chapter 3 survival distributions and life tables exam m notes f05
TRANSCRIPT
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
1/26
Chapter 3: Survival Distributions and Life Tables
Distribution function of X :
F X (x) = Pr( X x)
Survival function s(x):
s(x) = 1 F X (x)
Probability of death between age x andage y:
Pr( x < X z) = F X (z) F X (x)= s(x) s(z)
Probability of death between age x and
age y given survival to age x:
Pr( x < X z |X > x ) = F X (z) F X (x)
1 F X (x)
= s(x) s(z)
s(x)
Notations:
t q x = Pr[ T (x) t]= prob. ( x) dies within t years= distribution function of T (x)
t px = Pr[ T (x) > t ]= prob. ( x) attains age x + t= 1 tq x
t |u q x = Pr[ t < T (x) t + u]= t+ u q x tq x= t px t+ u px= t px uq x + t
Relations with survival functions:
t px = s(x + t)
s(x)
t q x = 1 s(x + t)
s(x)
Curtate future lifetime ( K (x) greatestinteger in T (x)):
Pr[K (x) = k] = Pr[ k T (x) < k + 1]= k px k+1 px
= k px q x + k= k| q x
Force of mortality (x):
(x) = f X (x)
1 F X (x)=
s (x)s(x)
Relations between survival functions andforce of mortality:
s(x) = exp x
0
(y)dy
n px = exp x + n
x(y)dy
Derivatives:
ddt t
q x = t px (x + t) = f T (x ) (t)
ddt t
px = t px (x + t)
ddt
T x = lxddt
Lx = dx
ddt
ex = (x)ex 1
Mean and variance of T and K :
E [T (x)] complete expectation of life
ex =
0
t px dt
E [K (x)] curtate expectation of life
ex
=
k=1 k p
x
V ar[T (x)] = 2
0
t t px dt e2x
V ar[K (x)] =
k=1
(2k 1) k px e2x
Total lifetime after age x: T x
T x =
0 lx + t dtExam M - Life Contingencies - LGD c 1
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
2/26
Total lifetime between age x and x + 1 : Lx
Lx = T x T x +1
=1
0lx + t dt =
1
0lx t px dt
Total lifetime from age x to x + n: n Lx
n Lx = T x T x + n =n 1
k=0
Lx + k
=n
0
lx + t dt
Average lifetime after x: ex
ex = T x
lx
Average lifetime from x to x + 1 : ex : 1
ex : 1 = Lxlx
Median future lifetime of (x): m (x)
P r [T (x) > m (x)] = s(x + m(x))
s(x) =
12
Central death rate: m x
mx = lx lx +1
Lx
n mx = lx lx + n
n LxFraction of year lived between age x andage x + 1 by dx : a(x)
a(x) =
1
0 t t px (x + t) dt1 0 t px (x + t) dt
= Lx lx +1lx lx +1
Recursion formulas:
E [K ] = ex = px (1 + ex +1 )E [T ] = ex = px (1 + ex +1 ) + q x a(x)
ex = ex : n + n px ex + nex = ex : n + n px ex + n
E [K (m + n)] = ex : m + n= ex : m + m px ex + m : n
E [T (m + n)] = ex : m + n= ex : m + m px ex + m : n
Exam M - Life Contingencies - LGD c 2
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
3/26
Chapter 4: Life Insurance
Whole life insurance: Ax
E [Z ] = Ax =
0 vt t px x (t)dt
V ar[Z ] = 2Ax ( Ax )2
n-year term insurance: A1x : n
E [Z ] = A1x : n =n
0
vt t px x (t)dt
V ar[Z ] = 2
A1x : n ( A
1x : n )
2
m-year deferred whole life: m | Ax
E [Z ] = m | Ax =
m
vt t px x (t)dt
m | Ax = Ax A1x : m
n-year pure endowment: A 1x : n
E [Z ] = A 1x : n = vn n px n E x
V ar[Z ] = 2A 1x : n ( A 1
x : n )2 = v2n n px n q x
n-year endowment insurance: Ax : n
E [Z ] = Ax : n =n
0vt t px x (t)dt + vn n px
V ar[Z ] = 2Ax : n ( Ax : n )2
Ax : n = A1x : n + A 1
x : n
m-yr deferred n-yr term: m |n Ax
m |n Ax = m E x A 1x + m : n= A1x : m + n A
1x : m
= m | Ax m + n | Ax
Discrete whole life: Ax
E [Z ] = Ax =
k=0
vk+1 k px q x + k
=
k=0
vk+1 k| q x
V ar[Z ] = 2Ax (Ax )2
Discrete n-year term: A1x : n
E [Z ] = A1x : n =n 1
k=0
vk+1 k px q x + k
V ar[Z ] = 2A1x : n (A1x : n )
2
Discrete n-year endowment: Ax : n
E [Z ] = Ax : n =n 1
k=0
vk+1 k px q x + k + vn n px
V ar[Z ] = 2Ax : n ( Ax : n )2
Ax : n = A1x : n + A 1
x : n
Recursion and other relations:
Ax =
A
1
x : n + n | Ax2Ax = 2A1x : n + n |2Ax
n | Ax = n E x Ax + nAx = A1x : n + n E x Ax + nAx = vq x + vpx Ax +1
2Ax = v2q x + v2 px 2AxA1x : n = vq x + vpx A
1x +1: n 1
m |n Ax = vpx (m 1)|n Ax +1Ax : 1 = v
Ax : 2 = vq x + v2 px
Exam M - Life Contingencies - LGD c 3
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
4/26
Varying benet insurances:
(I A)x =
0
t + 1 vt t px x (t)dt
(I A)1x : n =n
0
t + 1 vt t px x (t)dt
(IA)x =
0
t vt t px x (t)dt
(IA)1x : n =n
0
t vt t px x (t)dt
(D A)1x : n =
n
0 (n t )vt
t px x (t)dt
(DA )1x : n =n
0
(n t)vt t px x (t)dt
(IA)x = Ax + vpx (IA)x +1= vq x + vpx [(IA)x +1 + Ax +1 ]
(DA )1x : n = nvq x + vpx (DA ) 1
x +1: n 1
(IA)1x : n + ( DA )1x : n = n A
1x : n
(I A)1x : n + ( D A)1x : n = ( n + 1) A
1x : n
