎𝑎# = 𝑎𝑎̈ # − 1
7𝐸𝐸# = (𝑣𝑣 ) ⋅ 7𝑝𝑝#
b

𝑙𝑙#L7
(z)
𝐴𝐴# = t 𝑣𝑣 (RLv)/z
⋅ R v 𝑞𝑞#
Continuous
= (𝑣𝑣 7 )m ⋅ z | z b
𝑙𝑙# Ru0
𝑎𝑎ê# = \ 𝑣𝑣 7 ⋅ 7𝑝𝑝# d𝑡𝑡

= 𝑣𝑣 7 ⋅ 𝑣𝑣 7 7𝑝𝑝# = 𝑣𝑣 7 ⋅ 7𝐸𝐸# 0
(𝒎𝒎) í 𝒙𝒙
Relationship between 𝑨𝑨𝒙𝒙 , 𝑨𝑨𝒙𝒙 , and 𝑨𝑨 Due; Discrete
Calculating 𝑨𝑨𝟏𝟏𝒙𝒙:𝒏𝒏| from the SULT (Under UDD Assumption) 𝑎𝑎̈ #:X|
êêê = 𝑎𝑎̈ # − X𝐸𝐸# ⋅ 𝑎𝑎̈ #LX

𝐴𝐴v#:X| = 𝐴𝐴 − X𝐸𝐸# 𝑖𝑖 Immediate; Discrete

#:X| 𝐴𝐴̅# = 𝐴𝐴# Temporary
𝛿𝛿
𝐴𝐴v#:X| = 𝐴𝐴# − X𝐸𝐸# ⋅ 𝐴𝐴#LX 𝑖𝑖 v Life 𝑎𝑎#:X|
êêê = 𝑎𝑎̈ #:X|
êêê − 1 + X𝐸𝐸#
𝐴𝐴̅ v êêê = 𝐴𝐴 #:X| êêê

#:X| 𝛿𝛿 Continuous
𝑖𝑖 𝑎𝑎ê#:X|
êêê = 𝑎𝑎
ê# − X𝐸𝐸# ⋅ 𝑎𝑎ê#LX
Uniform ̅
X|𝐴𝐴# = X|𝐴𝐴#
Constant Force 𝛿𝛿
Distribution 𝑖𝑖 v Due; Discrete
𝐴𝐴̅#:X|
êêê = 𝐴𝐴 #:X|
v
êêê + 𝐴𝐴 #:X|êêê
𝜇𝜇 𝑎𝑎êúf#| 𝛿𝛿 X|𝑎𝑎̈ # = 𝑎𝑎̈ # − 𝑎𝑎̈ #:X|
êêê
𝐴𝐴̅# = 𝐴𝐴̅# =
𝜇𝜇 + 𝛿𝛿 𝜔𝜔 − 𝑥𝑥 (z) 𝑖𝑖 = X𝐸𝐸# ⋅ 𝑎𝑎̈ #LX
𝐴𝐴# = (z) 𝐴𝐴# Deferred
𝐴𝐴̅ v 𝑖𝑖
#:X| 𝑎𝑎êX| Whole Life Continuous

𝜇𝜇 ̅
𝐴𝐴 v =
= g1 − X𝐸𝐸# h #:X| 𝜔𝜔 − 𝑥𝑥 Percentiles X|𝑎𝑎
ê# = 𝑎𝑎ê# − 𝑎𝑎ê#:X|
êêê
𝜇𝜇 + 𝛿𝛿
The 100𝑝𝑝th percentile of Z is the value 𝑧𝑧ß = X𝐸𝐸# ⋅ 𝑎𝑎ê#LX
X𝐸𝐸#
such that: Certain-
X𝐸𝐸# = 𝑒𝑒 f({Lù)X 𝜔𝜔 − (𝑥𝑥 + 𝑛𝑛) 𝑎𝑎êêêêêêê
êêê = 𝑎𝑎
#:X|
êêêê
X| + X|𝑎𝑎
ê#
= 𝑣𝑣 X ⋅ Pr®𝑍𝑍 ≤ 𝑧𝑧ß © = 𝑝𝑝 and-Life
𝜔𝜔 − 𝑥𝑥
To calculate 𝑧𝑧ß : Varying Annuities
Calculate 𝐴𝐴 and 𝐴𝐴̅ similarly to 𝐴𝐴 and 𝐴𝐴̅,
m m
1. Draw a graph with Z on y-axis and 𝑇𝑇# b
but with double the force of interest, 𝛿𝛿. on x-axis. (𝐼𝐼𝑎𝑎̈ ) # = t(𝑘𝑘 + 1)𝑣𝑣 R ⋅ R𝑝𝑝#
Equivalently, replace 𝑣𝑣 with 𝑣𝑣 m , or 2. Identify the parts of the curve where Ru0
b
replace 𝑖𝑖 with 2𝑖𝑖 + 𝑖𝑖 m . 𝑍𝑍 ≤ 𝑧𝑧ß . Determine the value of 𝑇𝑇# that ̅ ê) # = \ 𝑡𝑡𝑡𝑡 7 ⋅ 7𝑝𝑝# d𝑡𝑡
(𝐼𝐼𝑎𝑎

0
Variances corresponds to those parts.
(𝐼𝐼𝐼𝐼) #:X|
êêê + (𝐷𝐷𝐷𝐷)#:X|
êêê = (𝑛𝑛 + 1)𝑎𝑎#:X|
êêê
3. Use the value of 𝑇𝑇# from Step 2
Discrete 𝑉𝑉𝑉𝑉𝑉𝑉[𝑍𝑍] X
to calculate 𝑧𝑧ß . ̅ ê) êêê = \ 𝑡𝑡𝑡𝑡 7 ⋅ 7𝑝𝑝# d𝑡𝑡
(𝐼𝐼𝑎𝑎
m #:X|
Whole Life 𝐴𝐴# − (𝐴𝐴# )m 0
X
m í 𝑎𝑎ê) êêê
Endowment Insurance m
𝐴𝐴#:X|
êêê − g𝐴𝐴#:X|
êêê h (𝐷𝐷 #:X| = \ (𝑛𝑛 − 𝑡𝑡)𝑣𝑣 7 7𝑝𝑝# d𝑡𝑡
0

̅ ê) #:X|
(𝐼𝐼𝑎𝑎 í 𝑎𝑎ê) #:X|
êêê + (𝐷𝐷 êêê = 𝑛𝑛𝑎𝑎
ê#:X|
êêê
Replace A with 𝐴𝐴̅ for continuous cases.

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 3

 Uniform Annuities with mthly Payments PREMIUMS PREMIUMS
Constant Force
Distribution (z)
𝐴𝐴# =
(z)
1 − 𝑑𝑑(z) 𝑎𝑎̈ #
Integrate directly, UDD Assumption Net Future Loss
1
𝑎𝑎ê# = or use (z)
𝑎𝑎̈ # = 𝛼𝛼(𝑚𝑚) ⋅ 𝑎𝑎̈ # − 𝛽𝛽(𝑚𝑚) 0𝐿𝐿 = 𝑃𝑃𝑃𝑃(f. benefits) − 𝑃𝑃𝑃𝑃(f. premiums)
𝜇𝜇 + 𝛿𝛿
̅
𝐴𝐴# = 1 − 𝛿𝛿𝑎𝑎ê# (z)

êêê = 𝛼𝛼(𝑚𝑚) ⋅ 𝑎𝑎̈ #:X|

𝑎𝑎̈ #:X| êêê − 𝛽𝛽(𝑚𝑚)(1 − X𝐸𝐸# ) For a whole life insurance policy of $b on
𝑎𝑎ê#:X|
êêê Integrate directly,
(z)
= 𝛼𝛼(𝑚𝑚) ⋅ X|𝑎𝑎̈ # − 𝛽𝛽(𝑚𝑚) ⋅ X𝐸𝐸# (𝑥𝑥) with level premiums P:
1 or use X|𝑎𝑎̈ #
= g1 − X𝐸𝐸# h ̅ 𝑖𝑖𝑖𝑖 Fully Discrete
𝜇𝜇 + 𝛿𝛿 𝐴𝐴#:X| = 1 − 𝛿𝛿𝑎𝑎ê#:X|
êêê
𝛼𝛼(𝑚𝑚) = (z) (z)
𝑖𝑖 𝑑𝑑 0𝐿𝐿 = 𝑏𝑏𝑣𝑣 ≥¥Lv − 𝑃𝑃𝑎𝑎̈ ≥
êêêêêêêêê
¥ Lv|


