MT105b Vle

Mathematics 2
M. Anthony
MT105b, 279005b
2011
Undergraduate study in
Economics, Management,
Finance and the Social Sciences
This subject guide is for a 100 course offered as part of the University of London
International Programmes in Economics, Management, Finance and the Social Sciences.
This is equivalent to Level 4 within the Framework for Higher Education Qualifications in
England, Wales and Northern Ireland (FHEQ).
For more information about the University of London International Programmes
undergraduate study in Economics, Management, Finance and the Social Sciences, see:
www.londoninternational.ac.uk
This guide was prepared for the University of London International Programmes by:
Martin Anthony, Department of Mathematics, London School of Economics and
Political Science.
This is one of a series of subject guides published by the University. We regret that due to
pressure of work the author is unable to enter into any correspondence relating to, or arising
from, the guide. If you have any comments on this subject guide, favourable or unfavourable,
please use the form at the back of this guide.
University of London International Programmes

Publications Office
Stewart House
32 Russell Square
London WC1B 5DN
United Kingdom
Website: www.londoninternational.ac.uk
Published by: University of London
University of London 2010
Reprinted with minor revisions 2011
The University of London asserts copyright over all material in this subject guide except where
otherwise indicated. All rights reserved. No part of this work may be reproduced in any form,
or by any means, without permission in writing from the publisher.
We make every effort to contact copyright holders. If you think we have inadvertently used
your copyright material, please let us know.
Contents
Contents
1 General introduction
1.1
Mathematics 1 and Mathematics 2 . . . . . . . . . . . . . . . . . . . . .
1.2
Studying mathematics . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3
Aims and objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4
Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5
How to use the subject guide
. . . . . . . . . . . . . . . . . . . . . . . .
1.6
Recommended books . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.1
Main text . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.2
Other recommended texts . . . . . . . . . . . . . . . . . . . . . .
Online study resources . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.1
The VLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.2
Making use of the Online Library . . . . . . . . . . . . . . . . . .
1.8
Examination advice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.9
The use of calculators . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7
2 Further differentiation and integration, with applications
Essential reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2
Revision of Mathematics 1 differentiation . . . . . . . . . . . . . . . . . .
2.3
Using derivatives for approximations . . . . . . . . . . . . . . . . . . . .
10
2.4
Taylors theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
2.5
Elasticities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14
2.6
The effects of taxation . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16
2.7
Revision of Mathematics 1 integration . . . . . . . . . . . . . . . . . . .
18
2.8
Definite integrals and areas
. . . . . . . . . . . . . . . . . . . . . . . . .
19
2.9
Consumer and producer surplus . . . . . . . . . . . . . . . . . . . . . . .
20
Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
Sample examination/practice questions . . . . . . . . . . . . . . . . . . . . . .
23
Answers to activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
Contents
Answers to Sample examination/practice questions . . . . . . . . . . . . . . .

3 Functions of several variables
33
33
Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.2
Partial derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.3
Homogeneous functions and Eulers theorem . . . . . . . . . . . . . . . .
34
3.4
Optimisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
37
3.4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
37
3.4.2
Unconstrained optimisation . . . . . . . . . . . . . . . . . . . . .
38
3.4.3
Applications of unconstrained optimisation . . . . . . . . . . . . .
38
3.4.4
Constrained optimisation . . . . . . . . . . . . . . . . . . . . . . .
40
41
42
43
45
4 Linear algebra and applications
ii
25
51
51
Recommended reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
4.2
Revision of Mathematics 1 . . . . . . . . . . . . . . . . . . . . . . . . . .
51
4.2.1
Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
4.2.2
Linear equations . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
4.3
Matrix inverses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
52
4.4
Inverse matrices and linear equations . . . . . . . . . . . . . . . . . . . .
53
4.5
Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
4.5.1
The determinant . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
4.5.2
Calculating determinants using row operations . . . . . . . . . . .
55
4.6
Cramers rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
4.7
The square linear system when A is not invertible . . . . . . . . . . . . .
58
4.8
Row operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
4.9
Calculating inverses using determinants . . . . . . . . . . . . . . . . . . .
62
4.10 Calculating inverses using row operations . . . . . . . . . . . . . . . . . .
64
4.11 Application: input-output analysis . . . . . . . . . . . . . . . . . . . . . .
66
Contents
4.12 Eigenvalues and eigenvectors . . . . . . . . . . . . . . . . . . . . . . . . .
67
4.13 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
4.14 Finding eigenvalues and eigenvectors . . . . . . . . . . . . . . . . . . . .
67
4.15 Diagonalisation of a square matrix
. . . . . . . . . . . . . . . . . . . . .
69
71
72
73
77
5 Differential equations
83
83
Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
5.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
5.2
Separable equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
84
5.3
Linear equations and integrating factors . . . . . . . . . . . . . . . . . .
88
5.4
Second-order equations . . . . . . . . . . . . . . . . . . . . . . . . . . . .
90
5.5
Behaviour of solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . .
92
5.6
Coupled differential equations . . . . . . . . . . . . . . . . . . . . . . . .
93
5.6.1
Reducing to a second-order equation . . . . . . . . . . . . . . . .
94
5.6.2
Using diagonalisation . . . . . . . . . . . . . . . . . . . . . . . . .
96
5.7
Applications of differential equations . . . . . . . . . . . . . . . . . . . .
99
5.8
Macroeconomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
99
5.9
Continuous cash flows . . . . . . . . . . . . . . . . . . . . . . . . . . . .
100
5.10 Continuous price adjustment . . . . . . . . . . . . . . . . . . . . . . . . .
101
5.11 Determining demand from elasticity . . . . . . . . . . . . . . . . . . . . .
102
5.12 Market trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
103
103
104
105
Solutions to Sample examination/practice questions . . . . . . . . . . . . . . .
109
6 Difference equations
117
117
Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
117
6.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
117
6.2
Revision of Mathematics 1 . . . . . . . . . . . . . . . . . . . . . . . . . .
117
iii
Contents
6.2.1
Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
117
6.2.2
Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
118
6.3
First-order difference equations . . . . . . . . . . . . . . . . . . . . . . .
118
6.4
Solving first-order difference equations . . . . . . . . . . . . . . . . . . .
120
6.5
Long-term behaviour of solution . . . . . . . . . . . . . . . . . . . . . . .
121
6.6
The cobweb model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
122
6.7
Financial applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
123
6.8
Homogeneous second-order difference equations . . . . . . . . . . . . . .
124
6.9
Non-homogeneous second-order equations . . . . . . . . . . . . . . . . . .
126
6.10 Behaviour of the solution . . . . . . . . . . . . . . . . . . . . . . . . . . .
127
6.11 Coupled difference equations . . . . . . . . . . . . . . . . . . . . . . . . .
127
6.11.1 Solving by reducing to a second-order equation
. . . . . . . . . .
127
6.11.2 Using diagonalisation . . . . . . . . . . . . . . . . . . . . . . . . .
129
6.12 Economic applications of second-order difference equations . . . . . . . .
130
132
132
134
135
A Sample examination paper
139
B Comments on the Sample examination paper
143
iv
Chapter 1
General introduction
1.1
Mathematics 1 and Mathematics 2
If you are studying this subject, you will already have studied (or be concurrently
studying) 05A Mathematics 1 or you will have a qualification which allowed you
exemption from that course due to its equivalence. This subject builds upon
Mathematics 1. Everything in Mathematics 1 is essential to Mathematics 2. So,
although Mathematics 2 is formally a separate subject, it is best thought of as an
extension of Mathematics 1. Given this, it is essential that you have a good
understanding of Mathematics 1. In this subject guide, we will briefly review some of
the important ideas and techniques from Mathematics 1 that we shall need for
Mathematics 2, but you should refer to the Mathematics 1 guide and the textbooks if
you feel the need to refresh yourself on some of the more basic Mathematics 1 topics.
In Mathematics 2, we explore further some topics introduced in Mathematics 1 and we
study some more applications to the social sciences, particularly economics. In
particular, we investigate further the applications of differentiation and integration, and
functions of several variables: we shall see some new applications and also some new
techniques.
This subject also introduces some important new topics: for example, we shall meet new
techniques for solving linear equations, and we introduce differential and difference
equations.
This half course may not be taken with 76 Management mathematics, 173
Algebra or 174 Calculus.
1.2
Studying mathematics
I make a number of points in the introductory chapter of the 05A Mathematics 1

guide about the nature of studying mathematics, and these are worth repeating here.
The study of mathematics can be very rewarding. It is particularly satisfying to solve a
problem and know that it is solved. Unlike many of the other subjects you will study, in
the mathematics subjects, there is always a right answer. Although there may be only
one right (final) answer, there could be a number of different ways of obtaining that
answer, some more complex than others. Thus, a given problem will have only one
answer, but many solutions (by which we mean routes to finding the answer).
Generally, a mathematician likes to find the simplest solution possible to a given
problem, but that does not mean that any other solution is wrong. (There may be
different, equally simple, solutions.)
1. General introduction
With mathematical questions, you first have to work out precisely what it is that the
question is asking, and then try to find a method (hopefully a nice, simple one) which
will solve the problem. This second step involves some degree of creativity, especially at
an advanced level. You must realise that you can hardly be expected to look at every
mathematics problem and write down a beautiful and concise solution, leading to the
correct answer, straight away. For obvious reasons, teachers, lecturers, and textbooks
rarely give that impression: they present the solution right there on the page or the
blackboard, with no indication of the time a student might be expected to spend
thinking or of the dead-end paths he or she might understandably follow before a
solution can be found. It is a good idea to have scrap paper to work with so that you
can try out various methods of solution. You must not get frustrated if you cant solve a
problem immediately. As you proceed through the subject, gathering more experience,
you will develop a feel for which techniques are likely to be useful for particular
problems. You should not be afraid to try different techniques, some of which may not
work, if you cannot immediately recognise which technique to use.
1.3
Aims and objectives
This half unit is designed to:

enable students to acquire further skills in the methods of calculus and linear
algebra (in addition to those in 05A Mathematics 1), as required for their use in
economics-based subjects
prepare students for further courses in mathematics and/or related disciplines.
1.4
Learning outcomes
At the end of this half course and having completed the Essential reading and activities,
you should have:
used the concepts, terminology, methods and conventions covered in the half course
to solve mathematical problems in this subject
the ability to solve unseen mathematical problems involving understanding of these
concepts and application of these methods
seen how mathematical techniques can be used to solve problems in economics and
related subjects.
1.5
How to use the subject guide
This subject guide is absolutely not a substitute for the textbooks. It is only what its
name suggests: a guide to the study and reading you should undertake. In each of the
subsequent chapters, brief discussions of the syllabus topics are presented, together with
pointers to recommended readings from the textbooks. It is essential that you use
textbooks. Generally, it is a good idea to read the texts as you work through a chapter
1.6. Recommended books
of the guide. It is most useful to read what the guide says about a particular topic, then
do the necessary reading, then come back and re-read what the guide says to make sure
you fully understand the topic. Textbooks are also an invaluable source of examples for
you to attempt.
You should not necessarily spend the same amount of time on each chapter of the guide:
some chapters cover much more material than others. The chapters are divided in the
way they are in order to group together topics on particular central themes.
The discussion of some topics in this guide is rather more extensive than others. Often,
this is not because those topics are more significant, but because the textbook
treatments are not as extensive as they might be.
Within each chapter of the guide you will encounter activities. You should carry out
these activities as you encounter them: they are designed to help you understand the
topic under discussion. Solutions to the activities are given near the ends of the
chapters, but do make a serious attempt at them before consulting the solutions.
To help your time management, the chapters and topics of the subject are converted
below into approximate percentages of total time. However, this is purely for
indicative purposes. Some of you will know the basics quite well and need to spend less
time on the earlier material, while others might have to work hard to comprehend the
very basic topics before proceeding onto the more advanced.
Chapter
2
3
4
5
6
Title
Further differentiation and integration
Functions of several variables
Linear algebra
Differential equations
Difference equations
% Time
15
10
25
25
25
At the end of each chapter, you will find a list of Learning outcomes. This indicates
what you should be able to do having studied the topics of that chapter. At the end of
each chapter, there are sample examination questions, which are based largely on past
exam questions. Some of the sample examination questions are really only samples of
parts of exam questions. Solutions are given at the ends of the chapters.
1.6
Recommended books
None of the books in the following reading list covers absolutely everything in the
syllabus. You should therefore ensure that you have access to a sufficient number of the
books. At the beginning of each chapter, appropriate references are made to the books
dealing with the material of that chapter.
The main recommended text is the book by Anthony and Biggs. This covers most of
the required material, and uses the same notations as this guide. This is the
recommended book for 05A Mathematics 1, so most of you will already have a copy.
1.6.1
Main text
Anthony, M. and N. Biggs, Mathematics for economics and finance. (Cambridge,

UK: Cambridge University Press, 1996.) [ISBN 9780521559133].1
1.6.2
Other recommended texts
Please note that as long as you read the Essential reading you are then free to read
around the subject area in any text, paper or online resource. To help you read
extensively, you have free access to the VLE and University of London Online Library
(see below). Other useful texts for this course include:
R
R
R
R
R
Binmore, K. and J. Davies, Calculus. (Cambridge, UK: Cambridge University

Press, 2001) [ISBN 9780521775410].
Bradley, T. Essential mathematics for economics and business. (Chichester:
Wiley, 2008) Third Edition. [ISBN 9780470018569].
Dowling, Edward T. Introduction to mathematical economics. Schaums Outline
Series. (New York; London: McGraw-Hill, 2000) Third Edition. [ISBN
9780071358965].
Ostaszewski, A. Mathematics in economics: models and methods. (Oxford, UK:
Blackwell, 1993) [ISBN 9780631180562].
Simon, C.P. and L. Blume, Mathematics for economists. (New York; London:
W.W. Norton and Company Ltd, 1994) [ISBN 9780393957334].
The book by Anthony and Biggs covers most, but not all, of the required material. (You
will need to refer elsewhere, as indicated in the guide, for the topics of Taylor series,
and diagonalisation and its applications in differential and difference equations.) Each
chapter of Anthony and Biggs has a large section of fully worked examples, and a
selection of exercises for the reader to attempt. Since this is the key text for 05A
Mathematics 1, it also will be useful for revision of that subject.
Binmore and Davies cover all the calculus you will need, and a lot more.
The book by Bradley covers most of the more basic material, and has plenty of worked
examples.
Dowlings book contains lots of worked examples. It is, however, less concerned with
explaining the techniques. It would not be suitable as your main text, but it is a good
source of additional examples.
The book by Ostaszewski is very suitable for a number of the topics, and provides many
examples.
The book by Simon and Blume is a large book covering everything in this subject and
also many topics outside the coverage of this subject.
There are many other books which cover the material of this subject, but those listed
above are the ones I shall refer to explicitly.
Detailed reading references in this subject guide refer to the editions of the set
textbooks listed above. New editions of one or more of these textbooks may have been
1
Recommended for purchase.
1.7. Online study resources
published by the time you study this course. You can use a more recent edition of any
of the books; use the detailed chapter and section headings and the index to identify
relevant readings. Also check the virtual learning environment (VLE) regularly for
updated guidance on readings.
The single most important point to be made about learning mathematics is that to
learn it properly, you have to do it. Do work through the worked examples in a
textbook and do attempt the exercises. This is the real way to learn mathematics. In
the examination, you are hardly likely to encounter a question you have seen before, so
you must have practised enough examples to ensure that you know your techniques well
enough to be able to cope with new problems.
1.7
Online study resources
In addition to the subject guide and the Essential reading, it is crucial that you take
advantage of the study resources that are available online for this course, including the
VLE and the Online Library.
You can access the VLE, the Online Library and your University of London email
account via the Student Portal at:
http://my.londoninternational.ac.uk
You should receive your login details in your study pack. If you have not, or you have
forgotten your login details, please email uolia.support@london.ac.uk quoting your
student number.
1.7.1
The VLE
The VLE, which complements this subject guide, has been designed to enhance your
learning experience, providing additional support and a sense of community. It forms an
important part of your study experience with the University of London and you should
access it regularly.
The VLE provides a range of resources for EMFSS courses:
Self-testing activities: Doing these allows you to test your own understanding of
subject material.
Electronic study materials: The printed materials that you receive from the
University of London are available to download, including updated reading lists
and references.
Past examination papers and Examiners commentaries: These provide advice on
how each examination question might best be answered.
A student discussion forum: This is an open space for you to discuss interests and
experiences, seek support from your peers, work collaboratively to solve problems
and discuss subject material.
Videos: There are recorded academic introductions to the subject, interviews and
debates and, for some courses, audio-visual tutorials and conclusions.
Recorded lectures: For some courses, where appropriate, the sessions from previous
years Study Weekends have been recorded and made available.
Study skills: Expert advice on preparing for examinations and developing your
digital literacy skills.
Feedback forms.
Some of these resources are available for certain courses only, but we are expanding our
provision all the time and you should check the VLE regularly for updates.
1.7.2
Making use of the Online Library
The Online Library contains a huge array of journal articles and other resources to help
you read widely and extensively.
To access the majority of resources via the Online Library you will either need to use
your University of London Student Portal login details, or you will be required to
register and use an Athens login:
http://tinyurl.com/ollathens
The easiest way to locate relevant content and journal articles in the Online Library is
to use the Summon search engine.
If you are having trouble finding an article listed in a reading list, try removing any
punctuation from the title, such as single quotation marks, question marks and colons.
For further advice, please see the online help pages:
www.external.shl.lon.ac.uk/summon/about.php
1.8
Examination advice
Important: the information and advice given here are based on the examination
structure used at the time this guide was written. Please note that subject guides may
be used for several years. Because of this we strongly advise you to always check both
the current Regulations for relevant information about the examination, and the VLE
where you should be advised of any forthcoming changes. You should also carefully
check the rubric/instructions on the paper you actually sit and follow those
instructions. Remember, it is important to check the VLE for:
up-to-date information on examination and assessment arrangements for this course
where available, past examination papers and Examiners commentaries for the
course which give advice on how each question might best be answered.
A Sample examination paper may be found at the end of this subject guide. You will
see that from 200910, all of the questions on the paper are compulsory. Any further
changes to exam format will be announced on the VLE.
It is worth making a few comments about exam technique. Perhaps the most important,
though obvious, point is that you do not have to answer the questions in any particular
1.9. The use of calculators
order; choose the order that suits you best. Some students will want to do easy
questions first to boost their confidence, while others will prefer to get the difficult ones
out of the way. It is entirely up to you.
Another point, often overlooked by students, is that you should always include your
working. This means two things.
First, do not simply write down the answer in the exam script, but explain your
method of obtaining it (that is, what I called the solution earlier).
Secondly, include your rough working. You should do this for two reasons:
If you have just written down the answer without explaining how you obtained
it, then you have not convinced the Examiners that you know the techniques,
and it is the techniques that are important in this subject. (The Examiners
want you to get the right answers, of course, but it is more important that you
prove you know what you are doing: that is what is really being examined.)
If you have not completely solved a problem, you may still be awarded marks
for a partial, incomplete, or slightly wrong, solution: if you have written down
a wrong answer and nothing else, no marks can be awarded. (You may have
carried out a lengthy calculation somewhere on scrap paper where you made a
silly arithmetical error. Had you included this calculation in the exam answer
book, you would probably not have been heavily penalised for the arithmetical
error.) It is useful, also, to let the Examiners know what you are thinking. For
example, if you know you have obtained the wrong answer to a problem, but
you cant see how to correct it, say so!
As mentioned above, you will find that, wherever appropriate, there are sample exam
questions at the end of the chapters. These are based in large part on the questions
appearing in past examination papers. As such, they are an indication of the types of
question that might appear in future exams. But they are just an indication. The
Examiners want to test that you know and understand a number of mathematical
methods and, in setting an exam paper, they are trying to test whether you do indeed
know the methods, understand them, and are able to use them, and not merely whether
you vaguely remember them. Because of this, you will probably encounter some
questions in your exam which seem unfamiliar. Of course, you will only be examined on
material in the syllabus. Furthermore, you should not assume that your exam will be
almost identical to the previous years: for instance, just because there was a question,
or a part of a question, on a certain topic last year, you should not assume there will be
one on the same topic this year. For this reason, you cannot guarantee passing if you
have concentrated only on a very small fraction of the topics in the subject. This may
all sound a bit harsh, but it has to be emphasised.
1.9
The use of calculators
You will not be permitted to use calculators of any type in the examination. This is not
something that you should panic about: the Examiners are interested in assessing that
you understand the key methods and techniques, and will set questions which do not
require the use of a calculator.
In this guide, I will perform some calculations for which a calculator would be needed,
but you will not have to do this in the exam questions. Look carefully at the answers to
the sample exam questions
to see how to deal with calculations. For example, if the
answer to a problem is 2, then leave the answer like that: there is no need to express
this number as a decimal (for which one would need a calculator).
Chapter 2
Further differentiation and integration,
with applications
Essential reading
(For full publication details, see Chapter 1.)
R
R
R
R
R
Anthony and Biggs (1996) Chapters 6, 7, 8, 9, 25 and 26.
Further reading
Binmore and Davies (2001) Section 2.13 (on Taylors theorem) and Section 10.5
(on producer and consumer surplus)
Bradley (2008) Chapters 6 and 8.
Dowling (2000) Chapters 3, 4, 16 and 17.
Ostaszewski (1993) Chapters 10 and 14.
Simon and Blume (1994) Sections 3.5, 3.6 and 20.1.
2.1
Introduction
You will have studied 05A Mathematics 1 or its equivalent, and be familiar with
differentiation and integration. In this chapter of the subject guide we will look at some
more applications of both differentiation and integration.
2.2
Revision of Mathematics 1 differentiation
From Mathematics 1, you should remember the meaning of the derivative, and know
how to calculate derivatives. You should also know about some of the applications of
the derivative.
Just to remind you, here are some of the important things we learned in Mathematics 1.
The derivative of a function f (x) at a point a is the instantaneous rate of change of
the function at a. Formally,
f (a + h) f (a)
.
h0
h
f 0 (a) = lim
2. Further differentiation and integration, with applications
We have the following standard derivatives:

f (x)
xk
ex
ln x
sin x
cos x
f 0 (x)
kxk1
ex
1/x
cos x
sin x
We have the following rules for calculating derivatives

The sum rule: If h(x) = f (x) + g(x), then
h0 (x) = f 0 (x) + g 0 (x).
The product rule: If h(x) = f (x)g(x), then
h0 (x) = f 0 (x)g(x) + f (x)g 0 (x).
The quotient rule: If h(x) = f (x)/g(x) and g(x) 6= 0, then
h0 (x) =
f 0 (x)g(x) f (x)g 0 (x)

.
g(x)2
The composite function rule: if f (x) = s(r(x)), then

f 0 (x) = s0 (r(x))r0 (x).
The derivative and second derivative can be used in optimisation and in
curve-sketching. Particular applications in economics include profit-maximisation.
2.3
Using derivatives for approximations
Another way of looking at the definition of f 0 is to think of it as an approximation,1

which tells us how a small change in the input x affects the output f (x). If we denote a
small change in x by x, then the resulting change in f (x) is
f = f (x + x) f (x).
Since f 0 (x) is the limit of f /x as x approaches zero, for small values of x we have
f 0 (x) '
f
,
x
or f ' f 0 (x)x,
where the symbol ' means is approximately equal to. In d notation (in which the
derivative is denoted by df /dx rather than f 0 ), we have
f '
1
See Anthony and Biggs (1996) Section 6.1.
10
df
x.
dx
2.3. Using derivatives for approximations
We made mention of this idea when we discussed marginal cost in Mathematics 1, and
also when we looked at the meaning of the Lagrange multiplier.
A couple of examples will serve to illustrate this simple use of the derivative for
approximations.
Example 2.1 Let us use the derivative to find the approximate change in the
function f (x) = x4 when x changes from 3 to 3.005.
The derivative of f is f 0 (x) = 4x3 and we therefore have the approximation
f ' f 0 (3)x = 4(3)2 0.005 = 0.540.
The actual value of the change is
f = f (3.005) f (3) = (3.005)4 34 = 0.54135,
so our approximation is correct to two decimal places.
Example 2.2 Suppose the demand equation for a good is p3 q = 8000 where q is
the number of units demanded (in thousands per week) and p is the price per unit,
in dollars. If p is increased from $20 to $21, what will be the approximate fall in
sales? If, on the other hand, production were to be increased from 1000 units per
week to 1100, what would be the approximate fall in price?
The demand function and its derivative are
q D (p) = 8000p3 ,
(q D )0 (p) = 8000 (3)p4 =
24000
.
p4
Therefore when p = 20 and p = 1 we have

q ' (24000/p4 )p = (24000/204 )(1) = 0.15.
Remembering that q is measured in thousands of units, it follows that 150 fewer
units will be sold per week.
For the second question we have to consider p as a function of q, and so we need the
inverse demand function and its derivative:
pD (q) = 20q 1/3 ,
(pD )0 (q) = 20 (1/3)q 4/3 .
So when q = 1 and q = 0.1 we have

p ' (20/3)q 4/3 0.1 = 2/3.
The conclusion is that the price falls by about 67 cents.
Activity 2.1
Suppose that the demand for a good is given by the equation

p2 q = 6000,
where q is the quantity (in thousands) and p is the price in dollars. If the price is
increased from $20 to $21, what is the approximate fall in expected sales?
11
2.4
Taylors theorem
We have just seen that the derivative may be used to give an approximation to a small
change in value of a function. A more precise result along these lines is Taylors
theorem. This relates the function value not just to the derivative, but higher-order
derivatives (second-order, third-order, and so on).
Under certain circumstances, functions can be approximated well by certain
polynomials, referred to as power series.2 The appropriate way to do this is by using
Taylor series. To explain this, we first introduce a piece of notation. The derivative of a
00
function f is denoted f 0 and the second derivative is denoted by f . If we differentiate
the second derivative, we obtain the third derivative f (3) , and if we differentiate this we
obtain the fourth derivative f (4) , and so on. In general, we have the nth derivative f (n) .
Taylors theorem states that if f is a function which can be differentiated n times, then,
for certain ranges of x, the approximation
x3
xn
x2
+ f 000 (0) + + f (n) (0) ,
2!
3!
n!
is valid, where n! = n(n 1)(n 2) . . . 2 (called n factorial) is the product of all the
positive integers up to and including n. (Generally, the larger n, the better the
approximation.) The right-hand side of this approximation is known as the Taylor series
for f . Strictly speaking, the stated approximation is Taylors theorem about 0, often
called Maclaurins theorem, and the right-hand side is the Maclaurin series. The
right-hand side is known as a power series expansion of the function f .
f (x) ' f (0) + f 0 (0)x + f 00 (0)
A more general form of Taylors theorem is often useful. This states that (again, with
certain qualifications on the set of x for which the approximation is true), we have
(x a)n
(x a)2
+ + f (n)
.
2!
n!
This is called the Taylor expansion of f about a. Taking n = 1 we obtain
f (x) ' f (a) + f 0 (a)(x a) + f 00 (a)
f (x) ' f (a) + f 0 (a)(x a).

Writing x = a + h, and f = f (x) f (a), this becomes the simple rule we met earlier:
f ' f 0 (a)x.
Thus, Taylors theorem, for n > 1, is a generalisation of our simple approximation rule
and will, in general, be more accurate. When we refer simply to Taylors theorem or
Taylor series we shall mean, for the sake of simplicity, those about 0 (that is, Maclaurins
theorem and Taylors theorem): so, unless it is otherwise specified, we take a = 0.
Example 2.3 The exponential function f (x) = ex has all its derivatives equal to
ex . Since e0 = 1, we therefore have that, for all n, f (n) (0)/n! = 1/n! and hence
ex ' 1 + x +
x2 x 3
xn
+
+ + .
2!
3!
n!
In fact, this approximation is valid for all x (though it requires mathematics outside
the content of this subject to explain why).
2
See, for example, Ostaszewski (1993) Sections 14.114.2, and Binmore and Davies (2001) Section
2.13, for discussion of Taylors theorem.
12
2.4. Taylors theorem
Other important Taylor series are as follows:

ln(1 + x) ' x
valid for 1 < x 1,
x2 x3 x4
+
+ ...,
2
3
4
sin x ' x
x3 x5 x7
+
+ ...,
3!
5!
7!
cos x ' 1
x2 x4 x6
+
+ ...,
2!
4!
6!
valid for all x, and
valid for all x.

You should convince yourself that the Taylor series for these functions are as stated, by
calculating the derivatives. These standard Taylor series should be remembered.
The power series expansions for more complicated functions can often be determined by
using the standard power series given above, as the following example illustrates.
Example 2.4
We expand, as a power series up to x4 , the function

f (x) = cos (ln(1 + x)) .
This means that we find the Taylor approximation with n = 4. One approach is to
calculate the first, second, third, and fourth derivatives of f and use Taylors
theorem directly. However, well take a different approach. (It is a good exercise to
check that the other approach gives the same answer.) Well use the facts that
ln(1 + x) ' x
and
cos y ' 1
x 2 x3 x4
+
,
2
3
4
y2 y4
+ .
2!
4!
It follows from these, by taking

y = ln(1 + x) ' x
x2 x3 x4
+
,
2
3
4
that
cos (ln(1 + x)) = cos y
1
1
' 1 y2 + y4
2
24

2
1
x2 x3 x4
'1
x
+
2
2
3
4

4
x2 x3 x4
1
x
.
+
+
24
2
3
4
13
Because we only need the terms involving x4 or lower powers of x, the only relevant
terms we get from the y 2 /2 term are given by
2
x2 x3 x4
+
2
3
4

2
3
x
x
x4
x2 x3 x4
+
x
+
2
3
4
2
3
4
!
2
2 2
3
x
x
x
(x)(x) 2
(x) +
+2
(x) +
2
2
3

1
x
2

1
=
x
2
=
1
2
x2 x3 11 4
+ x + ,
2
2
24
where indicates terms which involve powers of x greater than 4 (and terms
which we can therefore ignore).
Similarly, the only relevant term we get from the y 4 /24 term is x4 /24. To explain
why this is so, note that
4

x2 x3 x4
+
,
x
2
3
4
is the bracketed expression

x2 x3 x4
+
x
,
2
3
4
multiplied by itself four times. The terms which arise from this product are obtained
by multiplying together four objects, one from each occurrence of the bracketed
expression. Since the term with lowest power of x in each bracket is x, it is only by
taking the x from each bracket that we obtain a term with a degree of no more than
4; and since all terms of degree greater than 4 are to be ignored, this is therefore the
only term we consider from y 4 /24.
Therefore, putting this all together, we have

2
x3 11 4
x4
x
+ x +
cos (ln(1 + x)) ' 1
2
2
24
24
3
2
x
x
5
+
x4 ,
=1
2
2
12
and this is the required expansion.
Activity 2.2 By taking x = 0.1 in the expansion just obtained, find an
approximate value for cos(ln(1.1)).
2.5
Elasticities
When the price of a commodity increases from p to p + p, there is a change in the

quantity demanded (usually a decrease). Suppose that the quantity demanded changes
14
2.5. Elasticities
from q to q + q (where q can be negative). Then, q = q D (p) and

q + q = q D (p + p). It follows that
q = q D (p + p) q D (p) ' (q D )0 (p)p.