(IA)1x : n + ( DA )1x : n = ( n + 1) A
1x : n
Accumulated cost of insurance:
n kx =A1x : nn E x
Share of the survivor:
accumulation factor = 1n E x
= (1 + i)nn px
Interest theory reminder
a n = 1 vn
i
a n = 1 vn
=
i
a n
a = 1
, a =
1i
, a = 1d
(Ia ) n = a n nv n
i
(Da ) n = n a n
(Ia ) = 1
2(n + 1) a n = ( Ia ) n + ( Da ) n
s 1 = i
d = iv
(Ia ) = 1
id =
1 + ii2
Doubling the constant force of interest
1 + i (1 + i)2
v v2
i 2i + i2
d 2d d2
i
2i + i2
2
Limit of interest rate i = 0 :
Axi=0 1
A1x : ni=0 n q x
n | Axi=0 n px
Ax : ni=0 1
m |n Axi=0
m |n q x(IA)x
i=0 1 + ex(IA)x
i=0 ex
Exam M - Life Contingencies - LGD c 4
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
5/26
Chapter 5: Life Annuities
Whole life annuity: ax
ax = E [a T ] =
0 at t px (x + t)dt
=
0
vt t px dt =
0
t E x dt
V ar[a T ] =2Ax ( Ax )2
2
n-year temporary annuity: ax : n
ax : n =
n
0 vt
t px dt =
n
0t E x dt
V ar[Y ] =2Ax : n ( Ax : n )2
2
n-year deferred annuity: n | ax
n | ax =
n
vt t px dt =
n
t E x dt
n | ax = vn n px ax + n = n E x ax + n
V ar[Y ] = 2
v2n n px (ax + n 2ax + n ) (n | ax )2
n-yr certain and life annuity: ax : n
ax : n = ax + a n ax : n= a n + n | ax = a n + n E x ax + n
Most important identity1 = ax + Ax
ax = 1 Ax
Ax : n = 1 ax : n
2Ax : n = 1 (2 ) 2ax : n
ax = 1 Ax
d
ax : n = 1 Ax : n
d
1 = dax : n + Ax : n
Recursion relations
ax = ax : 1 + vpx ax +1ax = ax : n + n | axax = 1 + vpx ax +1
2ax = 1 + v2 px 2ax +1ax : n = 1 + vpx ax +1: n 1a (m )x : n = a
(m )x n E x a
(m )x + n
ax = vpx + vpx ax +1ax : n = vpx + vpx ax +1: n 1
(I a)x = 1 + vpx [(I a)x +1 + ax +1 ]= ax + vpx (I a)x +1
Whole life annuity due: ax
ax = E [a K +1 ] =
k=0
vk k px
V ar[a K +1 ] =2Ax (Ax )2
d2
n-yr temporary annuity due: ax : n
ax : n = E [Y ] =n 1
k=0
vk k px
V ar[Y ] =2Ax : n (Ax : n )2
d2
n-yr deferred annuity due: n | ax
n | ax = E [Y ] =
k= n
vk k px
= ax ax : n= n E x ax + n
n-yr certain and life due: ax : n
ax : n = ax + a n ax : n
= a n +
k= n
vk k px
= a n + n | ax
Exam M - Life Contingencies - LGD c 5
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
6/26
Whole life immediate: ax
ax = E [a K ] =
k=1
vk k px
ax =
1 (1 + i)Axi
m-thly annuities
a (m )x = 1 A(m )x
d(m )
V ar[Y ] =2A(m )x (A(
m )x )2
(d(m ) )2
a (m )x = a(m )x
1m
a (m )x : n
= a (m )x : n
1
m(1 n E x )
Accumulation function:
sx : n = ax : n
n E x=
n
0
1n t E x + t
Limit of interest rate i = 0 :
axi=0 ex
axi=0 1 + ex
axi=0 ex
ax : ni=0 ex : n
ax : ni=0 1 + ex : n 1
ax : ni=0 ex : n
Exam M - Life Contingencies - LGD c 6
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
7/26
Chapter 6: Benet Premiums
Loss function:Loss = PV of Benets PV of PremiumsFully continuous equivalence premiums(whole life and endowment only):
P ( Ax ) =Axax
P ( Ax ) = Ax1 Ax
P ( Ax ) = 1ax
V ar[L] = 1 +P
22Ax ( Ax )2
V ar[L] =2Ax ( Ax )2
( ax )2
V ar[L] =2Ax ( Ax )2
(1 Ax )2
Fully discrete equivalence premiums(whole life and endowment only):
P (Ax ) = Ax
ax= P x
P (Ax ) = dAx1 Ax
P (Ax ) = 1ax
d
V ar[L] = 1 + P d
22Ax (Ax )2
V ar[L] =2Ax (Ax )2
(dax )2
V ar[L] =2Ax (Ax )2
(1 Ax )2
Semicontinuous equivalence premiums:
P ( Ax ) =Axax
m-thly equivalence premiums:
P (m )# = A#a (m )#
h-payment insurance premiums:
h P ( Ax ) =A
xax : h
h P ( Ax : n ) =Ax : nax : h
h P x = Axax : h
h P x : n = Ax : n
ax : h
Pure endowment annual premium P 1x : n :it is the reciprocal of the actuarial accumulatedvalue sx : n because the share of the survivor whohas deposited P 1x : n at the beginning of each yearfor n years is the contractual $1 pure endow-ment, i.e.
P 1x : n sx : n = 1 (1)
P minus P over P problems:The difference in magnitude of level benet pre-miums is solely attributable to the investmentfeature of the contract. Hence, comparisons of
the policy values of survivors at age x + n maybe done by analyzing future benets:
( n P x P 1x : n )sx : n = Ax + n = n P x P 1x : n
P 1x : n
(P x : n n P x )sx : n = 1 Ax + n = P x : n n P x
P 1x : n
(P x : n P 1x : n )sx : n = 1 = P x : n P 1x : n
P 1x : n
Miscellaneous identities:
Ax : n = P ( Ax : n )
P ( Ax : n ) +
Ax : n = P x : nP x : n + d
ax : n = 1
P ( Ax : n ) +
ax : n = 1P x : n + d
Exam M - Life Contingencies - LGD c 7
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
8/26
Chapter 7: Benet Reserves
Benet reserve tV :The expected value of the prospective loss attime t.
