A.

1: Calculus Review

Derivatives Chain Rule

d x
e = ex D {f (g(x))} = f 0 (g(x)) · g 0 (x)
dx
d x
a = ax ln a Product Rule
dx
Integrals (uv)0 = u0 v + uv 0
Z
ex dx = ex + c Integration by Parts
ax
Z Z Z
ax dx = +c udv = uv − vdu
ln a

c
The Infinite Actuary, LLC 1
 A.2: Probability Review

Variance Normal Approximation

Var[X] = E[X 2 ] − E[X]2 S = X1 + X2 + . . . + Xn

2 2 where Xi ’s are iid RVs
Var[aX + bY ] = a Var(X) + b Var(Y )
+ 2abCov(X, Y ) S ∼ Normal(E[S] = nµx , Var[S] = nσx2 )
Pr(S ≤ s) = 0.95 ⇒ s = µS + σS Φ−1 (0.95)
If X and Y independent, then Cov(X, Y ) = 0
Discrete Uniform

domain = x1 , x2 , . . . , xn
Discrete R.V. 1
p(xi ) =
n
Pr(X = x) = p(x) n
1X
E[X] = xi
X
p(x) = 1 n
i=1
x
n
1
X
E[X k ] = xk · p(x)
X
E[X 2 ] = x2i
x n
i=1
Var[X] = E[X ] − E[X]2
2

Continuous R.V. if domain = 1, 2, 3, . . . , n

n+1
E[X] =
Z
f (x)dx = 1 2
x n2 − 1
Z Var[X] =
E[X k ] = xk f (x)dx 12
x
Bernoulli

p = probability of success
Mixed R.V.
Z b E[X] = p
p(a) + f (x)dx + p(b) = 1 Var[X] = p(1 − p)
a
Z b
Any R.V. that takes on two values is a scaled
E[X k ] = ak p(a) + xk f (x)dx + bk p(b)
a and translated Bernoulli
(
Multivariate Probability a p(a) = p
Y =
E[X] = EY [EX [X|Y ]] b p(b) = 1 − p
Var[X] = EY [VarX [X|Y ]] + VarY [EX [X|Y ]] Y = (a − b)X + b
E[Y ] = (a − b)p + b
Var[Y ] = (a − b)2 p(1 − p)

c
The Infinite Actuary, LLC 2
 Binomial Continuous Uniform

X ≡ # of successes in n independent trials domian : a < X < b

p = probability of a success 1 1
f (x) = =
 
n x b−a length interval
p(x) = p (1 − p)n−x a+b
x E[X] =
2
E[X] = np
(b − a)2
Var[X] = np(1 − p) Var[X] =
12
median = mean
Geometric
Exponential
X ≡ # of failures that occur before
first success f (x) = µe−µx , x > 0, µ > 0
p = probability of a success 1
E[X] =
q = 1 − p = probability of a failure µ
1
p(x) = p(1 − p)x Var[X] = 2
q µ
E[X] = F (x) = 1 − e−µx
p
q
Var[X] = 2 Percentiles
p
Let r = 90th percentile of X
Pr[X ≤ r] = 0.9 ⇒ F (r) = 0.9
Pr[X > r] = 0.1 ⇒ s(r) = 0.1

c
The Infinite Actuary, LLC 3
 A.3: Interest Theory Review

Interest Measures Annuities

d
1−d =i 1 − vn 1
d i an = a∞ =
i i
d i 1 − vn 1
i
=d än = ä∞ =
1+i d d
1 − vn 1
ān = ā∞ =
1 δ δ
ln(1 + i) = δ v= =1−d än − nv n 1+i 1
1+i (Ia)n = (Ia)∞ = =
2
1 + i = eδ v = e−δ i i di
än − nv n 1
(1 + i)n = enδ v n = e−nδ (Iä)n = (Iä)∞ = 2
d d
Double the force of interest än − nv n 1
(Iā)n = (Iā)∞ =
2 δ dδ
1 + i → (1 + i) ān − nv n 1
¯ n =
(Iā) ¯ ∞
(Iā) = 2
v → v2 δ δ
n − an
i → 2i + i2 (Da)n =
2
i
d → 2d − d

c
The Infinite Actuary, LLC 4
 B.1: Survival Distributions and Life Tables

Distribution function of X: Force of mortality µ(x):

FX (x) = Pr(X ≤ x) fX (x)
µ(x) =
1 − FX (x)
Survival function s(x):
s0 (x)
= −
s(x) = 1 − FX (x) s(x)
Probability of death between age x and Relations between survival functions and
age y: force of mortality:
Pr(x < X ≤ y) = FX (y) − FX (x)
 x 
Z
= s(x) − s(y) s(x) = exp − µ(y)dy 
0
Probability of death between age x and
x+n
 