The relative change in quantity is q/q and the relative change in price is p/p. The
ratio of these is
(q/q)
p q
p
=
' (q D )0 (p).
(p/p)
q p
q
Since, generally, q will be negative, and since p > 0,
p q
,
q p
will be positive. We define the point elasticity of demand to be

=
p dq
,
q dp
where q denotes the demand quantity q D (p).

For a typical demand function, the elasticity of demand is positive. It should be noted
that many texts omit the negative sign in the definition of elasticity of demand. The
demand is said to be elastic if the elasticity is greater than 1, and inelastic if the
elasticity is less than 1.
Example 2.5 Suppose that the demand curve has equation q = 2 p. Then the
point elasticity of demand is
=
p dq
p
p
= (1) = .
q dp
q
q
This can be written solely in terms of q as (2 q)/q or in terms of p as p/(2 p).

Example 2.6 Suppose that the demand function is given by q = 5/p2 . Then
dq/dp = 10/p3 and so the elasticity of demand is

p 10
p
10
p dq
=
=
= 2.
=
q dp
q
p3
(5/p2 )
p3
This is a constant, not depending on p or q.
Suppose a firm is producing a particular good, and let us think about how its total
revenue changes if the selling price increases. If the price p rises, the quantity sold q
falls; but the revenue R = qp is the product of these two things, and it may rise or fall.
We shall assume that a price rise applies uniformly to the entire supply of the good
under consideration, which would certainly be the case if the firm is a monopoly. Using
the product rule to differentiate total revenue T R = qp with respect to p, remembering
that q is a function of p, we get
T R0 = (qp)0 = q 0 p + q.
15
Thus the condition that total revenue increases, T R0 > 0, is equivalent to q 0 p + q > 0,
which is equivalent to
q0p
< 1,
q
or < 1. Thus the elasticity determines whether revenue increases or decreases as price
increases:
If < 1 (that is, demand is inelastic), then a small increase in price results in an
increase in revenue.
If > 1 (that is, demand is elastic), then a small increase in price results in an
decrease in revenue.
The point elasticity of supply is defined in a way similar to that used to define point
elasticity of demand, but using the supply function q S rather than q D , and omitting the
negative sign. That is, the point elasticity of supply is given by
p S 0
p dq S
(q ) (p) =
.
q
q dp
2.6
The effects of taxation
We now look at what happens in a market when a good is taxed. (This is not entirely a
differentiation topic, but this is as appropriate a chapter as any in which to place this
topic.)
When a fixed amount of tax is imposed on each unit of a good, we refer to this as an
excise tax or per-unit tax. The following example illustrates how we can determine the
new equilibrium price and quantity in the presence of such a tax.
Example 2.7
by
Suppose that the demand and supply functions for a good are given
15
p 10.
2
Then we can easily determine (by solving the equation q = q D (p) = q S (p)) that the
equilibrium price is p = 4. Suppose that the government imposes an excise tax of T
per unit. How does this affect the equilibrium price?
q D (p) = 40 5p
and
q S (p) =
The answer is found by noting that, if the new selling price is p, then, from the
suppliers viewpoint, it is as if the price were p T , because the suppliers revenue
per unit is not p, but p T . In other words the supply function has changed: when
the tax is T per unit, the new supply function q ST is given by
q ST (p) = q S (p T ) =
15
(p T ) 10.
2
Of course the demand function remains the same. Let us use q T and pT to denote
the new equilibrium values in the presence of the tax T . Then q T and pT satisfy the
equations
15
q T = 40 5pT and q T = q ST (pT ) = (pT T ) 10.
2
16
2.6. The effects of taxation
Eliminating q T we get
40 5pT =
15 T
(p T ) 10.
2
Rearranging this equation, we obtain

15
15 T
p = 50 + T,
5+
2
2
and so we have a new equilibrium price of
3
pT = 4 + T.
5
For example, if T = 1, the equilibrium price rises from 4 to 4.6. Unsurprisingly, the
selling price has risen (and, as you will discover from the Activity below, the
quantity sold has decreased). But note that, although the tax is T per unit, the
selling price has risen not by the full amount T , but by the fraction 3/5 of T . In
other words, not all of the tax is passed on to the consumer.
Activity 2.3
Show that the new equilibrium quantity is q T = 20 3T .
If we encountered a percentage tax rather than a per-unit tax, a similar sort of analysis
would apply. Again, the demand equation would remain unaltered, but the supply
equation would change: we would replace p by p(1 t) if the tax is 100t% of the selling
price. (This is because, from the suppliers point of view, the revenue obtained from
each item in other words, the effective price is not p, but p minus 100t% of p,
which is p(1 t).)
Suppose that a government wishes to raise revenue by imposing an excise tax on a

good. Clearly, a small tax will bring in little revenue but, on the other hand, if the tax
is too large consumption will fall dramatically and the revenue will also be hit. We can
often use differentiation to determine which level of tax will maximise the tax revenue
to the government. Suppose that, as in the example above, we have demand and supply
functions given by
q D (p) = 40 5p
and
q S (p) =
15
p 10.
2
We have seen that the equilibrium price and quantity in the presence of an excise tax T
are
3
pT = 4 + T
and
q T = 20 3T.
5
T
(The determination of q was Activity 2.3.) The revenue R(T ) is the product of the
quantity sold q T and the excise tax T . That is,
R(T ) = q T T = (20 3T )T = 20T 3T 2 .
To find the value of Tm where this is a maximum, we first set R0 (T ) = 0. We have
R0 (T ) = 20 6T , so that 10/3 is the only critical point. The second derivative is
R00 (T ) = 6, which is negative, so Tm = 10/3 is the maximum point. The maximum
revenue is R(Tm ) = 100/3.
17
2.7
Revision of Mathematics 1 integration
We now very briefly review the key topics in integration that were covered in
Mathematics 1.
Suppose the function f is given, and the function F is such that F 0 (x) = f (x). Then we
say that F is an anti-derivative of f . Any two anti-derivatives of a given function f
differ only by a constant. The general form of the anti-derivative of f is called the
indefinite integral of f (x), and denoted by
Z
f (x) dx.
Often we call it simply the integral of f . It is of the form F (x) + c, where F is any
particular anti-derivative of f and c is an arbitrary constant, known as a constant of
integration. The process of finding the indefinite integral of f is usually known as
integrating f , and f is known as the integrand.
Just as for differentiation, we shall have a list of standard integrals and some rules for
combining these.
R
f (x)
f (x) dx
n+1
x
+c
xn (n 6= 1)
n+1
1/x
ln |x| + c
ex
ex + c
sin x
cos x + c
cos x
sin x + c
Note that the integral of 1/x is ln |x| + c rather than ln x + c because one cannot take
the logarithm of a negative number.
We also have the simple rules
Z
Z
Z
(f (x) + g(x)) dx = f (x) dx + g(x) dx,
for any functions f and g, and
Z
Z
kf (x) dx = k
f (x) dx,
for any constant k.

If f is a function with an anti-derivative F , then the definite integral 3 of the function f
over the interval [a, b] is defined to be
Z b
f (x) dx = F (b) F (a).
a
In Mathematics 1, we met three important techniques for integration. First, we have the
substitution or change of variable method. Formally, this uses the fact that when we
3
18
2.8. Definite integrals and areas
change the variable by putting x = x(u), we have

Z
Z
f (x) dx = f (x(u)) x0 (u)du.
The rule for integration by parts is:

Z
Z
0
u (x)v(x) dx = u(x)v(x) u(x)v 0 (x) dx.
The third technique is partial fractions. This involves rewriting integrands of the form
p(x)/q(x), where p and q are polynomials, in a simpler form which makes them easier to
integrate. If p(x) is linear (that is, of the form ax + b) and q(x) is a quadratic with two
different roots, then the method of partial fractions applies. Suppose that
q(x) = (x a1 )(x a2 ), where a1 6= a2 and C is some number. Then it is possible to
write
p(x)
A1
A2
p(x)
=
=
+
,
(?)
q(x)
(x a1 )(x a2 )
x a1 x a2
for some numbers A1 and A2 . Cross-multiplying equation (?), we get
p(x) = A1 (x a2 ) + A2 (x a1 ).
The numbers A1 and A2 may be found by substituting x = a1 , x = a2 in turn into this
identity. When p(x)/q(x) is expressed in this way, it is easy to evaluate the integral. We
have

Z
Z
p(x)
A1
A2
dx =
+
dx
q(x)
x a1 x a2
= A1 ln |x a1 | + A2 ln |x a2 | + c.
2.8
Definite integrals and areas
There is a useful relationship between the definite integral and the area under a curve.
Suppose that f (x) 0 for x in the interval [a, b]. Then the area enclosed by the curve
Rb
y = f (x), the x-axis and the vertical lines x = a and x = b is equal to a f (x) dx.
The same result is essentially true even when the function is negative on part of the
interval, but in this case the area enclosed by that part of the graph of the function and
the x-axis is assigned a negative sign.
Example 2.8
What is the area under the curve y = x2 between x = 1 and x = 2?
To answer this, we first note that an anti-derivative of x2 is x3 /3. So the area is

Z
1
x3
x dx =
3
2
2
1
8 1
7
23 13
=
= = .
3
3
3 3
3
19
Activity 2.4 What is the area enclosed by the curve with equation y = sin x, the
x-axis, the y-axis and the line x = ?
2.9
Consumer and producer surplus
Figure 2.1 illustrates typical demand and supply curves for a good. Here, the
equilibrium price, p , and quantity, q , are also indicated.
p
Supply
p
Demand
q
Figure 2.1: Typical supply and demand curves
At equilibrium, the consumers buy q units of the good at price p per unit, and hence
the total amount they pay is p q . This is the area denoted by R in Figure 2.2
p
p
R
Demand
q
Figure 2.2: For q items at price p , consumers pay only p q , the area R
However, it can be argued4 that the total value the consumers place on q items of the
good is the area A of Figure 2.3, the area of the region that is bounded by the demand
curve, the q-axis, and the vertical lines q = 0 and q = q .
The difference between this area and the amount they actually pay is called the
consumer surplus. Since p q is the area of the rectangle R of length q and height p ,
the consumer surplus is the area of the region denoted CS in Figure 2.4, which is A R.
4
20
2.9. Consumer and producer surplus
2
p
A
Demand
q
Figure 2.3: Value of q items at price p is (approximately) the area A

p
CS
p
R
Demand
q
Figure 2.4: The consumer surplus is the area CS (i.e. it is just A R)
Now,RCS
= A R. Area A is the area under the demand curve from q = 0 to q = , so
q D
A = 0 p (q) dq, and area R is p q . Hence the consumer surplus is given by the
formula:

Z
q
pD (q) dq p q .
CS =
0
An alternative way of viewing the consumer surplus is as the region enclosed by the
demand curve, the p-axis, p = p , and p = pD (0). Thus, we also have
Z
pD (0)
CS =
q D (p) dp.
Note that here we need to express q as a function of p on the demand curve, whereas
the first formula expresses p as a function of q. Whichever formula you use, try not
simply to memorise it, but rather remember what the relevant region is, and compute
its area in the easiest way for a particular problem. (For example, if the demand curve
is a straight line, then the region is question is triangular. Then, to compute its area
and hence the consumer surplus, we dont need a complicated formula involving
integrals: we would simply use the well-known fact that the area of a triangle is one half
of the base times the height.)
There is a counterpart to the consumer surplus, known as the producer surplus. It can
be argued that the total cost to the manufacturers of producing q items is represented
by the area under the supply curve between q = 0 and q = q . Yet they receive total
21
receipts, in equilibrium, of p q for producing and selling q units. The difference

between these two values is known as the producer surplus. It is given by the formula
Z q

pS (q) dq.
PS = p q
0
Activity 2.5 The demand for a commodity is given by p(q + 1) = 231, and the
supply is given by p q = 11.
Calculate the equilibrium price and quantity and the consumer surplus.
Learning outcomes
At the end of this chapter and the relevant reading, you should be able to:
as in Mathematics 1, know what is meant by the derivative
as in Mathematics 1, state the standard derivatives
as in Mathematics 1, calculate derivatives using sum, product, quotient, and
composite function (chain) rules
as in Mathematics 1, calculate derivatives by taking logarithms
as in Mathematics 1, establish the nature of the critical/stationary points of a
function.
as in Mathematics 1, use the derivative to help sketch functions
as in Mathematics 1, know the terminology surrounding marginals in economics,
and be able to find fixed costs and marginal costs, given a total cost function
as in Mathematics 1, know what is meant by the breakeven point and be able to
determine this
as in Mathematics 1, make use of the derivative in order to minimise or maximise
functions, including profit functions
use the derivative as a means of approximating small changes in a function
quote Taylors theorem, quote the standard Taylor (Maclaurin) series (for ex ,
ln(1 + x), sin x, and cos x, where the series are about x = 0), and know the ranges
of x for which they are valid
determine power series approximations for functions using Taylors theorem and by
manipulating the standard Taylor series
quote the definitions of point elasticity of demand and point elasticity of supply
determine these elasticities given (respectively) the demand and supply equations
or sets
know what is meant by the terms elastic and inelastic, and be able to determine
when demand is elastic or inelastic
determine the new resulting equilibrium price and quantity in the presence of an
excise (per-unit) tax or a percentage tax
as in Mathematics 1, understand the meaning of an (indefinite) integral and a
definite integral
as in Mathematics 1, state the standard integrals
as in Mathematics 1, use integration by substitution
22
2.9. Sample examination/practice questions
as in Mathematics 1, use integration by parts

as in Mathematics 1, integrate using partial fractions
understand the connection between areas and definite integrals
compute areas using definite integration
know what is meant by consumer and producer surplus, and be able to calculate
consumer and producer surpluses
Sample examination/practice questions

Question 2.1
Use differentiation to find the approximate change in x as

x increases from 100 to 101.
Show, more generally, that when
n
is
large,
the
change
in
x as x increases from n to
n + 1 is approximately 1/(2 n).

Question 2.2
r
1x
Express ln
as a series of terms in ascending powers of x up to and including x5 .
1+x
Use your series to obtain an approximate value for ln(3/ 11), giving your answer
correct to five decimal places.
Question 2.3
Expand as a power series, in terms up to x4 , the function f (x) = cos(sin x).
Question 2.4
The demand quantity q for a good is given by
q(1 + p2 ) = 100,
where p is the price. Determine the point elasticity of demand as a function of p. For
what values of p is the demand elastic?
Question 2.5
The supply and demand functions for a good are
q S (p) = bp a, q D (p) = c dp,
where a, b, c, d are all positive, and bc > ad. Suppose the government wishes to raise as
much money as possible by imposing an excise tax on the good. What should be the
value of the excise tax? What is the resulting government revenue?
Question 2.6
Find the area enclosed by the curves y = 1/t2 , y = t3 , the t-axis and the lines t = 1/2
and t = 2.
23
Question 2.7
The demand relationship for a product is

p=
50
,
q+5
and the supply relationship is
q
9
+ .
10 2
On the same diagram, produce graphs of the supply and demand curves. Determine the
consumer and producer surpluses.
p=
Question 2.8
Suppose that the demand and supply functions for a good are
q S (p) = bp a
and
q D (p) = c dp,
where a, b, c, d are positive constants. Find an expression for the consumer surplus.
Answers to activities
Feedback to activity 2.1
The demand function and its derivative are
q D (p) = 6000p2
and (q D )0 (p) = 6000 (2)p3 =
12000
.
p3
Therefore when p = 20 and p = 1 we have

12000
12000
q ' 3
p =
= 1.5.
p
203
So the price rise will result in approximately 1500 fewer items being sold.
You should obtain 0.995458. This compares well with the true value of 0.995461. (Of
course, this needs a calculator, but I stress again that such calculations will not be
required in the examinations, in which the use of calculators is not permitted.)
The corresponding new equilibrium quantity is
q T = 40 5pT = 40 5(4 + (3/5)T ) = 20 3T.
R
The required area is the definite integral 0 sin x dx. This is calculated as follows:
Z
sin x dx = [ cos x]0 = cos() ( cos(0)) = (1) + 1 = 2.
0
24
2.9. Answers to Sample examination/practice questions

The inverse demand function is pD (q) = 231/(q + 1). The equilibrium quantity q is the
solution to the equation
231
= q + 11,
q+1
obtained by equating pD (q) and pS (q). So, q satisfies the equation
(q + 11)(q + 1) = 231
= q 2 + 12q 220 = 0
= (q + 22)(q 10) = 0.
Since q cannot be negative, q = 10. The equilibrium price is p = q + 11 = 21. The
consumer surplus is then
Z q
CS =
pD (q) dq p q
Z0 10
231
=
dq (21)(10)
q+1
0
h
i10
= 231 ln(q + 1) 210
0
= 231 ln(11) 231 ln(1) 210

= 231 ln(11) 210,
which is approximately 343.9.
Answers to Sample examination/practice questions

Answer to question 2.1
The derivative of f (x) = x is f 0 (x) = 1/(2 x). Therefore, by the approximation

f '
we have
df
x,
dx
1
1
f (101) f (100) ' f 0 (100)(1) =
= .
20
2 100
The
change is therefore 1/20 = 0.05. (Incidentally, the exact change is
approximate
101 100, which is 0.04987562, so the approximation is good.)

Generally, if n is large and x increases from n to n + 1, then the change is given
approximately by
1
f (n + 1) f (n) ' f 0 (n)((n + 1) n) = ,
2 n
as required.
25
We first note that
1x
1
1
= ln(1 x) ln(1 + x).
1+x
2
2
Now we use the following expansions:
ln
x2 x3 x4 x5
,
ln(1 x) = x
2
3
4
5
x2 x3 x4 x5
+
+ ,
2
3
4
5
6
where we have omitted powers of x higher than x (noting that the question requires us
to work with powers up to x5 only). It follows that we have, approximately,

x2 x3 x4 x5
1
x2 x 3 x4 x5
1
x
x
+
+
,
2
2
3
4
5
2
2
3
4
5
ln(1 + x) = x
and simplifying this, we see that the Taylor (Maclaurin) series is

r
1x
x3 x5
ln
' x
.
1+x
3
5
For the
p next part of the question, it would appear that we have to express 3/ 11 in the
form (1 x)/(1 + x). For this to be so, we need
(1 x)/(1 + x) = (3/ 11)2 = 9/11,

so 11(1 x) = 9(1 + x), or 20x = 2, so we take x = 0.1. Then,

3
(0.1)3 (0.1)5
ln
' (0.1)
3
5
11
0.001 0.00001
= 0.1
3
5
= 0.1 0.000333 0.000002
= 0.100335.
(Incidentally, the true value is 0.1003353480, so the approximation is correct to five
decimal places.)
The series for sin x is
x3 x5 x7
+
+ ,
3!
5!
7!
but since we are only interested in powers of x up to x4 , we shall use the approximation
sin x ' x
sin x ' x
The series for cos y is
cos y ' 1
26
x3
x3
=x .
3!
6
y2 y4 y6
+
+ .
2!
4!
6!
To get an approximation to cos(sin x) up to terms in x4 , we substitute

y =x
x3
6
in the expansion for cos y and ignore any powers of x higher than x4 . (Ignored terms
will be denoted by .) We get
2

4
x3
1
x3
x
+
x
+
6
24
6

2

4
x2
x2
x2
x4
1
1
+
+
2
6
24
6

x2 x4
x4
x2
1
+
+
(1 + ) +
2
3
36
24
x2 x4 x4
+
+
+
2
6
24
1 2
5
x + x4 + .
2
24
1
cos(sin x) ' 1
2
= 1
= 1
= 1
= 1
Be careful about what can be ignored and what cannot. Consider, for example, the term

x4
x2 x4
+
1
.
2
3
36
In expanding this, since we are ignoring powers of x higher than 4, we can ignore all the
terms inside the brackets apart from the first, since they lead to a power of x at least as
high as x6 . But you also have to be careful to include every term that needs to be
included.
We have
q = 100/(1 + p2 ) = 100(1 + p2 )1 ,
so
dq
= 100(2p)(1 + p2 )2 = 200p(1 + p2 )2 ,
dp
and the elasticity of demand is

p dq
q dp

p
200p
=
q (1 + p2 )2
200p2
=
q(1 + p2 )2
200p2
=
(100/(1 + p2 )) (1 + p2 )2
2p2
=
.
1 + p2
27
Note that in order to obtain the elasticity as a function of p, we needed to express q in

terms of p in the second-last of these equations. Now, the demand is elastic when > 1,
which means 2p2 /(1 + p2 ) > 1, or 2p2 > 1 + p2 . So we have elastic demand when p2 > 1
which, since p 0, means when p > 1.
The tax revenue is R(T ) = T q T . In order to calculate q T , we first note that when the
excise tax is imposed, the selling price at equilibrium, pT , is such that
q T = b(pT T ) a = c dpT .
Solving for pT , we obtain
pT =
Then
bT
c+a
+
.
b+d b+d
bc ad
bdT
,
b+d
b+d

bd
bc ad
T
T 2.
R(T ) =
b+d
b+d
q T = c dpT =
so that
Setting R0 (T ) = 0, we discover that there is only one critical point,

Tm =
bc ad
.
2bd
The second derivative R00 (T ) = (2bd)/(b + d) is constant, and negative (since b and d
are positive). Hence R00 (Tm ) < 0 and Tm is a maximum point. Therefore, Tm is the level
of excise tax the government should impose. The resulting government revenue is
R(Tm ) =
(bc ad)2
.
4bd(b + d)

It is a good idea to sketch the region described. Heres what it looks like.
Note that the curves y = 1/t2 and y = t3 intersect when t5 = 1 which, in the positive
quadrant, means t = 1. The region were interested in then divides naturally into two
parts. The first is for t from 1/2 to 1, where the curve 1/t2 lies above t3 and the region
is therefore bounded by t = 1/2, t = 1 and y = t3 . The second part is for t ranging from
1/2 to 1: here, the curve y = t3 lies above y = 1/t2 , and so the region is bounded by
t = 1, t = 2 and y = 1/t2 . The easiest way to compute the area, A, of the region is to
calculate each of the areas of the two parts, which well call A1 and A2 , separately. Then
A = A1 + A2 . We have
4 1
4
Z 1
t
1 4 1 1
15
3
A1 =
t dt =
= (1)
= ,
4 1/2 4
4 2
64
1/2
and
Z
A2 =
1
28
2

1
1
1
1
1
dt
=
=
.
t2
t 1
2
1
2
It follows that
A = A 1 + A2 =
47
.
64

The supply and demand curves are as follows. Note that the supply curve is an
upward-sloping straight line.
We need to find the equilibrium price and quantity. To find the equilibrium quantity, we
solve
50
q
9
=
+ .
q+5
10 2
Multiplying both sides by 10(q + 5), we obtain
500 = q(q + 5) + 45(q + 5) = q 2 + 50q + 225,
29
so
q 2 + 50q 275 = 0.
The solutions are 5 and 55 and clearly it is the positive solution were interested in. So
the equilibrium quantity is 5 and the equilibrium price is 50/(5 + 5) = 5.
Now, the consumer surplus is
5
50
dq (5)(5)
0 q+5
= [50 ln(q + 5)]50 25
= (50 ln 10 50 ln 5) 25
= 50 ln 2 25.
Z
CS =
The producer surplus is easy to calculate, since it is the area of the triangular region
bounded by the p-axis, the line p = 5 and the portion of the supply curve between q = 0
and q = 5. Since the supply curve intersects the p-axis at 4.5, this triangle has height 0.5
and base 5, so its area is (1/2)(0.5)(5), which is 1.25. So the producer surplus is 1.25.
We first calculate the consumer surplus by elementary methods, using only the fact that
the area of a triangle is half its base times its height. As is easily seen, the equilibrium
price is
c+a
,
p =
b+d
and the equilibrium quantity is
q = c dp =
bc ad
.
b+d
The inverse demand function is
cq
.
d
The consumer surplus is the area of the triangular region bounded by the lines p = p
and q = 0, and by the demand curve. (Sketch the curves to see that this is so! The fact
that the supply demand curve is a straight line means the region in question is
triangular.) Since the demand curve crosses the p-axis at (0, pD (0)), this area is

1 D
1 c c+a
bc ad

CS = (p (0) p )q =
.
2
2 d b+d
b+d
pD (q) =
That is,
1
CS =
2
cb + cd cd ad
d(b + d)

bc ad
b+d

=
(bc ad)2
.
2d(b + d)2
We can also calculate the consumer surplus by definite integration. (This turns out to
be more difficult in this particular case, but definite integration really is needed for
demand functions more complex than the simple linear one discussed here.) We have
Z
CS =
0
30
q
D
p (q) dq p q =
Z
0
cq
dq p q .
d
Now,
Z
0
cq
dq =
d
Z
0

q
q
c
q2
c
(q )2
dq =
q
= q
,
d d
d
2d 0
d
2d
c
so

CS =
=
=
=
=
=
c (q )2
q
d
2d
p q
q
(2c q 2p d)
2d

q
bc ad
c+a
2c
2
d
2d
b+d
b+d
q
(2cb + 2cd bc + ad 2cd 2ad)
2d(b + d)
q
(bc ad)
2d(b + d)
(bc ad)2
,
2d(b + d)2
as above.
31
32
Chapter 3
Functions of several variables
3
Essential reading
R
R
R
R
Anthony and Biggs (1996) Chapters 11, 12, 13, 21 and 22.
Further reading
3.1
Binmore and Davies (2001) Sections 6.6 and 6.8.

