Professional Documents
Culture Documents
MLC Review
MLC Review
MLC Review
LIFE CONTINGENCIES
x x
3. x q0 = Pr( X ≤ x ) = ∫ f X ( s )ds = ∫ s p0 μ ( s )ds
0 0
f X ( x + t)
(pdf): f T (t ) =
Pr( X > t )
Pr( X > x + t ) p0
(sf): sT (t ) = Pr(T > t )= t p x = = x +t
= 1− t q x
Pr( X > x ) x p0
1. f T (t )= t p x ⋅ μ ( x + t )
t
2. t q x = ∫0 s p x ⋅ μ ( x + s )ds
t x +t ∞
3. t p x = exp( − ∫ μ ( x + s )ds ) = exp( − ∫ μ ( s )ds ) Also, t p x = ∫ s p x ⋅ μ ( x + s )ds
0 x t
4. 1 q x = q x and 1 p x = p x
5. (survival factorization) n +t p x = n p x ⋅t p x + n
6. (deferred mortality)
t +u
t |u qx = Pr(t < T ( x ) ≤ t + u ) = ∫ s p x ⋅ μ ( x + s )ds = t p x − t + u p x = t + u qx − t qx = t p x ⋅u qx + t
t
d
− [t px ]
7.
d
[ t p x ] = − f T (t ) =− t p x ⋅ μ ( x + t ) and so μ ( x + t ) = dt
dt t px
8.
d
[ t p x ]= t p x [μ ( x ) − μ ( x + t )] (Use 2nd equality in 3. above and the FTC)
dx
∞ ∞
10. E[(T ( x )) 2 ] = ∫0 t 2 ⋅t p x ⋅ μ ( x + t )dt = ∫02t⋅t p x dt (Helps us get variance of T.)
11. (n-year temporary complete expectation of life for (x); this is the
expected number of years lived by (x) between ages x and x+n)
0 n n
e x :n | = E[T ( x ) ∧ n ] = ∫ t⋅t p x ⋅ μ ( x + t )dt + n⋅n p x = ∫ t p x dt
0 0
K Pr
0 Pr(0 ≤ T < 1) = Pr(0 < T ≤ 1) = q x ( = 0| q x )
1 Pr(1 ≤ T < 2) = Pr(1 < T ≤ 2) = 1| q x
2 Pr(2 ≤ T < 3) = Pr(2 < T ≤ 3) = 2| q x
M M
Note: n| q x = n p x − n +1 p x = n +1 q x − n q x = n p x ⋅ q x + n
k
(cdf): FK ( k ) = Pr( K ≤ k ) = ∑ n| q x (Notice that Pr( K ≤ k ) = Pr(T ≤ k + 1)= k +1 q x )
n =0
(The last equality is used often. Notice the index starts at k = 1.)
∞ ∞
2. E[( K ( x )) 2 ] = ∑ k 2 ⋅k | q x = ∑ (2k − 1)⋅k p x (Helps us get variance of K.)
k =0 k =1
3. (n-year temporary curtate expectation of life for (x); this is the expected
complete number of years lived by (x) between ages x and x+n)
n
e x:n| = ∑ k p x
k =1
4. (Recursion Formulas)
e x = e x:n| + n p x ⋅ e x + n and when n = 1 we get e x = p x ⋅ (1 + e x +1 )
e x:n| = e x:m| + m p x ⋅ e x + m:n −m|
l x+n d l −l d x +n l x +n − l x +n+m
1. n px = and n q x = n x = x x +n = 1− n p x and n|m qx = m
=
lx lx lx lx lx
−
d
[l x ]
, and so [l x ] = −l x ⋅ μ (x )
dx d
2. μ ( x ) =
lx dx
n n
3. n d x = ∫ l x +t ⋅ μ ( x + t )dt (follows since n q x = ∫ t p x ⋅ μ ( x + t )dt )
0 0
4. n Lx = the total number of years lived in the next n years by the l x people
alive at age x
n x+n
n L x = ∫ l x +t dt = ∫ l y dy = n ⋅ l x + n + n d x ⋅ E[T ( x) | T ( x) < n]
0 x
n n
d
5. [ n Lx ] = l x + n − l x =− n d x
dx
6. Tx = ∞ Lx the total number of years lived in the future by the l x people alive
at age x
d
7. n Lx = Tx − Tx + n and [Tx ] = −l x
dx
2Y x ∞
9. E[(T ( x )) 2 ] = where Y x = ∫0Tx +t dt
lx
dx
10. n m x = n
is the n-year central mortality weight
n Lx
qx qx
Then l x +t = l x − t ⋅ d x and μ ( x + t ) = and m x =
1 − t ⋅ qx 1 − 0.5q x
Defining V(x) to be the random variable representing the fraction of the year
lived in the year in which (x) dies, we can relate the random variables T and
K; namely, T(x) = K(x) + V(x), and K and V are independent.
