Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Contents
1 Survival Models 2
1.1 Survival, distribution, and density functions . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 The force of mortality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Parametric classes and the future lifetime random variable . . . . . . . . . . . . . . . . 6
1.4 Deferred mortality probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 Curtate and expected future lifetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6 Temporary expected future lifetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4 Life Annuities 45
4.1 Introduction to life annuities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.2 Temporary and deferred life annuities . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3 Guaranteed annuities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.4 The UDD assumption for life annuities . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.5 Life annuities payable continuously . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.6 Increasing Annuities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Chapter 1
Survival Models
Let X be a continuous and non-negative random variable, then X is an appropriate random variable
to model the age at which an individual dies and is therefore called the age at death random variable.
Definition: The survival distribution function of X is the probability that a newborn survives to at
least age x and is defined as
S0 (x) = Pr(X > x) (1.1)
This function is also known as the decumulative distribution function but more commonly it is simply
referred to as the survival function. For a function to be a legitimate survival function it must satisfy
all three of the following properties
• S0 (0) = 1,
• limx→∞ S0 (x) = 0,
Solution: We need to check that S0 (x) satisfies the three conditions previously given. By substituting
for x = 0 it can be seen that S0 (0) = 1, and similarly for x = 100, S0 (100) = 0. To determine whether
the function is non-increasing we consider the first derivative. Differentiating once with respect to x
gives
−1
1
x −2
S0′ (x) = 200
1 − 100
Since S0′ (x) is non-positive for all x in the given domain, S0 (x) must be non-increasing. All three
properties are satisfied so we conclude that S0 (x) is a legitimate survival function.
Note that in the previous example there is a slight alteration to the second property and the limit is
taken as x tends to 100 as opposed to infinity, this is becasue the survival function is only defined
for x ≤ 100. In survival models a limiting age is often defined since it is not necessary to consider
unrealistically large values of x. The limiting age is denoted by ω and is usually between 100 and 120,
the limiting age in the previous example is ω = 100.
Definition: The cumulative distribution function of X is the probability that a newborn dies before
age x and is given by
F0 (x) = Pr(X ≤ x) (1.2)
This function is more often referred to as the distribution function and as with the survival function
there are three properties that any legitimate distribution function must satisfy. These properties are
• F0 (0) = 0,
• limx→∞ F0 (x) = 1,
• F0 (x) is a non-decreasing function.
Since surviving beyond age x and dying before age x are complementary events, there exists a simple
relationship between the survival and distribution functions, namely,
The final function used to calculate mortality probabilities is the probability density function (p.d.f).
Definition: Assume that for the r.v. X there exists a continuously differentiable survival function
S0 (x), then X has a continuous distribution function F0 (x) and the probability density function may
be defined as
d
f0 (x) = F (x)
dx 0
Given the p.d.f there also exist integral expressions for the survival and distribution functions
R∞ Rx
S0 (x) = x
f0 (s) ds F0 (x) = 0
f0 (s) ds
Occasionally a question may ask to find the p.d.f given only the survival function, in which case an
obvious approach may be to first calculate the distribution function and then differentiate to find the
p.d.f. However, by using equation (1.3) we can obtain the following expression for the density in terms
of the survival function.
d d d
f0 (x) = F0 (x) = (1 − S0 (x)) = − S0 (x) (1.4)
dx dx dx
The following example demonstrates how the functions defined above can be used to calculate mortality
probabilities.
What is the probability that a newborn dies between the ages of 65 and 75?
Let µ(x) be a non-negative real-valued function, µ(x) may be considered as a force of mortality if and
only if the following properties are satisfied
From the definition of the force of mortality, we can obtain an equation linking it with both the survival
function and p.d.f. The following derivation uses the formal definition of a derivative seen in second
year analysis courses in addition to equation (1.4) from Section 1.
1
µx = lim Pr (X ≤ x + ∆x|X > x)
∆x→0 ∆x
1 Pr(X > x) − Pr(X > x + ∆x)
= lim
∆x→0 ∆x Pr(X > x)
1 S0 (x) − S0 (x + ∆x)
= lim
∆x→0 ∆x S0 (x)
1 S0 (x + ∆x) − S0 (x)
=− lim
S0 (x) ∆x→0 ∆x
1 d
=− S0 (x)
S0 (x) dx
f0 (x)
=⇒ µx = (1.6)
S0 (x)
By an application of the chain rule, an alternative equation for the force of mortality is
1 d d
µx = − S0 (x) = − log (S0 (x)) (1.7)
S0 (x) dx dx
Note that it is easier to solve this example using equation (1.6) instead of (1.7) so it is important to learn
both. Equation (1.7) is only particularly useful when the survival function is written as an exponent,
however, it can also be used to derive one of the most important formulae in the course. Integrating
equation (1.7) between 0 and y gives
Z y
µx dx = − [log (S0 (y)) − log (S0 (0))]
0
but since S0 (0) = 1, it follows that log (S0 (0)) = 0 and therefore
Z y Z y
µx dx = −log (S0 (y)) ⇐⇒ S0 (y) = exp − µx dx
0 0
In keeping with the notation previously used, a simple exchange of variables gives
Z x
S0 (x) = exp − µs ds (1.8)
0
Example: A survival model is defined with force of mortality µs = ksn for constants k and n such that
s ≥ n, k > 0 and n ≥ 1. Find an expression for the survival function S0 (x).
Solution: Substituting the expression for the force of mortality into equation (1.8) gives
Z x
n
S0 (x) = exp − ks ds
0
k n+1 x
= exp − s 0
n+1
n+1
kx
= exp −
n+1
The force of mortality in the previous example is in fact the force of mortality for the Weibull model,
one of a number of parametric models defined in the following section.
1
f0 (x) =
ω
Using this, along with relations defined in previous sections, expressions can be obtained for the survival
function, distribution function, and force of mortality.
Z x Z x
1 x
F0 (x) = f0 (s) ds = ds =⇒ F0 (x) =
0 0 ω ω
ω ω
1
Z Z h z iz=ω x
S0 (x) = f0 (z) dz = dz = =⇒ S0 (x) = 1 −
x x ω ω z=x ω
f0 (x) 1 ω 1
µx = = . =⇒ µx =
S0 (x) ω ω−x ω−x
1
Example: Given that µx = 100−x
,0 ≤ x ≤ 100, find an expression for the survival function S0 (x).
Solution: There are two ways to approach this question, either we can recognise that the given force
of mortality is in the required form for De Moivre’s model with a limiting age ω = 100 and hence
x
S0 (x) = 1 − 100 , or, a safer approach is to find the expression explicitly using equation (1.8).
Z x
1
S0 (x) = exp − dz = exp ([log(100 − z)]z=x
z=0 )
0 100 − z
= exp (log(100 − x) − log(100))
x
= exp log 1 −
100
x
=1−
100
The table on the following page provides a list of the key parametric models used in actuarial science
along with the corresponding survival function, force of mortality, and any limitations on the parameters
involved. Makeham’s law is another commonly used model since the force of mortality is comprised
of two components, a constant force that is independent of current age, and a force that increases
exponentially with age. Note that a special case of Makeham’s law is when A=0 and the model reduces
to Gompertz’s law.
Definition: If X is the age at death random variable and x is the current age of an individual, the
future lifetime can be defined by the random variable
Tx = X − x (1.9)
Consider an individual currently aged x who survives for another t years, the probability of this is given
by the survival function
Sx (t) = Pr(Tx > t)
Similarly, the probability that the individual dies within t years is given by the distribution function
Fx (t) = Pr(Tx ≤ t)
When we consider Sx (t) we make the underlying assumption that the life has survived from birth
through to age x before surviving from age x to x + t. This assumption allows us to write Sx (t) in
terms of the age at death random variable X and results in a very important equation relating S0 (x)
and Sx (t).
Pr(X > x + t)
Sx (t) = Pr(Tx > t) = Pr(X > x + t|X > x) =
Pr(X > x)
S0 (x + t)
=⇒ Sx (t) = (1.10)
S0 (x)
The actuarial notation used to define the survival and distribution functions is as follows, this notation
will be used in place of the existing notation throughout the remainder of the course.
Sx (t) = t px Fx (t) = t qx
In actuarial notation, the equation linking the survival and distribution functions may therefore be
written as
t p x + t qx = 1
It is also common practice that in the case where t = 1 the first subscript is omitted and we simply
write px and qx for the survival and distribution functions.
The p.d.f of the future lifetime random variable is given in terms of the distribution and survival
functions in a similar way as before.
d d
fx (t) = Fx (t) = − Sx (t) (1.11)
dt dt
The force of mortality may also be defined for the future lifetime random variable as
1
µx+t = lim Pr(t < Tx < t + ∆t|Tx > t)
∆t→0 ∆t
Recall that equations (1.6) and (1.7) give two different expressions for the force of mortality of a life
aged x. Equivalent expressions also exist for the force of mortality at any age x+t, t > 0, the derivations
of which are similar to before but instead begin with the definition of µx+t as opposed to µx . For this
reason the derivations are not included but the results are as follows
fx (t) d
µx+t = and µx+t = − log (Sx (t))
Sx (t) dt
Finally we state without proof the analogue to equation (1.8) for calculating the survival function Sx (t)
given only the force of mortality. The proof may be attempted as an exercise and is a simple application
of equation (1.10) and (1.8).
Z t
Sx (t) = exp − µx+s ds (1.12)
0
There also exists a second equation for the deferred mortality probability, but before we show this we
need to consider the multiplicative property of the survival function. Suppose we are interested in
the probability that a life currently aged x survives for another 2 years but only know the probability
of survival for each year separately. The probability of surviving for 2 years may be written as the
probability of surviving the first year, multiplied by the probability of surviving the second year. In
actuarial notation this may be written as
2 px = px · px+1
It is important to note that in general, the multiplicative rule only holds for the survival function and
not the distribution function, that is
t+u qx 6= t qx · u qx+t
The following example demonstrates how useful the multiplicative rule can be.
Example: You are given 5 p50 = 0.9, 10 p50 = 0.8, and q55 = 0.03. Find the probability that an individual
Consider again the deferred probability of an individual aged x dying between x + t and x + u + t. The
individual survives the deferred period with probability t px and dies in the period of mortality with
probability u qx+t , therefore using the multiplicative rule, the deferred probability may be written as
Definition: For a life aged x, if Tx is the future lifetime random variable, the curtate future lifetime
random variable is defined as
Kx = ⌊Tx ⌋
Proof:
Pr(Kx = k) = Pr(k ≤ Tx < k + 1)
= Pr(Tx > k) − Pr(Tx > k + 1)
= k px − k+1 px
= k |qx
The distribution function for Kx is given by
Proof:
Pr(Kx ≤ K) = Pr(Kx = 0) + Pr(Kx = 1) + ... + Pr(Kx = k)
= 0| qx + 1| qx + ... + k| qx
k
X k
X
= j| qx = (j px − j+1 px ) using equation (1.13)
j=0 j=0
k
X k
X
= j px − j+1 px
j=0 j=0
We do not derive an expression for the survival function of Kx as this can easily be deduced from
knowing the distribution function.