age y given survival to age x: Z
n px = exp − µ(y)dy 
FX (y) − FX (x)
Pr(x < X ≤ y|X > x) = x
1 − FX (x)
s(x) − s(y) Derivatives:
=
s(x) d
t qx = t px · µ(x + t) = fT (x) (t)
Notations: dt
d
t qx = Pr[T (x) ≤ t] t px = − t px · µ(x + t)
dt
= prob. (x) dies within t years
= distribution function of T (x) Mean and variance of T and K:
t px = Pr[T (x) > t]
E[T (x)] ≡ complete expectation of life
= prob. (x) attains age x + t Z∞
= 1 − t qx ≡ e̊x = t px dt
0
t|u qx = Pr[t < T (x) ≤ t + u]
E[K(x)] ≡ curtate expectation of life
= t+u qx − t qx X∞
= t px − t+u px ≡ ex = k px
k=1
= t px · u qx+t
Z∞
Relations with survival functions: V ar[T (x)] = 2 t · t px dt − e̊2x
s(x + t) 0
t px = ∞
s(x) X
s(x + t) V ar[K(x)] = (2k − 1) k px − e2x
t qx = 1 − k=1
s(x)
Curtate future lifetime (K(x) ≡ greatest Total lifetime after age x: Tx
integer in T (x)):
Z∞
Pr[K(x) = k] = Pr[k ≤ T (x) < k + 1]
Tx = `x+t dt
= k px − k+1 px
0
= k px · qx+k
= k| qx

c
The Infinite Actuary, LLC 5
Total lifetime between age x and x + 1: Lx Central death rate: mx

Lx = Tx − Tx+1 `x − `x+1
mx =
Z1 Z1 Lx
= `x+t dt = `x · t px dt `x − `x+n
n mx =
n Lx
0 0

Total lifetime from age x to x + n: Fraction of year lived between age x and
n Lx
age x + 1 by dx : a(x)
n−1
X
n Lx = Tx − Tx+n = Lx+k R1
k=0
t · t px · µ(x + t) dt
0 Lx − `x+1
Zn a(x) = =
R1 `x − `x+1
= `x+t dt t px · µ(x + t) dt
0
0

Average lifetime after x: e̊x Recursion formulas:

Tx E[K] = ex = px (1 + ex+1 )
e̊x =
`x E[T ] = e̊x = px (1 + e̊x+1 ) + qx a(x)
Average lifetime from x to x + 1: e̊x: 1 ex = ex: n + n px ex+n
e̊x = e̊x: n + n px e̊x+n
Lx
e̊x: 1 =
`x
E[K ∧ (m + n)] = ex: m+n
Median future lifetime of (x): m(x) = ex: m + m px ex+m: n

s(x + m(x)) 1 E[T ∧ (m + n)] = e̊x: m+n

P r[T (x) > m(x)] = =
s(x) 2 = e̊x: m + m px e̊x+m: n

c
The Infinite Actuary, LLC 6
 B.2: Life Insurance

Whole life insurance: Āx Discrete whole life: Ax

Z∞ ∞
X
E[Z] = Āx = t
v · t px µx (t)dt E[Z] = Ax = v k+1 · k px · qx+k
k=0
0 ∞
Āx − (Āx )2
2
X
V ar[Z] = = v k+1 · k| qx
k=0

1
V ar[Z] = Ax − (Ax )2
2

n-year term insurance: Āx: n

Zn
E[Z] = Āx: n = 1
v t · t px µx (t)dt Discrete n-year term: A1x: n
0 n−1
X
V ar[Z] = 2 1
Āx: n − (Āx: n ) 1 2 E[Z] = Ax: n = 1
v k+1 · k px · qx+k
k=0
V ar[Z] = Ax: n − (A1x: n )2
2 1

m-year deferred whole life: m| Āx

Z∞ Discrete n-year endowment: Ax: n

E[Z] = m| Āx = v t · t px µx (t)dt
n−1
m
X
1
E[Z] = Ax: n = v k+1 · k px · qx+k + v n · n px
m| Āx = Āx − Āx: m k=0
V ar[Z] = Āx: n − (Āx: n )2
2

n-year pure endowment: Āx: 1n Āx: n = Ā1x: n + Āx:1n

E[Z] = Āx:1n = v n · n px ≡ n Ex
V ar[Z] = Āx:1n − (Āx:1n )2 = v 2n · n px · n qx
2 Recursion and other relations:

Āx = Ā1x: n + n| Āx

2 2 1 2
n-year endowment insurance: Āx: n Āx = Āx: n + n| Āx

Zn n| Āx = n Ex · Āx+n
1
E[Z] = Āx: n = v t t px µx (t)dt + v n n px Āx = Āx: n + n Ex · Āx+n
0 Ax = vqx + vpx Ax+1
V ar[Z] = Āx: n − (Āx: n )2
2
Ax = v 2 qx + v 2 px 2Ax+1
2

Āx: n = Ā1x: n + Āx:1n A1x: n 1

= vqx + vpx Ax+1: n−1
m|n Ax = vpx · (m−1)|n Ax+1
Ax: 1 = v
m-yr deferred n-yr term: m|n Āx
Ax: 2 = vqx + v 2 px
1
m|n Āx = m Ex · Āx+m: n
= Āx: m+n − Ā1x: m
1

= m| Āx − m+n| Āx

c
The Infinite Actuary, LLC 7
Varying benefit insurances: Accumulated cost of insurance:
Z∞
Ā1x: n
(I Ā)x = bt + 1cv t · t px µx (t)dt n k̄x =
n Ex
0
Zn
(I Ā)1x: n = bt + 1cv t · t px µx (t)dt Share of the survivor:
0 1 (1 + i)n
accumulation factor = =
Z∞ n Ex n px
(I¯Ā)x = t · v t · t px µx (t)dt
0 Limit of interest rate i = 0:
Zn
i=0
(I¯Ā)1x: n = t · v t · t px µx (t)dt Ax −→ 1
i=0
0 A1x: n −→ n qx
Zn i=0
n| Ax −→ n px
(DĀ)1x: n = (n − btc)v t · t px µx (t)dt
i=0
0 Ax: n −→ 1
Zn i=0
m|n Ax −→ m|n qx
1 t
(D̄Ā)x: n = (n − t)v · t px µx (t)dt i=0
(IA)x −→ 1 + ex
0
i=0
(IA)x = Ax + vpx (IA)x+1 (I¯Ā)x −→ e̊x
= vqx + vpx [(IA)x+1 + Ax+1 ]
1 1
(DA)x: n = nvqx + vpx (DA)x+1: n−1
(I¯Ā)1
x: n + (D̄Ā)1x: n = nĀ1x: n
(I Ā)1x: n + (DĀ)1x: n = (n + 1)Ā1x: n
(IA)1x: n + (DA)1x: n = (n + 1)A1x: n