Continuous reserve formulas:
Prospective: tV ( Ax ) = Ax + t P ( Ax )ax + t
Retrospective: tV ( Ax ) = P ( Ax )sx : t A1x : n
t E xPremium diff.: tV ( Ax ) = P ( Ax + t ) P ( Ax ) ax + t
Paid-up Ins.: tV ( Ax ) = 1 P ( Ax )
P ( Ax + t ) Ax + t
Annuity res.: tV ( Ax ) = 1 ax + t
ax
Death ben.: tV ( Ax ) =Ax + t Ax
1 Ax
Premium res.: tV ( Ax ) =P ( Ax + t ) P ( Ax )
P ( Ax + t ) +
Discrete reserve formulas:
k V x = Ax + k P k ax + k
k V x : n = P x + k : n k P x : n ax + k : n k
k V x : n = 1 P x : n
P x + k : n kAx + k : n k
t V x : n = P x : n sx : n tkx
k V x = 1 ax + k
ax
k V x = P x + k P x
P x + k + d
k V x = Ax + k Ax
1 Ax
h-payment reserves:ht V = Ax + t : n t h P ( Ax : n )ax + t : h t
hk V x : n = Ax + k : n k h P x : n ax + k : h k
hk V ( Ax : n ) = Ax + k : n k h P ( Ax : n )ax + k : h k
hk V
1( m )x : n = Ax + k : n k h P
(m )x : n a
(m )x + k : h k
Variance of the loss function
V ar[ t L] = 1 + P
22Ax + t ( Ax + t )2
V ar[ t L] =2Ax + t ( Ax + t )2
(1 Ax )2 assuming EP
V ar[ t L] = 1 +P
22Ax + t : n t ( Ax + t : n t )
2
V ar[ t L] =2Ax + t : n t ( Ax + t : n t )
2
(1 Ax : n )2 assuming EP
Cost of insurance: funding of the accumu-lated costs of the death claims incurred betweenage x and x + t by the living at t , e.g.
4kx = dx (1 + i)3 + dx +1 (1 + i)2 + dx +2 (1 + i) + dx +
lx +4
=A1x : 44E x
1kx = dxlx +1
= q x px
Accumulated differences of premiums:
( n P x P 1x : n )sx : n = n
n V x n V 1
x : n )= Ax + n 0 = Ax + n
( n P x P x )sx : n = nn V x n V x= P x ax + n
(P x : n P x )sx : n = n V x : n n V x= 1 n V x
( m P x : n m P x )sx : m = mm V x : n mm V x= A
x + m : n m Ax + m
Relation between various terminal re-serves (whole life/endowment only):
m + n + pV x = 1 (1 m V x )(1 n V x + m )(1 pV x + m + n )
Exam M - Life Contingencies - LGD c 8
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
9/26
Chapter 8: Benet Reserves
Notations:b j : death benet payable at the end of year of death for the j -th policy year j 1: benet premium paid at the beginning of the j -th policy yearbt : death benet payable at the moment of death t : annual rate of benet premiums payable continuously at tBenet reserve:
h V =
j =0
bh + j +1 v j +1 j px + h q x + h + j
j =0
h + j v j j px + h
t V =
0
bt + u vu u px + t x (t + u)du
0 t+ u vu u px + t du
Recursion relations:
h V + h = v q x + h bh +1 + v px + h h+1 V (h V + h )(1 + i) = q x + h bh +1 + px + h h+1 V (h V + h )(1 + i) = h+1 V + q x + h (bh +1 h+1 V )
Terminology:policy year h+1 the policy year from time t = h to time t = h + 1h V + h initial benet reserve for policy year h + 1h V terminal benet reserve for policy year hh +1 V terminal benet reserve for policy year h + 1
Net amount at Risk for policy year h + 1
Net Amount Risk bh +1 h+1 V
When the death benet is dened as a function of the reserve:For each premium P , the cost of providing the ensuing years death benet , based on the net amount atrisk at age x + h, is : vq x + h (bh +1 h+1 V ). The leftover, P vq x + h (bh +1 h+1 V ) is the source of reservecreation. Accumulated to age x + n, we have:
n V =n 1
h =0
[P vq x + h (bh +1 h+1 V )] (1 + i)n h
= P s n n 1
h =0
vq x + h (bh +1 h+1 V )(1 + i)n h
If the death benet is equal to the benet reserve for the rst n policy years
n V = P s n
If the death benet is equal to $1 plus the benet reserve for the rst n policy years
n V = P s n
n 1
h =0 vq x + h (1 + i)n h
Exam M - Life Contingencies - LGD c 9
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
10/26
If the death benet is equal to $1 plus the benet reserve for the rst n policy years and q x + h q constant
n V = P s n vq s n = ( P vq )s n
Reserves at fractional durations:
( h V + h )(1 + i)s = s px + h h+ s V + s q x + h v1 s
UDD ( h V + h )(1 + i)s = (1 s q x + h ) h + s V + s q x + h v1 s
= h+ s V + s q x + h (v1 s h+ s V )h + s V = v1 s 1 s q x + h + s bh +1 + v1 s 1 s px + h + s h+1 V
UDD h+ s V = (1 s)( h V + h ) + s(h +1 V )i.e. h+ s V = (1 s)( h V ) + ( s)(h +1 V ) + (1 s)(h )
unearned premium
Next year losses:h losses incurred from time h to h + 1
E [h ] = 0V ar[h ] = v2(bh +1 h+1 V )2 px + h q x + h
The Hattendorf theorem
V ar[ h L] = V ar[h ] + v2 px + h V ar[ h +1 L]= v2(bh +1 h+1 V )2 px + h q x + h + v2 px + h V ar[ h +1 L]
V ar[ h L] = v2(bh +1 h+1 V )2 px + h q x + h
+ v4(bh +2 h+2 V )2 px + h px + h +1 q x + h +1+ v6(bh +3 h+3 V )2 px + h px + h +1 px + h +2 q x + h +2 +