𝑖𝑖 − 𝑖𝑖 (z) 𝑃𝑃 𝑃𝑃
Variances 𝛽𝛽(𝑚𝑚) = (z) (z) = µ𝑏𝑏 + ∂ 𝑣𝑣 ≥¥Lv −
𝑖𝑖 𝑑𝑑 𝑑𝑑 𝑑𝑑
Discrete 𝑉𝑉𝑉𝑉𝑉𝑉[𝑌𝑌] z fz
𝐸𝐸® 0𝐿𝐿© = 𝑏𝑏𝐴𝐴# − 𝑃𝑃𝑎𝑎̈ #
𝑖𝑖 (z) 𝑑𝑑(z)
m
𝐴𝐴# − (𝐴𝐴# )m [1 + ] = [1 − ]
Whole Life 𝑚𝑚 𝑚𝑚 𝑃𝑃 m
𝑑𝑑m 𝑉𝑉𝑉𝑉𝑉𝑉® 0𝐿𝐿© = µ𝑏𝑏 + ∂ ® m𝐴𝐴# − (𝐴𝐴# )m ©
m
= (1 + 𝑖𝑖) = (1 − 𝑑𝑑)fv 𝑑𝑑
m
Temporary Life 𝐴𝐴#:X|
êêê − g𝐴𝐴#:X|
êêê h Note: Fully Continuous
m

𝑑𝑑 = 𝑏𝑏𝑣𝑣 ∑¥ − 𝑃𝑃𝑎𝑎êêêêê

• The formulas for 𝛼𝛼(𝑚𝑚) and 𝛽𝛽(𝑚𝑚) are 0𝐿𝐿 ∑¥ |

Replace A with 𝐴𝐴̅ and 𝑑𝑑 with 𝛿𝛿 for provided on the LTAM Table. 𝑃𝑃 𝑃𝑃
= µ𝑏𝑏 + ∂ 𝑣𝑣 ∑¥ −
continuous cases. • The values for 𝛼𝛼(𝑚𝑚) and 𝛽𝛽(𝑚𝑚) when 𝛿𝛿 𝛿𝛿
𝑖𝑖 = 0.05 are also given in the LTAM Table. 𝐸𝐸® 0𝐿𝐿© = 𝑏𝑏𝐴𝐴̅# − 𝑃𝑃𝑎𝑎ê#
Recursive Formula
𝑃𝑃 m
𝑎𝑎̈ # = 1 + 𝑣𝑣𝑝𝑝# ⋅ 𝑎𝑎̈ #Lv Woolhouse’s Formula (3 terms) 𝑉𝑉𝑉𝑉𝑉𝑉® 0𝐿𝐿© = µ𝑏𝑏 + ∂ ® m𝐴𝐴̅# − (𝐴𝐴̅# )m ©
𝛿𝛿

(z) 𝑚𝑚 − 1 𝑚𝑚m − 1
Relationship between Insurances and 𝑎𝑎̈ # ≈ 𝑎𝑎̈ # − − (𝜇𝜇 + 𝛿𝛿)
2𝑚𝑚 12𝑚𝑚m # Equivalence Principle
Annuities If the question asks to use the Woolhouse’s
• 𝐸𝐸® 0𝐿𝐿© = 0
Discrete formula with two terms, just drop the
𝐴𝐴# = 1 − 𝑑𝑑𝑎𝑎̈ # last term. ⇒ 𝐸𝐸𝐸𝐸𝐸𝐸(f. premiums) = 𝐸𝐸𝐸𝐸𝐸𝐸(f. benefits)
m • Variance
Whole Life 𝐴𝐴#

If 𝜇𝜇# is not available, approximate 𝜇𝜇# as:

= 1 − (2𝑑𝑑 − 𝑑𝑑m ) m𝑎𝑎̈ #
1 𝑉𝑉𝑉𝑉𝑉𝑉® 0𝐿𝐿©
𝜇𝜇# ≈ − (ln 𝑝𝑝#fv + ln 𝑝𝑝# )
𝐴𝐴#:X|
êêê = 1 − 𝑑𝑑𝑎𝑎̈ #:X|
êêê 2 Discrete m
𝐴𝐴# − (𝐴𝐴# )m
Temporary m

= 𝑏𝑏m ⋅
𝐴𝐴#:X|
êêê
(z) 𝑚𝑚 − 1 (1 − 𝐴𝐴# )m
Life 𝑎𝑎̈ #:X|
êêê ≈ 𝑎𝑎̈ #:X|
êêê − g1 − X𝐸𝐸# h
m) m 2𝑚𝑚
= 1 − (2𝑑𝑑 − 𝑑𝑑 𝑎𝑎̈ #:X|
êêê
𝑚𝑚m − 1 𝑉𝑉𝑉𝑉𝑉𝑉® 0𝐿𝐿©
− ®𝜇𝜇 + 𝛿𝛿 − X𝐸𝐸# (𝜇𝜇#LX + 𝛿𝛿)©
12𝑚𝑚m # Continuous m
𝐴𝐴̅# − (𝐴𝐴̅# )m
Continuous = 𝑏𝑏m ⋅
(1 − 𝐴𝐴̅# )m

If the interest rate is 0:

𝐴𝐴̅# = 1 − 𝛿𝛿𝑎𝑎ê# ∘ 1 1
Whole Life m ̅ 𝑒𝑒# ≈ 𝑒𝑒# + − 𝜇𝜇# í to 𝐴𝐴′𝑠𝑠 and 𝑎𝑎′𝑠𝑠 for endowment
Add 𝑛𝑛|
𝐴𝐴# = 1 − (2𝛿𝛿) m𝑎𝑎ê# 2 12
insurance.
Temporary 𝐴𝐴̅#:X|
êêê = 1 − 𝛿𝛿𝑎𝑎
ê#:X|
êêê
Percentiles
Life m
𝐴𝐴̅#:X| m
êêê = 1 − (2𝛿𝛿) 𝑎𝑎
ê#:X|
êêê The 100pth percentile of Y is the value 𝑦𝑦ß Gross Premium
π
such that: 0𝐿𝐿
= 𝑃𝑃𝑃𝑃(f. benefits) + 𝑃𝑃𝑃𝑃(f. expenses)
Pr®𝑌𝑌 ≤ 𝑦𝑦ß © = 𝑝𝑝 − 𝑃𝑃𝑃𝑃(f. gross premiums)
Equivalence Principle
To calculate 𝑧𝑧ß :
• 𝐸𝐸® 0𝐿𝐿π © = 0
1. Draw a graph with Y on y-axis and 𝑇𝑇#
on x-axis. ⇒ 𝐸𝐸𝐸𝐸𝐸𝐸(f. gross premiums)
2. Identify the parts of the curve where = 𝐸𝐸𝐸𝐸𝐸𝐸(f. benefits) + 𝐸𝐸𝐸𝐸𝐸𝐸(f. expenses)

𝑌𝑌 ≤ 𝑦𝑦ß . Determine the value of 𝑇𝑇# that
corresponds to those parts.
3. Use the value of 𝑇𝑇# from Step 2 to
calculate 𝑦𝑦ß .

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 4

Variance RESERVES RESERVES Gross Premium Reserve
For a fully discrete whole life policy: Prospective Method
π
𝐺𝐺 − 𝑒𝑒£ ≥ Lv Net Premium Reserve 7𝑉𝑉
π
= 𝐸𝐸𝐸𝐸𝑉𝑉7 (f. ben.) + 𝐸𝐸𝐸𝐸𝑉𝑉7 (f. exp.)
0𝐿𝐿 = µ𝑏𝑏 + 𝐸𝐸 + ∂ 𝑣𝑣 ¥
𝑑𝑑 Prospective Method − 𝐸𝐸𝐸𝐸𝑉𝑉7 (f. pre.)
𝐺𝐺 − 𝑒𝑒£
− + (𝑒𝑒ª − 𝑒𝑒£ ) 7𝑉𝑉 = 𝐸𝐸𝐸𝐸𝑉𝑉7 (f. benefits)
𝑑𝑑
− 𝐸𝐸𝐸𝐸𝑉𝑉7 (f. premiums) 0𝑉𝑉
π
= 0 if the following 2 requirements
𝑉𝑉𝑉𝑉𝑉𝑉® 0𝐿𝐿π ©
are satisfied:
𝐺𝐺 − 𝑒𝑒£ m m If 7𝑉𝑉 occurs at the same time as a premium
= µ𝑏𝑏 + 𝐸𝐸 + ∂ ® 𝐴𝐴# − (𝐴𝐴# )m © 1. The gross premium is determined using
𝑑𝑑 or benefit, then be careful about which cash
where the equivalence principle.
flows to include in calculating the future
𝑒𝑒ª = initial expense in year 1 2. The assumptions used for calculating the
loss. Unless stated otherwise, assume:
𝑒𝑒£ =renewal expense in year 2+ reserve are the same as those used in
- all death benefits at time t occurred in the
𝑏𝑏 = benefit amount calculating the premium.
past
𝐸𝐸 = settlement expense
- all premium payments occur in the future
𝐺𝐺 = gross premium Expense Reserve
- endowment payments occur in the future
Expense Premium/Loading