Since t px → 0 as t → ∞, evaluating the first term at each end point yields zero and so the only non-zero
term is the second term. Therefore,
Z ∞
e̊x = t px dt (1.17)
0
If both the first and second moments are known, the variance of Tx may be calculated using
We may also be interested in the expected value of Kx , known as the curtate expectation of life and
denoted in actuarial notation by ex . To obtain an expression for ex we use the definition of the expected
value of a discrete random variable as follows
∞
X ∞
X
E[Kx ] = kPr(Kx = k) = k · k| qx
k=0 k=0
X∞
= k (k px − k+1 px )
k=0
∞
X ∞
X
= k · k px − k · k+1 px
k=0 k=0
= (px + 22 px + 33 px + ...) − (2 px + 23 px + 34 px + ...)
= px + 2 px + 3 px + ...
therefore,
∞
X
ex = k px (1.19)
k=1
There also exists a relationship linking the curtate and standard expectations of life based on methods
from numerical analysis for evaluating definite integrals. Recall that the trapezium rule is based on the
approximation Z b
f (a) + f (b)
f (x) dx ≈ (b − a)
a 2
1
e̊x ≈ + ex (1.20)
2
The temporary expectation of life is denoted by e̊x:n and the curtate temporary expectation of life by
ex:n . The formulae remain the same as in the previous section with the exception of the upper limit
which changes from infinity to n, the formulae are therefore
Z n n
X
e̊x:n = t px dt and ex:n = k px
0 k=1
The following equation relates the temporary and standard expectations of life.
Proof:
Z n Z ∞
e̊x:n + n px e̊x+n = t px dt + n px t px+n dt
0 0
Z n Z ∞
= t px dt + t+n px dt
0 0
Z n Z ∞
= s px ds + s px ds
0 n
Z ∞
= s px ds
0
= e̊x
We finish this chapter with a worked example that makes use of some of the key points and equations
provided so far.
1
Example: A survival model has force of mortality µx = 2(100−x) , 0 ≤ x ≤ 100
(i) Find an expression for the survival function S0 (x).
(ii) Find an expression for the survival function t px .
(iii) Compute e̊36 , the expected future lifetime for an individual aged 36.
Solution: To obtain the survival function for part (i) we need to use equation (1.8) with the force of
mortality given in the question.
Z x
1
S0 (x) = exp − dz
0 2(100 − z)
z=x
1 h 1 z=x
i
= exp − − log(100 − z) = exp log(100 − z) 2
2 z=0 z=0
12 !
100 − x
= exp log
100
x 12
= 1−
100
For part (ii) we can choose from one of two methods, we can either use equation (1.10) with our result
from part (i), or we can take a more direct approach using equation (1.12).
21
t
= 1−
100 − x
s=t !
1
= exp − log(100 − x − s)
2 s=0
12 !
100 − x − t
= exp log
100 − x
12
t
= 1−
100 − x
the same result as found using the first method. Note that in this case the first method is only quicker
because we had already calculated S0 (x) earlier in the question. Finally, for part (iii) we simply need
to substitute our expression for t px into equation (1.17) with x = 36.
Z 64 21
t
e̊36 = 1− dt
0 64
Z 0 Z 1
1 1
= −64 u du = 64 2 u 2 du
1 0
128 h 3 iu=1
= u2
3 u=0
128
= ≈ 42.67
3
where the substitution u = 1 − 64t has been used to evaluate the integral. It is important to note that
in the initial integral the upper limit is 64 instead of infinity as stated in the general formula for e̊x .
This is because the survival model is only defined for x ≤ 100 and therefore for an individual currently
aged 36, they can only survive for a maximum of 64 years.
Chapter 2
A life table is a tabular representation of mortality evolution for a group of lives. Initially the number
of lives is denoted by l0 , the radix of the life table, and is usually in excess of 1000. Let lx be the number
of individuals alive at age x, the probability that a newborn survives for x years is then the number of
individuals alive at age x as a fraction of the total number of lives at the start, that is,
lx
S0 (x) = x p0 = (2.1)
l0
The life table may not always begin with newborn lives and instead starts at some integer age x, if this
is the case we need an expression for the survival function Sx (t). We know that lx is the number of
individuals alive at age x, so lx+t is the number of individuals alive at age x + t. By the same reasoning
as before
lx+t
Sx (t) = t px = (2.2)
lx
The number of deaths that occur between x and x + 1 is calculated using
dx = lx − lx+1 (2.3)
and it is usual in integer tabulated life tables that this information is included in one of the columns.
The number of deaths between x and x + t is given by
t dx = lx − lx+t (2.4)
The probability that a life aged x dies within the next t years may be written as
lx+t lx − lx+t t dx
t qx = 1 − t px = 1 − = = (2.5)
lx lx lx
Some intuitive formulae that can be useful include
∞
X t−1
X
lx = dx+k and t dx = dx+k
k=0 k=0
The first formula states that the number of individuals alive at age x is equal to the total number of
deaths in each of the remaining years, the second states that the number of deaths that occur in the
next t years is equal to the sum of the number of deaths in each of the following t years.
The force of mortality can also be calculated using life tables. Recall from chapter 1 that the force of
mortality may be written as
d 1 d t px
µx+t = − log (t px ) = −
dt t px dt
Using equation (2.2) and noticing that lx is constant under differentiation with respect to t,
lx d lx+t
µx+t = −
lx+t dt lx
1 dlx+t
=−
lx+t dt
As with the force of mortality, the equations for the expected future lifetime remain the same as in
chapter 1 with our life table expression of the survival function substituted accordingly. The complete
expectation of life may therefore be written as
Z ∞ Z ∞
lx+t
e̊x = t px dt = dt
0 0 lx
1 ∞
Z
= lx+t dt
lx 0
1 ∞
Z
= lz dz
lx x
where the change of variable z = x + t is used in the final line.
Solution: We first need to find an expression for the survival function t px . Using equation (2.2) we
have
lx+t (1 + x)2
p
t x = =
lx (1 + x + t)2
The mean of Tx is then given by
∞ ∞
(1 + x)2
Z Z
e̊x = p
t x dt = dt
0 0 (1 + x + t)2
Z ∞
2
u=∞
u−2 du = −(1 + x)2 u−1 u=1+x
= (1 + x)
1+x
(1 + x)2
=0− −
1+x
=1+x
Solution: For the first part we use the equation for the curtate expectation of life, it is only necessary
to sum over the first 4 terms as these are the only non-zero terms given the age of the individual.
∞ 4
X X l95+k
e95 = k p95 =
k=1 k=1
l95
l96 + l97 + l98 + l99
=
l95
7212 + 4637 + 2893 + 1747
=
10902
= 1.5125
To calculate the variance of K95 we first need to find the second moment E[Kx2 ]. For this we do not
have a simple equation as we do for the first moment so we instead use the definition of the second
moment for a discrete random variable.
∞
X ∞
X 4
X
2 2 2
E[K95 ] = k Pr(K95 = k) = k k| q95 = k 2 (k p95 − k+1 p95 )
k=0 k=0 k=0
4
X l95+k − l95+k+1
= k2
k=0
l95
l96 − l97 16(l99 − l100 )
=0+ + ... +
l95 l95
7212 16(1747 − 0)
= 10902 + ... +
4637 10902
= 4.3861
Finally, the temporary curtate expectation of life can be calculated in the same way as the curtate
expectation of life from part (i) with the upper limit of the sum now set equal to 3.
3 3
X X l95+k
e95:3 = k p95 =
k=1 k=1
l95
l96 + l97 + l98
=
l95
7212 + 4637 + 2893
=
10902
= 1.3522
Equation (2.6) is one of the most important equations in this section as it allows us to derive expressions
for the survival, distribution, and density functions, and the force of mortality under the UDD
assumption. For integer x and 0 ≤ t ≤ 1 we have
lx+t lx − tdx tdx
t px = = =1− = 1 − tqx using equation (2.5)
lx lx lx
d t px d
fx (t) = − = − (1 − tqx ) = qx (2.7)
dt dt
f0 (x) qx
µx+t = =
t px 1 − tqx
It is not necessary to learn all four of the expressions above since by knowing any one of them the others
may be easily determined. The most useful expression is that for the distribution function as this is the
simplest to apply when calculating probabilities.
Example: You are given the following extract from a life table
x lx
35 97250
36 97126
37 96993
Under the UDD assumption, compute the probabilities (i) 0.5 p35 , (ii) 1.5 p35
The UDD assumption can now be applied to evaluate 0.5 p36 while p35 can be calculated using standard
results.
l37 96993
0.5 p36 = 1 − 0.5 q36 = 1 − 0.5q36 = 1 − 0.5 1 − = 1 − 0.5 1 − = 0.9993
l36 97126
l36 97126
p35 = = = 0.9987
l35 97250
The overall probability is therefore
As with equation (2.6) in the previous lecture, equation (2.9) is one of the key equations for the constant
force of mortality assumption and we now use it to derive expressions for the survival function, density
function, and force of mortality.
t
(lx )1−t (lx+1 )t
lx+t lx+1
t px = = = =⇒ t px = (px )t (2.10)
lx lx lx
=⇒ t qx = 1 − (px )t
To find the p.d.f we recall that when differentiating ax with respect to x we instead apply the chain
rule to exp(xln(a)), therefore,
d t px d d
fx (t) = − = − (px )t = − exp(tlog(px )) = −ptx log(px )
dt dt dt
Finally, the force of mortality is given by
fx (t) pt log(px )
µx+t = =− x t = −log(px )
t px px
Note that since 0 ≤ px ≤ 1, log(px ) ≤ 0 and so the force of mortality and p.d.f are non-negative as
required.