c
The Infinite Actuary, LLC 8
 B.3: Life Annuities

Whole life annuity: āx Recursion relations

Z∞ āx = āx: 1 + vpx āx+1
āx = E[ā T ] = ā t · t px µ(x + t)dt āx = āx: n + n| āx
0
äx = 1 + vpx äx+1
Z∞ Z∞
= v t · t px dt =
2
äx = 1 + v 2 px 2äx+1
t Ex dt
0 0
äx: n = 1 + vpx äx+1: n−1
2
Āx − (Āx )2 (m)
äx: n = ä(m)
x − n Ex · äx+n
(m)
V ar[ā T ] =
δ2 ax = vpx + vpx ax+1
ax: n = vpx + vpx ax+1: n−1
n-year temporary annuity: āx: n (Iä)x = 1 + vpx [(Iä)x+1 + äx+1 ]
Zn Zn = äx + vpx (Iä)x+1
āx: n = v t · t px dt = t Ex dt
Whole life annuity due: äx
0 0
2
Āx: n − (Āx: n )2 ∞
X
V ar[Y ] =
δ2 äx = E[ä K+1 ] = v k · k px
k=0
2
Ax − (Ax )2
n-year deferred annuity: n| āx V ar[ä K+1 ] =
d2
Z∞ Z∞
n| āx = v t · t px dt = t Ex dt n-yr temporary annuity due: äx: n
n n n−1
n
n| āx = v · n px āx+n = n Ex · āx+n
X
äx: n = E[Y ] = v k · k px
k=0
2
Ax: n − (Ax: n )2
n-yr certain and life annuity: āx: n V ar[Y ] =
d2
āx: n = āx + ā n − āx: n
= ā n + n| āx = ā n + n Ex · āx+n n-yr deferred annuity due: n| äx

∞
X
n| äx = E[Y ] = v k · k px
Most important identity
k=n
1 = δāx + Āx = äx − äx: n
1 − Āx = n Ex · äx+n
āx =
δ
Āx: n = 1 − δāx: n
2 n-yr certain and life due: äx: n
Āx: n = 1 − (2δ) 2āx: n
1 − Ax äx: n = äx + ä n − äx: n
äx =
d X ∞
1 − Ax: n = ä n + v k · k px
äx: n =
d k=n
1 = däx: n + Ax: n = ä n +n| äx

c
The Infinite Actuary, LLC 9
Whole life immediate: ax Accumulation function:
∞
X āx: n
ax = E[ä K ] = v k · k px s̄x: n =
k=1 n Ex

1 − (1 + i)Ax
ax = Limit of interest rate i = 0:
i
ax = äx − 1 i=0
ax −→ ex
i=0
äx −→ 1 + ex
m-thly annuities i=0
(m) āx −→ e̊x
1 − Ax
ä(m)
x = ax: n
i=0
−→ ex: n
d(m)
(m) (m) i=0
2
− (Ax )2
Ax äx: n −→ 1 + ex: n−1
V ar[Y ] =
(d(m) )2 āx: n
i=0
−→ e̊x: n
1
a(m)
x = ä(m)
x −
m
(m) (m) 1
ax: n = äx: n − (1 − n Ex )
m

c
The Infinite Actuary, LLC 10
 B.4: Benefit Premiums

Loss function: h-payment insurance premiums:

Loss = PVFB − PVFP
Āx
h P̄ (Āx ) =
Fully continuous equivalence premiums āx: h
(whole life and endowment only): Āx: n
h P̄ (Āx: n ) =
āx: h
Āx
P̄ (Āx ) = Ax
āx h Px =
äx: h
δ Āx
P̄ (Āx ) = Ax: n
1 − Āx h Px: n =
äx: h
1
P̄ (Āx ) = −δ
āx Pure endowment annual premium Px: n1 :
2
it is the reciprocal of the actuarial accumulated

P̄
Āx − (Āx )2
2 
V ar[L] = 1+ value s̈x: n because the share of the survivor who
δ
2
Āx − (Āx )2 has deposited Px: n1 at the beginning of each year
V ar[L] = assuming EP for n years is the contractual $1 pure endow-
(1 − Āx )2
ment, i.e.
Fully discrete equivalence premiums
(whole life and endowment only): Px: n1 s̈x: n = 1 (1)

Ax P minus P over P problems:

P (Ax ) = = Px The difference in magnitude of level benefit pre-
äx
dAx miums is solely attributable to the investment
P (Ax ) = feature of the contract. Hence, comparisons of
1 − Ax
1 the policy values of survivors at age x + n may
P (Ax ) = −d be done by analyzing future benefits:
äx