Note: = Gross Premium − Net Premium
Note:
√ π
1. Replace 𝐴𝐴 and d with their continuous - The time-0 net premium reserve is 0 7𝑉𝑉 = 7𝑉𝑉 − 7𝑉𝑉 X
counterparts for fully continuous policies. √
because the equivalence principle is 7𝑉𝑉 = 𝐸𝐸𝐸𝐸𝑉𝑉7 (f. exp.)
2. Add 𝑛𝑛| í to 𝐴𝐴′𝑠𝑠 for endowment insurance. assumed: − 𝐸𝐸𝐸𝐸𝑉𝑉7 (f. exp. premium)
3. This shortcut formula can only be used Expense reserve is usually negative.
7 𝑉𝑉 = 𝐸𝐸® 0𝐿𝐿© = 0
for fully discrete/continuous whole life - The time-n net premium reserve for an

Recursive Formula
and endowment insurance. For other n -year term insurance is 0 because there
products, use basic principles. • Net premium reserve
are no future benefits or premiums due
g 7𝑉𝑉 + 𝑃𝑃h(1 + 𝑖𝑖) = 𝑞𝑞#L7 ⋅ 𝑏𝑏 + 𝑝𝑝#L7 ⋅ 7Lv𝑉𝑉
at time n:
Portfolio Percentile Premium
X𝑉𝑉
= 0
𝑆𝑆 = 𝐿𝐿v + 𝐿𝐿m + ⋯ + 𝐿𝐿Ω • Gross premium reserve
- The time-n net premium reserve for an n- π
𝐸𝐸[𝑆𝑆] = 𝑁𝑁 ⋅ 𝐸𝐸[𝐿𝐿] g 7𝑉𝑉 + 𝐺𝐺 − 𝑒𝑒h(1 + 𝑖𝑖) = 𝑞𝑞#L7 ⋅ (𝑏𝑏 + 𝐸𝐸)
year endowment insurance right before
𝑉𝑉𝑉𝑉𝑉𝑉[𝑆𝑆] = 𝑁𝑁 ⋅ 𝑉𝑉𝑉𝑉𝑉𝑉[𝐿𝐿] π
the endowment benefit is paid is equal to + 𝑝𝑝#L7 ⋅ 7Lv𝑉𝑉

the endowment benefit, because there are
Using the portfolio percentile premium
no future premiums due at time n, and the Thiele’s Differential Equation
principle, the premium is set such that there
only future benefit due at time n is the d
is a specified probability (x%) that the total 𝑉𝑉 = 𝛿𝛿7 ⋅ 7𝑉𝑉 + 𝐺𝐺7 − 𝑒𝑒7
endowment benefit. d𝑡𝑡 7
loss is negative:
−g𝑏𝑏7 + 𝐸𝐸7 − 7𝑉𝑉h𝜇𝜇#L7
Pr[𝑆𝑆 < 0] = 𝑥𝑥% X𝑉𝑉 = endowment benefit
where
Percentile of 𝟎𝟎𝑳𝑳 Special Formulas 𝐺𝐺7 = annual rate of premium payable
The 100pth percentile of 0𝐿𝐿 is the value 𝜋𝜋ß For a fully discrete whole life insurance at time t
policy of $b, the net premium reserve can 𝑒𝑒7 = annual rate of premium-related
such that Pr® 0𝐿𝐿 ≤ 𝜋𝜋ß © = 𝑝𝑝. To determine
also be calculated as: expense payable at time t
𝜋𝜋ß : 𝑎𝑎̈ #L7 𝐸𝐸7 = expense of paying the face amount at
1. Graph 0𝐿𝐿 on y-axis and 𝑇𝑇# on x-axis. 7𝑉𝑉 = µ1 − ∂ ⋅ 𝑏𝑏
𝑎𝑎̈ # time t (e.g., settlement/claim expense)
2. Identify the parts of the curve where 𝐴𝐴#L7 − 𝐴𝐴#
7𝑉𝑉 = µ ∂ ⋅ 𝑏𝑏 𝑏𝑏7 = face amount payable at time t if the
0𝐿𝐿 ≤ 𝜋𝜋ß . Determine the value of 𝑇𝑇# that 1 − 𝐴𝐴#
insured dies at exact time t
corresponds to those parts. Note:

3. Use the value of 𝑇𝑇# from Step 2 to 1. Replace 𝐴𝐴 and 𝑎𝑎 with their continuous
calculate 𝜋𝜋ß . counterparts for fully continuous policies.
2. Add 𝑛𝑛|í to 𝐴𝐴′𝑠𝑠 and 𝑎𝑎′𝑠𝑠 for endowment

insurance.
3. These special formulas can only be used
for fully discrete/continuous whole life
and endowment insurance. For other
products, use basic principles.

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 5

To solve the differential equation, use MARKOV CHAINS
MARKOV CHAINS Premiums
Euler’s method: For an insurance on (𝑥𝑥) currently in
• Forward Euler Approximation: ª… state i that pays $1 immediately upon
7 𝑝𝑝# : probability someone in state 𝑖𝑖 at age 𝑥𝑥
d 7Lò𝑉𝑉 − 7𝑉𝑉 is in state 𝑗𝑗 (where 𝑗𝑗 may equal 𝑖𝑖) at age every transition to state j:
𝑉𝑉 ≈ b
d𝑡𝑡 7 ℎ 𝑥𝑥 + 𝑡𝑡 𝐴𝐴̅ª…
# = \ 𝑒𝑒
fù7 R…
t 7𝑝𝑝#ªR ⋅ 𝜇𝜇#L7 d𝑡𝑡
• Backward Euler Approximation: ªª 0
7𝑝𝑝# : probability someone in state 𝑖𝑖 at age 𝑥𝑥 RÃ…
d 7𝑉𝑉 − 7fò𝑉𝑉 remains in state 𝑖𝑖 until age 𝑥𝑥 + 𝑡𝑡
b
𝑉𝑉 ≈ = \ 𝑒𝑒 fù7
⋅ Pr[start in 𝑖𝑖, move into 𝑗𝑗]d𝑡𝑡
d𝑡𝑡 7 ℎ ªª
7𝑝𝑝# ≤ 7𝑝𝑝#ªª 0

For net premium reserve, drop For an annuity on (𝑥𝑥) currently in state i
expense-related terms and replace 𝐺𝐺7 Continuous Probabilities that pays $1 per year while the person is in
with net premium. • Direct Approach state j:
b
7 𝑎𝑎ê#ª… = \ 𝑒𝑒 fù7 7𝑝𝑝#ª… d𝑡𝑡
ª…
Interim Reserves (𝟎𝟎 ≤ 𝒔𝒔 < 𝟏𝟏) ªª
7𝑝𝑝# = exp À− \ t 𝜇𝜇#Li d𝑠𝑠Õ 0
0 …ê b
• Exact value: ª… ª…
𝑎𝑎̈ # = t 𝑣𝑣 R R𝑝𝑝#
7Li𝑉𝑉
For permanent disability model:
Ru0
7
g 7𝑉𝑉 + 𝑃𝑃h(1 + 𝑖𝑖)i − i𝑞𝑞#L7 ⋅ 𝑏𝑏 ⋅ 𝑣𝑣vfi ª… ªª ª… ……
= 7 𝑝𝑝# =\ i𝑝𝑝# ⋅ 𝜇𝜇#Li ⋅ 7fi𝑝𝑝#Li d𝑠𝑠
i𝑝𝑝#L7
0 Reserves

• Linear approximation: Thiele’s Differential Equation
• Approximation
7Li𝑉𝑉 = g 7𝑉𝑉 + 𝑃𝑃h(1 − 𝑠𝑠) + 𝑠𝑠 ⋅ 7Lv𝑉𝑉 d
Kolmogorov’s Forward Equations: 𝑉𝑉 (ª)
d𝑡𝑡 7
d
X
(ª)
Modified Reserve 𝑝𝑝ª… = tg 7𝑝𝑝#ªR ⋅ 𝜇𝜇#L7
R…
− 7𝑝𝑝#ª… ⋅ 𝜇𝜇#L7
…R
h = 𝛿𝛿7 7𝑉𝑉 (ª) − 𝐵𝐵7
d𝑡𝑡 7 # X
Ru0
A modified reserve is a reserve computed ª… ª…
RÃ…
− t 𝜇𝜇#L7 g𝑏𝑏7 + 7𝑉𝑉 (…) − 7𝑉𝑉 (ª) h
without expenses but adjusting the = Pr(Start in 𝑖𝑖, move into 𝑗𝑗) …u0
valuation premiums to allow implicitly for −Pr(Start in 𝑖𝑖, move out of 𝑗𝑗) …ê

initial expenses. where

To solve the differential equation, use (ª)
𝐵𝐵7 = rate of benefit payment while the
For any modified reserve method, the EPV Euler’s method: insured is in state 𝑖𝑖
of the modified premiums must equal the d
ª… ª…
7Lò𝑝𝑝# − 7𝑝𝑝# ª…
𝜇𝜇#L7 = force of transition from state 𝑖𝑖
EPV of the benefits, which also equals the 𝑝𝑝ª… ≈
d𝑡𝑡 7 # ℎ
to 𝑗𝑗 at age 𝑥𝑥 + 𝑡𝑡
EPV of net premiums.

