MTH5124 Actuarial Mathematics I Lecture Notes

Lecture Notes
MTH5124: Actuarial Mathematics I
Dr Adrian Baule
School of Mathematical Sciences
Queen Mary University of London
October 4, 2019
2
Contents
0 Prologue 5
0.1 What is an actuary? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
0.2 About this course . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
0.3 About these notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
0.4 Life Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
0.5 Books and tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
0.6 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1 Compound interest 11
1.1 Two types of interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.1.1 Simple interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.1.2 Compound interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2 Nominal and effective interest rates . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2.1 Accumulation factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2.2 Nominal interest rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2.3 Effective interest rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3 Force of interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3.1 Time-dependent interest rates . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3.2 Force of interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3.3 Special case of constant force of interest . . . . . . . . . . . . . . . . . . . . 21
1.4 Rates of discount . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.4.1 Relation between nominal rates of discount and interest . . . . . . . . . . . . 24
1.5 Discounting Cash Flows or Present Values . . . . . . . . . . . . . . . . . . . . . . . 26
1.5.1 Discrete cash flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.6 Annuities-certain: introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.7 Annuities-certain: more variations . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.8 Continuous cash flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.8.1 Continuous cash flow with variable force of interest . . . . . . . . . . . . . . 36
1.9 Repayment of Loans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.9.1 Schedule of payments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.9.2 Consolidating loans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.10 Investment project appraisal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.11 Fixed Interest Securities and Other Investments . . . . . . . . . . . . . . . . . . . . 43
1.11.1 Fixed Interest Securities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.11.2 Cash including Treasury Bills . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.11.3 Inflation Linked Bonds and Real Returns . . . . . . . . . . . . . . . . . . . . 48
1.11.4 Equities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
1.11.5 Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
1.11.6 Taxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3
2 Life tables and life-table functions 59
2.1 Lifetime as a random variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.2 Basic life-table functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.3 Force of mortality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.4 Analytical laws of mortality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.5 The expectation of life . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.6 Interpolation for fractional ages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.7 Select mortality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3 Life insurance and related functions 85

3.1 Introduction to life assurance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.2 Whole-life assurance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.3 Whole-life annuities payable annually . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.4 Policies of duration n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.5 p-thly payments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.6 Loss, policy value, prospective reserve and surrender values . . . . . . . . . . . . . . 106
3.7 Retrospective accumulation and reserves . . . . . . . . . . . . . . . . . . . . . . . . 111
3.8 Importance of variance, normal approximation . . . . . . . . . . . . . . . . . . . . . 114
4
Chapter 0
Prologue
0.1 What is an actuary?

“An actuary is someone who expects everyone to be dead on time.” is one definition you can
find on the internet.1 In fact, actuaries do much more than deal with the probabilities of peo-
ple dying! In short, they use financial and statistical theories to quantify and manage risk in
all areas of business. Traditionally actuaries specialize in consultancy, investment, life and gen-
eral insurance or pensions. However, their analytical skills are also increasingly valuable in other
areas—after all, “risk management” is of wide importance in financially turbulent times. Actuar-
ies are highly-regarded and (well-paid!) professionals. For more details of career paths, etc., see
http://www.actuaries.org.uk/becoming-actuary/.
0.2 About this course

The idea of this course is to introduce you to some of the basic mathematical ideas used in actuarial
work. MTH5124 is designed for second or third year undergraduates and assumes a background
in basic Calculus and Probability.2 All practicalities about the course itself (timetable, coursework,
assessment details), etc., can be found on the course QM+ website
http://qmplus.qmul.ac.uk/course/view.php?id=767.
Important announcements and corrections will also be posted there.
The course uses familiar mathematical concepts (e.g., geometric series, probability distributions)
but in an unfamiliar context. I hope you will be interested to see how maths you already know can
be used effectively in the “real world”. However, one of the problems in applying maths to a different
field is that you often have to learn new terminology, notation, etc. in order to communicate with
specialists in that area. This is certainly the case here; indeed you will soon be introduced to a
whole zoo of complicated actuarial symbols and new vocabulary. It is crucial for success that you
learn this “new language” so that the familiar mathematical objects do not become lost in the fog
of unfamiliar actuarial terminology.
The course has three main “chapters”:
1. Compound interest: Here we will cover various (interrelated) ways to quantify how compound
interest is added to a loan/investment. You will learn how to calculate accumulations given
the force of interest (or related parameters) and how to deal with series of payments, in
1
A quick search with Google yields a variety of other humorous, and not-so-humorous, definitions as well as more
useful career descriptions.
2
Formal prerequisites are Calculus II and Introduction to Probability; contact the lecturer if you have difficulty
meeting these.
5
particular annuities-certain and perpetuities. You will also learn about the most common
forms of investments and how to value investments using compound interest.
2. Life tables and life-table functions: Life tables are the actuary’s basic tool. You will learn
how to interpret them in order to find various probabilities related to life and death.
3. Life insurance and related functions: Here we will combine material from the first two
chapters to deal with situations involving payments (with compound interest) whose value
and/or timing may depend on a person’s survival or death! Life insurance is the classic
example here.
0.3 About these notes

These notes will cover the material in roughly the same order as the lectures but their style will
probably be slightly different. In particular, the printed notes may contain some longer explanations
and extra examples which time prevents me covering in class. The lectures, however will be important
for emphasizing the main points and giving exam tips—I strongly recommend that you attend!
To help you with revision, all important actuarial terms are printed in bold. You must be sure
to understand what these mean, both in everyday language (could you explain them to your grand-
mother or your next-door neighbour?) and in terms of the associated mathematical formulation.
The notes will also contain a number of “examples” and “exercises”. For the former you will find
full details of the working; for the latter (usually) only the answers. A good way to check your own
understanding would be to read the text and try the associated exercises (contact me if you have
any difficulty getting the stated answers). I intend the unstarred exercises to correspond roughly to
the Key Objectives for the course (i.e., everyone should be able to do them); starred exercises will
be somewhat harder and should be attempted by those aiming for a high grade.
0.4 Life Tables

The examples in this document are based on the following life tables:
• English Life Table No 17,
• AMC00 mortality table for assured lives.
These are available on the QM+ page for MTH5124. Copies will be provided in the examination.
Full details of mortality tables published by the CMI can be found at

https://www.actuaries.org.uk/learn-and-develop/continuous-mortality-investigation/cmi-mortality-and-
morbidity-tables.
The CMI is a subsidiary of the Institute and Faculty of Actuaries, and was previously known
as The Continuous Mortality Investigation. CMI tables relate to the mortality experience of life
insurance policyholders and the members of pension schemes.
The English Life Tables represent the mortality experience of the population of England and
Wales. Tables are published by the Office of National Statistics (ONS); further information on
ELT17 can be found at
https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/lifeexpectancies-
/bulletins/englishlifetablesno17/2015-09-01.
6
0.5 Books and tables
The course is designed to be fairly self-contained and does not follow any one textbook. However,
you may find the following useful for background reading:
• S. J. Garrett An Introduction to the Mathematics of Finance (Butterworth-Heinemann);

– cited in these notes as [Gar13].
• J. J. McCutcheon & W. F. Scott An Introduction to the Mathematics of Finance (Butterworth-
Heinemann);
– cited in these notes as [MS86].
• D. C. M. Dickson, Mary R. Hardy & Howard R. Waters Actuarial Mathematics for Life Con-
tingent Risks (Cambridge University Press);
– cited in these notes as [DHW13].
• A. Neill Life Contingencies (Heinemann);

– cited in these notes as [Nei77].
• N. L. Bowers, H. U. Gerber, J. C. Hickman, D. A. Jones and C. J. Nesbitt Actuarial Mathe-
matics (Society of Actuaries);
– cited in these notes as [BGH+ 97].
0.6 Acknowledgements
These notes have been produced by Jim Webber from notes for the predecessor module MTH6100
Actuarial Mathematics, a module which covered a very similar syllabus. Great credit and thanks to
Dr Rosemary Harris for her excellent work in producing the original draft of the MTH6100 notes.
Also thanks to Dr D. Stark and Dr W. Just for their additions and amendments to the notes since
the original draft. In her original introduction, Dr Harris gave credit to the previous notes of Prof.
B. Khoruzhenko and also the work of another previous lecturer, Dr L. Rass.
The changes I have made have been limited, but I take full responsibility for this edition of the
notes and the mistakes and contradictions that students will inevitably find. Please alert me to any
mistake that you find by sending an email to: a.baule@qmul.ac.uk.
7
8
Bibliography
[Ber89] J. Bernoulli. Tractatus de seriebus infinitis. manuscript, 1689.
[BGH+ 97] N. L. Bowers, H. U. Gerber, J. C. Hickman, D. A. Jones, and C. J. Nesbitt. Actuarial

Mathematics. Society of Actuaries, Schaumburg, 1997.
[DHW13] D. C. M. Dickson, M. R. Hardy, and H. R. Waters. Actuarial Mathematics for Life

Contingent Risks. Cambridge University Press, Cambridge, 2013.
[Gar13] S. J. Garrett. An Introduction to the Mathematics of Finance. Butterworth-Heinemann,

Oxford, 2013.
[Hal93] E. Halley. An estimate of the degrees of mortality of mankind, drawn from the curious
tables of the births and funerals at the city of Breslaw, with an attempt to ascertain the
price of annuities upon lives. Philosophical Transactions, 17:596–610, 1693.
[MS86] J. J. McCutcheon and W. F. Scott. An Introduction to the Mathematics of Finance.

Butterworth-Heinemann, Oxford, 1986.
[Nei77] A. Neill. Life Contingencies. Heinemann, London, 1977.
[Pac94] L. Pacioli. Summa de Arithmetica. Venice, 1494.
9
10
Chapter 1
Compound interest
The material in this Chapter is covered very well in ([Gar13]). It is also covered in Chapters 1–4 of
([MS86].)
1.1 Two types of interest

Often in the course of daily life and business, people need (or choose) to borrow money. For example,
you might have a student loan or, later in life, need a mortgage or business loan. On the other
hand, if you happen to have “spare” money you can lend it to a bank, for example, by investing in
a savings account or fixed term bond. In general, in all these situations the money lender receives a
kind of “reward” for lending the money; you can also think of this as the charging of “rent” for the
use of the money.
To be more specific the original loan/investment is called the capital (or principal) and the
“reward” to the lender/investor is the interest. The time-dependent value of the investment, i.e.,
the original loan plus the interest, is known as the accumulated amount (or accumulation).
Interest is expressed as a rate in two senses: per unit capital and per unit time. In practice,
the interest rate is often quoted in percent and usually, but not always the basic time unit is one
year. Note that “p.a.” is often used as an abbreviation for per annum (i.e., each year). To avoid
confusion you should always state the basic time period when giving an interest rate.
The interest rate on a transaction is affected by various factors including:
• the market rate for similar loans;
• the risk involved in the use to which the borrower puts the money (cf. mortgage loan rates
and unsecured personal loan rates);
• the anticipated rate of appreciation or depreciation in the value of the currency in which the
transaction is carried out (e.g., in times of high inflation the interest is higher).
For the present we assume that interest rates are constant in time and that there is no dependence
on the sum invested. We will later relax the first of these assumptions but the second will remain
throughout the course.
Now let’s consider a concrete example—a savings account with an interest rate of 5% p.a.. In
other words, one gets a return of £5 in one year for each £100 invested. Suppose you were to invest
£200 then you could close the account after one year and withdraw £210 made up of the principal
(£200) and the interest (£10). But what if the account were kept open for a period of time which
was longer (or shorter) than one year? In that case we would need to distinguish between simple
interest and compound interest.
11
1.1.1 Simple interest
For the bank account considered above then, in the case of simple interest, £10 would be added
each year to your original deposit of £200. In general, the accumulated amount is after one time
unit (n = 1)
Accumulation = P + iP = P (1 + i) (1.1)
After n time units you obtain interest of niP and thus
Accumulation = P + niP = P (1 + ni) (1.2)
where P = principal invested;
i = rate of interest;
n = duration of investment/loan.
Expression (1.2) applies for all non-negative values of n. The normal commercial practice in
relation to fractional periods of a year is to pay interest on a pro rata basis (i.e., proportional to the
time the account is open). For an account of duration of less than one year it is usual to allow for
the actual number of days the account is held.
What happens when n is greater than 1 year? Imagine investing the £200 of our example for
two years (again at 5% p.a. simple interest) then after these two years the accumulation will be
£200(1 + 2 × 0.05) = £220.
Is there a way to make more money?
Yes, you can close this account (Let’s call it account A.) after one year, at which time you will
withdraw £210 [see (1.2)]. Then place this sum on deposit in a new account, say B. When you
close account B after one further year, the sum withdrawn will be £210(1+1× 0.05)=£220.50. In
other words, you will have gained the princely sum of 50p!
The difference here is that, effectively, account B pays interest on the interest already earned.
In general, banks do not want people to be frequently opening and closing accounts. (Whilst you
might not bother to do that to gain 50p you probably would for £50,000!) This is one reason why,
for periods greater than one year, most bank accounts do not pay simple interest but instead the
interest is compounded...
1.1.2 Compound interest

In this case interest is paid on the previous interest accrued1 . After one time unit (n = 1) we have
the same result as for simple interest
Accumulation = P (1 + i). (1.3)
After two time units (n = 2) we also accumulate the interest on the interest accrued in the previous
time period
Accumulation = P (1 + i) + P (1 + i)i = P (1 + i)(1 + i) = P (1 + i)2 . (1.4)
The general formula for the accumulated amount thus becomes
Accumulation = P (1 + i)n . (1.5)
Another way to see this is that if An is the accumulation after n time units then An = An−1 (1+i) =
An−2 (1 + i)2 = . . . = A0 (1 + i)n , where A0 is the capital invested initially, i.e. P . This argument
obviously holds for integer n; the validity of (1.5) for non-integer n will be established in Section 1.3.
1
Here “interest accrued” simply means interest already earned / applied to the account.
12
Exercise 1.1.1: Effect of compound interest
Convince yourself that applying compound interest according to (1.5) has the same effect as closing an account
with simple interest after each time period and reinvesting the money in an account paying the same interest.
Exercise 1.1.2: Compound interest formula

Write a formal proof of (1.5) for integer n using induction.
From now on, “interest” always means compound interest unless explicitly stated otherwise.
However, simple interest calculations (which you should be able to do in your head!) can sometimes
be a useful tool for checking that compound interest results are in the right ballpark. Simple
interest is a good approximation for compound interest when n and i are small since in that case
(1 + i)n ≈ 1 + ni.
Example 1.1.1: Low interest rates
Suppose you invest £1,000 for three years in a bank account which pays interest annually at 1.5% p.a.. What
is the difference in accumulation if the bank pays compound interest compared to simple interest?
Solution
For simple interest then, using (1.2),
Accumulation = £1000 × (1 + 3 × 0.015) = £1045.00. (1.6)
On the other hand, if interest is compounded, then we use (1.5):
Accumulation = £1000 × (1 + 0.015)3 = £1045.68. (1.7)
[Unless stated otherwise, you should always give monetary answers rounded to the nearest penny.] So the
accumulation is only 68p greater if compound interest is paid.
In real-life there are many different kinds of transactions such as:

• payment of a series of premiums throughout a given time period in return for a lump sum at
the end of the period;
• mortgage loans, i.e., loans which are made for the specific purpose of house purchase (The
property to be purchased usually acts as security for the loan);
• fixed interest securities, such as regular income payments throughout a given time period and
an additional lump sum at the end of the period in return for a one-off payment.
Knowledge of compound interest can be useful in comparing the merits of such transactions.
Example 1.1.2: Investment advice
You have £10,000 to invest now and are being offered £22,500 after ten years as the return from the
investment. The market rate is 10% p.a. (compound interest). Ignoring complications such as the effect of
taxation, the reliability of the company offering the contract, etc., do you accept the investment?
Solution
Investing £10,000 for ten years with 10% compound interest will yield
Accumulation = £10000(1 + 0.1)10 = £25937. (1.8)
Since this is more than £22,500 you should reject the offered investment and just put the money in the bank.
Summary of 1.1
Accumulation of P units of money, simple interest: A = P (1 + ni)
Accumulation of P units of money, compound interest: A = P (1 + i)n
13
A13 A1 A2
0 i 0
1 2 3 t 1 2 3
(a) (b)
Figure 1.1: Schematic of investment and accumulation in two different scenarios. The PoC (1.10)
says that the accumulated amount at t3 is the same in both cases.
1.2 Nominal and effective interest rates

Now we generalize the set-up of the previous section and consider how to define interest rates when
the interest is compounded more frequently than annually. For example, many bank accounts pay
interest monthly or quarterly.
1.2.1 Accumulation factor

Let A(t1 , t2 ) be the accumulated value at time t2 of 1 unit of money invested at time t1 . A(t1 , t2 )
is called the accumulation factor and has the following useful properties:
1. A(t, t) = 1 (by definition)
2. Assuming that interest rates don’t depend on the size of the investment, then the accumulation
at t2 of an investment of P at t1 is P × A(t1 , t2 ).
3. In a “consistent” market we expect that the accumulation doesn’t depend on when, or how
often, money is withdrawn and re-invested. This assumption is called the principle of con-
sistency (PoC) and implies.
A(t1 , t3 ) = A(t1 , t2 )A(t2 , t3 ) ∀ t1 ≤ t2 ≤ t3 . (1.9)
You may find this easier to understand with the aid of the diagram in Fig. 1.1.
4. From (1.9), it follows that

A(t0 , t2 )
A(t1 , t2 ) = ∀ t0 ≤ t1 ≤ t2 . (1.10)
A(t0 , t1 )
Exercise 1.2.1:
Check Eq. (1.10).
Suppose the basic unit of time is one year and interest is compounded yearly at constant rate i.
Then we have A(0, 1) = (1 + i) or, more generally,
A(t, t + n) = (1 + i)n (1.11)
where n is (for now) an integer. This statement is just another way of writing (1.5).
Now consider what happens if interest is paid more frequently. In this case we can have accu-
mulations over fractional time periods. In general we say that interest is compounded p-thly (or
convertible p-thly) if it is paid p times in each unit time interval, i.e., paid after a “term” of length
1/p. For example, in the usual case where the basic unit of time is one year, then p = 4 means
interest paid quarterly, whereas p = 12 means interest paid monthly.
14
1.2.2 Nominal interest rates
A nominal interest rate is a rate, per unit time, of interest which applies over a different time
period. For example, “overnight money” means that a yearly rate of interest is applied daily (i.e.
interest is converted into capital daily, but interest is quoted as a rate per annum).
A nominal interest rate of i(p) per basic time unit is defined to mean that interest is compounded
p-thly with an interest rate of i(p) /p in a time interval of length 1/p. Equivalently, we have
i(p)

1
A t, t + =1+ . (1.12)
p p
For example, a nominal interest rate of 12% p.a. compounded monthly (p=12) means an interest
rate of 1% per month and therefore an accumulation factor A(t, t + 1/12) = 1.01.
Note on notation:
• i(p) does not mean i raised to the power p. The brackets are there to remind you that the p
here is just a label.
• In fact, a number in brackets to the top right of any actuarial symbol usually tells you about
the frequency of payments; we will see other examples later.
• In [MS86] you will find that i(p) is sometimes written as ih where h = 1/p; we will not use
this notation.
Nominal interest rates for different periods (i.e., terms of length 1/p for different p) are often
quoted in the financial press, for example, see the excerpt from the Financial Times presented in
Fig. 1.2.
Example 1.2.1: Nominal interest rates
Consider the nominal market interest rates (% p.a.) given in Fig. 1.2. Look at the line labelled “£ Libor”.2
Assuming these interest rates, find the accumulation of an investment of £10,000 on 7.12.16 after (a) one
quarter and (b) 1 day (overnight money).
Solution
Here the basic time unit is one year (interest rates are quoted p.a.) but we are considering interest paid after
fractions of this time.
(a) p = 4; from the table we have a nominal interest rate of 0.38363% per year which means i(4) = 0.0038363.
The accumulated amount is given by

1
Accumulation = P × A 0, (1.13)
4

0.0038363
= £10000 × 1 + [using (1.12)] (1.14)
4
= £10009.59 (to the nearest penny). (1.15)
(b) p = 365; i(365) = 0.0022625 and

0.0022625
Accumulation = £10000 × 1 + (1.16)
365
[Note that to get these answers correct to the nearest penny you need to keep all available decimal places for
the interest rates. Never round up until the end of a calculation!]
2
Libor stands for the “London Interbank Offered Rate” and is based on the interest demanded by banks in the
London wholesale money market when borrowing from each other.
15
Figure 1.2: Financial Times table of market interest rates on 7.12.2016.
These are nominal rates given as percentages per annum. Data courtesy of
http://markets.ft.com/ft/markets/researchArchive.asp.
16
1.2.3 Effective interest rates
The effective interest rate i is the total interest paid on one unit of money over one basic time
period. In other words, it is the interest rate converted from the nominal rate into the “equivalent”
rate if compounding were carried out at the end of one basic time unit (rather than p-thly).
By equivalent, we mean the rate which gives the same accumulation after unit time. Now, the
accumulation factor for one time unit with interest at rate i per unit time is just [see Eq. (1.11)]
A(0, 1) = (1 + i). (1.18)
On the other hand, for compounding p-thly we have

!p
i(p)

1 1 2 1
A 0, A , ...A 1 − ,1 = 1+ . (1.19)
p p p p p
So, if the accumulations are the same, we require

!p
i(p)
1+i= 1+ . (1.20)
p
This is a very important relationship between the effective interest rate i and the nominal interest rate
converted p-thly i(p) . You need to remember (1.20) or be able to reproduce quickly the argument
which gives it. Similarly, the rearranged formula
i(p) = p[(1 + i)1/p − 1], (1.21)
is often useful.
Note:
• i = i(1) .
• When the basic time period is 1 year, i is called the Effective Annual Rate (EAR) or the
Annual Equivalent Rate (AER).
• The AER is useful for comparing the annual cost of financial products with different periods
of compounding.
• Adverts for credit legally have to include the so-called Annual Percentage Rate (APR).3
This is defined as the effective annual rate of interest on a transaction obtained by taking into
account all the items entering the total charge for credit (i.e., including fees, etc.).
Example 1.2.2: AER
You wish to take out a loan and are offered two different deals. Bank A charges interest weekly at the nominal
rate of 11% per annum. Bank B charges interest quarterly at the rate of 3% per quarter. Calculate the AER
in each case and hence decide which bank offers the better deal.
Solution
Bank A: Setting the basic time unit as 1 year, we have p = 52 since there are (approximately) 52 weeks
in 1 year. We are given the nominal rate i(52) = 0.11 and therefore can use (1.20) to calculate the annual
equivalent rate:
52
0.11
i= 1+ −1 (1.22)
52
= 0.1161 (to 4 d.p.). (1.23)
3
Consumer Credit Act (1974)
17
Bank B: Here we are given that i(4) /4 = 0.03 (rate per quarter ) and, using again (1.20),
4
i = (1 + 0.03) − 1 (1.24)
= 0.1255 (to 4 d.p.). (1.25)
Notice that in both cases the effective interest rate is slightly higher than the nominal interest rate (due
to the compounding of interest). Bank A offers a lower AER (approximately 11.61% p.a.) than Bank B
(approximately 12.55% p.a.) so it is obviously the better choice for the loan.
An alternative strategy in problems which only involve one compounding time period (rather
than a comparison) is to set the basic time unit equal to the compounding period. This is usually
slightly quicker but, if you find it confusing, you may feel safer always setting unit time equal to one
year (i.e., always measuring time in years).
Example 1.2.3: Quarterly interest
If interest is paid quarterly at the rate of 2% per quarter, what is the accumulation of an investment of £500
after two years?
Solution
Let us set the basic time unit to be equal to one quarter. Then i(1) = i = 0.02 and (since there are 8 quarters
in two years) we have
Accumulation = P × (1 + i)n (1.26)

8
= £500 × (1 + 0.02) (1.27)
Exercise 1.2.2: Quarterly interest

Consider the same scenario as Example 1.2.3. Check that one gets the same answer by taking the basic time
unit to be equal to one year (so that i(4) /4 = 0.02); calculating i via (1.20) and then finding the accumulation
for n = 2.
Exercise 1.2.3: High APR

First Premier Bank offers a credit card with an “APR of 79.9%” (p.a. is assumed here). If the interest is
compounded monthly and there are no other charges/fees (i.e., the APR is just the annual effective interest
rate), determine how much a loan of $1000 increases to after just one month.
[Answer: $1050.15 (to the nearest cent)]
Summary of 1.2
For general rates:
1
i(p) (t)
Accumulation factor for interest converted p-thly: A t, t + p =1+ p
Only for time-independent rates: p
i(p)
Relation between nom. and effect. interest rates: 1+i= 1+ p
1.3 Force of interest

1.3.1 Time-dependent interest rates
So far we have assumed that interest rates are time-independent. In reality of course, interest
rates usually vary due to changing economic circumstances. For time-dependent rates we define a
time-dependent nominal rate i(p) (t), for transactions of term 1/p starting at time t, such that
i(p) (t)

1
A t, t + =1+ (1.29)
p p
18
or, on rearranging,
(p)
A(t, t + p1 ) − 1
i (t) = . (1.30)
1/p
Equation (1.29) means that if we know the nominal rates at some time t0 we can calculate the
accumulation for terms (of length 1/p) starting at time t0 but not for terms starting at any other
time. For example, the data in the table of Fig. 1.2 allow us to calculate the accumulations for time
periods starting on 7.12.16 just as was done in Example 1.2.1. However, since for time-dependent
rates the accumulation factors on the left-hand side of (1.19) are all different we can no longer
define an effective rate i via the simple relation (1.20).
What can we say in general about the function i(p) (t)? Inspection of data, such as that in
Fig. 1.2, suggest that i(p) (t) is usually a decreasing function of p. Generically, one also observes
that as p becomes very large i(p) (t) approaches a t-dependent limiting value (approximately 0.57
for the £ Libor data in Fig. 1.2). Motivated by this observation, in the next section we will see a
powerful general formalism for dealing with time-dependent rates.
Exercise 1.3.1: *Trend of i(p) as function of p

In the time-independent case, show from (1.21) that the nominal interest rate i(p) corresponding to constant
effective rate i decreases when the frequency of compounding p increases. [Hint: You may use, without proof,
the inequality e−x > 1 − x which holds for x > 0.]
Exercise 1.3.2: Limit of i(p)

In the time-independent case, prove from (1.21) that limp→∞ i(p) = ln(1 + i). [Hint: Use l’Hôpital’s rule.]
1.3.2 Force of interest

Following on from the remarks in the previous section about the trend of nominal interest rates, we
assume that, as p increases, i(p) (t) tends to a t-dependent limiting value and define
δ(t) = lim i(p) (t). (1.31)

p→∞
The quantity δ(t) is called the force of interest per unit time at time t. It can also be described
as the “instantaneous rate of interest per unit time at time t” or the “nominal rate of interest per
unit time at time t convertible momently”. (Here “convertible momently” means that interest is
compounded continuously.)
In many problems it is useful to treat δ(t) as the fundamental parameter and derive other
quantities from it. The following is a particularly important theorem.
Theorem 1.3.1:
If δ(t) and A(t0 , t) are continuous functions of t for t ≥ t0 , and the Principle of Consistency (PoC)
holds, then for t0 ≤ t1 ≤ t2 ,
Z t2
A(t1 , t2 ) = exp δ(s) ds . (1.32)
t1
Proof
We start by combining (1.31) and (1.30):
A(t, t + p1 ) − 1
δ(t) = lim . (1.33)
p→∞ 1/p
19
Now let h = 1/p, then
A(t, t + h) − 1
δ(t) = lim (1.34)
h→0+ h
A(t0 , t)A(t, t + h) − A(t0 , t)
= lim (1.35)
h→0+ hA(t0 , t)
1 A(t0 , t + h) − A(t0 , t)
= lim [using PoC]. (1.36)
A(t0 , t) h→0+ h
Next, for convenience, we let F (t) = A(t0 , t) (i.e., the value at t of one unit of money invested at
t0 ) and observe that (1.36) can be written as
1 F (t + h) − F (t)
δ(t) = lim . (1.37)
F (t) h→0+ h
Recognising here the derivative4 of F we have
1 dF (t)
δ(t) = (1.38)
F (t) dt
d
= [ln F (t)] [chain rule]. (1.39)
dt
Straightforward integration then yields
Z t
δ(s) ds = ln F (t) − ln F (t0 ) (1.40)
t0

F (t)
= ln , (1.41)
F (t0 )
and upon inverting both sides and using F (t0 ) = A(t0 , t0 ) = 1, we obtain
Z t
F (t) = A(t0 , t) = exp δ(s) ds , (1.42)
t0
which proves the theorem, since t0 , t are arbitrary.
Notice the appearance of the exponential function in this proof. In fact, the mathematical
constant e was first “discovered” by Jacob Bernoulli during his 1683 studies of interest compounded
continuously [Ber89].
By substituting (1.32) in (1.30), we obtain i(p) (t) in terms of δ(t)
hR i
t+1/p
exp t δ(s) ds − 1
i(p) (t) = , (1.43)
1/p
and the p = 1 case gives Z t+1

i(t) = exp δ(s) ds − 1. (1.44)
t
Example 1.3.1: Force of interest and accumulation factor

Assume that the force of interest varies with time and is given by δ(t) = a + bt . Find the formula for the
accumulation of one unit of money from time t1 to time t2 .
4
Strictly the right-sided derivative.
20
Solution
By Eq. (1.32),
Z t2
b
A(t1 , t2 ) = exp a+ ds (1.45)
t1 s
= exp [(at2 + b ln t2 ) − (at1 + b ln t1 )] (1.46)

t2
= exp a(t2 − t1 ) + b ln (1.47)
t1
b
t2
= exp [a(t2 − t1 )] . (1.48)
t1
Example 1.3.2: Force of interest and value of investment

1
Assume that the force of interest varies with time and is given by δ(t) = 0.02(1 + 1+t2 ) where t is measured
in years. Find the accumulation at time t = 1 of £1,000 invested at time t = 0.
Solution
Again using Eq. (1.32),
Accumulation = C × A(0, 1)
Z 1
1
= £1000 exp 0.02 1 + ds
0 1 + s2
n o
1
= £1000 exp 0.02 [s + arctan s]0
n π o
= £1000 exp 0.02 1 + − 0 − 0
4
= £1036.35 (to nearest penny).
Exercise 1.3.3: Force of interest and nominal interest rates

Assume that the force of interest varies as in Example 1.3.2. Use the relation (1.43) to find the value at
t = 0.5 of the nominal rate of interest per annum on transactions of term three months.
[Answer: 3.45% p.a. (to 2 d.p.)]
Exercise 1.3.4: *Stoodley’s formula

Stoodley’s formula is sometimes used to model gradually increasing or decreasing interest rates. It has the
form
s
δ(t) = p + (1.49)
1 + rest
where p, r and s are suitable parameters. Show that, if this formula holds, the accumulation of one unit of
money from time 0 to time t is given by
1+r
A(0, t) = exp[(p + s)t] . (1.50)
1 + rest
1.3.3 Special case of constant force of interest

In some situations (including many exam questions) a constant, i.e., time-independent force of
interest is assumed. For this special case δ(t) = δ for all t. Hence
A(t1 , t2 ) = eδ×(t2 −t1 ) (1.51)
which, in (1.43), leads straightforwardly to
i(p) = p(eδ/p − 1). (1.52)
The p = 1 case gives an important relationship between the effective rate i and the force of interest
i = eδ − 1 (1.53)
21
or, equivalently,
δ = ln(1 + i). (1.54)
Combining this last equation with (1.51) we obtain that, for any n,
A(t0 , t0 + n) = eδn (1.55)
n
= (1 + i) . (1.56)
In other words, if interest on fractional time periods is paid at the nominal rate corresponding to the
same effective rate i, then the compound interest formula (1.11) [or (1.5)] holds also for non-integer
n!
Example 1.3.3: Double your money!
If interest is compounded at an effective rate of 5% p.a., how long does it take an investment of £200 to
double in value?
Solution
We take the basic time unit to be 1 year and set i = 0.05. Hence the doubling time is n years where n
satisfies the equation
400 = 200(1 + 0.05)n . (1.57)
This simplifies to
2 = 1.05n , (1.58)
which is solved by taking logarithms of both sides to give
ln 2
n= = 14.21 (to 2 d.p.). (1.59)
ln 1.05
So the investment doubles in approximately 14.2 years or, to the nearest month, 14 years and 2 months.
[Note that the size of the original investment is irrelevant.]
Exercise 1.3.5: “Rule of 72”

A rough rule-of-thumb dating back to at least the 15th Century [Pac94] is that, if interest is compounded at
an effective rate of K% p.a., an investment takes approximately 72/K years to double. Check this against
the exact doubling time for interest rates of 4%, 8% and 12%. You should find that for these values the
relative error implied by the “Rule of 72” is less than 2%.
Example 1.3.4: Quadruple your money!