P 22

1
− Px:
n Px
1
n
Ax − (Ax )2 ( n Px − Px: n )s̈x: n = Ax+n =

V ar[L] = 1+ Px: n1
d
2
Ax − (Ax )2 Px: n − n Px
V ar[L] = assuming EP (Px: n − n Px )s̈x: n = 1 − Ax+n =
(1 − Ax )2 Px: n1
Two FCWL on (x): 1
Px: n − Px: 1
n
(Px: n − Px: n )s̈x: n = 1=
Px: n1
P̄1 → L1 P̄2 → L2
 2 Miscellaneous identities (whole life and
V ar[L2 ] P̄2 + δ
= endowment only):
V ar[L1 ] P̄1 + δ
P̄ (Āx: n )
Semicontinuous equivalence premiums: Āx: n =
P̄ (Āx: n ) + δ
Āx Px: n
P (Āx ) = Ax: n =
äx Px: n + d
m-thly equivalence premiums: 1
āx: n =
P̄ (Āx: n ) + δ
(m) A# 1
P# = (m) äx: n =
ä# Px: n + d

c
The Infinite Actuary, LLC 11
 B.5: Benefit Reserves

Benefit reserve t V :
The expected value of the prospective loss at h-payment reserves:
time t. h
t Vx = Ax+t − h Px äx+t:h−t
Reserve formulas (remove bars for dis- Variance of the loss function
crete versions):  2
P̄
Āx+t − (Āx+t )2
2 
V ar[ t L] = 1+
δ
• All 2
Āx+t − (Āx+t )2
V ar[ t L] = assuming EP
(1 − Āx )2
Prospective: PVFB - PVFP 
P̄
2 h i
V ar[ t L] = 1+ 2
Āx+t: n−t − (Āx+t: n−t )2
δ
t V̄ (Āx ) = Āx+t − P̄ (Āx )āx+t
2
Āx+t: n−t − (Āx+t: n−t )2
V ar[ t L] = assuming EP
Retrospective: AVPP - AVPB (1 − Āx: n )2

t V̄ (Āx ) = P̄ (Āx )s̄x: t −t k̄x Cost of insurance: funding of the accumu-

lated costs of the death claims incurred between
• Basic Level Insurance Only age x and x + t by the living at t, e.g.
dx (1 + i)3 + dx+1 (1 + i)2 + dx+2 (1 + i) + dx+3
Difference in Premium: 4 kx =
`x+4
  A1x: 4
t V̄ (Āx ) = P̄ (Āx+t ) − P̄ (Āx ) āx+t =
4 Ex
Paid-Up: dx qx
1 kx = =
`x+1 px
 
P̄ (Āx )
t V̄ (Āx ) = 1 − Āx+t
P̄ (Āx+t ) Accumulated differences of premiums: is
the difference between reserves.
• Whole Life and Endowment Ins. Only
1 n
( n Px − Px: n )s̈x: n = n Vx − n Vx:1 n
= Ax+n − 0 = Ax+n
Annuity: n
( n Px − Px )s̈x: n = n Vx − n Vx
āx+t = Px äx+n
t V̄ (Āx ) = 1 −
āx (Px: n − Px )s̈x: n = − n Vx
n Vx: n

Life Insurance: = 1 − n Vx

Āx+t − Āx
t V̄ (Āx ) = Relation between various terminal re-
1 − Āx
serves (whole life/endowment only):
Premium:
m+n+p Vx = 1−
P̄ (Āx+t ) − P̄ (Āx ) (1 − m Vx )(1 − n Vx+m )(1 − p Vx+m+n )
t V̄ (Āx ) =
P̄ (Āx+t ) + δ

c
The Infinite Actuary, LLC 12
 B.5: Benefit Reserves Continued

Notations:
bj : death benefit payable at the end of year of death for the j-th policy year
πj−1 : benefit premium paid at the beginning of the j-th policy year

Recursion relations:

hV + πh = v qx+h · bh+1 + v px+h · h+1 V

(h V + πh )(1 + i) = qx+h · bh+1 + px+h · h+1 V
(h V + πh )(1 + i) = h+1 V + qx+h (bh+1 − h+1 V )

Terminology:
“policy year h+1” ≡ the policy year from time t = h to time t = h + 1
“h V + πh ” ≡ initial benefit reserve for policy year h + 1
“h V ” ≡ terminal benefit reserve for policy year h
“h+1 V ” ≡ terminal benefit reserve for policy year h + 1

Net amount at Risk for policy year h + 1

Net Amount Risk ≡ bh+1 − h+1 V

Reserve Creation Formula:

For each premium P , the cost of providing the ensuing year’s death benefit , based on the net amount at
risk at age x + h, is : vqx+h (bh+1 − h+1 V ). The leftover, P − vqx+h (bh+1 − h+1 V ) is the source of reserve
creation. Accumulated to age x + n, we have:
n−1
X
nV = [P − vqx+h (bh+1 − h+1 V )] (1 + i)n−h
h=0
n−1
X
= P s̈ n − vqx+h (bh+1 − h+1 V )(1 + i)n−h
h=0

• If the death benefit is equal to the benefit reserve for the first n policy years

nV = P s̈ n

• If the death benefit is equal to $1 plus the benefit reserve for the first n policy years
n−1
X
n V = P s̈ n − vqx+h (1 + i)n−h
h=0

• If the death benefit is equal to $1 plus the benefit reserve for the first n policy years and qx+h ≡ q
constant

nV = P s̈ n − vqs̈ n = (P − vq)s̈ n

c
The Infinite Actuary, LLC 13
Reserves at fractional durations:

Exact method

( h V + πh )(1 + i)s = s px+h · h+s V + s qx+h v 1−s

Approximation

h+s V = (1 − s)( h V ) + (s)(h+1 V ) + (1 − s)(πh )

| {z }
unearned premium

Accounting Loss for year h + 1:

Λh ≡ losses incurred from time h to h + 1

V ar[Λh ] = v 2 (bh+1 − h+1 V )2 px+h qx+h

The Hattendorf theorem

V ar[ h L] = V ar[Λh ] + v 2 px+h V ar[ h+1 L]

= v 2 (bh+1 − h+1 V )2 px+h · qx+h + v 2 px+h V ar[ h+1 L]
V ar[ h L] = v 2 (bh+1 − h+1 V )2 px+h · qx+h
+v 4 (bh+2 − h+2 V )2 px+h · px+h+1 · qx+h+1
+v 6 (bh+3 − h+3 V )2 px+h · px+h+1 · px+h+2 · qx+h+2 + · · ·

c
The Infinite Actuary, LLC 14
 B.6: Multiple Life Functions

Joint survival function: Last survivor status T (xy):

t pxy = sT (x)T (y) (t, t) T (xy) + T (xy) = T (x) + T (y)

= P r[T (x) > t and T (y) > t] T (xy) · T (xy) = T (x) · T (y)
fT (xy) + fT (xy) = fT (x) + fT (y)
FT (xy) + FT (xy) = FT (x) + FT (y)
Joint life status T (xy):
t pxy + t pxy = t px + t py
FT (xy) (t) = P r[min(T (x), T (y)) ≤ t] Āxy + Āxy = Āx + Āy
= t qxy āxy + āxy = āx + āy
= 1 − t pxy e̊xy + e̊xy = e̊x + e̊y
exy + exy = ex + ey
Independant lives: n| qxy = n| qx + n| qy − n| qxy

t pxy = t px · t py
t qxy = 1 − t pxy Complete expectation of the last-survivor
status:
Complete expectation of the joint-life sta-
Z∞
tus:
e̊xy = t pxy dt
Z∞ 0
e̊xy = t pxy dt ∞
X
0 exy = k pxy
1

PDF joint-life status:

Variances:
fT (xy) (t) = t pxy · µxy (t) Z∞
fT (xy) (t) fT (xy) (t) V ar[T (u)] = 2 t · t pu dt − (e̊u )2
µxy (t) = =
1 − FT (xy) (t) t pxy 0
Z∞
Independant lives V ar[T (xy)] = 2 t · t pxy dt − (e̊xy )2
0
µxy (t) = µ(x + t) + µ(y + t) Z∞
fT (xy) (t) = t px · t py [µ(x + t) + µ(y + t)] V ar[T (xy)] = 2 t · t pxy dt − (e̊xy )2
0

Curtate joint-life functions:

Notes:
P r[K = k] = k| qxy For joint-life status, work with p’s:
= k pxy− k+1 pxy
∞ n pxy = n px · n py
X
E[K(xy)] = exy = k pxy
For last-survivor status, work with q’s:
1

n qxy = n qx · n qy

c
The Infinite Actuary, LLC 15
 Variance of insurance functions:
“Exactly one” status:
V ar[Z] = 2
Au − (Au )2
[1]
n pxy = n pxy − n pxy V ar[Z] = 2
Axy − (Axy )2
= n px + n py − 2 n pxy Cov [T (xy), T (xy)] = (e̊x − e̊xy ) (e̊y − e̊xy )
[1]
āxy = āx + āy − 2āxy Cov[v T (xy) , v T (xy) ] = (Āx − Āxy )(Āy − Āxy )

Common shock model:

Insurances:
Total force on (x) = µ∗x + λ
Total force on (y) = µ∗y + λ Āx = 1 − δāx
Total force on (xy) = µ∗x + µ∗x +λ Āxy = 1 − δāxy
Āxy = 1 − δāxy

µ∗x
Pr [(x) dies first] =
µ∗x + µ∗y + λ Premiums:
µ∗y 1
Pr [(y) dies first] = Px = −d
µ∗x + µ∗y + λ äx
λ 1
Pr [T (x) = T (y)] = Pxy = −d
µ∗x + µ∗y + λ äxy
1
Pxy = −d
äxy
µ∗x + λ
Āx =
µ∗x + λ + δ Annuity functions:
µ∗y + λ
Āy = Z∞
µ∗y + λ + δ
āu = v t · t pu dt
µ∗x + µ∗y + λ
Āxy = 0
µ∗x + µ∗y + λ + δ
Āu − (Āu )2
2

Āxy = Āx + Āy − Āxy V ar[Y ] =

δ2
Insurance functions:
∞
X Reversionary annuities:
Au = v k+1 · k pu · qu+k A reversioanry annuity is payable during the ex-
k=0 istence of one status u only if another status v
∞
X has failed. E.g. an annuity of 1 per year payable
= v k+1 · k| qu continuously to (y) after the death of (x).
k=0
∞
X āx|y = āy − āxy
Axy = v k+1 · k| qxy
k=0
∞
X
Axy = v k+1 · k| qxy
k=0

c
The Infinite Actuary, LLC 16
 Contingent Probabilities Contingent Insurance
1 1
n q xy = prob (x) dies before (y) Āxy → pays 1 at the death of (x)
and before n years from now if (y) is still alive
Z n Z ∞
= t px µ(x + t) t py dt = v t t px µ(x + t) t py dt
0 0

2
n q xy = prob (y) dies after (x) Āxy2 → pays 1 at the death of (y)
and before n years from now if predeceased by (x)
Z n Z ∞
= t py µ(y + t) t qx dt = v t t py µ(y + t) t qx dt
0 0