Alternatively, with Euler's method, a ª…

𝑏𝑏7 = lump sum benefit payable
Full preliminary term (FPT) continuous Markov chain can also be instantaneously at time 𝑡𝑡 on transition
FPT reserve is a modified reserve. The transformed to a discrete Markov chain from state 𝑖𝑖 to 𝑗𝑗
policy is treated as if it were issued one year with time increments of ℎ, and the

later, with the first year of the policy being transition probability can be In this model, premiums are treated as
treated as if it were a one-year term approximated as: negative benefits and expenses can be
insurance. ª… treated as additions to the benefits.
ª… ℎ𝜇𝜇 𝑖𝑖 ≠ 𝑗𝑗
ò𝑝𝑝# ≈ – # ª∙
1 − ℎ𝜇𝜇# 𝑖𝑖 = 𝑗𝑗 To solve the differential equation, use
For example, using this method, a whole life
insurance issued to (𝑥𝑥) would be treated as Euler’s method:
(ª) (ª)
if it were a one-year term insurance on (𝑥𝑥), d (ª) 7𝑉𝑉 − 7fò𝑉𝑉
7𝑉𝑉 =
followed by a whole life insurance on d𝑡𝑡 ℎ
(𝑥𝑥 + 1):
• FPT net premium
1st year modified net premium = 𝐴𝐴v#:v|
í
𝐴𝐴#Lv
Renewal modified net premium =
𝑎𝑎̈ #Lv
• FPT reserve
«»∑
v𝑉𝑉# = 0
«»∑
7𝑉𝑉# = 7fv𝑉𝑉#Lv

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 6

MULTIPLE DECREMENT MODELS
MULTIPLE DECREMENT MODELS Insurance Applications Key Relationships between
Consider a whole life policy: ÷(…) (…)
7𝑞𝑞# and 7𝑞𝑞#
Multiple Decrement Tables (MDT) No Assumption
Decrements are dependent on each other. Discrete
(“)
z ÷(…)
Discrete Probabilities 𝐸𝐸𝐸𝐸𝐸𝐸[benefits] ⎧ 7𝑝𝑝# = ∏…uv 7𝑝𝑝#
7fv 7fv z b (“) (…)
• 7𝑝𝑝 = 1 − ∑z …uv 7𝑞𝑞#

(…) (“)
7𝑞𝑞# = t R 𝑝𝑝# 𝑞𝑞#LR = t R|𝑞𝑞#
(…) (…)
= t t 𝑣𝑣 RLv 𝑏𝑏
(…) (“) (…)
R 𝑝𝑝# 𝑞𝑞#LR
⎨ #(“) (“)
Ru0 Ru0 …uv Ru0
⎩ 7𝑝𝑝# + 7𝑞𝑞# = 1
z b
(“) (…) (“)
7𝑞𝑞# = t 7𝑞𝑞# 𝐸𝐸𝐸𝐸𝐸𝐸[annuity] = t 𝑣𝑣 R R𝑝𝑝# d (‡)
⎧ 𝜇𝜇(…) = dfi fifl¥
…uv Ru0 ⎪ #L7 (·)
fiߥ
(“) (“) Continuous •
7𝑝𝑝# + 7𝑞𝑞# = 1 d ‚(‡)
⎨ (…) fl
dfi fi ¥
7L;fv b ⎪𝜇𝜇#L7 = ß‚(‡)
(…) (“) (…) (“) (…) 𝐸𝐸𝐸𝐸𝐸𝐸[benefits] = \ 𝑣𝑣 7 7𝑝𝑝# 𝜇𝜇#L7 d𝑡𝑡
(“) (…) ⎩ fi ¥

7|;𝑞𝑞# = t R 𝑝𝑝# 𝑞𝑞#LR = 7𝑝𝑝# ;𝑞𝑞#L7 0

Ru7 (…)
𝜇𝜇# UDD in Multiple-Decrement Tables
𝐸𝐸𝐸𝐸𝐸𝐸[benefits] = (“)
(Under CF)
Life Table 𝛿𝛿 + 𝜇𝜇 (UDDMDT) (0 ≤ 𝑠𝑠 < 1)
(…) (‡)
b ߴ
(…) 7𝑑𝑑# 7 (“)
7𝑞𝑞# = 𝐸𝐸𝐸𝐸𝐸𝐸[annuity] = \ 𝑣𝑣 7𝑝𝑝# d𝑡𝑡 ÷(…) (“) fl (·)
(“)
𝑙𝑙# 0 i𝑝𝑝# =r i𝑝𝑝# s ¥

(“) (“) (“)

(“) 7𝑑𝑑# 𝑙𝑙# − 𝑙𝑙#L7
𝑞𝑞
7 # = (“)
= (“)
Fractional Ages (0 ≤ 𝑠𝑠 < 1) UDD in Associated Single Decrement Tables
𝑙𝑙# 𝑙𝑙#
• UDD in the multiple decrement table: (UDDASDT)(0 ≤ 𝑠𝑠 < 1)
(“)
(“) 𝑙𝑙#L7 (…) (…) For 2 decrements:
𝑝𝑝
7 # = (“)
i𝑞𝑞# = 𝑠𝑠 ⋅ 𝑞𝑞#
𝑙𝑙# (“) (“)
÷(m)
𝑠𝑠 m 𝑞𝑞#
i𝑞𝑞# = 𝑠𝑠 ⋅ 𝑞𝑞# (v)
(…) i𝑞𝑞# = 𝑞𝑞#÷(v) „𝑠𝑠 − ‰
(…) ;𝑑𝑑#L7 2
7|;𝑞𝑞# = (“)
• Constant forces of decrement:
𝑙𝑙# (…) (…) (…) For 3 decrements:
𝜇𝜇# 𝑞𝑞# i𝑞𝑞#

= =
Continuous Probabilities (“)
𝜇𝜇# 𝑞𝑞#
(“) (“) 𝑠𝑠 m r𝑞𝑞#÷(m) + 𝑞𝑞#÷(à)s
i𝑞𝑞# (v)
i𝑞𝑞# = 𝑞𝑞#÷(v) Â𝑠𝑠 −
7 (…) 2
(…) (“) (…) 𝑞𝑞# (“) i
7𝑞𝑞# = \ i𝑝𝑝# 𝜇𝜇#Li d𝑠𝑠
(…)
0 i𝑞𝑞# = (“) °1
− r𝑝𝑝# s ¢
𝑞𝑞# 𝑠𝑠 à 𝑞𝑞#÷(m) ∙ 𝑞𝑞#÷(à)
d (…)
+ Á
7𝑞𝑞#
= d𝑡𝑡 (“) 3
(…)
⇒ 𝜇𝜇#L7
Associated Single Decrement Tables
7𝑝𝑝#
z (ASDT) CF in MDT or ASDT (0 ≤ 𝑠𝑠 < 1)
(“) (…) (‡)
𝜇𝜇#L7 = t 𝜇𝜇#L7 The associated single decrements are fl¥
÷(…) (“) fl (·)
…uv independent. i𝑝𝑝# =r i𝑝𝑝# s ¥
7
(“) (“) 7
7𝑝𝑝# = exp [− \ 𝜇𝜇#Li d𝑠𝑠] ÷(…) (…)
7𝑝𝑝# = exp [− \ 𝜇𝜇#Li d𝑠𝑠]
0 0
7
(“) (“) (“) (…) d
7𝑞𝑞# =\ i𝑝𝑝# 𝜇𝜇#Li d𝑠𝑠 ⇒ 𝜇𝜇#L7 =− ln r 7𝑝𝑝#÷(…)s
0 d𝑡𝑡
d (“) d (“) 7
÷(…) (…)
7𝑞𝑞# − 7𝑝𝑝#
÷(…)
(“) d𝑡𝑡 d𝑡𝑡 7𝑞𝑞# =\ i𝑝𝑝# 𝜇𝜇#Li d𝑠𝑠
⇒ 𝜇𝜇#L7 = (“)
= (“)
0
7𝑝𝑝# 7𝑝𝑝# d ÷(…)
7L; 7𝑞𝑞#
⇒ 𝜇𝜇#L7 = d𝑡𝑡
(…) (…)
(“) (…)
7|;𝑞𝑞# =\ i𝑝𝑝# 𝜇𝜇#Li d𝑠𝑠 ÷(…)
7 7𝑝𝑝#
÷(…)
7𝑝𝑝# + 7𝑞𝑞#÷(…) = 1
z
(“) ÷(…)
7𝑝𝑝# = ◊ 7𝑝𝑝#
…uv