Example: You are given the following extract from a life table.
x lx
80 53925
81 50987
82 47940
83 44803
Under the assumption of a constant force of mortality between integer ages, calculate the probabilities
(i) 0.75 p80 , (ii) 2.25 p80
Solution: Assuming a constant force of mortality, the most useful equation to calculate probabilities
is (2.10). Since for the first part t ≤ 1, this equation can be applied straight away to give
0.75 0.75
0.75 l81 50987
0.75 p80 = (p80 ) = = = 0.95885
l80 53925
For the second part we use the multiplicative property of the survival function to give
where
l82 47940
2 p80 = = = 0.8890
l80 53925
and 0.25 0.25
0.25 l83 44803
0.25 p82 = (p82 ) = = = 0.9832.
l82 47940
Solution: We know
l[35]+1
p[35] = =⇒ l[35]+1 = p[35] · l[35] = (1 − q[35] ) · l[35] = (1 − 0.013) · 1000 = 987
l[35]
and hence
l[35]+2
p[35]+1 = =⇒ l[35]+2 = p[35]+1 · l[35]+1 = (1 − q[35]+1 ) · l[35]+1 = (1 − 0.012) · 987 = 975.16
l[35]+1
Solution If we notice that since the select period is 2 years, after 3 years the life will no longer be select
and l[x]+3 = lx+3 , then
l[46]+3 l49 965
2 p[46]+1 = = = = 0.9415
l[46]+1 l[46]+1 1025
For the second part
l[45]+2.1 l47.1
0.4 p[45]+1.7 = =
l[45]+1.7 l([45]+1)+0.7
Now, using equation (2.6) and considering the numerator and denominator separately
and
l([45]+1)+0.7 = l[45]+1 − 0.7(l[45]+1 − l[45]+2 ) = l[45]+1 − 0.7(l[45]+1 − l47 )
= 1124 − 0.7(1124 − 1039)
= 1064.5
Example: A select and ultimate life table has a 3 year select period. You are given the following
information
1
• q[x] = 6
• lx+6 = 90000
4
• 5 p[x+1] = 5
9
• 3 p[x]+1 = p
10 3 [x+1]
Evaluate l[x]
1
Solution: We can start by considering which equations involve lx , since we know q[x] = 6
we can easily
determine that p[x] = 56 and so the most obvious choice is
l[x]+1 l[x]+1 6
p[x] = =⇒ l[x] = = · l[x]+1
l[x] p[x] 5
We therefore need to find an expression for l[x]+1 . Using the final bullet point, we have
l[x]+4 9 l[x+1]+3
= ·
l[x]+1 10 l[x+1]
But since the select period is 3 years, l[x]+4 = l[x+1]+3 = lx+4 , and these terms cancel on each side.
This leaves
1 9 1 10
= · =⇒ l[x]+1 = · l[x+1]
l[x]+1 10 l[x+1] 9
We now need to compute the value of l[x+1] which may be done using the remaining two bullet points
lx+6 lx+6 5
5 p[x+1] = =⇒ l[x+1] = = 90000 · = 112500
l[x+1] 5 p[x+1] 4
and therefore
10
l[x]+1 = · 112500 = 125000
9
Finally, our expression for l[x] is
6
l[x] = · 125000 = 150000
5
Chapter 3
Term insurance: A term insurance pays a benefit on the death of the policyholder provided they
die within a period of n years as specified in the contract, no payment is made if the life survives the
duration of the term. Premiums are often paid at the start of each year for the duration of the term.
Whole life insurance: A whole life insurance pays a benefit on the death of the policyholder whenever
this may occur. Premiums are generally paid at the start of each year given the individual is alive,
however, sometimes they are payable up to some maximum age to avoid the problems older lives might
face in paying them.
Pure Endowment insurance: A pure endowment insurance pays a benefit to the policyholder at the
end of an n year period provided they are still alive, similar to a term insurance, premiums are paid
annually for the duration of the term.
The following results should be familiar from previous courses in financial mathematics but are provided
anyway. Given an annual effective interest rate i, the annual interest rate convertible mthly satisfies
m
i(m)
1+i= 1+
m
and if the interest is compounded continuously we use the force of interest which is given by
δ = log(1 + i)
The annual effective discount rate can be defined in terms of the discount factor v as
1
d = iv = 1 − v where v= = e−δ
1+i
Finally, given d, the annual discount rate convertible mthly satisfies
m
d(m)
1−d= 1−
m
v x = e−δTx
T
if Tx ≤ n
Z=
0 if Tx > n
The expected value of Z, sometimes referred to as the actuarial present value (APV) and denoted by
1
Āx:n , is
Z n Z n
1 −δt
Āx:n = E[Z] = e fx (t) dt = e−δt t px µx+t dt (3.1)
0 0
In actuarial notation, the bar above the A denotes that the benefit is payable immediately on death
and the ‘1’ above the x indicates that for the benefit to be paid the life must die before the end of the
n year term. To find the second moment of the random variable Z, we consider the rule of moments.
Suppose we want to find the j th moment, we would use the following formula
Z n Z n Z n
−δt j −(δj)t
j
e−γt fx (t) dt
E[Z ] = e fx (t) dt = e fx (t) dt =
0 0 0
where γ = δj can be thought of as a new force of interest. Since the final equality gives our formula for
the expectation of Z, we have a simple rule for finding the second moment.
E[Z 2 ]δ = E[Z]|2δ
(3.2)
That is, if E[Z] is evaluated with a force of interest δ, E[Z 2 ] is equal to E[Z] evaluated with a force
of interest equal to 2δ. For a term insurance the second moment of Z is denoted by 2 Āx:n 1
and so the
formula for calculating the variance is
2
Var[Z] = 2 Āx:n
1 1
− Āx:n (3.3)
Example: A survival model is defined with a constant force of mortality µx+t = µ. For an n year
term insurance find the EPV of a unit benefit payable immediately on death and the variance of the
present value random variable.
Solution: To find the EPV of the benefit we need the survival function given a constant force of
mortality. Using results from the first chapter it can be shown that
t px = e−µt
Example: An individual currently aged 40 purchases a 15-year term insurance that pays a benefit of
£50,000 immediately on death. If the force of interest is δ = 0.05 and
x
S0 (x) = 1 − 0 ≤ x ≤ 100
100
S0 (x + t) 100 − x − t 100 t
t px = = · =1−
S0 (x) 100 100 − x 100 − x
Since we have not been given the force of mortality, it is much easier to compute the EPV of the benefit
using the formula involving the density, where
d 1
f40 (t) = − t p40 =
dt 60
The EPV of a unit benefit is therefore
15 15
1
Z Z
1 −δt
A40:15 = e f40 (t) dt = e−δt dt
0 0 60
1 −δt 15
=− e 0
60δ
1
1 − e−15δ = 0.1759
=
60δ
However, since the policy pays £50,000 on death, the EPV of the benefit is
1
£50, 000Āx:n = £50, 000 × 0.1759 = £8, 793.89
Using the rule of moments, the second moment of the present value random variable is
1
E[Z 2 ] = 1 − e−30δ = 0.1295
120δ
When finding the variance we need to make sure we find it for the total benefit and not just a unit
benefit
Var[50000Z] = 500002 Var[Z] = 500002 2 Ā40:15
1 1
)2
− (Ā40:15
= 500002 (0.1295 − 0.17592 )
= 246, 397, 975
Z = v Tx = e−δTx
The expected value of Z is denoted in actuarial notation by Āx and is given by the following formula
Z ∞ Z ∞
−δt
Āx = e fx (t) dt = e−δt t px µx+t dt (3.4)
0 0
The variance of Z is
Var[Z] = 2 Āx − (Āx )2
where the second moment can again be calculated using the rule of moments. In fact, for all the insurance
functions considered in this course the rule of moments can be applied to find higher moments, this will
not be the case with the annuity functions considered in chapter 4.
Example: A survival model has constant force of mortality µx+t = 0.01. A life currently aged 30
purchases a whole life insurance policy with sum insured £1. Find the EPV of the benefit if the force
of interest is δ = 0.05.
Solution: We could solve this question by using equation (3.4) and evaluating the integral, however,
there is an alternative method using a result from a previous example. In the first term insurance
example the survival model also had a constant force of mortality and we found that for the general
case
1 µ
1 − e−(µ+δ)n
Āx:n =
µ+δ
A whole life insurance is simply the limiting case of a term insurance as n → ∞, therefore, to find Āx
we only need to consider the limit as n → ∞ of the equation above.
1 µ
=⇒ Āx = lim Āx:n =
n→∞ µ+δ
Z is a discrete random variable that takes only two values, 0 with probability n qx , and v n with probability
n px , therefore, the expected value is given by
Ax:n1 = n Ex = v n n px (3.5)
The ‘1’ over the n indicates that for the benefit to be paid the term must expire before the life and there
is no bar over the A as there only exists a discrete time version for a pure endowment insurance. This
notation is not favoured due to its similarity to the notation for the EPV of a term insurance benefit,
so for this reason the more convenient notation n Ex is used instead. The second moment of Z can be
calculated in the same way as for any discrete random variable
2
Ax:n1 = (v n )2 · n px = v 2n n px
v x = e−δTx
T
if Tx < n
Z= n −δn
v =e if Tx ≥ n
Therefore, the expectation of Z is equal to the sum of the expectations of the term and pure endowment
present value random variables, that is
1
Āx:n = Āx:n + n Ex (3.6)
The variance of Z is 2
Var[Z] = 2 n| Āx − n| Āx
where the rule of moments can be used to calculate the expectation of the squared present value.
Remember also that since there is always a limiting age with the De Moivre model, we should never
have the upper limit of an integral set equal to infinity. If a life is currently aged x and the limiting
age is ω, they can only survive for a maximum of ω − x years so this would be the upper limit of the
integral.
1
Proof (i): For De Moivre’s model it can be shown that fx (t) = ω−x
, using this along with equation
(3.4) gives
∞ ω−x
1
Z Z
−δt
Āx = e fx (t) dt = e−δt dt
0 0 ω−x
1 −δt t=ω−x
=− e t=0
δ(ω − x)
1 − e−δ(ω−x)
=
δ(ω − x)
āω−x
=
ω−x
(ii) Starting with the definition of the EPV of a unit benefit for a term insurance
Z n Z n
1 −δt 1 1 − e−δn
Āx:n = e fx (t) dt = e−δt dt =
0 0 ω−x δ(ω − x)
1 t=n
e−δt t=0
=−
δ(ω − x)
ān
=
ω−x
Proof: Z n Z ∞ Z ∞
1 −δt −δt
Āx:n + n| Āx = e fx (t) dt + e fx (t) dt = e−δt fx (t) dt = Āx
0 n 0
The second relationship links whole life, deferred, and pure endowment insurances
Proof:
Z ∞ Z ∞
−δt
n| Āx = e fx (t) dt = e−δt t px µx+t dt
Zn ∞ n
Example: Consider a whole life insurance with a benefit of £1000 payable immediately on death, the
force of interest and force of mortality are
0.04 0 ≤ t ≤ 10 0.006 0 ≤ t ≤ 10
δ= , µ=
0.05 t > 10 0.007 t > 10
Solution: Since the force of interest and force of mortality change after 10 years, we need to consider
the EPV in two separate parts, from t = 0 to t = 10 and then from t = 10 onwards. For the first
part a benefit is paid provided the life dies within the first 10 years, this is how we define a 10-year
1
term insurance so the EPV is given by Āx:10 . For the second part a benefit is paid on death given the
life survives for the first 10 years, this is how we define a deferred life insurance so the EPV is 10| Āx .