t ⋅ qx qx
For 0 ≤ s + t ≤ 1 , t q x + s = and t p x + s ⋅ μ ( x + s + t ) =
1 − s ⋅ qx 1 − s ⋅ qx
For 0 ≤ s + t ≤ 1 , t p x + s = e − μ⋅t = ( p x ) t
mx = μ
1 1 ⎛ 1 1⎞
= + t ⎜⎜ − ⎟⎟
l x +t lx ⎝ l x +1 l x ⎠
qx t ⋅ qx
μ( x + t) = and for 0 ≤ s + t ≤ 1 , t qx + s = = t ⋅ μ( x + s + t)
1 − (1 − t ) ⋅ q x 1 − (1 − s − t ) ⋅ qx
Select mortality rates are used for a period (usually 3 years or less) and are
different than the ultimate (general population) rates.
[x] denotes an x-year old for which select rates are used starting at age x
[x] + k denotes an (x + k)-year old for which select rates are used starting at
age x
[x + k] denotes an (x + k)-year old for which select rates are used starting at
age x + k
(x + k) denotes an (x + k)-year old for which ultimate rates are used starting
at age x + k
Comments:
1. [x] + k = (x + k) if k exceeds the select period
Multiple-Life Models
For the moment, we consider only two independent lives (x) and (y). The
models can easily be extended to more than two lives.
t q xy = 1− t p xy
0 ∞
e xy = ∫ t p xy dt
0
∞
e xy = ∑ k p xy
k =1
t |u q xy = t p xy − t +u p xy
μ xy (t ) = μ x (t ) + μ y (t )
t p xy = 1− t q xy
t |u q xy = t +u q xy − t q xy
p x ⋅ μ x (t )+ t p y ⋅ μ y (t )− t p xy ⋅ μ xy (t )
μ xy (t ) =
t
t p xy
Contingent Probabilities
Notation:
n q x1 y = Pr((x) dies first and within the next n years)
1
The notation x y indicates that the joint-life status (xy) fails due to the failure
of the (x) status, ie the death of (x).
1. n q x = n q x1 y + n q x2 y
2. n q y = n q x y1 + n q x y2
3. n q xy = n q x1 y + n q x 1y
4. n q xy = n q x2 y + n q x 2y
1 1
Given: T * ( x ) ~ EX (mean = ) and T * ( y ) ~ EX ( mean = )
μx μy
− μ y ⋅t
Formulas: t p*x = e − μ x ⋅t and t p*y = e
− ( μ y + λ ) ⋅t
t p x = e − ( μ x + λ )⋅t and t py = e
− ( μ x + μ y + λ ) ⋅t
t p xy = e
Special Probability:
Pr((x) and (y) die within n years by the common shock)
=∫
0
n
t
n
p xy ⋅ λdt = ∫ λ ⋅ e
0
− ( μ x + μ y + λ ) ⋅t
dt =
λ
μx + μ y + λ
⋅ 1− e
−(μ
( x + μ y + λ )⋅n
)
Paris’s Exam MLC Seminar Page 12 of 52
www.steveparisseminars.com
Life Contingencies
Zx Probability
v Pr(K = 0) = q x
v2 Pr(K = 1) = 1| q x
v3 Pr(K = 2) = 2| q x
M M
PVRV = Z x = v K +1
2. n-year term insurance (benefit is paid if death occurs within next n years)
1
Z x:n | Probability
v Pr(K = 0) = q x
2
v Pr(K = 1) = 1| q x
v3 Pr(K = 2) = 2| q x
M M
vn Pr(K = n − 1 ) = n −1| qx
0 Pr(K ≥ n ) = n p x
⎧v K +1 L K < n
PVRV = Z x:n | = ⎨ 1
⎩0 L K ≥ n
1
The probability distribution table for the PVRV, Z x:n | , is
Z x:n |
1
Probability
0 Pr( K < n ) = n q x
v n
Pr( K ≥ n ) = n p x
1 ⎧0 L K < n
PVRV = Z x:n | = ⎨
⎩v L K ≥ n
n
Z x :n | Probability
v Pr(K = 0) = q x
v2 Pr(K = 1) = 1| q x
v3 Pr(K = 2) = 2| q x
M M
vn Pr(K = n − 1 ) = n −1| qx
vn Pr(K ≥ n ) = n p x
⎧v K +1 L K < n
PVRV = Z x:n | = ⎪⎨
1
= Z x:n | + Z x:n |
1
⎪⎩v n L K ≥ n
(iii) WARNING: Var ( Z x:n | ) ≠ Var ( Z x:n | ) + Var ( Z x:n | ) It is easy to show that if
1 1
n| Zx Probability
0 Pr( K < n ) = n q x
v n +1 Pr(K = n) = n| q x
v n+2 Pr(K = n + 1) = n +1| q x
M M
⎧0L K < n
PVRV = n | Z x = ⎨ = Z x − Z x :n |
1
K +1
⎩v LK ≥ n
(iii) The ideas above can be extended to an n-year deferred, j-year term
insurance, or an n-year deferred, j-year pure endowment, or a … If you
are tested on one of these other insurance types, just use the basic
principals learned by studying the above insurances.
