Exam LTAM

You have what it takes to pass updated 04/14/21

SURVIVAL DISTRIBUTIONS
SURVIVAL DISTRIBUTIONS Force of Mortality Moments
d d Complete Expectation
f! (t) dt #p! dt
l!&#
Probability Functions µ!&# = =− =− • First Moment
S! (t) #p!
l!&# -
Survival Function ∘
f! (t) = S! (t) ⋅ µ!&# = #p! ⋅ µ!&# e! = E[T! ] = P t ⋅ #p! µ!&# dt
T! : future lifetime or time-to-death of (x) "

-
S! (t): Probability that (x) survives t years Finding (p! Using Force of Mortality =P #p! dt
= Pr[T! > t] ( "
= Pr[T" > x + t|T" > x] (p! = exp O− P µ!&# dtQ • Second Moment
" -
S" (x + t) !&(
= E[T!* ] = P t * ⋅ #p! µ!&# dt
S" (x) = exp O− P µ, dyQ "
! -
S! (t) must satisfy:
= P 2t ⋅ #p! dt
• S! (0) = 1 Properties of Force of Mortality "
• S! (∞) = 0 • µ!&# ≥ 0 • Variance
- *
• S! (t) is a non-increasing function of t • ∫" µ!&# dt = ∞ ∘
Var[T! ] = E[T!* ] − Ze! [


Actuarial Notations Adding/Multiplying a Constant Curtate Expectation
#p! : Probability that (x) survives t years
• First Moment
• µ∗!&# = µ!&# + k ⇒ ∗
(p! = (p! ∙ e +'(
- -
= Pr(T! > t) '
• µ∗!&# = k ∙ µ!&# ⇒ ∗
(p! = U (p! V e! = E[K ! ] = \ k ⋅ '|q! = \ 'p!
= S! (t)
'1" '1)
#q! : Probability that (x) dies within t years • Second Moment
Extra Mortality Risk
= Pr(T! ≤ t) - -
• Age rating: Adding additional years to a
= F! (t) E[K *! ] = \ k * ⋅ '|q! = \(2k − 1) 'p!
person’s age, effectively treating the '1" '1)
#p! + #q! = 1
person as a different aged risk, e.g., 5-year • Variance
$|#q ! : Probability that (x) survives u years age rating to (50) means treating (50) as Var[K ! ] = E[K *! ] − (e! )*
and dies within the following t years if he/she is 55 years old.

= $ p! ⋅ # q!&$ • Adjusting force of mortality: Adding a Temporary Expectation

(
= $ p! − $&# p! constant to the force of mortality. ∘
e!:(| = P t ⋅ #p! µ!&# dt + n (p!
= $&# q! − $ q! • Adjusting mortality rates: Multiplying "
(

Curtate Future Lifetime the mortality rate for standard lives by a =P #p! dt
"
K ! : number of completed future years by constant. (+) (

(x) prior to death Express p’s or q’s in terms of µ e!:(| = \ k ⋅ '|q! + n ⋅ (p! = \ 'p!
K ! = ⌊T! ⌋ - '1" '1)
.
#p! =P /p! ⋅ µ!&/ ds
Pr[K ! = k] = ' p! ∙ q!&' = '|q! ∘
# Relationship between e! and e!
#
(q ! = )q! + )p! ⋅ )q!&) + *p! ⋅ )q!&* ∘
#q ! =P .
/p! ⋅ µ!&/ ds e! ≈ e! + 0.5
+ ⋯ + (+)p! ⋅ )q!&(+) "
$&# Recursive Formulas
= "|q! + )|q! + ⋯ + (+)|q! .
$|#q ! =P /p! ⋅ µ!&/ ds ∘ ∘ ∘
Life Table $ e! = e!:(| + (p! ⋅ e!&(
l!&# e! = e!:(| + (p! ⋅ e!&(
#p! =
l! ∘ ∘ ∘
e!:3&(| = e!:3| + 3p! ⋅ e!&3:(|
#d! l! − l!&#
#q ! = = e!:3&(| = e!:3| + 3p! ⋅ e!&3:(|
l! l!
e! = p! (1 + e!&) )
#d!&$ l!&$ − l!&$&#
$|#q ! = =
l! l!

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 1

Special Mortality Laws Fractional Ages INSURANCE INSURANCE
Constant Force of Mortality UDD (0 ≤ t ≤ 1)
µ! = µ Use linear interpolation: In general, to calculate the EPV of insurance
= e+4#
#p! l!&# = (1 − t) ⋅ l! + t ⋅ l!&) benefits, sum/integrate the product of the
∘ 1 = t ⋅ q!
#q ! following three components over all
e! =
µ tq! possible payment times:
#q !&/ = for 0 ≤ s + t ≤ 1
∘ 1 1 − s ⋅ q! 1. the benefit amount
e!:(| = (1 − e+4⋅( ) q!
µ 2. the appropriate discount factor
µ!&# =
1 − t ⋅ q! 3. the probability that the benefit will be
Uniform Distribution
f! (t) = #p! ∙ µ!&# = q! (a constant) paid at that time
The following formulas are good to know but ∘
e! = e! + 0.5
are rarely tested on current exams. ∘ ∘
e! = 0.5q! + p! Z1 + e!&) [ Present Value
l! = k(ω − x)
1 1
Type of
µ! = ⇒ µ!&# = Constant Force of Mortality (0 ≤ t ≤ 1) 𝐏𝐏𝐏𝐏
ω−x ω − (x + t) Insurance
l!&# ω − (x + t) Use exponential interpolation:
#p! = = l!&# = (l! ))+# ⋅ (l!&) )# Discrete
l! ω−x
#
#p! = (p! )

l! − l!&# t µ!&# = − ln(p! ) Z = v :!&) , K ! ≥ 0
#q ! = = Whole Life
l! ω−x
f! (t) = #p! ∙ µ!&# = e+4⋅# ⋅ µ of $1
l!&$ − l!&$&# t Continuous
$|#q ! = =
l! ω−x
∘ ω−x Select & Ultimate Mortality Z = v ;! , T! ≥ 0
e! = The age at which a person is selected is
2
∘ n denoted as [x]. Discrete
e!:(| = (p! (n) + (q! Z [
2
v :!&) , K ! < n

Select mortality is written as q[!]&# where x n-year Z=n
Beta Distribution 0 , K ! ≥ n
is the selected age and t is the number of Term Life
The following formulas are good to know but of $1 Continuous
years after selection.
are rarely tested on current exams. v ;! , T! ≤ n
Z=n
l! = k(ω − x)6 The mortality after the select period is 0 , T! > n
α α called the ultimate mortality, where:
µ! = ⇒ µ!&# = Discrete
ω−x ω − (x + t) q[!]&# = q!&#
6 n-year 0 , K ! < n
l!&# ω − (x + t) Z = n :!&)
#p! = =b c Deferred v , K! ≥ n
l! ω−x Common Approach
Whole Life Continuous
∘ ω−x Read from the left to the right and then
e! = of $1
α+1 continue downwards: 0 , T! ≤ n
Z = n ;!
q[!] q[!]&) q[!]&* v , T! > n
Gompertz’s Law x q!&9
µ! = Bc ! c > 1, B > 0 30 n-year
Bc ! # 31
#p! = exp g− (c − 1)h Pure 0 , T! ≤ n
ln c Z=n
32 Endowment v ( , T! > n
Makeham’s Law 33 of $1
µ! = A + Bc ! c > 1, B > 0, A ≥ −B Read in a diagonal direction until we reach
Bc ! # the ultimate column, and then proceed Discrete
#p! = exp g− (c − 1)h ⋅ exp(−At)
ln c
downwards: v :!&) , K ! < n
Z=n
x q[!] q[!+)]&) q[!+*]&* q! n-year v ( , K ! ≥ n
Endowment = v 3<((:!&),()
30
Insurance Continuous
31 of $1
v ;! , T ≤ n
32 Z=n ( !
v , T! > n
33
= v 3<((;!,()