Exam M - Life Contingencies - LGD c 10
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
11/26
Chapter 9: Multiple Life Functions
Joint survival function:
sT (x )T (y) (s, t ) = P r [T (x) > s &T (y) > t ]
t pxy = sT (x )T (y)(t, t )= P r [T (x) > t and T (y) > t ]
Joint life status:
F T (t) = P r [min(T (x), T (y)) t]= tq xy= 1 t pxy
Independant livest pxy = t px t pyt q xy = tq x + tq y tq x tq y
Complete expectation of the joint-life sta-tus:
exy =
0
t pxy dt
PDF joint-life status:
f T (xy )(t) = t pxy xy (t)
xy (t) =f T (xy ) (t)
1 F T (xy )(t) =
f T (xy ) (t)t pxy
Independant lives
xy (t) = (x + t) + (y + t)f T (xy ) (t) = t px t py [(x + t) + (y + t)]
Curtate joint-life functions:
k pxy = k px k py [IL]k q xy = kq x + kq y kq x kq y [IL]
P r [K = k] = k pxy k+1 pxy= k pxy q x + k :y+ k= k pxy q x + k :y+ k = k|q xy
q x + t :y+ t = q x + k + q y+ k q x + k q y+ k [IL]
exy = E [K (xy)] =
1k pxy
Last survivor status T (xy):
T (xy) + T (xy) = T (x) + T (y)
T (xy) T (xy) = T (x) T (y)f T (xy ) + f T (xy ) = f T (x ) + f T (y)
F T (xy ) + F T (xy ) = F T (x ) + F T (y)t pxy + t pxy = t px + t pyAxy + Axy = Ax + Ayaxy + axy = ax + ayexy + exy = ex + eyexy + exy = ex + ey
n |q xy = n |q x + n | q y n |q xy
Complete expectation of the last-survivorstatus:
exy =
0
t pxy dt
exy =
1k pxy
Variances:
V ar[T (u)] = 2
0 t t pu dt (eu )2V ar[T (xy)] = 2
0
t t pxy dt (exy )2
V ar[T (xy)] = 2
0
t t pxy dt (exy )2
Notes:
For joint-life status, work with ps:n pxy = n px n py
For last-survivor status, work with q s:
n q xy = n q x n q y
Exactly one status:
n p[1]xy = n pxy n pxy
= n px + n py 2 n px n py
= n q x + n q y 2 n q x n q ya [1]xy = ax + ay 2axy
Exam M - Life Contingencies - LGD c 11
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
12/26
Common shock model:
sT (x )(t) = sT (x ) (t) sz (t)
= sT (x ) (t) e t
sT (y)(t) = sT (y) (t) sz (t)
= sT (y) (t) e t
sT (x )T (y)(t) = sT (x ) (t) T (y) (t) sz (t)
= sT (x ) (t) sT (y)(t) e t
xy (t) = (x + t) + (y + t) +
Insurance functions:
Au =
k=0
vk+1 k pu q u + k
=
k=0vk+1 P r [K = k]
Axy =
k=0
vk+1 k pxy q x + k :y+ k
Axy =
k=0
vk+1 ( k px q x + k + k py q y+ k
k pxy q x + k :y+ k )
Variance of insurance functions:
V ar[Z ] = 2Au (Au )2
V ar[Z ] = 2Axy (Axy )2
Cov[vT (xy ) , vT (xy ) ] = ( Ax Axy )( Ay Axy )
Insurances:
Ax = 1 axAxy = 1 axyA
xy = 1 a
xy
Premiums:
P x = 1ax
d
P xy = 1axy
d
P xy = 1axy
d
Annuity functions:
au =
0
vt t pu dt
var [Y ] =2Au ( Au )2
2
Reversionary annuities:A reversioanry annuity is payable during the ex-istence of one status u only if another status vhas failed. E.g. an annuity of 1 per year payablecontinuously to ( y) after the death of ( x).
ax |y = ay axy
Covariance of T (xy) and T (xy):
Cov [T (xy), T (xy)] = Cov [T (x), T (y)] + {E [T (x)] E [T (xy)]} {(E [T (y)] E [T (xy)]}
= Cov [T (x), T (y)] + ( ex exy ) ( ey exy )= ( ex exy ) (ey exy ) [IL]
Exam M - Life Contingencies - LGD c 12
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
13/26
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
14/26
Basic relationships:
t p( )x = exp t
0
(1)x (s) + + (m )x (s) ds
t p( )x =m
i=1t p (i )x
t q ( j )x tq ( j )xt p ( j )x t p(
)x
UDD for multiple decrements:
t q ( j )x = t q ( j )
x
t q ( )x = t q ( )
x
q ( j )x = t p( )
x ( j )x (t)
( j )x (t) = q ( j )x
t p( )x
= q ( j )x
1 t q ( )x
t p ( j )x = t p( )x
q( j )x
qxt
Decrements uniformly distributed in theassociated single decrement table:
t q ( j )x = t q ( j )
x
q (1)x = q (1)x 1 1
2q (2)x
q (2)x = q (2)
x 1 12
q (1)x
q (1)x = q (1)x 1 12
q (2)x 12
q (3)x + 13
q (2)x q (3)x
Actuarial present values
A =m
j =1
0
B ( j )x + t vt t p( )x (
j )x (t)dt
Instead of summing the benets for each pos-sible cause of death, it is often easier to writethe benet as one benet given regardless of thecause of death and add/subtract other benetsaccording to the cause of death.
Premiums:
P ( )x =
k=0B ( )k+1 v
k+1 k p( )x q (
)x + k
k=0vk k p( )x
P ( j )x =
k=0B ( j )k+1 v
k+1 k p( )x q ( j )x + k
k=0vk k p
( )x
Exam M - Life Contingencies - LGD c 14
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
15/26
Chapter 15: Models Including Expenses
Notations:
G expense loaded ( or gross) premium
b face amount of the policyG/b per unit gross premium
Expense policy fee:The portion of G that is independent of b.