( IZ ) x Probability
v Pr(K = 0) = q x
2v 2 Pr(K = 1) = 1| q x
3v 3 Pr(K = 2) = 2| q x
M M
PVRV = ( IZ ) x = ( K + 1) ⋅ v K +1
( IZ ) x1 :n | Probability
v Pr(K = 0) = q x
2v 2 Pr(K = 1) = 1| q x
3v 3 Pr(K = 2) = 2| q x
M M
nv n Pr(K = n − 1 ) = n −1| qx
0 Pr(K ≥ n ) = n p x
⎧( K + 1) ⋅ v K +1 L K < n
PVRV = ( IZ ) x:n | = ⎨ 1
⎩0 L K ≥ n
SBP = APV =
E[( IZ ) x1 :n | ] = ( IA) x1 :n | = v ⋅ qx + 2v 2 ⋅ p x ⋅ qx +1 + 3v 3 ⋅2 p x ⋅ qx +1 + L + nv n ⋅n −1 p x ⋅ qn + n −1
( DZ ) x1 :n | Probability
nv Pr(K = 0) = q x
(n − 1)v 2
Pr(K = 1) = 1| q x
(n − 2)v 3 Pr(K = 2) = 2| q x
M M
vn Pr(K = n − 1 ) = n −1| qx
0 Pr(K ≥ n ) = n p x
⎧(n − K ) ⋅ v K +1 L K < n
PVRV = ( DZ ) x:n | = ⎨1
⎩0 L K ≥ n
SBP = APV = E[( DZ ) x:n | ] = ( DA) x:n | = Ax:n | + Ax:n −1| + Ax:n − 2| + L + Ax:1|
1 1
1 1 1 1
PVRV = Z x = v T = e −δ ⋅T
∞
SBP = APV = E[ Z x ] = Ax = ∫0 e −δ ⋅t ⋅t p x ⋅ μ x (t )dt
∞
E[ Z x ]= Ax = ∫ e − 2δ ⋅t ⋅t p x ⋅ μ x (t )dt
2 2
2
Therefore, Var ( Z x )= Ax − ( Ax ) 2 .
⎧v T LT < n
PVRV = Z x:n | = ⎨ 1
⎩0LT ≥ n
n n
SBP = APV = E[ Z x:n | ] = Ax:n | = ∫0 v t ⋅ f T (t )dt = ∫0 e −δ ⋅t ⋅t p x ⋅ μ x (t )dt
1 1
n
E[( Z x :n | ) 2 ]= Ax :n | = ∫ e − 2δ ⋅t ⋅t p x ⋅ μ x (t )dt
1 2 1
2
Therefore, Var ( Z x:n | ) = Ax:n | − ( Ax:n | ) 2 .
1 1 1
⎧v T LT < n
PVRV = Z x:n | = ⎨ = Z x1 :n | + Z x :n |
1
⎩v LT ≥ n
n
2 2
E[( Z x :n | ) 2 ]= Ax :n | = Ax :n | + 2 Ax :n |
1 1
2
Therefore, Var ( Z x:n | ) = Ax:n | − ( Ax:n | ) 2 .
Comments and Concepts: (Comments (ii) – (iv) are analogues to the same
numbered comments in the n-year endowment insurance payable at the end
of the year of death case – “the discrete case”.)
(i) Notice that the pure endowment random variable is always discrete,
1
and so we do not have a Z x:n | random variable. Subsequently, there are no
1 2 1
actuarial symbols Ax:n | or Ax:n | .
(ii) On a random variable level, ( Z x:n | ) 2 = ( Z x:n | ) 2 + ( Z x:n | ) 2 since Z x:n | ⋅ Z x:n | = 0 .
1 1 1 1
(iii) WARNING: Var ( Z x:n | ) ≠ Var ( Z x:n | ) + Var ( Z x:n | ) It is easy to show that if
1 1
(iv) We do not get the nice formula in the n = 1 case that we got in the
discrete case.