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 7

 MULTIPLE LIVES
MULTIPLE LIVES Relationships between (𝒙𝒙𝒙𝒙) Status and Relationships
(𝒙𝒙𝒙𝒙
êêêê) Status v
7𝑞𝑞#^
v
+ 7𝑞𝑞#^
= 7𝑞𝑞#^
Joint Life 𝑇𝑇#^ + 𝑇𝑇#^ = 𝑇𝑇# + 𝑇𝑇^ v m
= 7𝑞𝑞#
7𝑞𝑞#^ + 7𝑞𝑞#^
𝑇𝑇#^ = ming𝑇𝑇# , 𝑇𝑇^ h 𝑇𝑇#^ ⋅ 𝑇𝑇#^ = 𝑇𝑇# ⋅ 𝑇𝑇^ v m
7𝑞𝑞#^ = 7𝑞𝑞#^ + 7𝑞𝑞# ⋅ 7𝑝𝑝^
7𝑝𝑝#^ + 7𝑞𝑞#^ = 1 7𝑝𝑝#^ + 7𝑝𝑝#^ = 7𝑝𝑝# + 7𝑝𝑝^ m m
7𝑞𝑞#^ + 7𝑞𝑞#^ = 7𝑞𝑞#^
∘ ∘ ∘ ∘
;|7𝑞𝑞#^ = ;𝑝𝑝#^ ⋅ 7𝑞𝑞#L;:^L; 𝑒𝑒#^ + 𝑒𝑒#^ = 𝑒𝑒# + 𝑒𝑒^ v v
b𝑞𝑞#^ + b𝑞𝑞#^ = 1
∘ ∘ ∘ ∘
= ;𝑝𝑝#^ − ;L7𝑝𝑝#^ 𝑒𝑒#^:X|
êêê + 𝑒𝑒#^:X|
êêê = 𝑒𝑒#:X|
êêê + 𝑒𝑒^:X|
êêê m m
b𝑞𝑞#^ + b𝑞𝑞#^ = 1
= ;L7𝑞𝑞#^ − ;𝑞𝑞#^ 𝑒𝑒#^ + 𝑒𝑒êêêê
#^ = 𝑒𝑒# + 𝑒𝑒^ v m
b𝑞𝑞#^ = b𝑞𝑞#^

𝐴𝐴̅#^ + 𝐴𝐴̅#^ = 𝐴𝐴̅# + 𝐴𝐴̅^
Independent Lives
𝑎𝑎ê#^ + 𝑎𝑎ê#^ = 𝑎𝑎ê# + 𝑎𝑎ê^ Contingent Insurance
7𝑝𝑝#^ = 7𝑝𝑝# ⋅ 7𝑝𝑝^
b
X𝐸𝐸#^ + X𝐸𝐸#^ = X𝐸𝐸# + X𝐸𝐸^ 𝐴𝐴v̅#^ = \ 𝑣𝑣 7 ⋅ 7𝑝𝑝#^ ⋅ 𝜇𝜇#L7 d𝑡𝑡
7𝑞𝑞#^ = 7𝑞𝑞# + 7𝑞𝑞^ − 7𝑞𝑞# ⋅ 7𝑞𝑞^
0
𝜇𝜇#L7:^L7 = 𝜇𝜇#L7 + 𝜇𝜇^L7 Covariance of 𝑇𝑇#^ and 𝑇𝑇#^ b

𝐴𝐴̅m#^ = \ 𝑣𝑣 7 ⋅ 7𝑝𝑝# ⋅ 𝜇𝜇#L7 ⋅ 7𝑞𝑞^ d𝑡𝑡

Covg𝑇𝑇#^ , 𝑇𝑇#^ h = Covg𝑇𝑇# , 𝑇𝑇^ h 0
Moments ∘ ∘ ∘ ∘
b + r𝑒𝑒# − 𝑒𝑒#^ s r𝑒𝑒^ − 𝑒𝑒#^ s Relationships
∘
𝑒𝑒#^ = \ 7𝑝𝑝#^ d𝑡𝑡
0 Covg𝑇𝑇# , 𝑇𝑇^ h = 0 if 𝑇𝑇# and 𝑇𝑇^ are independent 𝐴𝐴v̅#^ + 𝐴𝐴̅ v ̅
#^ = 𝐴𝐴#^
b
m

𝐴𝐴̅m#^ + 𝐴𝐴̅ m ̅
#^ = 𝐴𝐴#^
𝐸𝐸 °g𝑇𝑇#^ h ¢ = 2 \ 𝑡𝑡 ⋅ 7𝑝𝑝#^ d𝑡𝑡 Exactly One Life Survives
0 𝐴𝐴#^ + 𝐴𝐴#^ = 𝐴𝐴̅#
̅
v ̅ m
b Pr(exactly one life survivies 𝑡𝑡 years)

𝑒𝑒#^ = t R𝑝𝑝#^ = 7𝑝𝑝#^ − 7𝑝𝑝#^
Reversionary Annuities
Ruv
= 7𝑝𝑝# + 7𝑝𝑝^ − 2 ⋅ 7𝑝𝑝#^ • Make payments to (y) after (x) has died:

𝑎𝑎ê#|^ = 𝑎𝑎ê^ − 𝑎𝑎ê#^
Last Survivor
Relationships between Insurance
𝑇𝑇êêêê
#^ = maxg𝑇𝑇# , 𝑇𝑇^ h • Make payments only when exactly one
Policies, Annuities, and Premiums
life is alive:
7𝑝𝑝#^ + 7𝑞𝑞#^ = 1 𝐴𝐴#^ = 1 − 𝑑𝑑𝑎𝑎̈ #^
𝐸𝐸𝐸𝐸𝐸𝐸(annuities) = 𝑎𝑎ê#^ − 𝑎𝑎ê#^
;|7𝑞𝑞#^ = ;𝑝𝑝#^ − ;L7𝑝𝑝#^ ̅#^
1 − 𝐴𝐴êêêê
êêêê =
𝑎𝑎ê#^
= ;L7𝑞𝑞#^ − ;𝑞𝑞#^ 𝛿𝛿
𝑎𝑎̈ #^:X|êêê = 𝑎𝑎̈ #^ − X𝐸𝐸#^ ∙ 𝑎𝑎̈ #LX:^LX
Independent Lives 1
𝑃𝑃êêêê
#^ = − 𝑑𝑑
7𝑞𝑞#^ = 7𝑞𝑞# ⋅ 7𝑞𝑞^ 𝑎𝑎̈ êêêê
#^

7𝑝𝑝#^ = 7𝑝𝑝# + 7𝑝𝑝^ − 7𝑝𝑝# ⋅ 7𝑝𝑝^ 𝛿𝛿𝐴𝐴̅#^

𝑃𝑃#^ =
1 − 𝐴𝐴̅#^
7𝑝𝑝# 𝜇𝜇#L7 ⋅ 7𝑞𝑞^ + 7𝑝𝑝^ 𝜇𝜇^L7 ⋅ 7𝑞𝑞#
𝜇𝜇#^ (𝑡𝑡) = Note: The list above is not exhaustive;
7𝑝𝑝# + 7𝑝𝑝^ − 7𝑝𝑝# ⋅ 7𝑝𝑝^
similar relationships can be applied to other
Moments forms of insurance/annuities with
b
∘ appropriate adjustments.
𝑒𝑒#^ = \ 7𝑝𝑝#^ d𝑡𝑡
0
b
Contingent Probabilities
𝑒𝑒#^ = t R𝑝𝑝#^ 7
v
Ruv 7𝑞𝑞#^ =\ i𝑝𝑝#^ ∙ 𝜇𝜇#Li 𝑑𝑑𝑑𝑑
0
7
v
7𝑞𝑞#^ =\ i𝑝𝑝#^ ∙ 𝜇𝜇^Li 𝑑𝑑𝑑𝑑
0
7
m
7𝑞𝑞#^ =\ i𝑝𝑝# ∙ 𝜇𝜇#Li ⋅ g1 − i𝑝𝑝^ h𝑑𝑑𝑑𝑑
0
7
m
7𝑞𝑞#^ =\ i𝑝𝑝^ ∙ 𝜇𝜇^Li ⋅ g1 − i𝑝𝑝# h𝑑𝑑𝑑𝑑
0

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 8

LONG-TERM LONG-TERM PROFIT TESTS PROFIT TESTS Gain by Source
INSURANCE COVERAGE
INSURANCE COVERAGE Gain in the order of expenses, interest, and
Profits for Traditional Products mortality ( ′ = actual):
Disability Income Insurance (DII) The profit per policy in force at time t is Expense: g𝑒𝑒7 − 𝑒𝑒7÷ h(1 + 𝑖𝑖7 )
Continuous Sojourn Annuity Pr7 = g7fv𝑉𝑉 + 𝑃𝑃7 − 𝐸𝐸7 h(1 + 𝑖𝑖) − 𝑞𝑞#L7fv 𝐷𝐷𝐵𝐵7 +g𝐸𝐸7 − 𝐸𝐸7÷ h𝑞𝑞#L7fv
The EPV of an n-year continuous sojourn − 𝑝𝑝#L7fv 7𝑉𝑉 Interest: g𝑖𝑖7÷ − 𝑖𝑖7 hg7fv𝑉𝑉 + 𝐺𝐺7 − 𝑒𝑒7÷ h
annuity on (x) in state i that pays $1 per