Bradley (2008) Chapter 7.
Dowling (2000) Chapters 5 and 6.
Ostaszewski (1993) Chapters 12 and 15.
Introduction
In this chapter of the guide, we briefly review the material from 05A Mathematics 1
concerning functions of several variables, and we introduce some new topics and
applications related to functions of several variables.
3.2
Partial derivatives
We briefly review some key ideas from Mathematics 1 (though in the slightly more
general context of n-variable functions rather than 2-variable functions). A function of
n variables, for n 2, takes inputs (x1 , x2 , . . . , xn ) and returns an output value
f (x1 , x2 , . . . , xn ). In Mathematics 1, we concentrated almost entirely on the case in
which n = 2, and we would often use x, y to denote the variables, rather than x1 , x2 . In
this chapter we shall sometimes consider specific examples where n > 2. In particular,
when n = 3, we shall often use x, y, z to represent the three variables rather than
x1 , x2 , x3 .
For a function f (x1 , x2 , . . . , xn ), the rate of change of f with respect to x1 , when the
other variables are fixed, is the partial derivative of f with respect to x1 , denoted
f /x1 . The partial derivatives with respect to the other variables are similarly defined.
As we saw in Mathematics 1, calculating partial derivatives is only slightly more
difficult than calculating standard derivatives, and you should be proficient in this from
your study of Mathematics 1. For example, to calculate the partial derivative of a
33
3. Functions of several variables
function f (x, y, z) with respect to x, you just treat y and z both as if they were fixed
numbers, and differentiate with respect to x. Here is an example, to refresh your
memory, and to indicate how to deal with 3-variable functions.
p
Example 3.1 Suppose f (x, y, z) = x2 y y 2 + z 2 . Writing this in the form
x2 y(y 2 + z 2 )1/2 makes it slightly easier to work with. We regard this as a product of
two functions, where the second one is a composition. Applying the product and
chain rules appropriately, we have
p
f
= 2xy(y 2 + z 2 )1/2 = 2xy y 2 + z 2 ,
x

1
f
2 2
2 1/2
2
= x (y + z ) + x y
(y 2 + z 2 )1/2 (2y)
y
2
= x2
f
= x2 y
z
3.3
x2 y 2
y2 + z2 + p
y2 + z2

1
x2 yz
(y 2 + z 2 )1/2 (2z) = p
.
2
y2 + z2
Homogeneous functions and Eulers theorem
Suppose we have a function of two variables, such as f (x, y) = 3x2 y + 7xy 2 , and we
multiply the inputs x and y by a constant c. In this case we get output
f (cx, cy) = 3(cx)2 (cy) + 7(cx)(cy)2 = c3 (3x2 y + 7xy 2 ) = c3 f (x, y).
Thus, for this particular f , multiplying the inputs by c results in the output being
multiplied by c3 . In general, if a function h is such that
h(cx, cy) = cD h(x, y),
then we say that h is homogeneous of degree D. The number D is called the degree of
homogeneity of h. The function f given by the formula above is homogeneous of degree
3. Note that many (indeed most) functions are not homogeneous. The notion of a
homogeneous function is related to the idea of returns to scale in economics. In the
case of a production function, we say that there are constant returns to scale if a
proportional increase in k and l results in the same proportional increase in q(k, l); that
is, if
q(ck, cl) = cq(k, l).
This means, for example, that doubling both capital and labour doubles the production.
This is the same as saying that q is homogeneous of degree 1. If q is homogeneous of
degree D > 1 then the proportional increase in q(k, l) will be larger than that in k and
l, and we say that there are increasing returns to scale. On the other hand, if D < 1 we
say that there are decreasing returns to scale.
34
3.3. Homogeneous functions and Eulers theorem
p
Example 3.2 Consider the function f (x, y) = x2 y 2 + y 2 x4 + y 4 . Then f is
homogeneous of degree 4 because
p
f (cx, cy) = (cx)2 (cy)2 + (cy)2 (cx)4 + (cy)4
p
= c2 x2 c2 y 2 + c2 y 2 c4 (x4 + y 4 )
p
= c4 x2 y 2 + c2 y 2 c2 x4 + y 4
p
= c4 x 2 y 2 + c4 y 2 x 4 + y 4
= c4 f (x, y).
Informally speaking, a function such as the one just considered, which involves only
powers of the variables and powers of combinations of the variables, is homogeneous if
the total degree or power of each constituent term of the function is the same. In the
example we just saw, the total degree of the first part of the function, x2 y 2 is 2 + 2 = 4,
the sum of the degree of x and the degree of y. The degree of the second term is
2 + (1/2)4 = 4 (where the factor 1/2 comes from the square root). Since these are the
same, and both equal to 4, the function is homogeneous of degree 4. To see why this
informal approach holds, just think about what happens when we substitute cx for x
and cy for y. Perhaps it would also be useful to see why a function with parts of
differing degrees cannot be homogeneous. Consider the function f (x, y) = x + y 2 . (This
has terms of differing degree 1 and 2.) Then f (cx, cy) = cx + c2 y 2 . But there is no way
in which cx + c2 y 2 can be written as cD (x + y 2 ), for all values of c, for some D.
Activity 3.1 Which of the following functions are homogeneous? If homogeneous,
what are their degrees?
x4
x2 y + p
x2
Example 3.3
y2
x
2
2
+ 5 + ex /y , x2 + x sin y.
y
Suppose we are told that the function

f (x, y) =
x y + x 2 y
,
xy + y 2
is homogeneous of degree 2. Then we can use this fact to determine the numbers
, and . There are several ways to do this. Think about what happens when x, y
are replaced by cx, cy. We have
c+1 x y + c2+ x2 y
f (cx, cy) =
.
c1+ xy + c2 y 2
Now, if the function is homogeneous of degree 2, this should simply equal c2 f (x, y).
But for this to be true, three things must hold:
35
1. The degree of each of the two terms on the numerator (top line) should equal
each other; lets say this common degree is d1 .
2. The degree of each of the two terms on the denominator (bottom line) should
equal each other; lets say this degree is d2 .
3. The degrees d1 and d2 should cancel to give degree 2; that is, the cd1 factor from
the top and the cd2 factor on the bottom should cancel to give c2 . This means
that d1 d2 = 2.
These conditions are equivalent to saying that

The numerator must be homogeneous.
The denominator must be homogeneous.
The degree of the numerator must be 2 more than the degree of the
denominator.
Respectively, these three conditions mean:
1. + 1 = 2 + .
2. 1 + = 2.
3. We can write four possible equations here. For example, since the degree of the
first term on the numerator is + 1 and the degree of the first term on the
denominator is 1 + , we have + 1 (1 + ) = 2. By considering other terms
(such as the second term on the numerator, and the first on the denominator,
and so on), we obtain three other equations: + 1 2 = 2, 2 + (1 + ) = 2,
and 2 + 2 = 2. But by the first and second observations above, these four
equations are all equivalent and we need only use one of them.
Choosing the second of the four equations in 3. (because it is simple) we obtain
= 3. Then the equation in 1. tells us that = 2, and the equation in 2. tells us
that = 1. So the function is
f (x, y) =
x3 y + x2 y 2
.
xy + y 2
There is a useful result about homogeneous functions known as Eulers theorem. This
states that a function f (x, y) is homogeneous of degree D if and only if
x
f
f
+y
= Df.
x
y
Example 3.4 Consider the function f (x, y) = (x2 + y 2 )3/2 x1/2 y 1/2 . We can show
that this is homogeneous of degree 4, and verify that
x
36
f
f
+y
= 4f.
x
y
3.4. Optimisation
To show that the function is homogeneous of degree 4, we observe that

f (cx, cy) = ((cx)2 + (cy)2 )3/2 (cx)1/2 (cy)1/2
= (c2 )3/2 (x2 + y 2 ))3/2 c1/2 x1/2 c1/2 y 1/2
= c4 f (x, y).
Now,

1 1/2 1/2
.
x
y
2

f
1 1/2 1/2
3
2
2 1/2 1/2 1/2
2
2 3/2
.
= 2y(x + y ) (x y ) + (x + y )
x y
y
2
2
f
3
= 2x(x2 + y 2 )1/2 (x1/2 y 1/2 ) + (x2 + y 2 )3/2
x
2
It follows that
x
f
f
+y
= (3x2 + 3y 2 )(x2 + y 2 )1/2 (x1/2 y 1/2 ) + (x1/2 y 1/2 )(x2 + y 2 )3/2
x
y
= 3(x2 + y 2 )3/2 (x1/2 y 1/2 ) + (x2 + y 2 )3/2 (x1/2 y 1/2 )
= 4(x2 + y 2 )3/2 (x1/2 y 1/2 ) = 4f (x, y),
as predicted by Eulers theorem.

Activity 3.2 That was a difficult example. Try an easier one for yourself, by
verifying Eulers theorem for the function f (x, y) = x2 y 3 + x4 y.
More generally, a function f (x1 , x2 , . . . , xn ) of n variables is said to be homogeneous of
degree D if for all c,
f (cx1 , cx2 , . . . , cxn ) = cD f (x1 , x2 , . . . , xn ).
Eulers theorem in this case is: f (x1 , x2 , . . . , xn ) is homogeneous of degree D if and only
if
f
f
f
x1
+ x2
+ + xn
= Df (x1 , x2 , . . . , xn ).
x1
x2
xn
3.4
3.4.1
Optimisation
Introduction
We now briefly review the optimisation techniques we met in Mathematics 1, and again
we shall describe them a little more generally, in the context of n-variable functions.
Here, we will look at constrained optimisation of three-variable rather than simply
two-variable functions. For unconstrained optimisation of such functions, there are
techniques for testing whether a critical point is a maximum, a minimum, or a saddle
point. However, for functions of more than two variables, these techniques are outside
the scope of this subject. Also not covered in this subject are tests for ensuring that the
solution to a constrained optimisation problem is a maximum or a minimum.
37
3.4.2
Unconstrained optimisation
The local maxima and minima of a function f (x1 , x2 , . . . , xn ) occur at points where all
f
(for i = 1, 2, . . . , n) are equal to 0. Such points are called
the partial derivatives
xi
critical points or stationary points. A critical point which is neither a local maximum
nor a local minimum is a saddle point. If n = 2, then there is a fairly simple test to
determine the nature of a critical point: Suppose that (a, b) is a critical point of f (x, y).
Then the following statements are true, where the partial derivatives are evaluated at
(a, b):
2f 2f
If
x2 y 2
2f 2f
If
x2 y 2
2f 2f
If
x2 y 2
3.4.3
2f
xy
2
2f
xy
2
2f
xy
2
> 0 and
f
< 0, it is a maximum.
x
> 0 and
f
> 0, it is a minimum.
x
< 0, it is a saddle point.
Applications of unconstrained optimisation
In Mathematics 1, we described the problem of maximising profit for a firm making two
products, X and Y . Generally, if pX and pY are the selling prices of one unit of X and
one unit of Y , then the total revenue obtained by producing amounts x and y is
T R(x, y) = xpX + ypY .
The joint total cost function T C(x, y) will tell us how much it costs the manufacturer to
produce x units of X and y of Y . Then, the profit function is
(x, y) = T R(x, y) T C(x, y) = xpX + ypY T C(x, y),
and we maximise this function of x and y using the techniques described above.1
There are many other interesting applications of unconstrained optimisation in
economics. We now consider the case of a firm producing just one good, but selling that
good in two markets, such as the domestic market and the export market.
Let us suppose that the firm is a monopoly in each of the two markets, and that the
demand curves for its product are different in the two markets. The firm may decide to
set different prices in the domestic market and in the export market, or it may set the
same price in both. In the former case, we say that we have price discrimination. The
following example demonstrates how to deal with such problems.
Example 3.5 Suppose the demand functions for a firms domestic and foreign
markets are given by
P1 = 30 4Q1
P2 = 50 5Q2 .
1
See Anthony and Biggs (1996) Chapter 13, for many examples.
38
3.4. Optimisation
and the total cost function is

T C = 10 + 10Q,
where Q = Q1 + Q2 . We shall determine the prices which maximise profit when we
have price discrimination, and when we have no price discrimination. First, we note
that the total revenue is P1 Q1 + P2 Q2 , and hence the profit function is
(Q1 , Q2 ) = (P1 Q1 + P2 Q2 ) (10 + 10Q)

= (30 4Q1 )Q1 + (50 5Q2 )Q2 (10 + 10Q1 + 10Q2 )
= 20Q1 + 40Q2 4Q21 5Q22 10.
To find the maximum, we solve
= 20 8Q1 = 0
Q1
and
= 40 10Q2 = 0,
Q2
obtaining Q1 = 5/2 and Q2 = 4, and hence P1 = 20, P2 = 30. We can see that this is
indeed a maximum of profit by using the second derivative test (check this!). The
maximum profit with price discrimination is therefore (5/2, 4) = 95. (Here, we have
found an expression for profit in terms of Q1 and Q2 . An alternative approach would
be to determine as a function of P1 and P2 and maximise the resulting function of
P1 and P2 .)
Now, when there is no price discrimination, we have P1 = P2 = P , say. From the
demand equations, we have
P = 30 4Q1 = 50 5Q2 ,
so
Q1 =
15 P
2
4
and Q2 = 10
P
,
5
and
35 9P
.
2
20
It follows that the profit, as a function of P , is given by
Q = Q1 + Q2 =
(P ) = P (Q1 + Q2 ) (10Q + 10)

= P Q 10Q 10

35 9P
35 9P
10
10
=P
2
20
2
20
= 22P
9 2
P 185.
20
This is a function of the single variable P . (We could just as easily have expressed
in terms of the single variable Q: both approaches are fine.) To maximise, we set
d/dP = 0. This yields
9
22 P = 0,
10
so P = 220/9. This gives a maximum of the profit function, since
d2 /dP 2 = (9/10) < 0. The corresponding profit is (220/9) = 755/9, which is
39
83.88 to two decimal places. This is, as we would expect, less than the maximum
profit when price discrimination is allowed.
It is certainly possible to use Lagrangean methods, with the constraint P1 P2 = 0
expressed in terms of Q1 and Q2 , to deal with the case of no price discrimination.
However, it is probably easier to take the approach indicated in the example just
given, where the equation P1 = P2 is used to reduce the optimisation problem to a
one-variable problem.
Activity 3.3 Follow up the comment just made by using the Lagrangean method
for the last part of the example, in which there is no price discrimination.
3.4.4
Constrained optimisation
Suppose that f (x1 , x2 , . . . , xn ) has to be minimised or maximised subject to the

constraint g(x1 , x2 , . . . , xn ) = 0. This means we want to find the maximum (or
minimum) value of the function f at points (x1 , x2 , . . . , xn ) which satisfy the condition
g(x1 , x2 , . . . , xn ) = 0. Then we may use the method of Lagrange multipliers.2 To find the
optimal points, we first find the critical points of the (n + 1)-variable function
L(x1 , x2 , . . . , xn , ) = f (x1 , x2 , . . . , xn ) g(x1 , x2 , . . . , xn ).
The function L is known as the Lagrangean (sometimes spelt Lagrangian) and is
known as the Lagrange multiplier. In other words, we find the points at which the
first-order conditions
L
= 0 (for i = 1, 2, . . . , n) and
xi
L
= 0.
Then the theory of Lagrange multipliers asserts that the required optimal points of f ,
subject to the constraint, are to be found among these critical points.
In Mathematics 1, you studied the Lagrangean method quite extensively for
two-variable functions. Here, we dont introduce much new material on this subject, but
for Mathematics 2, you should be able to deal with more challenging 2-variable
constrained optimisation problems, and with Lagrangean problems involving more than
two variables. The following example serves as an illustration of the technique for such
functions.
Example 3.6
Use Lagrange multipliers to find the minimum value of

1
1
1
+ 2 + 2,
2
2x
3y
6z
for x, y, z > 0 subject to
1
1
1
x + y + z = c,
2
3
6
where c > 0 is a constant.

2
See Anthony and Biggs (1996) Chapters 21 and 22.
40
3.4. Learning outcomes
To solve this, we form the Lagrangean

1
1
1
L= 2 + 2 + 2
2x
3y
6z

1
1
1
x+ y+ zc ,
2
3
6
and we solve the equations
L
1
1
= 3 =0
x
x
2
L
2
1
= 3 =0
y
3y
3
L
1
1
= 3 =0
z
3z
6
L
1
1
1
= x y z+c=0
2
3
6
The first three equations give three expressions for :
=
2
2
2
= 3 = 3,
3
x
y
z
from which it follows that x3 = y 3 = z 3 and hence x = y = z. Then, setting

y = z = x in the final equation (which is simply the constraint) gives
1
1
1
x + x + x = c,
2
3
6
which is x = c. Therefore the optimum (in this case minimum) is when
x = y = z = c, and the optimum value is
1
1
1
1
+ 2 + 2 = 2.
2
2c
3c
6c
c
Learning outcomes
as in Mathematics 1, understand the concept of a function of many variables
as in Mathematics 1, calculate partial derivatives, and use implicit differentiation
state precisely what is meant by a homogeneous function of n variables, and be
able to show that a given function is homogeneous
state Eulers theorem, and be able to verify it for given homogeneous functions
as in Mathematics 1, find and classify critical points
as in Mathematics 1, solve optimisation and constrained optimisation problems:
additionally, solve constrained optimisation problems with three or more variables
as in Mathematics 1, explain, and make use of, the meaning of the Lagrange
multiplier
41

Question 3.1
The production function Q(K, L) of a firm is given by
5L K
Q(K, L) = 2
.
L + K 2 L
If the firm has constant returns to scale, what are the values of and ?
Question 3.2
If the following function is homogeneous of degree 1 determine the values of , , :
g(x, y, z) =
3x y 4x+3
y 3/2 (x+ + 2z 2+ )1/2
Question 3.3
Show that the function
f (x, y) = 2y 3 x + 5y 4 (y 3/4 2x3/4 )4 x,
is homogeneous of degree 4. Verify that
x
f
f
+y
= 4f.
x
y
Question 3.4
A monopolist has a total cost function of the form T C = 20Q + 20 where Q is the
demand. He is aware of the possibility of separating his customers into two distinct
markets with the following demand equations:
Market 1: Q1 = 9 0.05P1 ,
Market 2: P2 = 80 5Q2 ,
where Q1 + Q2 = Q and Pi is the price in market i (i = 1, 2). If the monopolist wishes
to maximise his total profits, determine the prices he will charge in the two markets (i)
if he is permitted to choose different prices for the two markets, and, (ii) if he has to
choose the same price for the two markets.
Question 3.5
It is thought that a consumer measures the utility u of possessing a quantity x of apples
and a quantity of y of oranges by the formula:
u = u(x, y) = x y 1 .
It is known that when the consumers budget for apples and oranges is $1 he will buy 1
apple and 2 oranges when they are equally priced. Find . [Hint: First solve the utility
maximisation problem with as a parameter.]
42
3.4. Answers to activities
The price of apples falls to half that of oranges, with the price of oranges unchanged.
How many apples and oranges will the consumer buy for $10?
Question 3.6
Company headquarters holds 100 units of a raw material; it proposes to divide this into
three lots of x, y and z units of raw material to be sent to three area factories in the
North, the Midlands and the South of the country and there they are turned into a
finished product. Although the product sells at the same price in all three regions, the
factories are not equally productive and in fact their respective outputs are:
North: 3 x
Midlands: 4 y
South: 5 z.
How should the headquarters allocate x, y, z to maximise the revenue? [Hint: Use the
Lagrange multiplier method.]
Question 3.7
Use the Lagrange multiplier method to minimise
1 1 1
+ + ,
x y z
subject to
x + y + z = c,
for x, y, z > 0, where c is a positive constant. Use your result to deduce that for
x, y, z > 0

1
1
1 1 1 1
(x + y + z)
+ +
.
3
3 x y z
x4
, then
If f (x, y) = x2 y + p
x2 + y 2
c4 x 4
f (cx, cy) = c3 x2 y + p
c2 (x2 + y 2 )
c4 x 4
= c3 x 2 y + p
c x2 + y 2
x4
= c3 x 2 y + c3 p
x2 + y 2
which is c3 f (x, y), so the function is homogeneous of degree 3.
43
If f (x, y) =
x
2
2
+ 5 + ex /y , then
y
x
cx
2 2
2 2
2
2
+ 5 + ec x /(c y ) = + 5 + ex /y .
f (cx, cy) =
cy
y
In other words, f (cx, cy) = f (x, y), which means that f is homogeneous of degree 0
(since f (cx, cy) = c0 f (x, y)).
For f (x, y) = x2 + x sin y, we have f (cx, cy) = c2 x2 + cx sin(cy). Now, this cannot be
written as cD f (x, y) for some D. For, if this were the case then, clearly, from the x2
part, D would have to be 2. Then wed have to have sin(cy) = c sin y for all y, which
simply isnt true. So the function is not homogeneous.
Since
f (cx, cy) = c2 x2 c3 y 3 + c4 x4 cy = c5 (x2 y 3 + x4 y) = c5 f (x, y),
the function is homogeneous of degree 5. To verify Eulers theorem, we need to check
that
f
f
+y
= 5f.
x
x
y
Now,
f
f
= 2xy 3 + 4x3 y,
= 3x2 y 2 + x4 ,
x
y
so
x
f
f
+y
= x(2xy 3 + 4x3 y) + y(3x2 y 2 + x4 )
x
y
= 2x2 y 3 + 4x4 y + 3x2 y 3 + x4 y
= 5x2 y 3 + 5x4 y
= 5f (x, y),
as required.
The constraint is P1 = P2 , which is
30 4Q1 = 50 5Q2 ,
so we take as the constraint equation 20 + 4Q1 5Q2 = 0, and the Lagrangean is
L = 20Q1 + 40Q2 4Q21 5Q22 10 (20 + 4Q1 5Q2 ).
We then solve the equations
L
= 20 8Q1 4 = 0
Q1
L
= 40 10Q2 + 5 = 0
Q2
L
= 20 + 4Q1 5Q2 = 0.
44
From the first two equations,

=
20 8Q1
10Q2 40
=
,
4
5
so
100 40Q1 = 40Q2 160,
and Q2 = (13/2) Q1 . Substituting this into the third equation gives

13
Q1 = 0,
20 + 4Q1 5
2
which implies Q1 = 25/18. The corresponding value of Q2 is 46/9. To find the price, we
can use the equation for P1 or P2 . (Of course, the prices P1 and P2 are equal, to P , say:
that was our constraint.) Using the formula for P1 , we get

25
220
P1 = 30 4
=
,
18
9
as before.

To say the firm has constant returns to scale means that the function Q is homogeneous
of degree 1. The total degree of the numerator is + . Since the terms on the
denominator must have the same degree as each other, we must have 2 = 2 + 1, which
means = 1/2. Now, since Q is homogeneous of degree 1, the degree of the numerator
must be one more than that of the denominator. So, + = 1 + 2, which, since
= 1/2, means = 5/2.
Since the function is homogeneous of degree 1, we can make the following three
observations:
The numerator must be homogeneous, so + = + 3.
The denominator must be homogeneous, so + = 2 + .
The degree of the numerator must be one more than that of the denominator. Now,
because the bracketed term is raised to the power 1/2 in the denominator, the
degree of the denominator is 3/2 + (1/2)( + ) (or, equivalently, given the
previous observation, 3/2 + (1/2)(2 + )). The degree of the numerator is +
(or, equivalently, + 3). So we can say
+ =1+
3 1
+ (2 + ).
2 2
In fact, given the equivalences from the first two observations, there are four
different equations we could write down, and all would say the same thing.
45
So we have the system

+ = + 3, + = 2 + , + =
1
5
+ + ,
2
2
which simplifies to
= + 2, = 2,
1
5
+ = + .
2
2
There are many ways to solve this. From the first two equations, + 2 = 2, so
= 2. Then, = 2 = 4. Using these in the third equation, we obtain
1
5
+ 4 = + 2,
2
2
so = 1 and, then, = 4 and = 2.
We have
f (cx, cy) = 2c3 y 3 cx + 5c4 y 4 c3/4 y 3/4 2c3/4 x3/4
4
= 2c4 y 3 x + 5c4 y 4 c4 y 3/4 2x3/4 x
4
cx
= c4 f (x, y),
so the function is homogeneous of degree 4.
Now,

3 1/4
f
3
= 2y 2 x
4(y 3/4 2x3/4 )3 x (y 3/4 2x3/4 )4
x
4
= 2y 3 + 6(y 3/4 2x3/4 )3 x3/4 (y 3/4 2x3/4 )4 ,
and
f
= 6y 2 x + 20y 3
y

3 1/4
y
4(y 3/4 2x3/4 )3 x
4
= 6y 2 x + 20y 3 3y 1/4 (y 3/4 2x3/4 )3 x.

So,
x
f
= 2y 3 x + 6(y 3/4 2x3/4 )3 x7/4 (y 3/4 2x3/4 )4 x,
x
and
y
46
f
= 6y 3 x + 20y 4 3y 3/4 (y 3/4 2x3/4 )3 x.
y
Therefore,
x
f
f
+y
= 2y 3 x + 6(y 3/4 2x3/4 )3 x7/4 (y 3/4 2x3/4 )4 x
x
y
+ 6y 3 x + 20y 4 3y 3/4 (y 3/4 2x3/4 )3 x
= 8y 3 x 3x(y 3/4 2x3/4 )3 (y 3/4 2x3/4 )

(y 3/4 2x3/4 )4 x + 20y 4
= 8y 3 x + 20y 4 3x(y 3/4 2x3/4 )4 x(y 3/4 2x3/4 )4
= 8y 3 x + 20y 4 4(y 3/4 2x3/4 )4 x
= 4f (x, y),
as required.
With a view to making everything a function of Q1 and Q2 , we note first that
P1 = 180 20Q1 . The profit, as a function of Q1 and Q2 , is then
(Q1 , Q2 ) = (P1 Q1 + P2 Q2 ) (20Q + 20)
= (180 20Q1 )Q1 + (80 5Q2 )Q2 (20Q1 + 20Q2 + 20)
= 160Q1 + 60Q2 20Q21 5Q22 20.
To find the maximum, we solve
= 160 40Q1 = 0
Q1
= 60 10Q2 = 0,
Q2
and
obtaining Q1 = 4 and Q2 = 6, and hence P1 = 100, and P2 = 50. We can see that this is
indeed a maximum of profit by using the second derivative test. We have
2
= 40,
Q21
2
= 10,
Q22
2
= 0,
Q1 Q2
which gives
2
< 0,
Q21
2
Q21

2
Q22
2
Q1 Q2
2
> 0,
and so the critical point is indeed a maximum.

Now, when there is no price discrimination, we have P1 = P2 = P , say. From the
demand equations, we have
Q1 = 9 0.05P
and
Q2 = 16 0.2P,
so
Q = Q1 + Q2 = 25 0.25P.
47
It follows that the profit, as a function of P , is given by

(P ) = P (Q1 + Q2 ) (20Q + 20)
= P Q 20Q 20
= P (25 0.25P ) 20 (25 0.25P ) 20
= 30P 0.25P 2 520.

To maximise, we set d/dP = 0. This yields
30 0.5P = 0,
so P = 60. This gives a maximum of the profit function, since d2 /dP 2 = 0.5 < 0.
Suppose that oranges and apples are equally priced, at p. The budget equation is then
px + py = 1. The Lagrangean for the utility maximisation problem, subject to the
budget equation, is
L = x y 1 (px + py 1),
and the first-order conditions are
L
= x1 y 1 p = 0
x
L
= (1 )x y p = 0
y
L
= 1 px py = 0.
From the first two equations,

1
1
= x1 y 1 = (1 )x y ,
p
p
so
y 1 y =
(1 ) 1
x x ,
so
y=
(1 )
x,
when utility is maximised. Now, were told that when utility is maximised, we have
x = 1 and y = 2, so we must have (1 )/ = 2, which means 1 = 2, or = 1/3.
The budget equation tells us what p is, since p(1) = p(2) = 1, so p = 1/3.
Now, suppose the price of apples is halved, to 1/6. Then the Lagrangean is
L=x
48
1/3 2/3

1
1
x+ y1 ,
6
3
and the first-order conditions are

L
= (1/3)x2/3 y 2/3 (1/6) = 0
x
L
= (2/3)x1/3 y 1/3 (1/3) = 0
y
L
= 1 (1/6)x (1/3)y = 0.
The first two equations show that

= 2x2/3 y 2/3 = 2x1/3 y 1/3 ,
so x = y. By the third equation, x/2 = 1, so x = 2 and y = 2.
6. The constraint is given by the fact that there are only 100 units of the raw material,
so that x + y + z = 100. The function to be maximised is the revenue, which is equal to
the selling price (which is fixed across all three sectors) times the quantity
produced.
Thus, the revenue is proportional to the quantity produced, which is 3 x + 4 y + 5 z,

so it suffices to maximise this function subject to x + y + z 100 = 0. The Lagrangean is
L = 3 x + 4 y + 5 z (x + y + z 100),
and the equations to be solved are
3
L
= =0
x
2 x
L
2
= =0
y
y
L
5
= =0
z
2 z
L
= x y z + 100 = 0.
The first three equations give three expressions for :

2
5
3
= = = .
y
2 x
2 z
These relationships enable us to express all three variables x, y,
z in terms of any given
one of them.Lets express y and z in terms of x. We have 3/(2 x) = 2/ y, so
y=
on squaring
(4/3) x which,
both sides,
tells us that y = (16/9)x. Similarly, from
3/(2 x) = 5/(2 z), we obtain z = (5/3) x and z = (25/9)x. Then the constraint
equation x + y + z = 100 becomes
x+
16
25
x + x = 100,
9
9
so (50/9)x = 100 and x = 18. Then, y = (16/9)(18) = 32 and z = (25/9)(18) = 50. So

the optimal values of x, y, z are
x = 18,
y = 32 and z = 50.
49

The Lagrangean is
L=
1 1 1
+ + (x + y + z c) .
x y z
The first order conditions are

L
1
= 2 =0
x
x
1
L
= 2 =0
x
y
L
1
= 2 =0
x
z
L
= x y z + c = 0.
From the first three equations,

=
1
1
1
= 2 = 2.
2
x
y
z
So, x = y = z.
By the third equation, 3x = c, so the function is minimised when x = y = z = c/3. The
minimum value is
1
1
9
1
+
+
= .
c/3 c/3 c/3
c
Now, given any x, y, z > 0, let c = x + y + z. By what has just been shown, for any
X, Y, Z, with X + Y + Z = c,
1
1
9
9
1
+ + =
.
X Y
Z
c
x+y+z
In particular, this must be true for X = x, Y = y and Z = z. Therefore
9
9
1 1 1
+ + =
,
x y z
c
x+y+z
and so
x+y+z
3
3
or, in other words,

1
(x + y + z)
3
50
1
1/x + 1/y + 1/z

,

1
1 1 1 1
+ +
.
3 x y z
Chapter 4
Linear algebra and applications
Essential reading

Anthony and Biggs (1996) Chapters 1520.
Recommended reading
R
R
R
R
4.1
Bradley (2008) Sections 9.4 and 9.5.

Dowling (2000) Chapters 1011 and Section 12.7.
Ostaszewski (1993) Sections 6.5, 6.7, 6.9, 6.10 and 6.11.
Simon and Blume (1994) Sections 7.17.3, 8.18.5, 9.1, 23.1 and 23.3.
Introduction
You will have seen from 05A Mathematics 1 that matrices and linear equations play
an important role in analysing economics mathematically. In this chapter we develop
further the theory of matrices and linear equations and present some applications.
4.2
4.2.1
Revision of Mathematics 1
Matrices
You should recall how to add and multiply vectors and matrices. I will not reiterate all
the definitions here, but if you need to refresh your memory, consult the Mathematics 1
subject guide or the textbooks (particularly the Anthony and Biggs book).
4.2.2
Linear equations
The most important use of matrices we met in Mathematics 1 was their application to
solving systems of linear equations.
Recall that a system of m linear equations in n unknowns x1 , x2 , . . . , xn is a set of m
51
4. Linear algebra and applications
equations of the form
a11 x1 + a12 x2 + + a1n xn

a21 x1 + a22 x2 + + a2n xn
..
.
=
=
am1 x1 + am2 x2 + + amn xn
b1
b2
..
.
bm .
The numbers aij are usually known as the coefficients of the system. We say that
(x1 , x2 , . . . , xn ) is a solution of the system if all m equations hold true when x1 = x1 ,
x2 = x2 and so on. Sometimes a system of linear equations is known as a set of
simultaneous equations; such terminology emphasises that a solution is an assignment of
values to each of the n unknowns such that each and every equation holds with this
assignment.
In order to deal with large systems of linear equations we usually write them in matrix
form. First we observe that vectors are just special cases of matrices: a row vector or list
of n numbers is simply a matrix of size 1 n, and a column vector is a matrix of size
n 1. The rule for multiplying matrices tells us how to calculate the product Ax of an
m n matrix A and an n 1 column vector x. According to the rule, Ax is

a11 x1 + a12 x2 + + a1n xn
a11 a12 ... a1n
x1
a21 a22 ... a2n x2 a21 x1 + a22 x2 + + a2n xn

.
..
..
..
.. .. =
..
.
.
.
.
.
am1 x1 + an2 x2 + + amn xn
am1 am2 ... amn
xn
Note that Ax is a column vector with m rows, these being the left-hand sides of our
system of linear equations. If we define another column vector b, whose components are
the right-hand sides bi , the system is equivalent to the matrix equation
Ax = b.
We often use the phrase linear system to mean system of linear equations and we say
that a linear system is square if the number of equations is the same as the number of
unknowns; that is, if the matrix A is square.
In Mathematics 1, we used the method of row operations (or, as it is also known, the
Gauss-Jordan elimination method) to solve some systems of linear equations. In this
chapter of the guide, we shall use this method again to solve many more types of
systems of linear equations. (Revision of the row operations method is deferred until
later in this chapter, since we shall want to develop the method further than we did in
Mathematics 1.)
4.3
Matrix inverses
Recall that a matrix A is square if it has the same number of rows as columns. We say
that A has an inverse matrix1 if there is a square matrix B such that
AB = BA = I,
1
52
4.4. Inverse matrices and linear equations
the identity matrix. It turns out that a square matrix A either may have no inverse at
all. But, if it does, then it has only one, which we denote by A1 . We say that A is
invertible or nonsingular if it has an inverse. We say that A is non-invertible or singular
if it has no inverse.
For example, if

a b
A=
c d
where ad bc 6= 0,
then A has an inverse and

A
1
=
ad bc

d b
.
c a
Activity 4.1 Check that this is indeed the inverse of A by showing that its
products with A are the identity matrix I.
The number ad bc is called the determinant of A and is denoted |A|. (I shall say more
about determinants very soon.) Thus, a 2 2 matrix has an inverse if and only if its
determinant is not 0.
Activity 4.2
4.4

3 2
Let A =
. Show that
5 1

1 1 2
1
A =
.
7 5 3
Inverse matrices and linear equations
Suppose we have to solve the system of linear equations Ax = b, where A is a square

matrix. If A has an inverse then
A1 Ax = A1 b,
that is, x = A1 b. Clearly, A1 b is a solution since A(A1 b) = (AA1 )b = Ib = b.
Thus, if A is square and has an inverse then the system of linear equations Ax = b has
the unique solution x = A1 b.
Example 4.1
Consider the following very simple system of linear equations.