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 2

Expected Present Value (+)
Variances
(DA) ) B = \(n − k)v
!:(|
'&)
⋅ '|q!
Type of EPV / APV / Net Single Type of
'1"
Variance
Insurance Premium ( Insurance
pA
(D p ) ) #
B = P (n − t)v ⋅ #p! µ!&# dt
!:(|
"
Discrete Discrete
(IA) ) )
B + (DA) !:(|
!:(|
)
B = (n + 1)A!:(|
-
̅p) ) B + (D
pAp) ) p ) *
A! = \ v '&) ⋅ '|q! (IA !:(| B =n⋅A
!:(| B
!:(|
A! − (A! )*
Whole Life Whole Life
'1"
of $1 of $1 Continuous
Continuous SULT
-
*p p ! )*
p ! = P v # ⋅ #p! µ!&# dt
A A! − (A
#E! can be calculated as:
"
C
• #E! = v # ⋅ #p! = v # ⋅ !"# Discrete
Discrete C !

(+) • #&DE! = #E! ⋅ DE!&# *

* )
A)!:(| = \ v '&) ⋅ '|q! n-year A!:(| − ZA)!:(| [
*
'1" #E! can be calculated as: Term Life
n-year
= A! − (E! ⋅ A!&( • *#E! = (v # )* ⋅ #p! = (v # )* ⋅
C!"#
of $1 Continuous
Term Life C!
Continuous *
of $1 • *#E! = v # ⋅ v # #p! = v # ⋅ #E! *p)
A!:(| p) [
− ZA !:(|
(
p ) = P v # #p! µ!&# dt
A A)!:(| can be calculated as:
!∶(|
" Discrete
p ! − (E! ⋅ A
=A p !&( • A)!:(| = A
!:(|
− (E! n-year *
*
Deferred (|A! − U(|A! V
Discrete • A)!:(| = A! − (E! ⋅ A!&(
Whole Life Continuous
(|A! = A! − A)!:(| of $1
n-year *
Shortcuts *p
(|A!
p! V
− U(|A
Deferred = (E! ⋅ A!&(
The shortcuts for constant force can be
Whole Life Continuous n-year
useful on current exams. The shortcuts for *
of $1 Pure *
(E! − U (E! V
p =A
(|A!
p! − A
p ) uniform distribution are good to know but
!∶(| Endowment
are rarely tested on current exams. = v *( (p! (q!
p !&(
= (E! ⋅ A of $1
Uniform
n-year Constant Force Discrete
Distribution
n-year
Pure *
A!:(|
B − UA!:(|
BV
*
A ) = (E! = v ( (p! Endowment
Endowment !:(| µ auE+!|
p! =
A p! =
A Insurance
of $1 µ+δ ω−x Continuous
of $1
*p *
A!:(| p !:(|
B − UA BV
p )
A
Discrete !:(| au(|
n-year µ p ) =
A

A = A)!:(| + (E! = U1 − (E! V !:(| ω−x p similarly to A and A

Calculate *A and *A p,
Endowment !:(| µ+δ
Insurance but with double the force of interest, δ.
Continuous
of $1 Equivalently, replace v with v * , or
p = A
A p ) + (E! (E!
!:(| !:(| = e+(4&F)( replace i with 2i + i* .
(E! ω − (x + n)
= v( ⋅
Varying Insurance ω−x
Recursive Formulas
-
A! = vq! + vp! ⋅ A!&)
(IA)! = \(k + 1)v '&) ⋅ '|q!
'1" A)!:(| )
B = vq ! + vp! ⋅ A!&):(+)|
GGGGGGG
-
̅p)! = P tv # ⋅ #p! µ!&# dt
(IA (IA)! = vq! + vp! (A!&) + (IA)!&) )

"
(+)

(IA) ) B = \(k + 1)v

!:(|
'&)
⋅ '|q!
'1"
(
̅p) ) B
(IA = P tv # ⋅ #p! µ!&# dt
!:(|
"

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 3

1/mthly Insurance ANNUITIES ANNUITIES Expected Present Value
(3) 1 Type of EPV / APV / Net Single
K ! = ⌊mT! ⌋
m In general, to calculate the EPV of a life Annuities Premium
(3)
Pr wK ! = rx = Hp! ⋅ ) q !&H = ) q! annuity, sum/integrate the product of the Discrete; Due
3 H | 3
following three components over all -
Whole life insurance of $1 payable at the
possible payment times: ä ! = \ v ' ⋅ 'p!
end of the 1/mth year of death: Whole Life
(%) ) 1. the amount paid '1"
(3)
Z = v :! &
3, K! ≥ 0 2. the appropriate discount factor of $1
Continuous
-
(3) 3. the probability that the payment is made -
A! = \ v ('&))/3 ⋅ ' ) q! au! = P v # ⋅ #p! dt
|
3 3 at that time
'1" "
n-year term life insurance of $1 payable at
Discrete; Due
the end of the 1/mth year of death: Present Value
(%) )
(3) 1 Type of (+)
v :! &3 , K ! = 0, … , n − 𝐏𝐏𝐏𝐏 '
m Annuities ä !:(|
B = \ v ⋅ 'p!
Z=y
(3) 1 n-year '1"
0, K ! = n, n + , … Discrete; Due
m Temporary = ä ! − (E! ⋅ ä !&(
3(+)
(%) Y = ä :GGGGGGGGG , K ! ≥ 0 Life
A ) = \ v('&))/3 ⋅ ' ) q! Whole Life ! &)| Continuous
!:(| |
3 3 of $1
'1" of $1
where mn represents the number of periods Continuous (
#
au!:(|
B = P v ⋅ #p! dt