Change in Reserve Mortality: g𝑞𝑞#L7fv − 𝑞𝑞#L7fv

÷
h ×
year continuously while the life remains in
state i is: Δ 7𝑉𝑉 = 7fv𝑉𝑉 (1 + 𝑖𝑖) − 𝑝𝑝#L7fv 7𝑉𝑉
g𝐷𝐷𝐷𝐷7 + 𝐸𝐸7÷ − 7𝑉𝑉h

X Profit Vector
ªª
𝑎𝑎ê#:X| =\ ªª fù7
d𝑡𝑡 Actual Profit
êêê 7𝑝𝑝# 𝑒𝑒
0 Pr = (Pr0 Prv Prm … PrX ) Using the actual experience:

EPV of benefit of an n-year DII: Profit Signature Actual Profit = g 7fv𝑉𝑉 + 𝐺𝐺7 − 𝑒𝑒7÷ h(1 + 𝑖𝑖7÷ )
÷ (𝐷𝐷𝐷𝐷7 + 𝐸𝐸7÷ )

Profit per policy issued, Π −𝑞𝑞#L7fv
X
0v 00 0v vv = Pr7 ÷
𝑎𝑎ê#:X|
êêê = \ 7𝑝𝑝# 𝜇𝜇#L7 𝑎𝑎
ê#L7:Xf7|
êêêêêêê 𝑒𝑒
fù7
d𝑡𝑡 −𝑝𝑝#L7fv ⋅ 7𝑉𝑉
0 in force at time Expected Profit
⋅ Prob ° Ô in force at time 0¢
𝑡𝑡 − 1 Using the assumed experience:
With waiting period of w years, the EPV is:
Profit signature: (Π0 Πv Πm … ΠX )

Expected Profit = g 7fv𝑉𝑉 + 𝐺𝐺7 − 𝑒𝑒7 h(1 + 𝑖𝑖7 )
XfÎ
vv vv
where
\ 00 0v
7𝑝𝑝# 𝜇𝜇#L7 r𝑎𝑎
ê#L7:Xf7|
êêêêêêê − 𝑎𝑎
ê#L7:Î|
êêêê s 𝑒𝑒
fù7
d𝑡𝑡 −𝑞𝑞#L7fv (𝐷𝐷𝐷𝐷7 + 𝐸𝐸7 )
0 Π0 = Pr0

Π7 = Pr7 ⋅ 7fv𝑝𝑝# , 𝑡𝑡 = 1, 2, 3, … , 𝑛𝑛 −𝑝𝑝#L7fv ⋅ 7𝑉𝑉
With waiting period of w years and benefit
term of m years, the EPV is: Profit Measures Total Gain

Xf(zLÎ)
NPV Total Gain = Actual Profit − Expected Profit
00 0v vv vv fù7 b
\ 7𝑝𝑝# 𝜇𝜇#L7 r𝑎𝑎ê#L7:zLÎ|
êêêêêêêêê − 𝑎𝑎
ê#L7:Î|
êêêê s 𝑒𝑒 d𝑡𝑡 = Gain from Expenses
0 NPV = t Π… ⋅ 𝑣𝑣£…
XfÎ
vv vv
+ Gain from Interest
00 0v
+\ 7𝑝𝑝# 𝜇𝜇#L7 r𝑎𝑎ê#L7:Xf7|
êêêêêêê − 𝑎𝑎
ê#L7:Î|
êêêê s 𝑒𝑒
fù7
d𝑡𝑡 …u0
+ Gain from Mortalities
Xf(zLÎ) where 𝑟𝑟 = risk discount or hurdle rate


Reserve Recursion for Policies with Partial NPV
R
Multiple States
NPV(𝑘𝑘) = t Π… ⋅ 𝑣𝑣£… PENSION MATHEMATICS
PENSION MATHEMATICS
Assuming there are 𝑚𝑚 + 1 states and cash
…u0
flows are made every h years:
IRR Valuation of Benefits
(…)
r 7𝑉𝑉 (…) + ℎ𝑃𝑃7 s (1 + 𝑖𝑖)ò b
Motivations
z NPV = t Π… ⋅ 𝑣𝑣£… = 0 1. Attract potential employees
…R …R (R)
= t ò𝑝𝑝#L7 rℎ𝐵𝐵7Lò + 𝑏𝑏7Lò + 7Lò𝑉𝑉 (R) s …u0
2. Provide incentive for employees to stay
Ru0 DPP
3. Facilitate turnover of older employees
DPP = min[𝑡𝑡: NPV(𝑡𝑡) > 0]
If lump sum benefit is assumed to be paid in
4. Provide tax-friendly compensation
the middle of an interval: Profit Margin 5. Pressure from trade unions

NPV 6. Reward employees who have contributed
(…)
g 7𝑉𝑉 (…) + ℎ𝑃𝑃7 h(1 + 𝑖𝑖)ò Profit margin =
𝐸𝐸𝐸𝐸𝐸𝐸(f. premiums) to the company’s success
z
…R (R) …R
= t ò𝑝𝑝#L7 gℎ𝐵𝐵7Lò + 𝑏𝑏7Lò (1 + 𝑖𝑖)ò/m + 7Lò𝑉𝑉 (R) h

Replacement Ratio
Ru0 Zeroized Reserves


1. Begin with the last year and work pension income in the year after retirement
Activities of Daily Living (ADLs): 𝑅𝑅 =
backwards salary in the year before retirement
• Bathing
2. Set the profit for the year to zero then
• Dressing
solve for the beginning-of-year reserve
• Eating
3. If the reserve is negative, set to zero
• Toileting
and repeat this entire process again
• Continence
until time 0
• Transferring

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 9

Salary Projection Funding the Benefits Actuarial Liability at time t, tV
S: Salary 𝑆𝑆̅: Rate of salary Actuarial Liability Assuming linear accrual to the earliest
s: Salary scale 𝑠𝑠̅: Rate of salary function 7𝑉𝑉 = 𝐸𝐸𝐸𝐸𝐸𝐸(all accrued benefits at time 𝑡𝑡) possible retirement age (pro-rata accruals
• Constant percentage of increase
method):
Normal Contribution 1
𝑆𝑆^L7 = 𝑆𝑆^ (1 + 𝑥𝑥%)7
7𝑉𝑉 = ⋅ 𝐴𝐴𝐴𝐴𝐴𝐴𝐻𝐻𝐵𝐵7
• Salary Scale 𝑡𝑡 + 𝑘𝑘
7𝑉𝑉 + 𝐶𝐶7 =
𝑠𝑠^
𝑆𝑆^ = 𝑆𝑆# ⋅ EPV(benefits for mid-year exits)+𝑣𝑣𝑝𝑝#00 7Lv𝑉𝑉 Assuming linear accrual to each possible
𝑠𝑠#
retirement age:
• Rate of salary • Two methods of funding benefits: PUC
𝑠𝑠̅^ 7𝑉𝑉 =
𝑆𝑆^̅ = 𝑆𝑆#̅ ⋅ and TUC 4óf#
𝑠𝑠̅# 𝑡𝑡 𝑟𝑟#LR R
• If there are no mid-year exits: t ∙ 𝑣𝑣 𝐵𝐵(𝑥𝑥 + 𝑘𝑘, 𝑡𝑡 + 𝑘𝑘) ∙
v
𝑆𝑆̅ v
v 𝑡𝑡 + 𝑘𝑘 𝑙𝑙#
̅ 𝑑𝑑𝑑𝑑 = # ⋅ \ 𝑠𝑠̅^L7 𝑑𝑑𝑑𝑑
𝑆𝑆^ = \ 𝑆𝑆^L7 § PUC: 𝐶𝐶7 = ⋅ 7𝑉𝑉 Ru0
0 𝑠𝑠̅# 0 7
𝑎𝑎̈ 1 (𝑥𝑥 + 𝑘𝑘, 𝑡𝑡 + 𝑘𝑘)
7Lv .¥
• Relationships § TUC: 𝐶𝐶7 = r ⋅. − 1s 7𝑉𝑉
7 ¥/0
Rate of salary function to salary scale: Normal Cost

v
Retiree Health Benefits 7𝑉𝑉 + 𝐶𝐶7 = EPV(benefits for mid-year
𝑠𝑠^ = \ 𝑠𝑠̅^L7 d𝑡𝑡 exits)+𝑣𝑣𝑝𝑝#00 7Lv𝑉𝑉
0 Benefit Premium Annuity for age x at time t

Salary scale to rate of salary function: b
𝐵𝐵(𝑥𝑥 + 𝑘𝑘, 𝑡𝑡 + 𝑘𝑘) If there are no mid-year exits:
𝑠𝑠̅^ ≈ 𝑠𝑠^f0.ó 𝑎𝑎̈ 1 (𝑥𝑥, 𝑡𝑡) = t 𝑣𝑣 R R𝑝𝑝# [ ]
𝐵𝐵(𝑥𝑥, 𝑡𝑡)
Ru0 7𝑉𝑉