Therefore, the total EPV is
1
Ax = Āx:10 + 10| Āx
We now need to evaluate each of the insurance functions using the appopriate values for the force of
mortality and force of interest. Note that in both cases the force of mortality is constant so the density
function is given by fx (t) = µe−µt . For the term insurance we have
Z 10 Z 10
1 −δt −µt µ −(δ+µ)t t=10
Āx:10 = e µe dt = µ e−(δ+µ)t dt = − e t=0
0 0 µ+δ
µ
1 − e−10(δ+µ)
=
µ+δ
0.006
1 − e−0.46
=
0.046
= 0.04809
= 0.06945
Example: A life currently aged 40 purchases a whole life insurance policy that pays £10 should death
occur in the first 10 years and £5 in subsequent years, in both cases the benefits are paid immediately
on death. De Moivre’s model is used with limiting age ω = 100 and the force of interest is δ = 0.05.
Find the EPV of the death benefit.
Solution: As with the previous example we need to consider the EPV from t = 0 to t = 10 and then
t = 10 onwards, however, in this case there are two approaches that can be taken. The first method
treats the insurance as the sum of a 10-year term insurance paying £10 and a 10-year deferred insurance
paying £5 whereas for the second method we think of the insurance as the sum of a whole life insurance
paying £5 and a 10-year term insurance also paying £5. Graphs showing how each method splits up
the benefit are given below, the first method is to the left and the second is on the right.
Using equation (3.8) it may be shown that the two methods are equivalent
1 1 1 1
5Ā40 + 5Ā40:10 = 5 Ā40:10 + 10| Ā40 + 5Ā40:10 = 10Ā40:10 + 510| Ā40
Therefore, which method you use depends on whether you find it easier to visualise dividing the benefit
according to age or the amount of benefit.
1
The density function for De Moivre’s model with a life aged 40 is fx (t) = 60 , so
60
1 1 −δt t=60 1
Z
e−δt −3
Ā40 = dt = − e t=0
= 1 − e = 0.31674
0 60 60δ 3
10
1 1 −δt t=10 1
Z
1
e−δt −0.5
Ā40:10 = dt = − e t=0
= 1 − e = 0.13116
0 60 60δ 3
60
1 1 −δt t=60 1 −0.5
Z
e−δt −3
10| Ā40 = dt = − e t=10
= e − e = 0.18558
10 60 60δ 3
As before, we can find higher moments using the rule of moments but since for discrete insurances we
use the annual effective interest rate rather than the force of interest, the rule differs slightly. Starting
from the definition of the j th moment, we want to find a new rate α that allows us to write the higher
moments in terms of the first. The j th moment of Z is given by
∞ ∞
k+1 j
X X
j
v j(k+1) k| qx
E[Z ] = v k| qx =
k=0 k=0
Therefore, for discrete insurances using an effective rate of interest i, the rule of moments states that
the second moment satisfies
E[Z 2 ]i = E[Z]|(1+i)2 −1
(3.12)
Note that if a life dies in the last year of the term, the payment will be made at t = n but Kx = n − 1,
hence why the upper limit of the sum is n − 1. The variance of Z is
2
Var[Z] = 2 Ax:n
1 1
− Ax:n (3.15)
Example: A survival model is defined with constant force of mortality µx+t = µ, show that the EPV
of a unit benefit for a whole life insurance may be written as
eµ − 1
Ax =
eµ+δ − 1
Solution: Using the defintion of Ax
∞
X ∞
X
k+1
Ax = v k| qx = v k+1 k px qx+k
k=0 k=0
For a constant force of mortality we know that the survival function is given by k px = e−µk and since
this is independent of age, it follows that px+k = e−µ and hence qx+k = 1 − e−µ . Substituting these
expressions into the previous equation and writing the discount factor in exponential form gives
∞
X
−µ
Ax = (1 − e ) e−δ(k+1) e−µk
k=0
∞
X
−δ −µ
= e (1 − e ) (e−(δ+µ) )k
k=0
−δ −µ
e (1 − e )
=
1 − e−(δ+µ)
eµ − 1
= (δ+µ)
e −1
where the following formula for an infinite geometric series has been used to evaluate the sum
∞
X a
ark =
k=0
1−r
As with the continuous case, the EPV is equal to the sum of the EPV for a term and pure endowment
insurance benefit, that is
1
Ax:n = Ax:n + n Ex (3.16)
The variance of Z is then given by
The variance of Z is 2
Var[Z] = 2n| Ax − n| Ax (3.19)
The two relationships that we previously derived for continuous insurance functions also apply in the
discrete case, that is
1
Ax = Ax:n + n| Ax (3.20)
Proof:
n−1
X ∞
X ∞
X
1 k+1 k+1
Ax:n + n| Ax = v k| qx + v k| qx = v k+1 k| qx = Ax
k=0 k=n k=0
n| Ax = n Ex Ax+n (3.21)
Proof:
∞
X ∞
X
k+1
n| Ax = v k| qx = v j+n+1 j+n| qx
k=n j=0
∞
X ∞
X
n j+1 n
=v v j+n px · qx+j+n = v v j+1 j px+n · n px · qx+j+n
j=0 j=0
∞
X
= v n n px v j+1 j px+n · qx+j+n
j=0
= n Ex Ax+n
1 1
Ax:n = vqx + vpx Ax+1:n−1 (3.23)
Proof:
n−1
X n−1
X
1 k+1
Ax:n = v k| qx = v k+1 k px · qx+k
k=0 k=0
n−1
X
= vqx + v k+1 k px · qx+k
k=1
n−1
X
= vqx + vpx v k k−1 px+1 · qx+k
k=1
n−2
X
= vqx + vpx v j+1 j px+1 · qx+j+1 (using the substitution j = k − 1)
j=0
n−2
X
1
= vqx + vpx v j+1 j| qx+1 = vqx + vpx Ax+1:n−1
j=0
1
= vqx + vpx Ax+1:n−1 + n−1 Ex+1
There also exist recursive formulae for the second moments of the present value random variable,
however, since the proofs are similar to those for the first moment, we only prove the case for a whole
life insurance.
2
Ax = v 2 qx + v 2 px · 2 Ax+1 (3.26)
Proof:
∞
X ∞
X
2 2(k+1)
Ax = v k| qx = v 2k+2 k px · qx+k
k=0 k=0
∞
X
= v 2 qx + v 2k+2 k px · qx+k
k=1
∞
X
= v 2 qx + v 2(j+1)+2 j+1 px · qx+j+1 (using j = k − 1)
j=0
X∞
= v 2 qx + v 2j+4 j px+1 · px · qx+j+1
j=0
∞
2 2
X 2
= v qx + v p x v j+1 j px+1 · qx+j+1
j=0
= v 2 qx + v 2 px · Ax+1 2
For the other types of insurance the recursive formulae for the second moment are
2 1
Ax:n = v 2 qx + v 2 px · 2 Ax+1:n−1
1 2
Ax:n = v 2 qx + v 2 px · 2 Ax+1:n−1 2
n| Ax = v 2(n+1) n px · qx+n + 2n+1| Ax
Example: An individual aged 41 purchases a whole life insurance policy with benefit payable at the
end of the year of death. You are given
• i = 0.05
• p40 = 0.9972
Solution: We know that Var[Z] = 2 A41 − (A41 )2 so we need to find expressions for 2 A41 and A41 . Using
the recursive formula for a whole life insurance benefit,
Since we are told that A41 − A40 = 0.00822, we have a set of simultaneous equations that may be solved
to give A41 = 0.215169.
Similarly,
2
A40 = v 2 qx + v 2 px · 2 A41 = v 2 (1 − px ) + v 2 px · 2 A41
2 2
1 1
= (1 − 0.9972) + 0.9972 · 2 A41
1.05 1.05
2
=⇒ A40 = 0.002537 + 0.90452 A41
and since 2 A41 − 2 A40 = 0.00433, solving these equations gives 2 A41 = 0.071925. Substituting the values
found for A41 and 2 A41 into the formula for the variance gives
1
Kx(m) = ⌊mTx ⌋
m
(2)
For example, suppose a life aged x survives for a further 23.675 years, then Kx = 23, Kx = 23.5,
(4) (12) 8
Kx = 23.5, and Kx = 23 12 .
(m)
The probability mass function for Kx is given by
(m)
1 1
Pr Kx = k = Pr k < Tx < k + = Pr (Tx > k) − Pr Tx > k +
m m
= k px − k+ 1 px
m
= k| 1 qx
m
We now use this to give expressions for the insurance functions when a benefit is payable at the end of
the m1 th year of death, as before, the present value random variable is denoted by Z
(m) 1 (m) P∞ k+1
+m
Whole life insurance: Z = v Kx =⇒ E[Z] = Ax = k=0 v m · k 1
| qx
m m
( (m) 1
+m (m) 1
v Kx if Kx ≤ n − Pmn−1 k+1
Term insurance: Z= (m)
m =⇒ A(m)1x:n = k=0 v m k 1
| qx
0 if Kx ≥ n m m
( (m) 1
+m (m) 1
v Kx if Kx ≤ n − (m)
Endowment insurance: Z= (m)
m =⇒ Ax:n = A(m)1x:n + n Ex
vn if Kx ≥ n
These equations are not always efficient, for example, if we had a 10 year term insurance with benefit
payable at the end of the month of death our summation would include 120 terms. We therefore
introduce a set of equations based on the UDD assumption that gives an mthly death benefit EPV in
terms of an annual death benefit EPV. A proof is given for the whole life case, this can be extended for
the remaining equations.
i
Ax(m) = Ax (3.27)
i(m)
Proof: We can consider a whole life insurance as the sum of deferred one-year term insurances. In the
first year we have A(m)1x:1 but in the second year the term insurance A(m) x+1:1
1
is contingent on the life
survivng the first year and must be discounted by one year. Therefore, the whole life insurance may be
written as
∞
X
Ax(m) = v k k px A(m) x+k:1
1
k=0
where
m−1 m−1 m−1
X j+1 X j+1 X j+1
A(m) x+k:1
1
= v m j 1
| qx+k = v m j px+k − j+1 px+k = v m j+1 qx+k − j qx+k
m m m m m m
j=0 j=0 k=0
j j+1
From the UDD assumption we know that for 0 < s ≤ 1, t qx = tqx and since 0 < m
< m
≤ 1.