⎧0LT < n
PVRV = n | Z x = ⎨ = Z x − Z x:n |
1
⎩v LT ≥ n
T
2 2 2
E[( n | Z x ) 2 ]= n | Ax = Ax − Ax1 :n |
2
Therefore, Var ( n | Z x )= n | Ax −( n | Ax ) 2 .
2 2
(ii) Remembering the meaning of 2 A , we also get n| A x = v 2 n ⋅n p x ⋅ A x +n
(iii) The ideas above can be extended to an n-year deferred, j-year term
insurance, or an n-year deferred, j-year pure endowment, or a … If you
are tested on one of these other insurance types, just use the basic
principals learned by studying the above insurances.
Ax = A x:1| + vp x Ax +1 = vq x + vp x Ax +1
1
The next three are the continuous analogues to the previous three.
A x = A x:1| + vp x A x +1
1
⎧bT v T L T < n
Z =⎨ (n-year term)
⎩0 L T ≥ n
M
M
2
Z = bT2 v 2T (whole life)
2 ⎧bT2 v 2T L T < n
Z =⎨ (n-year term)
⎩0 L T ≥ n
M
M
Notice that E[Z2] will not equal 2A unless the benefit is 1.
Y&&x Probability
a&&1| Pr(K = 0) = q x
a&&2| Pr(K = 1) = 1| q x
M M
K +1
1− v 1 − Zx
PVRV = Y&&x = a&&K +1| = =
d d
1 − Z x ⎤ 1 − Ax
(i) a&&x = E[Y&&x ] = E ⎡⎢ ⎥= (equivalently, Ax = 1 − d ⋅ a&&x )
⎣ d ⎦ d
⋅ Var ( Z x ) = 2 ⋅ (2 Ax − ( Ax ) 2 )
1 1
(ii) Var (Y&&x ) = 2
d d
2. n-year temporary life annuity due (pays 1 until the earlier of the death of
the annuitant or an n-year certain period)
Y&&x :n | Probability
a&&1| Pr(K = 0) = q x
a&&2| Pr(K = 1) = 1| q x
a&&3| Pr(K = 2) = 2| q x
M M
a&&n| Pr(K = n − 1 ) = n −1| qx
a&&n| Pr(K ≥ n ) = n p x
⎧1 − v K +1
⎧a&&K +1| L K < n ⎪⎪ d L K < n 1 − Z x :n |
PVRV = Y&&x:n | = ⎪⎨ =⎨ =
⎪⎩a&&n| L K ≥ n ⎪1 − v L K ≥ n
n
d
⎩⎪ d
SBP = APV =
E[Y&&x :n | ] = a&&x :n | = a&&1| ⋅ qx + L + a&&n | ⋅n −1| qx + a&&n | ⋅n p x = 1 + vp x + v 2 ⋅2 p x + L + v n −1 ⋅n −1 p x
⎡1 − Z x :n | ⎤ 1 − Ax :n |
(i) a&&x:n | = E[Y&&x:n | ] = E ⎢ ⎥= (equivalently, Ax:n| = 1 − d ⋅ a&&x:n| )
⎣ d ⎦ d
n| x Y&& Probability
0 Pr(K < n) = n q x
n| a&&1| Pr(K = n) = n| q x
n| a&&2| Pr(K = n+1) = n +1| qx
M M
⎧⎪0L K < n
PVRV = n |Y&&x = n| a&&K +1−n| = ⎨ = Y&&x − Y&&x:n |
⎪⎩a&&K +1| − a&&n| L K ≥ n
4. n-year certain and life annuity due (pays 1 until the later of the death of
the annuitant or an n-year certain period)
Y&&x :n | Probability
a&&n| Pr(K < n) = n q x
a&&n +1| Pr(K = n) = n| q x
a&&n +2| Pr(K = n+1) = n +1| q x
M M
⎧a&& L K < n
PVRV = Y&&x:n | = a&&n | + n |Y&&x = ⎪⎨
n|
⎪⎩a&&K +1| L K ≥ n
Observe that an n-year certain and life annuity due pays until the failure of
the last-survivor status x : n | (here, the “life” y is n | , an n-year certain
period). That is, the annuity pays until the later of the death of (x) and the
death of n | , or equivalently, it pays as long as (x) is alive, with a minimum
of n years.
When dealing with annuities immediate, use the above formulas for
annuities due and relationships between annuities immediate and annuities
due.