𝐶𝐶7 =
Defined Contribution Pension Plans

𝑡𝑡
Value of retiree health benefit at retirement
𝐴𝐴𝐴𝐴(pension fund) = 𝐸𝐸𝐸𝐸𝐸𝐸(pension benefits) 1

for a life retiring at age xr in t years: 𝐶𝐶7 = ⋅ 𝐴𝐴𝐴𝐴𝐴𝐴𝐻𝐻𝐵𝐵7
𝐵𝐵(𝑥𝑥𝑥𝑥, 𝑡𝑡)𝑎𝑎̈ 1 (𝑥𝑥𝑥𝑥, 𝑡𝑡) 𝑡𝑡 + 𝑘𝑘
Defined Benefit Pension Plans
(for linear accrual to the earliest possible
Annual Retirement Benefit = 𝑛𝑛 ⋅ 𝑆𝑆˝˛ˇ ⋅ 𝛼𝛼
When healthcare premiums increase retirement age)
where
exponentially with age and at a constant
𝑛𝑛 = total number of years of service
annual inflation rate where:
𝑆𝑆˝˛ˇ = final average salary
• 𝑐𝑐 = 𝐵𝐵(𝑥𝑥 + 1, 𝑡𝑡)⁄𝐵𝐵(𝑥𝑥, 𝑡𝑡) SURVIVAL MODEL ESTIMATION
SURVIVAL MODEL ESTIMATION
𝛼𝛼 = accrual rate
• 𝑗𝑗 = annual rate of inflation for healthcare
• Two methods to calculate the amount of
costs Kaplan-Meier and Nelson-Aalen
retirement benefit:
Estimators
§ PUC: projects salary to retirement or 𝐵𝐵(𝑥𝑥𝑥𝑥 + 𝑘𝑘, 𝑡𝑡 + 𝑘𝑘) = 𝑐𝑐 R (1 + 𝑗𝑗)R 𝐵𝐵(𝑥𝑥𝑥𝑥, 𝑡𝑡)
Empirical Distribution
exit date 𝑎𝑎̈ 1 (𝑥𝑥𝑥𝑥, 𝑡𝑡) = 𝑎𝑎̈ #£|ª ∗
# of data points = 𝑥𝑥
§ TUC: calculates salary based on 1 + 𝑖𝑖 Pr(𝑋𝑋 = 𝑥𝑥) =
where 𝑖𝑖 ∗ = − 1 𝑛𝑛
employee’s current age 𝑐𝑐(1 + 𝑗𝑗) # of data points ≤ 𝑥𝑥
𝐹𝐹X (𝑥𝑥) = Pr(𝑋𝑋 ≤ 𝑥𝑥) =
• Early retirement 𝑛𝑛
Actuarial Value of Total Health Benefit
Annual Retirement Benefit 7 [𝑆𝑆X (𝑥𝑥)] = Var
Var 7 [𝐹𝐹X (𝑥𝑥)]
(AVTHB)
= 𝑛𝑛 ⋅ 𝑆𝑆˝˛ˇ ⋅ 𝛼𝛼 𝐹𝐹X (𝑥𝑥) ⋅ [1 − 𝐹𝐹X (𝑥𝑥)]
𝐴𝐴𝐴𝐴𝐴𝐴𝐻𝐻𝐵𝐵7 =
⋅ (1 − pension reduction factor) 𝑛𝑛
4óf#
• Withdrawal without COLA 𝑟𝑟#LR R
= t 𝑣𝑣 𝐵𝐵(𝑥𝑥 + 𝑘𝑘, 𝑡𝑡 + 𝑘𝑘) ∙ Kaplan-Meier Estimator
Annual Retirement Benefit = 𝑛𝑛 ⋅ 𝑆𝑆˝˛ˇ ⋅ 𝛼𝛼 𝑙𝑙#
Ru0 …
• Withdrawal with COLA 𝑎𝑎̈ 1 (𝑥𝑥 + 𝑘𝑘, 𝑡𝑡 + 𝑘𝑘) 𝑠𝑠ª
𝑆𝑆8 g𝑦𝑦… h = ◊ µ1 − ∂
Annual Retirement Benefit 4óf# 𝑟𝑟ª
𝑟𝑟#LR R ªuv
= 𝑛𝑛 ⋅ 𝑆𝑆˝˛ˇ ⋅ 𝛼𝛼 = 𝐵𝐵(𝑥𝑥, 𝑡𝑡) t 𝑣𝑣 ∗ 𝑎𝑎̈ ∗
𝑆𝑆8 g𝑦𝑦… h
𝑙𝑙# ª #LR|ª ℎ9 g𝑦𝑦… h = 1 −
Ru0
⋅ (1 + COLA)#$%˛#$&$ˇ% '($f)˛%*+#')', '($ 𝑆𝑆8 g𝑦𝑦…fv h
Tail Correction
• Efron’s tail correction:
𝑆𝑆8 (𝑡𝑡) = 0 for 𝑡𝑡 > 𝑢𝑢
• Klein and Moeschberger's tail correction:
𝑆𝑆8(𝑢𝑢), for 𝑢𝑢 < 𝑡𝑡 < 𝛾𝛾
𝑆𝑆8 (𝑡𝑡) = :
0, for 𝑡𝑡 ≥ 𝛾𝛾
• Brown, Hollander, and Korwar’s tail
correction:
𝑆𝑆8 (𝑡𝑡) = 𝑆𝑆8 (𝑢𝑢)7/; for 𝑡𝑡 > 𝑢𝑢

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 10

Nelson-Aalen Estimator Age-Based Estimators MORTALITY IMPROVEMENT
MORTALITY IMPROVEMENT
… Individual Data
𝑠𝑠
<g𝑦𝑦… h = t ª
𝐻𝐻 • Exact Exposure: Single-Factor Mortality Improvement
𝑟𝑟ª
ªuv
𝑞𝑞G… = 1 − 𝑒𝑒 fH‡/√‡ Scales
< g^‡ h
𝑆𝑆8 g𝑦𝑦… h = 𝑒𝑒 f=
𝑞𝑞(𝑥𝑥, 𝑡𝑡) = 𝑞𝑞(𝑥𝑥, 0)(1 − 𝜑𝜑# )7
• Actuarial Exposure:

𝑑𝑑…
Variance of Estimators 𝑞𝑞G… =
𝑒𝑒… Two-Factor Mortality Improvement
• Greenwood’s Approximation:
… Note that exposures for deaths are Scales
m 𝑠𝑠ª Cubic Spline
7 ®𝑆𝑆8g𝑦𝑦… h© = ®𝑆𝑆8 g𝑦𝑦… h© t
Var counted until the end of the age interval
𝑟𝑟ª (𝑟𝑟ª − 𝑠𝑠ª ) A cubic function that joins the short-term
ªuv
Used for Kaplan-Meier Assumption Description factors and long-term factors in a smooth

Advance birthday fashion
• Klein’s Estimation: Age Last Birthday
… to policy date Age-Based Cubic Spline
𝑠𝑠ª (𝑟𝑟ª − 𝑠𝑠ª )
7 ®𝐻𝐻
Var <g𝑦𝑦… h© = t Advance/retreat 𝐶𝐶L (𝑥𝑥, 𝑡𝑡) = 𝑎𝑎𝑡𝑡 à + 𝑏𝑏𝑡𝑡 m + 𝑐𝑐𝑐𝑐 + 𝑑𝑑
𝑟𝑟ªà Age Nearest
ªuv
birthday to nearest
… Birthday Cohort-Based Cubic Spline
𝑠𝑠ª (𝑟𝑟ª − 𝑠𝑠ª ) policy date
7 ®𝑆𝑆8g𝑦𝑦… h© = ®𝑆𝑆8 g𝑦𝑦… h©m t
Var 𝐶𝐶M (𝑥𝑥, 𝑡𝑡) = 𝑎𝑎∗ 𝑡𝑡 à + 𝑏𝑏∗ 𝑡𝑡 m + 𝑐𝑐 ∗ 𝑡𝑡 + 𝑑𝑑∗
𝑟𝑟ªà Study starts and
ªuv Anniversary-to-
Used for Nelson-Aalen ends on policy Solve for the constants in the age-based and
Anniversary
anniversary cohort-based cubic splines using these 4
In general, constraints:
7 ®𝑆𝑆8g𝑦𝑦… h© = ®𝑆𝑆8g𝑦𝑦… h©m ⋅ Var
Var 7 ®𝐻𝐻
< g𝑦𝑦… h© Variance of Estimators 1. The starting value of the spline matches