Applying this to the previous line gives
m−1 m−1
X j+1 j+1 j qx+k X j+1
A(m) x+k:1
1
= v m qx+k − qx+k = vm
j=0
m m m j=0
Using standard results about geometric series we can evaluate the sum as follows
m−1
" 1
#
m
q x+k 1
X 1 j q
x+k 1 1 − (v m)
A(m) x+k:1
1
= vm vm = · vm 1
m j=0
m 1 − vm
qx+k 1−v iv
= · −1 = qx+k 1
m vm −1 m((1 + i) m − 1))
iv
= qx+k (m)
i
(m)
Substituting this expression into the initial equation for Ax gives
∞ ∞
X i i X i
Ax(m) = v k+1
k px qx+k (m) = v k+1 k| qx = Ax
k=0
i i(m) k=0
i(m)
The formulae for the other insurance functions are
i i (m) i
A(m)1x:n = (m)
1
Ax:n (m)
n| Ax = A
(m) n| x
Ax:n = 1
Ax:n + n Ex (3.28)
i i i(m)
1
where i(m) is the interest rate convertible m
thly and satisfies
m
i(m)
1+i= 1+
m
Example: Consider a 3 year term insurance with sum insured £1000 issued to a select life aged 50.
The benefit is payable at the end of the quarter of death, you may assume that deaths are uniformly
distributed and mortality follows the life table given below. Calculate the EPV of the benefit given the
interest rate is i = 5%.
[x] l[x] l[x]+1 lx+2 x+2
50 9706 9687 9661 52
51 9680 9660 9630 53
52 9653 9629 9596 54
and therefore,
i 0.05
A(4) [50]:3
1
= 1
A[50]:3 = (0.0070523) = 0.0071832
i(4) 0.0490889
Finally, the EPV of the £1000 benefit is
£1000A(4) [50]:3
1
= £1000 × 0.0071832 = £7.1832
i
Āx = Ax (3.29)
δ
X∞ Z 1
−δk
= k px e e−δs s px+k µx+s+k ds
k=0 0
∞ 1
eδ
X Z
k+1 δ(1−s)
= k px v e s px+k µx+s+k ds multiplying by δ = 1
k=0 0 e
X∞ Z 1
k+1
= v k px qx+k eδ(1−s) ds
k=0 0
Ax −δ(s−1) 1 Ax δ Ax
=− e 0
= e − 1 = [(1 + i) − 1]
δ δ δ
i
= Ax
δ
To get from line 5 to 6 we recall equation (2.7) which states that under the UDD assumption fx (t) = qx ,
therefore, s px+k µx+s+k = fx+k (s) = qx+k . The corresponding equations for term, deferred, and
endowment insurance functions are
1 i 1 i i 1
Āx:n = · Ax:n n| Āx = · n| Ax Āx:n = · Ax:n + n Ex
δ δ δ
Example: For a fully continuous whole life insurance issued to a life aged 40 you are given
• The death benefit is £1000 in the first year and increases by £500 each year for the following 3
years before remaining level at £5000 thereafter.
• Mortality follows the Life Table at the end of the lecture note at i = 5%.
Calculate the EPV of this benefit. If we consider splitting the EPV as shown in the graph below
EPV = 1000Ā40 + 500E40 Ā41 + 5002 E40 Ā42 + 5003 E40 Ā43 + 25004 E40 Ā44
Note that we could have also split the EPV into a series of term insurances, however, since we are using
the Illustrative Life Table, whole life insurances are more suitable.
Since the UDD assumption holds
i
EPV = 500 · [2A40 + E40 A41 + 2 E40 A42 + 3 E40 A43 + 54 E40 A44 ]
δ
i l41 2 l42 3 l43 4 l44
= 500 · 2A40 + v A41 + v A42 + v A43 + 5v A44
δ l40 l40 l40 l40
Substituting the appropriate values from the Illustrative Life Table gives
500 × 0.05 1 99285.88 5 99104.33
EPV = 2(0.12106) + (0.12665) + ... + (0.14496)
ln(1.05) 1.05 99338.26 1.054 99338.26
= £613.4038
Continuous Case
We first consider an insurance policy with arithmetically increasing benefit Tx payable immediately on
death, that is, the benefit is equal to the number of years a life currently aged x survives for.
Since the benefit is equal to Tx , the present value random variable is
Z = Tx v Tx = Tx e−δTx
The I here implies the benefit is increasing and the bar above the I denotes that the increases are
continuous. For an increasing term insurance the corresponding formula is
Z n Z n
(I¯Ā)x:n
1
= te −δt
fx (t) dt = te−δt t px µx+t dt . (3.31)
0 0
An interesting case is when the increase in the benefit is geometric. In this case the expected present
value for the term assurance becomes (there is not much scope for whole life policies as the EPV will
fairly quickly become too high to be realistic). In this case the benefit paid at the time of death is
(1 + j)t . The EPV for the term insurance is then:
Z n
EPV = (1 + j)t v t fx (t) dt
Z0 n
1+j t
= ( ) fx (t) dt
1+i
Z0 n
= v ∗t fx (t) dt
0
1
= (Ā)x:n ,
i∗
1+i
where i∗ = 1+j
− 1.
Discrete Case
Now consider a discrete increasing benefit that increases by 1 each year, a benefit of 1 is paid at the
end of the year of death given death occurs in the first year, 2 in the second year, and k + 1 if death
occurs between the ages of x + k and x + k + 1. The present value random variable is therefore defined
as
(Kx + 1)v Kx +1
The EPV of the death benefit is then denoted in actuarial notation by (IA)x where
∞
X
(IA)x = v k+1 (k + 1)k| qx (3.32)
k=0
A useful relationship between term and whole increasing assurances is the following (proof is an exercise):
1
(IA)x:n = (IA)x − v n n px [nAx+n + (IA)x+n ] (3.34)
Remark: Equations for the EPV can be derived for a number of other patterns including benefits
that decrease over time, though for the latter only term assurances can be considered in order to avoid
the possibility of negative benefits.
Remark: One can consider benefit increases that take place more often than annually in a similar
way.
Example: Consider a 20-year increasing term assurance. The benefit is payable at the end of year
of death and it is £10, 000 in the first year increasing by £100 in every subsequent year. Assume i = 5%
and mortality that follows the tables at the end of the lecture notes. Find the EPV of this assurance.
1 1
EPV = (10000 − 100)A50:20 + 100IA50:20
l70
= 9900(A50 − v 20 A70 )
l50
l70
+ 100 (IA)50 − v 20 [20A70 + (IA)70 ]
l50
1 91082.43
= 9900(0.18931 − 0.42818)
1.0520 98576.37
1 91082.43
+ 100 5.825499653 − [20 · 0.42818 + 6.742450463]
1.0520 98576.37
= 1923.705
Example: Consider a 20-year decreasing term assurance. The benefit is payable at the end of year
of death and it is £10, 000 in the first year decreasing by £100 in every subsequent year. Assume i = 5%
and mortality that follows the tables at the end of the lecture notes. Find the EPV of this assurance.
1 1
EPV = (10000 + 100)A50:20 − 100IA50:20
l70
= 10100(A50 − v 20 A70 )
l50
l70
− 100 (IA)50 − v 20 [20A70 + (IA)70 ]
l50
1 91082.43
= 10100(0.18931 − 0.42818)
1.0520 98576.37
1 91082.43
− 100 5.825499653 − 20
[20 · 0.42818 + 6.742450463]
1.05 98576.37
= 11862.495.
Chapter 4
Life Annuities
An n-year annuity due is the present value of a series of unit payments made at the beginning of each
year from t = 0 to t = n − 1, we have the equation
n−1
X 1 − vn
än = vk = (4.1)
k=0
d
An n-year annuity immediate is the present value of a series of unit payments made at the end of each
year, although it is often easier to think of this as the beginning of the following year. The payments
are therefore made at t = 1 through to t = n and we have
n
X 1 − vn
an = vk = (4.2)
k=1
i
An n-year continuous annuity is the present value of continual payments such that one unit is paid
across each year, the formula is given by
Z n
1 − e−δn
ān = e−δt dt = (4.3)
0 δ
1
For both annuities due and immediate we can consider the case where a unit payment is made every m
years in installments of m1 , the formulae are
mn−1 m
1 − vn d(m)
(m) 1 X k
än = v m = (m) where 1 − d = 1− (4.4)
m k=0 d m
mn m
1 − vn i(m)
(m) 1 X k
an = v m = (m) where 1 + i = 1+ (4.5)
m k=1 i m
Since premiums tend to start immediately on issue we focus more on annuities due and continuous
annuities.
Since an annuity certain is not easy to evaluate when inside a sum, we can use equation (4.1) to find a
simpler formula for the EPV.
∞ ∞ ∞
1 − v k+1
X X 1X
1 − v k+1 k px − 1 − v k+1 k+1 px
äx = äk+1 · k| qx = (k px − k+1 px ) =
k=0 k=0
d d k=0
1
(1 − v) + (1 − v 2 )px + (1 − v 3 )2 px + ... − (1 − v)px + (1 − v 2 )2 px + (1 − v 3 )3 px + ...
=
d
1
(1 − v) + (v − v 2 )px + (v 2 − v 3 )2 px + ...
=
d
1
(1 − v) + v(1 − v)px + v 2 (1 − v)2 px + ...
=
d
1−v
1 + vpx + v 2 2 px + ...
=
d
X∞
= v k k px
k=0
We may also take a less formal approach to deriving equation (4.6) by representing the cash flow using
a timeline. By considering the amount paid at the start of each year, the appropriate discount factor,
and the probability that each payment occurs, we can find the EPV of each individual payment.
The total EPV is then the sum of the EPV for all individual payments, therefore, the EPV of a whole
life annuity due is
∞
X
2 3
äx = 1 + vpx + v 2 px + v 3 px + ... = v k k px
k=0
Although equation (4.6) is a useful equation for äx , it may not be used to find expressions for higher
moments or the variance of Y . This is because the individual terms are dependent, the probability that
the life survives two years depends on whether they survived the first year. If we needed to calculate
the second moment of Y we would need to use the original expression involving an annuity certain, that
is,
∞
X 2
E[Y 2 ] = äk+1 k| qx
k=0
As before, this is not an easy approach to take so we now show how the expectation and variance may
be calculated using the whole-life insurance functions defined in the previous chapter. Recall that the
present value random variable for a whole life insurance benefit payable at the end of the year of death
is Z = v Kx +1 , using equation (4.1)
1 − v Kx +1
Y = äKx +1 =
d
so the expectation of Y is
1 − E v Kx +1
1 − v Kx +1
1 − Ax
E[Y ] = E = =
d d d
1 − Ax
äx = =⇒ Ax = 1 − däx (4.7)
d
Now recall that the variance of the whole life present value random variable is Var[Z] = 2 Ax − (Ax )2 ,
since the variance of a constant is zero, we have
1 − v Kx +1
1 1
= 2 Var v Kx +1 = 2 2 Ax − (Ax )2
Var[Y ] = Var
d d d
For this life annuity it is much easier to find the EPV using the less formal approach and consider the
cash flow as a time-line.