Examples:
3. If n |Yx is the PVRV for an n-year deferred life annuity immediate, then
&& Then SBP = APV = E[ Y ]= a = a&& and
n |Yx = n +1|Yx . n| x n | x n +1| x
1 − vT 1− Z x
PVRV = Y x = a T | = =
δ δ
∞ ∞
SBP = APV = E[Y x ] = a x = ∫0 a t | ⋅t p x ⋅ μ x (t )dt = ∫0 v t ⋅t p x dt
⎡1 − Z x ⎤ 1 − A x
(i) a x = E[Y x ] = E ⎢ ⎥= δ (equivalently, A x = 1 − δ ⋅ a x )
⎣ δ ⎦
(ii) Var (Y x ) =
δ
1
2
⋅ Var ( Z x ) =
δ
1
2
(
2
⋅ Ax − ( Ax )2 )
0
(iii) If i = 0, then a x = e x .
⎧a T | LT < n 1 − Z x :n |
PVRV = Y x:n | = ⎪⎨ =
⎪⎩a n | LT ≥ n δ
n n
SBP = APV = E[Y x:n | ] = a x:n | = ∫0 a t | ⋅t p x ⋅ μ x (t )dt + a n | ⋅n p x = ∫0 v t ⋅t p x dt
⎡1 − Z x :n | ⎤ 1 − A x :n |
(i) a x:n | = E[Y x:n | ] = E ⎢ ⎥= δ (equivalently, A x:n| = 1 − δ ⋅ a x:n| )
⎣ δ ⎦
∞ ∞
SBP = APV = E[ n |Y x ]= n | a x = a x − a x:n | = ∫n v n ⋅ a t − n | ⋅t p x ⋅ μ x (t )dt = ∫n v t ⋅t p x dt
(i) n| a x = v n ⋅n p x ⋅ a x + n
(iii) E[( n |Y x ) 2 ] =
δ
2
( 2
⋅ v 2 n ⋅n p x ⋅ a x + n − a x + n ) (unlikely to see this on the exam)
⎧a n | LT < n
PVRV = Y x:n | = a n | + n |Y x = ⎪⎨
⎪⎩a T | LT ≥ n
These annuities also have an annual payment of 1. Thus the symbol C ⋅ a&&x( m )
C
represents the APV of a life annuity to (x) that pays at the beginning of
m
each of m periods per year for the life of (x). The total annual payment each
year is C.
where the values of α (m) and β (m) will be given in the tables
a&&x = 1 + v ⋅ p x ⋅ a&&x +1
a&&x:n | = 1 + v ⋅ p x ⋅ a&&x +1:n −1|
a x = a x:1| + v ⋅ p x ⋅ a x +1
a x:n | = a x:1| + v ⋅ p x ⋅ a x +1:n −1|
a x = v ⋅ p x + v ⋅ p x ⋅ a x +1 = v ⋅ p x ⋅ (1 + a x +1 )
Interchange the roles of (x) and (y) to get similar formulas for insurance
payable on the death of (y).
1. A x = A x1 y + A x2 y
2. A y = A x y1 + A x y2
3. A xy = A x1 y + A x 1y
4. A xy = A x2 y + A x y2
∞
A xy = E[v 2T ( xy ) ] = ∫ v 2 t ⋅t p xy ⋅ μ xy (t )dt
2
2. 0
2
So Var ( Z xy ) = Axy − ( Axy ) 2 = variance of PVRV for insurance that pays 1 at
the moment of the first death of (x) or (y)
∞ 1 − A xy
3. a xy = ∫0 v t ⋅t p xy dt =
δ
n 1 − A xy:n|
4. a xy:n| = ∫0 v t ⋅t p xy dt =
δ
This is the APV of an n-year temporary life annuity that pays continuously
at a rate of 1 per year for the joint lifetimes of (x) and (y). This annuity
pays until failure of the xy : n | status, which fails at the first of the death
of (x), the death of (y), and the death of n | . That is, the annuity pays until
the first of the death of (x) or the death of (y), up to a maximum of n years.
n
5. A xy:n| = ∫0 v t ⋅t p xy ⋅ μ xy (t )dt
1
6. Var (Y xy ) =
1
δ 2
(A 2
xy )
− ( Axy ) 2 = variance of PVRV for a continuous annuity
that pays 1 per year for as long as the joint-life status (xy) survives, i.e. as
long as both (x) and (y) are alive.
M
You get the idea.