"
UDD Assumption Y = auGGGGG
;! | , T! ≥ 0
= au! − (E! ⋅ au!&(
i
p ! = A!
A Discrete; Due
δ Discrete; Due
i )
p )
A !:(|B = A !:(| B ä :GGGGGGGGG
&)| , K ! < n -
δ Y=Ä !
ä (|B , K ! ≥ n (|ä ! = \ v ' ⋅ 'p!
i n-year
p
(|A! = (|A!
δ = ä GGGGGGGGGGGGGGGGGGG
3K((:! &),()|
'1(
Temporary n-year = ä ! − ä !:(|
B
i )
p !:(|
A B = A !:(|
)
B + A !:(| B Life Continuous
δ Deferred = (E! ⋅ ä !&(
i of $1 Whole Life
(3)
A! = (3) A! auGGGGG
;! | , T! ≤ n Continuous
i Y=Ä of $1
au(|
B , T! > n -
Calculate *A and *A p similarly to A and A
p,
= auGGGGGGGGGGGGGGG (|a
u! = P v # ⋅ #p! dt
but with double the force of interest, δ. 3K((;! ,()| (
Equivalently, replace v with v * , or = au! − au!:(|
B
Discrete; Due
replace i with 2i + i* . For example: = (E! ⋅ au!&(
i* + 2i * 0 , K ! < n
*p
A! = A! Y = nv ( ä Discrete; Due
:! &)+(| , K ! ≥ n
GGGGGGGGGGGGG
2δ n-year
n-year 0 , K ! < n ä GGGGG
= nä B = ä (|
B + (|ä !
:! &)| − ä (|
GGGGGGGGG B , K ! ≥ n Certain- !:(|
Percentiles Deferred
and-Life of Continuous
The 100pth percentile of Z is the value zJ Whole Life Continuous $1
such that: of $1 auGGGGG
B =a
u(|
B + (|a
u!
!:(|
0 , T! ≤ n
Pr|Z ≤ zJ } = p Y = nv ( au
;! +(| , T! > n
GGGGGGGGG
To calculate zJ : 0 , T! ≤ n Annuity Immediate
1. Draw a graph with Z on y-axis and T! = nau − au , T > n
GGGGG
; !|
B
(| ! a! = ä ! − 1
on x-axis. B = ä !:(|
a!:(| B − 1 + (E!
2. Identify the parts of the curve where Discrete; Due
Z ≤ zJ . Determine the value of T! that ä (|
B , K ! < n
n-year Y=Ä
corresponds to those parts. ä GGGGGGGGG
:! &)| , K ! ≥ n
Certain-
3. Use the value of T! from Step 2
and-Life of Continuous
to calculate zJ .
$1
au(|
B , T! ≤ n
Y=Ä
auGGGGG
;! | , T! > n

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 4

Varying Annuities Recursive Formula Woolhouse’s Formula (3 terms)
-
ä ! = 1 + vp! ⋅ ä !&) (3) m − 1 m* − 1
(Iä ) ! = \(k + 1)v ' ⋅ 'p! ä ! ≈ ä ! − − (µ + δ)

2m 12m* !
'1" Relationship between Insurances and • If the question asks to use the
-
(Ia̅u) ! = P tv # ⋅ #p! dt Annuities Woolhouse’s formula with two terms, just
" A! = 1 − dä ! drop the last term.
(+) *
Whole A! = 1 − (2d − d* ) *ä ! • If µ! is not available, approximate µ! as:
(Iä ) !:(| '
B = \(k + 1)v ⋅ 'p!
Life p ! = 1 − δau!
A 1
'1"
*p µ! ≈ − (ln p!+) + ln p! )
( A! = 1 − (2δ) *au! 2
(Ia̅u) !:(|
B = P tv # ⋅ #p! dt (3)
" A!:(|
B = 1 − dä !:(|
B • To find ä !:(|
B , use the relationship below
(+) * * *
Temp. B = 1 − (2d − d ) ä !:(|
A!:(| B and apply Woolhouse’s formula to
(Dä ) !:(| '
B = \(n − k)v ⋅ 'p! (3) (3)
Life p !:(|
A B = 1 − δa
u !:(|
B approximate ä ! and ä !&( :
'1"
( *p * (3) (3) (3)
B = 1 − (2δ) a
A!:(| u !:(|
B ä !:(| = ä ! − (E! ä !&(
p au) B = P (n − t)v # #p! dt
(D B
!:(|
" • To approximate the EPV of a continuous
(Iä ) !:(|
B + (Dä )!:(|
B = (n + 1)ä !:(|
B 1/mthly Annuity life annuity using Woolhouse’s formula,
(Ia̅u) B + (D

!:(|
p au) B = nau!:(|

!:(|B A!
(3)
= 1 − d(3) ä !
(3)
let m → ∞:
(3) (3) 1 1 1
a! = ä ! − au! ≈ ä ! − − (µ! + δ)
Shortcuts m 2 12
(3) (3) 1 • If the interest rate is 0, we have:
The following shortcuts for constant force ä !:(|
B − a!:(|
B = |1 − (E!}
can be useful on current exams. For uniform m ∘ 1 1
e! ≈ e! + − µ!
distribution, just use basic principles. UDD Assumption 2 12
(3)
ä ! = α(m) ⋅ ä ! − β(m)
Constant Force Percentiles
(3)
B = α(m) ⋅ ä !:(|
ä !:(| B − β(m)(1 − (E! ) The 100pth percentile of Y is the value yJ
1
au! = (3)
such that:
µ+δ (|ä ! = α(m) ⋅ (|ä ! − β(m) ⋅ (E!
1 where: Pr|Y ≤ yJ } = p
au!:(|
B = U1 − (E! V (3) To calculate zJ :
µ+δ id i−i
α(m) = β(m) =
i(3) d(3) i(3) d(3) 1. Draw a graph with Y on y-axis and T!
on x-axis.
Variances
Note: 2. Identify the parts of the curve where
Discrete Var[Y]
• The formulas for α(m) and β(m) are Y ≤ yJ . Determine the value of T! that
*
A! − (A! )* provided on the LTAM Table. corresponds to those parts.
Whole Life
d* • The values for α(m) and β(m) when 3. Use the value of T! from Step 2 to
* i = 0.05 are also given in the LTAM Table.
*
A!:(|
B − UA!:(|
BV calculate yJ .
Temporary Life
d*
Interest Conversion
p and d with δ for
Replace A with A 3 +3
i(3) d(3)
continuous cases. O1 + Q = O1 − Q
m m

= (1 + i)
= (1 − d)+)

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 5

PREMIUMS PREMIUMS For an endowment insurance policy of $b Variance
p to A′s
on (x) with level premiums P, add n| For a fully discrete whole life policy:
Fully continuous: Both benefits and and a′s. L
G − eH : &)
"L = Öb + E + Üv !
premiums are payable continuously.
d
The net premium for fully discrete G − eH
Fully discrete: Both benefits and premiums insurance will be represented by P − + (e< − eH )
d
are payable at discrete time points. Unless with the appropriate symbols attached. L
Var| "L }
stated otherwise, discrete time points are The following symbols may be used G − eH * *
the beginning of each year for premiums on the exam: = Öb + E + Ü | A! − (A! )* }
d
and the end of each year for death benefits. A!

P! = where
Semi-continuous: Premiums are paid at ä ! e< = initial expense in year 1
discrete time points and the death benefits A)!:(| eH =renewal expense in year 2+
)
are paid at the moment of death. P!:(| =
ä !:(| b = benefit amount

Special policy: Either non-level benefits or A ) E = settlement expense

) !:(|
non-level premiums, or both. P!:(| = G = gross premium

ä !:(|

Net Future Loss A Note:

!:(|
P!:(| =
"L = PV" (f. benefits) − PV" (f. premiums) ä !:(| 1. Replace A and d with their continuous

Equivalence Principle counterparts for fully continuous policies.