If we consider the EPV for each individual payment separately we see that the total EPV may be
written as
Therefore, our formula to calculate the EPV of a temporary life annuity due is
n−1
X
äx:n = v k k px (4.8)
k=0
As with the whole life annuity we can derive expressions relating the expectation and variance of Y
with the insurance functions defined previously. Recall that the present value random variable for an
endowment insurance benefit is given by Z = v min(Kx +1,n) , then
1 − E v min(Kx +1,n)
1 − v min(Kx +1,n)
h i 1 − Ax:n
E[Y ] = E ämin(Kx +1,n) = E = =
d d d
Hence,
1 − Ax:n
äx:n = =⇒ Ax:n = 1 − däx:n (4.9)
d
1
It is important to note that although equation (4.8) may resemble the equation for Ax:n , the EPV of
a term insurance benefit, the annuity is in fact related to the endowment insurance function Ax:n . An
expression for the variance of Y may be found in a similar way
1 − v min(Kx +1,n)
1 1
= 2 Var v min(Kx +1,n) = 2 2
Ax:n − (Ax:n )2
Var[Y ] = Var
d d d
A timeline showing the cash flow for a deferred annuity due is given on the following page, using this
approach the EPV of a deferred annuity due is
∞
X
=⇒ n| äx = v k k px (4.10)
k=n
The expression for the second moment of Y is too complex and since Y can not easily be written in
terms of the present value random variable for any insurance function we do not consider an equation
for the variance. However, if asked to find the variance we can do so as we would for any discrete
random variable.
Example: A life currently aged 95 purchases an insurance contract and pays premiums at the start of
each year starting in two years time. Mortality follows the life table given below and the interest rate
is 6%. Find the EPV of the annuity and the variance of the present value random variable.
x 95 96 97 98 99 100
lx 1000 750 400 225 75 0
Solution: Using equation (4.10) the EPV of a 2-year deferred annuity due is
∞ 4
X
k
X
k l95+k
2| ä95 = v k p95 = v
k=2 k=2
l95
1 400 1 225 1 75
= 2
· + 3
· + 4
·
1.06 1000 1.06 1000 1.06 1000
= 0.6043
1 − v Kx −1
0 if Kx ≤ 1
Y = where ä =
v 2 äKx +1−2 if Kx ≥ 2 Kx −1
d
and the variance may be calculated by evaluating the first and second moments using the following
table.
k Pr(Kx = k) = k| qx Y Y Pr(Kx = k) Y 2 Pr(Kx = k)
0 0.25 0 0 0
1 0.35 0 0 0
2 0.175 0.8900 0.1557 0.1386
3 0.15 1.7296 0.2594 0.4487
4 0.075 2.5217 0.1891 0.4769
5 0 3.6730 0 0
5
X 5
X
E[Y ] = Y Pr(Kx = k) = 0.60432 , E[Y 2 ] = Y 2 Pr(Kx = k) = 1.06428
k=0 k=0
2
=⇒ Var[Y ] = E[Y 2 ] − (E[Y ]) = 1.06428 − (0.60432)2 = 0.699075
Recall that the deferred insurance function may be written as a discounted whole life insurance function,
the same is also true for annuities, that is
Proof:
∞
X ∞
X
k
n| äx = v k px = v j+n j+n px (using j = k − n)
k=n j=0
∞
X
n
= v n px v j j px+n
j=0
= n Ex äx+n
We can also show that the sum of a deferred and temporary annuity is equal to a whole life annuity
Finally, as with the insurance functions, we can derive a recursive formula for a whole life annuity due
which could be used to create a table of the EPV for consecutive ages.
= 1 + Ex äx+1
By manipulating the expression for Y we can derive an equation for the EPV of a guaranteed annuity
än if Kx ≤ n − 1 än if Kx ≤ n − 1
Y = =
äKx +1 if Kx ≥ n än + äKx +1 − än if Kx ≥ n
0 if Kx ≤ n − 1
= än +
äKx +1−n if Kx ≥ n
= än + Y ∗
where Y ∗ is the present value random variable for a deferred annuity due. The expected value of Y is
denoted in actuarial notation by äx:n and is therefore
• ä50:20 = 18.0606
• ä60 = 14.134
• ä70 = 10.375
• i = 4%
Find ä60:10 .
Solution: We know from equation (4.12) that a temporary annuity may be written as a whole life
annuity minus a deferred annuity
We know the values of ä60 and ä70 so we only need to determine 10 E60 . Using equation (4.15) and the
first bullet point, we have
1 − v 10
ä50:10 = ä10 + 10 E50 ä60 = + 10 E50 ä60
d
1
10
1 − 1.04
=⇒ 17.5661 = 1
+ 14.134 · 10 E50
1 − 1.04
⇐⇒ 10 E50 = 0.6460
1 − v 20
ä50:20 = ä20 + 20 E50 ä70 = + 20 E50 ä70
d
1
20
1 − 1.04
=⇒ 18.0606 = 1
+ 10.375 · 20 E50
1 − 1.04
⇐⇒ 20 E50 = 0.3785
Pure endowment insurance functions satisfy a similar multiplicative property as survival functions, so
20 E50 0.3785
⇐⇒ 10 E60 = = = 0.5859
10 E50 0.6460
We now have the three values necessary to evaluate the temporary annuity, therefore
Alternatively, using the timeline given on the following page and considering the EPV of each individual
payment
1 1 2 3
äx(m) = 1 + v m 1 px + v m 2 px + v m 3 px + ...
m m m m
∞
1 X k
=⇒ äx(m) = v m k px (4.17)
m k=0 m
(m)
(m) 1 − Ax:n
äx:n = (4.18)
d(m)
If we consider the timeline approach the formula for the temporary annuity differs from the whole life
annuity only in the limit of the sum. For the temporary annuity we have a total of mn payments and
since the first payment is made immediately we sum from k = 0 to k = mn − 1. Therefore,
mn−1
(m) 1 X k
äx:n = v m k px (4.19)
m k=0 m
The variance of Y is
(m)
" 1
#
1 − v min(Kx +m ,n)
1
2 (m) (m)
Var[Y ] = Var = 2 Ax:n − (Ax:n )2
d(m) (d(m) )
As with the annual case, it is much easier to determine the EPV using the timeline approach, doing so
gives
∞
(m) 1 X k
n| äx = v m k px (4.20)
m k=mn m
We do not consider an expression for the variance of Y for a deferred annuity payable m times a year.
id i − i(m)
äx(m) = α(m)äx − β(m) where α(m) = , β(m) = (4.21)
i d(m)
(m) i(m) d(m)
(m)
Proof: Recall that under the UDD assumption we have the following equation relating Ax and Ax
i
Ax(m) = Ax
i(m)
We may then write a whole life annuity due payable m times a year as
(m) i
(m) 1 − Ax 1 − i(m) Ax
äx = (m)
= (m)
=⇒ i(m) d(m) ä(m)
x = i(m) − iAx
d d
= α(m)äx − β(m)
Example: (Q3 Jan. 2013) A 30-year temporary life annuity due is issued to a life aged 45 with
benefits payable monthly. For the first 10 years the benefit is £10, increasing to £30 thereafter.
Mortality follows the Tables at the end of the lecture notes with i = 5%. Calculate the EPV of
this annuity.
Solution: Since the benefit changes during the term of the contract we need to consider the EPV in
two parts. We can think of the first 10 years as a temporary annuity paying £10 a month and the
second 20 years also as a temporary annuity but instead paying £30 a month and starting from age
55. Note that we must multiply the second annuity by 10 E45 to ensure we find the present value and
consider the probability that the life survives to age 55. Remember also that a unit annuity due payable
m times a year pays m1 at the start of each period, not 1, so when evaluating the EPV we need to make
sure we use the total amount paid each year. Taking all of this into account the EPV may be written
as
(12) (12)
EPV = 120ä45:10 + 36010 E45 ä55:20
(12) (12) (12) (12)
= 120 ä45 − 10 E45 ä55 + 36010 E45 ä55 − 20 E55 ä75 (using equations (4.11) and (4.12))
(12)
ä75 = α(12)ä75 − β(12) = (1.000197 × 10.31778) − 0.466508 = 9.853305
1 − v1 E v Kx +1
1 − v Kx
1 − (1 + i)Ax
a x = E aK x =E = =
i i i
An alternative formula for the EPV of a whole life annuity immediate may be found by considering the
sum of the EPV of each individual payment, this gives
∞
X
ax = vpx + v 2 2 px + v 3 3 px + ... = v k k px (4.22)
k=1
This equation can now be used to derive an expression relating the EPV of a whole life annuity due
and annuity immediate.
∞
X ∞
X
k
äx = v k px = 1 + v k k px = ax + 1 =⇒ äx = ax + 1 (4.23)
k=0 k=1
Y = amin(Kx ,n)
If we compare this to a temporary annuity due we can see that the annuity due has an additional
payment at t = 0 whilst the annuity immediate has an additional payment at t = n, therefore, we have
the relationship
1 + ax:n = äx:n + v n n px (4.25)
Y = āTx
If we recall that the present value random variable for a whole life insurance benefit payable at the time
of death is Z = e−δTx , then the expected value of Y may be written as
1 − E e−δTx
1 − e−δTx
E[Y ] = E =
δ δ
Therefore, in actuarial notation the EPV of a whole life annuity payable continuously is
1 − Āx
āx = (4.26)
δ
1 − e−δTx
1 1
= 2 Var e−δTx = 2 2
Āx − (Ax )2
Var[Y ] = Var
δ δ δ
An integral formula may also be derived by considering the expectation of Y as we usually would for a
continuous random variable
Z ∞ Z ∞
E[Y ] = āt fx (t) dt = āt · t px µx+t dt
0 0
We can evaluate this integral using integration by parts and noting that
d 1 − e−δt
d
ā = = e−δt
dt t dt δ
Therefore, ∞ Z ∞
∞
1 − e−δt
Z
d
āx = āt (−t px ) dt = −t px + e−δt t px dt
0 dt δ 0 0
Z ∞
=⇒ āx = e−δt t px dt (4.27)
0
Example: You are given that for a particular survival model µx+t = 0.03 for all x, t > 0. If the force
of interest is δ = 5%, calulate
p
Pr Y ≥ E[Y ] − Var[Y ]
where Y is the present value random variable for a whole life annuity payable continuously.