∞
1. Axy = E[v K ( xy )+1 ] = ∑ v k +1 ⋅k | q xy
k =0
∞
2. 2
Axy = E[v 2 ( K ( xy ) +1) ] = ∑ v 2 ( k +1) ⋅k | q xy
k =0
So Var ( Z xy ) = Axy − ( Axy ) 2 = variance of PVRV for insurance that pays 1 at the
2
end of the year of the first of the death of (x) or the death of (y)
n −1
3. Axy:n| = A xy:n| + v n ⋅n p xy = ∑ v k +1 ⋅k | q xy + v n ⋅n p xy
1
k =0
∞ 1 − Axy
4. a&&xy = ∑ v k ⋅k p xy =
k =0 d
n −1 1 − Axy:n|
5. a&&xy:n| = ∑ v k ⋅k p xy =
k =0 d
6. Var (Y&&xy :n | ) =
1
d2
(A
2
xy : n |
)
− ( Axy :n | ) 2 = variance of PVRV for an n-year temporary
life annuity due that pays 1 at the beginning of each year that the joint-life
status xy : n | survives, i.e. as long as both (x) and (y) are alive, up to a
maximum of n years.
M
You get the idea.
Method 2: (This can make many computations much easier.) Make use of the
following relationships.
1. t p xy + t p xy = t p x + t p y
2. t q xy + t q xy = t q x + t q y
0 0 0 0
3. e xy + e xy = e x + e y
4. e xy + e xy = e x + e y
5. Axy + Axy = Ax + Ay
6. a xy + a xy = a x + a y
10. n| a xy + n| a xy = n| a x + n| a y
⎝ ⎠ ⎝ ⎠
There are formulas in the continuous case too ( a ’s and A ’s), and there are
other formulas, e.g. n-year deferred insurance, pure endowments, ….
We focus on the single life case with issue age x. The ideas can be extended
to the multiple life case, as will be seen in some of the examples that we’ll do.
Examples:
1a. Fully Continuous Whole Life Insurance of 1 with premiums of Q for life
1− Z x
⎛ Q⎞ Q
L = Z x − Q ⋅Y x = Z x − Q ⋅ = ⎜1 + ⎟ ⋅ Z x −
δ ⎝ δ⎠ δ
⎛ Q⎞ Q
E [ L] = A x − Q ⋅ a x = ⎜1 + ⎟ ⋅ A x −
⎝ δ⎠ δ
( ) ( )
2 2
⎛ Q⎞ ⎛ Q⎞ 2
Var ( L) = ⎜1 + ⎟ ⋅ Var Z x = ⎜1 + ⎟ ⋅ Ax − ( Ax ) 2
⎝ δ⎠ ⎝ δ⎠
1b. Fully Discrete Whole Life Insurance of 1 with premiums of Q for life
1 − Zx ⎛ Q ⎞ Q
L = Z x − Q ⋅ Yx = Z x − Q ⋅ = ⎜1 + ⎟ ⋅ Z x −
d ⎝ d⎠ d
⎛ Q⎞ Q
E[ L] = Ax − Q ⋅ a&&x = ⎜1 + ⎟ ⋅ Ax −
⎝ d⎠ d
2 2
Var ( L) = ⎜1 + ⎟ ⋅ Var (Z x ) = ⎜1 + ⎟ ⋅ (2 Ax − ( Ax ) 2 )
⎛ Q⎞ ⎛ Q⎞
⎝ d⎠ ⎝ d⎠
( ) ( )
2 2
⎛ Q⎞ ⎛ Q⎞ 2
Var ( L) = ⎜1 + ⎟ ⋅ Var Z x :n | = ⎜1 + ⎟ ⋅ Ax:n | − ( Ax :n | ) 2
⎝ δ⎠ ⎝ δ⎠
( ) ( )
2 2
⎛ Q⎞ ⎛ Q⎞
Var ( L) = ⎜1 + ⎟ ⋅ Var Z x:n | = ⎜1 + ⎟ ⋅ 2 Ax :n | − ( Ax:n | ) 2
⎝ d⎠ ⎝ d⎠
You get the idea. There are many combinations of the type of insurance
purchased and the method of paying premiums. We use one more example
to illustrate the concept of the loss-at-issue random variable.
L = Z x :n | − Q ⋅ Y&&x :h |
E [ L] = A x:n| − Q ⋅ a&&x:h|
1
Calculating Premiums
Method 1: Percentiles
The premium Q is found by solving a probability equation such as
Pr( L > 0) = 0.05 (probability of a positive loss is 5%). This equation can
be solved by writing the event first in terms of the random variable Z and
then in terms of the random variable T (or K). Then use the distribution
of T (or K) to solve the probability equation. We will illustrate this with
examples.
Ax
Px = = whole life insurance with premiums paid for life
a&&x
A x1 :n|
P x1 :n| = = n-year term insurance with at most n premiums
a&&x:n|
Ax:n|
h Px:n| = = h-payment n-year endowment insurance
a&&x:h|
M
You get the idea.
Ax
P( A x ) = = whole life insurance with premiums paid for life
ax
Ax
P ( n| A x ) =
n|
k = k-payment n-year deferred whole life insurance
a x:k |
M
You get the idea.