Asset shares notations:
G level annual contract premiumk AS asset share assigned to the policy at time t = k
ck fraction of premium paid for expenses at k (i.e. cG is the expenxe premium)ek expenses paid per policy at time t = k
q (d)x + k probability of decrement by death
q (w )x + k probability of decrement by withdrawalk CV cash amount due to the policy holder as a withdrawal benet
bk death benet due at time t = k
Recursion formula:
[ k AS + G(1 ck ) ek ] (1 + i) = q (d)x + k bk+1 + q
(w )x + k k+1 CV + p
( )x + k k+1 AS
= k+1 AS + q (d)x + k (bk+1 k+1 AS ) + q (w)x + k ( k+1 CV k+1 AS )
Direct formula:
n AS =n 1
h =0
G(1 ch ) eh vq (d)x + h bh +1 vq
(w )x + h h +1 CV
n h E ( )x + h
Exam M - Life Contingencies - LGD c 15
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
16/26
Constant Force of Mortality
Chapter 3
(x) = > 0, xs(x) = e x
lx = l0e x
n px = e n = ( px )n
ex = 1
= E [T ] = E [X ]
ex : n = ex (1 n px )
V ar[T ] = V ar[x] = 12
mx =
Median[T ] = ln2
= Median[ X ]ex =
pxq x
= E [K ]
V ar[K ] = px(q x )2
Chapter 4
Ax = +
2Ax =
+ 2 A1x : n = Ax (1 n E x )
n E x = e n (+ )
(IA)x = ( + )2
Ax = q q + i
2Ax = q
q + 2 i + i2
A1x : n = Ax (1 n E x )
Chapter 5
ax = 1 +
2ax = 1 + 2
ax = 1 + i
q + i
2ax = (1 + i)2
q + 2 i + i2
ax : n = ax (1 n E x )ax : n = ax (1 n E x )
(IA)x = Ax ax =
( + )2
(I A)x = Ax ax = (1 + i)
( + )(q + i)(IA)x = Ax ax =
q (1 + i)(q + i)2
(Ia )x = ( ax )2 = 1
( + )2
(I a)x = ( ax )2 =1 + iq + i
2
Chapter 6
P x = vq x = P 1x : nP (
Ax ) = =
P (
A
1
x : n )For fully discrete whole life, w/ EP,
V ar[Loss ] = p 2Ax
For fully continuous whole life, w/EP,
V ar[Loss ] = 2Ax
Chapter 7
t V ( Ax ) = 0 , t 0
kV
x = 0 , k = 0 , 1, 2, . . .
For fully discrete whole life, assuming EP,
V ar[k Loss ] = p 2Ax , k = 0 , 1, 2, . . .
For fully continuous whole life, assuming EP,
V ar[ t Loss ] = 2Ax , t 0
Chapter 9For two constant forces, i.e. M acting on ( x)and F acting on ( y), we have:
Axy = M + F
M + F +
axy = 1
M + F +
exy = 1M + F
Axy = q xyq xy + i
axy = 1 + iq xy + i
exy = pxy
q xy
Exam M - Life Contingencies - LGD c 16
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
17/26
De Moivres Law
Chapter 3
s(x) = 1 x
lx = l0 x ( x)
q x = (x) = 1 x
n |m q x = m x
n px = x n
xt px (x + t) = q x = (x) = f T (x), 0 t < x
Lx = 1
2(lx + lx +1 )
ex = x
2 = E [T ] = Median[ T ]
ex = x
2
12
= E [K ]
V ar[T ] = ( x)2
12
V ar[K ] = ( x)2 1
12
mx = q x1 12 q x
= 2dx
lx + lx +1
a(x) = E [S ] = 1
2ex : n = n n px +
n2 n
q x
ex : n = ex : n + n2 n
q x
Chapter 4
Ax =a x x
A1x : n = a n x
2Ax =a 2( x )2( x)
Ax =
a x x
A1x : n = a n x
(IA)x =(Ia ) x
x
(IA)x =(Ia ) x
x
(IA)1x : n = (Ia ) n
x
(IA)1x : n = (Ia ) n
x
Chapter 5No useful formulas: use ax = 1 A xd and thechapter 4 formulas.
Chapter 9
exx = x
3 ( MDML with = 2 / ( x))
exx = 2( x)
3exy = y x px eyy + y x q x ey
For two lives with different s, simply translateone of the age by the difference in s. E.g.
Age 30, = 100 Age 15, = 85
Modied De Moivres Law
Chapter 3
s(x) = 1 x
c
lx = l0 x
c
( x)c
(x) = c x
n px = x n
x
c
ex = x
c + 1 = E [T ]
V ar[T ] = ( x)2c(c + 1) 2(c + 2)
Chapter 9
exx = x
2c + 1
ex with = 2c x
Exam M - Life Contingencies - LGD c 17
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
18/26
Uniform Distribution of Deaths (UDD)
Chapter 3
t q x = t q x , 0 t 1
(x + t) = q x1 tq x , 0 t 1
lx + t = lx t dx , 0 t 1
s q x + t = sq x1 tq x
, 0 s + t 1
V ar[T ] = V ar[K ] + 112
mx = (x + 12
) = q x
1 12 q x
Lx = lx 12
dx
a(x) = 12
ex : 1 = px + 12
q x
t px (x + t) = q x , 0 t 1
Chapter 4
Ax = i
Ax
A(m )x = ii(m )
Ax
A1x : n = i A1x : n
(I A)1x : n = i
(IA)1x : n
Ax : n = i
A1x : n + A
1x : n =
i
A1x : n + n E x
n | Ax = i
n |Ax
2Ax = 2i + i2
2 2Ax
Chapter 5
a (m )x ax m 1
2ma (m )x = (m)ax (m)
a (m )x : n = (m)ax : n (m)(1 n E x )
with: (m) = id
i(m )d(m ) 1
and (m) = i i(m )
i(m )d(m )
m 12m
a (m )x : n = a(m )x n E x a
(m )x + n
Chapter 6
P ( Ax ) = i
P x
P ( A1x : n ) = i
P 1x : n
P ( Ax : n ) = i
P 1x : n + P
1x : n
P (m )x = P x
(m) (m)(P x + d)
P (m )x : n = P x : n
(m) (m)(P 1x : n + d)
n P (m )x = n P x
(m) (m)(P 1x : n + d)
h P (m ) ( A1x : n ) = i hP 1(m )x : n
Chapter 7
hk V
(m )x : n =
hk V x : n + (m) h P
(m )x : n kV
1x : h
hk V
(m ) ( Ax : n ) = hk V ( Ax : n ) + (m) h P (m ) ( Ax : n ) k V 1x :
hk V ( Ax : n ) =
i
hk V 1
x : h + hk V
1x : h
Chapter 10UDDMDT
t q ( j )x = t q ( j )xq ( j )x =
( j )x (0)
q ( )x = ( )
x (0)
t p ( j )x = t p( )
x
q ( j )xq ( )x
( j )x = q ( j )x1 t q ( )x
( )x = q ( )x
1 t q ( )x
UDDASDT
t q ( j )x = t q ( j )x
q (1)x = q (1)
x 1 12
q (2)x
q (2)x = q (2)x 1 12
q (1)x
q (1)x = q (1)x 1 12
q (2)x 12
q (3)x + 13
q (2)x q (3)x
Exam M - Life Contingencies - LGD c 18
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
19/26
Chapter 1: The Poisson Process
Poisson process with rate :
P r [N (s + t) N (s) = k] = e t(t )k
k!E [N (t)] = tV ar[N (t)] = t
Interarrival time distribution:The waiting time between events. Let T n de-note the time since occurence of the event n 1.Then the T n are independent random variablesfollowing an exponential distribution with mean1/ .
