Академический Документы
Профессиональный Документы
Культура Документы
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.
Contents
Contents
1 General introduction
1.1
1.2
Studying mathematics . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3
1.4
Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5
. . . . . . . . . . . . . . . . . . . . . . . .
1.6
Recommended books . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.1
Main text . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.2
1.7.1
The VLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.2
1.8
Examination advice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.9
1.7
Essential reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2
2.3
10
2.4
Taylors theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
2.5
Elasticities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14
2.6
16
2.7
18
2.8
. . . . . . . . . . . . . . . . . . . . . . . . .
19
2.9
20
Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
23
Answers to activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
Contents
33
Essential reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.2
Partial derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.3
34
3.4
Optimisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
37
3.4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
37
3.4.2
Unconstrained optimisation . . . . . . . . . . . . . . . . . . . . .
38
3.4.3
38
3.4.4
Constrained optimisation . . . . . . . . . . . . . . . . . . . . . . .
40
Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
42
Answers to activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
43
45
ii
25
51
Essential reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
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
53
4.5
Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
4.5.1
The determinant . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
4.5.2
55
4.6
Cramers rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
4.7
58
4.8
Row operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
4.9
62
64
66
Contents
67
4.13 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
67
. . . . . . . . . . . . . . . . . . . . .
69
Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
71
72
Answers to activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
77
5 Differential equations
83
Essential reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
5.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
5.2
Separable equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
84
5.3
88
5.4
Second-order equations . . . . . . . . . . . . . . . . . . . . . . . . . . . .
90
5.5
Behaviour of solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . .
92
5.6
93
5.6.1
94
5.6.2
Using diagonalisation . . . . . . . . . . . . . . . . . . . . . . . . .
96
5.7
99
5.8
Macroeconomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
99
5.9
100
101
102
103
Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
103
104
Answers to activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
105
109
6 Difference equations
117
Essential reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
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
118
6.4
120
6.5
121
6.6
122
6.7
Financial applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
123
6.8
124
6.9
126
127
127
. . . . . . . . . .
127
129
130
Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
132
132
Answers to activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
134
135
139
143
iv
Chapter 1
General introduction
1.1
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
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
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
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
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. General introduction
1.6.1
Main text
1.6.2
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
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
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
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.
1. General introduction
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
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
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
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.
1. General introduction
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
R
R
R
R
R
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
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
f 0 (x)
kxk1
ex
1/x
cos x
sin x
2.3
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
10
df
x.
dx
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 ,
24000
.
p4
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)
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
x2 x3 x4
+
+ ...,
2
3
4
sin x ' x
x3 x5 x7
+
+ ...,
3!
5!
7!
cos x ' 1
x2 x4 x6
+
+ ...,
2!
4!
6!
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!
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
14
2.5. Elasticities
p q
,
q p
p dq
,
q dp
p dq
p
p
= (1) = .
q dp
q
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
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
Eliminating q T we get
40 5pT =
15 T
(p T ) 10.
2
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).)
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
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,
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
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
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
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
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
p
A
Demand
q
CS
p
R
Demand
q
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
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
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
50
,
q+5
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
12000
.
p3
24
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
25
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
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
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.)
Answer to question 2.3
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!
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.
Answer to question 2.4
We have
q = 100/(1 + p2 ) = 100(1 + p2 )1 ,
so
dq
= 100(2p)(1 + p2 )2 = 200p(1 + p2 )2 ,
dp
27
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
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)
28
2
1
1
1
1
1
dt
=
=
.
t2
t 1
2
1
2
It follows that
A = A 1 + A2 =
47
.
64
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.
Answer to question 2.8
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
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
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
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
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
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
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.
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
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),
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
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
= 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 =
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
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
1
1
1
x + y + z = c,
2
3
6
40
1
1
1
x+ y+ zc ,
2
3
6
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
Learning outcomes
At the end of this chapter and the relevant reading, you should be able to:
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
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
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
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
Answers to activities
Feedback to activity 3.1
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
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.
Feedback to activity 3.2
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.
Feedback to activity 3.3
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
20 8Q1
10Q2 40
=
,
4
5
so
100 40Q1 = 40Q2 160,
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
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.
Answer to question 3.3
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
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
= 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
Q2 = 16 0.2P,
so
Q = Q1 + Q2 = 25 0.25P.
47
(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
L
= 1 (1/6)x (1/3)y = 0.
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.
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
y = 32 and z = 50.
49
1 1 1
+ + (x + y + z c) .
x y z
1
L
= 2 =0
x
y
L
1
= 2 =0
x
z
L
= x y z + c = 0.
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
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
Recommended reading
R
R
R
R
4.1
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
=
=
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
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,
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
(1)
(2)
3 2
x
2
=
,
5 1
y
2
53
i.e.,
Ax = b,
where
x
x=
y
2
and b =
.
2
1
=
7
1 2
,
5 3
4.5
4.5.1
Determinants
The determinant
a b
det d e
g h
54
4.5. Determinants
Activity 4.3
4.5.2
1 1 0
Calculate 2 0 1.