Solution: We first need to evaluate āx , Āx and 2 Āx so we can obtain values for E[Y ] and Var[Y ].
Z ∞ Z ∞ Z ∞
−δt −δt −µt
Āx = e t px µx+t dt = µe e dt = µ e−(µ+δ)t dt
0 0 0
µ −(µ+δ)t ∞ µ
=− e 0
=
µ+δ µ+δ
0.03
= = 0.375
0.03 + 0.05
The rule of moments may be used to find 2 Āx by evaluating Āx at 2δ.
2 µ 0.03 3
=⇒ Āx = = =
µ + 2δ 0.03 + 0.1 13
1 − Āx 1 − 0.375
āx = = = 12.5
δ 0.05
Therefore
1 2 2 1 3
− 0.3752
E[Y ] = āx = 12.5 and Var[Y ] = 2 Āx − (Āx ) = = 35.05769
δ 0.052 13
If we substitute these values into the probability the problem reduces to calculating
√
Pr(Y ≥ 12.5 − 35.05769) = Pr(Y ≥ 6.49519)
We know that the present value random variable Y is dependent on Tx so we can rearrange the
probability such that it is of the form Pr(Tx ≥ a) for some constant a. This is then just the definition
of the survival function which is already known.
1 − e−δTx
≥ 6.49519 = Pr e−δTx ≤ 1 − 6.49519δ
Pr(Y ≥ 6.49519) = Pr
δ
= Pr(−δTx ≤ ln(1 − 6.49519δ))
−ln(1 − 6.49519δ)
= Pr Tx ≥
δ
= Pr(Tx ≥ 7.8537)
= 7.8537 px = e−(0.03×7.8537) = 0.79
Hence p
Pr Y ≥ E[Y ] − Var[Y ] = 0.79
1 − e−δmin(Tx ,n)
Y = āmin(Tx ,n) =
δ
1 − E e−δmin(Tx ,n)
1 − e−δmin(Tx ,n)
E[Y ] = E =
δ δ
Using our knowledge of continuous time endowment insurances, the EPV is therefore
1 − Āx:n
āx:n = (4.28)
δ
0.05 0 ≤ t ≤ 10
Example: You are given that µx+t = 0.02 for all x, t > 0 and δ =
0.10 t > 10
Calculate ā40 .
Solution: We must first note that since the force of interest changes during the term of the contract
we need to consider the first 10 years separately as a temporary annuity, that is,
Equations (4.27) and (4.29) can then be used to evaluate the annuities as follows, remember also that
for a constant force of mortality µ, the survival function for a life aged x is given by t px = e−µt
Z 10 Z ∞
−δt −10δ
ā40 = e t p40 dt + e 10 p40 e−δt t p50 dt
Z0 10 0
Z ∞
−0.05t −0.02t −0.5 −0.2
= e e dt + e e e−0.1t e−0.02t dt
Z0 10 Z ∞ 0
= 11.3298
Note that the force of interest we use for 10 E40 is the the force of interest for the first 10 years as this
is the period we are discounting through.
Whole life
Consider a whole life annuity due where the first payment is equal to £1 and every year the amount
increases by £1 so that in the second year it is £2, £3 in the third year and so on.
The expected present value of such an annuity is given by the current payments formula:
∞
X
(Iä)x = v k (k + 1)k px . (4.30)
k=0
and
(Iä)40 = 1 + 2vp40 + 3v 2 2 p40 + 4v 3 3 p40 + · · ·
Subtracting the two equations we get:
So
(Ia)40 = (Iä)40 − ä40 = 288.17248 − 18.45776 = 269.71472.
From the above example it is clear that the following relationship holds for any x > 0.
Temporary
Similarly we have
n−1
X
(Iä)x:n = v k (k + 1)k px . (4.32)
k=0
The values of temporary increasing life annuity due are not usually tabulated but one can use the
following property in order to express the temporary into whole life annuities due and whole life
increasing annuity due the values of which are tabulated.
lx+n
(Iä)x:n = (Iä)x − v n [näx+n + (Iä)x+n ]. (4.33)
lx
Note that when considering temporary life annuities we can also consider decreasing ones. That is
something we could not define for whole life annuities due to the risk of having negative payments. The
Temporary decreasing life annuity due is an annuity due that pays £n in the first year decreasing
by £1 every year thereafter. We can calculate its EPV using the following decomposition:
Example: Consider a 10−year life annuity due issued to a select life aged exactly x = [40]. The
first payment is £50, 000 decreasing by £5, 000 per year thereafter. Calculate the EPV of this annuity
assuming mortality follows the tables at the back of the lecture notes and i = 5%.
Solution:
EPV = 55, 000ä[40]:10 − 5, 000(Iä)x:n
l50
= 55, 000(ä[40] − v 10 ä[40] )
l[40]
l50
− 5, 000((Iä)[40] − v 10 [10ä50 + (Iä)50 ].
l[40]
and
EPV = 55, 000(18.45956 − 0.609268783 · 17.02453)
− 5, 000(288.2026 − 0.609268783[10 · 17.02453 + 235.17974]
= 238, 838.6.
We close the discussion of increasing annuities with the note that there are many possibilities to define
different kinds of increasing annuities. Indicatively, we mention the case of continuous annuity that
increases continuously:
Z n
¯
(Iā)x:n = te−δt t px dt
0
or a continuous annuity that the annual amount paid increases by £1 at the end of each year:
n−1
X
(Iā)x:n = (k + 1)k| āx:1 .
k=0
In any case the approach would be to try and write down the EPV using well known EPVs of
annuities.
Chapter 5
An insurance policy is funded by contract premiums, this may be one single premium paid on issue or a
series of payments contingent on the survival of the policyholder. The premium will need to cover any
future benefits paid to the policyholder but also the expenses the insurance company incurs by creating
and managing the policy. There are two types of premium
Net premium: This premium only covers the benefit paid to the individual by the insurance company,
it does not allow for the insurance company’s expenses.
Gross premium: This premium covers both the benefit and the company’s expenses. There are three
main types of expense associated to a contract - initial expenses, renewal expenses and termination
expenses. These may include commission paid to an agent for selling a contract, staff salaries, and the
rent for the insurer’s premises.
Premiums are (usually) paid in advance, with the first premium payable when the policy is purchased.
For example, regular premiums for a policy on a single life cease to be payable on the death of the PH.
Sometimes, in the case of a whole life insurance policy, it would be usual for the death benefit to be
secured by regular premiums, and it would be common for premium payments to cease at a certain age,
perhaps at age 65, when the PH is assume to retire, or at age 80, when the PH’s real income may be
diminishing.
We can classify the types of premiums in the following way:
According to whether expenses are considered:
• Calculation of premium without IC’s expenses ⇒ net premium (risk and/or mathematical
premium).
• Calculation of premium with explicit IC’s expenses ⇒ gross premium (or office premium).
• regular series of payments (annually, monthly, weekly, daily, etc.) ⇒ multiple premiums.
• premium P (income).
Note that the three items above are generally life contingent. If a single premium is paid however, then
there is no uncertainty in P . We can now define the future loss random variable as future outgo - future
income. What is more useful however is its present value.
Definition: PV of future loss random variable is
Depending on whether expenses are included, we classify the PV of future loss random variable as net
or gross and we will denote it accordingly.
Let us assume that for a life insurance policy the cash flow for the company consists only of premium
income and benefit outgo. Since expenses are excluded we use the “net future loss” which is denoted
by LN0 and defined as
LN
0 = PV of future benefit outgo − PV of net premium income
If expenses are also included, we use the the “gross future loss” which is denoted LG
0 = and defined as
LG
0 = PV(benefit/outgo) + PV (expenses/outgo) − PV (gross premium/income)
Example: An insurer issues a whole life insurance contract to a life currently aged 60 with sum
insurerd S payable at the end of the year of death and premiums P payable annually in advance. Find
an expression for the net future loss random variable.
Solution: From chapter 3 we know that the present value random variable for a whole life insurance
benefit S payable at the end of the year of death to a life aged 60 is
Z = Sv K60 +1
Similarly, from chapter 4 the present value random variable for a whole life annuity due with annual
payments P is
Y = P äK60 +1
Therefore, the net future loss random variable is
LN
0 = PV of future benefits − PV of future premiums
= Sv K60 +1 − P äK60 +1
E[L0 ] = 0
Therefore, the equivalence principle calculates the premium by equating the expected outgo and income
of the insurance company. When we exclude expenses the equivalence principle takes the form:
E[LN
0 ] = 0 =⇒ E[PV of future benefits − PV of future premiums] = 0
and therefore
EPV of benefits = EPV of premiums (5.1)
When expenses are taken into account the equivalence principle takes the form:
E[LG
0] = 0 =⇒ E[PV(benefit/outgo) + PV (expenses/outgo) − PV (gross premium/income)] = 0
and therefore
Z = Sv Kx +1 Y = P äKx +1
SAx
=⇒ P = (5.3)
äx
S 2 (2 Ax − (Ax )2 )
= (using equation (4.7))
(däx )2
S 2 (2 Ax − (Ax )2 )
=⇒ Var[LN
0 ] = (5.4)
(1 − Ax )2
Endowment insurance
Consider an n-year endowment insurance policy issued to a life aged x with sum insured S payable at
the end of the year of death and premiums P payable annually in advance throughout the term of the
contract. The future loss random variable is given by
min(Kx +1,n)
LN
0 = Sv − P ämin(Kx +1,n)
SAx:n
=⇒ P = (5.5)
äx:n
The variance of the net future loss random variable is found in a similar way as in the case of the whole
life insurance.
1 − v min(Kx +1,n)
h i
min(Kx +1,n) min(Kx +1,n)
Var[LN
0 ] = Var Sv − P ämin(Kx +1,n) = Var Sv −P
d
2 2
P min(Kx +1,n) P 2
Ax:n − (Ax:n )2
= S+ Var v = S+
d d
2
SAx:n 2
Ax:n − (Ax:n )2
= S+
däx:n
2
2 däx:n + Ax:n 2
Ax:n − (Ax:n )2
=S
däx:n
S 2 (2 Ax:n − (Ax:n )2 )
=
(däx:n )2
S 2 (2 Ax:n − (Ax:n )2 )
=⇒ Var[LN
0 ] = (5.6)
(1 − Ax:n )2
We now give the net future loss random variable and premium formula for a term and pure endowment
insurance contract, they can be derived using a similar approach as for a whole life and endowment
contract.