Find the force of interest if an investment of 2000 Icelandic Krona quadruples in value in 10 years.
Solution
From Eq. (1.51), the force of interest δ must satisfy
2000eδ×10 = 8000 ⇔ e10δ = 4. (1.60)
Taking logarithms then yields
1
δ= ln 4 = 0.1386 p.a. (to 4 d.p.). (1.61)
10
Exercise 1.3.6: Limit of i(p)

Starting from Eq. (1.52), show that limp→∞ i(p) = δ, i.e., the special case of (1.31) for time-independent
rates. [Hint: Compare Exercise 1.3.2.]
Exercise 1.3.7: *Approximate expressions for δ and i in terms of i(p) when p is large
Starting from Eqs. (1.54) and (1.20) use Taylor’s expansion of ln(1 + x), |x| < 1 (Calculus I) to obtain:
[i(p) ]2 [i(p) ]3
δ = i(p) − + ε, where |ε| ≤ (1.62)
2p 3p2
(p) 2 (p) 3
[i ] [i ] [i(p) ]4
= i(p) − + + ε, where |ε| ≤ . (1.63)
2p 3p2 4p3
22
Formula (1.62) can be used to give another proof of the fact that the force of interest is the nominal rate of
interest compounded instantly. Try to find this proof.
Summary of 1.3
Force of interest: δ(t) = lim i(p) (t) = lim A(t,t+h)−1

h
p→∞
R t2 h→0+
Accumulation at t2 of a unit invest. at t1 : A(t1 , t2 ) = exp t1 δ(s) ds
(p) p
h i
Constant force of interest δ: eδ = 1 + i = 1 + i p
1.4 Rates of discount

Let us consider the case of a time-varying force of interest δ(t) (i.e., for now we do not assume a
constant force of interest).
Assume that interest is compounded p-thly and recall, from Section 1.2, that the interest for
use of one unit of money over one subperiod of time of length 1/p starting at time t is i(p) (t)/p.
Note that it is assumed that this interest is paid in arrears at the end of the subperiod, i.e., at time
t + 1/p. So, restating our previous definition, we have:
i(p) (t) is the rate per unit time and per unit capital at which interest for use of money over time
period [t, t + 1/p] is paid in arrears (i.e., at t + 1/p).
However, in some circumstances one might pay for use of money in advance at the start of
the subperiod. The amount of the equivalent payment, for use of one unit of money, is denoted by
d(p) (t)/p where d(p) (t) is called the nominal rate of discount compounded p-thly (or convertible
p-thly) and defined such that:
d(p) (t) is the rate per unit time and per unit capital at which interest for use of money over time
period [t, t + 1/p] is paid in advance (i.e., at t).
When p = 1 we drop the subscript, i.e., d(t) is the rate at which interest for use of money over unit
time period is paid in advance.
We have defined A(t1 , t2 ) as the accumulation at time t2 of one unit of money invested at time
t1 . Analogously, we now define D(t1 , t2 ) as the value of an investment at time t1 which gives a
return of one unit of money at time t2 . You should be able to convince yourself easily that
1
D(t1 , t2 ) = (1.64)
A(t1 , t2 )
Z t2
= exp − δ(s) ds . (1.65)
t1
Now, when we borrow one unit of money for use over one subperiod and pay interest in advance,
we actually receive 1 − d(p) (t)/p at time t and repay 1 at time t + 1/p. Hence it follows that
d(p) (t)

1
D t, t + =1− . (1.66)
p p
Exercise 1.4.1: Rates of discount for differing subperiods

By working backwards through one time unit (and assuming the Principle of Consistency) show that
" #" #
d(4) ( 43 ) d(2) ( 41 ) d(4) (0)

1 − d(0) = 1 − 1− 1− . (1.67)
4 2 4
23
1.4.1 Relation between nominal rates of discount and interest
The nominal rates of interest and discount are related since payment of d(p) (t)/p at time t is
equivalent to payment of i(p) (t)/p at time t + 1/p. Or, in other words, an investment of d(p) (t)/p
at time t gives a return of i(p) (t)/p at time t + 1/p. Therefore
1 d(p) (t) i(p) (t)

A t, t + = . (1.68)
p p p
The validity of this equation can be easily seen as follows. Imagine that we have instead
1 d(p) (t) i(p) (t)

A t, t + > . (1.69)
p p p
Then it would be possible to obtain a risk-free profit: 1. Borrow 1 unit of money at time t with
(p)
interest paid in arrears. For this you need to pay i p(t) at time t + 1/p. 2. Loan this money to
(p)
another person with interest paid in advance. You receive d p(t) at time t. But now you can invest
this money in a bank account over the time period [t, t + 1/p], and if Eq. (1.69) holds you can
(p)
pay i p(t) and still make a profit. A similar reasoning can be made for < in Eq. (1.69) switching
borrowing and loaning. In a financial market, such a risk-free profit is not possible, therefore the
equality has to hold in Eq. (1.69).
Using Eq. (1.29), we obtain
i(p) (t)
d(p) (t) p
= (1.70)
p i(p) (t)
1+ p
Notice that this equation implies that the nominal rate of discount compounded p-thly is always
smaller than the nominal rate of interest compounded p-thly. Trivial rearrangement of (1.70) gives
1 1 1
= + (1.71)
d(p) (t) i(p) (t) p
which may be easier to remember. From this form of the relation it follows obviously that
lim d(p) (t) = lim i(p) (t) = δ(t) (1.72)
p→∞ p→∞
which is consistent with the intuition that for compounding continuously it makes no difference
whether we pay interest in advance or in arrears.
Exercise 1.4.2: Relation between nominal rates of discount and interest
Give an alternative derivation of (1.70) starting from (1.64).
Special case of constant force of interest

For the remainder of Section 1.4 we specialize again to the practically important case of constant
force of interest. By substituting i(p) (t) = i(p) = p(eδ/p − 1) [see (1.52)] in (1.70) we obtain that
d(p) (t) = d(p) where
eδ/p − 1
d(p) = p (1.73)
1 + eδ/p − 1
= p(1 − e−δ/p ). (1.74)
For p = 1 we have
i
d= = 1 − e−δ , (1.75)
1+i
where d is called the effective rate of discount (or, if unit time is one year, the annual rate of
discount).
24
Exercise 1.4.3: Relation between d and i
Interpret the first equality in (1.75) in words.
Relation between nominal and effective rates of discount

There are many ways to derive a relation between d and d(p) . One possibility is to start from (1.74)
which can be rearranged to give
d(p)
e−δ/p = 1 − (1.76)
p
which leads to !p
−δ d(p)
e = 1− (1.77)
p
and, using (1.75), !p
d(p)
1−d= 1− . (1.78)
p
Equation (1.78) should be compared with (1.20).
Example 1.4.1: Interest and discount
Suppose interest is paid at an AER of 3% p.a.. You take out a loan of £1,000 for 6 months. How much
interest should you pay if the interest is paid in arrears? And how much if the interest is paid in advance?
Solution
Take the basic time unit equal to 1 year, then i = 0.03 and
i(2) 2[(1 + i)1/2 − 1]

= [using (1.21)] (1.79)
2 2
= 0.14889 (to 5 d.p.) (1.80)
Now
i(2)
Interest paid in arrears = £1000 × (1.81)
2
= £14.89 (to nearest penny), (1.82)
and
d(2)
Interest paid in advance = £1000 × (1.83)
2
i(2)
2
= £1000 × (2)
[using (1.70)] (1.84)
1 + i2
= £14.67 (to nearest penny). (1.85)
Noticee that, as we expect, the interest paid in advance is slightly less than the interest paid in arrears; indeed
£14.67 = £14.89/(1 + i)1/2 .
Exercise 1.4.4: Key relationships

Look back through Sections 1.2–1.4 and combine equations to show that, for constant force of interest,
p
i(p)

1 1
1+ = 1 + i = eδ = =h ip . (1.86)
p 1−d 1− d
(p)
p
If you remember this series of equalities then, given any one of i(p) , i, δ, d(p) or d, you can find any of the
others by simple rearrangement.
25
Example 1.4.2: Conversion from d to i(12)
Given that d = 0.0625, find i(12) .
Solution
From Eq. (1.86), we have
p
i(p)

1
1+ = (1.87)
p 1−d
which can be rearranged to yield
1
i(p) = p − 1 . (1.88)
(1 − d)1/p
Hence, for d = 0.0625,
h i
i(12) = 12 (1 − 0.0625)−1/12 − 1 (1.89)
= 0.0647 (to 4 d.p.). (1.90)
[Note that it should not be necessary to memorize (1.88), or to calculate intermediate quantities such as i or
δ.]
Exercise 1.4.5: *Inequalities for rates

Show that
d = d(1) < d(2) < d(3) < . . . < d(p) < . . . < δ < . . . < i(p) < . . . < i(3) < i(2) < i(1) = i (1.91)
[Hint: Compare Exercise 1.3.1.]
Summary of 1.4
For general force of interest:
Rt
Discounted value at t1 of a unit invst. at time t2 : D(t1 , t2 ) = exp − t12 δ(s)ds
(p)
D t, t + p1 = 1 − d p(t)

When interest is converted p-thly:
1
Relation between nom. discount and interest rates: d(p) (t)
= i(p)1(t) + p1
Only for constant force of interest: h ip
d(p)
Relation between nom. and eff. discount rates: 1−d= 1− p
1.5 Discounting Cash Flows or Present Values

From Section 1.3, hwe know that
R t2 i the value at time t2 of an investment of C at time t1 is C ×
A(t1 , t2 ) = C exp t1 δ(s)ds . On the other hand, from Section 1.4, the value at time t1 of an
h R i
t
investment of C at time t2 is C × D(t1 , t2 ) = C exp − t12 δ(s)ds . This enables us to find the
value at any time of an investment whose value we know at any other time. In this section, we will
use this approach to compare the values of cash flows restricting the analysis to the case of constant
force of interest (with the exception of Exercises 1.8.1 and 1.8.1).
For constant force of interest we have,
A(t1 , t2 ) = (1 + i)(t2 −t1 ) and D(t1 , t2 ) = (1 + i)−(t2 −t1 ) . (1.92)
For brevity of writing we now define the discounting factor

1
v= = 1 − d = e−δ (1.93)
1+i
26
and note that with this notation A(t, 0) = D(0, t) = v t .
The quantity Cv t is called the (discounted) present value of C due at time t. We will abbreviate
this to P.V..
In practice this means that you must remember the following simple rules. Suppose we invest
an amount of C at some time. Then...
• ...to find the value of the investment t years later you multiply C by (1 + i)t ;
• ...to find the value of the investment t years earlier you multiply C by v t .
We will now use these rules to evaluate and compare the values of cash flows. Here a cash flow
means a series of payments. When considering cash flows, it is important to recognise the timing of
each cash flow; is the cashflow at the beginning of the month or the end of the month or, perhaps,
in the middle of the month?
1.5.1 Discrete cash flows

This is the most common case. Consider payments of c1 , c2 , . . . cn due at times t1 , t2 , . . . , tn in the
future. The present value 5 (i.e. at time t = 0) of such a cash flow is given by
n
X
P.V. of discrete cash flow = c1 v t1 + c2 v t2 + . . . + cn v tn = cj v tj . (1.94)
j=1
In many problems we are interested in comparing cash inflow and cash outflow. To compare two
cash flows we must look at their values at the same time; it’s usually best to choose the present time
and to compare P.V.s. Two cash flows are equivalent if they have the same P.V.. If the present
value of cash inflows is equal to the present value of cash outflows at a particular rate of interest,
it means that the outgoing cash flows when invested with interest will generate the incoming cash
flows. Equality of inflows and outflows is expressed by the so-called equation of value
P.V. of outgoing cash flows = P.V. of incoming cash flows. (1.95)
The equation of value can be expressed at any point of time. It brings together three quantities:
amount(s), time(s) and rate of interest. If the others are known, an unknown quantity from this list
can be determined by the equation of value.
Example 1.5.1: Equation of value

A borrower is under an obligation to repay a bank £6,280 in four years’ time, £8,460 in seven years’ time and
£7,350 in thirteen years’ time. As part of a review of his future commitments the borrower now offers either
of the following:
• to discharge his liability for these three debts by making an appropriate single payment five years from
now (offer A);
• to repay the total amount owed (i.e., £22,090) in a single payment at an appropriate time (offer B).
On the basis of a constant rate of interest 8% per annum effectively, find the appropriate single payment
if offer A is accepted by the bank, and the appropriate time to repay the entire indebtedness if offer B is
accepted.
5
Many authors use the terminology Net Present Value, abbreviated to NPV, to denote the present value of a cash
flow.
27
Solution
We take 1 year as the basic time unit so that i = 0.08 and
1 1
v= = . (1.96)
1+i 1.08
Offer A:
We need to find £C such that the following two cash flows are equivalent.
Out: £C at t = 5,
In: £6280 at t = 4, £8460 at t = 7, and £7350 at t = 13.
The equation of value expressed at the present time, t = 0, is
Cv 5 = 6280v 4 + 8460v 7 + 7350v 13 (1.97)
but in fact the equation of value expressed at t = 4 has a simpler form
Cv = 6280 + 8460v 3 + 7350v 9 . (1.98)
This is easily solved for C
6280 + 8460v 3 + 7350v 9

C= (1.99)
v
= 18006.46 (to 2 d.p.). (1.100)
So the borrower should make a single payment of £18,006.46 (to the nearest penny).
Offer B:
We need to find tp such that the following two cash flows are equivalent.
Out: £22090 at t = tp ,
In: £6280 at t = 4, £8460 at t = 7, and £7350 at t = 13.
In this case, the equation of value expressed at the present time, t = 0, is
22090v tp = 6280v 4 + 8460v 7 + 7350v 13 , (1.101)
and solving for tp yields

4
+8460v 7 +7350v 13
ln 6280v 22090
tp = (1.102)
ln v
= 7.66 (to 2 d.p.). (1.103)
So the borrower should make the payment after approximately 7 years and 8 months.
Summary of 1.5
1
Discounting factor: v = 1+i =1−d
Discounted present value of C due in time t: Cv t
Equation of value: P.V. of outflow = P.V. of inflow
1.6 Annuities-certain: introduction

An annuity is a series of payments made at regular time intervals. We restrict ourselves here mainly
to level annuities where the payments are all equal.
There are two sorts of annuities: annuities-certain and life annuities. For an annuity certain
the number of payments is certain and specified in the contract. In contrast, the payments may
28
depend on the survival of one or more human lives, then we say life annuity. In that case, the number
of payments is uncertain. For example, pensions are life annuities.
In this section we look at annuities-certain; life annuities will be considered later. We will
derive the P.V.s, for annuities-certain where one unit of money is paid per unit time. Obviously for
annuities where payment is C units of money per unit time, the P.V. is obtained by multiplying the
corresponding expression by C.
The main mathematical result we will need is the well-known formula for the sum of a geometric
progression:
N −1
X 1 − qN
q j = 1 + q + q 2 + . . . + q N −1 = , for any q 6= 1. (1.104)
1−q
j=0
Immediate annuity
Consider n payments of one unit of money to be made at intervals of one unit of time with the first
payment due one unit of time from now (i.e., the first payment is at time 1 and the last at time n).
This situation is known as an immediate annuity, the symbol for its present value is an and
1 − vn
an = v + v 2 + . . . + v n = v .
1−v
OR
1 − vn
an = multiplying numerator and denominator by (1 + i) and simplifying.
i
Note that the payments are made in arrears, i.e., at the end of each time period and that here, as
throughout this section, we assume a constant (and strictly positive) force of interest. In actuarial
notation, a subscript enclosed in a right angle always indicates the (fixed) term of the given financial
object.
Annuity-due
Now consider the same series of payments as in the previous paragraph but with payments made in
advance, i.e., at the beginning of each time period (so that the first payment is at time 0 and the
last at n − 1). This situation is known as an annuity-due, the symbol for its present value is än
and
1 − vn
än = 1 + v + . . . + v n−1 = . (1.105)
1−v
Notice that an = vän , and än = 1 + an−1 . An immediate annuity is also known as an annuity
paid in arrears.
Exercise 1.6.1: Relations between annuities

Interpret the equalities in the last sentence in words.
Example 1.6.1: Annuity-due

A loan of £300,000 is to be repaid by ten equal annual instalments, the first is due now. Find the annual
payment if the interest on the loan is charged at the AER of 12.99% p.a..
Solution
Let £C be the annual payment. Take 1 year as the unit time. Then i = 0.1299 and the discounting factor is
given by
1 1
v= = . (1.106)
1+i 1.1299
29
Now the loan is to be repaid by an annuity-due payable annually at rate £C per annum. The present
value of the repayments is thus Cä10 and equating P.V.s we obtain the equation of value:
300000 = Cä10 . (1.107)
Hence
300000 300000(1 − v)
C= = = 48911.2067 (to 4 d.p.). (1.108)
ä10 1 − v 10
So the annual repayment is £48, 911.21 (to the nearest penny).
Note, if there is likely to be any confusion about the value of i (and especially for questions involving
more than one interest rate), it is wise to state explicitly the interest rate implicit in a given function. For
example, one could write (1.107) in the form
300000 = Cä10 at 12.99%. (1.109)
An alternative notation is to use a vertical bar and an @ symbol. For example

300000(1 − v)
C= . (1.110)
1 − v 10
@i=0.1299
An annuity-due is also known as an annuity paid in advance.
Perpetuities
Notice that an and än are monotone increasing functions of n. Considering the limit of n → ∞
corresponds to payments being made “in perpetuity”. One finds
v 1
a∞ = lim an = = , (1.111)
n→∞ 1−v i
and
1 1
ä∞ = lim än = = , (1.112)
n→∞ 1−v d
which give the present values of an immediate perpetuity and a perpetuity-due respectively.
The income from equity share or from a property investment will often be valued ”in perpetuity”
using the approach above.
Deferred annuities
Suppose a series of n unit payments starts at time m + 1, the last one due at time m + n. This may
be considered as an immediate annuity deferred m time periods. The symbol for the present
value is m |an and
m+1
m |an = v + . . . + v m+n = v m an (1.113)
Similarly one can define an annuity-due deferred m time periods whose present value is m |än =
v m än .
30
Exercise 1.6.2: More relations between annuities
Show that m |an = am+n − am and m |än = äm+n − äm .
Exercise 1.6.3: Deferred annuity

John Doe wishes to purchase a deferred annuity of £10,000 per annum paid out for ten years. Payments are
made annually, the first payment being due in two years’ time. What is the present value of the annuity if
the annual effective interest rate stays at 10% over this period?
[Answer: £55,859.70 (to nearest penny)]
Increasing annuities
Annuities where the payments are not equal are called varying annuities. In particular, an annuity
which pays k units of money at the end of the kth time period (i.e., 1 unit of money at the end of
the first time period, 2 units of money at the end of the second time period, and so on, up to n
units of money at the end of the nth time period) is called an increasing immediate annuity. Its
present value is denoted (Ia)n and it can be shown that
än − nv n
(Ia)n = v + 2v 2 + 3v 3 + . . . + nv n = . (1.114)
i
Similarly, an annuity paying k units of money at the beginning of the kth time period for n time
periods is an increasing annuity-due and has present value
(Iä)n = 1 + 2v + 3v 2 + 3v 3 + . . . + nv n−1 = (1 + i)(Ia)n . (1.115)
Summary of 1.6
n
P.V. of annuity-due: än = 1−v
1−v
P.V. of immediate-annuity: an = vän
m = v m an
P.V.s of deferred annuities: m |än = v än , m |an
Note that the expressions for än and an are given on the exam formula sheet; the corresponding
deferred values should be obvious.
1.7 Annuities-certain: more variations

Annuities payable p-thly
Consider an annuity of one unit of money per unit time payable over n units of time in instalments
of 1/p at p-thly intervals. In the case where the payments are made in arrears (so that the first one is
due one p-thly interval from now at time 1/p), we have an immediate annuity payable p-thly. Its
(p)
present value is denoted by the symbol an . By summing the present values of the n × p individual
payments we obtain
(p) 1 1/p 1 2/p 1
an = v + v + . . . + v np/p (1.116)
p p p
1 1/p
= v 1 + v 1/p + . . . + v (np−1)/p (1.117)
p
1 1 − vn
= v 1/p [using (1.104) with q = v 1/p ]. (1.118)
p 1 − v 1/p
Obviously the p = 1 case reduces to the present value of an “ordinary” immediate annuity paid
annually, as given by (??).
31
Example 1.7.1: Mobile phone contract
Suppose you sign a mobile phone contract agreeing to pay £15 at the end of each month for the next two
years. Assuming a constant AER of 2% p.a., what is the present value of the whole series of payments? (In
other words, what lump sum of money would you need to put in the bank now to cover all the future monthly
payments?)
Solution
Take 1 year as the basic time unit, then i = 0.02 and
1 1
v= = . (1.119)
1+i 1.02
The payments form an immediate annuity payable monthly (p = 12) at the rate of £180 per year. Hence the
required P.V. is given by
(12)
P.V. = £180 × a2 (1.120)
2
1 1/12 1 − v
= £180 × v (1.121)
12 1 − v 1/12
As expected, this is slightly less than £360 because of the interest paid over the two years.
Exercise 1.7.1: *Mobile phone contract (alternative solution)

An alternative solution for Example 1.7.1 is to choose 1 month as the basic time unit. Then i (the effective
interest rate per time unit) is given by (1 + 0.02)1/12 − 1 = 0.00165158 (to 8 d.p.) and
1 1
v= ≈ . (1.123)
1+i 1.00165158
With 1 month as the basic time unit and v as given by (1.123) the required P.V. is
P.V. = £15 × a24 . (1.124)
Check that this gives the same answer as (1.122), and that you understand why. [To avoid numerical errors,
be careful not to round up the value for v given in (1.123).]
Now consider the same situation (an annuity of one unit of money per unit time payable p-thly
over n time units) but with the first payment due at time 0, i.e., payment in advance. In this case
we have an annuity-due payable p-thly whose P.V. is given by
(p) 1 1 1 − vn
än = 1 + v 1/p + . . . + v (np−1)/p = . (1.125)
p p 1 − v 1/p
Analogously to the discussion at the end of Section 1.6 one can also consider annuities payable
(p) (p)
p-thly deferred by m time units. It should be obvious that their P.V.s are given by m |än = v m än
(p) (p) (p) (p)
and m |an = v m an . In particular, note that 1/p |än = an . (Why?)
One can also derive alternative formulae for the P.V.s of annuities payable p-thly which illustrate
the relation to nominal rates of interest/discount. For example, start from (1.118) and use that
v 1/p = e−δ/p [follows from Eq. (1.93)] to obtain
(p) 1 −δ/p 1 − v n
an = e (1.126)
p 1 − e−δ/p
1 − vn
= (1.127)
p(eδ/p − 1)
1 − vn
= (p) [using (1.52)]. (1.128)
i
(p)
Exercise 1.7.2: Alternative formula for än
Show that
(p) 1 − vn
än = . (1.129)
d(p)
32
Annuities payable continuously
If payments are frequent they can be approximated by the theoretical construction of a continuous
annuity as we discuss below.
Suppose payments are made continuously over n units of time at the rate of one unit of money
per unit time. The present value of this payment stream is denoted by ān . To evaluate it, we first
imagining partitioning each time unit into small subintervals of length 1/p. Then
np
X k−1 k
ān = P.V. of payment in between t = and t = (1.130)
p p
k=1
If p is large, so that each subinterval is small, then we can approximate the continuous payment
between t = (k − 1)/p and t = k/p as a discrete payment of 1/p at t = k/p. Hence, for large p,
we have,
np
X 1 (p)
ān ≈ v k/p = an . (1.131)
p
k=1
(p)
The approximate relation ān ≈ an becomes exact in the limit p → ∞ which gives one way of
calculating ān :
(p)
ān = lim an (1.132)
p→∞
1 − vn
= lim [using (1.128)] (1.133)
p→∞ i(p)
1 − vn
= . [using the time indep. form of (1.31)]. (1.134)
δ
Exercise 1.7.3: p → ∞ of annuity-due

(p)
Show that taking limp→∞ än gives the same result as (1.134), and explain why (in words).
Exercise 1.7.4: *Another derivation of the P.V. of a continuous annuity

Consider a continuous cash flow between t1 = 0 and t2 = n with rate ρ = 1 per unit time. The present value
of such a payment stream is given by the integral in (1.151). Show that this gives the same expression for
ān as in (1.134).
Example 1.7.2: Deferred perpetuities

The AER is 8% p.a.. Find the present value of a deferred annuity of £13,000 per year paid in perpetuity in
the following cases:
(i) Payments are made monthly with the first payment in 6 months’ time.
(ii) Payments are made continuously in time starting in 6 months’ time.
Solution
We take the basic time unit to be one year so that i = 0.08 and
1
v= . (1.135)
1.08
(i) Here we have a perpetuity-due payable 12-thly and deferred by half a time unit (6 months). Remember
that the symbol a (with associated labels) always refers to an annuity of one unit of money per unit time so
33
that here the required P.V. (in pounds) is expressed as:
(12)
P.V. = 13000 × 1/2 |ä∞ (1.136)
1/2 (12)
= 13000v lim ä (1.137)
n→∞ n
1 1 − vn
= 13000v 1/2 lim (1.138)
n→∞ 12 1 − v 1/12
13000 v 1/2
= (1.139)
12 1 − v 1/12
= 163061.8827 (to 4 d.p.). (1.140)
So the present value of the specified perpetuity is £163,061.88 (to the nearest penny).6
(ii) Here we have a perpetuity payable continuously, again deferred by half a time unit (6 months). The
required P.V. (in pounds) is given by
P.V. = 13000 × 1/2 |ā∞ (1.141)