3 1 1
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
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
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
57
1
2
3
2 0
x2 = 2 ,
5 4
x3
1
= 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
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
58
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
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
1
x2 =
0 .
1
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
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.
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
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
2
1
1
1 2 1 1
1
1
0 0 1 1 0 .
2 2 1
0 0 0 1
Activity 4.6
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
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
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
1
A12
A1 =
det(A)
+A13
Note carefully the alternating pattern
The matrix
+A11
A21
+A31
invertible and
A21 +A31
+A22 A32 .
A23 +A33
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
Let
2 1 3
A = 0 1 1 .
1
2 0
We shall calculate its inverse using the
determinant:
1
det(A) = 2
2
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
2 1 1
2 6 4
adj(A)T = 6 3 5 = 1 3 2 ,
4
2 2
1
5 2
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
Activity 4.8
1 2
A= 1 3
2 5
4.10
1
1 .
1
See Anthony and Biggs (1996) Section 18.3 for an explanation of why this technique works.
64
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
1 3
1 1 ,
2 0
2 1 3
0 1 1
(A | I) =
1
2 0
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
2 6
1
1
1 3
A =B=
8
1
5
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
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
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
d1
d = d2
d3
and
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
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
4.14
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
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
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,
4.15
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
...
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
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
70
Example 4.13
The 2 2 matrix
A=
4 1
,
1 2
Learning outcomes
At the end of this chapter and the relevant reading, you should be able to:
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
I = I0 ar
and Y = C + I + G,
and Ms = Md ,
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.
Answers to activities
Feedback to activity 4.1
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
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.
Feedback to activity 4.4
The augmented matrix is
1 2 3 5
2 1 1 1 .
4 3 5 11
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
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).
Feedback to activity 4.5
This is easily done by verifying that AA1 = I.
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.)
Feedback to activity 4.7
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
1
=
3
2 1
,
1 1
76
1
2
3
6
1 1
1 2
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
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)
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
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
x1
3s
x = x2 = 1 .
x3
s
5 5 5
3 4 2
2 1 3
250
1 1 1 50
1 1
150 3 4 2 150 0 1
100
2 1 3 100
0 1
1 50
1 1 1 50
1 0 0 1 1 0 .
1 0
0 0 0 0
78
25000 (in which case the same number of Break bars are produced and no Abracadabra
bars will be manufactured).
Answer to question 4.4
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)
79
1
(A | I) = 2
1
and we
1
2
1
1 5 1 0 0
3 1 0 1 0 ,
0 14 0 0 1
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.
Answer to question 4.6
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.)
Answer to question 4.7
Denoting the matrix by A, we have
2
1
= (2 )(1 ) 3 = 2 3 + 2 3 = 2 3 1.
|A I| =
3
1
80
3 13
,
=
2
81
82
Chapter 5
Differential equations
Essential reading
(For full publication details, see Chapter 1.)
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.
Dowling (2000) Chapters 18 and 20.
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
83
5. Differential equations
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
84
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
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
5. Differential equations
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
(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
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. Differential equations
5.3
For example, ex
2 /2
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
which is
d
((x)y) = (x)Q(x).
dx
(x)Q(x) dx.
Example 5.2
d
dy
+ (x)P (x)y =
((x)y) .
dx
dx
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
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. Differential equations
5.4
Second-order equations
b
4ac b2
=
and =
.
2a
2a
Then, in this case, the general solution of (?) is
y(x) = ex (A cos x + B sin x) ,
where A and B are arbitrary constants.
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
90
Consider
d2 y
dy
2y = 4x2 .
2
dx
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 .
91
5. Differential equations
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
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
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
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
5. Differential equations
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
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
=
Example 5.6
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
5. Differential equations
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,
1
z(t) =
2
Activity 5.6
dy
+ 4y
dt
=
9
+ t e2t + 4t2 6t + 8.
2
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
97
5. Differential equations
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
and
D=
6 0
,
0 1
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
There are many applications of differential equations in economics and finance, and this
section briefly discusses several of them.
5.8
Macroeconomics
99
5. Differential equations
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
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
100
(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
1
and
r2
A(t) = A(0)ert +
Activity 5.8
5.10
1
+ c,
r2
ert (1 + rt)
.
r2
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
5. Differential equations
Example 5.8
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
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
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
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
At the end of this chapter and the relevant reading, you should be able to:
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
5. Differential equations
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
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
Answers to activities
Feedback to activity 5.1
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
5. Differential equations
Feedback to activity
5.2
We have y(t) = 2/( 1 + x4 + k) and y(0) = 2, so
2
= 2,
1+0+k
2
.
1 + x4
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
and
2A B = 6.
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
so we need
4a = 8,
8a + 4b = 0 and 2a + 4b + 4c = 8,
rt
Z
=
rt
Z
f (t) dt =
1
ert dt = ert + c.
r
1
dp =
(28 8p)3
Z
1 dt,
1
1
= t + c.
16 (28 8p)2
107
5. Differential equations
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
1
.
256t + 1
Observe that, as t , the square root term tends to zero, and so p(t) approaches the
equilibrium price 7/2.