1 2
n q xy = n q xy + n qx · n py Āxy1 + Āxy2 = Āy
2 1
n q xy = n qy − n q xy

Prob (x) dies: SBP for a payment of 1 at the death of (x) if he

dies:
A. before (y):
2
∞ q xy A. before (y):
1
Āxy
B. more than n years after the death of (y):
2
n px ∞ q x+n:y B. more than n years after the death of (y):
2
C. within n years after the death of (y): n Ex Āx+n:y
2 2
∞ q xy − n px ∞ q x+n:y
C. within n years after the death of (y):
Pr(A) + Pr(B) + Pr(C) = 1 2
Āxy 2
−n Ex Āx+n:y

(A) + (B) + (C) = Āx

c
The Infinite Actuary, LLC 17
 B.7: Multiple Decrement Models

Notations:
(j) Probability density functions:
t qx = probability of decrement in the next
t years due to cause j Joint PDF: fT,J (t, j) = t p(τ ) (j)
x · µx (t)

(τ ) (j)
t qx = probability of decrement in the next Marginal PDF of J: fJ (j) = ∞ qx
t years due to all causes Z∞
Xm = fT,J (t, j)dt
(j)
= t qx 0
j=1
Marginal PDF of T : fT (t) = t p(τ ) (τ )
x · µx (t)
µ(j)
x = the force of decrement due only Xm
to decrement j = fT,J (t, j)
j=1

(j)
µ(τ
x
)
= the force of decrement due to all µx (t)
Conditional PDF: fJ|T (j|t) = (τ )
causes simultaneously µx (t)
Xm
= µ(j)
x Associated single decrement:
j=1
0 (j)
(τ ) t qx = rate of decrement from cause j only
t px = probability of surviving t years
Zt
 
despite all decrements 0 (j)
= exp − µ(j)
t px x (s)ds

= 1 − t qx(τ ) 0
Rt (τ ) 0 (j)
− µx (s)ds = 1 − t qx
= e 0

Force of mortality:
(τ )
Basic relationships:
d d (τ )
− dt `x+t dt t qx
µ(τ )
x (t) = =
 
(τ ) (τ )  Zt h
`x+t t px
i 
(τ )
t px = exp − µ(1) (m)
x (s) + · · · + µx (s) ds
d (j) d (j)
− dt `x+t dt t qx
 
0
µ(j)
x (t) = (τ )
= (τ ) m
`x+t t px (τ )
Y 0 (i)
t px = t px
i=1
Integral forms of t qx : 0 (j) (j)
t qx ≥ t qx
Zt
(j) (τ )
t qx = s px · µ(j)
x (s)ds
0
Zt
(τ ) (τ )
t qx = s px · µ(τ )
x (s)ds
0

c
The Infinite Actuary, LLC 18
 B.8: Models Including Expenses

Premiums:

G ≡ expense loaded ( or gross) premium

P ≡ benefit premium
e ≡ expense loading
G = P +e
ck ≡ percent of premium expense paid at time t = k
ek ≡ expenses paid per policy at time t = k

PV FP = PV FB + PV FE

Reserves

• Benefit Reserves (usually positive)

kV = PV FB − PV FP (prospective)
= AV P P − AV P B (retrospective)

• Expense Reserves (usually negative)

e
kV = PV FE − PV FL (prospective)
= AV P L − AV P E (retrospective)

• Expense-Augmented Reserve

= kV + kV e

Expense policy fee:

1. part of expense-loaded premium (therefore only paid when a premium is paid.
2. provides solely for payment of per policy expenses.
3. it is subject to expenses expressed as a percentage of premium.

c
The Infinite Actuary, LLC 19
Asset shares notations:

G ≡ level annual contract premium

k AS ≡ asset share assigned to the policy at time t = k
(d)
qx+k ≡ probability of decrement by death
(w)
qx+k ≡ probability of decrement by withdrawal
k CV ≡ cash amount due to the policy holder as a withdrawal benefit
bk ≡ death benefit due at time t = k

Recursion formula:
(d) (w) (τ )
[ k AS + G(1 − ck ) − ek ] (1 + i) = qx+k · bk+1 + qx+k · k+1 CV + px+k · k+1 AS
(d) (w)
= k+1 AS + qx+k (bk+1 − k+1 AS) + qx+k ( k+1 CV − k+1 AS)

Asset Share Creation formula:

n−1 (d) (w)
X G(1 − ch ) − eh − vqx+h bh+1 − vqx+h h+1 CV
n AS = (τ )
h=0 n−h Ex+h

c
The Infinite Actuary, LLC 20
 Constant Force of Mortality
Age does not matter
1
Future lifetime is exponential with mean µ
µ
• B.1 (I¯Ā)x = Āx āx =
(µ + δ)2
µ(x) = µ > 0, ∀x µ(1 + i)
s(x) = e−µx (I Ā)x = Āx äx =
(µ + δ)(q + i)
`x = `0 e−µx q(1 + i)
(IA)x = Ax äx =
= e−nµ = (px )n
n px (q + i)2
1 ¯ x = (āx )2 1
e̊x = = E[T ] = E[X] (Iā) =
µ (µ + δ)2
e̊x: n = e̊x (1 − n px ) 
1+i 2