• Exact Exposure: the improvement factor at 2007.
Confidence Interval
𝑑𝑑… 2. The starting derivative of the spline
Linear Confidence Interval for 𝑆𝑆(𝑡𝑡): 7 ®𝑞𝑞G… © = g1 − 𝑞𝑞G… hm ⋅
Var
𝑒𝑒…m matches the slope of the improvement
7 ®𝑆𝑆8 (𝑡𝑡)©
𝑆𝑆8 (𝑡𝑡) ± 𝑧𝑧(vLß)/m ?Var • Actuarial Exposure: function at 2007.
𝑞𝑞G… g1 − 𝑞𝑞G… h 3. The ending value of the spline matches
Log-transformed Confidence Interval: 7 ®𝑞𝑞G… © =
Var
𝑒𝑒… the improvement factor at 2027.
• For 𝑆𝑆(𝑡𝑡):
4. The ending derivative of the spline
g𝑆𝑆8(𝑡𝑡)v/@A , 𝑆𝑆8 (𝑡𝑡)@A h where
Interval-based Data matches the slope of the improvement
7 ®𝑆𝑆8(𝑡𝑡)©
𝑧𝑧(vLß)/m ?𝑉𝑉𝑉𝑉𝑉𝑉 Exposures without uniform assumption function at 2027.
𝑈𝑈. = exp ⎛ ⎞ 𝑒𝑒… = 𝑃𝑃… + 𝑛𝑛…
𝑆𝑆8(𝑡𝑡) ⋅ ln 𝑆𝑆8(𝑡𝑡) Improvement factor for age x in year t:
⎝ ⎠ Exposures with uniform assumption
𝜑𝜑(𝑥𝑥, 2007 + 𝑡𝑡) = 0.5𝐶𝐶L (𝑥𝑥, 𝑡𝑡) + 0.5𝐶𝐶M (𝑥𝑥, 𝑡𝑡)

• Exact Exposure:
• For 𝐻𝐻(𝑡𝑡):
𝑒𝑒… = 𝑃𝑃… + 0.5g𝑛𝑛… − 𝑤𝑤… − 𝑑𝑑… h where t is the number of years measured
< (𝑡𝑡)
𝐻𝐻
[ <(𝑡𝑡) ⋅ 𝑈𝑈= ] where
, 𝐻𝐻 • Actuarial Exposure: from 2007
𝑈𝑈=
𝑒𝑒… = 𝑃𝑃… + 0.5g𝑛𝑛… − 𝑤𝑤… h
Mortality Rate with Mortality Improvement
7 ®𝐻𝐻
𝑧𝑧(vLß)/m ?𝑉𝑉𝑉𝑉𝑉𝑉 < (𝑡𝑡)©
7
𝑈𝑈= = exp ⎛ ⎞ Multiple State Estimation
< (𝑡𝑡)
𝐻𝐻 𝑞𝑞(𝑥𝑥, 𝑡𝑡) = 𝑞𝑞(𝑥𝑥, 0) ◊g1 − 𝜑𝜑(𝑥𝑥, 𝑘𝑘)h
Transition intensity:
⎝ ⎠ Ruv
𝑑𝑑ª…
𝜇𝜇̂ ª… =
1 𝑇𝑇ª
𝑈𝑈. =
𝑈𝑈= 𝑑𝑑
7 g𝜇𝜇̂ ª… h = ª…m
Var
𝑇𝑇ª

www.coachingactuaries.com Copyright © 2019 Coaching Actuaries. All Rights Reserved. 11

The Lee Carter Model The Cairns-Blake-Dowd (CBD) Models
Central Death Rate Logit Function
v 𝑥𝑥
𝑞𝑞# ∫0 7𝑝𝑝# 𝜇𝜇#L7 d𝑡𝑡 logit(𝑥𝑥) = ln
𝑚𝑚# = v = v 1 − 𝑥𝑥
∫0 7𝑝𝑝# d𝑡𝑡 ∫0 7𝑝𝑝# d𝑡𝑡 𝑞𝑞(𝑥𝑥, 𝑡𝑡)
𝑙𝑙𝑙𝑙(𝑥𝑥, 𝑡𝑡) = logit[𝑞𝑞(𝑥𝑥, 𝑡𝑡)] = ln

1 − 𝑞𝑞(𝑥𝑥, 𝑡𝑡)
• Assuming constant force of mortality
𝑒𝑒 Yfl(#,7)
between integer ages: ⇒ 𝑞𝑞(𝑥𝑥, 𝑡𝑡) =
1 + 𝑒𝑒 Yfl(#,7)
𝑚𝑚# = 𝜇𝜇

𝑞𝑞# = 1 − 𝑒𝑒 fz¥ The Original CBD Model

(v) (m)
𝑙𝑙𝑙𝑙(𝑥𝑥, 𝑡𝑡) = 𝐾𝐾7 + 𝐾𝐾7 (𝑥𝑥 − 𝑥𝑥)
• Assuming UDD between integer ages:
𝑞𝑞#
𝑚𝑚# = where:
1
1 − 2 𝑞𝑞# • 𝑥𝑥 is the average age in the data set
𝑚𝑚# • 𝐾𝐾7
(v) (v) (v)
= 𝐾𝐾7fv + 𝑐𝑐 (v) + 𝜎𝜎R0 𝑍𝑍7
𝑞𝑞# =
1
1 + 2 𝑚𝑚# (m)
𝐾𝐾7
(m) (m)
= 𝐾𝐾7fv + 𝑐𝑐 (m) + 𝜎𝜎RS 𝑍𝑍7

(v) (m)
Normal and Lognormal Random Variables • 𝐸𝐸 °𝑍𝑍7 𝑍𝑍7 ¢ = 𝜌𝜌, − 1 ≤ 𝜌𝜌 ≤ 1
𝑋𝑋~𝑁𝑁(𝜇𝜇, 𝜎𝜎 m ) represents a normally (ª) (…)
𝐸𝐸 °𝑍𝑍7 𝑍𝑍; ¢ = 0 for 𝑡𝑡 ≠ 𝑢𝑢, 𝑖𝑖 = 1,2, 𝑗𝑗 = 1,2
distributed random variable, X, with mean 𝜇𝜇

and variance 𝜎𝜎 m . Advantages of the original CBD model over

the Lee Carter model:
𝑍𝑍~𝑁𝑁(0,1) represents a standard normal
• Fewer parameters
random variable with a mean of 0 and a
• Less parameter uncertainty
variance of 1.

Disadvantage of the original CBD model
𝑌𝑌 = 𝑒𝑒 Q ~logN(𝜇𝜇, 𝜎𝜎) is a lognormal random
over the Lee Carter model:
variable with parameters 𝜇𝜇 and 𝜎𝜎.
RS
• Fit to population is sometimes worse
• 𝐸𝐸[𝑌𝑌] = 𝑒𝑒 {L S

TS The CBD M7 Model

• 𝑉𝑉𝑉𝑉𝑉𝑉[𝑌𝑌] = (𝐸𝐸[𝑌𝑌])m g𝑒𝑒 − 1h (v) (m)
𝑙𝑙𝑙𝑙(𝑥𝑥, 𝑡𝑡) = 𝐾𝐾7 + 𝐾𝐾7 (𝑥𝑥 − 𝑥𝑥)
The Lee Carter Model (à)
+𝐾𝐾7 ((𝑥𝑥 − 𝑥𝑥)m − 𝑠𝑠#m ) + 𝐺𝐺7f#
ln 𝑚𝑚(𝑥𝑥, 𝑡𝑡) = 𝑙𝑙𝑙𝑙(𝑥𝑥, 𝑡𝑡) = 𝛼𝛼# + 𝛽𝛽# 𝐾𝐾7 + 𝜖𝜖#,7
where:
where: ∑¥¥[¥
V (#f#)S
• 𝑠𝑠#m = W

• 𝐾𝐾7 = 𝐾𝐾7fv + 𝑐𝑐 + 𝜎𝜎R 𝑍𝑍7 #V f#W Lv

• 𝜖𝜖#,7 is sufficiently small to be negligible • 𝐺𝐺7f# introduces a cohort time series

Constraints used to fix identifiability Advantages of the CBD M7 model:

problem: • Includes a cohort effect
#V 7X • Includes a quadratic age difference term
t 𝛽𝛽# = 1.0 and t 𝐾𝐾7 = 0.0
#u#W 7u7W

With these constraints:

∑77u7
X
W
𝑙𝑙𝑙𝑙(𝑥𝑥, 𝑡𝑡)
𝛼𝛼# =
𝑡𝑡X − 𝑡𝑡0 + 1

Central death rate improvement factor:

𝑚𝑚(𝑥𝑥, 𝑡𝑡)
𝜑𝜑z (𝑥𝑥, 𝑡𝑡) = 1 −
𝑚𝑚(𝑥𝑥, 𝑡𝑡 − 1)

In the Lee Carter model:

1 − 𝜑𝜑z (𝑥𝑥, 𝑡𝑡)~logN(𝜇𝜇 = 𝛽𝛽# 𝑐𝑐, 𝜎𝜎 = 𝛽𝛽# 𝜎𝜎R )

Copyright © 2019 Coaching Actuaries. All Rights Reserved. 12

www.coachingactuaries.com Personal copies permitted. Resale or distribution is prohibited.