Introduction to

survival models

Overview
A survival model is a probabilistic model of a random variable that represents the time until the
occurrence of an unpredictable event. For example, we may wish to study the life expectancy of
a newborn baby, or the future working lifetime of a machine until it fails. In both cases, we study
how long the subject may be expected to survive.
The theory that we will develop throughout this course can be applied in a wide range of
situations, in which the concept of “survival” may not be immediately obvious, for example:
• the time until a claim is made on an automobile insurance policy
• the time until a patient in a coma recovers from the coma, given that he recovers
• the time until a worker leaves employment.
The focus of our study—the time until the specified event—is known as a waiting time or a
random time-to-event variable. Probabilities associated with these models play a central role in
actuarial calculations such as pricing insurance contracts.

1
 Introduction to survival models Chapter 1

1.1 The role of a survival model in a contingent payment model

Let’s start by considering the most basic contingent payment model, in which a specified
amount is paid if and only if a particular event occurs.
Suppose that an amount P is to be paid in n years if a random event E occurs. Otherwise, if the
complementary event occurs, then nothing is to be paid.
At an effective annual rate of interest i , the random present value of the payment is:

 P ⋅ vn if E occurs
Z = 
 0 if E′ occurs

where v = (1 + i )−1 is the one-year present value discount factor.

The random present value of the payment, Z , is a discrete random variable. Its expected value is
known as the actuarial present value of the payment, which incorporates the amount of the
payment, the discount factor associated with the timing of the payment, and the probability of
the payment being made:

E [ Z ] = P ⋅ vn ⋅ Pr ( E ) + 0 ⋅ Pr ( E′ )
= P ⋅ vn ⋅ Pr ( E )
amount discount probability

Throughout Chapters 2 through 8 of this course, we will be concerned with the distribution of
random present value of payment variables. We will frequently calculate the mean, the variance,
a percentile, or the probability of some event regarding Z such as Pr(Z > E [ Z ]) .

In Chapters 2 and 3 we will introduce contingent payment models that arise in the context of life
insurance and life annuities. For these models, we will need to compute probabilities of events
that are expressed in terms of the random future lifetime after age x , when an insurance contract
has just been issued. For example, in order to calculate the appropriate life insurance premium
for a 30 year old policyholder, we need to calculate the probability that the policyholder will die
before age 31, or 32, or 33, and so on.
In this first chapter, our aim is to familiarize you with the theory of survival models, the notation
employed in these models, and standard terminology.
There are three principal variables, all of which are measured in years:
• the random lifetime (ie time until death) of a newborn life is denoted X
• the random future lifetime at age x , given that a newborn has survived to age x , is
denoted T ( x ) , ie T ( x ) = X − x X > x

• the curtate future lifetime at age x , given that a newborn has survived to age x , is the
complete number of years of future lifetime at age x and is denoted K ( x ) , ie
K ( x ) = T ( x )  (greatest integer)

The variables X and T are assumed to be continuous random variables, whereas K is obviously
discrete. For example, suppose a newborn life eventually dies at age 74.72. Then X = 74.72 ,
T ( 30 ) = 74.72 − 30 = 44.72 , and K ( 30 ) = [T ( 30 )] = [44.72] = 44 .

Notice that T is a function of X , and K is a function of T . So, the distributions of these three
variables are closely related. These relationships will be developed over the rest of this chapter.

2
 Chapter 1 Introduction to survival models

1.2 The life table – a discrete survival model

We begin by studying the life table, a discrete survival model commonly used in insurance
applications. This model gives us the opportunity to gain an intuitive understanding of some of
the most fundamental concepts before we study continuous time models.
We start by defining lx as the number of lives expected to survive to age x from a group of l0
newborn lives. A life table displays in a table format the values of lx at ages x equal to
0, 1, 2,… , ω , where ω is the first whole number age at which there are no remaining lives in the
group. If we are modeling human mortality, we may choose a value of ω of around 120 years.
We’ll start by taking a deterministic view of future mortality. By this, we mean that the table
tells us exactly how many of the l0 lives will be surviving at ages 1, 2, and so on.

Here is a portion of a hypothetical life table:

x 0 1 2 3 4 5 6 7 8 9
lx 1,000 991 985 982 979 976 972 968 964 959

dx 9 6 3 3 3 4 4 4 5 6

The row labeled dx represents the number of lives among l0 newborn lives that die in the age
range [ x , x + 1) . It is computed as:

dx = lx − lx + 1

For example, since l2 = 985 lives survive to age 2, and l3 = 982 lives survive to age 3, then exactly
d2 = l2 − l3 = 3 lives must die between age 2 and age 3.

A number of probabilities can be computed from the entries in such a table. Let’s begin by
introducing the standard notation for the most significant types of probabilities that you will see
in the actuarial models throughout Chapters 2 – 8.
The probability that a life currently age x will survive n years is denoted n px . With our
deterministic interpretation of the life table, along with the point of view that the probability of
an event is the relative frequency with which it occurs, we have:
lx+ n
n px =
lx

It is a standard convention to omit the n subscript when n = 1 , so the probability that a life
currently age x will survive 1 year is:
lx + 1
px =
lx

For example, the probability that a life age 5 survives for 2 years to age 7 is:
l7 968
2 p5 = =
l5 976

3
 Introduction to survival models Chapter 1

The probability that a life currently age x will die within n years is denoted n qx , and we
have:
lx+ n lx − lx + n
n qx = 1 − n px = 1 − =
lx lx

Intuitively, this is the probability that a life is one of the (lx − lx + n ) lives to die between age x and
age x + n , out of the lx lives age x .

For example, the probability that a life age 1 dies within 3 years is:
l −l 991 − 979 12
3 q1 = 1 4 = =
l1 991 991

Again, we omit the n subscript when n = 1 , so the probability that a life currently age x will
die within 1 year is:
lx − lx + 1 dx
qx = or qx =
lx lx

Finally, the probability that a life currently age x will survive for m years and then die within
the following n years is denoted m n qx , and we have:

lx + m − lx + m+ n
m n qx =
lx

Intuitively, m n qx is the probability that a life age x survives for m years, multiplied by the
probability that a life age x + m dies within n years:
l l −l l −l
m n qx = m px × n q x + m = x + m × x + m x + m + n = x + m x + m + n
lx lx + m lx

Again, we omit the n subscript when n = 1 , so the probability that a life currently age x will
survive for m years and then die within 1 year is:
lx + m − lx + m+ 1 d
m qx = or m qx = x +m
lx lx

For example, the probability that a life age 4 survives for 3 years and then dies within the
following 2 years is:
l −l 968 − 959 9
3 2 q4 = 7 9 = =
l4 979 979

Example 1.1

Compute the following probabilities from the life table above:

(a) 5 p0

(b) 5 q0

(c) 4 2 q1

(d) p1

(e) q2

4
 Chapter 1 Introduction to survival models

Solution

l 976
(a) 5 p0 = 5 =
l0 1, 000

d 4
(b) 5 q0 = 5 =
l0 1, 000

l5 − l7 976 − 968 8
(c) 4 2 q1 = = =
l1 991 991

l 985
(d) p1 = 2 =
l1 991

d 3
(e) q2 = 2 = ♦♦
l2 985

We can also take a stochastic (ie random) view of future mortality. Under a stochastic approach,
the number of survivors at age x is a random variable L ( x ) , and the life table function lx
represents the expected number of survivors, ie:

lx = E L ( x ) 

The random variable L ( x ) follows a binomial distribution. Each life is viewed as an

independent Bernoulli trial. The number of trials is n = l0 . We define “success” as survival to age
x , with probability p = x p0 .

For example, let’s consider the distribution of L ( 5 ) , the random number of survivors at age 5 in
this life table from the 1,000 newborn lives.
The random variable L ( 5 ) follows a binomial distribution with n = l0 = 1, 000 trials and a
probability of success of p = 5 p0 = 976 /1, 000 = 0.976 .

The mean and variance of L ( 5 ) are:

E  L ( 5 )  = np = l0 × 5 p0 = 1, 000 × 0.976 = 976 ( = l5 )

var ( L ( 5 ) ) = npq = n × 5 p0 × 5 q0 = 1, 000 × 0.976 × 0.024 = 23.424

Finally, we can deduce little from the life table about the continuous random lifetime X , but we
can identify the distribution of the curtate lifetime of a newborn, K (0) .

The probability Pr(K (0) = k ) is the probability that a newborn life dies in the age range [ k , k + 1) ,
which we have already defined as k q0 . Hence:

d
Pr ( K (0) = k ) = k q0 = k
l0

For example, the probability that a newborn life has a curtate lifetime of 3 years is:
d 3
Pr ( K (0) = 3 ) = 3 q0 = 3 = = 0.003
l0 1, 000

5
 Introduction to survival models Chapter 1

1.3 The theory of continuous survival models

In this section we will study five different mathematical functions that can all be employed to
specify the distribution of X , the random lifetime (ie age at death) of a newborn life:
• the cumulative distribution function of X
• the probability density function of X
• the survival function
• the life table function
• the force of mortality.
We will focus on the relations between these functions as well as their meaning.

The pdf and cdf of the random lifetime

The random lifetime (ie age at death) of a newborn life, X , is assumed to be a continuous
random variable. We will review the basic properties of continuous random variables and
explain their interpretation in the context of the random lifetime.
Let’s begin with the cumulative distribution function (cdf):
FX ( x ) = Pr( X ≤ x )

The cdf FX ( x ) represents the probability that a newborn life will die at or before age x .