1 (Iä)x = (äx )2 =
V ar[T ] = V ar[X] = 2 q+i
µ
mx = µ • B.4
ln2 1
Median[T ] = = Median[X] Px = vqx = Px: n
µ
px P̄ (Āx ) = µ = P̄ (Ā1x: n )
ex = = E[K]
qx
px For fully discrete whole life, w/ EP,
V ar[K] =
(qx )2 V ar[Loss] = p · 2Ax
• B.2
µ For fully continuous whole life, w/EP,
Āx =
µ+δ
2 µ V ar[Loss] = 2Āx
Āx =
µ + 2δ
1
Āx: n = Āx (1 − n Ex ) • B.5

n Ex = e−n(µ+δ) t V̄ (Āx ) = 0, t ≥ 0
µ
(I¯Ā)x = k Vx = 0, k = 0, 1, 2, . . .
(µ + δ)2
q For fully discrete whole life, assuming EP,
Ax =
q+i
2 q V ar[ k Loss] = p · 2Ax , k = 0, 1, 2, . . .
Ax =
q + 2i + i2
A1x: n = Ax (1 − n Ex ) For fully continuous whole life, assuming EP,
• B.3 V ar[ t Loss] = 2Āx , t ≥ 0
1
āx =
µ+δ
2 1
āx =
µ + 2δ
1+i
äx =
q+i
2 (1 + i)2
äx =
q + 2i + i2
āx: n = āx (1 − n Ex )
äx: n = äx (1 − n Ex )

c
The Infinite Actuary, LLC 21
 • B.6 Contingent
For two constant forces, i.e. µx acting on (x) 1 µx
and µy acting on (y), we have: n q xy = µx +µy n qxy
2 1
µx + µy n q xy = n qx − n q xy
Āxy =
µx + µ y + δ
1 1 µx
āxy = Āxy =
µx + µ y + δ µx + µy + δ
2 1
1 Āxy = Āx − Āxy
e̊xy =
µx + µ y
qxy • B.7
Axy =
qxy + i   qx(j)
(j) (τ )
1+i p0 x = p(τ
x
) qx
äxy =
qxy + i 1(j)
pxy Āx:t = Ā(j) (τ )
x (1 −t Ex )
exy =
qxy

c
The Infinite Actuary, LLC 22
 De Moivre’s Law
Out of ω births 1 dies every year
Future lifetime is uniform between 0 and ω − x

• B.1 ā ω−x
2
Āx = @ i∗ = 2i + i2
x ω−x
s(x) = 1 − a ω−x
ω Ax =
ω−x ω−x
`x = `0 an
ω A1x: n =
1 ω−x
qx = µ(x) = ¯
(Iā)
ω−x (I¯Ā)x =
ω−x
m ω−x
n|m qx =
ω−x (Ia) ω−x
ω−x−n (IA)x =
n px = ω−x
ω−x ¯
(Iā) n
t px µ(x + t) = qx = µ(x) = fT (x), 0 ≤ t < ω − x (I¯Ā)1x: n =
ω−x
1 (Ia) n
Lx = (`x + `x+1 ) (IA)1x: n =
2 ω−x
ω−x
e̊x = = E[T ] = Median[T ]
2
ω−x 1 • B.3
ex = − = E[K] No useful formulas: use B.2 formulae and
2 2
the most important identity.
(ω − x)2
V ar[T ] =
12
(ω − x)2 − 1
V ar[K] =
12 • B.6
qx 2dx ω−x
mx = 1 = e̊xx = (≡ MDML with µ = 2/(ω − x))
1 − 2 qx `x + `x+1 3
1 2(ω − x)
a(x) = E[S] = e̊xx =
2 3
n e̊xy = y−x x e̊yy + y−x qx e̊y
p
e̊x: n = n n p x + n qx
2
1 For two lives with different ω’s, simply translate
e̊x: n = ex: n + n qx one of the age by the difference in ω’s. E.g.
2
• B.2 Age 30, ω = 100 ⇔ Age 15, ω = 85
ā ω−x
Āx = Contingent
ω−x
ā n 2
n q xy = 1
if x + n and y + n ≤ ω
Ā1x: n = 2 n qxy
ω−x 1 2
n q xy = n qx − n q xy

1
āy:ω−x
Āxy =
ω−x
2 1
Āxy = Āx − Āxy

c
The Infinite Actuary, LLC 23
 Modified De Moivre’s Law

• B.1 • B.6
 x a
s(x) = 1− ω−x
e̊xx =
 ω a 2a + 1
ω−x 2a
`x = `0 ≡ e̊x with µ =
ω ω−x
a
µ(x) =
ω−x
ω−x−n a
 
n px =
ω−x
ω−x
e̊x = = E[T ]
a+1

c
The Infinite Actuary, LLC 24
 Uniform Distribution of Deaths (UDD)
for fractional ages

• B.1 • B.4
= t · qx
t qx i
qx P (Āx ) = Px
µ(x + t) = δ
1 − tqx i 1
P (Ā1x: n ) = P
`x+t = `x − t · dx δ x: n
sqx i 1
s qx+t = P (Āx: n ) = P + Px: n1
1 − tqx δ x: n
1 Px
V ar[T ] = V ar[K] + Px(m) =
12 α(m) − β(m)(Px + d)
1 qx (m) Px: n
mx = µ(x + ) = Px: n =
2 1 − 12 qx α(m) − β(m)(Px: 1
n + d)
1 n Px
Lx = `x − dx (m)
n Px =
2 α(m) − β(m)(Px: 1
n + d)
1
a(x) = (m) i 1(m)
2 hP (Ā1x: n ) = hP
1 δ x: n
e̊x: 1 = px + qx
2 • B.5
t px µ(x + t) = qx (m)
k Vx = k Vx + β(m)Px(m) k Vx
• B.2
i • B.7
Āx = Ax
δ UDDMDT
i
Ax(m) = Ax (j)
= t · qx(j)
i(m) t qx
i 1 qx(j) = µ(j)
x (0) for integral values of x
Ā1x: n = A
δ x: n qx(τ ) = µ(τ )
x (0) for integral values of x
i
(I Ā)1x: n = (IA)1x: n  qx(j)
δ 0 (j)