If P is not expressed as one of the symbols
2. Add n|p to A′s for endowment insurance.
• E| "L} = 0 shown above, then P will be defined in the
⇒ EPV" (f. premiums) = EPV" (f. benefits) 3. This shortcut formula can only be used
text of the question.
for fully discrete/continuous whole life
• The premium determined under the
and endowment insurance. For other
equivalence principle and excluding Gross Premium
products, use basic principles.
expenses is called the net premium. Gross premiums account for both benefits

and expenses.
For a whole life insurance policy of $b on Portfolio Percentile Premium
(x) with level premiums P: Gross Future Loss S = L) + L* + ⋯ + LM
L
Fully Discrete "L = PV" (f. benefits) + PV" (f. expenses) E[S] = N ⋅ E[L]
− PV" (f. gross premiums) Var[S] = N ⋅ Var[L]
"L = bv :!&) − Pä :GGGGGGGGG
! &)|

P P Equivalence Principle Using the portfolio percentile premium

= Öb + Ü v :!&) −
d d E| "LL } = 0 principle, the premium is set such that there
E| "L} = bA! − Pä ! ⇒ EPV" (f. gross premiums) is a specified probability (x%) that the total
P * = EPV" (f. benefits) + EPV" (f. expenses) loss is negative:
Var| "L} = Öb + Ü | *A! − (A! )* }
d
Pr[S < 0] = x%
Fully Continuous
= bv ;! − PauGGGGG Percentile of 𝟎𝟎𝐋𝐋
"L ;! |
P P The 100pth percentile of "L is the value πJ
= Öb + Ü v ;! −
δ δ such that Pr| "L ≤ πJ } = p. To determine
p ! − Pau!
E| "L} = bA πJ :
*
P 1. Graph "L on y-axis and T! on x-axis.
p ! − (A
Var| "L} = Öb + Ü | *A p ! )* }
δ 2. Identify the parts of the curve where

Under the equivalence principle, we have: "L ≤ πJ . Determine the value of T! that
Fully Discrete corresponds to those parts.
3. Use the value of T! from Step 2 to
A! 1 dA!
P=b⋅ = b g − dh = b g h calculate πJ .
ä ! ä ! 1 − A!
*
A! − (A! )*
Var| "L} = b* ⋅
(1 − A! )*
Fully Continuous
p!
A 1 p!
δA
P=b⋅ = b g − δh = b b p ! c
au! au! 1−A
*p p ! )*
A! − (A
Var| "L} = b* ⋅
p ! )*
(1 − A

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 6

RESERVES RESERVES Variance of Loss Thiele’s Differential Equation
For a fully continuous whole life or d
V = δ# ⋅ #V + G# − e#
Net Premium Reserve endowment insurance with premiums P, the dt #
Prospective Method variance of the net future loss at time t is: −Ub# + E# − #VVµ!&#

#V = EPV# (f. benefits) − EPV# (f. premiums) P *
Var| #L} = Öb + Ü Var[Z] where

δ
If #V occurs at the same time as a premium G# = annual rate of premium payable
where
or benefit, then be careful about which cash at time t
Var[Z] = *Ap !&# − (A
p !&# )*
flows to include in calculating the future *
e# = annual rate of premium-related
p
Var[Z] = *A p
− ZA [ expense payable at time t
loss. Unless stated otherwise, assume: !&#:(+#| !&#:(+#|

- all death benefits at time t occurred in the For fully discrete insurance, remove the E# = expense of paying the face amount at
past bars and replace δ with d. time t (e.g., settlement/claim expense)
- all premium payments occur in the future b# = face amount payable at time t if the
- endowment payments occur in the future Gross Premium Reserve insured dies at exact time t

Note: Prospective Method
L
To solve the differential equation
- The time-0 net premium reserve is 0 #V = EPV# (f. ben.) + EPV# (f. exp.) numerically, use Euler’s method:
because the equivalence principle is − EPV# (f. pre.) • Forward Euler Approximation:
assumed:

"V
L
= 0 if the following 2 requirements d #&DV − #V
#V = E| "L} = 0 #V ≈
are satisfied: dt h
- The time-n net premium reserve for an • Backward Euler Approximation:
1. The gross premium is determined using
n -year term insurance is 0 because there d #V − #+DV
the equivalence principle. V≈
are no future benefits or premiums due dt # h
2. The assumptions used for calculating the
at time n:

reserve are the same as those used in For net premium reserve, drop
= 0 (V
calculating the premium. expense-related terms and replace G#
- The time-n net premium reserve for an n-
with net premium.
year endowment insurance right before
Expense Reserve
the endowment benefit is paid is equal to
Expense Premium (a.k.a. Expense Loading) Interim Reserves (𝟎𝟎 ≤ 𝐬𝐬 ≤ 𝟏𝟏)
the endowment benefit, because there are
= Gross Premium − Net Premium
no future premiums due at time n, and the O L
#V = # V − #V (
only future benefit due at time n is the
O
endowment benefit. #V = EPV# (f. exp.)
− EPV# (f. exp. premium)
(V = endowment benefit
Expense reserve is usually negative. Exact Method
Special Formulas

Recursive Formula U #V + PV(1 + i)/ = /q!&# ⋅ bv)+/

For a fully discrete whole life insurance
policy of $b, the net premium reserve can • Net premium reserve only considers + /p!&# ⋅ #&/V
)+/
also be calculated as: benefits, not expenses: #&/V(1 + i) = )+/q!&#&/ ⋅ b
ä !&# U #V + PV(1 + i) = q!&# ⋅ b + p!&# ⋅ #&)V + )+/p!&#&/ ⋅ #&)V
#V = Ö1 − Ü ⋅ b
ä ! • Expense reserve only considers expenses, Linear Approximation
A!&# − A! not benefits: #&/V = U #V + PV(1 − s) + s ⋅ #&)V
#V = Ö Ü ⋅ b
1 − A!
U #VO + PO − eV(1 + i) = q!&# ⋅ E
Note:
+ p!&# ⋅ #&)V O
1. Replace A and a with their continuous
• Gross premium reserve considers both
counterparts for fully continuous policies.
p to A′s and a′s for endowment benefits and expenses:
2. Add n| L
insurance. U #V + G − eV(1 + i) = q!&# ⋅ (b + E)
L
3. These special formulas can only be used + p!&# ⋅ #&)V
for fully discrete/continuous whole life
and endowment insurance with ordinary
level premium and benefits. For all
others, use basic principles.

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 7

Modified Reserve MARKOV CHAINS
MARKOV CHAINS Boundary Conditions
A modified reserve is a reserve computed Boundary conditions are constraints that a
without expenses but adjusting the A Markov chain is a multiple state model differential equation must satisfy for it to be
valuation premiums to allow implicitly for that is memoryless; the probability that the valid. For any model:
initial expenses. process is in any given state at some future <<
"p! = 1

date depends only on the current state, not <R
For any modified reserve method: = 0 (j ≠ i)
"p!
on the history of the process.
EPV(modified prems) = EPV(net prems)

= EPV(benefits) <R Premiums

#p! : probability someone in state i at age x
For an annuity on (x) currently in state i
is in state j (where j may equal i) at age
Full Preliminary Term (FPT) that pays $1 per year while the person is in
x + t
FPT reserve is a modified reserve. The <<
state j:
#p! : probability someone in state i at age x -
policy is treated as if it were issued one year <R <R
remains in state i until age x + t au! = P e+F# #p! dt
later, with the first year of the policy being "
<<
#p! ≤ #p<<! -
treated as if it were a one-year term <R <R