P r [T i t] = 1 e t
f T (t) = e t
E [T ] = 1
V ar[T ] = 12
Waiting time distribution:Let S n be the time of the n-occurence of theevent, i.e. S n =
ni=1 T i .
S n has a gamma distribution with parameters nand = 1 /
S n GammaRV[ = n, = 1]
P r [S n t] =
j = ne t
(t ) j
j !
f S n (t) = et (t )n 1
(n i)!
E [S n ] = n
V ar[S n ] =
n2
Sum of Poisson processes:If N 1, , N k are independent Poisson processeswith rates 1, , k then, N = N 1 + + N kis a Poisson process with rate = 1 + + k .
Special events in a Poisson process:Let N be a Poisson process with rate .Some events i are special with a probabilityP r [event is special] = i and N i counts the spe-cial events of kind i. Then, N i is a Poissonprocess with rate i = i and the N i are inde-pendent of one another.If the probability i (t) changes with time, then
E [ N i (t)] = t
0
(s)ds
Non-homogeneous Poisson process:
(t) intensity functionm(t) mean value function
=t
0
(y)dy
P r [N (t) = k] = e m ( t )[m(t)]k
k!
Compound Poisson process:
X (t) =N (t )
i=1
Y i N (t) PoissonRV w/ rate
E [X (t)] = t E [Y ]V ar[X (t)] = t E [Y 2]
Two competing Poisson processes:Probability that n events in the Poisson process ( N 1,1) occur before m events in the Poisson process(N 2,2):
P r [S 1n < S 2m ] =n + m 1
k= n
n + m 1k
11 + 2
k 21 + 2
n + m 1 k
P r [S 1n < S 21 ] = 1
1 + 2
n
P r [S 11 < S 2
1] =
11 + 2
Exam M - Loss Models - LGD c 1
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
20/26
Chapter 2&3: Random Variables
k-th raw moment:
k = E [X k ]
k-th central moment:
k = E [(X )k ]
Variance:
V ar[x] = 2 = E [X 2] E [X ]2
= 2 = 2 2
Standard deviation:
= V ar[X ]Coefficient of variation:
/
Skewness:
1 = E [(X )3]
3 =
33
Kurtosis:
1 = E [(X )4]
4 =
44
Left truncated and shifted variable (akaexcess loss variable):
Y P = X d|X > d
Mean excess loss function:
eX (d) e(d) = E [Y P ]= E [X d|X > d ]
=
d S (x)dx1 F (d)
=
E [X ] E [X d]1 F (d)
Higher moments of the excess loss vari-able:
ekx (d) = E [(X d)k |X > d ]
=
d (x d)k f (x)dx1 F (d)
= x>d(x d)k p(x)
1 F (d)
Left censored and shifted variable:
Y L = ( X d)+
= 0 X < d
X d X d
Moments of the left censored and shiftedvariable:
E [(X d)k+ ] =
d
(x d)k f (x)dx
=x>d
(x d)k p(x)
= ek
(d)[1 F (d)]
Limited loss:
Y = ( X u) = X X < uu X u
Limited expected value:
E [X u] = 0
F (x)dx +u
0
S (x)dx
=u
0
[1 F (x)]dx if X is always positive
Moments of the limited loss variable:
E [(X u)k ] =u
xk f (x)dx + uk [1 F (u)]
=x u
xk p(x) + uk [1 F (u)]
Moment generating functions mX (t):
mX (t) = E [etX ]
Sum of random variables S k = X 1 + + X k :
mS k (t) =k
j =1
mX j (t)
Exam M - Loss Models - LGD c 2
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
21/26
Chapter 4: Classifying and Creating distributions
k-point mixture ( a i = 1 ):
F Y (y) = a1F X 1 (y) + + ak F X k (y)
f Y (y) = a1f X 1 (y) + + ak f X k (y)E [Y ] = a1E [X 1] + + ak E [X k ]
E [Y 2] = a1E [X 21 ] + + ak E [X 2k ]
Tail weight:
existence of moments light tailhazard rate increases light tail
Multiplication by a constant ,y > 0:
Y = X F Y (y) = F X ( y) and f Y (y) = 1
f X ( y)
Raising to a power:
Y = X 1
> 0 : F Y (y) = F X (y )f Y (y) = y 1f X (y )
< 0 : F Y (y) = 1 F X (y )f Y (y) = y 1f X (y )
> 0: transformed distribution = 1: inverse distribution
< 0: inverse transformed
Exponentiation:
Y = eX F Y (y) = F X (ln( y))f Y (y) = 1y f X (ln( y))
Mixing:The random variable X depends upon a para-meter , itself a random variable . For a givenvalue = , the individual pdf is f X | (x |).