Ax
P( A x ) = = whole life insurance with premiums paid for life
a&&x
A x1 :n|
h P ( A x:n| ) = = h-payment n-year term insurance
1
a&&x:h|
M
You get the idea. Notice the notation suggests, as is correct, that
the benefit is paid at the moment of death, and the premiums are
paid at the beginning of each year.
Ax
Px(12 ) = = fully discrete whole life insurance with premiums
a&&x(12 )
Px(12 )
of paid at the beginning of each month for life
12
A x:n|
P ( 4 ) ( A x:n| ) = = semi-continuous n-year endowment insurance
a&&x( 4:n)|
P ( 4 ) ( A x:n| )
with premiums of paid at the beginning
4
of each quarter until death of the insurance expires
M
You get the idea.
Ax
1. Fully Continuous Whole Life - Q = P( A x ) =
ax
⎛ Q⎞ Ax δ ⋅ a x + Ax 1 1
⎜1 + ⎟ = 1 + = = =
⎝ δ⎠ δ ⋅ax δ ⋅ax δ ⋅ a x 1 − Ax
A x:n|
2. Fully Continuous n-year Endowment Insurance - Q = P( A x:n| ) =
a x:n|
Ax
3. Fully Discrete Whole Life - Q = Px =
a&&x
⎛ Q⎞ Ax d ⋅ a&&x + Ax 1 1
⎜1 + ⎟ = 1 + = = =
⎝ d⎠ d ⋅ a&&x d ⋅ a&&x d ⋅ a&&x 1 − Ax
Ax:n|
4. Fully Discrete n-year Endowment Insurance - Q = Px:n| =
a&&x:n|
V = E[ t L | T ( x) ≥ t ]
t ( tV = E[ t L | K ( x) ≥ t ] in the discrete case)
Benefit reserves means the premiums are the benefit premiums. Benefit
reserves can be calculated either prospectively or retrospectively.
V x = Ax +t − Px a&&x + t
t
V x:n| = A x +t:n −t| − P x:n| a&&x +t:n −t| (Important Fact: V x:n| = 1 )
1 1 1 1
t n
V x:n| = Ax +t:n −t| − Px:n| a&&x + t:n −t| = t V x1 :n| + t V x:n| (Important Fact: nV x:n| = 1 )
1
V ( A x ) = A x +t − P ( A x ) ⋅ a&&x +t
t
M
You get the idea.
a&&x +t P − P x:t|
1
1. tV x = 1 − = ( Px + t − Px ) ⋅ a&&x +t = x
a&&x 1
P x:t|
a&&x + t :n − t |
2. tVx:n| = 1 − = ( Px + t :n − t | − Px:n| ) ⋅ a&&x + t :n − t |
a&&x:n|
P( A x ) − P( A x:t| )
1
a
3. t V ( A x ) = 1 − x +t = [ P( A x +t ) − P( A x )] ⋅ a x +t = 1
ax P( A x:t| )
a x + t :n − t |
4. t V ( A x:n| ) = 1 − = [ P ( A x + t :n − t | ) − P ( A x:n| )] ⋅ a x + t :n − t |
a x:n |
Calculate terminal reserves for the tth year by calculating reserves at the end
of the tth year. In a prospective calculation, the benefit for the tth year will
not be included, but the premium for the (t+1)st year is included.
Calculate initial reserves for the tth year by calculating reserves at the
beginning of the tth year. In a prospective calculation, the premium for the tth
year will not be included.
The initial reserves for the (t+1)st year will exceed the terminal reserves for
the tth year by the amount of premium paid at time t.
(A ) (A )
2
⎛ P( A x ) ⎞
2
⎛ 1 ⎞
Var ( t L | T ( x) ≥ t ) = ⎜⎜1 + ⎟ ⋅
2 2
− ( A x +t ) = ⎜2
⎟ ⋅ − ( A x +t ) 2
δ ⎟⎠
x +t x +t
⎝ ⎝1 − Ax ⎠
(A )
2
⎛ P ( A x :n | ) ⎞
Var ( t L | T ( x ) ≥ t ) = ⎜⎜ 1 + ⎟ ⋅
2
⎟ x + t :n − t | − ( A x + t :n − t | ) 2
⎝ δ ⎠
(A )
2
⎛ 1 ⎞ 2
= ⎜⎜ ⎟⎟ ⋅ x + t :n − t | − ( A x + t :n − t | ) 2
⎝ 1 − A x :n | ⎠
tV = bt +1 ⋅ v ⋅ q x +t − Qt + t +1V ⋅ v ⋅ p x +t , where
se = settlement expenses
Et = BOY expenses paid at time t
Qt = expense loaded premium at time t (usually more in 1st year)
1. Generally, part of the expense loaded premium will depend on the face
amount of the policy. The amount of the expense loaded premium that
does not depend on the face amount of the policy is called the policy fee.