1/2
= 13000v lim ān (1.142)
n→∞
n
1−v
= 13000v 1/2 lim (1.143)
n→∞ δ
v 1/2
= 13000 (1.144)
δ
v 1/2
= 13000 (1.145)
ln(1 + i)
= 162540.1066 (to 4 d.p.). (1.146)
So the present value of the specified perpetuity is £162,540.11 (to the nearest penny).
Accumulated values
Until now we have discussed how to calculate the present value of an annuity at time 0. Suppose
that instead, we are interested in the accumulated value (accumulation) at the end of the series of
payments, i.e., at time n if it’s an annuity of term n. In fact, this is very simple to obtain from the
corresponding present value but does involve the introduction of yet another symbol...
To be specific consider an immediate annuity (of one unit of money per unit time) payable
p-thly over n time units. The accumulated value at time n, after the last payment has been made,
(p)
is usually denoted by sn . From the discussion at the beginning of Section 1.5 it follows that
(p) (p) (p) (p)
sn = (1 + i)n an , or an = v n s n . (1.147)
Of course, there is a similar relation between the present and accumulated values of an annuity-due,
(p) (p) (p) (p)
s̈n = (1 + i)n än , or än = v n s̈n , (1.148)
and between the present and accumulated values of a continuous annuity,
s̄n = (1 + i)n ān , or ān = v n s̄n , (1.149)
Don’t worry too much about remembering the s symbol. It’s much more important that you
understand that the accumulated value of an annuity-certain, of term n, is always given by the
associated present value multiplied by (1 + i)n .
6
Note that if you’ve specified that you’re working in pounds (as we did here above (1.136)), then you don’t need
to include the £ sign in the intermediate calculations but you should always give units in the final answer.
34
Summary of 1.7
h i
(p) 1 1−v n
P.V. of annuity-due payable p-thly: än = p 1−v 1/p
(p) (p)
P.V. of immediate-annuity payable p-thly: an = v 1/p än
n
P.V. of continuous annuity: ān = 1−v
δ
(p) (p) (p) (p)
Accumulated values of annuities: sn = (1 + i)n an , s̈n = (1 + i)n än
1.8 Continuous cash flows

Continuously payable cash flows are also known as payment streams. They are a theoretical concept
but sometimes a useful approximation, e.g., payments made daily or weekly can be considered
practically continuous if we are interested in times on a scale of years. The cash flow can consist of
both incoming and outgoing transactions and so it’s possible that the net cash flow ρ(t) is negative.
We define ρ(t) as the rate (per unit time) at which payment is made at time t, i.e., ρ(t) =
limh→0+ { h1 ×(amount paid over [t, t + h])}. The present value of such a continuous payment stream
over time interval [t1 , t2 ] in the future is then given by
Z t2
P.V. of continuous cash flow = v t ρ(t) dt. (1.150)
t1
The integration in the case where ρ(t) = ρ is particularly simple:

Z t2
P.V. of constant continuous cash flow = v t ρ dt (1.151)
t1
Z t2
= e−δt ρ dt (1.152)
t1
h ρ it2
= − e−δt (1.153)
δ t1
ρ −δt1
= (e − e−δt2 ). (1.154)
δ
Example 1.8.1: Linear cash flow
Consider a continuous cash flow having a linear payment density of the form
(
at for t1 ≤ t ≤ t2
ρ(t) = (1.155)
0 otherwise
with a a time-independent constant. Find the present value of this cash flow at time t = 0.
Solution
Z t2
P.V. = v t at dt (1.156)
t1
Z t2
=a e−δt t dt (1.157)
t1
( t2 Z t2 )
1 −δt 1 −δt
=a t − e − − e dt [integration by parts] (1.158)
δ t1 t1 δ
( t2 )
−δte−δt − e−δt
=a (1.159)
δ2 t1
a{(1 + δt1 )e−δt1 − (1 + δt2 )e−δt2 }

= . (1.160)
δ2
35
1.8.1 Continuous cash flow with variable force of interest
In fact, it is not very hard to extend this approach to treat also the case of a time-dependent force
of interest δ(t). Consider a cash flow at rate ρ(t) units of money per unit time.
Let us revisit (1.65). Since
1
× A(0, t) = 1,
A(0, t)
the value at time 0 of 1 unit of money received at time t is
Z t
1
D(0, t) = = exp − δ(s)ds .
A(0, t) 0
The present value of the cash flow ρ(t) is therefore

Z t2 Z t2 Z t
ρ(t)D(t1 , t)dt = ρ(t) exp − δ(s)ds dt. (1.161)
t1 t1 0
Exercise 1.8.1: *P.V.s with piecewise constant force of interest

Suppose that the force of interest at time t years is given by
(
0.05 for t < 6
δ(t) = (1.162)
0.07 for t ≥ 6.
Use (1.161) to determine the present value of a continuous payment stream at a (constant) rate of £1000
p.a. for 10 years beginning at time 0.
[Answer: £7768.20]
Example 1.8.2: Increasing annuity payable continuously

Consider a continuous annuity which has a constant rate of payment k (per unit time) throughout the kth
time period. The present value of such an annuity is denoted by (Iā)n ; show that it is given by
än − nv n
(Iā)n = . (1.163)
δ
Solution
Obviously the total P.V. can be expressed as a sum over the P.V.s for each time period so we have
n
!
X Z k
t
(Iā)n = kv dt (1.164)
k=1 k−1
n
X k −δ(k−1)
= e − e−δk [using (1.154)] (1.165)
δ
k=1
Pn−1 −δj
Pn
j=0 (j + 1)e − k=1 ke−δk
= [using j = k − 1] (1.166)
δ
Pn−1 Pn−1
(k + 1)e−δk − k=0 ke−δk − ne−δn
= k=0 [relabelling j as k] (1.167)
δ
Pn−1 −δk
e − ne−δn
= k=0 (1.168)
δ
Pn−1 k
v − nv n
= k=0 (1.169)
δ
än − nv n
= . (1.170)
δ
36
Exercise 1.8.2: *Spot the difference!
Another form of increasing continuous annuity has a rate of payment t at time t (i.e., a linear payment
¯ n ; show that it is given by
density). The present value of this variation is denoted by (Iā)
¯ n = ān − nv n
(Iā) . (1.171)
δ
[Hint: Consider Example 1.8.1]
Accumulated value of a continuous cash flow

Suppose that for a continuous time cash flow payments begin at time t1 and end at time t2 and
assume time-dependent force of interest δ(t). The accumulated value of a continuous time cash
flow ρ(t) at the time t2 is
Z t2 Z t2 Z t2
ρ(t)A(t, t2 )dt = ρ(t) exp δ(s)ds dt.
t1 t1 t
1.9 Repayment of Loans

1.9.1 Schedule of payments
When a loan is repaid by a series of regular payments, these payments are an annuity. Payments will
normally be made in arrears so, to be more specific, they form an immediate annuity. A schedule
of payments details how much capital and interest is paid each time a payment is made and how
much of a loan remains outstanding.
If payments are annually in arrears for n years, and the loan is for C units of money, then the
annual repayment (often referred to as the premium) is found by equating P.V.s:
C = P an . (1.172)
Let us calculate the amount of the loan still outstanding after m years, i.e., after the mth payment
has just been made. A moment’s thought shows that the amount then outstanding is just the value
at time m of C paid at time 0 minus the total value at time m of the annual repayments made at
times 1,2,. . . ,m. In fact it is more convenient to calculate these values at time 0 (i.e., going back
m years) and then rescale by the accumulation factor (1 + i)m . In other words,
Amount outstanding at time m
= [(P.V. at time 0 of loan) − (P.V. at time 0 of first m payments)] × (1 + i)m (1.173)
m
= (C − P am )(1 + i) . (1.174)
The amount outstanding can also be calculated as the value at time m of the remaining repayments,
i.e.,
Amount outstanding at time m = P an−m . (1.175)
The expressions in (1.174) and (1.175) are equivalent. However, in practice they may give marginally
different answers because the premium P is always rounded to the nearest penny.
(p)
If payments are made p-thly one can write similar formulae in terms of an or, alternatively,
change the units of time, e.g., from years to months if p = 12 (cf. Exercise 1.7.1).
The difference between the amount outstanding at time m and at time m + 1 is
1 − v n−m 1 − v n−m−1
P an−m − P an−m−1 = Pv − Pv
1−v 1−v
v n−m−1 (1 − v)
= Pv
1−v
= P v n−m
37
which is less than P . The remainder equals
1 − v n−m 1 − v
P − P v n−m = P v ×
1−v v
= P an−m i,
which is interest on the amount outstanding at time m. This shows that each payment P consists
of an amount P an−m i which is interest on the amount outstanding and a remainder P − P an−m i
which is the reduction in principal.
The schedule of payments tabulates how much is outstanding at each time step and specifies
how much of each payment is interest and how much goes towards reducing the principal. It is easily
obtained by carrying out a series of simple calculations as illustrated by the following example.
Example 1.9.1: Schedule of payments
Determine the schedule of payments on a loan of £10,000 to be repaid by equal annual payments, at effective
interest rate 10% p.a., over five years with the first payment due in one year from now.
Solution
As usual, we take the basic time unit equal to 1 year (so that i = 0.10) and determine the annual premium
£P by equating P.V.s:
v(1 − v 5 ) 1
£10000 = £P a5 where a5 = , and v = . (1.176)
1−v 1.1
Hence
10000 10000(1 − v)
P = = = 2637.9748 (to 4 d.p.). (1.177)
a5 v(1 − v 5 )
So the annual premium is £2,637.97.
Then we complete the following table year-by-year.
Time Payment Interest paid Principal paid Amount outstanding

(in years) (in £) (in £) (in £) (in £)
0 10000.00
1 2637.97 1000.00 1637.97 8362.03
2 2637.97 836.20 1801.77 6560.26
3 2637.97 656.03 1981.94 4578.32
4 2637.97 457.83 2180.14 2398.18
5 2638.00 239.82 2398.18 0
At time 0 no payments are made and the amount outstanding is the original loan of £10,000. For years 1–4
we carry out the following steps:
1. Enter the annual premium (as calculated above) in the “Payment” column.
2. Calculate the interest paid by multiplying the amount outstanding in the previous year by i. For
example, at the end of Year 1 the interest paid is £10000 × 0.1 = £1000.
3. Calculate the principal paid by subtracting the interest paid from the premium. For example, at the
end of Year 1, the principal paid is £2637.97 − £1000 = £1637.97.
4. Calculate the amount outstanding at the end of the year by subtracting the principal paid from the
amount outstanding in the previous year. For example, at the end of Year 1, the amount outstanding
is £10000 − £1637.97 = £8362.03.
For the final year (here the 5th) one has to be careful. The premium paid is usually slightly different (because
the premium in previous years has been rounded to the nearest penny). To find the exact premium one
first calculates the interest paid and then adds it to the amount outstanding. In this example, the interest
paid at the end of Year 5 is £2398.18 × 0.1 = £239.82 (to the nearest penny) so the required premium is
£2398.18 + £239.82 = £2638.00.
38
Exercise 1.9.1: Schedule of payments
Compare the figures in the “Amount outstanding” column of the schedule of payments in Example 1.9.1 with
the values given by (1.174) and (1.175).
1.9.2 Consolidating loans

It is possible that a person is paying back two or more existing loans and makes an agreement that in
the future he will only need to make payments on one new loan. This is illustrated by the following
example.
Example 1.9.2: Consolidating loans

Joe Bloggs is repaying two loans. The first is a loan of £5,000 taken out exactly 3 years ago and being paid
back over a 10 year period by annual payments in arrears. An APR of 10% p.a. is being charged on this loan.
The second loan is for £10,000 taken out exactly 1 year ago and also being paid back over a 10 year period
by annual payments in arrears. An APR of 12% p.a. is being charged on this loan. There is no penalty for
repaying the loans early, and Joe takes out a secured loan to pay off the outstanding amount of the original
two loans. An APR of 6% p.a. is to be charged on this new loan. Find the repayment level required to pay
off the secured loan by equal annual payments in arrears for 10 years.
Solution
Throughout the question we set the basic time unit to be 1 year. The strategy is to find the outstanding
amounts on the first two loans; their sum gives the present value for the new loan so the new premium can
be calculated. [Since there are several different interest rates in this question, it is important to indicate at
each stage which is being applied.]
• 1st loan: Here i = 0.1 and so v = 1/1.1. Let the annual premium for this loan be £P1 , then the
equation of value (at the time the loan was taken out) gives
5000 = P1 a10 at 10%. (1.178)
Solving for P1 we obtain

5000(1 − v)
P1 = = 813.73 (to 2 d.p.). (1.179)
v(1 − v 10 ) @i=0.1
Using (1.175), the amount outstanding (in pounds) after 3 years is
v(1 − v 7 )

P1 a7 = 813.73 × = 3961.58 (to 2 d.p.). (1.180)
1 − v @i=0.1
[Using (1.174) gives 3961.55 outstanding.]

• 2nd loan: Here i = 0.12 and so v = 1/1.12. Let the annual premium for this loan be £P2 , then the
equation of value (at the time the loan was taken out) gives
10000 = P2 a10 at 12%. (1.181)
Solving for P2 we obtain

10000(1 − v)
P2 = = 1769.84 (to 2 d.p.). (1.182)
v(1 − v 10 ) @i=0.12
Using (1.175), the amount outstanding (in pounds) after 1 year is
v(1 − v 9 )

P2 a9 = 1769.84 × = 9430.15 (to 2 d.p.). (1.183)
1 − v @i=0.12
[Using (1.174) gives 9430.16 outstanding.]
39
• New loan: For the new loan i = 0.06 and so v = 1/1.06. Let the annual premium for this loan be £P .
The present value of the debt (in pounds) at the time this loan is taken out is 3961.58 + 9430.15 =
13391.73 so equating present values gives
13391.73 = P a10 at 6%. (1.184)
Solving for P we obtain

13391.73(1 − v)
P = = 1819.51 (to 2 d.p.). (1.185)
v(1 − v 10 ) @i=0.06
Hence the annual premium for the new loan is £1819.51.
Summary of 1.9
Repayment of loan (term n, paid in arrears):
Amount outstanding at time m: (C − P am )(1 + i)m = P an−m
(Schedule of payments tabulates how much is outstanding at each time step
and specifies how much of payment is interest/principal reduction.)
1.10 Investment project appraisal

We now consider discrete time cash flow streams (introduced in Section 1.5).
Payback periods
If the net cash flow changes sign only once with the change being from positive to negative (i.e.,
the outflows precede the inflows then the balance in the investor’s account changes from negative
to positive at a unique time t1 known as the discounted payback period (DPP).
To be precise, t1 is the smallest value of t such that the investor’s accumulation is positive, i.e.,
such that X
cs (1 + i)t−s ≥ 0. (1.186)
s≤t
For practical purposes, we observe that this equation is equivalent to a condition on the cumulative
discounted cash flow: X
cs v s ≥ 0. (1.187)
s≤t
We may think of the cash flow stream as an investment project. If the project is viable, i.e. if such a
t1 exists, then the accumulated profit when the project ends at time tn is the total accumulation
X X
cs (1 + i)tn −s = (1 + i)tn cs v s . (1.188)
s≤tn s≤tn
If one sets i = 0 in (1.186) (i.e., ignores interest) the resulting value of t1 is called the payback
period. It is the smallest value of t such that the investor’s accumulation is positive, i.e., such that
X
cs ≥ 0. (1.189)
s≤t
The payback period is a naive, and usually quite poor, approximation to the discounted payback
period.
The calculation of discounted payback period is best demonstrated by means of an example.
40
Example 1.10.1:
In return for an initial investment of $10,000 an investor will receive $3,500 at the end of the first year, $5,000
at the end of the second year and $1,500 at the end of each subsequent year. If the AER is 15%, determine
the discounted payback period and compare it with the payback period.
Solution
We construct a table as follows:
Time Cash flow Discounted cash flow Cumulative P discounted
t ct ct v t cash flow, s≤t cs v s
(in years) (in $) (in $) (in $)
0 -10000.00 -10000.00 -10000.00
1 3500.00 3043.48 -6956.52
2 5000.00 3780.72 -3175.80
3 1500.00 986.27 -2189.53
4 1500.00 857.63 -1331.90
5 1500.00 745.77 -586.13
6 1500.00 648.49 62.36
The discounted payback period is the smallest time for which the value of the cumulative discounted cash
flow, shown in the rightmost column, is positive. So here the discounted payback period is 6 years, whereas
the payback period (without discounting) is 3 years, because −10000 + 3500 + 5000 < 0 and −10000 +
3500 + 5000 + 1500 ≥ 0. If the project terminates after 6 years, then the accumulated profit at that time is
62.36 × (1.15)6 = 144.24 (in $).
The next example does not use a table.

Example 1.10.2:
Based on a question from the CT1 Exam for April 2013.
A car manufacturer is to develop a new model to be produced from 1 January 2018 to an indefinite time
in the future.
• The development costs will be £19 million on 1 January 2016, £9 million on 1 July 2016, and £5
million on 1 January 2017.
• It is assumed that 6,000 cars will be produced each year from 2018 onwards and that all will be sold.
• The production cost per car will be £9,500 during 2018 and will increase by 4% each year with the
first increase occurring in 2019. All production costs are assumed to be incurred at the beginning of
each calendar year.
• The sale price of each car will be £12,600 during 2018 and will also increase by 4% each year with the
first increase occurring in 2019. All revenue from sales is assumed to be received at the end of each
calendar year.
Given this information
(i) Calculate the discounted payback period at an effective rate of interest of 9% per annum.
(ii) Without doing any further calculations, explain whether the discounted payback period would be greater
than, equal to, or less than the period calculated in part (i) if the effective rate of interest were
substantially less than 9% per annum.
Solution
(i) Let the discounted payback period from 1 January 2016 be n. Working in millions of pounds, considering
the project at the end of year n but before the outgo at the start of year n + 1: the present value of the
development cost is 19 + 9v 1/2 + 5v; the present value of the production cost is

(6)(9.5) v 2 + 1.04v 3 + · · · + (1.04)n−3 v n−1 ;
= 57v 2 (1 + 1.04v + · · · + (1.04v)n−3 )
1 − (1.04v)n−2
= 57v 2 ;
1 − 1.04v
41
the present value of the revenue from sales is

(6)(12.6) v 3 + 1.04v 4 + · · · + (1.04)n−3 v n ;
= 75.6v 3 (1 + 1.04v + · · · + (1.04v)n−3 )
1 − (1.04v)n−2
= 75.6v 3 ;
1 − 1.04v
The discounted payback period n is the smallest integer n such that
1 − (1.04v)n−2

−19 + 9v 1/2 + 5v + (75.6v 3 − 57v 2 ) ≥0
1 − 1.04v
or, using v = (1.09)−1 ,
n−2
1.04
≤ 0.85796,
1.09
log(0.85796)
n−2≥
log(1.04/1.09)
n ≥ 5.262.
Therefore the discounted payback period is n = 6 years.
(ii) The discounted payback period would be shorter using an effective rate of interest less than 9% per
annum This is because the income (in the form of car sales) does not commence until a few years have elapsed
whereas the bulk of the outgo occurs in the early years. Using a lower rate of interest has a greater effect
on the present value of the income than on the present value of the outgo (although both values increase).
Hence the discounted payback period becomes shorter.
Yield
The yield on an investment is the rate of interest at which the outgoing payments are equivalent to
the incoming ones. It is sometimes called the internal rate of return or money-weighted rate of
return.
Typically, one determines the yield by solving the equation of value (1.95) for the unknown
interest rate. The following example illustrates the idea.
Example 1.10.3:
In return for an initial investment of £100 an investor will receive £60 at the end of each of the next two
years. Find the yield for this transaction.
Solution
We need to find an effective interest rate i such that the following two cash flows are equal.
Out: £100 at t = 0,
In: £60 at t = 1, and £60 at t = 2.
The equation of value expressed at the present time, t = 0, is
100 = 60v + 60v 2 . (1.190)
Simplifying, we have
3v 2 + 3v − 5 = 0, (1.191)
and solving this quadratic equation for v yields
√
−3 ± 69
v= . (1.192)
6
Since i > −1, then v must be positive. Hence we take the positive root of (1.192) and finally obtain
1
i= − 1 = 0.1306623 (to 7 d.p.) (1.193)
v
So the yield on the transaction is 13.07% p.a. (to 2 d.p.).
42
Linear Interpolation for yield
When the yield cannot be solved for explicitly, numerical techniques such as linear interpolation
can be useful. If we know that P.V. is positive for i = i1 and that P.V. is negative for i = i2 , where
i1 < i2 , then the yield i∗ must lie between i1 and i2 . If the function P.V.(i) is approximately linear
in i, then
P.V.(i∗ ) − P.V.(i1 ) i∗ − i1
≈
P.V.(i2 ) − P.V.(i1 ) i2 − i1
Since P.V.(i∗ ) = 0, we have
P.V.(i1 )
i∗ ≈ i1 + (i2 − i1 ) × (1.194)
P.V.(i1 ) − P.V.(i2 )
Example 1.10.4:
Based on a question from the CT1 Exam for April 2013.
For an investment of £97, an investor will receive £6 at the ends of the first two years and £109 at the end
of the third year. What is the yield?
Solution
The present value of the cash flow stream is
P.V.(i) = −97 + 6(1 + i)−1 + 6(1 + i)−2 + 109(1 + i)−3 . (1.195)
We note that for i1 = 8%, P.V.(i1 ) = 0.227 and that for i2 = 9%, P.V.(i2 ) = −2.277. Plugging into (1.194)
shows that the yield is approximately
0.227
0.08 + 0.01 × = 0.08091 (1.196)
0.227 + 2.277
or 8.09% p.a. (The exact answer is 8.089% p.a.)
Existence of yield
In some cases the equation of value, with v as the unknown, may have no roots or more than one
positive root and then the yield does not exist. However, the following statements can be made:
• If all negative cash flows precede all positive cash flows (or vice versa) the yield is always well
defined, i.e., there is a unique positive root for v (corresponding to a unique solution with
i > −1).
• For a cash flow consisting of amounts ct0 , ct1 , ctP
2 , . . . , ctn at times t0 , t1 , t2 , . . . , tn then
let us define the cumulative total amount Ai = ir=0 ctr . If A0 and An are both non-zero
and, excluding zero values, the sequence {A0 , A1 , . . . , An } contains precisely one change of
sign, then the equation of value has a unique positive solution for i (but there may be other
positive roots for v corresponding to −1 < i ≤ 0).
See pages 38–40 of [MS86] for further discussion of these important classes of transaction.
1.11 Fixed Interest Securities and Other Investments

We now turn to the different classes of assets that insurance companies, pension funds and individuals
invest in. For the purposes of this module, the classes considered are cash, fixed interest securities,
index-linked securities, equities and property. Individual investors may invest directly or through a
collective investment product offered by an investment management firm; examples of collective
investment products include ISAs (UK only), unit trusts, investment trusts and exchange traded
instruments. Large corporate investors are likely to hold most investments directly, using collective
investment vehicles only for more specialised investment holdings.
43
1.11.1 Fixed Interest Securities
Governments, local authorities and companies borrow longer term funds by selling bonds (or fixed
interest securities) carrying defined levels of interest payments and defined repayment dates. The
buyer of a bond is entitled to a fixed series of interest and capital repayments, but assumes the risk
that the borrower (sometimes termed the issuer) will be unable to make the contractual repayments
(eg perhaps because of the bankruptcy of a corporate issuer or due to the financial problems of a
sovereign borrower such as Greece in recent years). Government bonds sometimes have common
informal trading names such as gilts (UK), Treasuries (USA), Bunds (Germany) and JGBs (Japan).
Bonds issued by companies are known as corporate bonds.
Bonds issued by government and large companies are usually listed on a recognised stock ex-
change and may be freely traded between investors, meaning that prices of bonds fluctuate from
day to day according to fluctuations in market interest rates and changes in the perceived level of
risk attached to the repayment of interest and principle by the issuer of the bond. For example, at
the time of writing, many bonds issued by the Greek government trade at much lower prices than
equivalent Bunds reflecting continuing concerns over the ability of the Greek government to repay
debt.
It should be noted that many bonds are bought by investors such as insurance companies and
pension funds to match the liabilities of their business; these businesses often follow ”buy and hold”
strategies with the result that trading levels can be low for many bonds. This means that market
liquidity (the ease with which significant quantities of the investment can be bought or sold without
materially changing the price) can be limited for many bonds apart from those issued by major
governments and the largest companies.
Bond Terminology
It is important to understand the terminology used to describe fixed interest securities:
• The nominal amount of a holding is used to define the maturity and interest payments. The
price of UK gilts is usually quoted per £100 nominal. Bonds are usually traded in integer
multiples of the nominal amount.
• The coupon rate defines the interest payable on the bond. The annual interest paid to an
investor is found by multiplying the coupon rate by the nominal amount of the bond holding.
• The redemption price is the amount repaid at maturity (or redemption date) per unit
nominal. If the redemption price is 1.00 then the bond is said to be redeemed at par.
• The running yield is the annual interest paid per £100 nominal divided by the market value
of the bond per £100 nominal; this is a measure of the immediate cash return to the investor.
• The gross redemption yield (or GRY) on a bond is the average annual pre-tax return to an
investor who buys a bond at the current market price and holds the bond to maturity. An
example of the calculation of GRY is given below.
• A zero coupon bond is a bond that consists only of a redemption payment; there are no
interest payments. A Treasury Bill (see section 1.11.2) is a simple example of a zero coupon
bond.
• An undated bond or perpetuity has no redemption date.
44
Example of a Government Security
The 2.75% Treasury Gilt 2024 is a typical conventional UK government security; note that the
repayment year and the annual coupon rate are referenced in the name of the the security. Key
features include:
• The gilt is bought or sold in units of £100 nominal, in common with most gilts.
• The gilt is repayable at par (ie £100 per £100 nominal) on 7 September 2024.
• The gilt was first sold by the Debt Management Office on behalf of the UK Treasury on 12
March 2014, so the initial term was around 10.5 years. The initial price was £98.40 per £100
nominal; ie a little below par.
• The gilt was issued on a number of other occasions up to January 2015; the total nominal
amount issued to date is £27 billion.
• In common with most gilts, interest payments are made twice yearly in arrears; in this case on
7 March and 7 September. The last coupon is paid on the date of repayment of the bond.
The coupon is £2.75 per £100 nominal meaning that the six monthly interest payment is
£1.375 per £100 nominal.
• At the time of writing (April 2017) the price of the gilt is £114.34 per £100 nominal, so the
bond is standing above par. The capital appreciation of the gilt since issue reflects how market
interest rates have fallen since the date of issue; the current gross redemption yield is 0.766%
pa compared to approximately 2.75% pa at the time of the first issue.
Types of Corporate Bonds

There are many different types of corporate bonds available; investors need to pay attention to the
precise features of each bond as these will influence the bond’s risk characteristics and its value to
an investor. Particular types to be aware of include:
• Debenture A debenture is a corporate bond offering some additional security to the investor
over and above the ability of the company to continue to trade profitably.
• Unsecured loan stock Unsecured loan stock offers no additional security to investors and
repayments of interest and capital depend on the ability of the issuing company to generate
profits to make repayments. In the event that the issuer becomes insolvent, investors would
rank alongside other unsecured creditors such as suppliers.
• Convertible bonds A convertible bond is a bond that is convertible into a different form of
security in certain circumstances. For example, since the 2007/2008 financial crisis, banks and
insurance companies have been encouraged to issue bonds which convert into equity shares if
the bank or insurance company is unable to meet certain solvency thresholds.
• Asset backed securities There are a wide range of bonds backed by underlying cashflows
due to a bank or other financial institution; for example the repayments due under mortgage
loans, credit card loans or student loans. Non financial cashflows can also be used to back
bond investments; examples range from usage charges on infrastructure projects, the emerging
profits of a book of life insurance policies to the royalties arising from David Bowie’s music.
The process of taking an illiquid or non-financial asset and structuring it into one or more
securities is often termed securitisation.
45
Valuation of a Bond
In relation to a corporate bond let us define:

N =the notional amount of bond that is used to define interest and repayment amounts (ie £100);
C = the annual coupon rate on the bond, payable twice yearly in arrears (ie 4%);
R = the repayment amount as a percentage of the notional amount of the bond (ie 110%); and
t = the term to maturity of the bond.
The price of the bond per N nominal at a gross redemption yield (or interest rate) of i p.a.
is given by the equation of value:
Price = Present value of future coupon payments + Present value of future capital repayments
(1.197)
(2) 1
= CN at + RN v t , where v = (1.198)
1+i
Example 1.11.1: Valuation of a Bond

On 5th April 2017, the 3.75% UK Corporate Bond 2024 is trading at a gross redemption yield of 1.4% p.a.
The repayment date is 5th September 2024 and repayment is at 110% of nominal. The coupon is paid
semi-annually on the anniversary of the repayment date. Calculate the price of the bond per £100 nominal.
Solution
Note that at the valuation date, the bond has 7 years and 5 months (or 153 days) to run until maturity, and
that the next coupon will be paid in 5 months time.
Then, by substituting in (1.198), and adjusting for the fact that coupon payments will be received in respect
of 7.5 years, the price per £100 nominal is:
1 (2)
Price per £100 nominal = 100(.0375)(1 + i) 12 a7.5 + 100(1.10)v 7.41667 at i = 1.40% pa
−7.5
1 1 − (1.014)
= 3.75(1.014) 12 + 110(1.014)−7.41667
2((1.014)0.5 − 1)
= 26.65 + 99.22
= £125.87 to the nearest penny.
Example 1.11.2: Calculating the Gross Redemption Yield

On 5th January 2017, the UK Treasury 8.75% 2017 gilt is trading at £108.50. The repayment date is 25th
August 2017 and the coupon is paid semi-annually. Calculate the gross redemption yield available to an
investor, assuming that repayment will be at par (ie at £100 per £100 nominal).
Solution
The price quoted is per £100 nominal. Note that the gilt is trading above par and the the investor will face
a capital loss at maturity. This reflects how interest rates have fallen since the gilt was issued at a price of
£97.38 on 30th April 1992.
The six monthly interest payment per £100 nominal is £4.375 (= 0.0875 × 100/2). Interest payments
will be on February 24th (50 days) and August 25th (232 days).
The gross redemption yield(GRY) to the investor is then i where i is the solution to the equation of value
developed from (1.198):
F (i) = 108.50 = 4.375(1 + i)−50/365 + (100 + 4.375)(1 + i)−232/365
The GRY i will be found by iteration. By inspection, you can see that the GRY is close to zero since
F (0) = 108.75. So, we will take i1 = 0.00 as our first estimate of i. We aim to find a second estimate of
the GRY, i2 , such that F (i2 ) < 108.50. For our next estimate we take i2 = 0.005, as using an estimate of
1.0 for i is likely to reduce the calculated price by almost £1 which would be too much.
F (0.005) = 108.42 which is less than 108.50 as desired.
46
The next iteration i3 is therefore estimated by interpolation as:

108.50 − F (i1 )
i3 = i1 + (i2 − i1 )
F (i2 ) − F (i1 )

108.50 − 108.75
= 0.00 + .005
108.42 − 108.75
= 0.0038 to 2 significant figures
Testing this new trial value, we find F (0.0038) = 108.50 to 2 decimal places. Therefore the GRY is
0.38% to 2 significant figures.
Summary of 1.11.1
h i
(2)
Value of Bond per N nominal = N Can + Rv n
C is the annual coupon rate paid twice yearly in arrears;
R is the repayment amount as a multiple of the nominal amount (frequently R will be 1.00);
n is the term to maturity of the bond in years; and
i is the gross redemption yield (or pre tax return) on the bond.
1.11.2 Cash including Treasury Bills

As an investment class, cash includes not only money held in current accounts with banks but
also a wider class of investments known as money market instruments. These are short term
investments(typically less than 1 year and usually much shorter)with low risk and high liquidity (ie
readily traded). The most common types of instrument are:
• Certificates of Deposit issued by banks and building societies for terms usually ranging from
28 days to 6 months. Interest is paid on the maturity of the contract.
• Commercial Paper issued by major companies to finance their short term borrowing needs. In
the UK, the term of commercial paper is limited to one year; while in the USA the limit is 270
days.
• Treasury Bills issued by governments. These bills do not pay interest, but are usually sold by
governments for less than the nominal (or ”face”) value. The investor earns a positive return
by purchasing a bill for less than its nominal value or ”less than par” and receiving the full
par value at redemption. At times of very low interest rates, Treasury Bills sometimes trade
at ”over par” meaning that the investor earns a negative return for holding this low risk cash
investment.
In the UK, the Government’s Debt Management Office (DMO) sells Treasury Bills on at least a
weekly basis through an auction process. UK Treasury Bills are usually issued for terms of 1, 3 or 6
months.
Example 1.11.3: Returns on Treasury Bills

A investor buys a 6 month (182 day) Treasury Bill for £98.75 per £100 nominal. Find the annual compound
return that the investor receives if the bill is held to maturity.
Solution
Return earned over 182 days per £100 nominal = £1.25 (= 100.00 − 98.75) or 1.2658% (= 1.25/98.75 ×
100%).
Annual compound return = (1 + 1.25/98.75)365/182 - 1 = 2.5548%
47
1.11.3 Inflation Linked Bonds and Real Returns
Inflation linked bonds or index-linked bonds are a type of bond for which interest and capital
repayments are fixed not in currency terms but relative to an index of inflation. If inflation is positive,
the bond holder receives an increasing stream of coupon payments and a repayment amount that
will be greater than the nominal value of the bond. The attraction to issuers is that the initial costs
of servicing (paying the coupons) an index-linked bond will usually be lower than for a fixed interest
bond; while for an investor an index-linked bond is a hedge against inflation risk and perhaps a
natural match to some types of liabilities. In the UK, most UK Index-Linked gilts are held by
pension funds and insurance companies to help match the inflation linked payments made by many
pension schemes.
Financial Calculations in Real Terms

To date, all calculations in this document have been defined in terms of known monetary amounts.
However, with inflation, the amount of goods and services purchased by a fixed amount of money
payable at some future date is unknown. In general, prices tend to rise as a result of inflation in the
economy; in recent years, central bankers have typically tried to target inflation rates in the range
of 2% to 3% per annum. This means that the ”purchasing power” of a fixed sum of money declines
over time.
Sometimes, financial calculations are made in ”real terms” rather than monetary terms; some-
times this is referred to as assuming constant prices.
The first step is to adjust monetary calculations to a common price point, using an inflation
index Q(t). For example, if Ct2 is a monetary cashflow payable at time t2 , then the ”real value” of
the cashflow in terms of prices at time t1 is given by:
Q(t1 )
Real value of cashflow in terms of t1 prices = Ct2 (1.199)
Q(t2 )
This is often termed the cashflow in terms of ”t1 money”.
If Ct2 is a future cashflow, and the inflation index after t1 is not yet determined, then we assume
a future rate of inflation of j% p.a. and write:
Q(t2 ) = Q(t1 )(1 + j)t2 −t1 for t2 > t1 . (1.200)
Sometimes, we will use a combination of known values and assumed future inflation to determine
the inflation index. For example, if ta is the last date for which the inflation index is known, then:
Q(t2 ) = Q(ta )(1 + j)t2 −ta where t1 < ta < t2 and
(1.201)
Q(t1 )
Real value of cashflow in terms of t1 prices = Ct2 from 1.200 (1.202)
Q(ta )(1 + j)t2 −ta
Calculating the Real Rate of Return

The internal rate of return (IRR) is the rate of interest at which the present value of all (positive and
negative) cashflows is zero. Similarly, the real rate of return is the rate of interest at which the
present value of all real cashflows (or ”constant price” cashflows) is zero. Given a series of monetary
cashflows Ct for t = t0 to tn and an inflation index Q(t), the real rate of return is i0 where:
tn
X Q(t0 )
Ct (1 + i0 )−(t−t0 ) = 0 (1.203)
t=t
Q(t)
0
48
If all future values of the inflation index are based on an assumed rate of future price inflation
j, then 1.203 becomes:
tn
X Q(t0 )
Ct (1 + i0 )−(t−t0 ) = 0 from 1.200
t=t0
Q(t0 )(1 + j)t−t0
tn
X 1
⇒ Ct =0
t=t0
[(1 + j)(1 + i0 )]t−t0
tn
X 1
⇒ Ct = 0 where i = (1 + j)(1 + i0 ) − 1.
t=t0
(1 + i)t−t0
This equation is the usual equation for the present value of a series of cashflows, and from this
we can derive the relationship between:
• j, the assumed rate of future inflation;
• i0 , the real rate of return; and
• i, the monetary rate of return (usually just termed the interest rate or yield on an investment)
1 + i = (1 + i0 )(1 + j); or (1.204)

0 0
i = i + j(1 + i ); or (1.205)
i−j
Real rate of return, i0 = (1.206)
1+j
Example 1.11.4: Real rate of Return
An investor buys a bond with a gross redemption yield of 4.5% p.a. If future price inflation is expected to be
0.5% p.a. above the central bank target rate of 2.5%, calculate the expected real return on the bond.
Solution
Expected future inflation,j = .025 + 0.005 = 0.030

i−j
Expected real return,i0 = from 1.206
1+j
.045 − 0.030
=
1.03
= 0.0146 to 3 significant figures.
The expected real return from the investment is therefore 1.46% p.a. Note that if expected future inflation
is low then i0 ≈ i − j.
Valuation of Index-Linked Bonds

The biggest issuer of index-linked bonds in the UK is the government; these are termed ”index-linked
gilts”. For example, the UK Government first issued the 1.25% Index-Linked Treasury Gilt 2017 on
8th February 2006. The gilt is repayable on 22 November 2017 and coupon payments are made half
yearly on 22nd May and 22nd November. Redemption payments per £100 nominal are given by the
formula:
Index(22.11.2017)
RedemptionP ayment = 100 × (1.207)
Index(8.2.2006)
49
Similarly, each coupon payment is indexed using the formula:
Index(coupondate)
CouponP ayment(coupondate) = 100 × 0.0125/2 × (1.208)
Index(8.2.2006)
where Index(date) is the value of the Retail Prices Index 3 months before date. This reflects that
RPI is determined in arrears and that for practical reasons all bond payments must be determined
from a known index that will be (at least fractionally) out of date. This is referred to a 3 month
deferment or time-lag (or just lag).
The RPI Index for November 2005 was 193.6 and the index had increased to 259.8 by August
2015 (3 months prior to the November 2015 payment date). The coupon paid in November 2015
per £100 nominal was £0.8387 (= 0.0125/2 × 100 × 259.8 193.6 ).
To calculate the real return on an index-linked bond it is necessary to make assumptions about
the rate of future inflation. The Financial Times publishes real yields for UK index-linked benchmark
gilts assuming future inflation of 0% per year and 5% per year.
Assuming no time-lag (or deferment) in the calculation of indexation, the price of £N nominal
of an index-linked bond at time zero at an effective monetary yield of i can be found by discounting
the expected monetary payments at the assumed monetary yield:
2n
X D Q(k/2) k/2 Q(n) n
Price,P = N vi + N R v (1.209)
2 Q(0) Q(0) i
k=0
where
• D is the nominal annual coupon rate paid twice per year in arrears;
• R in the nominal repayment amount per N nominal;
• n is the term to run in years; and
• Q(t) is the value of the price index at time t.
On the assumption, that there is no lag in indexation, then the price of an index-linked bond at
a real return of i0 is simply
2n
X D k/2
Price,P = N v 0 + N Rvin0 (1.210)
2 i
k=0
The valuation of the bond is based on nominal real amounts at the real rate of return.
Example 1.11.5: Valuation of an Index-Linked Bond
On 1 January 2013, a government issued a 5 year index-linked bond with a nominal coupon of 3% per annum
payable half yearly in arrears. The nominal redemption price is 100%. A non-tax paying investor buys £1000
nominal on 20 January 2017 and holds to maturity. Bond payments are indexed with a 3 month time lag.
Relevant values of the Retail Prices Index are given in the table below:
RPI 2012 2013 2016 2017 2018

Jan 144.1 150.90 170.0 175.6
April 145.0 151.40 171.4
July 147.4 167.6 172.7
Oct 149.2 169.4 173.8
1. Calculate the coupons and the redemption payment that the investor receives.
2. Calculate the price that the investor pays for the bond if the real return to the investor is 3.5% pa.
50
Solution 1. The investor receives the coupons paid on July 1 2017 and the final coupon and redemption
payment on January 1 2018. Due to the 3 month indexation lag, the relevant values of the Index are
the values for October 2012 (for indexation of the nominal values at issue) and the values in April and
October 2017.

.03 Index(Apr17)
Coupon received July 2017 = 1000 × (1.211)
2 Index(Oct12)

171.4
= 15 (1.212)
149.2
= £17.42 to the nearer penny. (1.213)
and similarly, the final coupon and redemption payment are found from:

Index(Oct17)
Final Coupon and Redemption Payment = (1000 + 15) (1.214)
Index(Oct12)

173.8
= 1015 (1.215)
149.2
= £1182.35 to the nearer penny. (1.216)
2. To determine the price the investor paid for the bond, we:
• convert the monetary cashflows into constant (or real) price terms, using the January 2017 RPI
value ; and
• discount the real cashflows using the investor’s real investment return of 3.5% and the usual bond
pricing formula (1.210).
Cashflows are converted to ”January 17” prices in the table below:
Date Monetary Amount in Real

Amount ”Jan 17” Money Cashflow
Index(Jan17)
July17 17.42 = 17.42× Index(July17)
17.42 × 170.0
172.5 = 17.16
Jan18 1182.35 =1182.35× Index(Jan17)
Index(Jan18)
1182.35 × 170.0
175.6 =1144.64
The first cashflow is paid after 162 days, while the redemption proceeds and final coupon are received
after 346 days. Therefore, the price of bond to give a real return of 3.5% pa is given, using (1.210),
by:
162 346
Price = 17.16(1.035)− 365 + 1144.64(1.035)− 365
= 1124.81 to the nearer penny.
Summary of 1.11.3
i−j
Real rate of return, i0 = 1+j
where i is the monetary rate of return and j is the expected rate of future inflation.
To determine, the real rate of return from a series of monetary cashflows, Ct , first
convert monetary cashflows to real (or constant price) cashflows, using the
appropriate Indexvalues, allowing for any specified time-lag in indexation, and
then determine the real rate of return following usual iterative approaches.
51
1.11.4 Equities
Equities represent the shares of companies. These may be large household names such as BP,
Barclays Bank or Tesco plc whose shares are traded on the London Stock Exchange(LSE) and are
constituent companies in the FTSE100 index, which is the most widely quoted UK stock market
index. Smaller, lesser known companies are also traded on the LSE and on its sister market AIM
(formerly known as the Alternative Investment Market).
Shares listed on markets can be readily bought and sold, but there are also thousands of small
private companies that are not listed. Institutions will also invest in private unlisted companies
with the intention of being a shareholder for several years before the company is sold or listed on a
recognised stock exchange; this form of investment is usually termed venture capital.
Characteristics of Ordinary Shares
When we talk of shareholders, we are usually referring to the holders of the ordinary shares in the
company. The key characteristics of ordinary shares include:
• The holders of the ordinary shares are the owners of the company; they have the responsibility
for appointing the directors; approving the accounts; and agreeing all major decisions (such
as a decision to sell the company).
• The shareholders benefit from the residual profits of the company, after bond holders and any
preference shareholders have been paid, and have to decide on the level of profit to be held
back by the company to fund future growth (”retained earnings”) and the amount of profit
to be distributed to shareholders as dividends.
• Ordinary shareholders receive a return on their investment in the ordinary shares of a company
through dividend payments and through increases (or falls) in the value of the shares.
• Ordinary shares are usually considered to be the class of investment providing the highest risk
and, potentially, the highest return to an investor.
• In the event of the insolvency of the company, ordinary shareholders will only receive money
after all other creditors including bond holders, have been paid. In this situation, shareholders
frequently receive nothing.
• Shares in large quoted companies are highly marketable with low transaction costs.
Valuation of Shares by Discounting Future Dividends
Valuing ordinary shares is a matter of art as well as science and a full treatment of the many different
approaches is beyond the scope of this module. We will however, apply our knowledge of compound
interest to look at the discounted dividend model of share valuation.
In the simplest form, we assume that dividends are payable annually in perpetuity. If Dt is the
dividend payable at time t, and the next dividend is payable in 1 year’s time, then the value of one
share at an interest rate of i is found by by discounting future cashflows:
∞
X
Value of share = Ct v t (1.217)
t=1
52
It is common to assume that dividends grow at a constant rate of g pa. Then if D is the last
dividend paid, we can write Dt = D(1 + g)t . Discounting future cashflows we get:
∞
X
Value of share = Ct v t (1.218)
t=1
X∞
= D(1 + g)t (1 + i)−t (1.219)
t=1
0 1+i
i
= Dā∞ where (1 + i0 ) = (1.220)
1+g
1
=D using formula for value of a perpetuity (1.221)
i0
(1 + g)
Value of share = D (1.222)
(i − g)
This is sometimes termed the dividend growth model. In practice, the valuation of future dividends
will depend on the precise timing on cashflows. Shares can be sold ”cum div” meaning with the
next dividend payment or ”ex div” meaning without the next payment. For practical reasons, shares
often trade ”ex div” several weeks before the dividend payment is actually paid.
Sometimes, investors will make more complex assumptions about the rate of future dividend
growth as in example below.
Example 1.11.6: Share Valuation
A pension fund, not subject to tax, is considering buying a share whose last annual dividend was 15p per
share. The fund expects dividends to grow by 10% pa for the next 10 years and to be able to sell the shares
after 10 years at a price that will give the purchaser a yield of 5% on the purchase price ( a ”dividend yield”).
The next dividend is due in one year’s time. What price should the fund pay to achieve a return of 8% pa on
their investment if the assumptions are achieved in practice?
Solution
The value of the share to the investor is the sum of the present value of the expected dividend payments over
10 years plus the present value of the share price after 10 years (since the plan is to sell the shares then).
10
X
PV(Dividends for 10 years) = D(1 + g)t v t
t=1
10
X
= 0.15(1.10)t (1.08)−t
t=1
1 − (1.10)10 (1.08)−10

−1
= 0.15(1.10)(1.08)
1 − 1.10(1.08)−1
= 1.66 to 2 decimal places
Dividend in 10 Year’s Time

Share Price after 10 Years =
Yield at Sale Price
0.15(1.10)10
=
.05
= 7.78
Value of Share = 1.66 + 7.78v 10

= 1.66 + 7.78(1.08)−10
= 5.26 to 2 decimal place
The value of the share is £5.26 on the assumptions given.
53
Characteristics of Preference Shares
A lesser known class of shares is preference shares. For any given company, preference shares will
be considered as a higher risk investment than a corporate bond, but a lower risk investment than
ordinary shares. The key characteristics of preference shares include:
• Preference shares carry a fixed rate of dividend, meaning that shareholders do not usually
benefit from the profitable growth of the business.
• Dividends can only be paid from profits after all interest costs have been met. Preference
shareholders must be paid before any dividends can be distributed to the ordinary shareholders
in a company.
• In the event that profits are inadequate to pay dividends on preference shares, it is common
for any unpaid dividend to be carried forward to future years. This type of share is termed a
cumulative preference share.
• In the event of the insolvency or winding up of the company, the holders of preference shares
will only receive a return of their capital after the amounts due to all other creditors (other
than ordinary shareholders) have been paid.
• Taking these factors together, it can be seen that the preference shares of a company are
likely to be lower risk than the ordinary (or equity) shares of a company, but higher risk than
corporate bonds issued by the same company.
This class of share introduces no new mathematical questions about valuation. It is important to
remember that in the absence of any further information, dividends on preference shares are non
increasing and are paid in perpetuity
Tax changes in the UK mean that this class of shares is relatively unattractive at present for the
issuing company; it is usually cheaper to raise finance through issuing corporate bonds.
1.11.5 Property
Institutional investors, such as insurance companies and pension funds, have traditionally invested in
UK office blocks, shops and industrial units. In recent years, the range of property investments has
expanded to include international property assets and residential accommodation that is available
for rent (including student accommodation). Most assets are held on a freehold basis, although
some might be on a long leasehold basis. Freehold is a legal term used in property; a freeholder is
the absolute owner of a property. A freeholder may grant an individual the right to use a property
for a period of time through agreeing a lease; this is a contract giving the leaseholder a right to
occupy the property for a period of time. For example, new flats in London are frequently sold on
the basis of 99 or 125 year leases; with the developer retaining the freehold.
Investors in property receive a return through the rental income on the property and from changes
in the capital value of the property.
As an asset class, property is usually positioned between corporate bonds and equities in terms
of risk and return. However, within the property class, there are great variances between different
types of investment holding. For example, a prime London office building would be sold on a low
yield in the expectation of future growth in rent levels. In contrast, properties in less successful areas
of the UK might be sold on much higher yields reflecting the poor prospects for rental growth and
the difficulty of re-letting the property should the tenant fail or simply not renew the lease.
Other characteristics of property include:
54
• A balanced property portfolio is likely to have a higher running yield than an equivalent equity
portfolio.
• Some properties may be let on long leases, with rent reviews every 3 or 5 years in accordance
with the terms of the lease.
• Some leases may specify that rent reviews are ”upwards only”. Despite the terms of lease-
hold agreements, in areas where the demand for property is weak (eg some shopping areas)
leaseholders have sometimes been able to negotiate lower rental levels.
• Property investments are typically large, ”lumpy” (meaning that a portfolio valued at say £100
million may consist of only a few properties) and expensive, and sometimes difficult, to buy
and sell. This means that property is a relatively illiquid asset class.
• The illiquidity and individuality of property investments make this a difficult asset class to value;
professional independent valuation on a regular basis is required adding to the management
costs of this asset class.
• Investors must remember the risk of rent ”voids”, when there is no tenant or income for a
property.
• Properties may require some repair or improvement in order to re-let the property when a
tenant vacates.
• All of the factors above mean that property has materially higher management costs than
other asset classes.
• The complexity and ”lumpiness” of property means that only the largest pension funds and
insurance companies will manage their own property portfolios; many will contract this role
out to specialist property investment firms or invest via pooled property funds.
No new mathematical techniques are required to value a property investment, or to determine the
IRR from an investment. However, the fact that rents may increase every 3 or 5 years can add
complexity to calculations as shown in the example below.
Example 1.11.7:
An investor is considering buying a property with a 25 year lease. The property is tenanted with a current
rental income of £220,000 per year payable quarterly in advance. Rents are reviewed every 5 years and the
next review is in 5 years time. Expenses of management are £1,000 per month payable monthly in arrears.
Determine the maximum price that the investor should offer to achieve a pre-tax return of 6% pa, assuming
that the rate of future rental increase will be 2% pa.
Solution
The maximum price is determined by discounting future expected cashflows at a rate of 6% per annum:
Maximum price = PV(Rent) − PV(Expenses) at a rate of 6% pa
h i
(4) (4) (4) (4) (4)
PV(Rent) = 220000 ä5 + v 5 (1.02)5 ä5 + v 10 (1.02)10 ä5 + v 15 (1.02)15 ä5 + v 20 (1.02)20 ä5
(4)
= 220000ä5 1 + v 5 (1.02)5 + v 10 (1.02)10 + v 15 (1.02)15 + v 20 (1.02)20

−25
1 − 1.06−5 1 − ( 1.06

1.02 )
= 220000
4(1 − 1.06−0.25 ) 1 − ( 1.06
1.02 )
−5
= 961218.93(3.530623)
= 3394000 to the nearest £1,000
The maximum price to be paid is therefore £3,394,000 to the nearest £1,000.
55
1.11.6 Taxation
The rules of taxation are complex and different for individuals, companies, insurance companies and
pension funds. Detailed tax rules and tax rates are set by nation states and vary significantly across
the globe, and are outside of the scope of this course. This section introduces the concepts of the
taxation of income and the taxation of capital gains and includes an example illustrating how tax
can be included into the evaluation of investments.
In earlier sections, we have discussed the yield obtained from a bond investment. Different terms
are used depnding on whether a pre-tax or a post-tax return is being calculated:
• The Gross Redemption Yield on a bond is the average return to maturity calculated on
a pre-tax or a gross basis. No allowance is made for taxation when calculating the gross
redemption yield.
• The Net Redemption Yield on a bond is the average return to maturity calculated on a post-
tax or a net(of tax) basis. The net redemption yield is calculated allowing for the payment
of tax on income and, if applicable, any tax paid on capital gains.
It should be noted that pension funds are not usually subject to tax on income or capital gains, so
the pension fund receives the full gross return.
Income Tax
Taxpayers, or a company, will often be subject to tax on regular sources of income, including:
• interest on bank accounts;
• coupons paid on government securities and corporate bonds;
• dividends paid on ordinary or preference shares; and
• rental income from properties after allowing for applicable management expenses. ( Manage-
ment expenses will usually be tax deductible.)
Example 1.11.8: Valuation of a Bond allowing for Income Tax
If an investor is subject to income tax at the rate t1 then the formula for valuation of a bond with a coupon
C payable twice-yearly [see (1.198) above] becomes:
(2)
Value of bond = (1 − t1 )CN at + RN v t (1.223)
Example 1.11.9: Calculation of after-tax Coupon

An investor pays tax on income at the highest marginal rate of 45%. The investor holds £100,000 nominal
of a corporate bond with a coupon rate of 4.5% per annum payable twice yearly. What is the net periodic
coupon received by the investor?
Solution
Periodic (six-monthly) coupon = .045 × 100000/2

= £2, 250.
Investor’s tax rate = 45%
Net (after-tax) periodic coupon = (1 − 0.45) × 2250

= 1237.50 to the nearer penny.
The net periodic coupon received by the investor is £1,237.50 .
56
Tax on Capital Gains
Investors who sell an asset for a higher price than they purchased it for make (or realise) a capital
gain. Similarly, investors who sell an asset for a lower price than they purchased it realise a capital
loss. Individual investors are often subject to tax on capital gains.
In the UK, individual investors usually pay no tax on capital gains on gilts and other bond
holdings, but gains on equity and property holdings are subject to tax. If an investor makes a capital
loss, then that loss can often be used to reduce the taxable gains generated on other investments.
No detailed knowledge of the complex rules regarding taxation of capital gains and losses is required
for this module. Students should, however, be able to adjust calculations for tax on capital gains as
instructed.
Example 1.11.10: Calculation of Tax on a Capital Gain

An investor buys a shareholding in a company for £100,000 and sells the holding after 4 years for £250,000.
Calculate the tax payable by the investor on the capital gain, assuming that gains are taxable at a rate of
28%.
Solution
Capital gain = Selling price − Purchase price

= 250000 − 100000
= £150, 000
Tax payable on capital gain = Tax rate applicable × Capital gain

= 0.28(150000)
= £42, 000.
The tax payable by the investor on the realised gain is £42,000.
Example 1.11.11: Including Income Tax and Tax on Capital Gains

An investor pays income tax at the rate of 40% on income including dividend income from shares, and is
liable to tax at a rate of 28% on capital gains. The investor has the option to buy 10% of a company for
£300,000. The estimated dividend payable on the shareholding is £1,000 and this is payable in one year’s
time. The company is forecast to grow rapidly and it is expected that dividends will grow at the rate of 20%
per annum for the next 5 years. The company plans to list on the stock exchange in 5 years time and, at this
time, the sale price of the investor’s shareholding is expected to be £1 million. The investor will only invest
if the expected after tax return from the investment is over 25% per annum. Determine if the investment
meets the investor’s investment criterion.
Solution
The investment meets the investor’s criterion if the discounted value of the expected after tax payments at
an interest rate of 25% per annum is greater than zero.
After tax dividend in year 1 = (1 − 0.40)1000

= £600
Expected sale proceeds less capital gains tax = 1000000 − 0.28(1000000 − 300000)
= £804, 000
57
NPV after-tax cashflows = PV(Post-tax dividends) − PV(Post-tax sale proceeds) − Initial investment
.202 .203 .204

1 1.20
= 600 + + + + + 804000(1.25)−5 − 300000
1.25 1.252 1.253 1.254 1.255
1 − ( 1.20 5

600 1.25 )
= + 263455 − 300000
1.25 1 − 1.20
1.25
= 2216 − 36545
= £ − 34329
The net present value of the expected cashflows at 25% per annum is less than zero, indicating that the
opportunity does not meet the investment criterion.
58
Chapter 2
Life tables and life-table functions
The best references for this material are: Chapters 2 and 3 of ([DHW13]).
2.1 Lifetime as a random variable

In this chapter we will consider a population of individuals with each individual referred to as a “life”.
We use X to denote the lifetime (exact age-at-death) of an individual and make the following two
assumptions:
• For each life, X is a continuous random variable, so that,
P (x < X ≤ x + h) ≈ fX (x)h for small h. (2.1)
• The lifetimes of different individuals are mutually independent.

The distribution of X is completely defined by knowledge of the probability density function
fX (x) but actuaries often work instead in terms of the survival function s(x) which is defined as
the probability of a newborn to attain age x. We have,
s(x) = P (X > x) (2.2)
= 1 − P (X ≤ x) (2.3)
= 1 − FX (x). (2.4)
Obviously s(0) = 1 and s(x) is monotone decreasing to 0. In practice, s(x) vanishes for x sufficiently
large.
So the probability distribution of X can be described in three equivalent ways:
• Cumulative distribution function (c.d.f.),
Z x
FX (x) = P (X ≤ x) = fX (u) du; (2.5)
−∞
• Probability density function (p.d.f.),

fX (x) = FX0 (x) = −s0 (x); (2.6)
• Survival function,
Z ∞
s(x) = P (X > x) = 1 − FX (x) = fX (u) du. (2.7)
x
Unless explicitly stated otherwise, we will assume in this chapter that the basic time unit is one year,
i.e., that all lifetimes are measured in years. Following standard actuarial convention, we use (x) to
denote a life of age x.
59
Future lifetime
Of particular interest is the the time-until-death for (x). This is a random variable, known as the
future lifetime (or the complete further lifetime) at age x, and denoted by the symbol T (x). As
is obvious from a diagram,
T (x) = X − x [for a life of age x]. (2.8)
However, to obtain the PDF of T (x) we have to take into account that the person was observed
alive at age x, i.e., X > x or T (x) > 0. We first derive the CDF with this condition
FT (x) (t) = P (T (x) ≤ t|T (x) > 0) (2.9)

= P (X ≤ x + t|X > x) (2.10)
P (x < X ≤ x + t)
= (2.11)
P (X > x)
s(x) − s(x + t)
= . (2.12)
s(x)
The PDF follows as the derivative of FT (x) (t) with respect to t:
s0 (x + t)
fT (x) (t) = − . (2.13)
s(x)
Exercise 2.1.1: Alternative derivation of p.d.f. of future lifetime

Express FT (x) (t) = P (T (x) ≤ t) in terms of s(x). Hence obtain an alternative derivation of the p.d.f.
fT (x) (t).
Example 2.1.1: Exponentially distributed lifetime

Suppose that
X ∼ Exp(λ). (2.14)
In other words, X has an exponential distribution with mean 1/λ and p.d.f.
(
λe−λx if x ≥ 0
fX (x) = (2.15)
0 if x < 0.
[Note that by convention one often only states the p.d.f. in the region where it is non-zero so (2.15) could
be written fX (x) = λe−λx , x ≥ 0.]
Find the survival function s(x) and hence the distribution of T (x).
Solution
The survival function is obtained by integrating the p.d.f.:
Z ∞
s(x) = fX (u) du (2.16)
x
Z ∞
= λe−λu du (2.17)
x
∞
= −e−λu x

(2.18)
−λx
=e . (2.19)
Also
s0 (x) = −fX (x) = −λe−λx . (2.20)
60
Armed with this knowledge it’s trivial to find the p.d.f. of T (x) via (2.13). For t ≥ 0 we have
− −λe−λ(x+t)

fT (x) (t) = (2.21)
e−λx
−λt
= λe . (2.22)
Notice that this is the same function as the p.d.f. of X itself, i.e., T (x) ∼ Exp(λ). Since the right-hand
side of (2.22) does not contain x, the distribution of future lifetime does not depend on age x. For example,
fT (1) (t) = fT (101) (t) so the p.d.f. is the same for a 1-year-old as for an 101-year-old. This is clearly unrealistic
so we conclude that an exponentially distributed lifetime is not appropriate for modelling the human population.
Exercise 2.1.2: Uniformly distributed lifetime

Suppose the age at death is equally likely to be any number between 0 and 100. Then
X ∼ Uniform[0, 100] (2.23)
and
1
fX (x) = , 0 ≤ x ≤ 100. (2.24)
100
Show that, in this case, T (x) ∼ Uniform[0, 100 − x].
In Section 2.4 we will discuss some slightly more realistic models for human lifetimes.
Summary of 2.1
Lifetime of individual as continuous random variable: X
Survival function: s(x) = P (X > x)
Future lifetime at age x: T (x)
0 (x+t)
P.d.f. of T (x): fT (x) (t) = −ss(x)
2.2 Basic life-table functions

Definitions
Here we list some important life-table functions and show how they can be expressed in terms of
the survival function s(x). The following expressions are defined for t and u positive.
• t px is the probability that (x) will survive to age x + t. It is given by
t px = P (T (x) > t|T (x) > 0) (2.25)

= P (X > x + t|X > x) (2.26)
P (X > x + t)
= [since X > x + t implies X > x] (2.27)
P (X > x)
s(x + t)
= . (2.28)
s(x)
When t = 1, the left-hand subscript is conventionally omitted so that
s(x + 1)
px = 1 px = (2.29)
s(x)
is the probability that (x) will survive to age x + 1.
61
• t qx is the probability that (x) will die by age x + t. It is given by
t qx = P (T (x) ≤ t) (2.30)
= P (X ≤ x + t|X > x) (2.31)
= 1 − P (X > x + t|X > x) (2.32)
= 1 − t px (2.33)
s(x + t)
=1− (2.34)
s(x)
s(x) − s(x + t)
= . (2.35)
s(x)
In the special case where t = 1 we have
s(x) − s(x + 1)
qx = 1 qx = . (2.36)
s(x)
• t|u qx is the probability that (x) will survive to age x + t but die by x + t + u. It is given by
t|u qx = P (t < T (x) ≤ t + u) (2.37)

= P (x + t < X ≤ x + t + u|X > x) (2.38)
P (x + t < X ≤ x + t + u)
= (2.39)
P (X > x)
FX (x + t + u) − FX (x + t)
= (2.40)
s(x)
s(x + t) − s(x + t + u)
= [recalling that s(x) = 1 − FX (x)]. (2.41)
s(x)
Exercise 2.2.1: s(x) in terms of px

Show that it follows from (2.28) that s(x) = px−1 px−2 . . . p1 p0 .
Exercise 2.2.2: Relations between life-table functions

Show that
t+u px = 1 − t+u qx = t px × u px+t , (2.42)
and
t|u qx = t px − t+u px = t px × u qx+t . (2.43)
Interpret these identities in words.
You should be able to calculate t px , t qx and t|u qx for any given s(x). In real-life we don’t know
s(x) [although various approximate forms have been proposed, see Section 2.4] which leads to the
use of life tables which tabulate estimated/“observed” values of life-table functions for exact ages,
x = 0, 1, 2, . . ..
Life tables
The first life table was published in 1693 by Edmond Halley (he of the comet fame). He based
his table on the register of births and deaths of the city of Breslau in Germany (now Wroclaw,
Poland) [Hal93]. In this course we will be using two life tables: English Life Table No. 17 (usually
referred to as ELT17) and (later) AMC00 a mortality table for Male assured lives.
62
ELT17 mortality tables were produced from the population reords for England and Wales for the
three year period 2010 to 2012. AMC00 tables were produced by the CMI Limited, a subsidiary of the
Institute and Faculty of Actuaries. The CMI produces tables reflecting the mortality of individuals
covered by the life assurance policies and pension plans of contributing life assurance companies. The
AMC00 table covers assured (customers with whole of life policies or endowments - see Chapter 3)
males combined (both smoker and non-smoker) over the 4 year period 1999 to 2002; from the same
series of tables, the TFN00 table represents the mortality of female non smokers with temporary life
assurance policies over the same period of time. Given that the mortality rates were measured over
a similar period or time which table do you think has the heavier (”higher”) mortality rates - ELT17
Males or AMC00?
Most life tables include the life-table functions lx , dx , px , qx , and e̊x .1 We have already met px
and qx ; the remaining functions will be introduced and discussed in the next few pages. Perhaps
the most useful, since many other life-table functions can be easily derived from it, is lx :
• lx is the expected number out of l0 newborns who survive to age x; l0 is called the radix of
the table.
In fact, life tables are not constructed by observing l0 newborns until the last survivor dies (that
would take far too long!). Instead, they are estimated from death rates across the whole population.
For example, for ELT12 the mortality of the entire male population of England in 1960–1962 was
used.
Relation between lx and s(x)

We can obtain an important relationship between lx and s(x) by considering the concept of a
random survivorship group as described below.
Let us take a group of l0 newborns whom we label by the index j, with j = 1, . . . , l0 . Let Xj
be the time-until-death for newborn j. We are interested in how many of the group will survive to
age x and we denote this random variable by N (x). Now
l0
1j ,
X
N (x) = (2.44)
j=1
where 1j is the indicator for the survival of newborn j, i.e.,

(
1 if newborn j survives to age x
1j = (2.45)
0 otherwise.
We assume that Xj for all j has a common distribution specified by survival function s(x). Then
1j ∼ Bernoulli(p) where p = P (X > x) = s(x) for all j and, as follows from the standard properties
of a Bernoulli distribution, E(1j ) = p = s(x).
Therefore we conclude that the expected number of survivors to age x from the group is given
by  
l0 l0
1j  = E(1j ) = l0 s(x).
X X
lx = E(N (x)) = E  (2.46)
j=1 j=1
In fact, under our assumption that all the lifetimes are mutually independent, N (x) ∼ Bin(l0 , s(x))
which provides an alternative route to calculate E(N (x)).
1
Note that the original version of ELT17 use the symbol ex for what we shall call e̊x ; the versions of ELT17 that are
used for MTH5124 are annotated to show the column-heading as e̊x , consistent with standard actuarial terminology.
63
The final result
lx = l0 s(x), (2.47)
is almost obvious—the expected number of individuals alive at age x is just the number of newborns
we started with, l0 , multiplied by the probability, s(x), that each individual survives to age x.
Basic life-table functions in terms of lx

Using (2.47) we can express other life-table functions in terms of lx . First we introduce the new
function t dx :
• t dx is the expected number out of l0 newborns dying between ages x and x + t. It follows that
t dx = l0 [s(x) − s(x + t)] = lx − lx+t . (2.48)
dx = 1 dx is the expected number dying between age x and x + 1:
dx = lx − lx+1 . (2.49)
We can also express t px , t qx and t|u qx in terms of lx :
s(x + t) lx+t
t px = = , (2.50)
s(x) lx
s(x) − s(x + t) lx − lx+t
t qx = = , (2.51)
s(x) lx
s(x + t) − s(x + t + u) lx+t − lx+t+u
t|u qx = = . (2.52)
s(x) lx
Exercise 2.2.3: p’s and q’s

lx −lx+1
For t = 1, (2.50) and (2.51) become px = lx+1
lx and qx = lx respectively. Check that these relations are
obeyed by the data in the relevant columns of ELT17.
The relationship lx+1 = px lx is particularly useful in problems requiring one to construct a toy
life table from knowledge of px (or equivalently qx = 1 − px ) for all integer x.
Example 2.2.1: Animal life table

Consider an animal population with a lifespan of 5 years (i.e. the animals live at most 5 years) where
p0 = 0.5 = P (T (0) > 1) = P (X > 1)
p1 = 0.4 = P (T (1) > 1) = P (X > 2|X > 1)
p2 = 0.3 = P (T (2) > 1) = P (X > 3|X > 2)
p3 = 0.2 = P (T (3) > 1) = P (X > 4|X > 3)
p4 = 0.1 = P (T (4) > 1) = P (X > 5|X > 4)
p5 =0 = P (T (5) > 1) = P (X > 6|X > 5).
Construct a life table with columns for lx , dx , px , and qx .
Solution
Set the radix: l0 = 10000, say. Then...
l1 = l0 p0 = 10000 × 0.5 = 5000, l2 = l1 p1 = 5000 × 0.4 = 2000, etc.;
d0 = l0 − l1 = 10000 − 5000 = 5000; d1 = l1 − l2 = 5000 − 2000 = 3000, etc.
Hence the required life table is as follows.
64
x lx dx px qx x
0 10000 5000 0.5 0.5 0
1 5000 3000 0.4 0.6 1
2 2000 1400 0.3 0.7 2
3 600 480 0.2 0.8 3
4 120 108 0.1 0.9 4
5 12 12 0 1 5
Example 2.2.2: American presidents

“In terms of governance, it’s a disaster. You do the actuary tables, there’s a one out of three chance McCain
doesn’t survive his first term and it’ll be President Palin...”
(quote from Actor Matt Damon during the 2008 US Presidential Election Campaign)
Assume that the lifetime distributions of US Presidents are the same as the general population of England
and Wales and governed by ELT17 Male.
(a) John McCain was 72 in August 2008, shortly before Matt Damon’s comments. What is the probability
that a man of exact age 72 dies before age 76? Was Matt Damon right?
(b) Barack Obama was 47 in August 2008. What is the probability that a man of exact age 47 survives to
age 51 but dies before age 55.
Solution
(a)
P (T (72) ≤ 4) = 4 q72 (2.53)
l72 − l76
= (2.54)
l72
76841 − 68299
= (2.55)
76841
= 0.1112 (to 4 d.p.). (2.56)
So, at least according to a (slightly old) English life table, the chance of President McCain dying during a
four-year presidency would only have been about 1 in 9. Of course, in McCain’s case there are many other
factors (health history, etc.) which Matt Damon may, or may not, have taken into consideration...
(b)
P (4 < T (47) ≤ 8) = 4|4 q47 (2.57)
l51 − l55
= (2.58)
l47
95338 − 93825
= (2.59)
96396
= 0.0157 (to 4 d.p.). (2.60)
So using ELT17, we would estimate the probability of Obama surviving his first term but dying during a
second term (assuming re-election) as about 1 in 100.
Exercise 2.2.4: American presidents (revisited)

Repeat the calculations of Example 2.2.2 with the older life table ELT12 [available on QM+ MTH5124]. How
do the probabilities change? Even better, look online for an American life table and try to interpret that.
Donald Trump was born in 1946; what is the probability of him surviving until the end of his second term of
offices in 2025?
Example 2.2.3: Random survivorship group

A population is subject to mortality described by ELT17 Male. A group of 9000 newborns were born in a
specific year. Estimate, to 4 significant figures, the number of survivors to age 60 from the group.
Solution
Since the lifetimes of the 9000 newborns are assumed independent, the number of survivors to age 60 has
65
distribution Bin(9000, 60 p0 ) and hence
Expected number of survivors to age 60 = 9000 × 60 p0 (2.61)
l60
= 9000 (2.62)
l0
l60
= 9000 (2.63)
100000
90975
= 9000 (2.64)
100000
= 8188 (to 4 s.f.). (2.65)
Exercise 2.2.5: Age-doubling survivors

Using ELT17 Males find the expected number of men who will survive to age 60 out of 100 who are aged 30
now.
[Answer: 92.23 (to 4 s.f.)]
Summary of 2.2
Relation between lx and s(x): lx = l0 s(x)
Other life-table functions in terms of lx : t dx = lx − lx+t , t px = lx+t
lx ,
lx −lx+t lx+t −lx+t+u
t qx = lx , t|u qx = lx
Hint: For analytical questions, write all life-table functions in terms of s(x); for questions involving
numerical data from tables, write all functions in terms of lx .
2.3 Force of mortality

The force of mortality µ(x) is defined as
s0 (x) d
µ(x) = − = − ln s(x). (2.66)
s(x) dx
It follows from the properties of s(x) that µ(x) ≥ 0. It can be interpreted as the instantaneous
death rate, per person and per unit time. To see this, consider the mortality of N (x) individuals
alive at age x:
Expected number who die between x and x + h = N (x) × h qx . (2.67)
From which it follows that,
N (x) h qx
Death rate per person per unit time = (2.68)
N (x)h
1 s(x) − s(x + h)
= [using (2.35)] (2.69)
h s(x)
1 s(x) − s(x + h)
= . (2.70)
s(x) h
The instantaneous death rate is then obtained by taking the limit h → 0+ :
1 s(x) − s(x + h)
Instantaneous death rate per person per unit time = lim (2.71)
h→0+ s(x) h
1 d
= × − s(x) (2.72)
s(x) dx
= µ(x). (2.73)
66
Note:
• µ(x) is a rate, not a probability, and therefore can be greater than one. Typically it is, for
example, in the first hours following birth.
• The probability for (x) to die before reaching age x + h can be approximated by µ(x) × h
for small h. For example, the probability for a newborn to die within 1 hour of birth is
1
approximately equal to µ(0) × 24×365 .
• There is a formal similarity between the force of mortality and the force of interest which will,
hopefully, become increasingly evident in the remainder of this section.
Relation between µ(x) and other functions

Integrating (2.66) gives
Z x
µ(u) du = − [ln s(u)]x0 (2.74)
0
= − ln s(x) [since s(0) = 1]. (2.75)
Then taking exponentials of both sides yields
Z x
s(x) = exp − µ(u) du . (2.76)
0
Exercise 2.3.1: Constant force of mortality

If µ(x) = λ for all x > 0, show that X has exponential distribution with mean 1/λ. [Hint: Find s(x) and
hence the p.d.f. fX (x).]
Now we can use the relationship (2.76) to express other life-table functions in terms of µ(x).
For example, using (2.28), we have
h R i
x+t
exp − 0 µ(u) du
t px = Rx (2.77)
exp − 0 µ(u) du
Z x+t
= exp − µ(u) du (2.78)
x
which has a similar form to the discounted value at x of an investment at x + t [cf. Eq. (1.65)].
Exercise 2.3.2: Life-table functions in terms of µ(x)
Use (2.76) to write other life-table functions in terms of the force of mortality.
We can also obtain an alternative expression for the p.d.f. of T (x). Starting from (2.13) we have
s0 (x + t)
fT (x) (t) = − (2.79)
s(x)
s(x + t) s0 (x + t)
= ×− (2.80)
s(x) s(x + t)
= t px µ(x + t). (2.81)
Thus for small h, the probability for (x) to die between x + t and x + t + h is approximately
t px µ(x + t) × h. From the normalization of the p.d.f. it follows that
Z ∞
t px µ(x + t) dt = 1. (2.82)
0
67
The curve of deaths
By definition, h dx is the expected number of deaths in the age interval between x and x + h in a
group of l0 newborns.
h dx = lx − lx+h (2.83)
= l0 s(x) − l0 s(x + h) (2.84)

d
≈ −l0 s(x) × h [for small h] (2.85)
dx

lx d
=− s(x) × h (2.86)
s(x) dx
= lx µ(x)h. (2.87)
Therefore, for small h, the expected number of deaths in the age interval (x, x + h] is approximately
lx µ(x) × h. Hence, lx µ(x) can be interpreted as the rate of (expected) deaths at age x per l0
newborns. For this reason, the plot of lx µ(x) against x, is often called the curve of deaths.
A plot of the discrete function dx shows a similar patten to the continuous function lx µ(x) and
this is also sometimes referred to as the curve of deaths.
Exercise 2.3.3: Curve of deaths for ELT17 Females
Sketch the curve of deaths corresponding to ELT17 Females and interpret its features.
Summary of 2.3
0
Force of mortality: µ(x) = − ss(x)(x)
Rx
(and hence s(x) = exp − 0 µ(u) du )
Alternative expression for p.d.f. of T (x): fT (x) (t) = t px µ(x + t)
2.4 Analytical laws of mortality

Over the past few centuries, there have been many attempts to describe the mortality experienced
by human or animal populations by a mortality law, i.e., by a mathematical expression for µ(x) or,
equivalently, s(x). A few famous examples are given below.
• de Moivre’s Law (1725):

s(x) = k(ω − x), 0 ≤ x ≤ ω. (2.88)
• Gompertz’s Law (1825):

µ(x) = bcx , x ≥ 0. (2.89)
• Makeham’s Law (1860)

µ(x) = a + bcx , x ≥ 0. (2.90)
• Makeham’s 2nd Law (1889):
µ(x) = a + hx + bcx , x ≥ 0. (2.91)
Perhaps the most famous law of mortality is that of Makeham from 1860. In this law, the term
a represents the mortality due to accident and the term bcx represents the mortality due to aging.
When c = 1 Makeham’s Law assumes a constant force of mortality and hence predicts exponentially
distributed lifetimes (see Exercise 2.3.1) which we have already seen is unrealistic.
68
When a small range of ages is concerned it is often possible to find values of a, b and c such that
Makeham’s Law fits the observed rates of mortality well. However, neither this nor any other law of
mortality represent the observed mortality of human populations over a large interval of ages. Indeed
it seems unlikely that there exists a universal law expressing the mortality of human populations by
way of a simple formula.
Example 2.4.1: Survival function for Makeham’s Law
Find the survival function s(x) corresponding to Makeham’s Law.
Solution
Using (2.76) we obtain
Z x
s(x) = exp − µ(u) du (2.92)
Z0 x
= exp − a + bcu du (2.93)
0
x
b u
= exp − au + c (2.94)
ln c 0
b
= exp {−ax − m(cx − 1)} , where m = . (2.95)
ln c
Exercise 2.4.1: Survival probability for Makeham’s Law

Show that, for Makeham’s Law,
x t

t px = exp −at − mc (c − 1) , (2.96)
where m is defined as in (2.95).
Exercise 2.4.2: de Moivre’s Law

Use the constraint that s(0) = 1 to determine the relationship between k and ω in de Moivre’s Law. Show
that this survival function corresponds to a uniform distribution of lifetimes (compare Exercise 2.1.2). Derive
an expression for the force of mortality. [Answer: µ(x) = 1/(ω − x) for 0 ≤ x ≤ ω]
Summary of 2.4
No need to remember these mortality laws...
...but you should be able to find s(x) given µ(x), or vice versa.
2.5 The expectation of life

The complete expectation of life
Recall that T (x) is the exact time-until-death for (x). T (x) is a continuous random variable with
p.d.f. fT (x) (t); its expectation is called the complete expectation of further life at age x and is
denoted by the symbol e̊x . Hence, by definition, we have
Z ∞
e̊x = E(T (x)) = t fT (x) (t) dt. (2.97)
0
Using (2.81), we can also write Z ∞

e̊x = t t px µ(x + t) dt. (2.98)
0
In fact, there is a much simpler expression, viz.
Z ∞
e̊x = t px dt, (2.99)
0
69
which holds under the assumption that xs(x) → 0 as x → ∞. This assumption is true for any
realistic survival function, (e.g., for humans s(x) = 0 for x > 120, say). To derive Eq. (2.99) we
use (2.13) and integration by parts. Specifically,
Z ∞
e̊x = t fT (x) (t) dt (2.100)
Z0 ∞
−s0 (x + t)
= t× dt (2.101)
0 s(x)
Z ∞
1
=− t s0 (x + t) dt (2.102)
s(x) 0
Z ∞
1 t=∞
=− [t s(x + t)]t=0 − s(x + t) dt [integration by parts] (2.103)
s(x) 0
Z ∞
s(x + t)
= dt [assuming xs(x) → 0 as x → ∞] (2.104)
0 s(x)
Z ∞
= t px dt [using (2.28)]. (2.105)
0
Exercise 2.5.1: Another expression for e̊x
Show that Z ∞
1
e̊x = lx+t dt. (2.106)
lx 0
Example 2.5.1: Exponentially distributed lifetime revisited

Find the complete expectation of life in the model of exponentially distributed lifetimes introduced in Exam-
ple 2.1.1.
Solution
For X ∼ Exp(λ), then we showed in Example 2.1.1 that
s(x) = e−λx , for x ≥ 0. (2.107)
Substituting this in (2.104) we find
Z ∞
s(x + t)
e̊x = dt (2.108)
0 s(x)
∞
e−λ(x+t)
Z
= dt (2.109)
e−λx
Z0 ∞
= e−λt dt (2.110)
0 ∞
1 −λt
= − e (2.111)
λ 0
1
= for x ≥ 0. (2.112)
λ
Note that the result is independent of x, which is clearly unrealistic!
Exercise 2.5.2: Uniform distributed lifetime revisited

Show that, for the model of uniformly distributed lifetimes introduced in Exercise 2.1.2,
100 − x
e̊x = for 0 ≤ x ≤ 100. (2.113)
2
The curtate expectation of life

T (x) is a continuous-type random variable. It expresses the exact time-until-death for (x). One
can also associate with the future lifetime of (x), a discrete-type random variable—the number of
70
further years completed by (x) prior to death which is denoted by K(x) and known as the curtate
further lifetime. Its probability mass function can be easily computed:
P (K(x) = k) = P (k ≤ T (x) < k + 1) (2.114)

= P (k < T (x) ≤ k + 1) [since T (x) is continuous] (2.115)
= k|1 qx (2.116)
s(x + k) − s(x + k + 1)
= . (2.117)
s(x)
The expected value of K(x) is called the expectation of curtate further life at age x and is
denoted by ex (i.e., without a ˚ symbol). Hence, by definition, we have
∞
X
ex = E(K(x)) = k P (K(x) = k). (2.118)
k=0
We can express ex in terms of the survival function as follows:

∞
X
ex = k P (K(x) = k) (2.119)
k=0
∞
X s(x + k) − s(x + k + 1)
= k (2.120)
s(x)
k=0
"∞ ∞
#
1 X X
= k s(x + k) − k s(x + k + 1) (2.121)
s(x)
k=0 k=0
 
∞ ∞
!
1  X X
= 0+ k s(x + k) − (j − 1) s(x + j) [using j = k + 1] (2.122)
s(x)
k=1 j=1
"∞ ∞
#
1 X X
= k s(x + k) − (k − 1) s(x + k) [relabelling j as k] (2.123)
s(x)
k=1 k=1
∞
1 X
= [k − (k − 1)] s(x + k) (2.124)
s(x)
k=1
∞
1 X
= s(x + k). (2.125)
s(x)
k=1
Alternatively, we can write

∞
X
ex = k px , (2.126)
k=1
which should be compared with (2.99).

Exercise 2.5.3: Another expression for ex
Show that
∞
1 X
ex = lx+k . (2.127)
lx
k=1
Example 2.5.2: Exponentially distributed lifetime revisited (again!)

Find the curtate expectation of life in the model of exponentially distributed lifetimes introduced in Exam-
ple 2.1.1.
71
Solution
Substituting for s(x) in (2.125) we find
∞
1 X
ex = e−λ(x+k) (2.128)
e−λx
k=1
∞
X
= e−λk (2.129)
k=1
∞
X
= e−λ e−λk (2.130)
k=0
−λ
e
= [sum of geometric progression] (2.131)
1 − e−λ
1
= λ . (2.132)
e −1
Exercise 2.5.4: Uniform distributed lifetime revisited (again)

Show that, for the model of uniformly distributed lifetimes introduced in Exercise 2.1.2,
99 − x
ex = for 0 ≤ x ≤ 99. (2.133)
2
Relation between complete and curtate expectations of life

By its definition, K(x) satisfies the inequalities K(x) ≤ T (x) < K(x)+1, hence taking expectations
E(K(x)) ≤ E(T (x)) < E(K(x)) + 1 and therefore, for all ages x,
ex ≤ e̊x < ex + 1. (2.134)
Although e̊x and ex are related, there is no explicit relationship expressing one in terms of the the
other.
However, there is a useful approximate relation:
1
e̊x ≈ ex + , (2.135)
2
which we will derive in the next section using linear interpolation. Observe that (2.135) holds exactly
for the special case of a uniformly distributed lifetime as considered in Exercises 2.5.2 and 2.5.4.
Summary of 2.5
Exact time-until-death for (x): T (x) R∞
1
Complete expectation of life: e̊x = E(T (x)) = s(x) 0 s(x + t) dt
No. of years completed by (x) prior to death: K(x)
1 P∞
Curtate expectation of life: ex = E(K(x)) = s(x) k=1 s(x + k)
1
Approximate relationship: e̊x ≈ ex + 2
2.6 Interpolation for fractional ages

Life tables contain estimates of lx [or equivalently s(x)] at exact ages x = 1, 2, 3, . . . (curtate values).
However, calculations of µ(x) and e̊x , and many practical applications, require knowledge of the
values of s(x) for all x, not just integer values. Hence various interpolation schemes have been
proposed to estimate the values of life-table functions for non-curtate values. For our purposes, we
will be particularly interested in the approximations resulting from linear interpolation.
72
Linear interpolation on s(x) and lx
If we know the values of the survival function only for integer x we can approximate it in the interval
between x and x + 1 by the linear function:
s(x + t) ≈ (1 − t)s(x) + ts(x + 1), for x = 0, 1, 2, . . ., 0 ≤ t ≤ 1. (2.136)
Exercise 2.6.1: Linear interpolation

Draw a sketch graph of s(x) against x and argue that the line segment joining the two point (x, s(x)) and
(x + 1, s(x + 1)) is given by (1 − t)s(x) + ts(x + 1), for 0 ≤ t ≤ 1.
Notice that (2.136) is equivalent to the approximation
lx+t ≈ (1 − t)lx + tlx+1 , for x = 0, 1, 2, . . ., 0 ≤ t ≤ 1. (2.137)
To estimate other life-table functions for fractional ages, we simply write the required function
in terms of lx [or s(x)] and apply the approximation (2.137) [or (2.136)]. An example illustrates the
method.
Example 2.6.1: Survival for quarter of a year
Use ELT17 Male to estimate the probability that a man of exact age 95 survives the next three months.
Solution
The required probability is given by
l95.25
0.25 p95 = (2.138)
l95
(1 − 0.25) × l95 + 0.25 × l96
≈ [using (2.137)] (2.139)
l95
0.75 × 6535 + 0.25 × 4854
= (2.140)
6535
= 0.9357 (to 4 d.p.). (2.141)
In fact, using (2.50) and (2.51), one finds the general results
t px ≈ 1 − t + tpx , (2.142)
t qx ≈ tqx , (2.143)
which again hold for x integer and 0 ≤ t ≤ 1. Equation (2.143) is an equivalent way of stating the
linear interpolation approximation and is known as the assumption of uniform distribution of deaths
within each year of age.
Exercise 2.6.2: Approximate forms for t px and t qx
Show that linear interpolation on s(x) leads to the approximations (2.142) and (2.143).
Example 2.6.2: Death within a month

Use ELT17 Male to estimate the probability that a man of exact age 100 dies in the next month.
Solution
Here one can simply use (2.143) to give
1 0.361872
1 q100 ≈ × q100 = = 0.0302 (to 4 d.p.). (2.144)
12 12 12
Exercise 2.6.3: Equivalent statements of linear interpolation

Complete the demonstration of equivalence by showing that (2.143) implies (2.137).
73
To demonstrate explicitly that linear interpolation on lx implies a uniform distribution of deaths,
let us consider the expected number out of l0 newborns dying between age x + t and age x + t + u.
Taking x as an integer, 0 ≤ t ≤ 1, and u small (such that 0 ≤ u ≤ 1 − t), we have
u dx+t = lx+t − lx+t+u (2.145)

≈ [(1 − t)lx + tlx+1 ] − [(1 − t − u)lx + (t + u)lx+1 ] (2.146)
= ulx − ulx+1 (2.147)
= udx . (2.148)
In other words, under assumption (2.137) [or equivalently (2.143)], the expected number of deaths
in the interval (x + t, x + t + u] (where x is an integer and 0 ≤ t ≤ 1) is independent of t and
proportional to the length of the interval u. This means, for example, that the expected number of
1
deaths between ages 100 and 100 12 is the same as the expected number of deaths between ages
11
100 12 and 101.
More generally, we can define R(x) as the fractional part of a year lived by (x) in the year of
death. R(x) is a continuous random variable taking values in the interval (0, 1). By definition we
have
T (x) = K(x) + R(x). (2.149)
Let us investigate the conditional c.d.f. of R(x) given K(x) = k:
FR(x)|K(x)=k (r) = P (R(x) ≤ r|K(x) = k) (2.150)

P ({R(x) ≤ r} ∩ {K(x) = k})
= . (2.151)
P (K(x) = k)
Now we evaluate the joint probability in the numerator under the assumption of a uniform distribution
of deaths:
P ({R(x) ≤ r} ∩ {K(x) = k}) = P (k < T (x) ≤ k + r) (2.152)

= k|r qx (2.153)
= k px r qx+k [see (2.43)] (2.154)
= k px qx+k × r [using (2.143)] (2.155)
= (k px − k+1 px ) × r [see (2.43) again] (2.156)
= P (K(x) = k) × r. (2.157)
This factorization of the joint probability already leads to the important conclusion that, under
the uniform distribution of deaths assumption, K(x) and R(x) are independent random variables.
Substituting (2.157) back into (2.151) we then obtain
FR(x)|K(x)=k (r) = r, for 0 < r < 1, (2.158)
which is just the c.d.f. of a uniform distribution.
Exercise 2.6.4: Mean of R(x)

If R(x) is uniformly distributed between 0 and 1, determine the mean E(R(x)).
[Answer: E(R(x)) = 21 ]
We shall return shortly to the consequences of this observation.
74
Derivation of relation between complete and curtate expectations of life
Armed with our knowledge of linear interpolation we can now derive the approximate relationship
e̊x ≈ ex + 12 , introduced
R ∞ in the last section. The crucial step is to use linear interpolation on lx to
estimate the integral 0 lx+t dt in terms of a sum of lx over integer x:
Z ∞ ∞ Z
X k+1
lx+t dt = lx+t dt (2.159)
0 k=0 k
X∞ Z 1
= lx+k+u du [substitution t = k + u] (2.160)
k=0 0
X∞ Z 1
≈ (1 − u)lx+k + ulx+k+1 du [linear interpolation] (2.161)
k=0 0
X∞ Z 1 Z 1
= lx+k (1 − u) du + lx+k+1 u du (2.162)
k=0 0 0
∞ ∞
1 X 1X
= lx+k + lx+k+1 (2.163)
2 2
k=0 k=0
∞ ∞
" #
1 X 1X
= lx + lx+k + lx+k (2.164)
2 2
k=1 k=1
∞
lx X
= + lx+k . (2.165)
2
k=1
Then, starting from (2.106), we have

Z ∞
1
e̊x = lx+t dt (2.166)
lx 0
∞
!
1 lx X
≈ + lx+k (2.167)
lx 2
k=1
∞
1 1 X
= + lx+k (2.168)
2 lx
k=1
1
= + ex . (2.169)
2
In fact, there is a much easier way to see this, if you are first prepared to prove (2.158). Observe
that
e̊x = E(T (x)) = E(K(x) + R(x)) = E(K(x)) + E(R(x)). (2.170)
By definition,
E(K(x)) = ex , (2.171)
and since, under the uniform distribution of deaths assumption, R(x) has a uniform distribution on
(0, 1),
1
E(R(x)) = . (2.172)
2
Hence it straightforwardly follows that
1
e̊x = ex + . (2.173)
2
75
In reality, of course, the uniform distribution of deaths assumption is only approximately satisfied so
that (2.173) only holds approximately.
Exercise 2.6.5: *Lifetime variances
Under the assumption of a uniform distribution of deaths show that
1
Var(T (x)) = Var(K(x)) + . (2.174)
12
Other interpolation schemes on s(x) and lx

Two non-linear interpolation schemes in common use in actuarial science are:2
• Exponential interpolation where s(x + t) is approximated as
ln s(x + t) ≈ (1 − t) ln s(x) + t ln s(x + 1), for x = 0, 1, 2, . . ., 0 ≤ t ≤ 1. (2.175)
This is consistent with the assumption of a constant force of mortality within each year of age.
Note that the above approximation leads to the following approximation:
t px ≈ (px )t for x = 0, 1, 2, . . ., 0 ≤ t ≤ 1. (2.176)
This approximation is generally used when undertaking monthly modelling of life insurance
contracts and is usually a more appropriate approximation at higher ages.
• Harmonic interpolation which is based on the approximation

1 1−t t
≈ + , for x = 0, 1, 2, . . ., 0 ≤ t ≤ 1. (2.177)
s(x + t) s(x) s(x + 1)
This is also known as the hyperbolic or Balducci assumption.
Example 2.6.3: Death within a month - exponential interpolation

Use ELT17 Male to estimate the probability that a man of exact age 100 dies in the next month assuming
exponential interpolation..
Solution
Here one can simply use (2.176) to give
1 1
1
12
q100 = 1 − 1
12
p100 =≈ (1 − q100 ) 12 = 1 − (1 − 0.361872) 12 = 0.0367 (to 4 d.p.). (2.178)
Note that the answer is approximately 20% higher than the answer found assuming linear inter-
polation in Example (2.6.2). Why is this?
Assumptions to obtain µ(x)

The force of mortality cannot be observed. Its values (even those for integer x given in life tables3 )
have to be calculated from the values of other life-table functions. Linear interpolation on s(x) is
not suitable for this purpose as it gives two different values of µ(x) for integer x, one when it is used
over the interval [x − 1, x] and the other when it is used over [x, x + 1]. To overcome this problem,
other approximations are usually used. The four most common are:
2
You are not expected to remember these but you might be asked to use them to obtain approximations for other
life-table functions.
3
Note that for integer x, µ(x) is often denoted as µx .
76
R1
1. Starting from px = e− 0 µ(x+t)dt
(which follows from (2.78) via a change of variable), we have
Z 1
1
− ln px = µ(x + t)dt ≈ µ x + , (2.179)
0 2
i.e., the approximation

1
µ x+ ≈ − ln px . (2.180)
2
R1
2. Starting from 2 px−1 = px−1 px = e− −1 µ(x+t)dt
, we have
Z 1
− ln(px−1 px ) = µ(x + t)dt ≈ 2µ(x), (2.181)
−1
leading to the approximation
1
µ(x) ≈ − (ln px + ln px−1 ) . (2.182)
2
3. Based on the assumption that ly is a quadratic polynomial in y in the interval [x − 1, x + 1],

one obtains
lx−1 − lx+1
µ(x) ≈ . (2.183)
2lx
4. Based on the assumption that ly is a quartic polynomial in y in the interval [x − 2, x + 2], one
obtains
8(lx−1 − lx+1 ) − (lx−2 − lx+2 )
µ(x) ≈ . (2.184)
12lx
You don’t need to know these approximations but you should be able to apply them.
Example 2.6.4: Quartic polynomial approximation

Estimate µ(40) using approximation 4. and the ELT17 values of lx for ages 38, 39, 40, 41, and 42.
Solution
Applying (2.184) we have
8(l39 − l41 ) − (l38 − l42 )

µ(40) ≈ (2.185)
12l40
8(97803 − 97524) − (97928 − 97371)
= (2.186)
12 × 97668
= 0.00143 (to 5 d.p.). (2.187)
The obtained value coincides with the value of µ40 in the table.
Summary of 2.6
Linear interpolation, for x integer and 0 ≤ t ≤ 1,: lx+t ≈ (1 − t)lx + tlx+1

or equivalently (uniform distribution of deaths): t qx ≈ tqx
Exponential interpolation, for x integer and 0 ≤ t ≤ 1,: px+t ≈ (1 − px )t
77
2.7 Select mortality
Let us consider two different ways to calculate the probability that a person of age x dies within the
next t years
s(x) − s(x + t)
(A) t qx = = P (x < X ≤ x + t|X > x);
s(x)
(B) t qx = 1 − t px = 1 − P (T (x) > t)
In (A) we evaluate the probability that (x) will survive to age x + t under the single hypothesis
that the newborn has survived to age x; we disregard any additional information we might possess
about chances of further survival for (x). By doing this we implicitly assume that the probability
distribution of T (x) is determined by exactly the same survival function that describes X; we assume
that P (T (x) > t) = P (X > x + t|X > x) = s(x+t) s(x) . Hence we evaluate t qx through the values of
the survival function at ages x and x + t.
However, on the basis of additional information about (x) that we might have, we may decide
that the use of s(x) is no longer appropriate as it refers to newborns and does not contain any
particular information about (x).
Such additional information may be, for instance, that the life
• has just passed a medical examination for the purpose of life assurance;
• has just been treated for a serious illness or has just become disabled;
• has just taken out a life-annuity (higher chances of survival compared to the general population,
since purchase of an annuity is only worth considering in good health, otherwise it is likely to
be a poor investment);
• has just entered the population from the outside world.
In (B) we evaluate t qx directly from the probability that a life observed alive at age x will survive
a further t years; we may use whatever values of t px we decide are appropriate for the description
of the mortality of (x) in the future.
As mentioned above, (A) and (B) are equivalent under the assumption that mortality rates
depend only on age. In previous sections we assumed this implicitly and treated (A) and (B) as
being equivalent.
Now we consider the situation when the force of mortality, and mortality rates, are a function
of age and the time since a certain event known as selection. We assume that individuals who
“(re)joined” the population after selection at age x will experience mortality that will be different
over a certain time, known as the select period, from that of the general population.
As usual in actuarial mathematics, the introduction of a new concept requires the introduction
of new notation:
• [x] is used to denote the age of selection.
• ([x] + k) denotes a person age x + k that (re)joined the population after selection at age x.
• T ([x] + k) is the future lifetime (the time-until-death) for a person age x + k who was selected
at age x.
78
Select life tables
Select life tables contain the values of life-table functions for individuals who have undergone selection
of some sort. The main select life table used in this course is the AMC00 mortality table for male
assured lives which is based on the experience of UK life assurance companies within the years 1999–
2002. Here we will mainly consider the pages of this table with column headings l[x] , l[x]+1 and lx+2
since other life-table functions can be readily expressed in terms of these functions. Specifically, in
any select life table, (and for t > 0 and k ≥ 0):
• l[x]+k denotes the expected number of survivors to age x + k in a group of l[x] individuals
who (re)joined the population at age x (with the same sort of selection for all members of the
group).
• p[x] is the probability to survive at least one further year after selection at age x, i.e., the
probability that ([x]) will attain age x + 1. It is given by
l[x]+1
p[x] = P ( T ([x]) > 1 ) = . (2.188)
l[x]
• p[x]+k is the probability for a person age x + k, who was selected at age x, to survive to age
x + k + 1. It is given by
l[x]+k+1
p[x]+k = P ( T ([x] + k) > 1 ) = . (2.189)
l[x]+k
• More generally, t p[x]+k is the probability for a person aged x + k, who was selected at age x,
to survive to age x + t + k. It is given by
l[x]+k+t
t p[x]+k = P ( T ([x] + k) > t ) = . (2.190)
l[x]+k
• t q[x]+k is the probability that a person selected at age x and being currently observed alive at
age x + k will die within t years. It is given by
l[x]+k − l[x]+k+t
t q[x]+k = P ( T ([x] + k) ≤ t ) = . (2.191)
l[x]+k
• e̊[x]+k is the complete expectation of further life for a person selected at age x and being
currently observed alive at age x + k. It is given by
∞
1 1 X l[x]+k+j
e̊[x]+k ≈ + e[x]+k = + . (2.192)
2 2 l[x]+k
j=1
Notice the similarity between the forms of these expressions and (2.50), (2.51), (2.127) and (2.135).
The life-table functions are linked in the normal way over the range of select values. So, for example,
t q[x]+k = 1 − t p[x]+k . (2.193)
Exercise 2.7.1: Select life-table functions

Use life-table function(s) from the previous pages, together with the analogue of (2.43), to obtain an expression
for t|u q[x]+k .
79
Beyond the select period the mortality rates of selected lives are assumed to revert to those of the
general population (sometimes called ultimate mortality rates). For instance, passing a medical, at
age 20 say, is not expected to say anything about the chances of you dying 50 years later! Denoting
the duration of the select period by n, we have
l[x]+k = lx+k for k ≥ n. (2.194)
and hence, for example,

p[x]+k = px+k for k ≥ n. (2.195)
The general rule is: in all formulae, we can simply use x + k instead of [x] + k, provided the duration
of the select period is n years and k is greater than or equal to n. If k ≥ n there is no difference
in terms of mortality between (x + k) and ([x] + k). Note that for the AMC00 mortality table, the
select period is 2 years; ie n = 2.
Example 2.7.1: Death probabilities from AMC00
Find q[52] , q52 , q[52]+1 , 2 q[52]+1 on the basis of the life assurance table AMC00.
Solution
This table has select period of 2 years, n = 2, so we obtain
l[52] − l[52]+1
q[52] = (2.196)
l[52]
9710.9228 − 9692.4138
= (2.197)
9710.9228
= 0.0019 (to 4 d.p.), (2.198)
l52 − l53
q52 = (2.199)
l52
9716.0627 − 9692.4332
= (2.200)
9716.0627
= 0.0024 (to 4 d.p.), (2.201)
l[52]+1 − l[52]+2
q[52]+1 = (2.202)
l[52]+1
l[52]+1 − l52+2
= (2.203)
l[52]+1
l[52]+1 − l54
= (2.204)
l[52]+1
9692.4138 − 9666.1182
= (2.205)
9692.4138
= 0.0027 (to 4 d.p.), (2.206)
l[52]+1 − l[52]+1+2
2 q[52]+1 = (2.207)
l[52]+1
l[52]+1 − l52+3
= (2.208)
l[52]+1
l[52]+1 − l55
= (2.209)
l[52]+1
9692.4138 − 9636.7719
= (2.210)
9692.4138
= 0.0057 (to 4 d.p.). (2.211)
80
Notice that q[52] < q52 , i.e., 52-year-olds who have taken out life assurance (and passed any associated
medical tests) have a lower chance of dying in the next year than 52-year-olds in the general population.
Example 2.7.2: Retirement of divers

Divers working for a marine company retire at the age of 55 years. For each retiring diver, the probability to
survive to the age of 56 is 0.8. After the age of 56, the retired divers are subject to the mortality of ELT17
Males. Find the complete expectation of life on retirement. (You may assume the approximate relation
e̊x ≈ ex + 12 .)
Solution
The age of selection is 55 and the select period is 1 year (only during that period does mortality differ from
ELT17 Males). Hence ([55] + 1) = (56), ([55] + 2) = (57), etc. and it follows that
1
e[55] = (l56 + l57 + . . .) (2.212)
l[55]
l55 1
= (l56 + l57 + . . .) (2.213)
l[55] l55
l55
= e55 . (2.214)
l[55]
We can obtain the value of e55 from the value of e̊55 given in ELT17. We can also take l55 from ELT17 but
then we need to extrapolate the corresponding value of l[55] using p[55] = l56 /l[55] :
l56 93352
l[55] = = = 116690. (2.215)
p[55] 0.8
Putting everything together, we have
1
e̊[55] ≈ + e[55] (2.216)
2
1 l55
= + e55 (2.217)
2 l[55]

1 l55 1
≈ + e̊55 − (2.218)
2 l[55] 2
1 93825 1
= + × (26.60 − ) (2.219)
2 116690 2
= 21.49 years (to 2 d.p.). (2.220)
Example 2.7.3: Life assurance for a 20-year-old

Assume mortality rates of ELT17 Males. A life assurance is being taken out by a 20-year-old man whose
chances of survival during the following two years are known to be higher than those of the general population:
p[20] = 12 (1 + p20 ) and p[20]+1 = 21 (1 + p21 ). Find the complete expectation of life for this person at age 20.
(You may assume the approximate relation e̊x ≈ ex + 12 .)
Solution
This is very similar to the previous example except that now the select period is 2 years so the calculations
are a little more complicated.
Since the duration of the selection period is 2 years, we can use the ELT17 values of the life-table functions
for all ages from 22 upwards in this question. So we can use, for example, p22 , p23 , ... and e̊22 from ELT17.
We can also use l22 , l23 , ..., provided we extrapolate the values of l[20]+1 and l[20] from l22 with the help of
81
p[20] and p[20]+1 :
l[20]+2
l[20]+1 = (2.221)
p[20]+1
l22
= (2.222)
p[20]+1
l22
= 1 (2.223)
2 (1 + p21 )
99094
= 1 [using ELT17 Males] (2.224)
2 (1 + 0.999484)
= 99119.6 (to 1 d.p.) (2.225)
and, similarly,
l[20]+1
l[20] = (2.226)
p[20]
l[20]+1
= 1 (2.227)
2 (1 + p20 )
99119.6
= 1 [using ELT17 Males] (2.228)
2 (1 + 0.999504)
= 99144.2 (to 1 d.p.). (2.229)
Therefore
1
e̊[20] ≈+ e[20] (2.230)
2
∞
1 X l[20]+k
= + (2.231)
2 l[20]
k=1
1 1
= + l[20]+1 + l[20]+2 + l[20]+3 + . . . (2.232)
2 l[20]
1 1
= + l[20]+1 + l22 + l23 + . . . (2.233)
2 l[20]

1 1 l21
= + l[20]+1 + (l22 + l23 + . . .) (2.234)
2 l[20] l21
1 1
= + l[20]+1 + l21 e21 (2.235)
2 l[20]
1 l[20]+1 l21
= + + e21 (2.236)
2 l[20] l[20]

1 l21 1
≈ + p[20] + e̊21 − (2.237)
2 l[20] 2
1 99119.6
= 0.5 + (1 + 0.999504) + (58.60 − 0.5) (2.238)
2 99144.2
= 59.59 years (to 2 d.p.). (2.239)
As expected, this is slightly greater that the ELT17 value of e̊20 .
Exercise 2.7.2: Life assurance for a 20-year-old (continued)

For the scenario of the previous example, construct an excerpt from the select life table giving lx . px and e̊x
for ages [20], [20] + 1, 22 and 23. [Hint: The only new quantity you will need to calculate is e̊[20]+1 .]
Exercise 2.7.3: *Alternative approach

Use the law of total probability to find a more direct route to (2.236) and (2.258). [Hint: You may find it
useful to first show the general relation ex = px (1 + ex+1 ).]
82
Summary of 2.7
Age of selection: [x]
Life age x + k that was selected at age x: ([x] + k)
For select period duration n: l[x]+k = lx+k for k ≥ n
(and similarly in other formulae)
83
84
Chapter 3
Life insurance and related functions
The best references for this material are: Chapters 4–7 of ([DHW13]) and Chapters 2–6 of ([Nei77]);
both of these resources cover the material in greater breadth than required for this module. You will
also need to thoroughly understand the contents of the preceding two chapters of these notes...
3.1 Introduction to life assurance

Here we put together ideas from Chapters 1 and 2 and consider payments (with interest) which
depend somehow on whether a person is alive or dead. There will be two main topics: life assurance
(e.g., death benefits, endowments) and life annuities (e.g., pensions, lottery wins, life assurance
premiums). Within the field of life assurance there is a lot of technical terminology which is used to
distinguish different types of policy as summarized below.
Life assurance glossary

A life assurance policy is a contract which pays a specified amount of money (called the benefit
payment or the sum assured) on the death of a specified person (called the life assured).
We will consider 4 types of life assurance policies.
1. Whole-life assurance: this is a life assurance policy which pays on the death of the life
assured at any future time (i.e., the policy remains in force for the whole life of the assured
person).
2. n-year term life assurance: this is a life assurance policy which pays on the death of the life
assured only if the death occurs within n years from the start of the contract (i.e., the policy
remains in force only for a fixed term of n years).
3. Pure endowment policy: this is a contract which pays a benefit on the survival of the life
assured to a certain age/date.
4. n-year term endowment assurance policy: this is a policy which combines an n-year term
life assurance with a pure endowment policy, i.e., it provides for a benefit either on death or
on survival (of the life assured) to the end of the n-year term whichever event occurs first.
The benefits may be level (constant) or they may decrease or increase in a way specified in
the contract. Another distinction is that for with profit policies the benefits may be increased
by additions called bonuses, whereas for without profit (non-profit) policies the benefits are
completely specified in money terms in the contract. In this course, we consider only non-profit
policies with level benefits.
85
To purchase life assurance, a person makes a one-off payment or, more usually, payment in
regular installments to the insurance company (the life office) in return for payment of the benefit.
Normally, the life office invests these collected premiums into a fund. This fund earns interest and
is used to pay out benefits. Premiums normally include charges. The charges are used to cover the
life company’s expenses. Premiums are worked out by applying the equation of value: when invested
into the fund they should generate a return which will then be used to pay out the benefit and to
cover the company’s expenses.
Exercise 3.1.1: Schematic of life assurance set-up
Draw a flow diagram showing the inflows and outflows for a life assurance fund.
For simplicity we assume no expenses. Under this assumption we should equate the present
values of benefit and premiums. However, these present values are random variables, as they depend
on the survival of the life assured: the benefit is paid on the death of the life assured and the
premiums are normally only paid whilst the life assured is alive.
As the exact time-until-death for the life assured is unknown, the premiums are worked out by
equating the expected present values (E.P.V.s) of the benefit and premiums:
E.P.V. of benefit(s) = E.P.V. of premiums (3.1)
So the aim of the present chapter is to learn

• How to calculate the E.P.V. of benefit(s);
• How to calculate the E.P.V. of premiums.
The payment of premiums can be regarded as a life annuity as we shall discuss in Section 3.3.
For calculating the E.P.V. of life assurance benefits, we shall consider three modes of payment of
the death benefit:
• Death benefit payable on the moment of death (Section 3.2),
• Death benefit payable at the end of the year of death (Section 3.2),
• Death benefit payable at the end of the month (or quarter, week, etc.) of death (Section 3.5).
Commutation functions
We will assume throughout this chapter that the life assurance fund earns interest at a constant
effective rate of i per annum. Hence the P.V. of one unit of money due in t years is v t , where
v = 1/(1 + i).
Two functions of lx and v will appear frequently and, for convenience, are given the symbols Dx
and Nx :
Dx = lx v x = l0 v x s(x), (3.2)
X∞ X∞ ∞
X
Nx = Dx+k = lx+k v x+k = l0 v x+k s(x + k). (3.3)
k=0 k=0 k=0
Notice that they depend on both mortality (through lx ) and the prevailing interest rate (through
v). Dx and Nx are examples of so-called commutation functions which have historically played an
important role in actuarial work.1 Values of Dx and Nx are available in tables for different interest
rates; we’ll use AMC00 with interest at 4% per annum.
1
Nowadays their use has diminished somewhat since computer algorithms based on recursion relations provide an
alternative route for rapid calculation of many quantities.
86
Exercise 3.1.2: Commutation functions
Show that Nx+1 = Nx − Dx .
Two other commutation columns are often found in life tables (including the tables used for this
module) are Cx and Mx which are sometimes used in the calculation of the present value of life
assurance functions. These are defined as:
Cx = (lx − lx+1 ) v x+1 , (3.4)

X∞ X ∞
Mx = Cx+k = (lx+k − lx+k+1 ) v x+k+1 . (3.5)
k=0 k=0
Probability reminder
The following fact, which should already be familiar, will be particularly useful in the forthcoming
analysis.
If T is a continuous random variable, with p.d.f. fT (t), then
Z ∞
E(h(T )) = h(t)fT (t) dt. (3.6)
−∞
Simple applications of this formula include the calculation of moments, e.g.,

Z ∞
E(T ) = tfT (t) dt, (3.7)
Z−∞
∞
E(T 2 ) = t2 fT (t) dt, (3.8)
−∞
but we shall use it to calculate expectations of slightly more complicated functions.
Summary of 3.1
For life assurance: E.P.V. of benefit(s) = PE.P.V. of premiums

Commutation functions: Dx = lx v and Nx = ∞
x
k=0 Dx+kP
Cx = (lx − lx+1 ) v x+1 and Mx = ∞ k=0 Cx+k
3.2 Whole-life assurance

Death benefit payable at instant of death
Consider a person of age x taking out whole-life assurance with unit death benefit payable at the
instant of death. What is the cost of this policy (ignoring expenses)? To answer this, we first note
that since the sum assured is paid immediately on the death of (x), its present value is
Z = v T (x) , (3.9)
where T (x) is the exact time-until-death for (x). T (x) is a random variable, with p.d.f. given by
fT (x) (t) = t px µ(x + t), and hence Z is also a random variable. The cost of the policy is just the
87
expected value of this random variable which is denoted by Āx and is given by
Āx = E(Z) (3.10)

= E(v T (x) ) (3.11)
Z ∞
= v t fT (x) (t) dt [using (3.6)] (3.12)
0
Z ∞
= v t t px µ(x + t) dt [using (2.81)] (3.13)
0
Z ∞
s0 (x + t)
=− vt dt [using (2.13)]. (3.14)
0 s(x)
Note that the “A” stands for “assurance” and the bar indicates that the sum assured is paid
immediately on the death of (x). By definition, Āx is the expected present value of unit death
benefit payable immediately on the death of (x). If the death benefit is S units of money, then by
proportion its E.P.V. is
E(SZ) = S Āx . (3.15)
In order to quantify the risk for the life company (an issue to which we shall return later), it
is also important to know the variance of the present value which, for unit death benefit, is simply
calculated as follows:
h i2
Var(Z) = E(v 2T (x) ) − E(v T (x) ) (3.16)
Z ∞
= v 2t fT (x) (t) dt − (Āx )2 (3.17)
0
Z ∞
= (v ∗ )t fT (x) (t) dt − (Āx )2 [setting v ∗ = v 2 ] (3.18)
0
= Ā∗x − (Āx )2 . (3.19)
Here, and elsewhere, the * superscript denotes that the quantity is evaluated at the modified rate
of interest defined by v ∗ = v 2 .
Exercise 3.2.1: Relationship between i∗ and i
Show that the condition v ∗ = v 2 implies i∗ = 2i + i2 .
If the death benefit is S units of money then the variance of its present value is given by
Var(SZ) = S 2 [Ā∗x − (Āx )2 ].
Example 3.2.1: Whole-life assurance with constant force of mortality

Assume a constant force of mortality µ and a constant force of interest δ. Express the expectation and
variance of the present value of unit death benefit, payable immediately on death under a whole-life assurance
policy, in terms of µ and δ.
Solution
For a constant force of mortality
µ(u) = µ for all u ≥ 0. (3.20)
Therefore Rx
s(x) = e− 0
µdu
= e−µx for all x ≥ 0,
and
s0 (x + t)
fT (x) (t) = t px µ = − = e−µt µ for all t ≥ 0.
s(x)
88
Notice that fT (x) (t) does not depend on x which is, of course, a property specific to the case of a constant
force of mortality (exponentially distributed lifetime).
The present value of the death benefit is the random variable Z = v T (x) and, using the above expression
for fT (x) (t), we can easily calculate its expectation:
Āx = E(v T (x) ) (3.21)

Z ∞
= v t e−µt µ dt (3.22)
0
Z ∞
1
= e−δt e−µt µ dt since v = = e−δ (3.23)
0 1+i
µ
= . (3.24)
µ+δ
The variance of the P.V. (i.e., the variance of v T (x) ) is
Var(Z) = Ā∗x − (Āx )2 , (3.25)

∗
where e−δ = v ∗ = v 2 = e−2δ . Hence δ ∗ = 2δ and therefore
2
µ µ
Var(Z) = − . (3.26)
µ + 2δ µ+δ
Note that we can also write Āx in terms of commutation functions. To see this we start
from (3.14) and use integration by parts
∞
s0 (x + t)
Z
Āx = − vt dt (3.27)
0 s(x)
Z ∞
1
=− v t s0 (x + t) dt (3.28)
s(x) 0
Z ∞
dv t

1 t t=∞
=− v s(x + t) t=0 − s(x + t) dt (3.29)
s(x) 0 dt
Now, since v = e−δ ,

dv t
= −δe−δt = −δv t . (3.30)
dt
Also, for any δ > 0,
v t s(x + t) → 0 as t → ∞. (3.31)
Hence (3.29) gives
Z ∞
1 t
Āx = − −s(x) + δ v s(x + t) dt (3.32)
s(x) 0
Z ∞
s(x + t) l0 v x
=1−δ vt dt × (3.33)
s(x) l0 v x
Z0 ∞
Dx+t
=1−δ dt [using (3.2)]. (3.34)
0 Dx
Death benefit payable at end of year of death

Now let’s consider the slightly more realistic case of whole-life assurance for (x) in which the benefit
is payable at the end of the year of death – on what would have been the next birthday. The symbol
Ax (no bar, no dots!) denotes the cost of such a policy, specifically, the expected present value of
89
unit death benefit payable at the end of (x)’s year of death. Ax can be written in an analogous
form to (3.34) as
∞
X Dx+k Nx
Ax = 1 − d =1−d . (3.35)
Dx Dx
k=0
To prove (3.35) we note that the present value of the death benefit in this case is the random
variable
Z = v K(x)+1 (3.36)
where K(x) is the curtate further life. Hence
Ax = E(Z) (3.37)
∞
X
= v k+1 P (K(x) = k) (3.38)
k=0
X∞
= v k+1 k|1 qx [see (2.116)] (3.39)
k=0
∞ ∞
X lx+k X k+1 lx+k+1
= v k+1 − v [using (2.52)]. (3.40)
lx lx
k=0 k=0
Now the first sum gives

∞ ∞
X lx+k X lx+k vx
v k+1 =v vk × x (3.41)
lx lx v
k=0 k=0
∞
X Dx+k
=v (3.42)
Dx
k=0
Nx
=v , (3.43)
Dx
while the second is
∞ ∞
X lx+k+1 X 0 lx+k 0
v k+1 = vk [change of variable k 0 = k + 1] (3.44)
lx 0
lx
k=0 k =1
∞
X Dx+k0
= (3.45)
Dx
k0 =1
∞
X Dx+k0
= −1 (3.46)
0
Dx
k =0
Nx
= − 1. (3.47)
Dx
Substituting these expressions back into (3.40) we have

Nx Nx
Ax = v − −1 (3.48)
Dx Dx
Nx
= 1 − (1 − v) (3.49)
Dx
Nx
=1−d [recall d = 1 − v, see (1.93)]. (3.50)
Dx
90
The variance of the P.V. can also be calculated:
h i2
Var(Z) = E(v 2(K(x)+1) ) − E(v K(x)+1 ) (3.51)
h i2
= E((v ∗ )K(x)+1 ) − E(v K(x)+1 ) (3.52)
= A∗x − (Ax )2 . (3.53)
where again the superscript ∗ indicates evaluation at v ∗ = v 2 .

Obviously, for a death benefit of S units of money paid at the end of the year of death, the
expectation of the present value is SAx and its variance is S 2 [A∗x − (Ax )2 ]. Since values of Nx
and Dx are readily available in tables we can thus easily calculate the E.P.V. (equal to policy cost)
and/or variance for a given situation. One caveat is that it is important to read the question carefully
to determine whether select or ultimate life-table values are to be used.
Exercise 3.2.2: Worth faking your death?

Find the cost (ignoring expenses) of a whole-life assurance with a death benefit of £160,000, payable at the
end of the year of death, for John Darwin age 30. Assume 4% interest and select mortality from the AMC00
table.
Alternative Calculation of Ax
As you become familiar with life assurance functions and commutation columns, you will see that
there often 2 or more equivalent formulae that can be used to value a life assurance benefit.
From (3.40) we see:
∞ ∞
X lx+k X k+1 lx+k+1
Ax = v k+1 − v (3.54)
lx lx
k=0 k=0
∞
X (lx+k − lx+k+1 )
= v k+1 (3.55)
lx
k=0
∞
X (lx+k − lx+k+1 ) v x+k+1
= (3.56)
lx v x
k=0
∞
X Cx+k
= [using (3.4)] (3.57)
Dx
k=0
Mx
= [using (3.5)]. (3.58)
Dx
Common tables of life insurance functions and commutation columns will often include tabulated
values of the most common functions such as Ax ; that is true for the tables in this module. You
can safely assume that in an exam you will not be asked to calculate a function that can be copied
from a table!
Relation between Āx and Ax
By linear interpolation on lx (i.e. asumming a uniform distribution of deaths) we can obtain a useful
approximate relationship between Āx and Ax . The derivation proceeds in a similar spirit to that for
the relationship between e̊x and ex given in Section 2.6.
91
Starting from (3.14) we have
Z ∞
1
Āx = − v t s0 (x + t) dt (3.59)
s(x) 0
∞ Z
1 X k+1 t 0
=− v s (x + t) dt [partitioning integration interval] (3.60)
s(x)
k=0 k
∞ Z
1 X 1 u+k 0
=− v s (x + k + u) du [substituting u = t − k]. (3.61)
s(x) 0
k=0
Now linear interpolation on lx is equivalent to the approximation s(x + k + u) ≈ (1 − u)s(x + k) +

us(x + k + 1) and taking the derivative with respect to u, we obtain s0 (x + k + u) ≈ s(x + k + 1) −
s(x + k). Therefore
∞ 1
s(x + k) − s(x + k + 1)
X Z
k
Āx ≈ v v u du (3.62)
s(x) 0
k=0
Z 1 ∞
X
u
= v du v k k|1 qx . (3.63)
0 k=0
Now
1 1
1−v
Z Z
iv
u
v du = e−δu du = = , (3.64)
0 0 δ δ
so
∞
i X k+1
Āx ≈ v k|1 qx , (3.65)
δ
k=0
and by comparison with (3.39) we have the simple relationship
i
Āx ≈ Ax . (3.66)
δ
In Section 3.5 we will see some different E.P.V. approximations resulting from linear interpolation
on Dx .
Exercise 3.2.3: *Alternative derivation of Āx ≈ δi Ax
Linear interpolation on lx is equivalent to the assumption that T (x) = K(x)+R(x) where K(x) and R(x) are
independent random variables and R(x) is uniform on (0, 1) (see again Section 2.6). Provide an alternative
derivation of (3.66) starting from the observation that
v T (x) = v K(x)+R(x) = v K(x)+1 × v R(x)−1 . (3.67)
An alternative approximation for Āx

Another approach is to consider that the difference between Āx and Ax is only one of the timing of
the payment of the death benefit. If the benefit is paid immediately on death, then, assuming the
UDD once again, the benefit will be paid approximately 6 months earlier than if paid at the end of
the policy year. This leads naturally to the following approximation:
Āx ≈ (1 + i)0.5 Ax (3.68)
Both approximations lead to similar answers at common interest rates; for example if i = 10%
then δi = 1.0492 and (1 + i)0.5 = 1.0488.
92
Summary of 3.2
E.P.V.s of whole-life assurance for (x):

R∞ R∞ Dx+t
Unit benefit at instant of death: Āx = 0 v t fT (x) (t) dt = 1 − δ 0 Dx dt
Nx
Unit benefit at end of yr of death: Ax = 1 − d D x
Approximate relationships: Āx ≈ δi Ax
Āx ≈ (1 + i)0.5 Ax
3.3 Whole-life annuities payable annually
Since the cost of life assurance is typically quite high (see, e.g., Example 3.2.2), a policy is usually
purchased not with a lump sum but with a series of payments at regular intervals. The premiums
are generally only paid while the person is still alive so they form a life annuity. The life assurance
context provides one motivation for the study of life annuities in this section but other applications
include pensions and some types of lottery prize. For now we will consider only whole-life annuities
with annual payments; extensions to the cases of limited duration and more frequent payments will
be covered in Sections 3.4 and 3.5 respectively.
Whole-life annuity-due
Consider a series of annual payments, each of one unit of money, made in advance to a life of age
x. The payments are only made while x is alive. This situation is a whole-life annuity-due. The
present value of these payments (when the policyholder is exact age x) is, of course, a random
variable. Its expectation is denoted by äx and will be calculated below.
First we notice that if K(x) is the curtate further lifetime of (x) (i.e., the number of complete
years still to be lived) then payments will be made at ages x, x + 1, x + 2, . . . , x + K(x). Hence
there are K(x) + 1 payments in total and
äx = E(äK(x)+1 ) (3.69)

∞
X
= äk+1 P (K(x) = k) (3.70)
k=0
∞
X
= äk+1 (k px − k+1 px ) (3.71)
k=0
∞
X ∞
X
= äk+1 k px − äk0 k 0 px [substituting k 0 = k + 1 in 2nd sum] (3.72)
k=0 k0 =1
∞
X ∞
X
= ä1 0 px + äk+1 k px − äk0 k0 px [taking out k = 0 term in 1st sum] (3.73)
k=1 k0 =1
Now ä1 = 1 = v 0 and also, since the only difference between äk+1 and äk is the (k + 1)th payment,
we have
äk+1 − äk = v k (3.74)
93
as is easily checked from the explicit formula for än . Returning to (3.73), we then find
∞
X
äx = v 0 0 px + v k k px (3.75)
k=1
∞
X
= v k k px (3.76)
k=0
∞
X lx+k
= vk (3.77)
lx
k=0
∞
X Dx+k
= , (3.78)
Dx
k=0
giving finally a simple expression for the E.P.V. of a whole-life annuity-due:
Nx
äx = . (3.79)
Dx
Substituting in (3.50) then gives the “conversion relationship”
Ax = 1 − däx . (3.80)
It’s also straightforward to find a relationship between the variance of the P.V. of a whole-life annuity-
due and the variance of the P.V. of a whole-life assurance (with payment at the end of the year of
death):
!
1 − v K(x)+1
Var(äK(x)+1 ) = Var (3.81)
1−v
!
1 v K(x)+1
= Var − (3.82)
1−v 1−v
2
1
= Var(v K(x)+1 ) [Var(α + βX) = β 2 Var(X)] (3.83)
1−v
1
= 2 (A∗x − (Ax )2 ) (3.84)
d
Obviously, if the annual payments are P rather than 1, the E.P.V. and the variance of the P.V. are
scaled by P and P 2 respectively.
Finally, remember that publishd tables often include common values such as äx .
Exercise 3.3.1: Alternative derivation of äx

Starting from (3.72), use the fact that ä0 = 0 to add a k 0 = 0 term into the second sum and hence provide
an alternative (simpler?) derivation of (3.79).
Example 3.3.1: Valuing a pension

Fred Goodwin (formerly Sir Fred Goodwin) retired at age 50 with a pension of £693,000 per year. If the
payments were to be paid annually in advance, what was the expected present value (on his retirement) of
Fred’s pension? Assume AMC00 ultimate (non-select) values with 4% interest. Why do you think your answer
is different to the estimates of the Royal Bank of Scotland and Standard Life (see Sunday Times article of
01.03.09)...?
94
Solution
Remembering that äx is defined for payments of one unit of money per year, we have
E.P.V. = £693000 × ä50 (3.85)

N50
= £693000 × (3.86)
D50
24774.94
= £693000 × (3.87)
1372.86
= £12506034. (3.88)
So we estimate the market value of his pension as £12.5million (to 3 significant figures). This is rather
less than the estimates of the Royal Bank of Scotland (£16.6m) and Standard Life (£32.7m). Reasons for
the discrepancy might include that life expectancy has increased since AMC00 and that prevailing interest
rates were lower than 4%; both these factors would increase the E.P.V. compared to our estimate. The
pension is probably also paid more frequently than annually; we will learn how to deal with p-thly payments
in Section 3.5. Finally, the pension might increase once it is in payment and may continue after his death,
usually at a lower rate, to his partner; these factors would further increasing the value of the pension.
Exercise 3.3.2: Rich for life

Camelot used to offer a “Rich For Life” scratchcard which cost £5 and offered a top prize of £40,000 a year,
every year, for the rest of your life. Suppose that you buy a ticket on your 21st birthday and happen to win
the top prize. Assuming AMC00 ultimate (non-select) values with 4% interest, what would be the expected
present value (on that day) of your winnings? You may assume that the yearly payments are made in advance
and ignore complications such as tax and validation fees.
[Answer: £928,240 (to nearest multiple of £10)]
Whole-life immediate annuity

Let us now consider a series of annual payments, each of one unit of money, made in arrears to a life
of age x. The payments are only made while (x) is alive. This situation is a whole-life immediate
annuity. The expected present value of the payments is denoted by ax .
In fact, this is just the same as the previous case except that the payment at age x is not made.
Since the P.V. of 1 due at age x is just 1, and the payment is guaranteed because the person is
certainly still alive at age x, we must have
ax = äx − 1. (3.89)
This leads to an explicit expression for ax in terms of commutation functions:

Nx
ax = −1 (3.90)
Dx
Nx − Dx
= (3.91)
D
P∞ x
k=0 Dx+k − Dx
= (3.92)
Dx
P∞
k=0 Dx+k+1
= (3.93)
Dx
Nx+1
= . (3.94)
Dx
Exercise 3.3.3: Deferred whole-life annuities-due
Show that the E.P.V. of a whole-life annuity-due, for a life aged x, deferred by m years (i.e., with the first
payment at age x + m) is given by
Nx+m
m |äx = . (3.95)
Dx
95
Exercise 3.3.4: Valuing a pension (revisited)
Suppose Fred Goodwin had originally intended to retire at age 60 with a yearly pension of £693,000. Had
he been “dismissed” rather than taken “early retirement”, his pension would have been reduced from the
full amount of £693,000. This implies that he would have received a reduced annual payment so that the
E.P.V. of the pension payments at age 50 was unchanged (and equal to the value of his pension if it started
at age 60). Estimate this reduced payment. Assume again yearly payments in advance and AMC00 ultimate
(non-select) values with 4% interest.
Solution
Let the reduced premium be £P then
Value at age 50 of pension starting at age 60 = 10 |ä50 (3.96)

⇒ P ä50 = 69300010 |ä50 . (3.97)
In terms of commutation functions this gives
N50 N60
P = 693000 , (3.98)
D50 D50
so
N60
P = 693000 (3.99)
N50
13323.07
= 693000 (3.100)
24774.94
= 372670 to the nearer multiple of £10 (3.101)
i.e., we estimate a yearly pension of £372,670 (to 3 s.f.). [The Royal Bank of Scotland said it would have
been £416,000 per year.]
Life assurance premiums

Consider a whole-life assurance to a life of age x, with unit death benefit payable at the end of
the year of death, which is purchased by annual payments in advance for life. The annual premium
can be worked out by equating the E.P.V. of the death benefit with the E.P.V. of a whole-life
annuity-due. If the annual premium is P we must have
Ax = P äx , (3.102)
and hence, using the conversion relationship (3.80)
1 − däx = P äx , (3.103)
which is easily rearranged to give
1
P = −d (3.104)
äx
Dx
= − d. (3.105)
Nx
It’s hardly worth remembering this equation since it’s so easily derived.
Example 3.3.2: Worth faking your death? (revisited)

Suppose that John Darwin (see Example 3.2.2) pays for his life assurance by a whole-life annuity-due. What
is the amount of each payment?
96
Solution
Let the annual premium be £P . Then the equation of value is:
Present Value of Premiums = Present Value of Benefits (3.106)
P ä[30] = £160000x × A[30] (3.107)
P ä[30] = 23720 useing result from 3.2.2 (3.108)
(3.109)
and hence
23730
P = (3.110)
ä[30]
23720
= (3.111)
N[30] /D[30]
23720
= (3.112)
67835.43/3063.17
= 1071.10 (to 2 d.p.) (3.113)
So the annual premium is £1071.10.
The same answer is obviously
obtained
by using (3.105) and multiplying by the required amount of death
D[30]
benefit, i.e., P = 160000 N[30] −d .
Exercise 3.3.5: Life assurance premiums as immediate annuity

Show that if the life assurance premiums are paid in arrears, then (3.105) is replaced by
Dx − dNx
P = . (3.114)
Nx+1
Summary of 3.3
E.P.V.s of whole-life annuities for (x):
Nx
Unit payments, annually in advance: äx = Dx
Unit payments, annually in arrears: ax = äx − 1
Conversion relationship: Ax = 1 − däx
3.4 Policies of duration n

In this section we will examine a variety of policies of duration n years, i.e., policies which apply
only over some finite term. They should be contrasted with the whole-life assurances and whole-life
annuities we discussed in the previous two sections.
n-year life assurance for life aged x

Let us consider life assurance policies in which the death benefit is only paid if death occurs within
n years from the start of the contract. (For example, a father with young children might take out
a life assurance policy with term 20 years so that if he should die before the children had grown up
they would be taken care of.) We restrict ourselves to policies which pay at the end of the year of
death. Our aim is to express the E.P.V. of unit death benefit in a policy of term n, taken out by a
life of age x, in terms of the commutation functions Dx and Nx .
Let Z be the P.V. of unit death benefit, then Z is a random variable taking the values
(
v K(x)+1 if K(x) < n
Z= (3.115)
0 if K(x) ≥ n.
97
1
The E.P.V. of n-year life assurance for (x) with unit death benefit is denoted by Ax:n (be especially
1
careful with the position of the superscript ) and given by
1
Ax:n = E(Z) (3.116)
n−1
X
= v k+1 P (K(x) = k) (3.117)
k=0
n−1
X
= v k+1 (k px − k+1 px ) (3.118)
k=0
n−1
X n−1
X
= v k+1 k px − v k+1 k+1 px . (3.119)
k=0 k=0
Now, considering each of the two sums separately, we have

n−1 n−1 n−1
X
k+1
X
k
X Dx+k
v k px =v v k px = v , (3.120)
Dx
k=0 k=0 k=0
and
n−1 n
0
X X
v k+1 k+1 px = vk k0 px [substituting k = k 0 − 1] (3.121)
k=0 k0 =1
n
X Dx+k0
= (3.122)
Dx
k0 =1
n−1
X Dx+k0 Dx+n
= −1+ . (3.123)
Dx Dx
k0 =0
where, in the last line, we have included the k 0 = 0 term in the sum and taken the k 0 = n term out.
Substituting back in (3.119) then gives
n−1 n−1
!
X Dx+k X Dx+k Dx+n
1
Ax:n = v − −1+ (3.124)
Dx Dx Dx
k=0 k=0
n−1
Dx+n X Dx+k
=1− − (1 − v) (3.125)
Dx Dx
P∞ k=0 P∞
Dx+n k=0 Dx+k − k=n Dx+k
=1− −d (3.126)
Dx Dx
P∞ P∞
Dx+n k=0 Dx+k − k0 =0 Dx+n+k0
=1− −d [substituting k = k 0 + n] (3.127)
Dx Dx
Dx+n Nx − Nx+n
=1− −d . (3.128)
Dx Dx
Exercise 3.4.1: Limiting behaviour of n-year life assurance
Check that taking n → ∞ in (3.128) recovers the result (3.50) for whole-life assurance.
n-year pure endowment for life aged x

A pure endowment policy pays out on survival of (x) to age x + n. The cost of such a policy is
given (ignoring expenses) by the expected present value of the survival benefit which we shall now
calculate.
98
Following the now familiar strategy, we first write down the possible values of the random variable
Z denoting the P.V. of unit survival benefit:
(
v n if K(x) ≥ n
Z= (3.129)
0 if K(x) < n.
and then find the expectation of this random variable. In the present case, this is particularly easy
since we observe that the p.m.f. of Z is
P (Z = v n ) = n px (3.130)
P (Z = 0) = 1 − n px = n qx , (3.131)
which implies that
Z ∼ v n Bernoulli(n px ). (3.132)
The symbol for the E.P.V. of an n-year pure endowment for (x) with unit survival benefit is
Ax:n1 (note the position of the 1 ). It can easily be written in terms of commutation functions:
Ax:n1 = E(Z) (3.133)

n
= v × n px (3.134)
lx+n
= vn (3.135)
lx
Dx+n
= . (3.136)
Dx
Similarly, the variance of the P.V. is easily obtained:
Var(Z) = Var[v n Bernoulli(n px )] (3.137)

= v 2n n px (1 − n px ) (3.138)

2n lx+n lx+n
=v 1− (3.139)
lx lx

l
n x+n n n lx+n
=v v −v (3.140)
lx lx

Dx+n n Dx+n
= v − . (3.141)
Dx Dx
Example 3.4.1: Granny’s contribution to buying a house
Mrs Case wants to help her 18-year old grandson Justin but a house a few years after he graduates from
University. She therefore pays £5,000 for a pure endowment which will pay a benefit to Justin on his survival
to age 25. Assuming AMC00 select values with an effective interest rate of 4% p.a., how much will this
survival benefit be?
Solution
Let the survival benefit of the policy be £S. Ignoring expenses, the cost of the policy must be equal to the
E.P.V. of the benefit. Hence
Present Value of Premiums = Present Value of Benefits (3.142)

5000 = S × A[18]:71 (3.143)
D[18]+7
=S× (3.144)
D[18]
99
and, by trivial re-arrangement
5000D[18]
S= (3.145)
D[18]+7
5000D[18]
= [AMC00 has select period of 2 years] (3.146)
D25
5000 × 4932.79
= (3.147)
3737.18
= 6599.62 (to 2 d.p.). (3.148)
So Justin will receive £6599.62 on his 25th birthday.
Exercise 3.4.2: Granny’s contribution to buying a house (revisited)

Suppose that Mrs Case (see the previous example) had just invested her £5,000 in a bank account paying
interest at a constant AER of 4%. How much would be the accumulation after 7 years? Explain why it is
(VERY slightly) less than the survival benefit from the pure endowment considered above.
n-year endowment policy for life aged x

An n-year endowment policy is a combination of an n-year life assurance and an n-year pure endow-
ment. Hence it pays:
• Death benefit if the life assured dies within n years,
• Survival benefit if the life assured survives n years.
(Such an endowment policy can be used, for example, to pay off a mortgage on a house; unsurpris-
ingly, this is called an endowment mortgage.) We will assume that the death benefit is paid at the
end of the year of death, whereas the survival benefit is paid immediately on survival to the end of
the term. Furthermore we assume here that the death and survival benefits are both of one unit of
money.
If Z1 is the P.V. of the death benefit and Z2 is the P.V. of the survival benefit, then it is clear
that the E.P.V. of the total benefits is just E(Z1 ) + E(Z2 ). The actuarial symbol for the E.P.V. of
an n-year endowment policy for (x) with unit death and survival benefits is Ax:n . Hence we have
1
Ax:n = Ax:n + Ax:n1 (3.149)
Dx+n Nx − Nx+n Dx+n
=1− −d + (3.150)
Dx Dx Dx
Nx − Nx+n
=1−d . (3.151)
Dx
n-year life annuities for life aged x

Often an n-year life assurance (or endowment) is paid for by an n-year life annuity. In that case,
ignoring expenses, the annual premium is worked out by equating the E.P.V.s of assurance and
annuity. To facilitate this, of course, we now need to derive expressions for the E.P.V.s of n-year life
annuities.
Let us first consider an n-year life annuity-due which consists of annual payments of one unit of
money in advance, whilst (x) is alive, but for at most n years. Denoting the P.V. of the series of
payments by Z, we have (
äK(x)+1 if K(x) < n
Z= (3.152)
än if K(x) ≥ n.
100
The E.P.V. of an n-year life annuity-due for (x) is denoted by äx:n (defined, as usual, for payments
of one unit of money per unit time) and given by
äx:n = E(Z) (3.153)

n−1
X
= äk+1 P (K(x) = k) + än P (K(x) ≥ n) (3.154)
k=0
n−1
X
= äk+1 (k px − k+1 px ) + än n px . (3.155)
k=0
Now,
n−1
X n−1
X
äk+1 k px = (äk + v k ) k px , (3.156)
k=0 k=0
and
n−1
X n
X
äk+1 k+1 px = äk0 k0 px [substituting k = k 0 − 1] (3.157)
k=0 k0 =1
n−1
X
= äk0 k0 px + än n px [note that ä0 = 0]. (3.158)
k0 =0
Substituting back into (3.155) we obtain

n−1 n−1
!
X X
äx:n = (äk + v k ) k px − äk k px + än n px + än n px (3.159)
k=0 k=0
n−1
X
= v k k px (3.160)
k=0
n−1
X Dx+k
= (3.161)
Dx
k=0
Nx − Nx+n
= . (3.162)
Dx
Furthermore, by comparison with (3.151) we have the conversion relationship
Ax:n = 1 − däx:n . (3.163)
Similarly, one can derive an expression for the E.P.V. of an n-year life immediate annuity for (x),
with payments of one unit of money per unit time:
Nx+1 − Nx+1+n
ax:n = . (3.164)
Dx
Exercise 3.4.3: n-year life immediate annuity
Explain why
ax:n = äx:n+1 − 1. (3.165)
Hence, or otherwise, prove (3.164).
Exercise 3.4.4: Limiting behaviour of n-year life annuities

Check that taking the limit n → ∞ in (3.162) and (3.164) one recovers the corresponding results for whole-life
annuities.
101
Exercise 3.4.5: Relations between life annuities
Use the result (3.95) to quickly obtain (3.162) by arguing with the aid of a diagram that
äx:n = äx − n |äx . (3.166)
Summary of 3.4
E.P.V.s for policies of term n for (x):

n-year life annuity-due with unit payment annually: äx:n = Nx −N
Dx
x+n
n-year endowment policy with unit benefits: Ax:n = 1 − däx:n

n-year pure endowment with unit survival benefit: Ax:n1 = DDx+n
x
n-year life assurance with unit death benefit: 1
Ax:n = Ax:n − Ax:n1
3.5 p-thly payments
In most situations payments are only made once per year so, in this section, we discuss how to
deal with more frequent payments. Although we restrict ourselves here to the cases of whole-life
assurance and whole-life annuities, the corresponding n-year policies can be treated in an analogous
way.
Whole-life assurance paid p-thly
Consider a whole-life assurance for (x) which pays unit death benefit at the end of the p-thly period
of death. (For example, if p = 12 the benefit is paid at the end of the month of death, whereas if
p = 4 it is paid at the end of the quarter.)
We denote, as usual, the present value of the death benefit (at the time the policy is taken out)
by Z. Then Z is a discrete random variable taking values v 1/p , v 2/p , v 3/p , etc. and indeed it is easy
to see that
j j+1
Z = v (j+1)/p if < T (x) ≤ (j = 0, 1, 2, . . .). (3.167)
p p
102
(p)
The E.P.V. is denoted by Ax and given by
A(p)
x = E(Z) (3.168)
∞
X
(j+1)/p j j+1
= v P < T (x) ≤ (3.169)
p p
j=0
∞
X
= v (j+1)/p j | 1 qx (3.170)
p p
j=0
X∞
(j+1)/p
= v j px − j+1 px (3.171)
p p
j=0
∞
X ∞
X
1/p j/p
=v v j px − v (j+1)/p j+1 px (3.172)
p p
j=0 j=0
∞ ∞
0
X X
= v 1/p v j/p j px − v j /p j 0 px [substituting j = j 0 − 1] (3.173)
p p
j=0 j 0 =1
 
∞ ∞
0
X X
= v 1/p v j/p j px −  v j /p j 0 px − 1 (3.174)
p p
j=0 j 0 =0
∞
X
= 1 − (1 − v 1/p ) v j/p j px (3.175)
p
j=0
∞
X Dx+j/p
= 1 − (1 − v 1/p ) . (3.176)
Dx
j=0
Now, noticing that

1/p
1/p 1
(1 − v )=1− (3.177)
1+i
(1 + i)1/p − 1
= (3.178)
(1 + i)1/p
i(p)
p
= (p)
[using (1.20)] (3.179)
1 + ip
d(p)
= [see (1.70)], (3.180)
p
and defining
∞
1X
Nx(p) = Dx+j/p , (3.181)
p
j=0
we can write (3.176) in the more elegant form

(p)
Nx
A(p)
x =1−d
(p)
(3.182)
Dx
which can be expressed asA(p)
x
(p) (p)
= 1 − d äx (3.183)
where
(p)
Nx
ä(p)
x = (3.184)
Dx
103
(p)
By now, it should be clear that äx the present value of a whole-life p-thly annuity-due paid at
the rate of one unit of money per unit time paid in p-thly payments of 1/p units of money in advance
for life. However, for practical applications this expression is still not very useful since it involves a
sum over Dx+j/p whereas tables only give Dx for integer x. We resolve this difficulty in the next
subsection.
Using 3.183 and taking the limit p → ∞ we obtain the expression for the present value of a
whole life assurance function payable on death:
Āx = 1 − δāx [recall lim d(p) = δ]. (3.185)

p→∞
Whole-life p-thly annuities

In the last section we obtained expressions for the E.P.V.s of whole-life policies paid p-thly. However,
as hinted there, such expressions are of limited practical value since they involve sums over Dx+j/p
while tables, such as AMC00, only include Dx for integer values of x. Hence our strategy now is to
(p)
obtain an approximate expression for äx (and other related E.P.V.s) by using linear interpolation.
To be specific, we shall use linear interpolation on Dx , i.e.,
Dx+t ≈ (1 − t)Dx + tDx+1 , for x = 0, 1, 2, . . ., 0 ≤ t ≤ 1, (3.186)
Note that this is not the same approximation as using linear interpolation on lx as we did in
(p)
Section 2.6. We can use (3.186) to approximate Nx starting from its definition (3.181):
∞
1X
Nx(p) = Dx+j/p (3.187)
p
j=0
∞ p−1
1XX
= Dx+k+j 0 /p [arranging into yearly sums] (3.188)
p 0 k=0 j =0
∞ p−1
j0 j0

1XX
≈ 1− Dx+k + Dx+k+1 [interpolation]. (3.189)
p 0
p p
k=0 j =0
To proceed further we utilize the well-known formula for the sum of an arithmetic progression:
p−1
X p(p − 1)
j 0 = 0 + 1 + . . . + (p − 1) = , (3.190)
2
j 0 =0
to obtain
∞
1X p(p − 1) p(p − 1)
Nx(p) ≈ p− Dx+k + Dx+k+1 (3.191)
p 2p 2p
k=0
∞ ∞
p+1X p−1X
= Dx+k + Dx+k+1 (3.192)
2p 2p
k=0 k=0
p+1 p−1
= Nx + Nx+1 (3.193)
2p 2p
p+1 p−1
= Nx + (Nx − Dx ) (3.194)
2p 2p

p−1
= Nx − Dx . (3.195)
2p
104
Substituting into (3.184) then yields
(p)
Nx Nx p − 1
ä(p)
x = ≈ − (3.196)
Dx Dx 2p
or, the oft-quoted form,
p−1
ä(p)
x ≈ äx −
. (3.197)
2p
Obviously, via (3.183), this can be used to give an approximation for the E.P.V. of a whole-life
assurance paid p-thly.
Note also that, on taking the limit p → ∞, we have
1
āx ≈ äx − , (3.198)
2
giving, via (3.185), a corresponding approximation for Āx .
Exercise 3.5.1: Approximation for p-thly immediate annuity
Combine (3.197) with (3.89) to show
p−1
ax(p) ≈ ax + . (3.199)
2p
Example 3.5.1: Death benefit at end of month of death

Prof. Smith, age 40, takes out a whole-life assurance with sum assured of £50,000 payable at the end of the
month of death. What is the cost of this assurance? Assume a 4% p.a. effective interest rate and select
mortality of tables AMC00. If he pays for the assurance monthly in advance for life, what is the monthly
premium?
Solution
Assuming, as always, no expenses then:
(12)
E.P.V. of benefit = 50000A[40] (3.200)
h i
(12)
= 50000 1 − d(12) ä[40] (3.201)
h i
(12)
= 50000 1 − 12(1 − v 1/12 ) × ä[40] [see (3.180)]. (3.202)
Now
(12) 12 − 1
ä[40] ≈ ä[40] − (3.203)
2 × 12
N[40] 11
= − (3.204)
D[40] 24
42052.51 11
= − (3.205)
2056.56 24
= 19.989653 (to 8 s.f.). (3.206)
Substituting (3.206) in (3.202) and using v = 1/(1 + i) = 1/1.04 we finally have
E.P.V.of benefit = 10863.57 (to 7 s.f.) (3.207)
So the cost of the assurance is £10863.57 (to the nearest penny).
Now let the monthly premium be £P , then the equation of value is:
(12) (12)
12P ä[40] = 50000A[40] (3.208)
and, hence,
10863.57
P = (3.209)
12 × 19.989653
= 45.29 (to the nearer £0.01). (3.210)
So the monthly premium is £45.29.
105
Exercise 3.5.2: Which prize?
A competition prize can be taken either as a lump sum of £10,000 or as payments of £50 per month in arrears
for life. If the annual effective interest rate is 4% and mortality is that of the ultimate values of AMC00,
which prize should be chosen by a 40-year-old?
[Answer: The monthly payments since they have E.P.V. £11,941.]
Alternative approximation for A(p)

x
(p)
An alternative approach to calculating Ax is to consider the timing of the payment of the death
benefit and compare this to payment at the end of the policy year as measured in Ax .
Assuming a Uniform Distribution of Deaths, then p1 of deaths in each year of age will occur in
1 p−1
each p-thly interval. Payments in the first p of a year will be paid p before the end of year of
1 p−2
death, payments in the second of a year will be paid
p before the end of year and so on. The
p
average acceleration of benefits, relative to end of policy year, is given by:
p
1Xp−j
Average acceleration of benefit = (3.211)
p p
j=1
p−1
= by summing the arithmetic series. (3.212)
2p
This leads to the common approximation:
p−1
A(p)
x ≈ (1 + i)
2p Ax (3.213)
Taking the limit as p → ∞ we obtain:
1
Āx = (1 + i) 2 Ax (3.214)
Summary of 3.5
Under linear interpolation on Dx :
(p) p−1
Approximate relationship for E.P.V.s of life annuities-due: äx ≈ äx − 2p
(p) p+1
Approximate relationship for E.P.V.s of life annuities-in arrears: ax ≈ äx − 2p
(p) (p)
Approximate relationship for E.P.V.s of life assurance payable p − thly: Ax ≈ 1 − d(p) äx
p−1
(p)
Ax ≈ (1 + i) 2p Ax
3.6 Loss, policy value, prospective reserve and surrender values

In this important section, we will formalize the notion of loss and introduce the idea of policy value
and its importance when an individual wants to “surrender” their policy. We will concentrate on
the case of fully-discrete whole-life assurance which means death benefit payable at the end of
the year of death and premiums paid annually in advance for life. However, the underlying method
is easily extended to other policy types.
Using the notation Px for the annual premium of fully-discrete whole-life assurance with unit
death benefit,2 we have the Equation of Value
Ax = Px äx , (3.215)
2
In actuarial work, subscripts/superscripts are often added to P to indicate exactly what type of policy the premiums
correspond to. However, in this course, we shall make only limited used of such notation.
106
which, combined with the conversion relationship
Ax = 1 − däx , (3.216)
yields
1
Px = − d. (3.217)
äx
We shall use both (3.216) and (3.217) as we examine the loss, policy value and surrender value for
fully-discrete whole-life assurance.
Loss for a Contract with Unit Death Benefit

The present value, at the time the policy is taken out, of the insurance company’s loss on the
contract to (x) is a random variable given by
Lx = v K(x)+1 − Px äK(x)+1 , (3.218)
i.e., the difference between the P.V. of the death benefit and the P.V. of the premiums. If Lx > 0
the insurer loses money while if Lx < 0 the insurer gains. Note that
E(Lx ) = 0 (3.219)
since that was just the condition used to determine the premiums. Now the values which can be
taken by Lx are obviously3
lk = v k+1 − Px äk+1 (k = 0, 1, 2, ...). (3.220)
Substituting in the formula for äk+1 we have
1 − v k+1

k+1
lk = v − Px (3.221)
1−v

Px Px
= 1+ v k+1 − (3.222)
1−v 1−v

Px Px
= 1+ v k+1 − . (3.223)
d d
A moment’s inspection of this expression reveals:
• lk is monotone decreasing with k. [Remember that v < 1.]
• The k = 0 case gives

l0 = v − Px > 0. (3.224)
[To see the inequality, note from (3.217) that v − Px = 1 − 1/äx where äx = E(1 + v + v 2 +
. . . + v K(x)+1 ) > 1.]
• Taking the limit k → ∞ gives
Px
l∞ = lim lk = − < 0. (3.225)
k→∞ d
3
This is unfortunate duplication of notation; lk here has nothing to do with the life-table function lx .
107
Taken together, these three facts imply that there exists a unique k0 such that lk > 0 for all k < k0
and lk ≤ 0 for all k ≥ 0. In other words, if (x) does not survive to age x + k0 , the insurer loses on
the contract to (x).
We know that the expectation of the present value of the loss is zero, but what about the
variability of the losses? The variance of Lx is easily calculated as follows:
Var(Lx ) = Var(v K(x)+1 − Px äK(x)+1 ) (3.226)

Px K(x)+1 Px
= Var 1 + v − [cf. (3.223)] (3.227)
d d
2
Px
= 1+ Var(v K(x)+1 ) (3.228)
d
!2
d + ä1x − d
= [A∗x − (Ax )2 ] [see (3.53)] (3.229)
d
A∗x − (Ax )2
= . (3.230)
(däx )2
where, as usual, the superscript ∗ indicates evaluation at v ∗ = v 2 .
108
Exercise 3.6.1: *Variability of losses
Show that däx ≤ 1 and hence that
Var(Lx ) ≥ A∗x − (Ax )2 . (3.231)
Note that A∗x − (Ax )2 is exactly the variance of the present value of the loss for the case where whole-life
assurance is purchased by a single down-payment rather than a series of premiums. Why does this case
minimize the variability of the losses?
Policy value or Prospective Reserve

Although the expected value of the future loss at the time the policy is taken out (i.e., the expected
present value) is zero by construction, the expected value of the future loss at some later time is
expected to be non-zero. To explore this issue further we define m Lx as the value of the prospective
(i.e., future) loss to the insurer m years after the initiation of a fully-discrete whole-life assurance
for (x), with unit death benefit. m Lx is a random variable which only makes sense if K(x) ≥ m
(i.e., (x) survives to age x + m); it is given by
m Lx = v K(x)+1−m − Px äK(x)+1−m . (3.232)
Here the first term is the value, m years after the start of the policy, of the death benefit while the
second term is the value, m years after the start of the policy, of the future premiums.
The conditional expectation of m Lx given that K(x) ≥ m is denoted by m Vx and called the
policy value m years after the initiation of the policy (provided the policyholder is still alive).
To be precise, the notation m Vx , indicates the policy value for a fully-discrete whole-life assurance
for (x), with unit death benefit. By definition, we have
m Vx = E(m Lx |K(x) ≥ m) (3.233)

K(x)+1−m
= E(v − Px äK(x)+1−m | K(x) ≥ m) (3.234)
= E(v K(x)+1−m |K(x) ≥ m) − Px E(äK(x)+1−m | K(x) ≥ m) (3.235)
= Ax+m − Px äx+m (3.236)

1
= (1 − däx+m ) − − d äx+m [using (3.216) and (3.217)] (3.237)
äx
äx+m
=1− . (3.238)
äx
It’s useful to remember this last form but probably more important to understand (3.236) and be
able to derive (3.238) from it. Remembering the equation in word form can also help understanding,
and application to different policy types:
Policy Value at time m = Present Value of Future Benefits − Present Value of Future Premiums
(3.239)
An analogous analysis can be carried out for other policy types. For example, consider an n-year
endowment assurance policy (with unit death and survival benefits) paid for by annual premiums of
Px:n in advance during the term of the policy, contingent on survival. In this case, the policy value
m years after the initiation (with m < n) is just
m Vx:n = Ax+m:n−m − Px:n äx+m:n−m .

Exercise 3.6.2: *Policy value for n-year endowment policy
Derive an expression for Px:n and hence show that (3.6) can be written as
äx+m:n−m
m Vx:n =1− .
äx:n
109
Surrender value
Suppose that you are paying annual premiums for a whole-life assurance policy but then your cir-
cumstances change and you decide to stop paying the premiums and cancel the policy. You will
no longer receive the agreed death benefit but you might reasonably expect some kind of a refund
to reflect the money you have already paid in. This refund is called the surrender value. Since
we always assume no expenses, the surrender value is exactly the same as the policy value. So,
for a fully-discrete whole-life assurance to (x) with unit death benefit, the amount the policyholder
receives on termination of the contract after m complete years is m Vx . In practice of course, the
life insurance company will try to make a profit by paying out less than this!
Example 3.6.1: Surrender value
Robert Lee takes out a whole-life assurance on his 50th birthday with a sum assured of £20,000 (payable at
the end of the year of death), with premiums payable annually in advance. Suppose that he is still alive at
the age of 60. Find the surrender value of his policy at that time. Use select values from the AMC00 table
with a 4% p.a. effective interest rate.
Solution
Note that the select period is only 2 years, so for all variables after age 52 ultimate tables should be used.
The first step is to calculate the net annual premium (P ) payable for the contact. This is found from:
E.P.V.of Premiums = E.P.V. of Benefits (3.240)

P ä[50] = 20000 × A[50] (3.241)
A[50]
⇒ P = 20000 (3.242)
ä[50]
0.30564
P = 20000 (3.243)
18.053
= 338.60 to the nearer £0.01 (3.244)
So, the annual premium is £338.60.

Ignoring expenses, the surrender value (in pounds) is given by:
= Present Value of Future Benefits − Present Value of Future Premiums (3.245)

= 20000 × A60 − 338.60 × ä60 (3.246)
= 20000 × 0.42834 − 338.60 × 14.863 (3.247)
= 3534.19 (to the nearer £0.01). (3.248)
So the surrender value is £3534.19(to nearest penny). Note that in this example, we were able to use vales
of Ax and äx directly from the life table. This will not not be the case for more realistic questions.
The answer could have been derived more rapidly from 3.238 as surrender value = 20000(1 − ää[50]
60
) (check
this for yourself). However, if you can’t remember the formula or with a more complex scenario, then it can
be better to calculate the premium and then use (3.236).
Paid-up value
If a policyholder stops paying premiums on a whole of life or an endowment assurance contract, it is
common for the policyholder to receive a paid-up (or PUP) benefit. This option can sometimes be
chosen as an alternative to a surrender value. The term ”paid-up” indicates that no further premiums
are payable. Ignoring expenses, the benefit under the paid-up contract is found by equating the value
of the new policy to the surrender value payable; e.g. the surrender value is used to purchase the
paid-up life assurance.
Example 3.6.2: Paid-up whole-life assurance
Find the death benefit if Robert Lee’s surrender value (see Example 3.6.1) is used to provide a paid-up
whole-life assurance (again with benefit paid at the end of the year of death).
110
Solution
Letting the death benefit be £S then the Equation of Value is
3534.19 = SA60 , (3.249)
and hence
3534.195
S=
A60
3534.195
=
0.42834
= 8250.90 (to 6 s.f.).
So the surrender value would purchase £8,250.90 (to nearest penny) of fully-paid up life assurance. As
expected, this is considerably less than the original death benefit of £20,000.
Summary of 3.6
Policy Value or Prospective Reserve
= Present Value of Future Benefits
- Present Value of Future Premiums
Policy value m years after initiation of
äx+m
a) a whole life policy: m Vx = Ax+m − Px äx+m = 1 − äx
äx+m:n−m
b) an n year Endowment Assurance: m Vx:n = Ax+m:n−m − Px:n äx+m;n−m = 1 − äx:n
(Surrender value same as policy value)
3.7 Retrospective accumulation and reserves

Retrospective accumulation
We have been considering cash flows prospectively, that is using present values. We can also consider
them retrospectively.
Conside a group of N lives, each of age (x) and each of which purchases the same financial
product (insurance or annuity). We imagine benefits accrueing to the group and saved in a bank at
interest rate i. What is the expected share of the common fund per survivor?
Example 3.7.1:
Suppose each member takes out a pure endowment of term n paying one unit of money. After n uears, tere
are N1 survivors and the total benefit is N1 . Therefore the total benefit per survivor is N1 /N1 = 1, assuming
that N is large enough so that P (N1 = 0) is small. In this simple example the interest rate i has no effect
and the total benefit per survivor is not random.
In general, suppose the number of survivors to x + n is N1 and that the accumulated fund is the
random variable F (N ). The retrospective accumulation R.A. of the financial product is defined
to be
F (N )
R.A. = lim .
N →∞ N1
By the Strong Law of Large Numbers, limN →∞ F (N )

N = E(F (1)), which is the expected benefit for
one person. The Strong Law of Large Numbers also gives limN →∞ NN1 = n px . Therefore
F (N )/N E(F (1))

R.A. = = .
N1 /N n px
111
Suppose the group takes out n-year life assurance. Then,
(1 + i)n−(k+1) if K(x) = k, where 0 ≤ k ≤ n − 1;

F (1) =
0 else.
Therefore,
n−1
X
E(F (1)) = (1 + i)n−(k+1) k|1 qx
k=0
n−1
X
= (1 + i) n
(1 + i)−(k+1) k|1 qx
k=0
n−1
X
= (1 + i)n v k+1 k|1 qx
k=0
= (1 + i)n Ax:n
1
and
(1 + i)n Ax:n
1
R.A. = .
n px
Suppose now that the group takes out an n year life annuity with payment in advance. Then,
(1 + i)n−(k+1) s̈k+1 if K(x) = k, where 0 ≤ k ≤ n − 1;

F (1) =
s̈n else,
(where s̈k is the accumulated value of an annuity-due of term k). Therefore,

n−1
X
E(F (1)) = (1 + i)n−(k+1) s̈k+1 k|1 qx + s̈n n px
k=0
n−1
!
X
= (1 + i)n äk+1 k|1 qx + än n px
k=0
= (1 + i)n äx:n
and
(1 + i)n äx:n
R.A. = .
n px
For n year life annuity we introduce the symbol
(1 + i)n äx:n
s̈x:n =
n px
for retrospective accumulation. Insurances do not have special symbols for retrospective accumula-
tion.
Net Premium Reserves

An insurance company will want to have cash available to pay its customers’ benefits. This available
cash is called reserves. In general, reserves will be established to eliminate any future losses (i.e.
such that the expected present value of future reserves is zero as in Section 3.6).
In this module, we consider only pet premium reserves; that is reserves calculated on the same
actuarial basis as the net premium and without any allowance for expenses or other charges.
112
The surrender value of whole life assurance payable at end of year of death given by the ex-
pression (3.236) is also called the prospective reserve because it looks into the future to estimate
future premiums and benefits. We define the retrospective reserve to be the difference between
the retrospective accumulated premiums and benefits. Thus, for whole life assurance payable at end
of year of death the retrospective reserve is given by
(1 + i)m Ax:m
1
Px s̈x:m − ,
m px
as, retrospectively, the premiums and benefits are for a term of m years.
In words, the retrospective reserve is found from:
Retrospective reserve = Accumulated value of Premiums paid−Accumulated value of Benfits provided

(3.250)
Note that while no benefits have been paid, policyholders surviving to age x + m have had the
benefit of life insurance cover.
The assumptions made when pricing financial products, such as mortaltiy are called the actuarial
basis or basis. If the basis for calculating prospective and retrospective reserves are the same, then
the reserves are also the same. We will illustrate this fact with whole life assurance payable at end
of year of death. We sill need two identities which are left as exercises:
m |äx = v m m px äx+m (3.251)
and
Ax − m px v m Ax+m = Ax:m
1
. (3.252)
We start with the prospective reserve for a whole of life assurance and show it equals the retrospective
reserve. We have
m Vx = Ax+m − Px äx+m
= Ax+m − Px äx+m − Px s̈x+m + Px s̈x+m
(1 + i)n äx:n
= Ax+m − Px äx+m − Px + Px s̈x+m
n px
(1 + i)n äx:n

= Ax+m − Px äx+m + + Px s̈x+m
n px
(1 + i)n m |äx (1 + i)n äx:n

= Ax+m − Px + + Px s̈x+m using (3.251)
n px n px
(1 + i)n
= Ax+m − Px (m |äx + äx:n ) + Px s̈x+m
n px
(1 + i)n
= Ax+m − Px äx + Px s̈x+m
n px
(1 + i)n
= Ax+m − Ax + Px s̈x+m
n px
(1 + i)m
= − (−m px v m Ax+m + Ax ) + Px s̈x+m
m px
(1 + i)m 1
= − Ax:m + Px s̈x+m using (3.252),
m px
which is the retrospective reserve.

This identity of prospective and retrospective net premium reserves is true for all contracts
providing, of course, that the actuarial bases are the same.
113
Calculation of reserves recursively
We will prove a recursive formula for reserves for whole life assurance payable at end of year of
death. By using commutation functions, one can show the identities
Ax+m = vqx+m + vpx+m Ax+m+1 (3.253)
and
äx+m = 1 + vpx+m äx+m+1 ; (3.254)
they are also intuitvely true by conditioning on what happens in the first year. We have
m Vx + Px = Ax+m − Px äx+m + Px
= (vqx+m + vpx+m Ax+m+1 ) − Px (1 + vpx+m äx+m+1 ) + Px
= v(qx+m + px+m (Ax+m+1 − Px äx+m+1 ))
= v(qx+m + px+m × m+1 Vx ).
Multiplying both sides gives the formula
(m Vx + Px )(1 + i) = qx+m + px+m × m+1 Vx .
The interpretation of the recursive formula is that the reserve at age x + m plus the premium from
that year, plus interest, equals the expected benefit to be paid in the following year plus the expected
cost of setting up the reserve for age x + m + 1.
Summary of 3.7
Retrospective Reserves
Accumulated value after n years
(1+i)n ä
x:n
of whole-life annuity due: s̈x:n = n px
.
Retrospective reserve = Accumulated value of premiums
- Accumulated value of benefits
Retrospective reserve for
m
whole of life policy : Px s̈x+m − (1+i)
m px
1
Ax:m
Recursive formula for reserves: (m Vx + Px )(1 + i) = qx+m + px+m × m+1 Vx .
3.8 Importance of variance, normal approximation

For several of the policy types in the preceding sections, we have calculated not just the mean of
the present value but also its variance. In this final section, we investigate the importance of the
variance and its use to estimate how much a life company must invest to ensure it has a good
chance of being able to pay out the agreed benefits. We work within the framework of a normal
approximation.
Central limit theorem

Let Zj , j = 1, 2, . . ., be a sequence of mutually independent and identically distributed random
variables each having mean µ and variance σ 2 . Then the Central limit theorem states that the
distribution of
Z1 + . . . + ZN − N µ
√ (3.255)
σ N
114
tends to the standard normal distribution as N → ∞. That is
u2
x
e− 2

Z1 + . . . + ZN − N µ
Z
lim P √ ≤ b = Φ(b), where Φ(x) = √ du. (3.256)
N →∞ σ N −∞ 2π
It follows from the Central Limit Theorem that, if Y is the sum of N independent and identically
distributed random variables, Y = Z1 + . . . + ZN , then
!
Y − E(Y )
P p ≤ b ≈ Φ(b) for large N . (3.257)
Var (Y )
Investment requirement for life company

Let us consider the case when N identical policies are sold by a life insurance office. Let Zj be the
present value
P of policy j, (j = 1, . . . , N ) and Y be the total present value of this block of policies,
i.e., Y = N j=1 Zj . The random variables Zj are mutually independent since they relate to different
individuals. Hence, for large N , the distribution of Y can be approximated by the Central Limit
Theorem. First we note that, if the expected value and variance of the P.V. for each individual
policy are4
E(Zj ) = E(Z), Var(Zj ) = Var(Z) = σZ2 for j = 1, 2, . . . , N , (3.258)
then
E(Y ) = N × E(Z) and Var(Y ) = N × Var(Z). (3.259)
If the life company is to have sufficient funds to make the benefit payments, it must obviously
invest at least Y units of money. However, since Y is a random variable one cannot say exactly how
much money is needed, but only speak about probabilities. To be precise, we would like to know
how much money, C, needs to be invested by the life company in order to have probability 1 − α of
generating funds to pay the benefit payments (α will typically be small). In other words, we want
to find C which satisfies the condition
P (Y ≤ C) = 1 − α. (3.260)
Now !
Y − E(Y ) C − E(Y )
P (Y ≤ C) = P p ≤ p , (3.261)
Var(Y ) Var(Y )
Y −E(Y )
and, by the Central Limit Theorem, for large N the probability distribution of √ can be
Var(Y )
approximated by the standard normal distribution [see (3.257)], so that,
! !
Y − E(Y ) C − E(Y ) C − E(Y )
P p ≤ p ≈Φ p . (3.262)
Var(Y ) Var(Y ) Var(Y )
Hence C can be approximately determined from the equation

!
C − E(Y )
Φ p ≈ 1 − α. (3.263)
Var(Y )
4
Some authors use the symbol µ for E(Z), as was done above in the definition of the Central Limit Theorem. In
general we shall refrain from using this notation in order to avoid confusion with the force of mortality.
115
Now let zα be the upper 100α percentage point of the standard normal distribution, so that Φ(zα ) =
1 − α. Then, from (3.263),
C − E(Y ) p
p ≈ zα , or, equivalently, C ≈ E(Y ) + zα Var(Y ). (3.264)
Var(Y )
Recalling that Y is the total present value of N identical policies and using (3.259), we finally obtain
√
C ≈ N E(Z) + zα N σZ . (3.265)
Note that
√ the amount that the life company needs to invest per policy is approximately E(Z) +
zα σZ / N which tends to E(Z) as N tends to infinity.
Once E(Z) and σZ are calculated, then they could just be substituted into expression (3.265)
if you have remembered it. However, it is far more important to understand the steps leading to
this formula, particularly since some questions ask you to “show your working”. An example should
illustrate the idea.
Example 3.8.1: Whole-life assurance with constant force of mortality (revisited)
Suppose that a life office is about to sell a whole-life assurance policy with £100,000 death benefit to 100
individuals, age x, who were born on the same day. The death benefits are to be paid on the moment of death
and withdrawn from an investment fund subject to a constant force of interest δ = ln(1 + i) = 0.06. Assume
that the lives assured are subject to a constant force of mortality µ = 0.04 and calculate the minimum total
amount to be invested at time t = 0, so that the probability is approximately 0.95 that sufficient funds will
be on hand to withdraw the benefit payment at the death of each individual.
Solution P100
The total P.V. of the benefit payments is Y = j=1 Zj where Zj is the P.V. of the benefit in policy j. The
2
first step is to use the results from Example 3.2.1 to obtain the mean E(Z) and variance Var(Z) = σZ of the
P.V. for each individual policy. Working in pounds, we have
E(Z) = 100000 × Āx (3.266)

µ
= 100000 × (3.267)
µ+δ
0.04
= 100000 × (3.268)
0.04 + 0.06
= 40000, (3.269)
and
Var(Z) = 1000002 Ā∗x − (Āx )2

(3.270)
" 2 #
2 µ µ
= 100000 − (3.271)
µ + 2δ µ+δ
" 2 #
0.04 0.04
= 1000002 − (3.272)
0.04 + 0.12 0.04 + 0.06
= 900000000. (3.273)
Now we require to find C such that

!
Y − E(Y ) C − E(Y )
P (Y ≤ C) = P p ≤ p = 0.95. (3.274)
Var(Y ) Var(Y )
By the Central Limit Theorem, the random variable Y is approximately normally distributed with mean
E(Y ) = 100 × E(Z) and variance Var(Y ) = 100 × Var(Z). Now the upper 5% point of the standard
normal distribution is found from statistical tables to be 1.6449, i.e., Φ(1.6449) = 0.95 so that, by definition,
116
z0.05 = 1.6449. So one either substitutes numbers directly into formula (3.265) or, if working is requested
Y −E(Y )
(or the formula forgotten!), argues as follows. Since the distribution of √ is approximately standard
Var(Y )
normal, C is approximately determined by
!
C − E(Y ) C − E(Y )
Φ p ≈ 0.95 ⇐⇒ p ≈ 1.6449, (3.275)
Var(Y ) Var(Y )
and hence
p
C ≈ E(Y ) + 1.6449 × Var(Y ) (3.276)
√
= 100 × E(Z) + 1.6449 × 100 × σZ (3.277)
= 100 × 40000 + 1.6449 × 10 × 30000 (3.278)
= 4493470. (3.279)
So the minimum amount the company needs to invest so that there is a probability of (approximately) 0.95
that it will be sufficient to pay the benefit payments incurred by the assurance policies is £4.49 million (to 3
significant figures).
Summary of 3.8
For N identical policies...
P.V. of benefits from policy j: Zj
with mean and variance: E(Zj ) = E(Z), Var(Zj ) = Var(Z) = σZ2
Y = N
P
P.V. of sum of benefits: j=1 Zj
with mean and variance: ) = N × E(Z), Var(Y
E(Y )= N × Var(Z)

Y −E(Y ) C−E(Y ) C−E(Y )
Normal approximation (for large N ): P √ ≤√ ≈Φ √
Var(Y ) Var(Y ) Var(Y )
117

MTH5124 Actuarial Mathematics I Lecture Notes

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

MTH5124 Actuarial Mathematics I Lecture Notes

Transféré par

Droits d'auteur :

Formats disponibles

Lecture Notes

MTH5124: Actuarial Mathematics I

3 Life insurance and related functions 85

0.1 What is an actuary?

0.2 About this course

0.3 About these notes

0.4 Life Tables

• English Life Table No 17,

• AMC00 mortality table for assured lives.

Full details of mortality tables published by the CMI can be found at

• S. J. Garrett An Introduction to the Mathematics of Finance (Butterworth-Heinemann);

• A. Neill Life Contingencies (Heinemann);

[Ber89] J. Bernoulli. Tractatus de seriebus infinitis. manuscript, 1689.

[BGH+ 97] N. L. Bowers, H. U. Gerber, J. C. Hickman, D. A. Jones, and C. J. Nesbitt. Actuarial

[DHW13] D. C. M. Dickson, M. R. Hardy, and H. R. Waters. Actuarial Mathematics for Life

[Gar13] S. J. Garrett. An Introduction to the Mathematics of Finance. Butterworth-Heinemann,

[MS86] J. J. McCutcheon and W. F. Scott. An Introduction to the Mathematics of Finance.

[Nei77] A. Neill. Life Contingencies. Heinemann, London, 1977.

[Pac94] L. Pacioli. Summa de Arithmetica. Venice, 1494.

1.1 Two types of interest

• the market rate for similar loans;

1.1.2 Compound interest

Exercise 1.1.2: Compound interest formula

Accumulation = £1000 × (1 + 3 × 0.015) = £1045.00. (1.6)

On the other hand, if interest is compounded, then we use (1.5):

Accumulation = £1000 × (1 + 0.015)3 = £1045.68. (1.7)

In real-life there are many different kinds of transactions such as:

Accumulation = £10000(1 + 0.1)10 = £25937. (1.8)

1.2 Nominal and effective interest rates

1.2.1 Accumulation factor

A(t1 , t3 ) = A(t1 , t2 )A(t2 , t3 ) ∀ t1 ≤ t2 ≤ t3 . (1.9)

4. From (1.9), it follows that

A(t, t + n) = (1 + i)n (1.11)

(b) p = 365; i(365) = 0.0022625 and

A(0, 1) = (1 + i). (1.18)

On the other hand, for compounding p-thly we have

So, if the accumulations are the same, we require

i(p) = p[(1 + i)1/p − 1], (1.21)

Accumulation = P × (1 + i)n (1.26)

Exercise 1.2.2: Quarterly interest

Exercise 1.2.3: High APR

1.3 Force of interest

Exercise 1.3.1: *Trend of i(p) as function of p

Exercise 1.3.2: Limit of i(p)

1.3.2 Force of interest

δ(t) = lim i(p) (t). (1.31)

Recognising here the derivative4 of F we have

which proves the theorem, since t0 , t are arbitrary.

and the p = 1 case gives Z t+1 

Example 1.3.1: Force of interest and accumulation factor

Example 1.3.2: Force of interest and value of investment

Exercise 1.3.3: Force of interest and nominal interest rates

Exercise 1.3.4: *Stoodley’s formula

1.3.3 Special case of constant force of interest

Exercise 1.3.5: “Rule of 72”

Example 1.3.4: Quadruple your money!

Exercise 1.3.6: Limit of i(p)

Force of interest: δ(t) = lim i(p) (t) = lim A(t,t+h)−1

1.4 Rates of discount

Exercise 1.4.1: Rates of discount for differing subperiods

Special case of constant force of interest

Relation between nominal and effective rates of discount

i(2) 2[(1 + i)1/2 − 1]

and the p = 1 case gives Z t+1