Term insurance:
1
Sv Kx +1 − P äKx +1
if Kx ≤ n − 1 SAx:n
LN
0 = =⇒ P =
−P än if Kx ≥ n äx:n
E[LN
Tx
0 ] = SE v − P E āTx = S Āx − P āx = 0
S Āx
=⇒ P = (5.7)
āx
The variance of the net future loss random variable is then defined as
1 − v Tx
N
Tx Tx
Var[L0 ] = Var Sv − P āTx = Var Sv − P
δ
2
P
Var v Tx
= S+
δ
2
S Āx 2
Āx − (Āx )2
= S+
δāx
2
2 δāx + Āx 2
Āx − (Āx )2
=S
δāx
S2 2
Āx − (Āx )2
=
(δāx )2
S2 2
Āx − (Āx )2
=⇒ Var[LN
0 ] = 2 (5.8)
1 − Āx
For other fully continuous insurance policies we have
Endowment:
Term insurance:
1
Sv Tx − P āTx
if Tx ≤ n S Āx:n
LN
0 = =⇒ P =
−P ān if Tx > n āx:n
Example: A life aged 45 purchases a 20-year endowment insurance where a benefit of £10,000 is paid
at the end of the year of death should death occur within 20 years and £20,000 at the end of 20 years
if the insured is still alive. Annual level premiums are paid at the beginning of each year for the first
10 years only and mortality follows the Illustrative Life Table at i=6%. Calculate the annual premium
using the equivalence principle.
Solution: We first need to calculate the present value of the benefits and premiums. Recall that an
endowment insurance is the sum of a pure endowment and term insurance, in this case £10,000 is the
benefit for the term insurance and £20,000 is the benefit for the pure endowment insurance.
10000v K45 +1
if Kx ≤ 19
=⇒ PV of benefits =
20000v 20 if Kx ≥ 20
The present value of the premiums is equal to the present value of a 10-year temporary annuity due,
that is,
PV of premiums = P ämin(K45 +1,10)
where P denotes the annual premium. The EPV of the premiums is then given by
EPV(premiums) = P ä45:10
If we now equate the EPV of benefits and premiums as stated by the equivalence principle, we find that
1
1
10000A45:20 + 2000020 E45
10000A45:20 + 2000020 E45 = P ä45:10 =⇒ P =
ä45:10
However, we must now write our temporary annuity and term insurance in terms of whole life annuities
and insurances since these are the only values provided in the Tables.
l65 1 94, 579.73
20 E45 = v 20 = = 0.3599383
l45 1.0520 99, 033.94
l55 1 97, 846.2
10 E45 = v 10 = = 0.6065504
l45 1.0510 99, 033.94
1
A45:20 = A45 − 20 E45 A65 = 0.15161 − (0.3599383 × 0.35477) = 0.0239147
ä45:10 = ä45 − 10 E45 ä55 = 17.81621 − (0.6065504 × 16.05987) = 8.075089
Substituting these values into our expression for the premium gives
• Kevin calculates non-level premiums of £608 for the first year and £350 for the second year
• d = 0.05
Calculate π.
Solution: If we consider the EPV of the premiums for each actuary, for Kevin we have
Since both actuaries calculated the premium on the same policy, by the equivalence principle both of
these expressions should equal the EPV of the benefit, in this case given by
EPV(benefits) = 1000Ax:2
and so we need to evaluate äx:2 using the information we know from Kevin. Using the equivalence
principle and the expression for the EPV of Kevin’s premiums
Hence the annual level premium on the contract is £489.0836. Note that this value seems plausible since
we would expect the level premium to lie between the two non-level premiums calculated by Kevin.
Example: An insurance company issues a 15-year deferred life annuity contract to a life aged 50. You
are given
• level monthly premiums P are paid in advance during the deferred period
• an annuity benefit of £25,000 is paid at the beginning of each year the insured is alive starting
from age 65
• mortality follows the Life Table at the end of the notes with i = 5%
(i) Write down an expression for the net future loss random variable
(ii) Calculate the monthly premium P
(iii) If an additional benefit of £10,000 is to be paid immediately on death during the deferred period,
how much will the monthly premium increase by?
Solution: (i) To find the net future loss random variable we need to consider the present value of the
benefits and premiums. The benefit is a 15-year deferred annuity due so the present value is
0 if K50 ≤ 14
PV(benefits) =
25000v 15 äK50 +1−15 if K50 ≥ 15
The premiums are payable monthly for the first 15 years unless death occurs beforehand which is how
we define a 15-year temporary annuity due. Remember also that since P is the monthly premium, the
amount paid each year is 12P . The present value of the premiums may then be written as
(12)
12P äK (12) +1
if K50 ≤ 14
50
PV(premiums) =
(12)
12P ä15 if K50 ≥ 15
(ii) To determine the premium we use the equivalence principle so we need to find the EPV of the
benefits and premiums. Consider the benefits first, the EPV of a 15-year deferred annuity due is
denoted by 15| ä50 but since this does not appear in the Illustrative Life Table, we must write it in terms
of a whole life annuity due. We calculate:
l65 1 94, 579.73
15 E50 = v 15 = = 0.4915149.
l50 1.0515 98, 576.37
(12) (12)
Since the UDD assumption holds, we may apply equation (4.21) to ä50 and ä65 to give
(12)
ä50 = α(12)ä50 − β(12) = (1.000197 × 17.02453) − 0.466508 = 16.56138
(12)
ä65 = α(12)ä65 − β(12) = (1.000197 × 13.54979) − 0.466508 = 13.08595
(iii) If an additional benefit is payable immediately on death during the deferred period, this only has
an effect on the EPV of the benefits and not the premiums. There are also no changes to the benefit
after the deferred period so the total EPV of the benefits will include just a single additional term.
10000v T50
if K50 ≤ 14
PV(benefits) = 15
25000v äK50 −14 if K50 ≥ 15
10000v T50
0 if K50 ≤ 14 if K50 ≤ 14
= +
25000v 15 äK50 −14 if K50 ≥ 15 0 if K50 ≥ 15
1 i 1 i
Ā50:15 = A50:15 = (A50 − 15 E50 A65 )
δ δ
0.05
= (0.18931 − (0.4915149 × 0.35477))
ln(1.05)
= 0.0262126
If we equate the EPV of the benefits and the premiums and let P ∗ denote the new monthly premium
1 (12)
=⇒ 2500015| ä50 + 10000Ā50:15 = 12P ∗ ä50:15
1
∗
2500015| ä50 + 10000Ā50:15 (25000 × 7.637975) + (10000 × 0.0262126)
⇐⇒ P = (12)
= = 1514.173
12ä50:15 12 × 10.5229
Therefore, the revised monthly premium is P ∗ = £1514.173, an increase of £2 from the premium
without the additional benefit.
Example: For a fully continuous n-year endowment insurance of £1 issued to a life aged x you are
given
• Z is the present value random variable of the benefit for this insurance
• E[Z] = 0.5198
• Var[Z] = 0.1068
• Continuous premium is paid at a rate P per year during the term of the contract.
Calculate Var[LN N
0 ] where L0 is the net future loss random variable.
Solution: For a fully continuous n-year endowment insurance with unit benefit and annual premium
rate P , the net future loss random variable is
min(Tx ,n)
LN
0 = v − P āmin(Tx ,n)
Var[Z] is given in the question so we need to find the value of P and δ, however, we are instead going
to find an expression for Pδ . Since the premium is calculated using the equivalence principle
P
=⇒ P āx:n = Āx:n ⇐⇒ (1 − Āx:n ) = Āx:n
δ
Hence,
2
Var[LN
0 ] = (1.0825 + 1) × 0.1068 = 0.4632
Every life insurance company needs a balance sheet showing the value of its assets, A, and its liabilities
L. If A ≥ L ⇒ company is solvent whereas if A < L ⇒ company is insolvent.
The insurer has to quantify its assets and liabilities. This is not always straightforward.
• (i) assets: are easy to quantify (market value). Investments (such as bonds, equities and property).
They have been purchased with premiums and they will be held in a fund from which benefits will
be paid out. (An insurer will also be concerned about liquidity risk, i.e. how easily and non-costly
investments are turned to cash.)
• (ii) liabilities: are difficult to quantify. Promises made to pay benefits in the future, against which
can be set the policyholder’s promises to pay premiums in the future.
E v Kx +1 − Px äKx +1 = 0.
Kx +1
Remember that LN
0 = v − Px äKx +1 = future loss at the initial time, t = 0.
At any time t > 0 (time in the future), and if the policy is still in place, let us introduce and define
LN
t ,
– 1) if policy alteration,
– 2) demonstrating solvency,
– 3) share holder ⇒ determine dividends to be paid,
– 4) if with-profit policies ⇒ determine bonus levels,
– 5) if policy terms allow the payment of a surrender benefit, needed to pay surrender values
to PH who surrenders (⇒ pay surrender values to PH).
Back to Example 1:
E LN N
t=0 = 0 V = E[L0 ] = Ax − P äx .
Ax
⇒P = äx
and 0 V = 0.
• (vi) boundary conditions for policy values at outset and at expiry of the policy:
The prospective policy value (or prospective reserve) for a life insurance contract, which is in force (not
yet expired through claim or reached the end of the term) is defined, for a given basis.
Definition: expected PV (future outgo) - expected PV (future income).
We observe that it is the minimum funds the IC needs to hold at any time during the term of the
contract.
E (PV future outgo)t - E (PV future income)t , where t > 0, t = 0 at the start of the contract.
def
LNt = B v X − P äX ,
X = min(Kx+t + 1, n − t) ,
E[LN
t ] = t Vx:n = B Ax+t:n−t − P äx+t:n−t .
We know that:
Px:n Px:n B Px:n
LN X X
vX −
t = B v −B 1−v =B 1+ ,
d d d
1 − vX
äX = .
d
2
2 Px:n
V [LN
t ] = B 1+ [2 Ax+t:n−t − (Ax+t:n−t )2 ] .
d
G
tV = EPV(future benefits)t + EPV (future expenses/outgo)t − EPV(future premiums)t
Example: Consider a 20−year endowment issued at a life aged exactly 40 years old. The benefit is
£100, 000 payable at the end of year of death (or maturity).The policy incurs initial expenses of £250,
renewal expenses equal to 10% of the annual premium and £100 termination (at maturity or at benefit
payment) expenses. Calculate the gross prospective reserve at t = 5.
We first need to calculate the gross premium using the equivalence principle.
Appendix