ä ! = \ v ' 'p!
insurance. If reentry to state i is impossible, then:
'1"
<<

#p! = #p<<! For an annuity on (x) currently in state i
For example, using this method, a fully
that pays $1 per year for a maximum of n
discrete whole life insurance issued to (x)
Continuous Probabilities years while the person is in state j:
would be treated as if it were a one-year (
Direct Approach <R <R
term insurance on (x), followed by a whole au = P e+F# #p! dt
!: (|
# "
life insurance on (x + 1): << <R
#p! = exp í− P \ µ!&/ dsì (+)
" RS< ä
<R
= \ v ' 'p!
<R
!: (|
• 1st year modified net premium, α:
For permanent disability model: '1"
EPV" (modified prems) = EPV" (benefits) # For an insurance on (x) currently in
<R << <R RR
α = A)!:)|
B = vq ! #p! =P /p! ⋅ µ!&/ ⋅ #+/p!&/ ds state i that pays $1 immediately upon
"
• Renewal modified net premium, β: every transition to state j:
Approximation -
EPV) (modified prems) = EPV) (benefits) p<R! = P e+F# \ #p<'
A
'R
! ⋅ µ!&# dt
βä !&) = A!&) Kolmogorov’s Forward Equations: " 'SR
(
A!&) d <R 'R <R R'
-
β= p = \ Z #p<'
! ⋅ µ!&# − #p! ⋅ µ!&# [ = P e+F# ⋅ Pr[start in i, move into j]dt
ä !&) dt # ! "
'1"
• FPT reserves 'SR
For an insurance on (x) currently in
PQ; = Pr(Start in i, move into j)
"V = 0 state i that pays $1 immediately upon
PQ; −Pr(Start in i, move out of j)
)V = 0 every transition to state j within a fixed

To solve the differential equation term of n years:
(
numerically, use Euler’s method: p<R = P e+F# \ #p<' 'R
A !: (| ! ⋅ µ!&# dt
<R <R "
d <R #&Dp! − #p! 'SR
p ≈ (
dt # ! h
= P e+F# ⋅ Pr[start in i, move into j]dt
"
Alternatively, with Euler's method, a

continuous Markov chain can also be Note:

transformed to a discrete Markov chain • The insurance pays upon every transition
with time increments of h, and the into state j, regardless of which state the
transition probability can be transition occurs from.
approximated as: • As long as it is possible to go from state i
<R to state j, even if there is no direct
<R hµ i ≠ j
Dp! ≈ Ä ! <∙
1 − hµ! i = j transition from state i to state j,
<R <R p<R
#p! , a
u ! , A! are not 0.

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 8

Helpful Relationships between Insurance To solve the differential equation Insurance Applications
and Annuities numerically, use Euler’s method: Consider a whole life policy:
1. Suppose there are m+1 states in a (<) (<)
d #V − #+DV
multiple state model, numbered 0, 1, ..., V(<) = Discrete
dt # h
and m. Then, for any state i, we have: EPV[benefits]
3
<R
3 -
\ au = au(| (R) (U) (R)
B = \ \ v '&) b
MULTIPLE DECREMENT MODELS 'p! q !&'
R1"
!: (| MULTIPLE DECREMENT MODELS
R1) '1"

2. For all possible states at time n, we -
Multiple Decrement Tables (MDT) (U)
subtract all available insurance policies EPV[annuity] = \ v ' 'p!
Decrements are dependent on each other. '1"
and annuities:
3 Discrete Probabilities
#+) #+) Continuous
p<R = A
A p<R! − v ( \ (p<' p'R
! ⋅ A!&( (R) (U) (R) (R)
#q ! =\ 'p! q !&' =\ '|q!
!: (|
'1" -
(U) (R)
3 '1" '1" EPV[benefits] = P v # #p! µ!&# dt
<R <R 'R 3 "
au
!: (|
= au! − v ( \ (p<'
! ⋅a
u !&( (U) (R) (R)
'1" #q ! = \ #q ! µ!
R1) EPV[benefits] = (U)
(Under CF)
3. Suppose that state k is an absorbing δ+µ
(U) (U)
state in a model with m+1 states. For any #p! + #q ! = 1 -
(U)
non-absorbing state i, where i ≠ k:
#&$+) EPV[annuity] = P v # #p! dt
(R) (U) (R) (U) (R) "
3 #|$q ! = \ 'p! q !&' = #p! $q !&#
p<' <R '1#
A ! =1−δ⋅ \ a
u! Fractional Ages (0 ≤ s ≤ 1)
R1",RS'
Life Table • UDD in the multiple decrement table:
Woolhouse’s Formula for State-Dependent (R)
#d!
(R) (R)
Annuities (R) /q ! = s ⋅ q !
#q ! = (U)

l! (U) (U)
From annual annuity EPVs: /q ! = s ⋅ q !
(U) (U) (U)
(3)<R <R m* − 1 <R (U) #d! l! − l!&# • Constant forces of decrement:
ä ! ≈ ä ! + µ for i ≠ j #q! = (U) = (U)

12m* ! l! l! (R)
µ! q!
(R) (R)
/q !
m − 1 m* − 1 <∙ (U) (U)
= (U)
= (U)

(3)<< l!&#
ä ! ≈ ä <<! − − Uµ! + δV (U) µ! q! /q !
2m 12m* #p! = (U)
l! (R)
From continuous annuity EPVs: (R) q! (U) /
(R) /q ! = (U) w1
− Zp! [ x
(3)<R <R 1 <R (R) $d!&# q!
ä ! ≈ au! − µ for i ≠ j q
#|$ ! =
12m* ! (U)
l!
1 1
(3)<<
ä ! ≈ au<<! + + Uµ<∙ + δV

Associated Single Decrement Tables
2m 12m* ! Continuous Probabilities
(ASDT)
(R)
#
(U) (R)
#q ! = P /p! µ!&/ ds The associated single decrements are
Reserves " independent.
Thiele’s Differential Equation d (R) #
dt #q ! V(R) (R)
d (<)
(R)
⇒ µ!&# = #p! = exp O− P µ!&/ dsQ
V (<) = δ# #V (<) − B# (U)
dt # #p!
"
( 3 (R) d V(R)
<R <R (U) (R) ⇒ µ!&# = − ln Z #p! [
− \ µ!&# Zb# + #V(R) − #V(<) [ µ!&# = \ µ!&# dt
R1" #
R1) V(R) V(R) (R)
RS< # #q ! = P /p! µ!&/ ds
(U) (U) "
where #p! = exp O− P µ!&/ dsQ
d V(R)
#q !
(<) "
B# = rate of benefit payment while the # ⇒ µ!&# = dt
(R)

(U) (U) (U) V(R)
insured is in state i #q ! = P /p! µ!&/ ds #p!
"
<R V(R) V(R)
µ!&# = force of transition from state i d (U) d (U) #p! + #q! = 1
#q ! − #p!
= dt (U) = dt (U)
(U) 3
to j at age x + t ⇒ µ!&# (U) V(R)
<R #p! #p! #p! = ó #p!
b# = lump sum benefit payable
#&$ R1)
(R) (U) (R)
instantaneously at time t on transition #|$q! =P /p! µ!&/ ds
#
from state i to j

In this model, premiums are treated as
negative benefits and expenses can be
treated as additions to the benefits.