3. The expense loaded premium pays expenses, but not any profit. The
contract premium is charged in order to expect a profit.
V
t ben = reserves as before in the non-expense model (using the premiums Qt)
V
t exp = expense reserves
= APVF(expenses)x+t – APVF(expense loadings)x+t
Notation:
t q x( j ) = Pr((x) departs within t years by decrement j)
t p x( j ) has no meaning
l x(τ ) = the total number that have survived all decrements to age x
∞ d x(τ ) = l x(τ )
∞ d x( j ) = l x( j )
1. f T ,J (t , j )= t p x(τ ) ⋅ μ x( j ) (t )
2. f T (t )= t p x(τ ) ⋅ μ x(τ ) (t )
∞
3. f J ( j ) = ∫0 t p x(τ ) ⋅ μ x( j ) (t )dt = ∞ q x( j )
t d x(τ )
4. t q x(τ ) = ∫0 s p x(τ ) ⋅ μ x(τ ) ( s )ds = t
l x(τ )
t d x( j )
5. t q x( j ) = ∫0 s p x(τ ) ⋅ μ x( j ) ( s )ds = t
l x(τ )
⎧ t p x(τ ) − t +u p x(τ )
⎪ d (τ )
6. t |u q x(τ ) = ⎨ t +u q x(τ ) − t q x(τ ) = u (τx)+t
⎪ p (τ ) ⋅ q (τ ) lx
⎩ t x u x +t
(τ ) d x( +j )t
7. t |u q =t p ⋅ q
( j)
x x u
( j)
x +t = (τ ) u
lx
0 (τ ) ∞
8. E[T ] = e x = ∫0 t p x(τ ) dt = expected time until decrement
μ x( j ) (t )
10. Pr( J = j | T = t ) = f J |T ( j | t ) =
μ x(τ ) (t )
q x( j )
11. Pr( J = j | T ≤ t ) = t
(τ )
t qx
f T , J (t , j )
12. f T |J (t | j ) =
f J ( j)
∞ f T , J (t , j )
13. E[T | J = j ] = ∫0 t ⋅ dt = expected time until departure, given the cause
f J ( j)
is by decrement j
Probability Formulas:
d
t
− [ t p ′x( j ) ]
p ′x( j ) = exp(− ∫ μ x( j ) ( s )ds) ⇒ μ x( j ) (t ) = dt
t
0
t px
′( j )
t
t q ′x( j ) = 1− t p ′x( j ) = ∫ s p ′x( j ) ⋅ μ x( j ) ( s )ds
0
t p x(τ ) = ∏ t p ′x( j )
j
Two Cases:
Case 1: MUDD
( t qx( j ) = t ⋅ qx( j ) ; UDD assumption in the multiple decrement model)
q x( j )
(τ )
Important Formula: s p′
x
( j)
=( s p )
x
q x( t )
Case 2: SUDD
( t q′x( j ) = t ⋅ q′x( j ) ; UDD assumption in the associated single decrement model)
⎛ 1 1 ⎞
q x(1) = q′x(1) ⋅ ⎜1 − ( q′x( 2 ) + q′x( 3) ) + q′x( 2 ) ⋅ q′x( 3) ⎟ (3 decrement case)
⎝ 2 3 ⎠
⎛ 1 1 ⎞
q x( 2 ) = q′x( 2 ) ⋅ ⎜1 − ( q′x(1) + q′x( 3) ) + q′x(1) ⋅ q′x( 3) ⎟
⎝ 2 3 ⎠
Timing of Decrements
If some decrements happen at a fixed point in time, then calculate 1-year
mortality probabilities in the associated single decrement table by using the
d x( j )
formula q ′x( j ) = ( j)
, where NAR x( j ) is the number at risk for decrement j, at
NAR x
d x( j )
age x. Note that q ( j)
x = always.
lx
Asset Shares
This concept is based on a double decrement model, death and withdrawal. The
death benefit at time h is denoted by h DB . If withdrawal occurs at time h, then
a cash value, denoted h CV , is paid at that time. We let h AS denote the asset
share at time h. Assume 0 AS = 0 . There are two main formulas.
Recursion Formula:
APV0 ( n AS ) = v n ⋅n p x(τ ) ⋅n AS
= APV0(Premiums paid from time 0 through n-1, backing
out non-settlement expenses)
– APV0(Benefits + settlement expenses from time 1 through n)