f X (x) = f X | (x |)f ()d If X is a Poisson distribution with parame-
ter , and follows a Gamma distributionwith parameters ( , ), then the resultingdistribution is a Negative Binomial distri-bution with parameters ( r = , = )
If X is an Exponential distribution withmean 1/ and is a Gamma distributionwith parameters ( , ), then the resulting
distribution is a 2-parameter Pareto distri-bution with parameters ( = , = 1 )
If X is a Normal distribution (, V ) and follows a Normal distribution (, A),then the resulting distribution is a Normaldistribution (, A + V )
Splicing ( a j > 0, a j = 1 ):
f (x) =a1f 1(x) c0 < x < c 1
... ...
ak f k (x) ck 1 < x < c k
Discrete probability function (pf):
pk = P r [N = k]
Probability generating function (pgf):
P N (z) = E [zN ] = p0 + p1z + p2z2 + P (1) = E [N ]P (1) = E [N (N 1)]
Poisson distribution:
pk = e k
k!P (z) = e (z 1)
E [N ] = V ar[N ] =
Negative Binomial distribution:Number of failures before success r with proba-bility of success p = 1 / (1 + )
pk = k + r 1
k
1 +
k 11 +
r
P (z) = [1 (z 1)] r
E [N ] = r V ar[N ] = r (1 + )
If N 1 and N 2 are independent NegativeBinomial distributions with parameters(r 1, ) and ( r 2, ), then N 1 + N 2 is a Neg-
ative Binomial distribution with parame-ters ( r = r1 + r 2, )
Exam M - Loss Models - LGD c 3
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
22/26
If N is a Negative Binomial distributionwith parameters ( r, ) and some events jare special with probability j , then N j ,the count of special events j , is a Nega-tive Binomial distribution with parame-ters ( r j = r, j = j )
Geometric distribution:Special case of the negative binomial distribu-tion with r = 1
pk = k
(1 + )k+1
P (z) = [1 (z 1)] 1 = 1
1 (z 1)
E [N ] = V ar[N ] = (1 + )
Binomial distribution:Counting number of successes in m trials givena probability of sucess q
pk = mk q
k (1 q )m k
P (z) = [1 + q (z 1)]m
E [N ] = mq V ar[N ] = mq (1 q )
The (a,b, 0) class:
pk pk 1
= a + bk
a > 0 negative binomial or geometricdistribution
a = 0 Poisson distributiona < 0 Binomial distribution
Compound frequency models:
P (z) = P prim (P sec (z))
Compound frequency and (a,b, 0) class:Let S be a compound distribution of primaryand secondary distributions N and M with pn =P r [N = n] and f n = P r [M = n]. If N is a mem-ber of the ( a,b, 0) class
gk = 11 af 0
k
j =1
a + j b
kf j gk j
g0 = P prim (f 0)
gk =
k
k
j =1
j f j gk j
Mixed Poisson models:If P (z) id the pgf of a Poisson distribution with
rate drawn from a discrete random variable ,then
P (z |) = e (z 1)
P (z) = p(1)e 1 (z 1) + + p( n )e n (z 1)
Exposure modications on frequency:n policies in force. N j claims produced by the j -th policy. N = N 1 + + N n . If the N j are in-dependent and identcally distributed, then theprobability generating function for N is
P N (z) = [ P N 1 (z)]n
If the company exposure grows to n , let N represent the new total number of claims,
P N (z) = [ P N 1 (z)]n = [ P N (z)]nn
Conditional expectations:
E [X | = ] = xf X |(x |)dxE [X ] = E [E [X |]]
V ar[X ] = E [V ar[X |]] + V ar[E [X |]]
The variance of the random variable X is thesum of two parts: the mean of the conditionalvariance plus the variance of the conditionalmean.
Exposure modications on frequency for common distributions:N Parameters for N Parameters for N
Poisson = n
n Negative Binomial r, r = nn r, =
Exam M - Loss Models - LGD c 4
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
23/26
Chapter 5: Frequency and Severity with Coverage Modications
Notations:
X amount of loss
Y L amount paid per lossY P amount paid per payment
Per loss variable:
f Y L (0) = F X (d)f Y L (y) = f x (y + d)F Y L (y) = F X (y + d)
Per payment variable:
f Y P (y) = f
x(y + d)
1 F X (d)
F Y L (y) = F X (y + d) F X (d)
1 F X (d)
Relations per loss / per payment variable:
Y P = Y L |Y L > 0
E [Y P ] = E [Y L ]P r [Y L > 0]
E [(Y P )k ] = E [(Y L )k ]
P r [Y L
> 0]Simple (or ordinary) deductible d:
X = X Y L = ( X d)+Y P = X d|X > d
E [Y L ] = E [(X d)+ ]= E [X ] E [X d]
E [Y P ] = E [X d|X > d ]
= E [X ] E [X d]
1 F X (d)
Franchise deductible d:
X = X Y L = 0 if X < d, X if X > dY P = X |X > d,
Y P franchise = Y P
ordinary + d
E [Y L ] = E [X ] E [X d] + d[1 F X (d)]E [Y P ] = E [X |X > d ]
= E [X ] E [X d]1 F X (d)
+ d
Effect of deductible for various distribu-tions:
X : exp[mean ] Y P
: exp[mean ]X : uniform[0, ] Y P : uniform[0, d]X : 2Pareto[ , ] Y P : 2Pareto[ = , = +
Loss elimination ratio:
LER X (d) = E [X d]
E [X ]
relations:
c(X d) = ( cX ) (cd)
c(X d)+ = ( cX cd)+A = ( A b) + ( A b)+
Ination on expected cost per loss:
E [Y L ] = (1 + r ) E (X ) E X d1 + r
Ination on expected cost per payment:
E [Y P ] =(1 + r ) E (X ) E X d1+ r
1 F X d1+ r
u-coverage limit:
Y L = X uY P = X u |X > 0
E [Y L ] = E [X u]
=u
0
[1 F x (x)]dx
u-coverage limit with ination:
E [X u] (1 + r )E X u1 + r
u-coverage limit and d ordinary de-ductible:
Y L = ( X u) (X d)Y P = ( X u) (X d)|X > 0
= ( X u) d|X > dE [Y L ] = E [X u] E [X d]
E [Y P ] = E [X u] E [X d]1 F X (d)
Exam M - Loss Models - LGD c 5
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
24/26
u-coverage limit and d ordinary deductiblewith ination:
E [Y L ] = (1 + r ) E X u1 + r
E X d1 + r
E [Y P ] = (1 + r )E X u1+ r E X d1+ r
1 F X d1+ r
Coinsurance factor :First calculate Y L and Y P with deductible andlimits. Then apply
Y L = Y L
Y P = Y P
Probability of payment for a loss with de-ductible d:
v = 1 F X (d)
Unmodied frequency:The number of claims N (frequency) does notchange, only the probabilities associated withthe severity are modied to include the possi-bility of zero payment.