3x + 2y = 2
5x + y = 2.
This can be written as
(1)
(2)

3 2
x
2
=
,
5 1
y
2
53
i.e.,
Ax = b,
where

x
x=
y

2
and b =
.
2
Now, |A| = 3 1 2 5 = 7 6= 0, so the linear system has precisely one solution,

x = A1 b.
Now, by the Activity above,
1
=
7

1 2
,
5 3
and the solution is therefore

1 1 2
x
2
2/7
=
=
.
y
4/7
5
3
2
7
That is, x = 2/7 and y = 4/7.
4.5
4.5.1
Determinants
The determinant
The determinant 2 of a square matrix A is a particular number associated with A,

written det(A) or |A|. When A is a 2 2 matrix, the determinant is given by the formula

a b
a b
= ad bc.
det
=
c d
c d
For example,

1 2
det
= 1 3 2 2 = 1.
2 3
For a 3 3 matrix, one way
a b
det d e
g h
of calculating the determinant is as follows:

a b c
c

f = d e f
g h i
i

e f
d f
d e
b

= a
g i + c g h
h i
= aei af h + bf g bdi + cdh ceg.
We have used expansion by first row to express the determinant in terms of 2 2

determinants. Note how this works and note, particularly, the minus sign in front of the
2
See Chapter 20 of Anthony and Biggs (1996).
54
4.5. Determinants
second 2 2 determinant. (Expansion by other rows or by columns is also possible: see

the textbooks for details.)
Example 4.2

1 1 2

1 1
0 1
0 1
0 1 1 = 1

1 2 1 3 2 + 2 3 1 = 1(2 1) 1(0 3) + 2(0 3) = 2.
3 1 2
Determinants of larger matrices (4 4 and so on) can be evaluated recursively in terms

of 3 3 determinants, but we present an easier method below. We refer to computing
determinants in this way as computation by expansion.
Activity 4.3
4.5.2

1 1 0

Calculate 2 0 1.
3 1 1
Calculating determinants using row operations
Determinants of large matrices can often be more easily calculated by using a technique
based on row operations. Recall that there are three basic types of row operation (which
are often combined together):
(i) multiply every entry of a row of the matrix by a non-zero constant,
(ii) add a multiple of one row of the matrix to another,
(iii) interchange two rows of the matrix.
It turns out that performing row operations on a matrix has the following effects on its
determinant:
(i) If any row of a matrix is multiplied by a constant c, its determinant is also
multiplied by c.
(ii) If a multiple of one row is added to another, the determinant is unchanged.
(iii) If two rows are interchanged, the determinant is multiplied by 1.
These observations, combined with the fact that the identity matrix has determinant
equal to 1, will enable us to compute determinants.
Note first that if a matrix is a diagonal matrix, by which we mean it has non-zero
entries only on its main diagonal, then its determinant is the product of these numbers.
That is, if
c1 0 0 . . . 0
0 c2 0 . . . 0
C = 0 0 c3 . . . 0 ,
.. .. .. . .
.
. . .
. ..
0 0 0 . . . cn
55
then det(C) = c1 c2 . . . cn . This is because C is obtained from I by multiplying the first

row by c1 , the second row by c2 , and so on. Hence, using the observations above,
det(C) = c1 c2 . . . cn det(I) = c1 c2 . . . cn .
Now suppose we have a matrix in which every entry below the main diagonal is zero,
known as an upper triangular matrix. That is, the matrix takes the form
c1 . . .
0 c2 . . .
0 0 c3 . . .
C=
,
.. .. .. . .
..
. . .
. .
0 0 0 . . . cn
where the s are any numbers. Such a matrix can be reduced to the diagonal matrix
c1 0 0 . . . 0
0 c2 0 . . . 0
0 0 c3 . . . 0
D=
,
.. .. .. . .
.
. . .
. ..
0 0 0 . . . cn
by adding multiples of one row to another, without altering the diagonal entries. (Note
that adding a negative multiple of a row is equivalent to subtracting a multiple of the
row.) Hence, again by the observations above,

c1 . . . c1 0 0 . . . 0

0 c2 . . . 0 c2 0 . . . 0

0 0 c3 . . . 0 0 c3 . . . 0

=
= c1 c2 . . . c n .
.. .. .. . .
.. .. .. .. . .
..
. . .
. . . . .
. .

0 0 0 . . . cn 0 0 0 . . . cn
In other words, we have proved that:
The determinant of an upper triangular matrix is equal to the product of its
diagonal entries.
It follows that we can calculate the determinant of a matrix by reducing it to upper
triangular form, using row operations of types (ii) and (iii).
Example 4.3
The matrix
1
0
A=
1
1
is reduced
1 3 2
0 0 3
1 5 4
1 2 1
56
to an upper
5
1
2
0
0
0
1
0
3
0
5
2
2
3
4
1
triangular matrix T as
3
2
5
1
0
3
2
0
0
2
2 5
1 1 4
0
5
2
,
0
1
follows:
3
2
5
1
0
2
2 5
0
0
3
2
1 1 4
0
3
2
0
0
2 5
2 5
= T.
3 2
0 13
2
4.6. Cramers rule
Here, the second step is an interchange of rows (rule (iii)) and the other steps are
covered by rule (ii). It follows that

13
= 39.
det(A) = det(T ) = 1 2 3
2
Activity 4.4
4.6

1 1 0

Calculate 2 0 1 by using the method just described.
3 1 1
Cramers rule
In Mathematics 1, you met the row operations method for solving systems of linear
equations. An alternative technique for solving Ax = b, where A is an invertible n n
matrix, is Cramers rule. Consider first a 2 2 system. Let us write the general system
of two equations in two unknowns in the matrix form Ax = b, that is,

a11 a12
x1
b
= 1 .
a21 a22
x2
b2
Using the formula for the inverse we get
1

b1
a11 a12
x1
=
b2
a21 a22
x2

1
a22 a12
b1
=
b2
a11 a22 a12 a21 a21 a11

1
a22 b1 a12 b2
=
.
a11 a22 a12 a21 a21 b1 + a11 b2
After a minor rearrangement, this tells us that the solutions x1 and x2 can each be
written in terms of determinants as:

a11 b1
b1 a12

a21 b2
b2 a22
and x2 =

x1 =

a11 a12 .
a
a
11
12

a21 a22
a21 a22
Cramers rule in the 2 2 case can be stated in the following way: xi = i /, where
is the determinant of A and i is the determinant of the matrix obtained by replacing
the ith column of A by b. The result in the n n case is just the same, so it is a neat
way of writing down the solution.
Example 4.4
Let us find the solution of the linear system

x1 + 2x2 + x3 = 1
2x1 + 2x2 = 2
3x1 + 5x2 + 4x3 = 1,
57
by using Cramers rule. First, we
1
2
3
write it in matrix form Ax = b as

2 1
x1
1
2 0
x2 = 2 ,
5 4
x3
1
Using the notation above, we have

1 2 1

2 0
2 0
2 2

= |A| = 2 2 0 = 1
2 3 4 + 1 3 5
5
4
3 5 4
= 1 8 2 8 + 1 4 = 4.
The other relevant determinants are as follows. (Check the calculations.)

1 2 1
1 1 1
1 2 1

1 = 2 2 0 = 0, 2 = 2 2 0 = 4, 3 = 2 2 2 = 4.
1 5 4
3 1 4
3 5 1
So the solution is given by
x1 =
1
= 0,
x2 =
2
= 1,
x3 =
3
= 1.
It is very simple to verify that we have the right answer just by checking that all
three equations hold with these values of x1 , x2 , x3 . It is a good idea always to check
your answer in this way.
Activity 4.5 Check that we do indeed have the correct answer by substituting the
values back into the original equations.
4.7
The square linear system when A is not invertible
It is the case that a matrix is invertible if and only if it has non-zero determinant.
When the matrix A has non-zero determinant, the system Ax = b has just one solution,
x = A1 b. When the matrix A has zero determinant, the system Ax = b might have no
solutions at all, or infinitely many solutions.3 For example, the matrix

1 1
A=
,
2 2
has determinant 0. Consider the system
x1 + x2 = 1
2x1 + 2x2 = 2.
Clearly the second equation, being just the first equation doubled, adds no new
restriction on x1 and x2 , so this system is equivalent to the single equation x1 + x2 = 1.
3
See Anthony and Biggs (1996) Sections 16.1 and 17.1.
58
4.8. Row operations
But this has infinitely many solutions; let be any number: then x1 = , x2 = 1 is
a solution. On the other hand, consider the system
x1 + x2 = 1
2x1 + 2x2 = 1.
This has no solutions, for the second equation is equivalent to x1 + x2 = 1/2 and this is
in contradiction with the first equation, since 1 6= 1/2.
The examples just given show that we have to be careful if a square linear system is
defined by a matrix which is not invertible. But things are far simpler in the very
special case in which b is the zero vector,

0
..
0 = . .
0
This always has at least one solution (namely the zero vector itself). The following
simple fact is important:
The square linear system Ax = 0 has solutions other than x = 0 precisely
when |A| = 0.
4.8
Row operations
The inverse matrix method and Cramers rule for solving square linear systems can be
generalised to square linear systems of any size. But they only work when A is
invertible. We need a different technique for dealing with systems where A is not
invertible. It is here that we find row operations again to be very useful. By way of
revision, we briefly review the reasoning behind row operations.
It is a simple observation that the set of solutions of a system of linear equations is
unaltered by the following three operations, since the restrictions on the variables
x1 , x2 , . . . given by the new equations imply, and are implied by, the restrictions given
by the old ones (that is, we can undo the manipulations made on the old system):
multiply both sides of an equation by a non-zero constant
add a multiple of one equation to another
interchange two equations.
These observations form the motivation behind the row operations method4 to solve
linear equations.
To solve a linear system Ax = b (where A need not be square), we first form the
augmented matrix (Ab), which is A with column b tagged on. For example, if (as in our
Example of Cramers rule) the system is

1 2 1
x1
1
2 2 0 x2 = 2 ,
3 5 4
x3
1
4
See Mathematics 1 and Anthony and Biggs (1996) Chapters 16 and 17.
59
then the augmented matrix is the 3 4 matrix
1 2 1 1
(Ab) = 2 2 0 2 .
3 5 4 1
We use this form because of the important fact that elementary operations on the
equations of the system correspond to the same operations on the rows of the augmented
matrix. The method now proceeds as follows: we use a sequence of elementary row
operations on the augmented matrix until we have changed it into a matrix of the form
1 ...
0 0 1 . . .
(Cd) = 0 0 0 1 . . . ,
.. .. .. .. . . .. ..
. . . .
. . .
0 0 0 0 ... 0 0
which is said to be in echelon form. Here, the symbols merely indicate some numbers.
Note that in an echelon matrix, the first non-zero entry in each row is 1 (we call this the
leading 1), the position of the leading 1 moves to the right as we go down the rows, and
any rows which consist entirely of zeros are located at the bottom of the matrix.
To see why this will be useful, let us carry out the procedure for the linear system given
above. We have, by the use of certain row operations,
1 2 1 1
1 2
1
1
2 2 0 2 0 2 2 0
3 5 4 1
0 1 1 2
1 2 1 1
0 1 1 0
0 1 1 2
1 2 1 1
0 1 1 0
0 0 2 2
1 2 1 1
0 1 1 0 .
0 0 1 1
Therefore the initial system has the
1 2
0 1
0 0
same set of solutions as the system

1
x1
1
1
x2 =
0 .
1
x3
1
This system of equations is

x1 + 2x2 + x3 = 1,
x2 + x3 = 0,
x3 = 1.
But it is easy to solve these equations by working backwards from the third equation to
the first one. Immediately, we have x3 = 1. The second equation then gives x2 = 1,
and then the first gives x1 = 1 2x2 x3 = 0.
60
4.8. Row operations
The method based on row operations is more useful than Cramers rule, since, for
instance, it applies to systems of linear equations in which the matrix A is not square.
Moreover, even when A is square, this method may work when Cramers rule does not.
The following two examples demonstrate this.5
Example 4.5
Consider
2x2 + x3 = 1
2x1 + 2x2 = 2
3x1 + 4x2 + x3 = 3.
As usual, we form the matrix
1 2 1 1
2 2 0 2 ,
3 4 1 3
and apply elementary row operations to reduce it to echelon form. (Cramers rule
cannot be applied here, because as one would quickly discover, the determinant of
the coefficient matrix is |A| = 0.)
1 2 1 1
1 2
1 1
1 2
1 1
1 2 1 1
2 2 0 2 0 2 2 0 0 1
1 0 0 1 1 0 .
3 4 1 3
0 2 2 0
0 2 2 0
0 0 0 0
This last matrix represents the system
x1 + 2x2 + x3 = 1
x2 + x3 = 0
0x1 + 0x2 + 0x3 = 0.
The third of these equations conveys no information at all about x1 , x2 and x3 , for it
simply tells us that 0 = 0. Consequently, the original system has the same solutions
as the following system of two equations in three unknowns.
x1 + 2x2 + x3 = 1
x2 + x3 = 0.
These equations tell us first that, given x3 , we have x2 + x3 = 0, so x2 = x3 . Then
we have x1 = 1 2x2 x3 = 1 + 2x3 x3 = 1 + x3 , so both x1 and x2 are determined
in terms of x3 . Indeed, if we let x3 be any real number s, then
x1 = 1 + s,
x2 = s and x3 = s,
is a solution to the system. So there are infinitely many solutions in this example.
See Anthony and Biggs (1996) Chapter 17, for general discussion and further examples.
61
Example 4.6
Consider the system of equations

x1 + 2x2 + x3 = 1
2x1 + 2x2 = 2
3x1 + 4x2 + x3 = 2.
Using row operations to reduce the augmented
1 2 1 1
1 2
1
1
1
2 2 0 2 0 2 2 0 0
3 4 1 2
0 2 2 1
0
matrix to echelon form, we obtain
2
1
1
1 2 1 1
1
1
0 0 1 1 0 .
2 2 1
0 0 0 1
Thus the original system of equations is equivalent to the system

x1 + 2x2 + x3 = 1
x2 + x3 = 0
0x1 + 0x2 + 0x3 = 1.
But this system has no solutions, since there are no values of x1 , x2 , x3 which satisfy
the last equation. It reduces to the false statement 0 = 1, whatever values we give
the unknowns. We deduce, therefore, that the original system has no solutions. Such
a system is said to be inconsistent.
Activity 4.6
Use row operations to show that the system

x1 2x2 + 3x3 = 5
2x1 + x2 x3 = 1
4x1 3x2 + 5x3 = 11,
has infinitely many solutions. Find a formula for the general solution. Use row
operations to show that if the number 11 on the right-hand side of the third
equation is replaced by the number 2, then the resulting system has no solutions.
4.9
Calculating inverses using determinants
We already

have noted a formula for the inverse of an invertible 2 2 matrix: if
a b
A=
and det(A) = ad bc 6= 0, then
c d
62
1
=
det(A)

d b
.
c a
4.9. Calculating inverses using determinants
There is also a formula for the inverse of an invertible 3 3 matrix, though it is much
more complicated.6 Suppose we have the 3 3 matrix
a11 a12 a13

A = a21 a22 a23 .
a31 a32 a33
Then A is invertible precisely when det(A) 6= 0. For each i and j between 1 and 3, let
Aij be the determinant of the 2 2 matrix obtained from A by deleting row i and
column j. For example,

a11 a12
A23 =
.
a31 a32
Then it turns out that if det(A) 6= 0, then A is
+A11
1
A12
A1 =
det(A)
+A13
Note carefully the alternating pattern
The matrix
+A11
A21
+A31
invertible and
A21 +A31
+A22 A32 .
A23 +A33
of signs attached to the entries of the matrix.
A12 +A13
+A22 A23 ,
A32 +A33
with (i, j) entry equal to (1)i+j Aij is often called the adjugate matrix, denoted adj(A),
and the matrix that appears in the above formula for the inverse is
+A11 A21 +A31

A12 +A22 A32 .
+A13 A23 +A33
Note that this is the transpose of the adjugate matrix, denoted adj(A)T , i.e. it is
obtained from the adjugate matrix by interchanging the rows and columns. For this
reason you might sometimes see the formula
1
adj(A)T .
A1 =
det(A)
Calculating the inverse this way requires care, since we have to compute a 3 3
determinant, i.e. det(A), then 9 different 2 2 determinants, then we have to attach the
correct signs to these, and place them in the right positions. Here is an example.
Example 4.7
Let
2 1 3
A = 0 1 1 .
1
2 0
We shall calculate its inverse using the
determinant:

1
det(A) = 2
2
method just described. First, we find its

0 1
0 1
1

1
+ 3
0
1 0
1 2
= 2(2) 1(1) + 3(1) = 8.

6
See Ostaszewski (1993) Section 6.11.4 for an explanation.
63
Since this is non-zero, we know that the inverse exists. We now calculate the relevant
2 2 determinants.

1 1
0 1
0 1
= 2,
= 1,
= 1,
A11 =
A12 =
A13 =
2 0
1 0
1 2

1 3
= 6,
A21 =
2 0

1 3

= 4,
A31 =
1 1
A22

2 3
= 3,
=
1 0

2 1
= 5,
A23 =
1 2
A32

2 3

= 2,
=
0 1

2 1

= 2.
A33 =
0 1
Then, the adjugate matrix is
+A11 A12 +A13

2 1 1
adj(A) = A21 +A22 A23 = 6 3 5 ,
+A31 A32 +A33
4
2 2
and the transpose of the adjugate matrix is given by
T
2 1 1
2 6 4
adj(A)T = 6 3 5 = 1 3 2 ,
4
2 2
1
5 2
which means that
A1
2 6 4
1
1
adj(A)T = 1 3 2 .
=
det(A)
8
1
5 2
You can easily check that this is correct by checking that the product of A with this
matrix is the identity matrix.
Activity 4.7
Check that this answer is correct.
Activity 4.8
Use this method to determine
1 2
A= 1 3
2 5
4.10
the inverse of the matrix
1
1 .
1
Calculating inverses using row operations
Calculation of inverses using determinants is a rather difficult technique, easy to get

wrong, and very impractical for large matrices. Instead we show how one can use
elementary row operations to find the inverse of a matrix.7 We start with the matrix A
7
See Anthony and Biggs (1996) Section 18.3 for an explanation of why this technique works.
64
4.10. Calculating inverses using row operations
and we form a new, larger, matrix by placing the identity matrix to the right of A,
obtaining the matrix denoted (A I). We then use row operations to reduce this to
(I B). If this is not possible (which will become apparent) then the matrix is not
invertible. If it can be done, then A is invertible and A1 = B.
Example 4.8 We use the same matrix as
determine whether the matrix
0
A=
1
in the previous example. In order to
1 3
1 1 ,
2 0
is invertible and, if so, to determine its inverse,
2 1 3
0 1 1
(A | I) =
1
2 0
we form the matrix

1 0 0

0 1 0 .

0 0 1
(We have separated A from I by a vertical line just to emphasise how this matrix is
formed.) Then we carry out elementary row operations.
2 1
3
1 0 0
2 1 3 1 0 0
0 1 1 0 1 0 0 1 1
0 1 0
1
2 0 0 0 1
0 5/2 3/2 1/2 0 1
2 1 3 1
0 0
1 0
0 1 1 0
0
0 4 1/2 5/2 1
1 1/2 3/2 1/2 0

0
1
1
0
1 0
0
1/8 5/8 1/4
0
0
1
1 1/2 3/2 1/2

0
0
1
0
1/8 3/8 1/4
0
1/8
5/8 1/4
0
0
1
1 0 3/2 7/16 3/16 1/8

0
1/8
3/8 1/4
0 1
0 0
1
1/8
5/8 1/4
1 0 0 1/4 3/4 1/2

0 1 0 1/8 3/8 1/4 .
0 0 1 1/8
5/8 1/4
This is now in the form (I | B), so we deduce that A
2 6
1
1
1 3
A =B=
8
1
5
is invertible and that
4
2 .
2
65
Activity 4.9 Look carefully at the calculation just carried out, and write down
which row operations have been used at each stage.
Activity 4.10
matrix
Use the row operations method to determine the inverse of the
1 2 1
A = 1 3 1 .
2 5 1
(This is the matrix in Activity 4.7 above, whose inverse you should have found using
the other method.)
4.11
Application: input-output analysis
Suppose an economy has n interdependent production processes, manufacturing

commodities C1 , C2 , . . ., Cn . The production process for any one of the commodities
requires an input, which uses part of the output of some of the others. In addition,
there is an external demand for each commodity. The problem is to determine the
production schedule which enables each process to meet all the demands for its product.
The following example illustrates how such a problem can be described as a system of
linear equations.8
Example 4.9
as follows:
The production processes for three goods, C1 , C2 , C3 are interlinked,
to produce one unit of C1 requires the input of 0.2 of C1 , 0.4 of C2 and 0.1 of C3 ,
to produce one unit of C2 requires 0.3 of C1 , 0.1 of C2 and 0.3 worth of C3 ,
to produce one unit of C3 requires 0.2 of each of C1 , C2 and C3 .
Suppose that, in a given time period, there is an external demand for d1 of C1 , d2 of
C2 and d3 of C3 . We wish to know the production levels x1 , x2 , x3 of C1 , C2 , C3
required to satisfy all demands in the given period. Consider first the total demand
for C1 . This is d1 , the external demand, plus the quantity required to produce C1 , C2
and C3 . Each unit of C1 requires 0.2 units of C1 , each unit of C2 requires 0.3 of C1
and each unit of C3 requires 0.2 of C1 . Since the quantities of C1 , C2 , C3 being
produced are x1 , x2 , x3 , the total demand for C1 is therefore
x1 = d1 + 0.2x1 + 0.3x2 + 0.2x3 .
Similarly, considering the total demands for C2 and C3 shows that
x2 = d2 + 0.4x1 + 0.1x2 + 0.2x3
x3 = d3 + 0.1x1 + 0.3x2 + 0.2x3 .
8
See Anthony and Biggs (1996) Sections 19.1 and 19.2 for a general analysis.
66
4.12. Eigenvalues and eigenvectors
The system of equations is simply

x = d + Ax,
where

d1
d = d2
d3
and
0.2 0.3 0.2

A = 0.4 0.1 0.2 .
0.1 0.3 0.2
A is known as the technology matrix. Rearranging we get

x Ax = d,
and using the fact that Ix = x, where I is the identity matrix, we see that we have
to solve the linear system
(I A)x = d.
4.12
Eigenvalues and eigenvectors
One of the most useful techniques in applications of matrices and linear algebra is
diagonalisation. Before discussing this, we have to look at the topic of eigenvalues and
eigenvectors. In this subject, we shall be primarily interested in eigenvalues,
eigenvectors and diagonalisation in the case of 2 2 matrices.
4.13
Definitions
Suppose that A is a square matrix. The number is said to be an eigenvalue of A if for

some non-zero vector x, Ax = x. Any non-zero vector x for which this equation holds
is called an eigenvector for eigenvalue or an eigenvector of A corresponding to
eigenvalue .
4.14
Finding eigenvalues and eigenvectors
To determine whether is an eigenvalue of A, we need to determine whether there are

any non-zero solutions to the matrix equation Ax = x. Note that the matrix equation
Ax = x is not of the standard form, since the right-hand side is not a fixed vector b,
but depends explicitly on x. However, we can rewrite it in standard form. Note that
x = Ix, where I is, as usual, the identity matrix. So, the equation is equivalent to
Ax = Ix, or Ax Ix = 0, which is equivalent to (A I)x = 0. Now, a square linear
system Bx = 0 has solutions other than x = 0 precisely when |B| = 0. Therefore, taking
B = A I, is an eigenvalue if and only if the determinant of the matrix A I is
zero. This determinant, p() = |A I|, is known as the characteristic polynomial of A,
since it turns out to be a polynomial in the variable . To find the eigenvalues, we solve
the equation |A I| = 0. Let us illustrate with a very simple 2 2 example.
67
Example 4.10
Let

1 1
A=
.
2 2
Then

1 1
1 0
1
1
A I =
=
,
2 2
0 1
2
2
and the characteristic polynomial is

1

1

|A I| =
2
2
= (1 )(2 ) 2
= 2 3 + 2 2
= 2 3.
So the eigenvalues are the solutions of 2 3 = 0. To solve this, one could use either
the formula for the solutions to a quadratic, or simply observe that the equation is
( 3) = 0 with solutions = 0 and = 3. Hence the eigenvalues of A are 0 and 3.
To find an eigenvector for eigenvalue , we have to find a solution to (A I)x = 0,
other than the zero vector. (I stress the fact that eigenvectors cannot be the zero vector
because this is a mistake many students make.) This is easy, since for a particular value
of , all we need to do is solve a simple linear system We illustrate by finding the
eigenvectors for the matrix of the example just given.
Example 4.11
We find eigenvectors of

1 1
A=
.
2 2
We have seen that the eigenvalues are 0 and 3. To find an eigenvector for eigenvalue
0 we solve the system (A 0I)x = 0: that is, Ax = 0, or

1 1
x1
0
=
.
2 2
x2
0
This could be solved using row operations. (Note that it cannot be solved by using
inverse matrices since A is not invertible. In fact, inverse matrix techniques or
Cramers rule will never be of use here since being an eigenvalue means that
A I is not invertible.) However, we can solve this fairly directly just by looking at
the equations. We have to solve
x1 + x2 = 0
and
2x1 + 2x2 = 0.
Clearly both equations are equivalent. From either one, we obtain x1 = x2 . We can
choose x2 to be any number we like. Lets take x2 = 1; then we need x1 = x2 = 1.
It follows that an eigenvector for 0 is

1
x=
.
1
68
4.15. Diagonalisation of a square matrix
The choice x2 = 1 was arbitrary; we could have chosen any non-zero number, so, for
example, we could also have taken

2
5.2
or
,
2
5.2
as eigenvectors for 0. There are infinitely many eigenvectors for 0: for each 6= 0,

is an eigenvector for 0. But be careful not to think that you can choose = 0; for
then x becomes the zero vector, and this is never an eigenvector, simply by
definition.
To find an eigenvector for 3, we solve (A 3I)x = 0, which is

2 1
x1
0
=
.
2 1
x2
0
This is equivalent to the equations
2x1 + x2 = 0
and
2x1 x2 = 0,
which are together equivalent to the single equation x2 = 2x1 . If we choose x1 = 1,

we obtain the eigenvector

1
x=
.
2
(Again, any non-zero scalar multiple of this vector is also an eigenvector for
eigenvalue 3.)
4.15
Diagonalisation of a square matrix
Square matrices A and B are similar if there is an invertible matrix P such that
P 1 AP = B.
The matrix A is diagonalisable if it is similar to a diagonal matrix; in other words, if
there is a diagonal matrix D and an invertible matrix P such that P 1 AP = D.
Suppose that matrix A is diagonalisable, and that P 1 AP = D, where D is a diagonal
matrix
1 0 0
0 2 0
.. . .
.. .
D = diag(1 , 2 , . . . , n ) = ...
.
.
.
0 0 0
0 0 n
(Note the useful notation for describing the diagonal matrix D.) Then we have
AP = P D. If the columns of P are the vectors v1 , v2 , . . . , vn , then
AP = A(v1 . . . vn ) = (Av1 . . . Avn ),
69
and
1 0
0 2
..
P D = (v1 . . . vn ) ...
.
0 0
0 0
0
0
.. = ( v . . . v ).
1 1
n n
.
0
n
...
So this means that

Av1 = 1 v1 ,
Av2 = 2 v2 ,
...,
Avn = n vn .
The fact that P 1 exists means that none of the vectors vi is the zero vector. So this
means that (for i = 1, 2, . . . , n) i is an eigenvalue of A and vi is a corresponding
eigenvector.
So we have established that an n n matrix A is diagonalisable if we can find an
invertible matrix P whose columns are eigenvectors of A. It is not always the case that
such a matrix can be found, but in the 2 2 case the issue is much simpler: if v and w
are eigenvectors which are not multiples of each other (which will certainly be true if A
has two different eigenvalues and these eigenvectors correspond to exactly one of these
eigenvalues each), then the matrix P with columns v and w will be as required.
Example 4.12
Consider the matrix