FX ( x ) is continuous and non-decreasing with FX ( 0 ) = 0 and FX (ω ) = 1 where ω is the first age

at which death is certain to have occurred for a newborn life.

The probability density function (pdf) is:

f X ( x ) = FX′ ( x ) wherever the derivative exists

The pdf f X ( x ) is non-negative and continuous on the interval [0 , ω ) .

Recall that a value of f X ( x ) is not a probability in itself. The probability that a newborn life dies
between ages a and b is:
b
Pr ( a ≤ X ≤ b ) = ∫a f X ( x ) dx = F(b ) − F( a )

You should note that since X is assumed to be a continuous random variable, all of the intervals
 a , b  , [ a , b ) , ( a , b ] , and ( a , b ) have the same probability, F ( b ) − F ( a ) .

Hence:
ω
∫0 f X ( x ) dx = 1

and:
x
FX ( x ) = Pr ( X ≤ x ) = ∫0 f X ( u ) du

Finally, the probability that a newborn life dies in the interval [ x , x + ∆x ] can be estimated as:

Pr ( x ≤ X ≤ x + ∆x ) ≈ f X ( x ) ⋅ ∆x

6
 Chapter 1 Introduction to survival models

Example 1.2

Suppose that the lifetime X of a newborn life is uniformly distributed on the interval  0 , 100  .

(a) Identify the probability density function.

(b) Identify the cumulative distribution function.
(c) Calculate the probability of death occurring between ages 60 and 80.

Solution

(a) The pdf for a uniform distribution is constant and equal to the reciprocal length of the interval:
1
fX ( x ) = = 0.01 for 0 ≤ x ≤ 100
100
(b) The cdf is:
x x
FX ( x ) = ∫0 f X ( u ) du = ∫ 0 0.01 du = 0.01 x for 0 ≤ x ≤ 100

(c) The probability of death between ages 60 and 80 is:

Pr ( 60 ≤ X ≤ 80 ) = FX ( 80 ) − FX ( 60 ) = 0.01(80 − 60) = 0.20

Note that the uniform distribution is not particularly well suited as a model of human mortality,
but it is useful as a simple context to illustrate the theory. This mortality model is commonly
known as de Moivre’s law. It was actually the first mortality model to be used in insurance
practice. ♦♦

Example 1.3

Suppose that the lifetime X of a newborn life is exponentially distributed with mean 75 years.
(a) Identify the probability density function.
(b) Identify the cumulative distribution function.
(c) Calculate the probability of death between ages 60 and 80.

Solution
1
The pdf for an exponential distribution with mean θ is f X ( x ) = e − x /θ for x > 0 . Hence:
θ
(a) The pdf is:
1 − x /75
fX ( x ) = e for x > 0
75
(b) The cdf is:

( )0
x 1 x
−u /75
FX ( x ) = ∫ 0 75 e du = − e − u /75 = 1 − e − x /75

(c) The probability of death between ages 60 and 80 is:

Pr ( 60 ≤ X ≤ 80 ) = FX ( 80 ) − FX ( 60 ) = e −60 /75 − e −80 /75 = 0.10518 ♦♦

7
 Introduction to survival models Chapter 1

The survival function

In actuarial mathematics it is common to describe a survival model by giving the survival

function rather that the density function or distribution function. The survival function is
denoted sX ( x ) and is defined as:

sX ( x ) = Pr ( X > x )

The survival function gives the probability that a newborn life dies after age x . This is the same
as saying that the newborn survives to age x , or is alive at age x .
From the preceding discussion of the lifetime variable X , we can deduce the following
properties of the survival function.

Key properties of the survival function

1. sX ( x ) is continuous and non-increasing with sX ( 0 ) = 1 and sX (ω ) = 0

2. sX ( x ) = 1 − FX ( x )

b
3. Pr ( a ≤ X ≤ b ) = ∫ a f X ( x ) dx = sX ( a ) − sX ( b )
4. f X ( x ) = − sX
′ (x)

Example 1.4

Suppose that the lifetime X of a newborn is exponentially distributed with mean 75 years.
(a) Identify the survival function sX ( x ) .

(b) Calculate the probability that a newborn is still alive at age 100.
(c) Calculate the probability that a newborn dies between ages 60 and 75.

Solution

(a) In Example 1.3 we saw that the cdf is FX ( x ) = 1 − e − x /75 . So, we have:

sX ( x ) = 1 − FX ( x ) = e − x /75 for x > 0

(b) The probability a newborn is still alive at age 100 is:

sX ( 100 ) = e −100 /75 = 0.26360

(c) The probability a newborn dies between ages 60 and 75 is:

sX ( 60 ) − sX (75) = e −60 /75 − e −75 /75 = 0.08145 ♦♦

8
 Chapter 1 Introduction to survival models

The life table function

In contrast with the discussion of the discrete case in Section 1.2, here we will define the life table
function lx for all ages between 0 and w .

As before, let L ( x ) denote the random number of survivors at any age x from a group of l0
newborn lives. The random variable L ( x ) follows a binomial distribution with n = l0 trials. We
define “success” as survival to age x , with probability p = Pr ( X > x ) = sX ( x ) .

The life table function lx is defined as the expected number of survivors at age x . Hence:

lx = E  L ( x )  = np = l0 sX ( x )

Example 1.5

Suppose that the lifetime X of a newborn is uniformly distributed on  0 , 100  .

(a) Identify the survival function sX ( x ) .

(b) Identify the life table function lx if l0 = 100 .

Solution

(a) The survival function is given by:

100 100
sX ( x ) = Pr ( X > x ) = ∫x f X ( u ) du = ∫x 0.01 du

= 0.01 ( 100 − x ) = 1 − 0.01x for 0 ≤ x ≤ 100

(b) The life table function is:

lx = l0 sX ( x ) = 100(1 − 0.01x ) = 100 − x for 0 ≤ x ≤ 100 ♦♦

Let’s summarize some of the key properties of the life table function, lx .

Key properties of the life table function

1. lx is the expected number of survivors at age x from a group of l0 newborn lives

2. lx = l0 sX ( x ) is continuous and non-increasing with lω = 0

lx
3. sX ( x ) =
l0

Note that the value of l0 (sometimes called the radix of the life table) is not important to the
survival model, since (by property 3) the survival function is independent of this quantity. So,
we can choose the value of l0 for convenience. The survival function will be identical whether
we choose l0 = 100 or l0 = 1, 000, 000 .

9
Introduction to survival models Chapter 1

Example 1.6
2
Suppose that the life table function is lx = 10, 000 ( 100 − x ) for 0 ≤ x ≤ 100 . Identify the
cumulative distribution function and the probability density function for the associated lifetime
variable X .

Solution

It is elementary to compute the survival function from the life table by property 3:
2
l 10, 000 ( 100 − x ) ( 100 − x )2
sX ( x ) = x = = for 0 ≤ x ≤ 100
l0 10, 000 × 100 2 100 2

We can then calculate FX ( x ) using the relationship FX ( x ) = 1 − sX ( x ) :

( 100 − x )2
FX ( x ) = 1 − for 0 ≤ x ≤ 100
100 2

Finally, we can calculate f X ( x ) using the relationship f X ( x ) = FX′ ( x ) :

 ( 100 − x )2 ′ 2 ( 100 − x ) 100 − x

f X ( x ) = FX′ ( x ) =  1 −  = = for 0 ≤ x ≤ 100 ♦♦
 100 2  100 2 5, 000
 

Example 1.7

Suppose that there are 1,000 newborn lives whose lifetime follows the survival model given in
Example 1.6. Determine the interval that lies within two standard deviations either side of the
mean for L ( 10 ) , the random number of survivors at age 10.

Solution

L ( 10 ) follows a binomial distribution with:

n = l0 = 1, 000

( 100 − 10 )2
p = sX ( 10 ) = = 0.81
100 2
So we have:

E  L ( 10 )  = np = 1, 000 × 0.81 = 810

var ( L ( 10 ) ) = npq = 1, 000 × 0.81 × (1 − 0.81) = 153.90
σ L( 10 ) = 153.90 = 12.406

Hence the required interval is:

 810 − 2 × 12.406, 810 + 2 × 12.406  = 785.19, 834.81 ♦♦

10
Chapter 1 Introduction to survival models

The force of mortality

We can also specify a survival model in terms of the force of mortality.

The force of the mortality is denoted µ ( x ) . It is an instantaneous measure of mortality at age x ,
and it can be defined in several equivalent ways:
fX ( x ) ′ (x)
sX l′
µ (x) =
sX ( x )
=−
sX ( x )
(
= − ln ( sX ( x ) ) ′ = − x
lx
)
These equalities can be verified using simple calculus.
For example, using the information in Example 1.6, we have:

′ (x)
sX −2 ( 100 − x ) /100 2 2
µ (x) = − =− = for 0 ≤ x < 100
sX ( x ) ( 100 − x ) 2
/100 2 100 − x

Or, using the information in Example 1.4, we have:

 x ′ 1
( )
µ ( x ) = − ln ( sX ( x ) ) ′ = −  −  =
 75  75
for x ≥ 0

Let’s now see how to calculate the survival function from the force of mortality.