(τ ) qx(τ )
i 1 i t px = t px
Āx: n = Ax: n + Ax:1n = A1x: n + n Ex (j)
δ δ qx
i µ(j)
x (t) = (τ )
n| Āx = Ax 1 − t qx
δ n|
(τ )
2i + i2 2 qx
2
Āx = Ax µ(τ )
x (t) = (τ )
2δ 1 − t qx
• B.3
UDDAST
ä(m)
x = α(m)äx − β(m)(1 − 0)
0 (j) 0
(m)
äx: n = α(m)äx: n − β(m)(1 − n Ex ) t qx = t · qx(j)
 
0 (1) 1 0 (2)
(m)
n| äx = α(m)n| äx − β(m)(n Ex − 0) m=2: qx(1) = qx 1− q
2 x
symbol with (m) = α(m) × symbol w/o (m)  
0 (1) 1 0 (2) 1 0 (3) 1 0 (2) 0 (3)
−β(m)(“start” − “finish”) m = 3 : qx(1) = qx 1− q − qx + qx · qx
2 x 2 3
id
with: α(m) = (m) (m) ≈ 1 q 0 (j)
x
i d µ(j)
x (t) = (j)
i − i(m) m−1 1 − t q0x
and β(m) = (m) (m) ≈
i d 2m

c
The Infinite Actuary, LLC 25
 Hyperbolic Assumption (aka Balducci)
for fractional ages

• B.1
t · qx
t qx =
1 − (1 − t)qx
qx
µ(x + t) =
1 − (1 − t)qx
qx p x
t px µ(x + t) =
[1 − (1 − t)qx ]2
1 1 1
= (1 − t) + (t)
`x+t `x `x+1
2`x `x+1
`x+0.5 =
`x + `x+1
sqx
s qx+t =
1 − (1 − s − t)qx

c
The Infinite Actuary, LLC 26
 Markov Chains

Properties

1. discrete time

2. finite number of states

3. history independence - probability distribution at time n may depend on state in at time n, but not
states prior to time n

Notation

• Mn ≡ state # at time n

• Qn is the transition probabilities from time n to n + 1

– If Qn = Q for all n, then chain said to be homogeneous

(i,j)
• k Qn = Pr [Mn+k = j | Mn = i]

– probability that subject in state # i at time n is in state # j at time n + k

– (i, j) entry of k Qn where k Qn = Qn × Qn+1 × . . . × Qn+k−1
(i)
• k Pn = Pr [Mn+1 = Mn+2 = . . . = Mn+k = i | Mn = i]

– probability a subject in state # i at time n remains in that state through time n + k

• (i,j)
`+1 C - denotes the cash flow at time ` + 1 if the subject is state # i at time ` and state # j at
time ` + 1

• ` C (i) - denotes the cash flow at time ` if subject is in state # i at time `

c
The Infinite Actuary, LLC 27
 Poisson Process

Counting Process

• N (t) ≡ total # of “events” that occur by time t

• N (t) must satisfy:

1. N (t) ≥ 0
2. N (t) is integer valued
3. If s < t, then N (s) ≤ N (t)
4. For s < t, N (t) − N (s) equals the # of events that occur in (s, t]

Poisson Process

1. counting process N (t)

2. N (0) = 0

3. independent increments

4. # of events in any interval of length t is distributed Poisson with mean = λt

(λt)n
Pr {N (t + s) − N (s) = n} = e−λt · n = 0, 1, 2, . . .
n!

Interarrival Times

• Tn denotes the time between (n − 1)st and nth event

• Tn ∼ exponential (mean = λ1 )

Waiting Time

• Sn = ni=1 Ti n ≥ 1
P

• Sn ∼ gamma (α = n, θ = λ1 )
n
• mean = αθ = λ
n
• var = αθ2 = λ2

• Pr{Sn ≤ t} = Pr {N (t) ≥ n} = 1 − Pr {N (t) < n}

Different Classes for each Event

• Events occur at a Poisson rate λ. Each event is classified as Type I, II, III, etc

• Each Type of event occurs at a Poisson rate = λ × Pr {Type | Event}

Markov Chains
If the states of Markov Chain have Poisson distributions at time 0, then they maintain their Poisson distri-
butions at each duration.

c
The Infinite Actuary, LLC 28
Probability of Classes vary by Time

• Pi (y) = probability classified as i at time y

• Ni (t) = # of type i events occurring by time t

Rt
• Ni (t) ∼ Poisson ( mean = λ 0 Pi (s)ds)

Non-homogeneous Poisson Process

• λ(t) varies by time t

• For interval (t1 , t2 ) # of events Poisson with mean

Z t2
m(t) = λ(t)dt
t1

Compound Poisson Process

• Number of events (N ) is a Poisson random variable and amount of each event (X) is independent and
identically distributed
N
X
• S= Xi
i=1

• E[S] = λt E[X]

• Var[S] = λt E[X 2 ]

c
The Infinite Actuary, LLC 29