1 1
A=
.
2 2
We have seen above that the eigenvalues of A are 0 and 3 and that examples of
corresponding eigenvectors are

1
1
and
,
1
2
respectively. If we let

P =

1 1
,
1 2
then P 1 exists and we should have P 1 AP = D where

0 0
D=
.
0 3
It is easy to check that this is so, by calculating P 1 and then evaluating P 1 AP ,
seeing that this product equals D.
Activity 4.11
Check this: determine P 1 and calculate P 1 AP .
Not all n n matrices have n eigenvectors v1 , v2 , . . . , vn that can be used to find an

invertible matrix P , as the following example shows.
70
4.15. Learning outcomes
Example 4.13
The 2 2 matrix

A=

4 1
,
1 2
has characteristic polynomial 2 6 + 9 = ( 3)2 , so there is only one eigenvalue,

= 3. The eigenvectors are the non-zero solutions to (A 3I)x = 0: that is,

x1
0
1
1
=
.
1 1
x2
0
This is equivalent to the single equation x1 + x2 = 0, with general solution x1 = x2 .
Setting x2= r,we see that the solution set of the system consists of all vectors of
r
the form
as r runs through all non-zero real numbers. So the eigenvectors are
r

1
precisely the non-zero scalar multiples of the fixed vector
. Any two
1
eigenvectors are therefore multiples of each other and hence we cannot form an
invertible matrix P with eigenvectors as its columns. So the matrix is not
diagonalisable.
Learning outcomes
demonstrate the learning outcomes of the Matrices and linear equations chapter of
Mathematics 1
calculate determinants of 2 2 and 3 3 matrices by expansion
calculate determinants by using row operations to reduce a matrix to upper
triangular form
know what is meant by the inverse of a matrix, and what it means to say that a
matrix is invertible
understand why when a system of linear equations has an invertible coefficient
matrix, then there is a unique solution to the system
state and use the formula for the inverse of an invertible 2 2 matrix
use Cramers rule
solve linear systems using row operations
determine, using row operations, when a system is inconsistent
solve, using row operations, systems with infinitely many solutions
calculate inverses of 3 3 matrices by computing the adjugate
calculate matrix inverses using row operations
solve input-output problems
understand what is meant by an eigenvalue and a corresponding eigenvector
determine eigenvalues and eigenvectors of 2 2 matrices
understand what is meant by saying that a matrix is diagonalisable
diagonalise a 2 2 matrix, or determine that it cannot be diagonalised
71

Question 4.1
Suppose
C = C0 + bY,
I = I0 ar
and Y = C + I + G,
where a, b, C0 and I0 are positive constants. Suppose also that

Md = M0 + f Y gr
and Ms = Md ,
where f , g and M0 are positive constants. Show that

f Y gr = Ms M0
(1 b)Y + ar = C0 + I0 + G.
Use matrix algebra to prove that the solutions Y , r to these equations are
Y =
r =
a(Ms M0 ) + g(C0 + I0 + G)
f a + g(1 b)
(1 b)(Ms M0 ) + f (C0 + I0 + G)
.
f a + g(1 b)
Question 4.2
For which value of c is the following system of equations consistent? Find all the
solutions when c has this value.
x1 + x2 + x 3 = 2
2x1 + x2 + 2x3 = 5
4x1 + 3x2 + 4x3 = c.
Question 4.3
A firm manufactures 3 different types of chocolate bar, Abracadabra, Break and
Choca-mocha. The main ingredients in each are cocoa, milk and coffee. To produce
1000 Abracadabra bars requires 5 units of cocoa, 3 units of milk and 2 units of coffee.
To produce 1000 Break bars requires 5 units of cocoa, 4 of milk and 1 of coffee, and the
production of 1000 Choca-mocha bars requires 5 units of cocoa, 2 of milk and 3 of
coffee. The firm has supplies of 250 units of cocoa, 150 of milk and 100 of coffee each
week (and as much as it wants of the other ingredients, such as sugar). Show that if the
firm uses up its supply of cocoa, milk and coffee, then the number of Break bars
produced each week equals the number of Choca-mocha bars produced. How does the
number of Abracadabra bars produced relate to the production level of the other two
bars? Find the maximum possible weekly production of Choca-mocha bars.
72
Question 4.4
Determine the inverse of the matrix
3 2 1
A = 1 7 5 .
1 0 1
Question 4.5
Determine whether the following matrix is invertible, and find its inverse if it is.
1 1 5
2 3 1 .
1 0 14
Question 4.6
Consider an economy with three industries: coal, electricity, railways. To produce $1 of
coal requires $0.25 worth of electricity and $0.25 rail costs for transportation. To
produce $1 of electricity requires $0.65 worth of coal for fuel, $0.05 of electricity for the
auxiliary equipment, and $0.05 for transport. To provide $1 worth of transport, the
railway requires $0.55 coal for fuel and $0.10 electricity. Each week the external demand
for coal is $50 000 and the external demand for electricity is $25 000. There is no external
demand for the railway. Find a system of linear equations which determines the weekly
production schedule for each of the three industries. Express the system in matrix form.
Question 4.7
Find the eigenvalues of the matrix

2 1
.
3 1
Question 4.8
Find the eigenvalues of the matrix

2 4
A=
,
3 3
and determine an eigenvector corresponding to each eigenvalue. Find an invertible
matrix P and a diagonal matrix D such that P 1 AP = D.
We have det(A) = 3(1) 2(5) = 7. This is non-zero, so the inverse exists. Using the
formula for the inverse of a 2 2 matrix, we have

1 1 2
1
1 2
1
=
.
A =
7 5 3
7 5 3
73

We have

1 1 0

0 1
2 1
2 0
2 0 1 = 1

1 1 (1) 3 1 + 0 3 1 = 1(1) (1)(1) = 2.
3 1 1

Using row operations, we have
1 1 0
1 1 0
1 1 0
2 0 1 0 2 1 0 2
1 .
3 1 1
0 4 1
0 0 1
In the first step, we subtracted multiples of the first row from the second and third and
in the second step we subtracted twice the second row from the third. These operations
do not change the determinant, so the determinant is the determinant of the resulting
upper triangular matrix, which is 1(2)(1) = 2.
The augmented matrix is
1 2 3 5
2 1 1 1 .
4 3 5 11
We reduce this to echelon form as follows:
1 2 3 5
1 2 3
5
1 2 3
5
2 1 1 1 0 5 7 9 0 5 7 9 .
4 3 5 11
0 5 7 9
0 0
0
0
It follows that the system is consistent, and equivalent to the system
x1 2x2 + 3x3 = 5
5x2 7x3 = 9.
(The final row translates simply into the statement 0 = 0.) There are infinitely many
solutions. Letting x3 = r, where r can be any number at all, we have 5x2 7x3 = 9, so
x2 = (7r 9)/5. From the first equation,
7 r
2
x1 = 5 + 2x2 3x3 = 5 + (7r 9) 3r = .
5
5 5
So the general solution is
x1 =
7 r
,
5 5
7
9
x2 = r
5
5
and x3 = r.
When 11 on the right-hand side of the third equation is replaced by 2, the augmented
matrix we start with is
1 2 3 5
2 1 1 1 .
4 3 5 2
74
It is easy to see that this reduces to
1 2 3
5
0 5 7 9 .
0 0
0 9
The last row shows that the system is inconsistent in this case (for if it was not, then
we would have 0x1 + 0x2 + 0x3 = 9, or 0 = 9, which is impossible).
This is easily done by verifying that AA1 = I.

The determinant of the matrix is 1 and the adjugate matrix is
2 1 1
adj(A) = 7 3 1 ,
5 2 1
so
A1
2 7 5
1
3 2 .
adj(A)T = 1
=
det(A)
1 1 1
(It can easily be checked that this is correct by calculating AA1 and noting that the
product equals the identity matrix.)
The sequence of row operations performed (in each step) is:
1
R3 R3 + R1 ,
2
5
R3 R3 + R2 ,
2
1
R1 R1 ,
2
R2 R2 ,
1
R3 R3 ,
4
R2 R2 + R3 ,
1
R1 R1 + R2 ,
2
3
R1 R1 + R3 .
2
(Here, R2 R2 + R1 , for example, means that the first row has been added to the
second; in other words, the second row R2 has become what was R2 , with R1 added.)
75
1 2 1 1 0 0
1
1 3 1 0 1 0 0
2 5 1 0 0 1
0
0
0
1
0
0
0
0
0
0
It follows that
2 1 1 0 0
1 2 1 1 0
1 3 2 0 1
0 0
2 1 1
1 2 1 1 0
0 1 1 1 1
2 1 1
0
0
1
3 2
1 0
0 1 1 1 1
2 0 0 1 1
3 2
1 0 1
0 1 1 1 1
0 0 2 7 5
3 2
1 0 1
0 1 1 1 1
A1
2 7 5
3 2 ,
= 1
1 1 1
in agreement with the answer obtained using the other method.

We easily get
P
1
=
3

2 1
,
1 1
using the formula for the inverse. Then,

1 2 1
1
1
P AP =
2
3 1 1

1 2 1
0
=
0
3 1 1

1 0 0
=
3 0 9

0 0
=
.
0 3
76
1
2

3
6
1 1
1 2

Because Y = C + I + G, we have
Y = (C0 + bY ) + (I0 ar) + G or (1 b)Y + ar = C0 + I0 + G.
Because Ms = Md ,
Ms = M0 + f Y gr
or f Y gr = Ms M0 .
We therefore have
f Y gr = Ms M0
(1 b)Y + ar = C0 + I0 + G.
In matrix terms this system is

f
g
Y
Ms M0
=
,
1b a
r
C0 + I0 + G
and the solution is

1

Y
f
g
Ms M0
=
r
1b a
C0 + I0 + G

1
a
g
Ms M0
.
=
C0 + I0 + G
f a + g(1 b) (1 b) f
Thus we have explicit formulae for the solutions Y and r :

Y =
r =
a(Ms M0 ) + g(C0 + I0 + G)
f a + g(1 b)
(1 b)(Ms M0 ) + f (C0 + I0 + G)
.
f a + g(1 b)

The augmented matrix is
1 1 1 2
2 1 2 5 .
4 3 4 c
Performing row operations to
1 1 1 2
1 1
2 1 2 5 0 1
4 3 4 c
0 1
reduce this to echelon form,
1
2
1 1 1
2
1 1 1
2
0
1 0 1 0 1 0 1 0 1 .
0 c8
0 1 0 8c
0 0 0 9c
Looking at the bottom row of the last matrix, we see that the system of equations is
inconsistent unless 9 c = 0; that is, it is consistent only when c = 9. When c = 9, the
system is equivalent to
x1 + x2 + x3 = 2
x2 = 1.
77
Setting x3 = s, we have x2 = 1 and

x1 = 2 x2 x3 = 3 s.
The general solution when c = 9 is therefore

x1
3s
x = x2 = 1 .
x3
s

Let us denote the weekly production levels of Abracadabra, Break and Choca-mocha
bars by a, b, c (respectively), measured in thousands of bars. Then, since 5 units of
cocoa are needed to produce one thousand bars of each type, and since 250 units of
cocoa are used each week, we must have
5a + 5b + 5c = 250.
By considering the distribution of the 150 available units of milk,
3a + 4b + 2c = 150.
Similarly, since 100 units of coffee are used, it must be the case that
2a + b + 3c = 100.
In other words, we have the system of equations
5a + 5b + 5c = 250
3a + 4b + 2c = 150
2a + b + 3c = 100.
We solve this
5 5 5
3 4 2
2 1 3
using row operations, starting with the
250
1 1 1 50
1 1
150 3 4 2 150 0 1
100
2 1 3 100
0 1
augmented matrix, as follows.
1 50
1 1 1 50
1 0 0 1 1 0 .
1 0
0 0 0 0
Therefore, the system is equivalent to

a + b + c = 50
b c = 0,
from which we obtain b = c and a = 50 b c = 50 2c. In other words, the weekly
production levels of Break and Choca-mocha bars are equal, and the production of
Abracadabra is (in thousands) 50 2c where c is the production of Choca-mocha (and
Break) bars. Clearly, none of a, b, c can be negative, so the production level, c, of
Choca-mocha bars must be such that a = 50 2c 0; that is, c 25. Therefore, the
maximum number of Choca-mocha bars which it is possible to manufacture in a week is
78
25000 (in which case the same number of Break bars are produced and no Abracadabra
bars will be manufactured).
We start with the matrix (A | I) and reduce this, using row operations, to (I | A1 ). We
have
1 0 1 0 0 1
3 2 1 1 0 0
1 7 5 0 1 0 1 7 5 0 1 0
1 0 1 0 0 1
3 2 1 1 0 0
1 0 1 0 0 1
1 7 5 0 1 0
3 2 1 1 0 0
1 0 1 0 0 1
0 7 6 0 1 1
0 2 2 1 0 3
1 0 1 0 0 1
0 2 2 1 0 3
0 7 6 0 1 1
1 0 1 0 0 1
0 1 1 1/2 0 3/2
0 7 6
0 1 1
1 0 1
0
0
1
1/2 0 3/2
0 1 1
0 0 1 7/2 1 19/2
1 0 1 0
0
1
0 1 1 1/2 0 3/2
0 0 1 7/2 1 19/2
1 0 0 7/2 1 17/2
8
0 1 0 3 1
0 0 1 7/2 1 19/2
So
A1
7/2 1 17/2
8 .
= 3 1
7/2 1 19/2
(Of course, there is nothing special about the sequence of row operations used here:
there are many possible routes to the required final matrix.) Another approach is to
determine det(A) and the adjugate matrix adj(A) and then use the fact that
A1 =
1
adj(A)T .
det(A)
This will, of course, give the same answer!
79

Suppose we were to try to use the determinant-based method to determine the inverse.
The first thing we would do (before finding the adjugate matrix) is to calculate the
determinant of the matrix. Let us call the matrix A. Then

3 1
2 1
2 3
(1)

det(A) = 1
1 14 + 5 1 0 = 1(42) (27) + 5(3) = 0.
0 14
We can stop now and conclude that the matrix has no inverse, because we know that a
matrix is invertible if and only if it has non-zero determinant.
Alternatively, we can use row operations.
1
(A | I) = 2
1
and we
1
2
1
We start with the 3 6 matrix

1 5 1 0 0
3 1 0 1 0 ,
0 14 0 0 1
reduce this as follows:

1 5 1 0 0
1 1
5 1 0 0
1 1 5 1 0 0
3 1 0 1 0 0 1 9 2 1 0 0 1 9 2 1 0 .
0 14 0 0 1
0 1 9 1 0 1
0 0 0 3 1 1
There is no need to proceed further. Since this last matrix has a row whose first half is
all-zero, we deduce that A is not invertible.
Let x1 be the amount of coal production required (in dollars), x2 the amount of
electricity and x3 the amount of transportation. Then, by considering in turn the total
demand for each of the three commodities,
x1 = 50000 + 0.65x2 + 0.55x3
x2 = 25000 + 0.25x1 + 0.05x2 + 0.10x3
x3 = 0.25x1 + 0.05x2 .
Therefore, we have
1
0.65 0.55
x1
50000
0.25 0.95 0.10 x2 = 25000 .
0.25 0.05
1
x3
0
(The question did not ask you to solve the equations (which is a fairly messy business),
but you can find a solution in Anthony and Biggs (1996) Worked Example 19.3 of
Chapter 19.)
Denoting the matrix by A, we have

2

1
= (2 )(1 ) 3 = 2 3 + 2 3 = 2 3 1.
|A I| =
3
1
80
To determine the eigenvalues, we therefore seek the solutions to 2 3 1 = 0. By the

formula for the roots of a quadratic, we have eigenvalues
3 13
,
=
2
so the eigenvalues are (3 + 13)/2 and (3 13)/2.

Denoting the matrix by A, we have

2
4

|A I| =
= (2 )(3 ) 12 = 2 5 6.
3
3
To determine the eigenvalues, we therefore seek the solutions to 2 3 1 = 0. You

can use the formula for the solutions of a quadratic, but in this case it is quite easy to
spot that the quadratic factorises as ( 6)( + 1), so that the eigenvalues are 6 and 1.
To find an eigenvector for eigenvalue 6 we solve the system (A 6I)x = 0: that is,

4 4
x1
0
=
.
3 3
x2
0
This is clearly equivalent to the single equation x1 = x2 , so a suitable eigenvector is

1
.
1
To find an eigenvector for eigenvalue

3
3
1 we solve the system (A (1)I)x = 0: that is,

4
x1
0
=
.
4
x2
0
This is equivalent to the single equation 3x1 + 4x2 = 0, so a suitable eigenvector is

4
,
3
if we take x1 = 4.
To diagonalise A, we let P be the matrix whose columns are these eigenvectors,

1 4
P =
.
1 3
Then P 1 AP = D, where D is the diagonal matrix

6 0
D=
.
0 1
81
82
Chapter 5
Differential equations
Essential reading
Anthony and Biggs (1996) Chapters 27 and 28. (This does not cover linear
first-order differential equations. For those, consult the other texts.)
Further reading
Binmore and Davies (2001) Chapters 12 and 14.
Ostaszewski (1993) Chapter 16. (Section 16.5 discusses linear differential
equations.)
Simon and Blume (1994) Sections 24.124.3 and 25.125.3.
5.1
Introduction
A differential equation1 for a function y = f (x) is an equation involving the derivatives

of y. For example,
dy
+ y = 0,
dx
is a differential equation. This is said to be a first-order equation since it involves only
the derivative dy/dx. An equation such as
d2 y
dy
+ 3 + 6y = 0,
2
dx
dx
involving the second derivative d2 y/dx2 , is a second-order equation. We consider here
several different types of first-order equations and second-order equations.
Sometimes we shall use t rather than x to denote the independent variable, particularly
when the independent variable is time.
We will need integration to solve differential equations, so it is a good idea to brush up
on that topic, including the techniques of substitution, integration by parts, and partial
fractions.
1
83
5. Differential equations
Differential equations occur very often in many different types of mathematical

modelling. We shall see applications in economics and finance later in this chapter, but
another very interesting field of application is population dynamics, which we will
briefly touch on now.
Suppose that y(t) is the size of a population of a species of animal at time t. (The time
is measured with respect to some reference time, t = 0, and the size need not be an
integer, since it may, for example, measure the population in thousands.) Modelling the
population as a continuous-time, differentiable, function, we interpret the derivative
dy/dt to be the rate of increase of the population. If the population has a constant
growth rate, r, then we have the Malthus equation
dy
= ry.
dt
This is an extremely simple model of population growth: what it says is that the rate at
which the population grows is proportional to the size of the current population. The
growth rate, (dy/dt)/y, will be a constant equal to r. (Think of the constant r as the
difference between the birth and death rates in the population.)
It turns out that the solution of the Malthus equation is y(t) = Kert , where K is the
population y(t) at time t = 0 (the starting time). Thus, this simple model of population
growth leads to an exponentially increasing population size.
Activity 5.1 Show that y(0)ert solves the Malthus equation dy/dt = ry by
calculating its derivative.
More realistically, as the population increases, growth-inhibiting factors may come into
action. (These might arise, for instance, from limitations on space and scarcity of
natural resources.) So we should modify the above model. We may do this by assuming
that the growth rate (dy/dt)/y is not constant, but is a decreasing function of y. The
simplest type of decreasing function is the linear one, a by, for a, b > 0. We then have
dy/dt
= a by,
y
or
dy
= (a by)y,
dt
known as the logistic model of population growth.
5.2
Separable equations
A first-order differential equation is separable2 if it can be written in the form

dy
= f (y)g(x),
dx
2
84
5.2. Separable equations
where f and g are given functions. The point is that the right-hand side is the product
of two functions, one depending only on y and the other only on x. For example, the
equation
dy
= yex ,
dx
is separable, whereas
dy
= x3 y + xy 3 ,
dx
is not, because it is not possible to write x3 y + xy 3 in the form f (y)g(x). To solve the
equation
dy
= f (y)g(x),
dx
we rewrite the equation as
1
dy = g(x) dx.
f (y)
This is called separating the variables. We then integrate each side. The following
example illustrates this technique.
Example 5.1
Solve the differential equation

x3 y 2
dy
=
.
dx
1 + x4
We separate:
x3
1
dx,
dy =
y2
1 + x4
and integrate:
Z
It follows that
1
dy =
y2
x3
dx.
1 + x4
1
1
=
1 + x4 + c,
y
2
where c is some constant. (We have used substitution to integrate the integral on the
right.) Note that we have not indicated a constant of integration on the left-hand
side of the equation. This is because any such constant can be taken over to the
right-hand side, so the constant there covers those arising from both integrations.
Thus, for some constant k (which equals c/2),
y=
2
,
1 + x4 + k
and this is the general solution3 : each value of the constant k gives a function y(x)
satisfying the differential equation.
Often, we know some extra information about the function, such as its initial value
y(0), or some other particular value. We can use this information to find y. For
3
85
example, suppose that y satisfies the differential equation of the example just given, and
that, furthermore, y(0)
= 1. Then, substituting x = 0, y = 1 into the general solution of
the equation, y = 2/( x4 + 1 + k), we obtain
1=
2
,
1+k
so k = 1 and hence
y(x) =
x4
2
.
+1+1
Activity 5.2 Suppose that, in the example just given, we have y(0) = 2 rather than
y(0) = 1. What then would be the solution?
Lets return to the equations we met above in the our discussion of simple population
dynamics. First, we have the Malthus equation
dy
= ry.
dt
We saw that the solution to this was y(0)ert , but how would we have found this? Well,
the equation is easily seen to be separable, and we can proceed as follows.
dy
= r dt,
y
so, integrating both sides, ln y = rt + c. Taking the exponential of each side,
y = ert+c = ert ec = Kert , where K is some (positive) constant (equal to ec ). Now, we
know that this expression should equal y(0) when t = 0, so Ke0 = y(0), which means
K = y(0). Therefore the solution is y(t) = y(0)ert .
The logistic equation
dy
= (a by)y,
dt
is also separable, but trickier to deal with. Lets make it slightly easier by dealing with a
particular case. Suppose that a = 1000, b = 2 and that the initial population is
y(0) = 100. Then
dy
= (1000 2y)y.
dt
Separating,
dy
= dt.
(1000 2y)y
To integrate the left-hand side we can use partial fractions. This tells us that
1
A
B
=
+ ,
(1000 2y)y
1000 2y
y
86
5.2. Separable equations
for some numbers A and B. We need to have Ay + B(1000 2y) = 1. Taking y = 0 we

see that B = 1/1000. Taking y = 500, we see that A = 1/500. Therefore

Z
Z
dy
1
1
2
=
+
dy
(1000 2y)y
1000
1000 2y y
1
1
ln(1000 2y) +
ln y
1000
1000

1
y
=
ln
1000
1000 2y
=
(We omit the constant of integration because, as explained above, it suffices to have a
constant of integration on the other side of the equation.) So the solution y(t) of the
differential equation satisfies

y
1
ln
= t + c,
1000
1000 2y
R
for some constant c. (We have used the fact that dt = t + c.) With a little bit extra
work, we can solve explicitly for y. For, we have,

y
ln
= 1000(t + c) = 1000t + k,
1000 2y
say, and then, taking exponentials,
y
= e1000t+k = Ae1000t ,
1000 2y
where A = ek is some positive constant.
Now, we can rearrange this to obtain
y = (1000 2y)Ae1000t ,
and so
(1 + 2Ae1000t )y = 1000Ae1000t ,
so that
y(t) =
1000
1000
1000Ae1000t
=
=
,
1000t
1000t
1 + 2Ae
2 + (1/A)e
2 + Be1000t
where we have divided numerator and denominator by Ae1000t and denoted 1/A by B.
Now, we know that y(0) = 100, so
100 = y(0) =
1000
,
2+B
and B = 8. So we have now found the population: it is

y(t) =
1000
.
2 + 8e1000t
Notice that, as t tends to infinity, y(t) tends to the limiting population 500, since as
t , e1000t 0. This is in marked contrast to the (perhaps over-simplistic) Malthus
model, where the solution increases exponentially without bound.
87
5.3
Linear equations and integrating factors
A linear first-order differential equation4 is one of the form

dy
+ yP (x) = Q(x).
dx
Note that the right-hand side is a function of x alone. The name linear is used because
y, rather than y 2 , for example, occurs.
There is a general technique for solving linear first-order differential equations, based on
the notion of an integrating factor. This is a function (x) of x with the property that if
we multiply our equation through by (x), then the left-hand side takes the form
d
((x)y) .
dx
For example, ex
2 /2
is an integrating factor for the equation

dy
+ xy = x,
dx
since if we multiply through by ex

ex
2 /2
2 /2
, we obtain
dy
2
2
+ xyex /2 = xex /2 ,
dx
which is
d x2 /2
2
ye
= xex /2 .
dx
When the left-hand side can be written in this form, the equation is easy to solve. The
solution to the equation just given follows from the fact that
Z
2
2
x2 /2
ye
= xex /2 dx = ex /2 + c,
so its solution is
ex /2 + c
2
y=
= 1 + cex /2 .
2 /2
x
e
(We have used substitution to determine the integral.)
The natural question to ask is: how do we find an integrating factor? Fortunately, there
is a simple recipe. If we have the linear equation
dy
+ yP (x) = Q(x),
dx
then an integrating factor is
(x) = e
P (x)dx
Thus, we integrate P and take the exponential of the result. There is no need to include
a constant of integration when integrating P , since we just need one integrating factor.
When we multiply through by (x), the equation becomes
(x)
4
dy
+ (x)P (x)y = (x)Q(x),
dx
See, for example, Ostaszewski (1993) Section 16.5., or Dowling (2000) Chapter 18.
88
5.3. Linear equations and integrating factors
which is
d
((x)y) = (x)Q(x).
dx
(Try to see why: perform the Learning activity below.) Therefore,