Bernoulli random variable:
N = 0 @ 1 p1 @ pE [N ] = p
V ar[N ] = (1 p) p
Deductible, limit, ination and coinsurance:
d = d1 + r
u = u1 + r
E [Y L ] = (1 + r ) {E [X u ] E [X d ]}
E [Y P ] = (1 + r )E [X u ] E [X d ]
1 F X (d )E [Y L 2] = 2(1 + r )2 E (X u )2 E (X d )2 2d E [X u ] + 2d E [X d ]
Modied frequency:N (the modied frequency) has a compound distribution: The primary distribution is N , and the secondarydistribution is the Bernoulli random variable I . The probability generating function is
P N (z) = P N [1 + v(z 1)]
The frequency distribution is modied to include only non-zero claim amounts. Each claim amount prob-
ability is modied by dividing it by th eprobability of a non-zero claim. For common distributions, thefrequency distribution changes as follows (the number of positive payments is nothing else than a specialevent with probability = v)
N Parameters for N Parameters for N Poisson = vBinomial m, q m = m, q = vq Negative Binomial r, r = r, = v
Exam M - Loss Models - LGD c 6
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
25/26
Chapter 6: Aggregate Loss Models
The aggregate loss compound randomvariable:
S =
N
j =1X j
E [S ] = E [N ] E [X ]V ar[S ] = E [N ] V ar[X ] + V ar[N ] E [X ]2
S (E [S ], Va r [S ]) if E [N ] is large
Probability generating function:
P S (z) = P N [P X (z)]
Notations:
f k = P r [X = k] pk = P r [N = k]gk = P r [S = k]
Compound probabilities:
g0 = P N (f 0) if P r [X = 0] = f 0 = 0gk = p0 P r [sum of X s = k]
+ p1 P r [sum of one X = k]+ p2 P r [sum of two X s = k]
+ If N is in the (a,b, 0) class, then
g0 = P N (f 0)
gk = 11 af 0
k
j =1a +
j bk
f j gk j
n-fold convolution of the pdf of X :
f (n )X (k) = P r [sum of n X s = k]
f (n )X (k) =k
j =0f (n 1)X (k j )f X ( j )
f (1)X (k) = f X (k) = f k
f (0)X (k) = 1 , k = 00 , otherwise
n-fold convolution of the cdf of X :
F (n )X (k) =k
j =0
F (n 1)X (k j )f X ( j )
F (0)X (k) = 0 , k < 01 , k 0
Pdf of the compound distribution:
f S (k) =
n =0 pn f
(n )X (k)
Cdf of the compound distribution:
F S (k) =
n =0 pn F (n )X (k)
Mean of the compound distribution:
E [S ] =
k=0
k f S (k) =
k=0
[1 F S (k)]
Net stop-loss premium:
E [(S d)+
] =
d(x d)f
S (x)dx
E [(S d)+ ] =
d+1
(x d)f S (x)
E [(S d)+ ] = E [S ] d d 1
x =0(x d)f S (x)
E [(S d)+ ] =
d
[1 F S (d)]
= E [S ]
d 1
x =0[1 F S (d)]
Linear interpolation a < d < b :
E [(S d)+ ] = b db a
E [(S a)+ ]+d ab a
E [(S b)+ ]
Recursion relations for stop-loss insurance:
E [(S d 1)+ ] = E [(S d)+ ] [1 F S (d)]E [(S s j )] = E [(S s j +1 )] + ( s j +1 s j )P r [S s j +1 ]
where the s j s are the possible values of S , S : s0 = 0 < s 1 < s 2 < andP r [S s j +1 ] = 1 P r [S = s0 or S = s1 or or S = s j ]
Exam M - Loss Models - LGD c 7
-
7/22/2019 Chapter 3 Survival Distributions and Life Tables Exam M Notes F05
26/26
Last Chapter: Multi-State Transition Models
Notations:
Q( i,j )n = P r [M n +1 = j |M n = i]
k Q( i,j )n = P r [M n + k = j |M n = i]P ( i )n = Q(
i,i )n
k P ( i )n = P r [M n +1 = = M n + k = i|M n = i]
Theorem:
k Qn = Qn Qn +1 Qn + k 1k Qn = Qk for a homogeneous Markov chain
Inequality:
k P ( i )n = P (i)
n P ( i )n +1 P
( i )n + k 1 kQ
( i,i )n
Cash ows while in states:
lC ( i ) = cash ow at time l if subject in state i at time l
k vn = present value of 1 paid k years after time t = nAP V s@n (C ( i ) ) = actuarial present value at time t = n of all future payments to be
made while in state i , given that the subject is in state s at t = n
AP V s@n (C ( i ) ) =
k=0k Q(s,i )n n + k C (
i ) kvn
Cash ows upon transitions:
l+1 C ( i,j ) = cash ow at time l + 1 if subject in state i at time l and state j at time l + 1
lQ(s,i )n Q( i,j )n + l = probability of being in state i at time l and in state j at time l + 1
AP V s@n (C ( i,j ) ) = actuarial present value at time t = n of a cash ow to be paid upontransition from state i to state j
AP V s@n (C ( i,j ) ) =
k=0k Q(s,i )n Q
( i,j )n + k n + k+1 C
( i,j ) k+1 vn