Feedback to activity 5.10
The fact that the elasticity is constant at 3 means that
p dq
= 3.
q dp
dq
= 3
q
dp
,
p
so
ln q = 3 ln p + c.
Taking exponentials,
q = e3 ln p + c = p3 ec =
A
,
p3
108
60
.
p3
= 1 + t 2 + c2 ,
1
+ c1 = 1 + t2 + c2 ,
y
so
y(t) =
1
,
1 + t2 + c
1
x
dx = ln(1 + x2 ) + c,
2
1+x
2
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
5. Differential equations
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
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
p2
8
.
+1
2
8
p 4 = 24 p .
3
3
111
5. Differential equations
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
3
.
16t + 3/64
dP
d2 P
+ 2 = 8P 6,
dt
dt
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).
Answer to question 5.9
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)
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
5. Differential equations
and
1
2
y2 (t) = y1 (0) + y2 (0) +
3
3
2
2
y1 (0) + y2 (0) e3t .
3
3
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
The auxiliary equation is
z 2 6z + 8 = 0,
df
= 3Ae2t + 3Be4t (2Ae2t + 4Be4t ) = Ae2t Be4t .
dt
114
It follows that if
1 1
P =
,
1 1
then
P
2 0
AP = D =
.
0 4
1 1 1
h
f
1 f
=P
.
=
k
1
1
g
g
2
dh
= 2h and
dt
dk
= 4k,
dt
and hence
h(t) = h(0)e2t
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
5. Differential equations
116
Chapter 6
Difference equations
Essential reading
R
R
Further reading
6.1
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
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
t(2a + (t 1)d)
t(y0 + yt1 )
=
.
2
2
6.3
1 xt
.
1x
118
119
6. Difference equations
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
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
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
1
b
=
= 1,
1a
12
so the solution is
yt = y + (y0 y )at = 1 + (1 (1))2t = 1 + 2t+1 ,
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
The behaviour of the general solution (or time path) for yt given by
yt = y + (y0 y )at ,
121
6. Difference equations
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
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.
pt = 24 qt .
6.7
Financial applications
123
6. Difference equations
I
I
= .
1 (1 + r)
r
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
124
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
(z 5)(z 1) = 0,
Let us find yt if
yt 2yt1 + 4yt2 = 0,
where y0 = 1 and y1 = 1
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
6. Difference equations
1
3
+F
,
1 3 = 2 (E cos(/3) + F sin(/3)) = 2
2
2
so we have 1
6.9
3=1+
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,
126
6.10
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
6.11.1
127
6. Difference equations
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 ,
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 .
6.11.2
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
6. Difference equations
Finally, then,
yt = 6t + 4(1)t
as before.
6.12
and zt = 6t 3(1)t ,
130
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
(1/2)
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
2 2
t
E cos
t + F sin
t .
4
4
2 2
0
(E cos(0) + F sin(0)) = 64 + E,
131
6. Difference equations
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
At the end of this chapter and the relevant reading, you should be able to:
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.
132
and
q D (p) = 20 0.15p.
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
6. Difference equations
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 .
Answers to activities
Feedback to activity 6.1
We do indeed have y1 = 22 1 = 3, y2 = 23 1 = 7 and y3 = 24 1 = 15.
Feedback to activity 6.2
For yt = (2/3)yt1 + 5, we have a = 2/3 and b = 5. Then,
y =
5
b
=
= 15,
1a
1 (2/3)
t
t
2
2
yt = y + (y0 y )a = 15 + (2 15)
= 15 13
.
3
3
yt = y + (y0 y )(1)t ,
yt 5yt1 14yt2 = 0,
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
b
N
N
=
=
= 20N,
1a
1 0.95
0.05
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
160
= 120,
1 (1/3)
135
6. Difference equations
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.
Answer to question 6.4
The auxiliary equation is
z 2 z 6 = (z 3)(z + 2) = 0,
so for some constants A and B
yt = A(3t ) + B(2)t .
Answer to question 6.5
(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
p
3/2 and
p
p
cos = (2)/(2 3/2) = 2/3.
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
6. Difference equations
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.
Answer to question 6.6
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
1 1 1
wt
yt
1 yt
=P
=
.
xt
zt
zt
2 1 1
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.)
1.
and
q = 2400 2p,
139
2.
xy + x y 3
,
x3 + y 3
A sequence xt satisfies
xt+1 =
a
axt xt1 ,
4
a,
SECTION B
Answer both questions from this section (20 marks each)
7. (a)
140
A
(b)
8.
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
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
ex ' 1 + x2 +
x4 x 6 x8
+
+ ,
2
6
24
and
2
ex 1 ' x2 +
x4 x6 x8
+
+ .
2
6
24
y3
.
6
So, taking care to ensure we have enough terms in each approximation so that our
143
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
xy D+2 + xD y 3
,
x3 + y 3
f
f
+y
= Df (x, y).
x
y
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
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
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
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,
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
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
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
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.
p dq
= 1.
q dp
90
.
p
147
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
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
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 .
=
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
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.