Z
(x)y =
(x)Q(x) dx.
To find y, we integrate (x)Q(x), remembering the constant of integration, and then

divide by (x).
Activity 5.3
Verify that, with the notations just used,

(x)
Example 5.2
d
dy
+ (x)P (x)y =
((x)y) .
dx
dx
Consider the equation

x
dy
+ 2y = x2 .
dx
This is not in linear form as there is a function (x) in front of the dy/dx. But we
can re-write the equation in linear form as
dy 2
+ y = x,
dx x
and we see that P (x) = 2/x and Q(x) = x. To calculate an integrating factor, we
first note that
Z
Z
2
dx = 2 ln x = ln x2 .
P (x) dx =
x
So an integrating factor is
2
(x) = eln x = x2 ,
and the solution y satisfies
2
xy=
1
x3 dx = x4 + c,
4
for some constant c of integration. Hence,

c
1
y = x2 + 2 .
4
x
Activity 5.4 Suppose that in the example just given, we have the additional
information that when x = 1, y = 3.
What is the value of the constant c?
89
5.4
Second-order equations
We shall now consider constant-coefficient second-order differential equations;5 these are

of the form
d2 y
dy
()
a 2 + b + cy = f (x).
dx
dx
First, we find the general solution of the related equation
d2 y
dy
(?)
+ b + cy = 0.
2
dx
dx
with 0 on the right, rather than f (x). (This is often known as the corresponding
homogeneous equation, but it should be realised that this use of the word
homogeneous is unrelated to its earlier use, in the discussion of homogeneous functions
and Eulers theorem.) We shall denote the general solution by yc (x), as it is often called
the complementary function.
a
To solve the homogeneous equation, we first consider the auxiliary equation

az 2 + bz + c = 0.
The form of yc (x) will depend on whether this equation has two different solutions, just
one solution, or no solutions.
If the auxiliary equation has two different solutions, r1 , r2 , then the general solution
of (?) is
y(x) = Aer1 x + Ber2 x ,
where A and B are arbitrary constants.
If the auxiliary equation has the same root r repeated twice (that is, it has just one
solution) then (?) has general solution
y(x) = (Ax + B)erx ,
If the auxiliary equation has negative discriminant b2 4ac, then it has no (real)
solutions. In this case, the general solution of (?) involves trigonometrical
functions: we define
b
4ac b2
=
and =
.
2a
2a
Then, in this case, the general solution of (?) is
y(x) = ex (A cos x + B sin x) ,
Now, in order to solve our original equation, which had f (x) on the right, we try to find
any particular solution, yp (x), of the original equation (). Then, the general solution to
() is
y(x) = yc (x) + yp (x),
where yc is (as above) the general solution to equation (?).
We find yp (x) by substitution, and this substitution depends on the nature of f (x).
5
See Anthony and Biggs (1996) Chapter 28.
90
5.4. Second-order equations
If f (x) = k is a constant, we try to find a particular solution of the form yp = c

where c is a (possibly) different constant.
If f (x) = kx, we try yp = cx + d where c and d are constants.
If f (x) = kx2 , we try yp = cx2 + dx + e where c, d and e are constants.
And so on. . .
If f (x) = k sin x or k cos x, we try yp = a sin x + b cos x.
If f (x) = ekx , we try yp = cekx .
(This last one works only if k is not a root of the auxiliary equation.)
Example 5.3
Consider
d2 y
dy
2y = 4x2 .
2
dx
dx
The auxiliary equation is

z 2 z 2 = 0 so (z 2)(z + 1) = 0.
Therefore it has solutions 2, 1 and
yc (x) = Ae2x + Bex ,
for constants A, B. To find a particular solution, we try
yp = cx2 + dx + e,
since the right hand side of the differential equation is a quadratic. (Note that it is
not sufficient to look for a solution of the form ax2 .) Now,
dyp
= 2cx + d and
dx
d2 y p
= 2c,
dx2
so, substituting into the equation, we see that yp is a solution of the equation if and
only if
2c (2cx + d) 2(cx2 + dx + e) = 4x2 ,
that is,
(2c)x2 + (2c 2d)x + (2c d 2e) = 4x2 .
Equating coefficients of x2 , this means that 2c = 4, so c = 2. Then, equating the

coefficients of x, 2c 2d = 0, so d = c = 2. Finally, 2c d 2e = 0, so e = 3.
Therefore, a particular solution is
yp (x) = 2x2 + 2x 3,
and the general solution of the differential equation is
y = Ae2x + Bex 2x2 + 2x 3.
91
Once we have determined the general solution of a second-order equation, the two
unknown constants can be determined if we have some additional information. Often
this is given in the form of the two initial conditions, the values of y(0) and y 0 (0). (Here
y 0 denotes the derivative of y.)
Example 5.4 Let us continue with the example just given. Suppose we want to
find the particular solution satisfying the initial conditions y(0) = 3 and y 0 (0) = 1.
We know that for some constants A and B,
y = Ae2x + Bex 2x2 + 2x 3.
Therefore,
y 0 (x) = 2Ae2x Bex 4x + 2,
and the two initial conditions therefore imply
A + B 3 = 3, 2A B + 2 = 1.
We therefore have the simultaneous equations
A + B = 6, 2A B = 3.
Adding these together gives 3A = 3, so A = 1 and then B = 5. So the particular
solution is
y = e2x + 5ex 2x2 + 2x 3.
Activity 5.5
Suppose, instead, that y(0) = 3 and y 0 (0) = 4.
Determine the particular solution in this case.
5.5
Behaviour of solutions
In practice, we are often less interested in the exact solution of a differential equation
than in how the solution evolves as x tends to infinity.6
If, for example, the auxiliary equation has two solutions and , then the general
solution takes the form
y(x) = yp (x) + Aex + Bex
and its behaviour depends on the particular solution yp and on the functions ex and
ex . For instance, if and are both negative then, as x , ex and ex both tend
to 0, and the general solution of the differential equation behaves in the same way as
yp (x), in the long run.
If the auxiliary equation has no solutions (or, to be precise for those who know about
complex numbers, no real solutions), so that the general solution is of the form
y(x) = yp (x) + ex (E cos x + F sin x) ,
6
92
5.6. Coupled differential equations
then (provided at least one of E and F is non-zero) the solution will exhibit some form
of oscillatory behaviour. For example, suppose the solution is
y(x) = e2x + ex (5 cos 3x + 4 sin 3x) .
Then, since ex 0 as x , the oscillatory part of the solution decreases in
significance as x increases; in other words, the solution is e2x with some oscillations
which fade away in the long run. On the other hand, if the solution were
y(x) = e2x + e4x (5 cos 3x + 4 sin 3x) ,
then the oscillations would increase in magnitude.
5.6
Coupled differential equations
We now turn our attention to coupled differential equations (or systems of differential
equations). First, let us motivate this topic by thinking again about population growth.
Suppose that we have two species of animal, and that they compete with each other (for
food, for instance). Denote the corresponding populations at time t by y1 (t) and y2 (t).
Here is one way to model the populations. We assume that, in the absence of the other,
the population of either species would exhibit logistic growth, as above. But given that
they compete with each other, we assume that the presence of each has a negative effect
on the growth rate of the other. That is, we assume that for some positive numbers
a1 , a2 , b1 , b2 , c1 , c2 ,
1 dy1
= a1 b1 y1 c1 y2
y1 dt
1 dy2
= a2 b2 y2 c2 y1 .
y2 dt
Then we have the coupled system of differential equations
dy1
= a1 y1 b1 y12 c1 y1 y2
dt
dy2
= a2 y2 b2 y22 c2 y1 y2 .
dt
In general, a (square) system of differential equations for the functions y1 , y2 , . . . , yn is of
the form
dy1
= f1 (y1 , y2 , . . . , yn , t)
dt
dy2
= f2 (y1 , y2 , . . . , yn , t)
dt
..
.
dyn
= fn (y1 , y2 , . . . , yn , t).
dt
93
The solution of systems of the generality just described is outside the scope of this
subject. In particular, we shall consider only the case where n = 2, and the functions
f1 , f2 are linear in y1 and y2 . (This means f1 (y1 , y2 ) = ay1 + by2 + g(t) for some constants
a, b and some function g of the variable t, and that the same property holds for f2 .)
We shall explore two ways of solving such systems. One is through reducing the system
of two first-order equations to a second-order equation, and the other involves the use of
diagonalisation. (We shall only consider this second technique for slight specialised
systems of equation, which can be expressed in a straightforward matrix manner.)
5.6.1
Reducing to a second-order equation
Sometimes we can use the methods of solving second-order differential equations to

solve coupled systems of differential equations. The following example illustrates the
technique.7
Example 5.5 Suppose the functions y1 (t) and y2 (t) satisfy y1 (0) = 5, y2 (0) = 2,
and the system
dy1
= 2y1 + 4y2
dt
dy2
= 3y1 + 3y2 .
dt
We will manipulate these equations to obtain a second-order equation for y1 .
Differentiating the first equation, we obtain
d2 y 1
dy1
dy2
=
2
+
4
.
dt2
dt
dt
Now, the second equation gives an expression for dy2 /dt, and substituting this in, we
obtain
dy1
dy1
d2 y 1
=2
+ 4 (3y1 + 3y2 ) = 2
+ 12y1 + 12y2 .
2
dt
dt
dt
Now, we want to express the 12y2 in terms of y1 . From the first equation, we have

1 dy1
y2 =
2y1 ,
4 dt
so
dy1
d2 y1
=2
+ 12y1 + 12y2
2
dt
dt

dy1
1 dy1
+ 12y1 + 12
2y1
=2
dt
4 dt
=5
7
dy1
+ 6y1 .
dt
Anthony and Biggs (1996) Chapter 28, Worked Example 28.6 provides another example, with an
interesting application.
94
Therefore,
d2 y 1
dy1
6y1 = 0.
5
2
dt
dt
The auxiliary equation, z 2 5z 6 = 0, has solutions 6 and 1, and so, for some
constants A and B,
y1 (t) = Ae6t + Bet .
To find y2 , we can use the expression we obtained above:

1 dy1
2y1
y2 =
4 dt

1
6Ae6t Bet 2Ae6t 2Bet
4
3
= Ae6t Bet .
4
=
The initial conditions y1 (0) = 5 and y2 (0) = 2 imply that

3
A + B = 5 and A B = 2.
4
Subtracting the second equation from the first gives (7/4)B = 7, so B = 4, and the
first equation then tells us A = 1. Hence,
y1 (t) = e6t + 4et
y2 (t) = e6t 3et .
Example 5.6
Suppose that the two functions y(t) and z(t) satisfy

dy
= 4y + 2z
dt
dz
= 2y + 4t2 + 4,
dt
and suppose that y(0) = 1 and z(0) = 7/2. We want to find the functions y and z.
(We use y and z to denote the functions rather than y1 and y2 .)
We can obtain a second-order equation from the system as follows. First, if we
differentiate the first equation, we obtain
d2 y
dy
dz
= 4 + 2 .
2
dt
dt
dt
The second equation gives an expression for dz/dt which we may substitute to
obtain:

d2 y
dy
2
2y
+
4t
+
4
,
=
4
+
2
dt2
dt
which is
d2 y
dy
+ 4 + 4y = 8t2 + 8,
2
dt
dt
95
a second order equation for y(t). The general solution (which you should work out
for yourself) is
y(t) = (Ct + D)e2t + 2t2 4t + 5.
We are given that y(0) = 1 and z(0) = 7/2. The second of these initial conditions
may be used to determine y 0 (0) through
y 0 (0) = 4y(0) + 2z(0) = 3.
Thus, having y(0) and y 0 (0), we can determine C and D. For, the condition y(0) = 1
means that D + 5 = 1 (simply putting t = 0) and, since
y 0 (t) = (2Ct + (C 2D)) e2t + 4t 4,
y 0 (0) = 3 reveals that C 2D 4 = 3. These two equations for C and D have

solution C = 1, D = 4. Therefore,
y(t) = (4 + t)e2t + 2t2 4t + 5.
We could find z by using the second of the two equations given at the start, but it is
easier to use the first. This tells us that
dy
= 4y + 2z,
dt
so
1
z(t) =
2
Activity 5.6
dy
+ 4y
dt

=

9
+ t e2t + 4t2 6t + 8.
2
Show that the general solution in the last example is indeed

y(t) = (Ct + D)e2t + 2t2 4t + 5.
Activity 5.7
infinity?
5.6.2
How does the solution y(t) in this last example behave as t tends to
Using diagonalisation
We shall consider how to use matrix diagonalisation to solve systems of the form
dy
= Ay,
dt
where
96

y1
y2

y = ..
.
yn
and
y10 (t)
0
dy
y2 (t)
= .. ,
dt
.
yn0 (t)
and A is a square n n matrix. (The examples well work with will all have n = 2, and
it is only such systems you will have to master, but there is some advantage in
describing the technique for general values of n.) This is a slightly more restricted type
of system than we considered above, because the right-hand sides of the equations
involve only the unknown functions yi (t) and nothing else. That is, we consider
equations of the form
dy
= Ay,
dt
rather than

dy
g1 (t)
= Ay +
.
g2 (t)
dt
This latter type of equation can be tackled using diagonalisation, but it is rather more
complicated to do so, and we shall not require you to do so for this subject.
Suppose that the matrix A can be diagonalised, meaning that there is an invertible
matrix, P , and a diagonal matrix, D, such that
P 1 AP = D.
(Refer to Chapter 4 of the guide: you will note that the columns of P must be
eigenvectors of A.) Let us introduce n new functions z1 , z2 , . . . , zn which are related to
y1 , y2 , . . . , yn by the equation
y = P z,
where

z1
z2

z = .. .
.
zn
(In other words, z = P 1 y.) Then, it can be seen that
dy
d
dz
=
(P z) = P ,
dt
dt
dt
and so
P
dz
= Ay = AP z.
dt
Multiplying both sides on the left by P 1 we obtain

dz
= P 1 AP z = Dz.
dt
Now, the point is that this is an easy system to solve for z because D is diagonal.
So much for the theory. Lets look at an example which will hopefully make things
clearer. (This is the same example we dealt with above by reduction to a second order
differential equation.)
97
Example 5.7 Suppose the functions y1 (t) and y2 (t) satisfy y1 (0) = 5, y2 (0) = 2,
and the system
dy1
= 2y1 + 4y2
dt
dy2
= 3y1 + 3y2 .
dt
In matrix form
dy
= Ay, this is
dt
dy1

dt
y1
= 2 4
.
dy
3
3
y2
2
dt
Now, the matrix

2 4
A=
,
3 3
was considered in the last sample examination question of the previous chapter. We
saw there that the eigenvalues are 6 and 1, and that corresponding eigenvectors are

1
4
and
,
1
3
respectively. As a result, if we take

1 4
P =
1 3

and
D=

6 0
,
0 1
then P 1 AP = D where D is a diagonal matrix. We introduce functions z1 , z2 which

satisfy

y1
1 4
z1
=
,
y2
1 3
z2
then, as above, the equation dy/dt = Ay implies dz/dt = Dz. That is, we now have
dz1

dt
6
0
z1
.
dz = 0 1
z2
2
dt
Unravelling this system, we obtain
dz1
= 6z1
dt
dz2
= z2 .
dt
Now, do you see whats happened? The system we have now is not coupled in the
same way as the original system: the equation for z1 involves only z1 and the
98
5.7. Applications of differential equations
equation for z2 involves only z2 . This uncoupling has occurred because D is a

diagonal matrix. This system is much, much easier to solve. Immediately, we have
z1 = z1 (0)e6t from the first equation, and z2 = z2 (0)et from the second. To find
z1 (0) and z2 (0), we use the known values of y1 (0) and y2 (0) and the fact that
z = P 1 y. Calculating P 1 in the usual way and setting t = 0 we obtain

1 3 4
5
1
z1 (0)
1 y1 (0)
=
=
.
=P
y2 (0)
1
1
2
1
z2 (0)
7
So, z1 = e6t and z2 = et . But we need y1 , y2 , not z1 , z2 . However, y = P z, so

6t 6t

y1
1 4
z1
1 4
e
e + 4et
=
=
= 6t
.
y2
1 3
z2
1 3
et
e 3et
Finally, then,
y1 (t) = e6t + 4et

y2 (t) = e6t 3et .
This is, of course, the same answer as we obtained using the other method above.
You might wonder what the point of the diagonalisation method is if we can in any case
solve such systems by reduction to a second-order equation. One answer to this is that
the diagonalisation method generalises to systems involving more than 2 unknown
functions (the case n > 2).
5.7
Applications of differential equations
There are many applications of differential equations in economics and finance, and this
section briefly discusses several of them.
5.8
Macroeconomics
In macroeconomics, one is concerned with the relationships between investment,

production, income and consumption.8 In the continuous model these are represented
by I(t), Q(t), Y (t) and C(t), respectively. We have the equilibrium conditions
Q(t) = Y (t)
Y (t) = C(t) + I(t).
In addition, we have certain other equations describing reasonable hypotheses about the
relationships between the various quantities. For example, suppose that the
consumption C(t) and income Y (t) are linked by the equation
C(t) = c + bY (t),
8
99
where b and c are constants. This implies that

dY
dC
=b
.
dt
dt
If we think of C as a function of Y , this means that
dC
= b,
dY
which explains why economists refer to the constant b as the marginal propensity to
consume. Using Y = C + I it follows that
I =Y C
= Y c bY
= c + sY,
where s = 1 b is equal to dI/dY and is known as the marginal propensity to invest. In

order to determine how the economy grows, we make the simple assumption that
production increases at a rate proportional to investment. That is,
dQ
= I,
dt
where is a constant. Using the results written down above, we have:
dI dY
dY
dQ
dI
=
=s
=s
= sI.
dt
dY dt
dt
dt
We have obtained the equation
dI
= sI,
dt
which is a differential equation for I. This very simple equation has solution I = Aest
for some constant A.9
5.9
Continuous cash flows
One way of describing how the balance A(t) of an account varies with time under
continuous compounding is to say that the rate of increase is proportional to the
amount. This translates into the differential equation
dA
= rA,
dt
where r is the interest rate. This separable equation has solution A = A(0)ert , a formula
we have seen before. More complicated analysis is possible using differential equations.
For instance, suppose that money is continuously added to the account at rate f (t).
Then the amount A(t) of money in the account at time t satisfies the differential
equation
dA
= f (t) + rA(t).
dt
9
See Ostaszewski (1993) Section 16.4 for more on macro-economic applications.
100
5.10. Continuous price adjustment
(Of course, a negative rate of flow f (t) represents the situation where money is removed
from the account.) Now, this can be written as
dA
rA(t) = f (t),
dt
which is a linear differential equation. The integrating factor is exp
and so A satisfies
Z
rt
A(t)e = ert f (t) dt.

r dt = ert ,
If f (t) is known, then the integral on the right, and hence A, can be determined. For
example, if f (t) = t, then we have:
Z
rt
A(t)e = tert dt
1
= tert +
r
=
1 rt
e dt
r
tert ert
2 + c,
r
r
where we have used integration by parts. Thus,

1
t
A(t) = 2 + cert .
r r
Now, given that the balance at time t = 0 is A(0), we have
A(0) =
so c = A(0) +
1
and
r2
A(t) = A(0)ert +
Activity 5.8
5.10
1
+ c,
r2
ert (1 + rt)
.
r2
Find A(t) if f (t) = 1 for all t.
Continuous price adjustment
Suppose we consider the price of a commodity to vary continuously with time, and that
initially the price does not equal the equilibrium price. We might expect that the price
perhaps converges to the equilibrium price in time, but to be sure we need to have a
model of how the price changes. Suppose that, initially, the price p is less than the
equilibrium value. We might expect that, since the consumers wish to buy more units,
q D (p), than the suppliers will provide, q S (p), the price will increase; and a reasonable
assumption is that the rate of increase depends in some way on the excess of demand
over supply, x(p) = q D (p) q S (p). This assumption means that the derivative p0 is equal
to some given function of p; in other words we have a differential equation for p(t).
101
Example 5.8
Suppose the demand and supply are given, respectively, by

3q + 2p = 10
and
q 2p = 6.
The usual calculation shows that the equilibrium price is 7/2. Suppose also that
when t = 0 we have p = 3, and the law of price adjustment is p0 = (3x(p))3 .
Explicitly, we have
3
dp
= 27 q D (p) q S (p)
dt

3
10 2p
= 27
(2p 6)
3
= (28 8p)3 .
This is a (rather trivially) separable equation, which can be solved by the methods
of this chapter.
Activity 5.9 Solve the equation just obtained, to find an expression for p(t). How
does p(t) behave as t ?
5.11
Determining demand from elasticity
Recall that the elasticity of demand is defined to be

(p) =
q0p
,
q
where q = q D (p) is the demand function. Suppose that there is a constant r such that
(p) = r for all p. (That is, the elasticity is constant.) Then we have
or
Separating and integrating,
p dq
= r,
q dp
dq
q
= r .
dp
p
Z
1
dq =
q
r
dp.
p
It follows that
ln q = r ln p + c,
where c is some constant. That is, ln q = ln (pr ) + c. Taking the exponential of each
side,
K
q = pr ec = Kpr = r ,
p
c
where we have denoted the positive constant e by K.
102
5.12. Market trends
Activity 5.10 A good has a constant elasticity of demand equal to 3, and the
demand function satisfies q D (1) = 60.
Find an expression for q D (p).
5.12
Market trends
In markets such as the housing market consumers often try to anticipate trends.10 They
base their demand not simply on the current selling price, but on how fast this price has
been rising (or falling) and whether its rate of change is slowing down or speeding up.
For example, suppose that the consumers demand is given by
q = 9 6p + 5
d2 p
dp
2 2.
dt
dt
This corresponds to a linear demand function, modified by taking account of the trend.
Suppose also that the quantity supplied is determined in a similar way, and is given by
q = 3 + 4p
dp d2 p
2.
dt
dt
If the market is in equilibrium at any given time, then we may equate the supply and
demand to obtain an equation for the market price p:
9 6p + 5
dp
d2 p
dp d2 p
2 2 = 3 + 4p
2,
dt
dt
dt
dt
or
dp
d2 p
6 + 10p = 12,
2
dt
dt
a fairly simple second-order differential equation for p.
Learning outcomes
know what is meant by a differential equation
solve a separable equation
solve a linear first-order equation using integrating factors
solve second-order equations
solve coupled differential equations (both by reduction to a second-order equation,
and by diagonalisation)
solve problems involving the applications of differential equations
10
103

Question 5.1
Suppose that the function y(t) satisfies y(0) = 2 and
dy
ty 2
=
.
dt
1 + t2
Find y(t).
Question 5.2
Use an integrating factor to solve the equation.
dy
x
+
y = 1 + x2 .
2
dx 1 + x
5
Question 5.3
Find the general solution of
dy
d2 y
4 + 3y = 0.
2
dt
dt
Question 5.4
Find the general solution of
d2 y
dy
+
2
+ 5y = 0.
dt2
dt
Question 5.5
Suppose that the function y(t) satisfies
d2 y
dy
5
+ 6y = 18 sin t,
dt2
dt
and that y = 0 and dy/dt = 1 when t = 0. Find an expression for y(t).
Question 5.6
The elasticity of demand for a good is
(p) =
2p2
.
p2 + 1
Given that q = 4 when p = 1, find the demand function q D (p).

Question 5.7
2
q S (p) = p 4
3
q D (p) = 20 2p.
104
Assuming that the initial price is p(0) = 6, and the price adjusts over time according to
the equation p0 = x(p)3 , where x(p) is the excess demand x(p) = q D (p) q S (p), find a
formula for p(t).
Question 5.8
Suppose consumers anticipate market trends, according to the demand
d2 P
dP
+ 2.
Q = 42 4P 4
dt
dt
d
If the supply is Qs = 8P 6, write down the equilibrium condition Qd = Qs as a

differential equation and determine P (t) if P (0) = 6 and P 0 (0) = 4. Describe the
behaviour determined by this model. Under what initial conditions will P (t) approach a
finite limit as t ?
Question 5.9
Use matrix diagonalisation to solve the system

0
1 1
y1
y1 (t)
=
.
y20 (t)
2 2
y2
Question 5.10
The functions f (t) and g(t) are related as follows:
df
= 3f (t) g(t)
dt
dg
= 3g(t) f (t).
dt
Find f (t) and g(t) if f (0) = 2 and g(0) = 0.
If f (t) = Kert , where K is a constant, then

df
= rKert = r Kert = rf (t),
dt
so f (t) is a solution of the Malthus equation. This is true for any constant K, but for
the correct answer, the population when t = 0 should equal y(0). So we had better have
f (0) = Ke0 = K equal to y(0), so K = y(0) and the population is y(0)ert . (This
illustrates the difference between the general solution to the differential equation, which
in this case is Kert , and the particular solution given the initial value y(0). This is
something discussed later in the chapter.)
105
Feedback to activity
5.2
We have y(t) = 2/( 1 + x4 + k) and y(0) = 2, so
2
= 2,
1+0+k
and so k = 0 and the solution is simply

y(t) =
2
.
1 + x4

Lets start with the right-hand side. We have
d
d
dy
((x)y) =
y+ .
dx
dx
dx
R
Now, (x) = exp( P (x) dx) so, by the chain rule,
Z

R
d
P (x) dx d
=e
P (x) dx
dx
dx
=e
P (x) dx
P (x)
= (x)P (x),
since, clearly, the derivative of the integral of P (x) is P (x). So,
d
dy
d
((x)y) =
y+
dx
dx
dx
= (x)P (x)y +
dy
,
dx
which is exactly the left-hand side.

We have y = x2 /4 + c/x2 and y(1) = 3, so 3 = 1/4 + c/1 and hence c = 3 1/4 = 11/4.
In this case (see the working in the example), we have the equations A + B 3 = 3 and
2A B + 2 = 4, so
A+B =6
and
2A B = 6.
Adding these shows that 3A = 0, so A = 0 and B = 6, which gives us

y(x) = 6ex 2x2 + 2x 3.
The auxiliary equation is z 2 + 4z + 4 = 0. This has just one solution, namely z = 2, so
the general solution of the homogeneous equation takes the form (Ct + D)e2t . We now
have to find a particular solution of the equation
dy
d2 y
+
4
+ 4y = 8t2 + 8.
dt2
dt
106
Since the right hand side is quadratic in t, we try for a solution of the form
y = at2 + bt + c.
With this y, we would have
d2 y
= 2a.
dt2
dy
= 2at + b and
dt
For y to be a solution, we therefore need
2a + 4 (2at + b) + 4(at2 + bt + c) = 8t2 + 8.

Collecting together like terms,
4at2 + (8a + 4b)t + (2a + 4b + 4c) = 8t2 + 8,
so we need
4a = 8,
8a + 4b = 0 and 2a + 4b + 4c = 8,
from which we obtain a = 2, b = 4 and c = 5. Therefore a particular solution of the

differential equation is 2t2 4t + 5, and the general solution is
y(t) = (Ct + D)e2t + 2t2 4t + 5.
Since e2t tends to 0 (quite quickly) as t tends to infinity, it follows that for large values
of t, y(t) is very close in value to 2t2 4t + 5. So, for large t, the solution behaves like
this quadratic.
We have
A(t)e
rt
Z
=
rt
Z
f (t) dt =
1
ert dt = ert + c.
r
Now, the balance at time t = 0 is A(0), giving

1
A(0) = + c,
r
so c = A(0) + 1/r and we have

1
1
1 rt
(ert 1)
rt
A(t) = + ce = + A(0) +
e = A(0)ert +
.
r
r
r
r
Separating and integrating, we have
Z
1
dp =
(28 8p)3
from which we obtain
Z
1 dt,
1
1
= t + c.
16 (28 8p)2
107
Rearranging,
(28 8p)2 =
1
,
16t + 16c
so that
r
either 28 8p =
1
16t + 16c
r
or 28 8p =
1
.
16t + 16c
We are given that p = 3 when t = 0. Substituting these values, it is clear that the first
of the alternatives holds, because 28 8 3 is positive. Furthermore, we have
r
28 8 3 = 4 =
1
,
16 0 + 16c
so c = 1/256 and
s
28 8p =
1
.
16t + 1/16
Rewriting this equation we get the price function

7 1
p(t) =
2 2
1
.
256t + 1
Observe that, as t , the square root term tends to zero, and so p(t) approaches the
equilibrium price 7/2.
The fact that the elasticity is constant at 3 means that
p dq
= 3.
q dp
This equation is separable. Separating and integrating, we obtain

Z
dq
= 3
q
dp
,
p
so
ln q = 3 ln p + c.
Taking exponentials,
q = e3 ln p + c = p3 ec =
A
,
p3
where A = ec is a positive constant. Now, q = 60 when p = 1, so A/13 = 60. Thus

A = 60 and the demand function is
q = q D (p) =
108
60
.
p3
5.12. Solutions to Sample examination/practice questions
Solutions to Sample examination/practice questions

The equation is separable. Separating and integrating we obtain
Z
Z
1
t
dy =
dt.
2
y
(1 + t2 )1/2
The integral on the left is 1/y + c1 , where c1 is a constant of integration. To evaluate
the integral on the right, let r = 1 + t2 . Then dr = 2t dt and
Z
Z
t
1 1
dr
dt =
2
1/2
1/2
(1 + t )
r 2
= r1/2 + c2
= 1 + t 2 + c2 ,
where c2 is a constant of integration. We therefore have
1
+ c1 = 1 + t2 + c2 ,
y
so
y(t) =
1
,
1 + t2 + c
where c = c2 c1 . Since y(0) = 2, we must have 2 = 1/(1 + c), so c = 1/2 and

1
2
y(t) =
=
.
2
1 + t 1/2
2 1 + t2 1
Here, P (x) = x/(1 + x2 ) and
Z
Z
P (x) dx =
1
x
dx = ln(1 + x2 ) + c,
2
1+x
2
so the integrating factor is

1
2
exp
ln(1 + x ) = exp ln( 1 + x2 ) = 1 + x2 .
2
Therefore
and
Finally, then,
d
y 1 + x2 = 1 + x2 1 + x2 = (1 + x2 ),
dx
x3
y 1 + x2 = x +
+ c.
3
x
x3
c
y=
+
+
2
2
1+x
3 1+x
1 + x2
109
for some constant c.

z 2 4z + 3 = 0.
Since z 2 4z + 3 = (z 1)(z 3) there are two solutions, 1 and 3. The general solution
is therefore
y(t) = Aet + Be3t ,
The auxiliary equation,

z 2 + 2z + 5 = 0, has no roots. In the notation used earlier,
= 1 and = 20 4/2 = 2. The general solution is therefore

y(t) = et (E cos 2t + F sin 2t) ,
where E and F are constants.
The auxiliary equation is z 2 5z + 6 = 0, with solutions 2 and 3. So the solution to the
homogeneous version of the equation is Ae2t + Be3t . Now, the right hand side is 18 sin t,
so we try for a particular solution of the form
y = a sin t + b cos t.
Then,
dy
= a cos t b sin t and
dt
d2 y
= a sin t b cos t,
dt2
so we need
(a sin t b cos t) 5 (a cos t b sin t) + 6 (a sin t + b cos t) = 18 sin t.
On grouping together terms on the left, this becomes
(5a + 5b) sin t (5a 5b) cos t = 18 sin t,
so we need
5a + 5b = 18 and 5a 5b = 0.
This means a = b = 9/5. The solution to the differential equation therefore takes the
form
9
9
y = Ae2t + Be3t + sin t + cos t,
5
5
for some constants A and B. To find these, we use the initial conditions given. First,
because y = 0 when t = 0, we have A + B + 9/5 = 0. Now,
dy
9
9
= 2Ae2t + 3Be3t + cos t sin t.
dt
5
5
110
Therefore the condition dy/dt = 1 when t = 0 means 2A + 3B + 9/5 = 1. So we need to

solve the two equations
A+B =
9
5
4
and 2A + 3B = .
5
Twice the first equation subtracted from the second shows B = 14/5 and then
9
23
A= B = .
5
5
So,
y=
23 2t 14 3t 9
9
e + e + sin t + cos t.
5
5
5
5
We have
p dq
2p
= 2
,
q dp
p +1
where q = q D (p) is the demand function. This may be written as a separable differential
equation
dq
2p
= 2
q.
dp
p +1
Separating and integrating,
Z
Z
1
2p
dp.
dq = 2
q
p +1
We can work out the integrals (using the substitution r = p2 + 1 in the second one), and
obtain
ln q = ln(p2 + 1) + c,
which is valid since q and p2 + 1 are positive. Taking the exponential of both sides
q = ec e ln(p
2 +1)
K
ec
= 2
,
2
p +1
p +1
where K = ec is some positive constant. Given that q(1) = 4, we have 4 = K/2, so

K = 8. Therefore the demand function is
q D (p) =
p2
8
.
+1

We have
D
x(p) = q (p) q (p) = 20 2p

2
8
p 4 = 24 p .
3
3
Hence the differential equation for p(t) is

3
dp
8
= 24 p ,
dt
3
111
which is separable. Separating and integrating,

Z
Z
1
3 dp = 1 dt,
24 83 p
so
3
1
= t + c.
16 24 8 p 2
3
Therefore either
or
3
,
16t + 16c
r
8
3
24 p =
.
3
16t + 16c
8
24 p =
3
Now, p(0) = 6 and so when t = 0, 24

p (8/3)p = 8. This shows that we must take the
positive square root, and that 8 = 3/16c. Therefore, 16c = 3/64 and
8
24 p =
3
3
.
16t + 3/64
Rearranging, we obtain the following formula for p(t):

r
192
3
.
p(t) = 9
8 3 + 1024t
Note that as t the price approaches 9, the equilibrium price.
Setting Qd = Qs we obtain
42 4P 4
dP
d2 P
+ 2 = 8P 6,
dt
dt
which we can write in the standard form as

d2 P
dP
4
12P = 48.
dt2
dt
The auxiliary equation is z 2 4z 12 = 0, with solutions 6 and 2. So the general
solution to the homogeneous equation is
Ae6t + Be2t .
A particular solution is easily seen to be the constant 4, and hence the solution for P
takes the form
P (t) = Ae6t + Be2t + 4.
The condition P (0) = 6 means A + B + 4 = 6. Since
P 0 (t) = 6Ae6t 2Be2t ,
112
the initial condition P 0 (0) = 4 becomes 6A 2B = 4. Solving these two equations for A
and B gives A = B = 1, and so
P (t) = e6t + e2t + 4.
As t , e6t and e2t 0. So the only way in which the function
Ae6t + Be2t + 4,
will tend to a finite limit is if A = 0. In terms of initial conditions, this will mean that
for some constant B, P (0) = B + 4 and P 0 (0) = 2B. Equivalently, P (t) will approach
a finite limit if and only if P 0 (0) = 8 2P (0).
Using the standard method (see Chapter 4, in which this is given as an example), we
can diagonalise A to get