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 9

Key Relationships between MULTIPLE LIVES
MULTIPLE LIVES Relationships between (𝐱𝐱𝐱𝐱) Status and
V(R) (R)
#q ! and #q !
(𝐱𝐱𝐱𝐱
uuu) Status
No Assumption Joint Life T!, + T!, = T! + T,
T!, = minUT! , T, V T!, ⋅ T!, = T! ⋅ T,
(U) 3 V(R)
⎧ #p! = ∏R1) #p!
#p!, + #q!, = 1 #p!, + #p!, = #p! + #p,
(U) (R)
• #p! = 1 − ∑R1) #q!
3
∘ ∘ ∘ ∘
⎨ $|#q !, = $p!, ⋅ #q!&$:,&$ e!, + e!, = e! + e,
(U) (U)
⎩ #p! + #q ! = 1 = $p!, − $&#p!, ∘ ∘ ∘ ∘
B + e!,:(|
e!,:(| B = e!:(|
B + e,:(|
B

d (')
= $&#q !, − $q!, e!, + eG!,
GGG = e! + e,
⎧ µ(R) = d# #W!
p p p! + A
p,
⎪ !&# (() A!, + A!, = A
• #J!
Independent Lives
⎨ (R) W
d )(')
au!, + au!, = au! + au,
d# # !
⎪µ!&# = J)(') #p!, = #p! ⋅ #p,
⎩ # !
(E!, + (E!, = (E! + (E,
#q !, = #q ! + #q , − #q ! ⋅ # q ,

UDD in Multiple-Decrement Tables µ!&#:,&# = µ!&# + µ,&# Covariance of T!, and T!,
(UDDMDT) (0 ≤ s ≤ 1)

(') Moments CovUT!, , T!, V = CovUT! , T, V

W!
V(R) (U) W(() - ∘ ∘ ∘ ∘
∘ + Ze! − e!, [ Ze, − e!, [
/p! =Z p
/ ! [ !
e!, = P #p!, dt
"

- CovUT! , T, V = 0 if T! and T, are independent
UDD in Associated Single Decrement Tables *
E wUT!, V x = 2 P t ⋅ #p!, dt
(UDDASDT)(0 ≤ s ≤ 1) " Exactly One Life Survives
-
For 2 decrements: Pr(exactly one life survivies t years)
V(*) e!, = \ 'p!,
()) V()) s* q! = #p!, − #p!,
/q ! = q ! ûs − ü '1)
2 = #p! + #p, − 2 ⋅ #p!,
For 3 decrements: Last Survivor
())
s* ZqV(*)
! + qV(9)
! [ T!,
GGGG = maxUT! , T, V Relationships between Insurance
/q ! = qV())
! †s −
2 #p!, + #q!, = 1 Policies, Annuities, and Premiums
A!, = 1 − dä !,
$|#q !, = $p!, − $&#p!,
s9 qV(*)
! ∙ qV(9)
! 1−A p G!,
+ ¢ =
GGG
3 $&#q !, − $q!, GGG =
auG!,
δ

CF in MDT or ASDT (0 ≤ s ≤ 1) Independent Lives ä !,:(|

B = ä !, − (E!, ∙ ä !&(:,&(
(') 1
#q !, = #q ! ⋅ # q ,
W!
PG!,
GGG = − d
V(R) (U) (()
W! ä G!,
/p! =Z /p! [ GGG
#p!, = #p! + #p, − #p! ⋅ #p, p !,
δA
P!, = p!,
#p! µ!&# ⋅ #q, + #p, µ,&# ⋅ #q! 1−A
µ!, (t) =
#p! + #p, − #p! ⋅ #p, Note: The list above is not exhaustive;

similar relationships can be applied to other
Moments
- forms of insurance/annuities with
∘
e!, = P #p!, dt appropriate adjustments.
"
-
e!, = \ 'p!, Contingent Probabilities
'1) #
)
#q !, = P /p!, ∙ µ!&/ ds
"
#
)
#q !, = P /p!, ∙ µ,&/ ds
"
#
*
#q !, = P /p! ∙ µ!&/ ⋅ U1 − /p, Vds
"
#
*
#q !, = P /p, ∙ µ,&/ ⋅ U1 − /p! Vds
"

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 10

Relationships With waiting period of w years and benefit Profit Vector
#q !, + #q !, = #q !,
) ) term of m years, the EPV is: The expected profit emerging at the end of
)
+ #q*!, = #q !

each year given the policy is in force at the
#q !, -+('#()

) * ! "" "%
$p! µ!#$ $a&%%
!#$:'#(| & %%
********* − a **** ( e
!#$:(|
+,$
dt start of the year can be expressed as a
#q !, = #q !, + #q ! ⋅ #p, "
-+( vector:
*
#q !, + #q *
!, = #q !, +! "" "%
$p! µ!#$ $a&%% & %%
****** − a **** ( e
+,$
dt
-+('#()
!#$:-+$| !#$:(| Pr = (Pr" Pr) Pr* … Pr( )
) )
-q!, + -q !, = 1 Profit Signature
* * The expected profit emerging at the end of
-q !, + -q !, = 1 Reserve Recursion for Policies with
)
= -q * Multiple States each year given a policy in force at issue is:
-q !, !,
Assuming there are m + 1 states and cash in force in force

Π = Pr# ⋅ Prob ß at time ® at time ©
Contingent Insurance flows are made every h years: t−1 0
- (R)
p)!, = P v # ⋅ #p!, ⋅ µ!&# dt Z #V (R) + hP# [ (1 + i)D Profit signature: (Π" Π) Π* … Π( )
A
3
" where
- R' (') R'
p*!, = P v # ⋅ #p! ⋅ µ!&# ⋅ #q, dt = \ Dp!&# ZhB#&D + b#&D + #&DV(') [ Π" = Pr"
A
" '1"
Π# = Pr# ⋅ #+)p! , t = 1, 2, 3, … , n
Relationships If lump sum benefit is assumed to be paid in
p)!, + A
A p ) p the middle of an interval: Profit Measures
!, = A!,

p*!, + A
A p * p NPV
!, = A!, (0)
- $V (0) + hP$ 1(1 + i)1 -
p) p* p
A!, + A!, = A! ' R
02 (2) 02 NPV = \ ΠR ⋅ vH
= 7 1p!#$-hB$#1 + b$#1 (1 + i)1/4 + $#1V (2) 1
R1"
25"
Reversionary Annuities where r = risk discount or hurdle rate
• Make payments to (y) after (x) has died: Activities of Daily Living (ADLs): Partial NPV
au!|, = au, − au!, • Bathing '
R
• Make payments only when exactly one • Dressing NPV(k) = \ ΠR ⋅ vH
life is alive: • Eating R1"

• Toileting IRR
EPV(annuities) = au!, − au!, -
• Continence R
NPV = \ ΠR ⋅ vH = 0
• Transferring R1"
LONG-TERM LONG-TERM DPP
INSURANCE COVERAGE DPP = min[t: NPV(t) > 0]
INSURANCE COVERAGE
PROFIT TESTS PROFIT TESTS

Profit Margin
Disability Income Insurance (DII)
Profits for Traditional Products NPV
Continuous Sojourn Annuity Profit margin =
EPV(f. premiums)
The EPV of an n-year continuous sojourn The expected profit emerging at time t,

annuity on (x) in state i that pays $1 per given that the policy is in force at time t-1 is:
Zeroized Reserves
year continuously while the life remains in Pr#
1. Begin with the last year
state i is: = #+)V + P# − E# + I# − EDB# − ESB# − E# V
and work backwards

= U #+)V + P# − E# V(1 + i)
( 2. Set the profit for the year to zero then
au<<!:(|
B =P
<< +F#
#p! e dt (Z)
− q!&#+) (DB# + S# )
"
solve for the beginning-of-year reserve
(Y)
− q!&#+) (CV# ) 3. If the reserve is negative, set to zero
EPV of benefit of an n-year DII: (U) and repeat this entire process again
− p!&#+) #V
( until time 0
= NCF# + Δ #V
au")B =P
!:(|
"" ") ))
#p! µ!&# a
u !&#:(+#|
GGGGGG e
+F#
dt
" where:

With waiting period of w years, the EPV is: E# V = p!&#+) #V

Δ #V = #+)V(1 + i) − p!&#+) #V
(+Y
P "" ")
u))
#p! µ!&# Za GGGGGG
!&#:(+#| − au)) GGGG [ e
!&#:Y|
+F#
dt
"

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 11

Gain by Source PENSION MATHEMATICS
PENSION MATHEMATICS

• Two methods to calculate the amount of

• Expenses: Actual expenses being less than retirement benefit:
assumed is a source of profit. Valuation of Benefits § PUC: projects salary to retirement or
• Interest: Actual interest earned being Motivations exit date
more than assumed is a source of profit. 1. Attract potential employees § TUC: calculates salary based on
• Mortality: Actual mortality being less 2. Provide incentive for employees to stay employee’s current age
than assumed is a source of profit for 3. Facilitate turnover of older employees • Early retirement
insurance products but a source of loss 4. Provide tax-friendly compensation Annual Retirement Benefit
for annuity products. 5. Pressure from trade unions = n ⋅ SP<( ⋅ α
Analysis of surplus in the order of expenses, 6. Reward employees who have contributed ⋅ (1 − pension reduction factor)
interest, and mortality ( ′ = actual): to the company’s success • Withdrawal without COLA

Gain from Expenses Annual Retirement Benefit = n ⋅ SP<( ⋅ α
Replacement Ratio
= Assumed Expenses − Actual Expenses • Withdrawal with COLA
= Ue# − eV# V(1 + i# ) + UE# − E#V Vq!&#+) pension income in the year after retirement Annual Retirement Benefit
R=
salary in the year before retirement = n ⋅ SP<( ⋅ α
Gain from Interest
= Actual Interest − Assumed Interest Salary Projection ⋅ (1 + COLA)HO#<HO3O(# \LO+Y<#DZH\Y\C \LO
= UiV# − i# VU #+)V + G# − eV# V S: Salary Su: Rate of salary
Gain from Mortality (for insurance) s: Salary scale s̅: Rate of salary function Funding the Benefits
= Assumed Mortality − Actual Mortality • Constant percentage of increase Actuarial Liability
= Uq!&#+) − qV!&#+) V × UDB# + E#V − #VV S,&# = S, (1 + x%)# #V = EPV(all accrued benefits at time t)

Note: • Salary Scale
Normal Contribution
s,
• Notice order matters. The analysis can be S, = S! ⋅
s!
performed in a different order, which #V + C# =
• Rate of salary
would yield slightly different results. EPV(benefits for mid-year exits)+vp""
! #&)V
s̅ ,
Su, = Su! ⋅

• The key in this analysis is: if a source
s̅ ! • Two methods of funding benefits: PUC
has been accounted for, use the )
Su! )
and TUC
actual experience; otherwise, use S, = P Su,&# dt = ⋅ P s̅ ,&# dt
" s̅! " • If there are no mid-year exits:
the assumed experience. )
• Relationships § PUC: C# = ⋅ #V
Actual Profit #
Rate of salary function to salary scale: #&) ]!
Use the actual experience. ) § TUC: C# = Z ⋅] − 1[ #V
# !*+
Actual Profit = U #+)V + G# − eV# V(1 + iV# ) s, = P s̅ ,&# dt
"

−qV!&#+) (DB# + E#V ) Retiree Health Benefits
Salary scale to rate of salary function:
−pV!&#+) ⋅ #V s̅ , ≈ s,+".[ Benefit Premium Annuity for age x at time t
Expected Profit
-
B(x + k, t + k)
Use the assumed experience. Defined Contribution Pension Plans ä ^ (x, t) = \ v ' 'p! O Q
B(x, t)
AV(pension fund) = EPV(pension benefits) '1"
Expected Profit = U #+)V + G# − e# V(1 + i# )

−q !&#+) (DB# + E#) Value of retiree health benefit at retirement
Defined Benefit Pension Plans
−p!&#+) ⋅ #V for a life retiring at age xr in t years:
Annual Retirement Benefit = n ⋅ SP<( ⋅ α
Total Gain B(xr, t)ä ^ (xr, t)
where

Total Gain = Actual Profit − Expected Profit n = total number of years of service When healthcare premiums increase
= Gain from Expenses SP<( = final average salary exponentially with age and at a constant
+ Gain from Interest α = accrual rate annual inflation rate where:
+ Gain from Mortality • c = B(x + 1, t)⁄B(x, t)
• j = annual rate of inflation for healthcare
costs

B(xr + k, t + k) = c ' (1 + j)' B(xr, t)

ä ^ (xr, t) = ä !H|<∗
1+i
where i∗ = − 1
c(1 + j)

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 12

Actuarial Value of Total Health Benefit Tail Correction Age-Based Estimators
(AVTHB) • Efron’s tail correction: Individual Data
AVTHB# S≤(t) = 0 for t > u • Exact Exposure:
_[+!
r!&' ' • Klein and Moeschberger's tail correction: qªR = 1 − e+Z'/O'
= \ v B(x + k, t + k) ∙
l! S≤(u), for u < t < γ • Actuarial Exposure:
'1" S≤(t) = n
0, for t ≥ γ dR
ä ^ (x + k, t + k) qªR =
_[+!
• Brown, Hollander, and Korwar’s tail eR
r!&' ' correction:
= B(x, t) \ v ∗ ä ∗ Note that exposures for deaths are
l! < !&'|< S≤(t) = S≤(u)#/$ for t > u
'1" counted until the end of the age interval


Actuarial Liability at time t, tV Nelson-Aalen Estimator
R
Assumption Description
Assuming linear accrual to the earliest s
possible retirement age (pro-rata accruals ¥ UyRV = \ <
H Advance birthday
r< Age Last Birthday
<1)
method): to policy date
t S≤UyR V = e+à b,' c
#V = ⋅ AVTHB# Advance/retreat
t+k Age Nearest
Variance of Estimators birthday to nearest
Assuming linear accrual to each possible Birthday
• Greenwood’s Approximation: policy date
retirement age: R
s< Study starts and
#V =
± |S≤UyR V} = |S≤UyR V}* \
Var Anniversary-to-
r< (r< − s< ) ends on policy
_[+! <1)
t r!&' ' Anniversary
\ ∙ v B(x + k, t + k) ∙ Used for Kaplan-Meier anniversary
t + k l!
'1"
• Klein’s Estimation:
ä ^ (x + k, t + k)
R Variance of Estimators
s< (r< − s< )
Normal Cost ± |H
Var ¥ UyR V} = \ • Exact Exposure:
r<9 dR
#V + C# = EPV(benefits for mid-year
<1) *
R
± |qªR } = U1 − qªRV ⋅
Var
exits)+vp"" e*R
! #&)V s< (r< − s<)
± |S≤UyR V} = |S≤UyR V}* \
Var

r<9 • Actuarial Exposure:
If there are no mid-year exits: <1)
Used for Nelson-Aalen qªR U1 − qªR V
± |qªR } =
Var
#V eR
C# =
t In general,