(
µ ( x ) = − ln ( sX ( x ) ) ′)
x
∫ 0 µ ( y ) dy = − ln ( sX ( y ) ) 0 = − ln ( sX ( x ) ) + ln ( sX ( 0 ) )
x
⇒

But since sX ( 0 ) = 1 and ln ( 1 ) = 0 , we have:

x
∫ 0 µ ( y ) dy = − ln ( sX ( x ) )
 x 
⇒ sX ( x ) = exp  − ∫ µ ( y ) dy 
 0 

Example 1.8

Suppose that the force of mortality for a survival model is given by the formula:
0.9
µ (x) = for 0 ≤ x < 90
90 − x
Calculate the survival function.

Solution

The survival function is calculated as:

 x   x 0.9 
sX ( x ) = exp  − ∫ µ ( y ) dy  = exp  − ∫ dy 
 0   0 90 − y 
  90 − x  
= exp  0.9 ln ( 90 − y )  = exp  0.9 ln 
x

 0   90  
0.9
 90 − x 
=  for 0 ≤ x < 90 ♦♦
 90 

11
Introduction to survival models Chapter 1

Since we have already studied simple relationships between the survival function, the life table
function, and the pdf and cdf of the lifetime function, we can easily calculate any of these
functions from the force of mortality.
For example, using the information in Example 1.8, we can calculate the life table function (with
l0 = 1, 000 ) as:
0.9
 90 − x 
lx = l0 sX ( x ) = 1, 000   for 0 ≤ x < 90
 90 
It is clear from the definition that the force of mortality is not a probability, so how should it be
interpreted? In order to understand the meaning of µ ( x ) , it is useful to rewrite the defining
formula in the form:
f X ( x ) = sX ( x ) µ ( x )

Now, recall that:

f X ( x ) ∆x ≈ Pr ( x ≤ X ≤ x + ∆x )

Rewriting the probability term using a conditional probability, we have:

f X ( x ) ∆x ≈ Pr ( x ≤ X ≤ x + ∆x )

( )
= Pr X ≤ x + ∆x X ≥ x Pr ( X ≥ x )
= Pr ( X ≤ x + ∆x X ≥ x ) sX ( x )

Substituting f X ( x ) = sX ( x ) µ ( x ) , we have:

(
sX ( x ) µ ( x ) ∆x ≈ Pr X ≤ x + ∆x X ≥ x sX ( x ) )
(
⇒ µ ( x ) ∆x ≈ Pr X ≤ x + ∆x X ≥ x )
So, µ ( x ) ∆x is approximately equal to the conditional probability that a newborn that has
survived to age x subsequently dies during the next ∆x years. For example, µ ( 20 ) multiplied
by ∆x = 1/365 (a day), is approximately equal to the conditional probability that a newborn that
has survived to exact age 20 will then die during the next day.

Example 1.9

Suppose that the force of mortality for a survival model is given by the formula:
0.9
µ (x) = for 0 ≤ x < 90
90 − x
Calculate the approximate probability that a life age 40 dies within the next week.

Solution

Setting ∆x = 7 /365 , the required probability is:

0.9 7
µ ( 40 ) ∆x = × = 0.00035 ♦♦
90 − 40 365

12
Chapter 1 Introduction to survival models

Let’s summarize the main properties of the force of mortality.

Key properties of the force of mortality

fX ( x ) ′ (x)
sX l′
1. µ (x) =
sX ( x )
=−
sX ( x )
( )
= − ln ( sX ( x ) ) ′ = − x
lx

 x 
2. sX ( x ) = exp  − ∫ µ ( y ) dy 
 0 
3. µ ( x ) ⋅ ∆x ≈ Pr ( X ≤ x + ∆x | X ≥ x )

4. µ ( x ) is non-negative and piece-wise continuous where defined

ω
5. ∫ 0 µ ( y ) dy = ∞ in order that sX (ω ) = 0

Standard probabilities in a continuous survival model

Let’s now reconsider the ideas we met in Section 1.2 (in the context of a discrete life table) in the
form of a continuous algebraic function defined for all x in  0 , ω  .

First, the probability that a life currently age x will survive t years is:

lx +t sX ( x + t ) Pr ( X > x + t )
t px = = = = Pr ( X > x + t| X > x )
lx sX ( x ) Pr ( X > x )

The probability that a life currently age x will die within the next t years is:

lx − lx +t Pr ( X > x ) − Pr ( X > x + t )
t qx = = = Pr ( X ≤ x + t| X > x )
lx Pr ( X > x )

Finally, the probability that a life currently age x will survive s years but die within the
following t years is:
lx + s − lx + s +t
s t qx = = Pr ( x + s < X ≤ x + s + t| X > x )
lx

These three functions are defined for all ages x in  0 , ω  and for 0 ≤ t ≤ ω − x .

It should be emphasized now that all of these probabilities are conditional, ie we are given that a
newborn has survived to age x.
The symbol ( x ) is commonly used to denote a newborn life that has survived to age x . So, 5 px
is the probability that ( x ) will still be alive in 5 years’ time, at age x + 5 , and 5 qx is the
probability that ( x ) will die within the next 5 years.

As in Section 1.2, the general convention is to drop the subscript t from the symbol when t = 1 .
So, for example, 3|qx is the probability that ( x ) will die between ages x + 3 and x + 4 .

13
 Introduction to survival models Chapter 1

Example 1.10

Suppose that the force of mortality for a survival model is given by the formula:
0.9
µ (x) = for 0 ≤ x < 90
90 − x
Compute the following probabilities:
(a) 2.5 p20

(b) 2.5 q20

(c) 2.5 |q20

Solution

Note that this is the force of mortality in Example 1.8. The life table function is:
0.9
 90 − x 
lx = l0 sX ( x ) = l0   for 0 ≤ x < 90
 90 
Recall that we can choose any convenient value of l0 without changing the distribution of X . So
let’s simplify our computations by choosing l0 = 900.9 , which gives:

lx = l0 sX ( x ) = (90 − x )0.9 for 0 ≤ x < 90

We can now calculate the required probabilities as follows.

0.9
l22.5 ( 90 − 22.5 )
(a) 2.5 p20 = = = 0.96780
l20 ( 90 − 20 )0.9
l20 − l22.5
(b) 2.5 q 20 = = 1 − 2.5 p20 = 0.03220
l20

l22.5 − l23.5 67.50.9 − 66.50.9

(c) 2.5|q20 = = = 0.01291 ♦♦
l20 700.9

The challenge in dealing with these standard probabilities is that there are so many relationships
that involve them. The key relations are listed below without proof. Most of the proofs rely on
simple probability theory – you may like to attempt them to improve your understanding.

Key relations concerning standard probabilities

1. t px + t qx = 1

2. s + t px = s px ⋅ t px + s

3. s|t qx = s px ⋅ t qx + s = s px − s +t px = s +t qx − s qx

4. n px = px ⋅ px + 1 ⋅ ⋅ px + n−1 when n is an integer

5. n qx = 0|qx + 1|qx + + n−1|qx when n is an integer

14
 Chapter 1 Introduction to survival models

For example, if the probability that ( 20 ) survives for 10 years is 0.97, and if the probability that
( 30 ) survives for 10 years is 0.95, then the probability that ( 20 ) is still alive at age 40 is:
20 p20 = 10 p20 ⋅ 10 p30 = 0.97 × 0.95 = 0.92150

Or, the probability that ( 20 ) dies between ages 30 and 40 is:

10|10 q 20 = 10 p20 ⋅ 10 q30 = 0.97 ( 1 − 0.95 ) = 0.04850

On the other hand, if p0 = 0.99 , p1 = 0.98 , and p2 = 0.97 , then the probability that a newborn dies
within three years is:

3 q0 = 1 − 3 p0 = 1 − p0 ⋅ p1 ⋅ p2 = 1 − 0.99 × 0.98 × 0.97 = 0.05891

Or the probability that a newborn dies during the second year of life is:

1|q0 = p0 ⋅ q1 = 0.99 × ( 1 − 0.98 ) = 0.01980

Relations 4 and 5 are useful in constructing a discrete life table for human lives. A statistical
study conducted over a time span of several years could be used to produce estimates of the
mortality rates q0 , q1 , q2 , … and so on. Values of lx at whole number ages can then be produced
as follows:
ln
n p0 = ⇒ ln = l0 ⋅ n p0 = l0 ⋅ p0 ⋅ p1 ⋅ ⋅ pn −1 = l0 ( 1 − q0 )( 1 − q1 ) ( 1 − qn − 1 )
l0

1.4 The continuous future lifetime after age x

Let the continuous random variable X again denote the random lifetime of a newborn. Now
suppose that we are given that a newborn has survived to age x , that is, X > x . The future time
lived after age x is X − x .
The conditional distribution of the time lived after age x , given survival to age x, is:
T (x) = X − x|X > x

The continuous random variable T ( x ) is a survival model defined on the interval [0 , ω − x ] . As

such, it can also be specified in the same ways that we specified the survival model for a newborn
life. It should be clear that the distribution of T ( x ) is closely related to the distribution of X .

The quickest way to see the relation between the distributions of T ( x ) and X is to calculate the
survival function for T ( x ) , sT ( x ) ( t ) = sT ( t ) = Pr (T ( x ) > t ) . In fact, we have already computed
this survival function in terms of the distribution of X , since the event T ( x ) > t is equivalent to
saying that ( x ) is alive at age x + t . The probability of this event is simply t px . So, we have:

l s (x + t)
sT ( x ) ( t ) = Pr (T ( x ) > t ) = t px = x +t = X since lx = l0 sX ( x )
lx sX ( x )

Note: When there is no ambiguity, we will write T for T ( x ) . However, subscripts are often
important, for example to distinguish sX ( 20 ) , the probability that a newborn survives to age 20,
from sT ( 10 ) ( 20 ) , the probability that ( 10 ) survives to age 30.