1 1
0 0
P =
and D =
,
1 2
0 3
where P 1 AP = D.
dz
= Dz; that is,
dt
!

z10 (t)
0 0
z1
0
=
=
.
0
0 3
z2
3z2
z2 (t)
Let z1 , z2 satisfy z = P 1 y. Then
This is now an uncoupled system, with solution

z1 (t) = z1 (0) and z2 = e3t z2 (0).
Then,
so

1 1
z1
z1
y1
=P
=
,
z2
y2
z2
1 2
y1 = z1 + z2 = z1 (0) + z2 (0)e3t ,
and
y2 = 2z2 z1 = z1 (0) + 2z2 (0)e3t .
Lets get the answer in terms of y1 (0), y2 (0). Observe that

1 2 1
z1 (0)
y1 (0)
1 y1 (0)
=P
=
,
z2 (0)
y2 (0)
y2 (0)
3 1 1
so
2
1
1
1
z1 (0) = y1 (0) y2 (0) and z2 (0) = y1 (0) + y2 (0).
3
3
3
3
Finally, then,
2
1
y1 (t) = y1 (0) y2 (0) +
3
3

1
1
y1 (0) + y2 (0) e3t ,
3
3
113
and
1
2
y2 (t) = y1 (0) + y2 (0) +
3
3

2
2
y1 (0) + y2 (0) e3t .
3
3

We demonstrate both methods for solving such systems (namely, reduction to a second
order equation, and the use of diagonalisation). The system is
df
= 3f (t) g(t)
dt
dg
= 3g(t) f (t).
dt
Differentiating the first equation and using the second to express dg/dt in terms of f
and g, we have
dg
df
d2 f
df
= 3 3g + f.
=3
2
dt
dt
dt
dt
To get an equation entirely in terms of f , we have to substitute for g. From the first
equation,
df
g = 3f ,
dt
so

df
df
d2 f
df
=
3
+
f
3
3f
8f.
=
6
dt2
dt
dt
dt
Writing this in the standard form,
d2 f
df
6 + 8f = 0.
2
dt
dt
z 2 6z + 8 = 0,
which has roots 2 and 4, So, for some constants A and B,

f (t) = Ae2t + Be4t ,
and
g(t) = 3f (t)
df
= 3Ae2t + 3Be4t (2Ae2t + 4Be4t ) = Ae2t Be4t .
dt
Since f (0) = 2 and g(0) = 0, we have A + B = 2 and A B = 0, so A = B = 1 which

means that
f (t) = e2t + e4t and g(t) = e2t e4t .
Now for the diagonalisation method. In matrix form, the system is
0

f (t)
f
3 1
f
=A
=
.
0
g (t)
g
1 3
g
To find the eigenvalues of A, we solve

3 1
2

1 3 = 6 + 8 = 0.
This has solutions = 2 and = 4. Then:
114
An eigenvector for = 2 will satisfy

1 1
x
0
=
,
1 1
y
0

1
and its easy to see that
will do.
1
An eigenvector for = 4 will satisfy

1 1
x
0
=
,
1 1
y
0

1
and its easy to see that
is a solution.
1
It follows that if

1 1
P =
,
1 1
then
P

2 0
AP = D =
.
0 4
Let h and k be functions defined by

f
h
=P
,
g
k
so that

1 1 1
h
f
1 f
=P
.
=
k
1
1
g
g
2
Then we will have, by the theory outlined in the chapter, that

0

h (t)
h
2 0
h
2h
=D
=
=
.
k 0 (t)
k
0 4
k
4k
So
dh
= 2h and
dt
dk
= 4k,
dt
and hence
h(t) = h(0)e2t
and k(t) = k(0)e4t .
Now,
Finally,

1 1
1 1 1
h(0)
f (0)
1
2
1
=
=
.
=
k(0)
g(0)
0
1
2 1 1
2 1 1

2t 2t

1 1
e
e + e4t
f
1 1
h
=
=
= 2t
.
g
1 1
k
1 1
e4t
e e4t
115
116
Chapter 6
Difference equations
Essential reading
R
R
Anthony and Biggs (1996) Chapters 3, 4, 5, 23 and 24.
Further reading
6.1
Binmore and Davies (2001) Chapter 14.
Introduction
In this chapter of the guide, we consider difference equations (also known as recurrence
equations). Many problems in economics and finance involve difference equations,
particularly those involving quantities which change with time, but not continuously
(such as the balance of a deposit account where interest is paid once a year, at the end
of the year).
6.2
6.2.1
Revision of Mathematics 1
Sequences
A sequence1 of numbers y0 , y1 , y2 , . . . is an infinite and ordered list of numbers with one

term, yt , corresponding to each non-negative integer, t. We call yt1 the tth term of the
sequence. Notice that, in our notation, the first term is y0 . Thus yt is actually the
(t + 1)st term of the sequence. (Be careful not to be confused by this, as some texts
differ.) Often, a sequence is defined explicitly by a formula. For instance, the formula
yt = t2 generates the sequence
y0 = 0, y1 = 1, y2 = 4, y3 = 9, y4 = 16, . . .
and the sequence 3, 5, 7, 9, . . . may be described by the formula
yt = 2t + 3,
1
117
6. Difference equations
for t 0. Two important types of sequence are the arithmetic and geometric
progressions.
The arithmetic progression with first term a and common difference d has its terms
given by the formula yt = a + dt. Note that yt is obtained from yt1 by adding the
common difference d. In symbols, yt = yt1 + d.
The geometric progression with first term a and common ratio x is given by the
formula yt = axt . Notice that successive terms are related through the relationship
yt = xyt1 .
We saw in Mathematics 1 that geometric progressions are important in the study of
compound interest.2 If the annual percentage rate of interest in a savings account is
R%, then the interest rate is r = R/100. If P dollars are deposited in the account, and
if no money is taken from or added to the account, then after t years we have a balance
of P (1 + r)t dollars.
6.2.2
Series
Given a sequence y0 , y1 , y2 , y3 , . . . , a finite series is a sum of the form

y0 + y2 + + yt1 ,
the first t terms added together, for some number t. There are two important results
about series, concerning the cases where the corresponding sequence is an arithmetic
progression (in which case the series is called an arithmetic series) and where it is a
geometric progression (in which case the series is called a geometric series).
If yt = a + dt describes an arithmetic progression and St is the sequence
St = y0 + y1 + y2 + + yt1 ,
then
St =
t(2a + (t 1)d)
t(y0 + yt1 )
=
.
2
2
If x 6= 1 and yt = axt , then the geometric series

St = y0 + y2 + + yt1 = a + ax + ax2 + + axt1
is given by
St = a
6.3
1 xt
.
1x
First-order difference equations
We have noted that if

y0 , y1 , y2 , y3 , . . .
2
118
6.3. First-order difference equations
is an arithmetic progression with common difference d then yt may be related to yt1

through the equation yt = yt1 + d. In fact, this equation, together with the value of y0 ,
tells us precisely what yt is. For example, suppose we have yt = yt1 + 3 and y0 = 5.
Then we can calculate y1 directly as y1 = y0 + 3 = 5 + 3 = 8, and, similarly,
y2 = y1 + 3 = 8 + 3 = 11, and so on. In general, yt will be yt = 5 + 3t. The relationship
yt = yt1 + d is an example of a (first-order) difference equation (or first-order recurrence
equation),3 and the value of y0 is known as an initial condition. The difference equation
yt = yt1 + 3, with y0 = 5, has, as we have seen, the solution yt = 5 + 3t. The question
considered here is how to determine such explicit solutions to difference equations.
The general first-order difference equation is of the form
yt = ayt1 + b,
where a and b are numbers. It is easy to see (and you should convince yourself of this)
that if a = 1 then the sequence of numbers yt is an arithmetic progression with common
difference b and first term y0 . Therefore we shall discuss the solution to such a general
difference equation when a 6= 1.
If b = 0, then yt = ayt1 and so y1 = ay0 , y2 = ay1 = a2 y0 , and, in general, yt = at y0 . In
this case, then, finding the solution is easy; the sequence of numbers yt is a geometric
progression with first term y0 and common ratio a. (In a specific application, y0 would
be given also.)
Difference equations in which a 6= 1 and b 6= 0 occur often in economics. Suppose, for
example, that yt represents the balance in a savings account at the end of t years and
that the interest rate is r. Suppose that, from this account, each year, the investor
withdraws an amount I. Then we have the difference equation
yt = (1 + r)yt1 I.
This is of the form yt = ayt1 + b where a = (1 + r) and b = I.
Lets work out a few terms of a difference equation.
Example 6.1 Suppose that y0 = 1 and for t 1, yt = 2yt1 + 1. The sequence of yt

is not an arithmetic or a geometric progression. (We do not get from one term to the
next simply by adding a fixed constant as we would were it an arithmetic
progression; nor by simply multiplying by a fixed constant as we would were it a
geometric progression. Instead, we both multiply by a fixed constant and then add a
fixed constant.) Working out successive terms, we have
y1 = 2y0 + 1 = 2(1) + 1 = 3
y2 = 2y1 + 1 = 2(3) + 1 = 7
y3 = 2y2 + 1 = 2(7) + 1 = 15,
and so on.
Now, suppose you wanted to know the value of y312 . Do you really want to have to carry
out 312 calculations of the type we have just seen? Certainly not! Which is why we
3
119
want a formula for yt . What we mean by a formula is some expression for yt involving
only t (and not yt1 ), and we call this formula the solution of the difference equation.
For example, we know that the difference equation
yt = yt1 + 2,
where y0 = 1 has solution yt = 1 + 2t (because we recognise it as an arithmetic
progression). To find any particular yt we simply substitute into the formula the desired
value of t. For instance, y312 = 1 + 2(312) = 625.
6.4
Solving first-order difference equations
It is possible to use our knowledge about geometric progressions to find the solution of a
difference equation. Although we shall shortly present a simpler method, lets use
geometric progressions to solve the equation we considered above.
Example 6.2 We consider the equation yt = 2yt1 + 1, with y0 = 1. Lets see how
successive terms are built up. We can see that
y1 = 2y0 + 1 = 2(1) + 1 = 2 + 1
y2 = 2y1 + 1 = 2(2 + 1) + 1 = 22 + 2 + 1
y3 = 2y2 + 1 = 2(22 + 2) + 1 = 23 + 22 + 2 + 1
y4 = 2y3 + 1 = 2(23 + 22 + 2 + 1) + 1 = 24 + 23 + 22 + 2 + 1.
In general, it would appear that
yt = 2t + 2t1 + + 22 + 1.
But this is just a geometric series: perhaps this is clearer if we write it as
yt = 1 + 2 + 22 + + 2t1 + 2t ,
from which it is clear that this is the sum of the first (t + 1) terms of the geometric
progression with first term 1 and common ratio 2. By the formula for the sum of a
geometric series, we have
1 2t+1
= 2t+1 1,
yt =
12
and this is the solution to the difference equation.
Activity 6.1 Check that this formula correctly fits the three values y1 , y2 , y3 worked
out in the example above.
That was a fairly simple equation to solve using geometric series. There is, however, a
more straightforward method for solving first-order difference equations. We now
describe this method.
120
6.5. Long-term behaviour of solution
Suppose we want to find the general solution of the equation yt = ayt1 + b (with a 6= 1)
is. First, note that the constant
b
,
y =
1a
satisfies y = ay + b. This means that if y0 happened to be y , then we would have
y1 = y , y2 = y and, generally, yt = y for all t. For this reason, the number y is known
as the constant solution or time-independent solution. The general solution to the
equation yt = ayt1 + b is
yt = y + (y0 y )at ,
where y = b/(1 a).
Example 6.3
Consider again yt = 2yt1 + 1 with y0 = 1. We have a = 2, b = 1 and

y =
1
b
=
= 1,
1a
12
so the solution is
yt = y + (y0 y )at = 1 + (1 (1))2t = 1 + 2t+1 ,
exactly as we found above.

Example 6.4
We find the solution of the equation

yt = 5yt1 + 6,
given that y0 = 5/2. If we take a = 5 and b = 6 in the standard form

yt = ayt1 + b,
we have exactly the equation given. The first thing to do is to find the constant
solution. By the formula, this is
y =
b
6
3
=
= .
1a
15
2
We can now write down the general solution and insert the given value of y0 , i.e.
3
yt = y + (y0 y )at = + 4(5t ).
2
Activity 6.2
6.5
2
Suppose that yt = yt1 + 5 and that y0 = 2. Find yt .
3
Long-term behaviour of solution
The behaviour of the general solution (or time path) for yt given by
yt = y + (y0 y )at ,
121
depends4 simply on the behaviour of at . For example, if at 0, then the formula tells
us that yt y . We can tabulate the results as follows. (For this table, we assume
y0 6= y because in this case it is clear that the solution is constant, and equal to y .)
Value of a
a>1
0a<1
1 < a < 0
a < 1
Behaviour of at
at
at 0 (decreasing)
at 0 (oscillating)
oscillates increasingly
Behaviour of yt
yt or yt
yt y
yt y
oscillates increasingly
In the first of these cases (a > 1), whether yt or yt will, of course, depend
on the sign of (y0 y ).
Activity 6.3 The case a = 1 is not covered in the table just given. How does the
solution yt behave in this case?
6.6
The cobweb model
Consider an agricultural product for which there is a yearly crop. If there are no
disturbances, the equilibrium price p and quantity q will prevail. Suppose that one
year, however, for some external reason such as drought, there is a shortage, so that the
quantity falls and the price rises to p0 . During the winter the farmers plan their
production for the next year on the basis of this higher price, and so an increased
quantity appears on the market in the next year: specifically q1 = q S (p0 ). Because the
quantity is greater, the price consumers pay falls, to the value p1 = pD (q1 ). Overall, the
effect of the disturbance on the price is that it goes from p0 , which is greater than p , to
p1 , which is less than p . The process is repeated again in the following year: this time
the lower price p1 leads to a decrease in production q2 and that in turn means a higher
price p2 . The next year a similar process takes place, and so on. When the sequences
p0 , p1 , p2 , . . . , and q1 , q2 , . . . , are plotted on the supply and demand diagram, we get a
figure like a cobweb.5 This is the reason for the name cobweb model.
Example 6.5 Generalising the argument above, we see that pt1 determines qt ,
which in turn determines pt , according to the rules
qt = q S (pt1 ),
pt = pD (qt ),
where q S is the supply function and pD the inverse demand function. Suppose that
the demand and supply equations are, respectively, as follows:
q + p = 24, 2q + 18 = p.
Then the equilibrium quantity and price are q = 2, p = 22, and
q S (p) = 0.5p 9,
4
5

122
pD (q) = 24 q.
6.7. Financial applications
The equations linking pt1 , qt and pt are thus

qt = 0.5pt1 9,
pt = 24 qt .
Eliminating qt we obtain a first-order difference equation for pt :

pt = 33 0.5pt1 .
This is in the standard form
yt = ayt1 + b,
with pt replacing yt and a = 0.5, b = 33. The time-independent solution is
b
33
=
= 22,
1a
1 (0.5)
and the explicit solution in terms of p0 is
pt = 22 + (p0 22)(0.5)t .
Note that the time-independent solution is the equilibrium price p = 22, and that in
this case the sequence approaches p in an oscillatory way. We say that we have a
stable cobweb. However, it is possible, for other supply and demand curves, that the
price oscillates about the equilibrium price with ever-increasing magnitude. In such
cases, the price does not approach p and we say we have an unstable or exploding
cobweb.6
6.7
Financial applications
First-order difference equations are very useful in the mathematics of finance.7

We consider how capital accrues under compound interest. In particular, we consider
the situation in which there is a fixed annual interest rate r available to investors, and
interest is compounded annually. In this case, if we invest P then after t years we have
an amount P (1 + r)t . This same result can be derived very simply via difference
equations. If we let yt be the capital at the end of the tth year we have y0 = P and the
difference equation
yt = (1 + r)yt1 ,
for t = 1, 2, 3, . . .. This is in the standard form, with a = 1 + r and b = 0. The solution is
fairly obvious (since this is just a geometric progression).
It might seem unnecessary to use difference equations for such a simple investment
scenario, when it is very easy to determine by elementary means the amount of capital
after t years. However, suppose that we withdraw an amount I at the end of each year
for N years. Then what is the balance of the account after t years? This is less obvious,
but difference equations provide an easy means of determining the answer. As we noted
6
7

123
above, in this case, the difference equation is

yt = (1 + r)yt1 I,
where y0 = P . This is another case of the first-order linear difference equation, in
standard form with a = 1 + r and b = I. The time-independent solution is therefore
y =
I
I
= .
1 (1 + r)
r
The general solution is

and since y0 = P we obtain
yt = y + (y0 y )at ,

I
I
yt = + P
(1 + r)t .
r
r
This formula enables us to answer a number of questions. First, we might want to know
how large the withdrawals I can be given an initial investment of P , if we want to be
able to withdraw I annually for N years. The condition that nothing is left after N
years is, yN = 0. This is

I
I
+ P
(1 + r)N = 0,
r
r
and rearranging, we get

I
(1 + r)N 1 = P (1 + r)N ,
r
so that

r(1 + r)N
I(P ) =
P.
(1 + r)N 1
An inverse question is: what principal P is required to provide an annual income I for
the next N years? Rearranging the equation gives the result

I
1
P (I) =
1
.
r
(1 + r)N
6.8
Homogeneous second-order difference equations
An equation of the form

yt + a1 yt1 + a2 yt2 = 0,
in which a1 and a2 are constants, is a homogeneous linear second-order difference (or
recurrence) equation, with constant coefficients.8 Solving these equations is a lot like
solving second-order differential equations.
The general solution to the equation depends on whether the auxiliary equation
z 2 + a1 z + a2 = 0,
has two distinct solutions, or just one solution, or no (real) solutions. Thus, the form of
general solution depends on the value of the discriminant, a21 4a2 . We consider each
case in turn.
8
124
6.8. Homogeneous second-order difference equations
When the auxiliary equation has two distinct solutions, and , the general
solution is
yt = At + B t ,
where A and B are arbitrary constants. In any specific case, A and B are
determined by the initial values y0 and y1 , as in the following example.
When the auxiliary equation has just one solution, , the general solution is
yt = Ctt + Dt = (Ct + D)t ,
where C and D are arbitrary constants. As in the previous case, the values of the
constants C and D can be determined by using the initial values y0 and y1 .
When the auxiliary equation has no solutions when the quantity a21 4a2 is
negative. In that case, 4a2 a21 is positive, and hence so is a2 . Thus there is a
positive square root r of a2 ; that is, we can define r = a2 . In order to write down
the general solution in this case we define the angle by
a1
a1
cos = = .
2r
2 a2
Then the general solution in this case is

yt = rt (E cos t + F sin t),
where E and F are arbitrary constants.
Example 6.6
We find the general solution of the difference equation

yt 6yt1 + 5yt2 = 0.

z 2 6z + 5 = 0
(z 5)(z 1) = 0,
with solutions z = 1 and z = 5. The general solution is therefore

yt = A(1t ) + B(5t ) = A + B(5t ),
for arbitrary constants A and B.
Example 6.7
Let us find yt if
yt 2yt1 + 4yt2 = 0,
where y0 = 1 and y1 = 1
3. Here, the auxiliary equation,

z 2 2z + 4 = 0,
has no
solutions, so we are in the third case. In the notation used above, we have
r = 4 = 2. It follows that
cos =
2
1
2
= = ,
2(2)
4
2
125
so = /3. The general solution is therefore

t
yt = 2 E cos
t + F sin
t ,
3
3
for arbitrary constants E and F . Putting t
= 0, and using the given initial condition
y0 = 1, we have E = 1. Similarly, y1 = 1 3 implies that
!
1
3
+F
,
1 3 = 2 (E cos(/3) + F sin(/3)) = 2
2
2
so we have 1
6.9
3=1+
3F . Therefore F = 1 and the required solution is

t
yt = 2 cos
t sin
t .
3
3
Non-homogeneous second-order equations
We now consider how to solve an equation of the form

yt + a1 yt1 + a2 yt2 = f (t).
Suppose first that f (t) = k where k is a constant. By analogy with the first-order case,
we start by looking for a constant solution9 where yt = y for all t. This requires that
y + a1 y + a2 y = k
or y =
k
.
1 + a1 + a2
We call y a particular solution and we find that the general solution of the
non-homogeneous equation is equal to the
particular solution + general solution of the homogeneous equation.
This principle holds when f (t) is any other function of t, and in these cases a particular
solution is found by substitution (rather as we did when solving non-homogeneous
second-order differential equations). For example, if f (t) were at , it would be
appropriate to try and find a particular solution of the form cat .
Activity 6.4
Suppose that
yt 5yt1 14yt2 = 18,
where y0 = 1 and y1 = 8. Find yt .

9
126
6.10. Behaviour of the solution
6.10
Behaviour of the solution
You should know how to discuss the behaviour of the solution, or time path.10 For
example, suppose that the solution to a second-order difference equation is
yt = 3t 2t .
How does this behave as t ? As t , we know that both 3t and 2t tend to
infinity, but what happens to their difference? It appears that 3t grows much faster than
2t and thus we might expect that yt . This is indeed the case, as can be seen by
writing
t !
2
3t 2t = 3t 1
.
3
Since (2/3)t tends to zero, it follows that 1 (2/3)t tends to 1. The other factor 3t tends
to infinity, so the product tends to infinity.
6.11
Coupled difference equations
Suppose the sequences yt and zt are related as follows: y0 = 5, z0 = 2, and

yt = 2yt1 + 4zt1
zt = 3yt1 + 3zt1 .
This is an example of a coupled system of difference equations, and our treatment of
such systems will parallel our treatment of coupled systems of differential equations in
the previous chapter. There are, just as for coupled differential equations, two different
methods that we will present. The first involves reducing the system to a second-order
difference equation, and the second involves the use of diagonalisation.
The general system we shall consider will take the form Xt = AXt1 where A is a 2 2
matrix,

yt
yt1
Xt =
and Xt1 =
.
zt
zt1
For example, the system above can be written as

yt
2 4
yt1
=
.
zt
3 3
zt1
6.11.1
Solving by reducing to a second-order equation
Staying with the system y0 = 5, z0 = 2, and

yt = 2yt1 + 4zt1
zt = 3yt1 + 3zt1 ,
heres one way to solve it.
10
See Anthony and Biggs (1996) Chapter 24, for instance.
127
We have, from the first equation, and using the second equation (where we have
replaced all occurrences of t with t 1),
yt = 2yt1 + 4zt1
= 2yt1 + 4(3yt2 + 3zt2 )
= 2yt1 + 12yt2 + 12zt2 .
Now, to get an equation just involving yt , yt1 and yt2 , we need to express zt2 in
terms of the sequence yt . From the first of the two original equations, we have

1
yt 2yt1 .
zt1 =
4
It follows (by replacing all occurrences of t by t 1) that

1
zt2 =
yt1 2yt2 .
4
Therefore,
yt = 2yt1 + 4zt1
= 2yt1 + 12yt2 + 12zt2

12
= 2yt1 + 12yt2 +
yt1 2yt2
4
= 5yt1 + 6yt2 ,
and we have the following second-order difference equation for yt ,

yt 5yt1 6yt2 = 0.
The auxiliary equation, z 2 5z 6 = 0, has solutions 6 and 1, so, for some constants
A and B,
yt = A(6t ) + B(1)t .
Now, as already observed,

1
yt 2yt1 .
zt1 =
4
and so (by replacing all occurrences of t by t + 1)

1
zt =
yt+1 2yt
4

1
=
[A(6t+1 ) + B(1)t+1 ] 2[A(6t ) + B(1)t ]
4

1
=
6A(6t ) B(1)t 2A(6t ) 2B(1)t
4

1
=
4A(6t ) 3B(1)t
4
3
= A(6t ) B(1)t .
4
Now we use the initial conditions y0 = 5 and z0 = 2. The first gives A + B = 5 and the
second A (3/4)B = 2. Subtracting the second of these equations from the first, we
obtain (7/4)B = 7, so B = 4, and A = 1. The solutions are therefore
yt = 6t + 4(1)t
128
and zt = 6t 3(1)t .
6.11. Coupled difference equations
6.11.2
Using diagonalisation
We can use diagonalisation as the key to a general method for solving systems of
difference equations. Given a system Xt = AXt1 , in which A is diagonalisable, we
perform a change of variable, as follows. Suppose that P 1 AP = D (where D is
diagonal). We introduce two new sequences wt , xt , related to yt , zt . The relationship is
given by Xt = P Zt , where

wt
Zt =
.
xt
Equivalently, the new sequences are related to the initial ones by Zt = P 1 Xt . Then the
equation Xt = AXt1 becomes
P Zt = AP Zt1 ,
which means that
Zt = P 1 AP Zt1 = DZt1 ,
so, since D is diagonal, this is very easy to solve for Zt . To find Xt we then use the fact
that Xt = P Zt . This is very similar indeed to the diagonalisation method for solving
coupled differential equations. We now show how this technique works for the system we
considered above.
We saw that the system we are considering as our example can be written in matrix
form as

yt
2 4
yt1
=
.
zt
3 3
zt1
We considered the matrix

2 4
A=
,
3 3
in Chapter 4 (it was the last sample examination question and, indeed, we considered it
in our example of solving coupled differential equations in the previous chapter). There
we saw that, if we take

1 4
6 0
P =
and
D=
,
1 3
0 1
then P 1 AP = D where D is a diagonal matrix.
Let wt , xt be given by Xt = P Zt , where

wt
Zt =
.
xt
Then, as described above, we have Zt = DZt1 , which means

wt
6 0
wt1
=
,
xt
0 1
xt1
so we have
wt = 6wt1
and
xt = xt1 ,
as the (uncoupled) equations for wt and xt . The solutions of these are immediate, i.e.
wt = 6t w0
and
xt = (1)t x0 .
129
Now, Zt = P 1 Xt , so taking t = 0 we have

1 3 4
5
1
w0
1 y0
=
.
=P
=
2
1
x0
z0
7 1 1
Then,

t t
6 + 4(1)t
yt
1 4
wt
1 4
6
.
=
=
=
6t 3(1)t
(1)t
zt
1 3
xt
1 3
Finally, then,
yt = 6t + 4(1)t
as before.
6.12
and zt = 6t 3(1)t ,
Economic applications of second-order

difference equations
Second-order difference equations occur quite naturally in macro-economic modelling.

We shall consider a closed national economy, with no government (for the sake of
simplicity). Three important quantities tell us something about the state of the
economy:
Investment, I.
Income, Y .
Consumption, C.
Suppose we can measure each of the quantities in successive time periods of equal
length (for example, each year). Denote by It , Yt , Ct the values of the key quantities in
time-period t. Then we have a sequence of values I0 , I1 , I2 , . . . , and similarly for the
other quantities. We shall assume that the equilibrium condition Yt = Ct + It holds for
each t.
In the multiplier-accelerator model, we assume that the following equations link the key
quantities:
Ct = c + bYt1 , where c and b are positive constants.
It = i + v(Yt1 Yt2 ), where i and v are positive constants.
Using the equilibrium condition Yt = Ct + It , we can obtain a difference equation

involving only the Y s.
Yt = Ct + It
= (c + bYt1 ) + (i + v (Yt1 Yt2 ))
= (c + i) + (b + v)Yt1 vYt2 .
In other words, we have the second-order difference equation
Yt (b + v)Yt1 + vYt2 = c + i.
130
6.12. Economic applications of second-order difference equations
Example 6.8
Suppose
3
Ct = Yt1
8
1
and It = 40 + (Yt1 Yt2 ),
8
and let us assume the equilibrium condition Yt = Ct + It holds. Let us also suppose
that Y0 = 65 and Y1 = 64.5, and try to determine an expression for Yt .
Arguing as above, we have
Yt = Ct + It

3
1
= Yt1 + 40 + (Yt1 Yt2 )
8
8
1
1
= 40 + Yt1 Yt2 ,
2
8
so
1
1
Yt Yt1 + Yt2 = 40.
2
8

1
1
z 2 z + = 0,
2
8
which has discriminant (1/2)2 4(1/8) = 1/4. This is negative, so there are no
solutions. We are therefore in the third case of a second-order
difference
p
equation. To
proceed, we use the recipe given above to see that r = 1/8 = 1/(2 2), and
(1/2)
2 2
1
cos =
=
= ,
2r
4
2
so = /4. Thus, the general solution to the homogeneous equation in this case is

2 2
t
E cos

t + F sin
t .
4
4
We need a particular solution of

1
1
Yt Yt1 + Yt2 = 40.
2
8
Trying Yt = k, a constant, we see that k (1/2)k + (1/8)k = 40, so k = 64. It follows
that for some constants E and F ,

Yt = 64 +
2 2
t

E cos
t + F sin
t .
4
4
To find E and F we use the initial conditions, Y0 = 65 and Y1 = 64.5. Now,

Y0 = 64 +
2 2
0
(E cos(0) + F sin(0)) = 64 + E,
131
so E = 1. Also,

Y1 = 64 +

= 64 +
2 2

2 2

1
E
2
E cos

4

+ F sin

+F

4

1
= 64 + (E + F ),
4
and since this is 64.5, we have E + F = 2 and hence F = 1. The final answer is
therefore

t

1
t
t
Yt = 64 +
cos
+ sin
.
4
4
2 2
Learning outcomes
6
solve problems involving first-order difference equations
solve problems involving second-order difference equations
analyse the behaviour of solutions to difference equations
solve certain coupled difference equations by reduction to second-order equations,
and by diagonalisation
solve problems involving the application of difference equations.