1 ± |S≤UyR V} = |S≤UyR V} ⋅ Var
± |H *
¥ UyR V} Interval-based Data
Var
C# = ⋅ AVTHB# Exposures without uniform assumption
t+k

(for linear accrual to the earliest possible Confidence Interval eR = PR + nR

retirement age) Linear Confidence Interval for S(t): Exposures with uniform assumption
• Exact Exposure:
± |S≤(t)}
S≤(t) ± z()&J)/* ∂Var
eR = PR + 0.5UnR − wR − dR V
SURVIVAL MODEL ESTIMATION
SURVIVAL MODEL ESTIMATION Log-transformed Confidence Interval:
• Actuarial Exposure:
• For S(t):
eR = PR + 0.5UnR − wR V
Kaplan-Meier and Nelson-Aalen US≤(t))/d. , S≤(t)d. V where

Estimators Multiple State Estimation

± |S≤(t)}
z()&J)/* ∂Var
Empirical Distribution U] = exp ⎛ ⎞ Transition intensity:
# of data points = x S≤(t) ⋅ ln S≤(t) d<R
Pr(X = x) = ⎝ ⎠ µª<R =
n
T<
# of data points ≤ x • For H(t): d
F( (x) = Pr(X ≤ x) = ± Uµª<R V = <R*
Var
n ¥ (t) T<
H
± [S( (x)] = Var
Var ± [F( (x)] O ¥ (t) ⋅ U` Q where
,H
U`
F( (x) ⋅ [1 − F( (x)]
=
n ± |H
z()&J)/* ∂Var ¥ (t)}
U` = exp ⎛ ⎞

Kaplan-Meier Estimator ¥ (t)

H
R ⎝ ⎠
s<
S≤UyR V = ó Ö1 − Ü 1
r< U] =
<1)
U`
S≤UyR V
h≤UyR V = 1 −
S≤UyR+) V

www.coachingactuaries.com Copyright © 2021 Coaching Actuaries. All Rights Reserved. 13

 MORTALITY IMPROVEMENT
MORTALITY IMPROVEMENT Note: With these constraints:
• Mortality reduction factors, r! , is the ∑##1#
3
lm(x, t)
2
Deterministic Model: Single-Factor term used in AMLCR to reflect mortality α! =
t( − t" + 1
Mortality Improvement Scales improvement. The single factor mortality

Depend only on age. Mortality improvement improvement factor in the study note Central death rate improvement factor:
factors are not a function of calendar year: m(x, t)
uses the symbol φ! . Note that r! = 1 − φ3 (x, t) = 1 −
m(x, t − 1)
q(x, t) = q(x, 0)(1 − φ! )# φ! .

In the Lee Carter model:
• When the symbol φ! is used, the
Deterministic Model: Two-Factor 1 − φ3 (x, t)~logN(µ = β! c, σ = β! σ')
mortality improvement is simply a
Mortality Improvement Scales
function of age x.
Cubic Spline Stochastic Model: The Cairns-Blake-
• When the symbol φ(x, t) is used, the Dowd (CBD) Models
A cubic function that joins the short-term
mortality improvement is simply a Logit Function
factors and long-term factors in a smooth x
function of both age x and calendar year t. logit(x) = ln
fashion 1−x

q(x, t)
Age-Based Cubic Spline Stochastic Model: The Lee Carter Model lq(x, t) = logit[q(x, t)] = ln
1 − q(x, t)
C\ (x, t) = at 9 + bt * + ct + d Central Death Rate
) eCW(!,#)
q! ∫" #p! µ!&# dt ⇒ q(x, t) =
Cohort-Based Cubic Spline m! = = 1 + eCW(!,#)
) )
∫" #p! dt ∫" #p! dt
Ce (x, t) = a∗ t 9 + b∗ t * + c ∗ t + d∗
The Original CBD Model
Solve for the constants in the age-based and Assuming constant force of mortality ()) (*)
lq(x, t) = K # + K # (x − x)
cohort-based cubic splines using these 4 between integer ages:

m! = µ where:
constraints:
1. The starting value of the spline matches q! = 1 − e+3! • x is the average age in the data set
()) ()) ())
the improvement factor at 2007. Assuming UDD between integer ages: • K# = K #+) + c()) + σ'+ Z#
q! (*) (*) (*)
2. The starting derivative of the spline m! = K# = K #+) + c(*) + σ'0 Z#
1
matches the slope of the improvement 1 − 2 q! ()) (*)
• E wZ# Z# x = ρ, − 1 ≤ ρ ≤ 1
function at 2007. m!
q! = (<) (R)
3. The ending value of the spline matches 1 E wZ# Z$ x = 0 for t ≠ u, i = 1,2, j = 1,2
1 + 2 m!
the improvement factor at 2027.

Advantages of the original CBD model over

4. The ending derivative of the spline Normal and Lognormal Random Variables
the Lee Carter model:
matches the slope of the improvement X~N(µ, σ* ) represents a normally
distributed random variable, X, with mean µ • Fewer parameters
function at 2027.

and variance σ* . • Less parameter uncertainty
Improvement factor for age x in year t:

Z~N(0,1) represents a standard normal Disadvantage of the original CBD model

φ(x, 2007 + t) = 0.5C\ (x, t) + 0.5Ce (x, t)

random variable with a mean of 0 and a over the Lee Carter model:
where t is the number of years measured • Fit to population is sometimes worse
variance of 1.
from 2007

Y = ef ~logN(µ, σ) is a lognormal random The CBD M7 Model
Mortality Rate with Mortality Improvement ()) (*)
lq(x, t) = K # + K # (x − x)
#
variable with parameters µ and σ.
/0 (9)
+K # ((x − x)* − s!* ) + G#+!
q(x, t) = q(x, 0) óU1 − φ(x, k)V • E[Y] = e4& 0

'1) 0
• Var[Y] = (E[Y])* Ueg − 1V where:

∑!!4!
1 (!+!)0

The Lee Carter Model • s!* = 2

!1 +!2 &)

ln m(x, t) = lm(x, t) = α! + β! K # + ϵ!,# • G#+! introduces a cohort time series

where: Advantages of the CBD M7 model:
• K # = K #+) + c + σ' Z# • Includes a cohort effect
• ϵ!,# is sufficiently small to be negligible • Includes a quadratic age difference term

Constraints used to fix identifiability
problem:
!1 #3

\ β! = 1.0 and \ K # = 0.0

!1!2 #1#2

Copyright © 2021 Coaching Actuaries. All Rights Reserved. 14

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