15
 Introduction to survival models Chapter 1

Example 1.11
2
Suppose that the life table function is lx = 10, 000 ( 100 − x ) for 0 ≤ x ≤ 100 .

(a) Compute the survival function for newborn lives.

(b) Compute the survival function for lives currently aged 20.

Solution

(a) The survival function for newborn lives, sX ( x ) , is:

2
l 10, 000 ( 100 − x )  100 − x 
2
sX ( x ) = x p0 = x = =  100  for 0 ≤ x ≤ 100
l0 10, 000( 100 − 0)2  

(b) The survival function for lives currently aged 20, sT (20) (t ) , is:

l20 +t 10, 000(100 − (20 + t ))2  80 − t 2

sT ( 20 ) ( t ) = t p20 = = =  for 0 ≤ t ≤ 80 ♦♦
l20 10, 000(100 − 20)2  80 

When we deal with the future lifetime after age x , we’ll frequently see expressions of the type
lx + t and µ ( x + t ) . In these expressions the value of x is fixed, and the value of t is allowed to
vary so that we can view the expressions as being functions of t .

For example, let’s see how to relate the pdf and cdf for the distributions of X and T .

FT ( t ) = Pr (T ( x ) ≤ t ) = Pr ( X − x ≤ t | X > x )

Pr ( x < X ≤ x + t ) F ( x + t ) − FX ( x )
= = X
Pr ( X > x ) 1 − FX ( x )

d  F ( x + t ) − FX ( x )  f X ( x + t ) f X ( x + t )
fT ( t ) = ( FT ( t ) )′ =  X  = =
dt  1 − FX ( x )  1 − FX ( x ) sX ( x )

In the figure below, we have areas A = FX ( x + t ) − FX ( x ) and A + B = sX ( x ) .

fX(x)

Area Area
=A =B

x x+t ω

Notice that we have:

FX ( x + t ) − FX ( x ) A
FT ( t ) = =
sX ( x ) A+B

If we view the age x as being fixed, then as the value of t increases, the value of A increases, the
value of B decreases, while the sum A + B remains constant.

16
 Chapter 1 Introduction to survival models

Furthermore, if we examine the relation:

f (x + t)
fT ( t ) = X for 0 ≤ t ≤ ω − x
sX ( x )

in light of the figure above, we can see how the graph of the pdf of T is related to the graph of
the pdf of X .
If we take the portion of the graph of f X ( x ) to the right of x , divide by the total area under the
remainder of the graph, A + B = sX ( x ) , and then relabel the horizontal axis as t and the vertical
axis as fT ( t ) , we have a graph of the pdf of T :

fT(t)

Area = Area =
A / sX(x) B / sX(x)

0 t ω -x

One final point worth noting is the similarity between the pdf’s for the distributions of X and T .
For X we have:
fX ( x )
µ (x) = ⇒ f X ( x ) = sX ( x ) µ ( x ) = x p0 µ ( x )
sX ( x )

and for T we have:

fX ( x + t ) p µ ( x + t ) x p0 t px µ ( x + t )
fT ( t ) = = x +t 0 = = t px µ ( x + t )
sX ( x ) x p0 x p0

Example 1.12
2
Suppose that the life table function is lx = 10, 000 ( 100 − x ) for 0 ≤ x ≤ 100 .

(a) Compute the distribution function for the future lifetime of a life aged 20.
(b) Compute the density function for the future lifetime of a life aged 20.

Solution

(a) In Example 1.11 we computed:

2
 80 − t 
t p20 = sT ( 20 ) ( t ) =   for 0 ≤ t ≤ 80
 80 

Hence, the distribution function for T ( 20 ) is:

2
 80 − t 
FT ( 20 ) ( t ) = 1 − sT ( 20 ) ( t ) = 1 −   for 0 ≤ t ≤ 80
 80 

17
 Introduction to survival models Chapter 1

(b) We have two options for calculating the density function.

The first option is to differentiate the distribution function:

 2 ′ ( 80 − t )
(
fT ( 20 ) ( t ) = FT ( 20 ) ( t ) )′ =  1 −  8080− t   =2
 80 2
 
80 − t
= for 0 ≤ t ≤ 80
3, 200
The other option is to use the relation:
fT ( t ) = t px µ ( x + t )

It is consistent with the formula for lx that:

2
µ (x) =
100 − x
Hence, we have:
2
 80 − t  2 80 − t
fT (20) ( t ) = t p20 µ ( 20 + t ) =   ⋅ = ♦♦
 80  80 − t 3, 200

Key results concerning the relation of the distributions of X and T(x)

l s (x + t)
1. sT ( t ) = Pr (T ( x ) > t ) = t px = x +t = X
lx sX ( x )

FX ( x + t ) − FX ( x )
2. FT ( t ) = t qx = 1 − sT ( t ) =
sX ( x )

fX ( x + t )
3. fT ( t ) = = t px µ ( x+ t )
sX ( x )

1.5 The curtate future lifetime after age x

In addition to computing the distribution of the continuous future lifetime T ( x ) , we may also
wish to derive the distribution of the curtate future lifetime after age x .
The curtate lifetime is a discrete random variable that is defined by:

K ( x ) = T ( x )  ie the integer part (or greatest integer) of T ( x )

Since it is a function of T ( x ) , it is simple to calculate the probability function of K ( x ) from what

we know about T ( x ) . The possible values of K ( x ) are the numbers 0, 1, 2, ,ω − x − 1 .

For example, if x = 70 and ω = 90 , then the possible values of K ( 70 ) are the twenty whole
numbers 0 through 19. If the life ( 70 ) eventually dies at age 85.8, then the continuous future
lifetime is T (70) = 15.8 and the curtate lifetime is K (70) = [ 15.8] = 15 .

18
Chapter 1 Introduction to survival models

The key observation is that if K ( x ) = k , then we must have:

k ≤ T (x ) < k + 1

This leads to the following formula for the probability function:

Pr ( K ( x ) = k ) = Pr ( k ≤ T ( x ) < k + 1 )

d
= k|qx = x + k for k = 0 , 1 , 2, ,ω − x − 1
lx

Example 1.13

Suppose that the life table function is given by the formula:

lx = 100 − x for 0 ≤ x ≤ 100

Compute the probability function for K ( 75 ) .

Solution

The probability function for K ( 75 ) is:

d75+ k l75+ k − l75+ k + 1

Pr ( K ( 75 ) = k ) = =
l75 l75
( 100 − 75 − k ) − ( 100 − 75 − k − 1) 1
= =
( 100 − 75 ) 25

So K ( 75 ) has 25 possible values ( 0, 1, 2, , 24 ) that are equally likely to occur. ♦♦

Example 1.14

Suppose that the life table function is given by the formula:

lx = 1, 000 e −0.015x for 0 ≤ x < ∞

Compute the probability function for K ( 75 ) .

Solution

The probability function for K ( 75 ) is:

d75+ k l75+ k − l75+ k + 1

Pr ( K ( 75 ) = k ) = =
l75 l75
−0.015( 75 + k ) −0.015( 75+ k + 1 )
=
e −e
e −0.015 × 75
(
= e −0.015 k 1 − e −0.015 )
Note that this is a geometric distribution. ♦♦

It is also useful to develop formulas for the cumulative distribution function and survival
function of the curtate future lifetime.

19
Introduction to survival models Chapter 1

Recall that for any random variable FX ( x ) = Pr( X ≤ x ) , hence:

FK ( x ) ( k ) = Pr ( K ( x ) ≤ k ) = Pr ( K ( x ) = 0 ) + Pr ( K ( x ) = 1 ) + + Pr ( K ( x ) = k )
d d d d l − lx + k + 1
= x + x +1 + x + 2 + + x+ k = x
lx lx lx lx lx
= k + 1 qx for k = 0 , 1 , , ω − x − 1

The survival function of the curtate future lifetime is then easily derived as:

sK ( x ) ( k ) = Pr ( K ( x ) > k ) = 1 − FK ( x ) ( k ) = 1 − k + 1 qx = k + 1 px for k = 0 , 1 ,... , ω − x − 1

Example 1.15

Suppose that the life table function is given by the formula:

lx = 100 − x for 0 ≤ x ≤ 100

Compute the survival function for K ( 75 ) .

Solution

The survival function for K ( 75 ) is:

l 100 − (75 + k + 1) 24 − k
sK ( 75 ) ( k ) = k + 1 p75 = 75+ k + 1 = = for k = 0 , 1 , , 24 ♦♦
l75 100 − 75 25

Example 1.16

Suppose that the life table function is given by the formula:

lx = 1, 000 e −0.015x for 0 ≤ x < ∞

Compute the survival function for K ( 75 ) .

Solution

The survival function for K ( 75 ) is:

l e −0.015(75+ k + 1)
sK ( 75 ) ( k ) = k + 1 p75 = 75+ k + 1 = − 0.015(75)
= e −0.015( k + 1) for k = 0 , 1 , 2 ,
l75 e ♦♦

Let’s conclude this section with a summary of the key relations concerning the curtate future
lifetime, K ( x ) .

Key relations concerning the curtate future lifetime

d
1. f K ( x ) ( k ) = Pr ( K = k ) = k|qx = x + k for k = 0 , 1 ,... , ω − x − 1
lx
lx+k+ 1
2. sK ( x ) ( k ) = k + 1 px = for k = 0 , 1 ,... , ω − x − 1
lx
3. FK ( x ) ( k ) = k + 1 qx = 1 − k + 1 px for k = 0 , 1 ,... , ω − x − 1

20
 Chapter 1 Introduction to survival models

1.6 Important life table functions

Now that we have developed the basic properties of X , T ( x ) , and K ( x ) , it is time to study
additional features of these distributions, such as the life expectancy for a newborn.

The life table functions Lx and Tx

The functions Lx and Tx are useful devices in the calculation of life expectancy. They are
defined in terms of the life table function, lx , as follows:
x +1
Lx = ∫x ly dy

ω
Tx = ∫x ly dy = Lx + Lx + 1 + + Lω −1

These functions have interpretations in terms of the aggregate future lifetime of a group of lives
that die exactly as scheduled in the life table (ie we take a deterministic view of the life table).
Consider a brief time interval [ y , y + ∆y ] that is part of the interval [ x , ω ] . At the start of this brief
period there are ly survivors. We can estimate the total people-years lived by the survivors
during this brief period by ly ∆y . This approximation ignores the possibility that anyone dies in
the short time available.
If we now sum these people-years lived over a set of disjoint sub-intervals of length ∆y
ω
comprising the age interval [ x , ω ] , we will have a Riemann sum for the integral ∫x ly dy .

This Riemann sum can be interpreted as an approximation to the total number of people-years
lived after age x by the survivors to age x .
Taking a limit as ∆y goes to zero, then we have the integral:
ω
Tx = ∫x ly dy
which can be interpreted as the total people-years lived after age x by the survivors to age x .
Beware of confusing Tx with T ( x ) , the random future lifetime of a single life age x.

We can break the interval [ x , ω ] into subintervals of length one year, ie [ x , x + 1] ,

[ x + 1, x + 2], , [ω − 1, ω ] . The function Lx is calculated over just one of these one-year periods:
x+1
Lx = ∫x ly dy

which is the number of people-years lived by the survivors to age x during the next year.

Example 1.17

Compute the function Tx for the following life table functions:

(a) lx = 1, 000 − 10 x for 0 ≤ x ≤ 100

(b) lx = 1, 000 e −0.015x for 0 ≤ x < ∞

21
 Introduction to survival models Chapter 1

Solution

(a) We have:
100
100  ( 1, 000 − 10 y )2  ( 1, 000 − 10x )2 2
Tx = ∫x 1, 000 − 10 y dy =  −  = = 5 ( 100 − x )
 20  20
  x

(b) We have:
∞
∞ −0.015 y
 1, 000 e −0.015 y  1, 000 e −0.015x
Tx = ∫x 1, 000 e dy =  −
 0.015


=
0.015
♦♦
  x

Complete life expectancy

Now let’s develop some formulas to calculate life expectancies. Before we do that, though, it’s
useful to develop an alternative formula for expected value calculations. The work we do now
will make the subsequent derivations simpler.
Let’s assume that X is a continuous, positive-valued random variable whose mean and variance
both exist. The usual formula for calculating the expected value is:
∞
E[X ] = ∫ 0 x f X ( x ) dx
We can develop an alternative formula using integration by parts:
u = x , dv = f X ( x ) dx ⇒ du = dx , v = − sX ( x )

∫ 0 x f X ( x ) dx = ( −x sX ( x ) 0 ) + ∫ 0 sX ( x ) dx
∞ ∞ ∞
E[X ] =

∞ ∞
=  − lim x sX ( x ) + 0  + ∫ 0 sX ( x ) dx = ∫ 0 sX ( x ) dx
 x →∞ 

In the above calculation, the limit is zero. Apply L’Hopital’s rule writing x sX ( x ) as sX ( x ) / x −1 .
Then use the fact that lim x 2 f ( x ) = 0 if E  X 2  exists.
x →∞  

In other words, we can calculate E[ X ] as the area under the graph of the survival function.

If the density function is supported on the finite interval [0, ω ] , then we can replace ∞ in the
above integrals by ω , since both of the functions f X ( x ) and sX ( x ) are zero for x > ω .

So we have:
ω
E[ X ] = ∫ sX ( x ) dx
0

Another type of expected value that occurs in both survival model theory and loss distribution
theory is the limited expected value.
Let’s define X ∧ n as the minimum of the random variable X and the number n :

X if X ≤ n
X ∧ n = min {X , n} = 
n if X > n

22
 Chapter 1 Introduction to survival models

Since X ∧ n is a function of X , we can compute its expected value as above with limits 0 and n:
n ∞ n
E  X ∧ n  = ∫ 0 x f X ( x ) dx + ∫n n f X ( x ) dx = ∫ 0 x f X ( x ) dx + n sX ( n )
(
= − x sX ( x )
n
0 )+∫ n
0
sX ( x ) dx + n sX ( n ) =
n
∫ 0 sX ( x ) dx
Now we’re ready to calculate the two types of life expectancy: complete and curtate.

The complete expected future lifetime at age x is denoted by ex , and is defined as:

ex = E T ( x ) 

Using results derived earlier in this section, we can develop several methods to calculate this
expected value:
ω −x
ex = E T ( x )  = ∫0 t fT ( x ) ( t ) dt

ω −x
= ∫0 sT ( t ) dt (integration by parts)
ω −x ω
ω − x lx +t ∫0 lx +t dt ∫x ly dy
Tx
= ∫0 dt = = = ( substitute y = x + t )
lx lx lx lx

So, life expectancy can be computed in three distinct ways, as:

• the standard expected value of the continuous random variable T ( x )
• the area beneath the graph of the survival function
• the ratio of people-years lived after age x by the group to the number of people in the
group at age x (an average number of years lived by a member of the group of survivors
at age x ).

Example 1.18

Compute the complete life expectancy at age x for the following life table functions:
(a) lx = 1, 000 − 10 x for 0 ≤ x ≤ 100

(b) lx = 1, 000 e −0.015x for 0 ≤ x < ∞

Solution

Using the results from Example 1.17, we have:

(a)
T
ex = x =
(( 1, 000 − 10x ) 2
20 ) = 100 − x
lx 1, 000 − 10x 2

Tx 1, 000 e −0.015x /0.015 1

(b) ex = = −0.015 x
= = 66.67 ♦♦
lx 1, 000 e 0.015

23
 Introduction to survival models Chapter 1

The temporary complete life expectancy at age x is denoted ex:n , and is used to represent the
expected number of years lived by ( x ) in the next n years.

The random number of years lived by ( x ) in the next n years is:

T ( x ) if T ( x ) ≤ n (death within n years)

T ( x ) ∧ n = min {T ( x ) , n} = 
 n if T ( x ) > n (survives n years)

As a result, we have the following:

ex:n = E T ( x ) ∧ n 

n
= ∫ 0 sT ( x ) ( t ) dt
n x +n
n lx +t ∫0 lx +t dt ∫x ly dy
Tx − Tx + n
= ∫ 0 lx
dt = = = ( substitute y = x + t )
lx lx lx

Example 1.19

Compute the 10-year temporary complete life expectancy at age 50 for the following life table
functions:
(a) lx = 1, 000 − 10 x for 0 ≤ x ≤ 100

(b) lx = 1, 000 e −0.015x for 0 ≤ x < ∞

Solution

Using the results from Example 1.17, we have:

(a)
T −T
e50:10 = 50 60 =
( 5002 − 4002 ) /20 = 9
l50 500

(b) e50:10 =
T50 − T60
=
(
1, 000 e −0.015×50 − 1, 000 e −0.015×60 /0.015
= 9.280
) ♦♦
l50 1, 000 e −0.015×50

Key formulas for complete life expectancy

ω −x ω −x Tx
1. ex = E T ( x )  = ∫0 t fT ( x ) ( t ) dt = ∫0 sT ( t ) dt =
lx
where sT ( x ) ( t ) = t px

n Tx − Tx + n
2. ex:n = E T ( x ) ∧ n  = ∫ 0 sT ( t ) dt = where sT ( x ) ( t ) = t px
lx

3. ex = ex:n + n px ex + n

ω −x 2 ω −x
E T ( x )  =
2
4.
  0 ∫
t fT ( x ) ( t ) dt =
0
2t sT ( t ) dt ∫

24
Chapter 1 Introduction to survival models

The proof of the third formula goes as follows:

Tx (Tx − Tx +n ) + Tx +n Tx − Tx +n lx +n Tx +n
ex = = = + ×
lx lx lx lx lx + n

= ex:n + n px ex + n

This result has an intuitive interpretation. The life ( x ) expects to live ex:n years in the next n

years, and, given survival to age x + n , expects to live an additional ex + n more years.

The proof of the fourth formula relies on integration by parts with u = t 2 and dv = fT ( x ) ( t ) dt .

Curtate life expectancy

The curtate expected future lifetime at age x is denoted by ex , and is defined in terms of the
curtate future lifetime K ( x ) :

ex = E  K ( x ) 

So, ex is the expected number of full years (the fractional part of the final year of life is not
counted) lived by the life ( x ) after age x .

We have:
ω − x −1
ex = E  K ( x )  = ∑ k Pr ( K ( x ) = k )
k =0

ω − x −1 ω − x −1
k dx + k dx + 1 + 2 dx + 2 + 3 dx + 3 + + (ω − x − 1 ) dω −1
= ∑ k ⋅ k|qx = ∑ lx
=
lx
k =0 k =0

lx + 1 + lx + 2 + lx + 3 + + lω −1
= (note: lx + k = dx + k + dx + k + 1 + + dω −1 )
lx
the survivors at age x + k
die at age x + k or later

= px + 2 px + 3 px + + ω − x − 1 px

The temporary curtate life expectancy at age x is denoted e , and is defined as:
x:n

ex:n = E  K ( x ) ∧ n 

It is the expected number of full years (the fractional part of the final year of life is not counted)
lived by the life ( x ) in the next n years. In a similar fashion, we can show that:

e = E  K ( x ) ∧ n  = px + 2 px + + n px
x:n

There is a recursive relation:

ex = ex:n + n px ex + n
ex = ex:1 + px ex + 1 = px + px ex + 1 (the special case with n = 1)

25
 Introduction to survival models Chapter 1

Example 1.20

Suppose that the life table function is given by the formula:

lx = 100 − x for 0 ≤ x ≤ 100

(a) Compute the curtate life expectancy for a newborn.

(b) Compute the 10-year temporary curtate life expectancy at age 50.

Solution

(a) The curtate life expectancy for a newborn is:

e0 = p0 + 2 p0 + + 99 p0

l +l + + l99 99 + 98 + +1 4950
= 1 2 = = = 49.5
l0 100 100

(b) The 10-year temporary curtate life expectancy at age 50 is:

e = p50 + 2 p50 + + 10 p50

50:10

l51 + l52 + + l60 49 + 48 + + 40 445

= = = = 8.9 ♦♦
l50 50 50

Key formulas for curtate life expectancy

lx + 1 + lx + 2 + lx + 3 + + lω −1
1. e x = px + 2 px + 3 px + + ω − x − 1 px =
lx

lx + 1 + lx + 2 + + lx + n
2. ex:n = px + 2 px + + n px =
lx

3. ex = e + n px ex + n
x:n

l + 3 lx + 2 + 5 lx + 3 +
E K ( x )  = x +1
2
4.
  lx

The proof of the second moment formula above could be established using summation by parts, a
technique that will not be covered since it is not often useful in contingent payment model
theory. It can also be proven with some fancy algebraic manipulation starting with the formula
for a second moment:
ω − x −1
E K 2  =
  ∑ k 2 Pr ( K = k )
k =0
ω − x −1
= ∑ k 2 k|qx
k =0
ω − x −1
lx + k − lx + k + 1
= ∑ k2 ×
lx
k =0

26
Chapter 1 Introduction to survival models

The central mortality rate

The n-year central mortality rate denoted by n mx computes a weighted average of the force of
mortality over the range from age x to age x + n .
It can be calculated in several ways:
n
∫ 0 t px µ ( x + t ) dt n qx n dx / lx n dx n dx
n mx = = = = =
n
ex:n (Tx − Tx +n ) / lx Tx − Tx + n Lx + Lx + 1 + + Lx + n−1
∫ 0 t px dt
The symbol n dx is defined as:

n dx = lx − lx + n

ie it is the expected number of deaths in the age range from x to x + n .

When n = 1 we simply write mx instead of 1 mx . In this case, we have:

dx
mx =
Lx

We now have three annual measures of mortality at age x : µ ( x ) , qx , and mx .

The force of mortality µ ( x ) is not a probability, but µ ( x ) ∆x is approximately the probability

that a life currently aged x will die in the next instant ∆x .
The probability that a life currently aged x will die within the next year is qx :

dx
qx =
lx

where lx is the total number of survivors aged x at the start of the year.

The only difference between qx and mx is the denominator. The quantity Lx in the definition of
mx , is often called an exposure, since it represents the total number of people-years lived by the
lx lives age x over the next year (when they are exposed to death).

Example 1.21

Suppose that the life table function is given by the formula:

0.5
lx = ( 100 − x )

Compute q50 , µ ( 50 ) and m50 .

Solution

For m50 we will need the value of e50:1 :

1
1 1 l50 +t 1  ( 1 − 0.02t )1.5 
0.5
e50:1 = ∫ 0 t p50 dt = ∫ 0 l50
dt = ∫ 0 ( 1 − 0.02t ) dt =  −
 1.5 × 0.02


= 0.99498
  0

27
Introduction to survival models Chapter 1

Now we have the necessary information to complete the three calculations:

0.5
l51 ( 49 )
q50 = 1 − =1− = 0.01005
l50 ( 50 )0.5
−0.5
′
l50 − ( 100 − 50 ) /2 0.5
µ ( 50 ) = − =− 0.5
= = 0.01000
l50 ( 100 − 50 ) 100 − 50
q50 0.01005
m50 = = = 0.01010 ♦♦
e50:1 0.99498

The function a ( x )

The function a ( x ) is defined as the average part of the final year of life lived by someone who
dies between ages x and x + 1 . From this verbal description we can see that it is the conditional
expected value:

E T ( x )| T ( x ) ≤ 1

since the event T ( x ) ≤ 1 indicates that the life ( x ) dies within a year.

Rather than calculating a( x ) by integrating a conditional density, let’s look at an intuitive way.

Consider the lx lives aged x . Let’s split them into the dx lives that die in the next year, and the
lx + 1 lives that survive the next year. Now consider ex:1 , the average time lived in the next year.

• For the lives that die in the next year, the average time lived is a( x ) , by definition.

• For the survivors to age x + 1 , the average time lived in the next year is 1.

The number ex:1 is a weighted average of time lived in the next year by these two groups, so:

d ⋅ a ( x ) + lx + 1 ⋅ 1 d ⋅ a ( x ) + lx + 1 ⋅ 1
ex:1 = x = x = qx a ( x ) + px
dx + lx + 1 lx

ex:1 − px L −l
⇒ a(x) = = x x +1
qx dx

For example, if lx = 100 − x , we have:

x +1 x +1 ( x + 1 )2 − x 2
Lx = ∫x ly dy = ∫ ( 100 − y ) dy = 100 − = 100 − x − 0.5
x 2
and:
L −l 100 − x − 0.5 − (100 − x − 1)
a( x ) = x x + 1 = = 0.5
dx 1

So, in this case the average time of death is midway through the final year of life.

28
 Chapter 1 Introduction to survival models

1.7 Extending a life table to a continuous survival model

In Sections 1.3 through 1.6 we have developed the theory of continuous survival models,
beginning with the distribution of the lifetime of a newborn. Now let’s consider the practical
application of this theory.
Suppose that we have a discrete life table that lists the number of survivors lx at whole number
ages 0, 1, 2, , ω , where lω = 0 .

How can we compute life table functions such as 2.5 p3 , µ ( 2.5 ) , or e25 , which rely on a
continuous model?
To calculate these functions, we’ll need to start by extending lx to a continuous model defined
for all ages x between 0 and ω . In order to do this, we’ll have to make an assumption about the
pattern in which deaths occur between the integer ages.
A common assumption is that the dx deaths that occur between age x and age x + 1 are spread
uniformly over the year. This is known as the uniform distribution of deaths assumption, or the
UDD assumption for short.
The implication of the UDD assumption is that we define the life table function at fractional ages
by interpolating linearly between the values at integer ages.
So, if x is an integer age and 0 ≤ t ≤ 1 , then we have:
lx + t = ( 1 − t ) lx + t lx + 1 (linear interpolation)
= lx − t ( lx − lx + 1 ) = lx − t dx

The resulting piecewise linear graph of lx is continuous and non-increasing (usual properties of
the life table function). The same must be true of the survival function, sX ( x ) .

•
lx
•
•
•
•

0 1 2 3 4 x

The UDD assumption has several important consequences.

For integer x , and 0 ≤ t ≤ 1 , we have:
lx +t l − t dx
t px = = x = 1 − t ⋅ qx
lx lx

t qx = 1 − t px = t ⋅ q x

lx′ +t ( lx − t dx )′ dx qx
µ (x + t) = − =− = =
lx + t lx − t dx lx − t dx 1 − t qx

29
Introduction to survival models Chapter 1

We can now perform some of the calculations that would have been impossible from discrete
data without the UDD assumption. For example, we have:
l5.5 l5 − 0.5 d5 d2 q2
2.5 p3 = = and µ ( 2.5 ) = =
l3 l3 l2 − 0.5 d2 1 − 0.5 q 2

Example 1.22

Suppose that q90 = 0.4, q91 = 0.7, and q92 = 1.0 . Calculate e90 under the UDD assumption.

Solution

A standard method of calculation is to compute the area beneath the graph of the survival
function. This can be accomplished with the help of calculus (Method 1 below), or it can be done
geometrically (Method 2).
Method 1.
We have:

t p90 = 1 − t ⋅ q90 = 1 − 0.4t for 0 ≤ t ≤ 1

t p91 = 1 − t ⋅ q91 = 1 − 0.7 t for 0 ≤ t ≤ 1
t p92 = 1 − t ⋅ q92 = 1 − t for 0 ≤ t ≤ 1

Hence:

e90 = e90:1 + p90 e91 = e90:1 + p90 ( e91:1 + p91 e92 )

1  1 1 
=∫ t p90 dt + p90  ∫ t p91 dt + p91 ∫ t p92 dt  (note that 3 p90 = 0)
0  0 0 
1  1 1 
= ∫ 0 ( 1 − 0.4t ) dt + 0.6  ∫ ( 1 − 0.7 t ) dt + 0.3 ∫ ( 1 − t ) dt 
 0 0 

(
= ( 1 − 0.2 ) + 0.6 ( 1 − 0.35 ) + 0.3 ( 1 − 0.5 ) = 1.28 )
Method 2.
Begin by drawing the graph of the continuous and piecewise linear survival function.

t p90

0 1 2 3 t

The area of a trapezoid whose base is 1 unit long is equal to the height at the midpoint:
3
e90 = ∫0 t p90 dt = sum of three trapezoidal areas

= 0.5 p90 + 1.5 p90 + 2.5 p90 = ( 1 − 0.4 × 0.5 ) + 0.6 ( 1 − 0.7 × 0.5 ) + 0.6 × 0.3 ( 1 − 0.5 )

= 1.28 ♦♦

30
Chapter 1 Introduction to survival models

Example 1.23

Suppose that q90 = 0.4, q91 = 0.7, and q92 = 1.0 . Draw a graph of the pdf for T ( 90 ) under the
UDD assumption.

Solution

We should start by looking at the picture of the survival function in the Solution to Example 1.22.
In general, the pdf is equal to the negative derivative of the survival function. This derivative
(the tangent slope) is undefined at t equal to 1 and 2 due to the sharp corners on the graph at these
locations.
So, we just need to look at the survival function and note the slope on each of the three linear
segments. The result is:

− ( 1 − 0.4t )′ for 0 < t < 1



fT ( 90 ) ( t ) = − ( t p90 )′ = − ( 1.02 − 0.42t )′ for 1 < t < 2

− ( 0.54 − 0.18t )′ for 2 < t < 3


note: for 1 ≤ t ≤ 2 we have t p90 = p 90 t −1 p91 = 0.6 ( 1 − ( t − 1) 0.7 ) = 1.02 − 0.42t

Hence:
0.40 for 0 < t < 1

fT ( 90 ) ( t ) = 0.42 for 1 < t < 2
 for 2 < t < 3
0.18
The graph is as shown below:

f(t)
ο ο
° ο

ο ο

0 1 2 3 t

Note: A quicker way to calculate the pdf form is to note that:

fT ( x ) ( t ) = k|qx when k = T ( x )  .

For example, notice in the example above that:

0|q90 = q90 = 0.4

1|q90 = p90 q91 = 0.42

2|q90 = p90 × p91 × q92 = 0.18 ♦♦

31
 Introduction to survival models Chapter 1

Key formulas for the UDD assumption (if x is an integer age and 0 ≤ t ≤ 1 )
1. lx + t = ( 1 − t ) lx + t lx + 1 = lx − t dx (linear interpolation)

2. t px = 1 − t ⋅ qx , t qx = t ⋅ qx

dx qx
3. µ (x + t) = =
lx − t dx 1 − t qx

4. fT ( x ) ( t ) = k|qx for k < t < k + 1

5. a ( x ) = 0.5 for all integer ages x

6. ex = ex + 0.5 for all integer ages x

Here is a brief explanation of the final two results in the summary. The UDD assumption gets its
name from the fact that linear interpolation in effect spreads the deaths out uniformly and
continuously over each year of age. For example, we assume that in the age interval
[ x + 0.40, x + 0.65] there will be 25% of the deaths between ages x and x + 1 .

With a little effort it can be shown that the fractional part of the final year of life is uniformly
distributed on the interval from 0 to 1, regardless of the (integer) age at the start of that year. This
should make it easy to see where the final two results come from.
A simple application of the final point provides an alternative solution to Example 1.22. It is easy
to see from the given mortality rates that:
e90 = p90 + 2 p90 + 3 p90 = 0.6 + 0.6 × 0.3 + 0.6 × 0.3 × 0 = 0.78
UDD ⇒ e90 = e90 + 0.5 = 0.78 + 0.5 = 1.28

1.8 Select and ultimate life tables

When an individual applies to buy a life insurance policy, there will typically be some sort of
medical screening process, either by filling out a questionnaire or visiting a doctor for a physical
examination. This protects the insurance company from people who know that they are in poor
health and likely to result in a claim on the policy. This process (known as underwriting) acts as
a filter. The healthy lives are accepted for the policy, and the unhealthy lives are either not
accepted or are issued a policy at a higher premium rate.
As a result, the average mortality of individuals who have just passed the medical screening
process will be lower than the mortality of the general population, which includes a mix of
healthy and unhealthy lives. It’s important to note that the differential in mortality rates is
greatest immediately after the underwriting process, ie as soon as the policy is sold. Within, say,
3 years of the medical screening process, the mortality rates of the policyholders will tend back
towards those of the general population as some of the policyholders contract illnesses and their
health deteriorates.
The underwriting process has an interesting implication. The mortality of a 40-year-old
policyholder who has just bought a policy (and has therefore just passed the medical screening
process) should be lower than the mortality of a 40-year-old policyholder who bought his policy
several years ago.

32
Chapter 1 Introduction to survival models

The lower mortality rates that apply in the years following the “selection” process (ie
underwriting) are called select mortality rates. The general rates that apply once the filtering
effects of underwriting have worn off are called ultimate mortality rates. The period of time for
which the differential exists (in this case 3 years) is called the select period.
Other events can give rise to mortality differentials too, eg giving up smoking. The average
mortality of people immediately after giving up smoking is higher than that of the general
population. After 10 years, the harmful effects of smoking may have reduced to such an extent
that the ex-smokers’ average mortality returns to that of the general population. The select
period here is 10 years.
In a general select and ultimate model, the mortality rate qx (integer age x ) for a life subject to
the model lx is adjusted to a select mortality rate for several years (the select period) after some
event takes place. At the end of the select period, the mortality rate reverts to the ultimate
mortality rates.
Let’s now look at the notation with the help of a select and ultimate table.

[x] l[ x ] l[ x ]+ 1 l[ x ]+ 2 lx + 3 x+3
22 35,628 35,619 35,607 35,596 25
23 35,615 35,604 35,590 35,581 26
24 35,602 35,589 35,575 35,562 27

The select portion of the table refers to the columns headed l[ x ] , l[ x ] + 1 , and l[ x ]+ 2 .

The subscript [ x ] + k indicates that a life “selected” at age x is currently age x + k , ie the life was
“selected” k years ago. In these models both x and k are assumed to be whole numbers.
The ultimate portion of the table refers to the column headed lx + 3 . The select period is 3 years in
length, so we start to use ultimate mortality rates 3 years after the life was selected.
What we have in effect is a distinct survival model for the curtate future lifetime after each
integer age at which selection could take place.
For example, if a life is selected at age 23, then the curtate future lifetime for this individual at age
23 is based on the numbers l[ 23] , l[ 23] + 1 , l[ 23] + 2 , l26 , l27 , surviving to each whole number age.

In general, we follow the table across the row for the age of selection until we arrive at the
ultimate column (at the end of the select period). At that point we continue down the ultimate
column.
The age subscript in other life table functions conforms to the [ x ] + k notation. For example, for a
life selected at age 23 we would have the following results according to the table given above.
The probability that a newly selected life aged 23 survives one year is:
l[ 23] + 1 35, 604
p[ 23] = = = 0.99969
l[ 23] 35, 615

The probability that a newly selected life aged 23 dies within one year is:
q[23] = 1 − p[ 23] = 1 − 0.99969 = 0.00031

33
 Introduction to survival models Chapter 1

The probability that a life aged 24, selected at age 23 (one year ago) survives one year is:
l[ 23] + 2 35, 590
p[ 23]+ 1 = = = 0.99961
l[ 23]+ 1 35, 604

The probability that a life aged 25, selected at age 23 (two years ago) survives one year is:
l 35, 581
p[ 23]+ 2 = 26 = = 0.99975
l[ 23]+ 2 35, 590

The probability that a life aged 26, selected at age 23 (three years ago) survives one year is:
l27 35, 562
p26 = = = 0.99947
l26 35, 581

We can calculate the curtate life expectancy as:

( )
e[ 23] = e[ 23]:3 + 3 p[ 23] e26 = p[ 23] + 2 p[ 23] + 3 p[ 23] + 3 p[ 23] ( p26 + 2 p26 + )
To compute a long-term survival probability such as 25 p[ 23] , we must consider the split of 25-
year period between the 3-year select period and the following 22-year ultimate period:
l26 l l
25 p[ 23] = 3 p[ 23] 22 p26 = × 48 = 48
l[ 23] l26 l[ 23]

Example 1.24

Suppose we have a select and ultimate model based on the life table lx = 100 − x at integer ages x,
where the select period is 2 years and the select mortality rates are calculated according to the
rule:

q[ x ] + k = ( 1 − 0.05 ( 2 − k ) ) qx + k for k = 0 , 1

Determine the following values:

(a) p[ 25]

(b) 2 p[ 25]

(c) 3 p[ 25]

(d) e[ 25] :3

Solution

This formula indicates that select mortality at the age of selection (ie when k = 0 ) is 90% of
ultimate mortality at the same age. On the other hand, one year after selection, the select rate is
95% of the ultimate rate at the same age. The effect of selection wears off after 2 years.
Notice first that the ultimate mortality and survival rates at age x are:
lx − lx + 1 1 100 − x − 1
qx = = and px = 1 − qx =
lx 100 − x 100 − x

34
 Chapter 1 Introduction to survival models

As a result, we have:

 1 
(a) p[ 25] = 1 − q[ 25] = 1 − 0.9 q25 = 1 − 0.9   = 0.98800
 75 

 1 
(b) p[ 25] + 1 = 1 − q[ 25]+ 1 = 1 − 0.95 q26 = 1 − 0.95   = 0.98716 ⇒ 2 p[ 25] = 0.97532
 74 
l 100 − 28
(c) 3 p[ 25] = 2 p[ 25] p27 = 2 p[ 25] × 28 = 0.97532 × = 0.96196
l27 100 − 27

(d) e[ 25]:3 = p[ 25] + 2 p[ 25] + 3 p[ 25] = 0.98800 + 0.97532 + 0.96196 = 2.92527 ♦♦

Note: In this text it is a convention that the numbers displayed are often rounded, but the un-
rounded form of the number is used in subsequent calculations.

Example 1.25

Using the same model as in Example 1.24, determine the following values:
(a) l[ 25] + 1

(b) l[ 25]

Solution

When the model is specified by select mortality rates in terms of ultimate mortality rates, to
compute the select portion of the life table we have to work backwards from the ultimate portion
of the table.
We’ll use the select survival rates calculated in Example 1.24. Hence:
l27 73
(a) l[ 25] + 1 = = = 73.94935
p[ 25]+ 1 0.98716

l[ 25]+ 1 73.94935
(b) l[ 25] = = = 74.84752
p[ 25] 0.98800

1.9 Mortality laws

Historically, various mathematical formulas have been used to model the force of mortality for
human lives. In this section we’ll look at some of these formulas and compute the associated
survival functions.
Before we get started, it’s useful to derive a formula for the effect on the survival function of a
linear transformation on the force of mortality. Suppose that we have a force function:
µ * (x) = a µ (x) + b

where the constants are chosen so that µ * ( x ) ≥ 0 for all x > 0 .

Let s * ( x ) be the survival function corresponding to the force µ * ( x ) .

35
Introduction to survival models Chapter 1

Then the relation between s * ( x ) and sX ( x ) is as follows:

 x   x 
s * ( x ) = exp  − µ * ( y ) dy  = exp  −
 0 ∫   0 ∫ (
aµ ( y ) + b dy 

)
  x     x 
= exp  − a  µ ( y ) dy  − bx  = exp  − a 
∫ µ ( y ) dy   exp ( − bx )
∫
  0     0 
a
  x 
=  exp  − ∫ µ ( y ) dy   exp ( − bx )
  0 
= e −bx ( sX ( x ) )
a

For example, if female mortality is given by µF ( x ) and male mortality is given by µ M ( x ) where:

µ M ( x ) = 0.01 + 1.005 µF ( x )

then the survival function for males is:

s M ( x ) = e −0.01x ( sF ( x ) )
1.005

Gompertz’s law of mortality

Under Gompertz’s law, the force of mortality is determined by the formula:

µ ( x ) = Bc x where B > 0 , c > 1 , x ≥ 0

This formula has a geometrically increasing force of mortality to model the effect of aging.
The corresponding survival function is:

 x 

sG ( x ) = exp  − ∫ Bc y dy  = exp  −
Bc y
x
 = exp  − (
 B cx − 1
) 
 0   ln ( c )   ln ( c ) 
 0  



( (
= exp −m c x − 1 )) where m =
B
ln ( c )

Makeham’s law of mortality

Makeham’s law adds an extra constant to Gompertz’s law to reflect an additional constant level
of risk at all ages (eg accidental death). Under Makeham’s law, the force of mortality is
determined by the formula:

µ ( x ) = A + Bc x where A > 0 and B > 0

Since this force formula is a linear transformation of Gompertz’s formula, the Makeham survival
function, s M ( x ) , is closely related to the Gompertz survival function, sG ( x ) :

s M ( x ) = e − Ax sG ( x )

Weibull’s law of mortality

Under Weibull’s law, the force of mortality is determined by the formula:

µ ( x ) = kx n where k > 0 , n > 0 , and x ≥ 0

The Weibull survival function is:

 − kx n+ 1 
sW ( x ) = exp  
 n+1
 
36
Chapter 1 Introduction to survival models

De Moivre’s law of mortality

De Moivre’s law is based on a life table function that is linear:
lx = ω − x for 0 ≤ x ≤ ω

This leads to a survival function of:

lx x
sX ( x ) = =1− for 0 ≤ x ≤ ω
l0 ω
and a force of mortality of:
−lx′ 1
µ (x) = = for 0 ≤ x < ω
lx ω−x
Since this law will be used extensively in examples illustrating theory in succeeding chapters, it is
worth listing some more properties of this mortality law.
The pdf of the lifetime function is:
ω −x 1 1
f X ( x ) = sX ( x ) µ ( x ) = × = for 0 ≤ x < ω
ω ω−x ω
and the pdf of the future lifetime function is:
ω −x −t 1 1
fT ( x ) ( t ) = sT ( x ) ( t ) µ ( x + t ) = × = for 0 ≤ t < ω − x
ω−x ω −x −t ω −x
It is important to note that these final two formulas indicate that:
• the lifetime of a newborn follows a uniform distribution on the interval [0, ω ]

• the future lifetime after age x follows a uniform distribution on the interval [0, ω − x ] .

Constant force of mortality

Finally, it is worthwhile to mention the constant force model:
µ (x) = µ for all ages x

While not suitable as a model of human mortality, it is again used in later chapters to illustrate
the theory being presented.
The survival function is:

sX ( x ) = e − µ x for all x > 0

The pdf of the lifetime function is:

f X ( x ) = sX ( x ) µ ( x ) = µ e − µ x for all x > 0

and the pdf of the future lifetime function is:

e − µ ( x +t )
fT ( x ) ( t ) = sT ( x ) ( t ) µ ( x + t ) = × µ = µ e − µ t for all t > 0
e− µ x
The lifetime of a newborn and the future lifetime after age x follow an identical exponential
distribution with mean 1/ µ . This distribution will also be used extensively in loss model
theory.

37
 Introduction to survival models Chapter 1

Chapter 1 Practice Questions

Question 1.1
Compute the following probabilities from the life table in Section 1.2:

2 p0 , 2| q0 , 4| 2 q3 , p4 , q 5

Question 1.2
Suppose that 2|q1 = 0.015 . Discuss the distribution of the random number of deaths between
ages 3 and 4 for a group of 20 lives currently age 1.

Question 1.3
The life table below is a survival model for a group of newborns suffering from a certain heart
impairment.

x 0 1 2 3 4
lx 100 46 19 6 0

Give the probability function for the curtate future lifetime of a member of this group of impaired
lives.

Question 1.4
Suppose that µ ( x ) = 3 / ( 2 + x ) for x > 0 . Determine the pdf, cdf, and survival function for the
∞
lifetime of a newborn. Check your work by verifying properties such as ∫0 f ( x ) dx = 1 ..

Question 1.5
Using the survival model in Question 4, determine the values of p1 and 2|q1 .

Question 1.6
Suppose that µ ( x ) = 0.5 / ( 100 − x ) for 0 ≤ x < 100 . Determine the pdf, cdf, and survival function
for the lifetime of a newborn.

Question 1.7
Using the survival model in Question 6, determine the values of 20 p40 and 20|20 q 40 .

38
Chapter 1 Introduction to survival models

Question 1.8
Suppose that µ ( x ) = 1.1/ ( 100 + x ) for x ≥ 0 . Compute t p20 , the survival function for the future
lifetime after age 20.

Question 1.9
For the survival model in Question 1.8, determine the probability that the curtate future lifetime
at age 20 is less than 2.

Question 1.10
Assuming that µ ( x ) = 0.015 for all x ≥ 0 , determine the pdf for the continuous lifetime after age
20.

Question 1.11
For the survival model in Question 1.10, determine the probability function for the curtate
lifetime after age 20. What is the probability that the curtate lifetime exceeds 1?

Question 1.12
0.95
For the life table function lx = 1, 000 ( 100 − x ) for 0 ≤ x ≤ 100 , develop a formula for Tx .

Question 1.13
For the survival model in Question 1.12, determine e20:5 .

Question 1.14
Suppose that µ ( x ) = 0.015 for all x ≥ 0 . Determine ex:1 and a ( x ) .

Question 1.15
For the survival model in Question 1.14, determine q20 and m20 .

Question 1.16
The life table function lx µ ( x ) is called the curve of deaths. It approximates the expected number
of deaths in the year of age from x to x + 1 . Calculate the curve of deaths for the life table given
0.75
by lx = 1, 000 ( 100 − x ) for 0 ≤ x ≤ 100 .

39
Introduction to survival models Chapter 1

Question 1.17
You are given: q30 = 0.01 , q31 = 0.02 , q32 = 0.03 . Computing the following using the UDD
assumption to calculate the number living at fractional ages:
µ ( 31.4 ) , 1.4 p30 , fT ( 30 ) ( 1.4 )

Question 1.18
Assuming the same mortality rates as in Question 1.17, calculate e . Then see if you can
30:3
reason how to adapt the UDD relation ex = ex + 0.5 to calculate e30:3 .

Question 1.19
Suppose that lx = 1, 000 e −0.02 x for whole number ages only. A select and ultimate model uses this
function for the ultimate portion and the relations below for the 2-year select period:

q[ x ] + k = ( 1 − 0.1 ( 2 − k ) ) qx + k .

Determine 3 q[ 20 ] .

Question 1.20
For the select and ultimate model in Question 1.19, determine l[ 20] .

40