Question 6.1
Planners believe that, as a result of a recent government grant scheme, the number of
new high technology businesses starting up each year will be N . There are already 3000
such businesses in existence in the country, but it is expected that each year 5% of all
those in existence at the beginning of the year will fail (shut down). Let yt denote the
number of businesses at the end of year t. Explain why
yt = 0.95yt1 + N.
Solve this difference equation for general N . Find a condition on N which will ensure
that the number of businesses will increase from year to year.
Question 6.2
q S (p) = 0.05p 4
132
and
q D (p) = 20 0.15p.
6.12. Sample examination/practice questions
Find the equilibrium price. What is the inverse demand function pD (q)? Suppose that
the sequence of prices pt is determined by pt = pD (q S (pt1 )) (as in the cobweb model).
Find an expression for pt .
Question 6.3
A market for a commodity is modelled by taking the demand and supply functions as
follows:
D(p) = 1 p
and
S(p) = p,
so that when the price p prevails the amount of commodity demanded by the market is
D(p) and the amount which producers will supply is S(p). Price adjusts over time t in
response to the excess of the demand over the supply according to the equation:
pt+1 pt = a(D(pt ) S(pt )),
where a is a positive constant. Initially the price p is p0 = 34 . Solve this equation and
show that over time the price adjusts towards the clearing value (i.e. the price at which
supply and demand are equal) if and only if
0 < a < 1.
Under what circumstances does the price tend towards the equilibrium price in an
oscillatory fashion? What happens to the price if a = 12 ?
Question 6.4
Find the general solution of the difference equation
yt yt1 6yt2 = 0.
Question 6.5
(a) Suppose that consumption this year is the average of this years income and last
years consumption; that is,
1
Ct = (Yt + Ct1 ).
2
Suppose also that the relationship between next years income and current investment is
Yt+1 = kIt , for some positive constant k. Show that, if the equilibrium condition
Yt = Ct + It holds, then

k+1
k
Yt
Yt1 + Yt2 = 0.
2
2
(b) In the model set up in part (a), suppose that k = 3 and that the initial value Y0 is
positive. Show that Yt oscillates with increasing magnitude.
(c) Find the values of k for which the model set up in part (a) leads to an oscillating Yt ,
and determine whether or not the oscillations increase in magnitude. (Remember we are
given that k > 0.)
133
Question 6.6
Suppose the sequences yt and zt satisfy
yt = 3yt1 zt1
zt = yt1 + 3zt1 .
If y0 = 2 and z0 = 1, find expressions for yt and zt .
We do indeed have y1 = 22 1 = 3, y2 = 23 1 = 7 and y3 = 24 1 = 15.
For yt = (2/3)yt1 + 5, we have a = 2/3 and b = 5. Then,
y =
and the solution is
5
b
=
= 15,
1a
1 (2/3)
t
t
2
2
yt = y + (y0 y )a = 15 + (2 15)
= 15 13
.
3
3

When a = 1, we have
yt = y + (y0 y )(1)t ,
and y alternately takes two values: it flips between the value

y + (y0 y ) = y0 ,
when t is even, and the value
y (y0 y ) = 2y y0 ,
when t is odd.
z 2 5z 14 = (z + 2)(z 7) = 0,
with solutions 2 and 7. The homogeneous equation
yt 5yt1 14yt2 = 0,
therefore has general solution yt = A(2)t + B(7t ). A particular solution of the

non-homogeneous equation is the constant solution y = 18/(1 5 14) = 1, so this
equation has general solution
yt = 1 + A(2)t + B(7t ).
To find the values of A and B we use the given values of y0 and y1 . Since y0 = 1, we
must have 1 + A + B = 1 and since y1 = 8, 1 2A + 7B = 8. Solving these, we
obtain A = 1 and B = 1, and therefore
yt = 1 (2)t + 7t .
134

Since 5% of the yt1 businesses in operation at the start of year t fail during that year,
it follows that 95% of these survive. Additionally, N new businesses are created, so
yt = 0.95yt1 + N.
This is a first-order difference equation with, in the standard notation, a = 0.95 and
b = N . Also, from the given information, y0 = 3000. The time-independent solution is
y =
b
N
N
=
=
= 20N,
1a
1 0.95
0.05
and the solution is

yt = y + (y0 y )(0.95)t = 20N + (3000 20N )(0.95)t .
There are several ways to solve the last part of the question. Perhaps the easiest way is
to notice that since (0.95)t decreases with t, yt will increase with t if and only if the
number, (3000 20N ) multiplying (0.95)t is negative. So we need 3000 20N < 0, or
N > 150.
The equilibrium price is given by 0.05p 4 = 20 0.15p, so p = 120. The inverse
demand function is obtained by solving the equation q = 20 0.15p for p, so
pD (q) = p =
400 20
q.
3
3
Now,
pt = pD (q S (pt1 ))
= pD (0.05pt1 4)
400 20
(0.05pt1 4)
3
3
1
= 160 pt1 .
3
This has time-independent solution

p =
160
= 120,
1 (1/3)
which is the equilibrium price. The solution for pt is

t
1
.
pt = 120 + (p0 120)
3
135

We have
pt+1 pt = a ((1 pt ) pt ) = a 2apt ,
so pt+1 = (1 2a)pt + a, which is entirely equivalent to the equation
pt = (1 2a)pt1 + a.
Now, the time-independent solution is
p =
1
a
= ,
1 (1 2a)
2
and so
1
pt = p + (p0 p )(1 2a) = +
2
3 1
4 2
(1 2a)t =
1 1
+ (1 2a)t .
2 4
The equilibrium price is given by 1 p = p, and so is 1/2. From our expression for pt ,
we see that pt 1/2 as t if and only if (1 2a)t 0. For this to be true, we need
1 < 1 2a < 1, which is equivalent to 0 < a < 1. The price will oscillate towards 1/2
when, additionally, 1 2a is negative. So this happens when 1/2 < a < 1. When
a = 1/2, 1 2a = 0 and the price pt equals 1/2 for all t.
z 2 z 6 = (z 3)(z + 2) = 0,
so for some constants A and B
yt = A(3t ) + B(2)t .
(a) There are a number of ways of deriving the difference equation. We note first that
substituting It = Yt+1 /k in the equation Yt = Ct + It gives
1
Yt = Ct + Yt+1 .
k
Thus,
C t = Yt
and so (replacing t by t 1),
Yt+1
k
Yt
.
k
Substituting in the equation Ct = (Yt + Ct1 )/2, we get

Yt+1
1
Yt
Yt
=
Yt + Yt1
.
k
2
k
136
Ct1 = Yt1
Rearranging and replacing t by t 1, we obtain the second-order difference equation

k
k+1
Yt1 + Yt2 = 0.
Yt
2
2
(b) When k = 3, the difference equation is
3
Yt 2Yt1 + Yt2 = 0.
2
The auxiliary equation z 2 2z + (3/2) = 0 has no solutions, so the solution for Yt is
Yt = rt (E cos t + F sin t) ,
where r =
p
3/2 and
p
p
cos = (2)/(2 3/2) = 2/3.
Since E = Y0 , and Y0 is positive, we have E > 0. Also r > 1, so Yt oscillates with

increasing magnitude.
(c) In general, the auxiliary equation for the difference equation is

k
k+1
2
z + = 0.
z
2
2
This has no solutions if
k+1
2
2

k
<4
,
2
that is,
(k + 1)2 < 8k.
In this case the general solution is of the form
r !t
k
Yt =
(E cos t + F sin t) .
2
This solution is oscillatory.
Suppose that (k + 1)2 > 8k. Then the auxiliary equation has the roots
p
(k + 1)/2 (k + 1)2 /4 2k
, =
,
2
and both of these are positive. Then the solution is of the form
Yt = At + B t ,
and since and are positive, in this case there can be no oscillatory behaviour. The
same holds true when (k + 1)2 = 8k.
We have shown that oscillations occur when (k + 1)2 < 8k, in other words when k lies
strictly between the roots of the equation (k + 1)2 = 8k. Rewriting this as the quadratic
equation k 2 6k + 1 = 0, we find that the roots are
3 2 2 and 3 + 2 2.
137
So the model predicts that, when k is between these two numbers, the national income
Yt will oscillate. (In economics language, it will exhibit business cycles.)
Whether the oscillations increase
p ort decrease in magnitude depends
p on k. Since the
solution involves the factor ( k/2) , the oscillations decrease if k/2 < 1, that is, if
k < 2, and increase if k > 2.
In matrix form, this system is

yt
yt1
3 1
yt1
=A
=
.
zt
zt1
1 3
zt1
We will solve it using matrix diagonalisation, though of course we could also reduce it
to a second-order difference equation.
The matrix A is the same as the one in the last sample exam question of the previous
chapter. There, we saw that if

1 1
P =
,
1 1
then

2 0
AP = D =
.
0 4
Let wt and xt be functions defined by

wt
yt
,
=P
xt
zt
so that

1 1 1
wt
yt
1 yt
=P
=
.
xt
zt
zt
2 1 1
Then, we will have (by the theory outlined in this chapter)

wt
wt1
2 0
wt1
2wt1
=D
=
=
.
xt
xt1
0 4
xt1
4xt1
So
wt = 2wt1
and xt = 4xt1 ,
and hence
wt = w0 (2t ) and xt = x0 (4t ).
Now,
Finally,

1 1 1
1 1 1
y0
2
3/2
w0
=
=
.
=
1
1/2
z0
x0
2 1 1
2 1 1

yt
1 1
wt
1 1
=
=
zt
1 1
xt
1 1
so
138
3 t
(2 )
2
1 t
(4 )
2
3 t
(2 )
2
3 t
(2 )
2
+ 12 (4t )
21 (4t )

1
1
t
t
t
t
yt =
3(2 ) + 4
and zt =
3(2 ) 4 .
2
2
Appendix A
Sample examination paper
Important note: This Sample examination paper reflects the examination and
assessment arrangements for this course in the academic year 200910. The format and
structure of the examination may have changed since the publication of this subject
guide. You can find the most recent examination papers on the VLE where all changes
to the format of the examination are posted.
The purpose of this Sample examination paper is to give you an idea of the format of
the paper and the type of questions that might be asked. Note that, unlike most of your
other subjects, this is a two-hour paper rather than a three-hour paper and that all of
the questions on the paper should be attempted.
Not all questions in Section A will necessarily carry the same number of marks. (These
are compulsory questions, so you should do them all anyway.)
Sample examination paper

Time: 2 Hours
This examination has two sections, Section A and Section B.
Attempt ALL questions in Section A.
Attempt BOTH questions in Section B.
Section A represents 60% of the available marks, and each question from Section B
represents 20% of the available marks.
SECTION A
Answer all six questions from this section (60 marks in total).
1.
The supply and demand equations for a good are, respectively,

q = 4p
and
q = 2400 2p,
where q denotes quantity and p the price in dollars. Determine the

equilibrium price and quantity.
Suppose an excise (or per-unit) tax of $300 is imposed. Determine the
new equilibrium price and quantity.
What percentage tax would result in the same equilibrium price?
139
A. Sample examination paper
2.
Expand as a power series, in terms up to x8 , the function

2
f (x) = sin(ex 1).

3.
The function f (x, y) takes the form

f (x, y) =
xy + x y 3
,
x3 + y 3
for some numbers and . If f is homogeneous of degree D, what

must the values of and be (in terms of D)?
Verify explicitly that this function satisfies Eulers equation.
4.
Find an invertible matrix P and a diagonal matrix D such that

P 1 AP = D, where A is the matrix

7 1
.
3 5
Hence find an expression for the nth power of A, i.e.
An = AA
A},
| {z
n times
simplifying your answer as far as possible.

5.
A sequence xt satisfies
xt+1 =
a
axt xt1 ,
4
for all t 1, where a > 0 is a fixed number. If x0 = 1 and x1 =

find a formula (in terms of t and a) for xt .
6.
a,
The supply equation for a good is p = 3 + 3q + 2q 2 where q denotes the

quantity supplied and p the price. The equilibrium price is 30.
Determine the equilibrium quantity and the producer surplus.
If the elasticity of demand for the good is equal to 1 for every value of
the price, determine the demand function.
SECTION B
Answer both questions from this section (20 marks each)
7. (a)
140
Determine the maximum value of xy 2 z 3 subject to the constraint

px + qy + rz = M , where p, q, r, M > 0 are fixed numbers.
A
(b)
Consider the following system of equations.

x y + 2z = 4
3x y z = 0
x + y + az = b.
Use matrix methods to determine what values a and b must take if this
system is consistent and has infinitely many solutions.
What must the value of a not be if the system has precisely one
solution?
What can be said about a and b if the system has no solutions?
8.
A manufacturer adjusts at time t the price p = p(t) of his product by

reference to his current inventory (level of stock) I(t) according to the
equation
dp
= (I(t) I0 ),
dt
where I0 = I(0) is his initial stock. The inventory satisfies the equation
dI
= Q(t) S(t),
dt
where the level of production Q(t) and the level of sales S(t) satisfy,
respectively,
dp
Q(t) = a bp c ,
dt
and
dp
S(t) = p ,
dt
where a, b, c, , , are constants and 6= b.
Show that there is a constant solution p(t) = p for which Q(t) = S(t).
Show that
d2 p
dp
+ ( c) + ( b)p = a.
2
dt
dt
Solve this equation when c = 3, b = 2, a = 1 and when the
initial price is p(0) = 1.
Show that, in general, if > b and > c, then the price tends to p as
t .
141
A. Sample examination paper
142
Appendix B
Comments on the Sample
examination paper
1. To find the equilibrium price, we solve 2400 2p = 4p, which gives p = 400 dollars.
The equilibrium quantity is q = 4(400) = 1600.
With an excise tax of 300 dollars, the demand equation stays the same but the supply
equation becomes q = 4(p 300). Make sure you understand this (see Chapter 2,
Section 6). It is a common mistake to change the demand equation rather than the
supply equation. We now solve
2400 2p = 4(p 300),
which gives a new equilibrium price of p = 600 dollars. The corresponding quantity is
either 2400 2(600) or 4(600 300). (It is not 4(600).) This gives q = 1200.
When there is an percentage tax of 100r% of the selling price, the supply equation is
q = 4p(1 r) and the demand equation is unchanged, i.e. q = 2400 2p. Now, we could
solve 2400 2p = 4p(1 r) to obtain an expression for r in terms of p, and then set
p = 600 to find r. But we can find r more quickly by noting that it must be the case that
300 = rp = 600r, so r = 0.5, representing a 50% tax. It is not right to argue that since
the price rose from 400 to 600, and because this is a 50% increase, it must be the case
that the percentage tax is 50%. This is merely coincidental and not a valid argument.
2. Because ex has series
1+x+
x2 x3 x4
+
+
+ ,
2
6
24
it follows (putting x2 in place of x) that

2
ex ' 1 + x2 +
x4 x 6 x8
+
+ ,
2
6
24
and
2
ex 1 ' x2 +
x4 x6 x8
+
+ .
2
6
24
Also, we know that

sin y ' y
y3
.
6
So, taking care to ensure we have enough terms in each approximation so that our
143
B. Comments on the Sample examination paper
answer is valid up to powers x8 , we have

x2
sin e

3

x4 x6 x8
1
x4
2
2
1 ' x +
+
+
x +
2
6
24
6
2

3 8
x4 x6 x8 1
6
2
+
+
x + x
'x +
2
6
24 6
2
= x2 +
x4
5
x8 .
2
24
3. As the function is homogeneous, the numerator must be homogeneous, and so

1 + = + 3.
Also, the degree of homogeneity is D and since the denominator has degree 3, we also
have
(1 + ) 3 = D.
Solving these equations, we see that = D + 2 and = D.
This means that we have
f (x, y) =
xy D+2 + xD y 3
,
x3 + y 3
and we need to verify Eulers equation, i.e.

x
f
f
+y
= Df (x, y).
x
y
In this case, the quotient rule gives us

(y D+2 + DxD1 y 3 )(x3 + y 3 ) (xy D+2 + xD y 3 )(3x2 )
f
=
,
x
(x3 + y 3 )2
and
f
([D + 2]xy D+1 + 3xD y 2 )(x3 + y 3 ) (xy D+2 + xD y 3 )(3y 2 )
=
.
y
(x3 + y 3 )2
These, in turn, give us

x
f
(xy D+2 + DxD y 3 )(x3 + y 3 ) (xy D+2 + xD y 3 )(3x3 )
=
x
(x3 + y 3 )2
xy D+2 + DxD y 3
3x3
D+2
D 3
=
(xy
+
x
y
)
x3 + y 3
(x3 + y 3 )2
and
y
f
([D + 2]xy D+2 + 3xD y 3 )(x3 + y 3 ) (xy D+2 + xD y 3 )(3y 3 )
=
y
(x3 + y 3 )2
=
144
3y 3
[D + 2]xy D+2 + 3xD y 3
D+2
D 3
(xy
+
x
y
)
x3 + y 3
(x3 + y 3 )2
Adding these together, we get

x
3
3
f
(xy D+2 + DxD y 3 ) + ([D + 2]xy D+2 + 3xD y 3 )
f
D+2
D 3 3x + 3y
+y
=
(xy
+
x
y
)
x
y
x3 + y 3
(x3 + y 3 )2
[D + 3]xy D+2 + [D + 3]xD y 3

x3 + y 3
D+2
D 3
3(xy
+
x
y
)
x3 + y 3
(x3 + y 3 )2
= [D + 3]
=D
xy D+2 + xD y 3
xy D+2 + xD y 3
3
x3 + y 3
x3 + y 3
xy D+2 + xD y 3
x3 + y 3
= Df (x, y),
as required.
4. The eigenvalues of A are found by solving |A I| = 0, which is
2 12 + 32 = 0
( 4)( 8) = 0,
so the eigenvalues are 4 and 8. For an eigenvector corresponding to 4, we solve

(A 4I)x = 0, which is

3 1
x
0
=
,
3 1
y
0
and this system is equivalent to the single equation 3x + y = 0, a solution to which is
x = 1, y = 3. So, an eigenvector corresponding to 4 is

1
,
3
or any non-zero scalar multiple of this recall that the zero vector is never an
eigenvector. Similarly, for an eigenvector corresponding to 8, we solve (A 8I)x = 0,
which is

1 1
x
0
=
,
3 3
y
0
and this system is equivalent to the single equation x + y = 0, a solution to which is
x = 1, y = 1. So, an eigenvector corresponding to 4 is

1
,
1
or, again, any non-zero scalar multiple of this. It follows from the standard theory that
if we let

1 1
4 0
P =
and
D=
,
3 1
0 8
then P 1 AP = D. Of course, this is only one of the many possible pairs of matrices
that we could choose for P and D: others are possible depending on what eigenvectors
145
we choose to form the columns of P and the order in which we choose to place them.
For instance, another answer is that P 1 AP = D when

1 1
8 0
P =
and
D=
.
1 3
0 4
For the next part, we use the fact that, since P 1 AP = D, we have A = P DP 1 . So,
An = AA
A} = (P DP 1 )(P DP 1 ) (P DP 1 ),
| {z
|
{z
}
n times
n times
and then we use the fact that P 1 P = I, to get

An = P |DIDI{z ID} P 1 = P Dn P 1 .
D occurs n times
which is a general expression for An . Now, using the standard formula for the inverse of
a 2 2 matrix, we have

1 1 1
1
,
P =
4 3 1
so we get

n

n

1 1 1
1 1
4 0 1 1 1
4
0
1 1
A =
=
.
1 3
0 8 4 3 1
0 8n
3 1
4 1 3
n
On calculating this matrix product (details omitted) we obtain

1
4n + 3(8n )
4n + 8n
n
.
A =
4 3(4n ) + 3(8n ) 3(4n ) + 8n
5. This is a second-order equation. In standard form, we have
a
xt+1 axt + xt1 = 0,
4
so the auxiliary equation is

2
a
a
z az + = 0
=
z
= 0,
4
2
so there is just one solution, a/2. Therefore, for some constants A and B, the general
solution is
t
a
.
xt = (At + B)
2
The facts that x0 = 1 and x1 = a show that A = 3 and B = 1, so the particular

solution is
t
a
xt = (3t 1)
,
2
2
a formula for xt in terms of t and a.

6. For the equilibrium quantity, noting that the equilibrium price is 30, we solve
3 + 3q + 2q 2 = 30,
146
which simplifies to
2q 2 + 3q 27 = 0
(2q + 9)(q 3) = 0,
and hence q = 3 is the only economically meaningful solution. (Or, we could have used
the formula for the solution to a quadratic equation.) To find the producer surplus, we
evaluate
Z
3
(30)(3)
(3 + 3q + 2q 2 ) dq.
After integrating and simplifying, this becomes 99/2.

Given that the elasticity of demand for the good is equal to 1 for every value of p, we
can find the demand function by solving
p dq
= 1.
q dp
This is a separable differential equation so we separate and integrate to see that

Z
Z
dp
dq
=
=
ln q = ln p + c,
q
p
and hence q = A/p with A = ec . Using the fact that, at equilibrium, we have q(30) = 3,
we find A = 90 and so the demand function is
q(p) =
90
.
p
(A common error is to forget the +c when solving the differential equation.)

7. (a) The Lagrangean is
L = xy 2 z 3 (px + qy + rz M ),
and the first order conditions are
y 2 z 3 p = 0,
2xyz 3 q = 0,
3xy 2 z 2 r = 0,
px + qy + rz = M.
Eliminating , it follows that
y2z3
2xyz 3
3xy 2 z 2
=
=
.
p
q
r
This means that y = (2p/q)x and z = (3p/r)x. (As, clearly, the other solution, i.e.
x = y = z = 0, will not satisfy the constraint if M > 0.) Then, px + qy + rz = M shows
that x = M/(6p). It follows that y = M/(3q) and z = M/(2r) yielding a maximum
value of xy 2 z 3 subject to the constraint given by
M6
M6
=
.
(6p)(3q)2 (2r)3
432pq 2 r3
147
(b) Reducing the augmented matrix, we have
1 1 2 4
1 1
2
4
1 1
2
4
3 1 1 0 0 2
7 12 0 2
7 12 .
1 1
a b
0 2 a2 b4
0 0 a+5 b+8
From this we see that the system will:
be consistent with infinitely many solutions if and only if the last row only contains
zeroes, i.e. we must have a = 5 and b = 8.
have precisely one solution if and only if a + 5 6= 0, i.e. we must have a 6= 5.
have no solutions if and only if a + 5 = 0 and b + 8 6= 0, i.e. we must have a = 5
and b 6= 8.
8. To find a constant solution p(t) = p that makes Q(t) = S(t), we must have p0 (t) = 0
and so
a
a bp = p
=
p =
,
b
as 6= b. This is the required constant solution.

Now, given that
dp
= (I(t) I0 ) = I(t) + I0 ,
dt
where I0 is a constant, we differentiate both sides with respect to t to get
dI
dp
d2 p
= = (Q(t) S(t)) = ( a) ( b)p ( c) ,
2
dt
dt
dt
and hence
d2 p
dp
+ ( c) + ( b)p = a,
2
dt
dt
as required.
When c = 3, b = 2 and a = 1, we get the equation
d2 p
dp
+ 3 + 2p = 1.
2
dt
dt
Solving this in the usual way, we see that
1
p(t) = Aet + Be2t + ,
2
for some constants A and B. How do we determine A and B? We are told that p(0) = 1
and, also, because
dp
= (I(t) I0 ),
dt
where I0 = I(0), we have, when t = 0,
dp
= (I(0) I0 ) = 0,
dt
148
so p0 (0) = 0 too. Now, as

dp
= Aet 2Be2t ,
dt
this means that p(0) = 1 gives A + B + 1/2 = 1 and p0 (0) = 0 gives A 2B = 0. So,
solving these equations, we get A = 1 and B = 1/2. Therefore, our particular solution
is
1
1
p(t) = et e2t + .
2
2
The next part of the problem is more theoretical. The auxiliary equation of the
differential equation is
z 2 + ( c)z + ( b) = 0,
and a particular solution is p . As the solution depends on the roots of the auxiliary
equation, there are three cases to consider, depending on the value of the discriminant
= ( c)2 4( b).
If > 0, then the auxiliary equation has two solutions
( c)
r, s =
.
2
If > b and > c, then ( c) and we see that both r and s are negative.
As such, p(t) takes the form Aert + Best + p where r, s < 0 which means that
p(t) p as t .
If = 0, the roots of the auxiliary equation are equal, and p(t) takes the form
(A + Bt)ert + p where the single root, r, is negative if > c. As such, we again
have p(t) p as t .
If < 0, then the auxiliary equation has no real roots and p(t) takes the form
et (A cos(t) + B sin(t)) + p ,
where, if > c, we have
c
< 0.
2
As such, we again have p(t) p as t .
=
Consequently, if > b and > c, we see that p(t) p as t which is what we

were asked to show.
149
Notes
Untitled-3 8
23/12/2008 10:39:12
Notes
Untitled-3 9
23/12/2008 10:39:12
Notes
Untitled-3 8
23/12/2008 10:39:12
Comment form 2010.qxp
10/11/2010
10:58
Page 1
Comment form
We welcome any comments you may have on the materials which are sent to you as part of your
study pack. Such feedback from students helps us in our effort to improve the materials produced
for the International Programmes.
If you have any comments about this guide, either general or specific (including corrections,
non-availability of Essential readings, etc.), please take the time to complete and return this form.
Title of this subject guide: ..............................................................................................................................
................................................................................................................................................................................
Name ......................................................................................................................................................................
Address ..................................................................................................................................................................
................................................................................................................................................................................
Email ......................................................................................................................................................................
Student number ......................................................................................................................................................
For which qualification are you studying? ..............................................................................................................
Comments
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
................................................................................................................................................................................
Please continue on additional sheets if necessary.
Date: ......................................................................................................................................................................
Please send your comments on this form (or a photocopy of it) to:
Publishing Manager, International Programmes, University of London, Stewart House, 32 Russell Square, London WC1B
5DN, UK.

MT105b Vle

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

MT105b Vle

Загружено:

Авторское право:

Доступные форматы

Mathematics 2

University of London International Programmes

Mathematics 1 and Mathematics 2 . . . . . . . . . . . . . . . . . . . . .

Aims and objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

How to use the subject guide

Other recommended texts . . . . . . . . . . . . . . . . . . . . . .

Online study resources . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Making use of the Online Library . . . . . . . . . . . . . . . . . .

The use of calculators . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 Further differentiation and integration, with applications

Revision of Mathematics 1 differentiation . . . . . . . . . . . . . . . . . .

Using derivatives for approximations . . . . . . . . . . . . . . . . . . . .

The effects of taxation . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Revision of Mathematics 1 integration . . . . . . . . . . . . . . . . . . .

Definite integrals and areas

Consumer and producer surplus . . . . . . . . . . . . . . . . . . . . . . .

Sample examination/practice questions . . . . . . . . . . . . . . . . . . . . . .

Answers to Sample examination/practice questions . . . . . . . . . . . . . . .

Homogeneous functions and Eulers theorem . . . . . . . . . . . . . . . .

Applications of unconstrained optimisation . . . . . . . . . . . . .

Sample examination/practice questions . . . . . . . . . . . . . . . . . . . . . .

Answers to Sample examination/practice questions . . . . . . . . . . . . . . .

4 Linear algebra and applications

Inverse matrices and linear equations . . . . . . . . . . . . . . . . . . . .

Calculating determinants using row operations . . . . . . . . . . .

The square linear system when A is not invertible . . . . . . . . . . . . .

Calculating inverses using determinants . . . . . . . . . . . . . . . . . . .

4.10 Calculating inverses using row operations . . . . . . . . . . . . . . . . . .

4.11 Application: input-output analysis . . . . . . . . . . . . . . . . . . . . . .

4.12 Eigenvalues and eigenvectors . . . . . . . . . . . . . . . . . . . . . . . . .

4.14 Finding eigenvalues and eigenvectors . . . . . . . . . . . . . . . . . . . .

4.15 Diagonalisation of a square matrix

Sample examination/practice questions . . . . . . . . . . . . . . . . . . . . . .

Answers to Sample examination/practice questions . . . . . . . . . . . . . . .

Linear equations and integrating factors . . . . . . . . . . . . . . . . . .

Coupled differential equations . . . . . . . . . . . . . . . . . . . . . . . .

Reducing to a second-order equation . . . . . . . . . . . . . . . .

Applications of differential equations . . . . . . . . . . . . . . . . . . . .

Continuous cash flows . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5.10 Continuous price adjustment . . . . . . . . . . . . . . . . . . . . . . . . .

5.11 Determining demand from elasticity . . . . . . . . . . . . . . . . . . . . .

5.12 Market trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Sample examination/practice questions . . . . . . . . . . . . . . . . . . . . . .

Solutions to Sample examination/practice questions . . . . . . . . . . . . . . .

First-order difference equations . . . . . . . . . . . . . . . . . . . . . . .

Solving first-order difference equations . . . . . . . . . . . . . . . . . . .

Long-term behaviour of solution . . . . . . . . . . . . . . . . . . . . . . .

The cobweb model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Homogeneous second-order difference equations . . . . . . . . . . . . . .

Non-homogeneous second-order equations . . . . . . . . . . . . . . . . . .

6.10 Behaviour of the solution . . . . . . . . . . . . . . . . . . . . . . . . . . .

6.11 Coupled difference equations . . . . . . . . . . . . . . . . . . . . . . . . .

6.11.1 Solving by reducing to a second-order equation

6.11.2 Using diagonalisation . . . . . . . . . . . . . . . . . . . . . . . . .

6.12 Economic applications of second-order difference equations . . . . . . . .

Sample examination/practice questions . . . . . . . . . . . . . . . . . . . . . .

Answers to Sample examination/practice questions . . . . . . . . . . . . . . .

A Sample examination paper

B Comments on the Sample examination paper

Mathematics 1 and Mathematics 2

I make a number of points in the introductory chapter of the 05A Mathematics 1

Aims and objectives

This half unit is designed to: