Introductory Mathematics For Economics and Business by K Holden

Introductory Mathematics for Economics and Business
Introductory
Mathematics for
Economics and Business
Second Edition
K. Holden and A. W. Pearson
M
© K. Holden and A. W. Pearson 1975, 1983, 1992
Softcover reprint of the hardcover 2nd edition 1992 978-0-333-57649-6
All rights reserved. No reproduction, copy or transmission of

this publication may be made without written permission.
No paragraph of this publication may be reproduced, copied or

transmitted save with written permission or in accordance with
the provisions of the Copyright, Designs and Patents Act 1988,
or under the terms of any licence permitting limited copying
issued by the Copyright Licensing Agency, 90 Tottenham Court
Road, London W1P 9HE.
Any person who does any unauthorised act in relation to this

publication may be liable to criminal prosecution and civil
claims for damages.
First published 1975 by

David & Charles (Holdings) Ltd
Published as Introductory Mathematics for Economists

1983 by The Macmillan Press Ltd
Reprinted 1986, 1988, 1990
This edition published 1992 by

THE MACMILLAN PRESS LTD
Houndmills, Basingstoke, Hampshire RG21 2XS
and London
Companies and representatives
throughout the world
ISBN 978-0-333-57650-2 ISBN 978-1-349-22357-2 (eBook)

DOl 10.1007/978-1-349-22357-2
A catalogue record for this book is available

from the British Library
Contents
Preface xi
1 Linear Equations 1
1.1 Some preliminaries 1
1.2 Graphical representation 4
1.3 Intercept and slope 6
1.4 Interpolation and extrapolation 7
1.5 Exercises 8
1.6 Simultaneous linear equations 9
1.7 Demand and supply 13
1.8 Exercises 16
1.9 The effects of taxation on demand 17
1.10 Simple national income models 21
1.11 IS-LM analysis 24
1.12 Exercises 26
1.13 Determinant notation 28
1.14 Exercises 35
1.15 Applications of determinants 36
1.16 Exercises 39
1.17 Revision exercises for Chapter 1 40
2 Elementary Matrix Algebra 43

2.1 The matrix notation 43
2.2 Elementary matrix operations 44
2.3 Exercises 49
2.4 The inverse matrix 50
2.5 Exercises 57
2.6 Inversion by Gaussian elimination 57
2.7 Exercises 63
2.8 Linear dependence and rank 64
2.9 Exercises 68
2.10 Further properties of matrices 69
2.11 Exercises 75
v
vi Contents
2.12 Input-output analysis 76

2.13 The input-output matrix 81
2.14 Exercises 88
3 Non-linear Equations 92
3.1 The quadratic 92
3.2 Exercises 97
3.3 The roots of a quadratic equation 98
3.4 Exercises 105
3.5 Other non-linear functions 105
3.6 Exercises 113
3.7 Breakeven point 115
3.8 Simultaneous quadratic equations 119
3.9 Net revenue 120
3.10 Exercises 124
3.11 Discontinuous functions 125
3.12 Exercises 127
4 Series 130
4.1 Introduction 130
4.2 Arithmetic progressions 130
4.3 Exercises 133
4.4 Geometric progressions 134
4.5 Exercises 136
4.6 Discounting 137
4.7 Exercises 140
4.8 Annuities and sinking funds 140
4.9 Exercises 145
4.10 Interest paid continuously 146
4.11 The binomial theorem 148
4.12 The exponential series 151
4.13 Exercises 155
4.14 Logarithms 156
4.15 Exercises 161
5 Differential Calculus 165

Contents vii
5.2 Some general rules 169

5.3 Exercises 176
5.4 Elasticity of demand 177
5.5 Marginal analysis 180
5.6 Exercises 183
5.7 Maxima and minima 184
5.8 Cost and revenue analysis 194
5.9 Exercises 197
5.10 Profit maximisation in several markets 198
5.11 Maximising tax revenue 200
5.12 Inventory models 202
5.13 Exercises 206
5.14 Differentials 207
5.15 Taylor's and MacLaurin's theorems 209
5.16 Euler relations 212
5.17 Newton's method 213
5.18 Plotting graphs 217
5.19 Exercises 220
6 Integral Calculus 224

6.2 Exercises 225
6.3 Techniques of integration 226
6.4 Exercises 228
6.5 The calculation of areas 229
6.6 Exercises 233
6.7 Absolute area 234
6.8 Exercises 235
6.9 Consumer's surplus and producer's surplus 236
6.10 Areas between curves 238
6.11 Exercises 242
6.12 Numerical methods of integration 242
6.13 Exercises 247
6.14 More difficult integration 248
6.15 Exercises 254
7 Partial Differentiation 257

7.1 Functions of more than one variable 257
viii Contents
7.2 Marginal analysis 263

7.3 Elasticity of demand 264
7.4 Exercises 266
7.5 Differentials 267
7.6 Total derivatives 273
7.7 Exercises 274
7.8 Implicit differentiation 275
7.9 Exercises 277
7.10 Homogeneous functions 278
7.11 Euler's theorem 280
7.12 Exercises 282
7.13 Maxima and minima 282
7.14 Exercises 285
7.15 Maxima and minima for functions of more than two
variables 286
7.16 Constrained maxima and minima 289
7.17 A linear expenditure system 292
7.18 Exercises 295
8 Linear Programming 297

8.2 A product-mix problem 297
8.3 A graphical approach 299
8.4 The simplex method 301
8.5 A summary of the simplex method 308
8.6 Incremental values 310
8.7 The general problem 312
8.8 Exercises 315
9 Differential Equations 319

9.2 First-order linear differential equations 321
9.3 Applications of first-order differential equations 332
9.4 Exercises 336
9.5 Second-order homogeneous differential equations 338
9.6 Exercises 345
9.7 Second-order non-homogeneous differential
equations 345
9.8 Exercises 349
Contents ix
9.9 Graphical presentation 350

9.10 Exercises 354
9.11 An inflation-unemployment model 354
9.12 Exercises 360
10 Difference Equations 364

10.2 Compound interest and the addition of capital at
yearly intervals 365
10.3 First-order linear difference equations 367
10.4 The cobweb model 368
10.5 The Harrod-Domar growth model 374
10.6 A consumption model 375
10.7 Exercises 376
10.8 Samuelson's multiplier-accelerator model 377
10.9 Second-order linear difference equations 378
10.10 A consumption-investment model 384
10.11 An inflation-unemployment model 387
10.12 Exercises 391
Appendix A Trigonometric functions 395

A.l Definitions 395
A.2 Compound angles 396
A.3 Degrees and radians 398
A.4 General angles 399
A.5 Differentiation 403
A.6 Integration 404
A.7 Inverse functions 406
Appendix B Set Theory 408

B.l Introduction 408
B.2 Combining sets 409
B.3 Venn diagrams 410
B.4 Relations and functions 414
Appendix C Answers to Exercises 417

Further Reading 506
Index 509
Preface
This book introduces to students with a limited mathematics back-
ground the essential mathematics needed for economics and business
courses. It is not intended to replace more formal mathematics texts
and, by design, does not include proofs and derivations of all the
formulae used; these are included only where they aid understanding.
We hope that this approach will be suitable for those students who, as a
result of their earlier experiences in this subject area, do not regard
themselves as having any mathematical ability. While reading the book
they are advised to check the algebra by jotting the working down on
paper and to use a calculator to check the numbers. Many examples are
included in the text and the exercises (with worked answers given in
Appendix C) are intended to help such students. At the end of each
chapter there are also some revision exercises (without answers).
The material is selected so as to increase in difficulty as the book
progresses. Chapter 1 revises many basic concepts in algebra and intro-
duces linear equations, with immediate applications in simple economic
models of markets and the national economy. Chapters 2 and 3 are the
natural generalisations of elementary matrix algebra and non-linear
equations. Chapter 4, on series, covers many applications in finance as
well as providing the groundwork for calculus in Chapters 5-7. Chap-
ter 5, on differential calculus, includes profit maximisation for a firm,
simple inventory models and other applications of marginal concepts.
Integration, the subject of Chapter 6, covering both standard analytical
techniques and numerical methods, is also applied to simple marginal
concepts, including consumer's and producer's surplus. Relationships
with many variables are examined in Chapter 7, on partial differentia-
tion, which ends with maximisation subject to a constraint. A variation
on this problem is treated in Chapter 8, on linear programming. Chap-
ters 9-10 consider dynamic relationships - in continuous terms in Chap-
ter 9 and in discrete terms in Chapter 10. There is an extensive treat-
ment of trigonometric functions in Appendix A, and an introduction to
set theory in Appendix B. Each of these can be treated as optional, with
little loss of the main text, except for parts of Chapters 9 and 10 which
xi
xii Preface
need trigonometric functions. Detailed numerical answers to the exer-

cises are provided in Appendix C.
The order in which the material is covered can be varied. For
example, integral calculus may be followed by differential equations, or
matrix algebra may be deferred to precede linear programming.
We have intentionally excluded any coverage of statistics, since we
believe that any brief treatment would be inadequate for most courses.
We have also ignored computer programming because the wide variety
of languages and packages now available mean that any general discus-
sion is unhelpful to students and teachers. The omission of these topics
should not be taken as evidence that we believe they are unimportant.
We are grateful to Roger Latham (Queen's University, Ontario,
Canada) and David Peel (University College, Aberystwyth), formerly
of the University of Liverpool, for reading through an early draft of the
manuscript. Our thanks are particularly due to John Thompson (Liver-
pool Polytechnic) who made detailed comments on the whole manu-
script which helped us to remove many errors. Peter Stoney, Tim
Worrall and Ken Cleaver, of the University of Liverpool, kindly read
particular sections and made helpful comments which improved the
manuscript. Since none of these saw the final manuscript then, as
always, the responsibility for any errors lies with the authors. Thanks
are also due to Simon Blackman, for help with the graphs and Jenny
Holden, for valuable assistance with preparation of the manuscript and
the index.
This book is a much revised and expanded version of Introductory
Mathematics for Economists (Macmillan, 1983) which was originally
published by David and Charles (Holdings) Ltd in 1973.
Finally, a word of warning. This book is introductory and inten-
tionally uses simple models. These are not meant to represent the real
world and many of the numbers chosen- for example, in supply and
demand analysis - and the assumptions imposed are arbitrary and for
illustrative purposes only. The world is complicated and it is only by
simplifying it that we can hope to gain an understanding of it. But once
we do master simple models, more complex and realistic models can be
considered.
K.H.
A.W.P
Chapter 1
Linear Equations
1.1 Some preliminaries
Since the purpose of this book is to help those students who have a
weak background in mathematics to understand applications in econ-
omics and business studies, it is useful to start by reviewing some basic
concepts. One fundamental idea is that of a number. The type of
number that is familiar in everyday life is known as a real number. That
is, it can be used to represent something 'real', such as 2 and 3 rep-
resenting the numbers of people in two cars. In this case the numbers
are positive whole numbers or positive integers, and are part of the
familiar sequence 1, 2, 3, 4, ... where the three dots indicate 'and so
on'. The positive integers increase without limit and are said to be
unbounded. They are used in counting. If the sequence is continued in
the negative direction the values are 0, -1, -2, -3, ... Below we will
see that the value zero is rather special in that it does not satisfy some of
the standard rules of algebra, and can be interpreted as being positive or
negative or neither. The numbers -1, -2, -3, ... are the negative
integers and while their interpretation is less obvious than for positive
integers, they are familiar for measuring sub-zero temperatures and in
economics they can represent debts (negative assets) or losses (negative
profits).
While the integers are a useful starting point in discussing real
numbers, another way of looking at real numbers is to classify them as
rational and irrational numbers. Rational numbers are defined as being
any numbers which can be expressed as the ratio of two integers, where
it is assumed that the second integer is not zero. Since any integer n,
say, can be expressed as nil then all the integers are rational numbers.
Also, fractions, such as 2/5 and- 516 are rational numbers. Some frac-
tions can be represented precisely as decimals, such as 112 = 0.5000,
while others are non-terminating, repeating decimals, such as
1/3 = 0.333333 . . . and 2/11 = 0.181818 . . . Irrational numbers are
1
K. Holden et al., Introductory Mathematics for Economics and Business
© K. Holden and A. W. Pearson 1992
2 Introductory Mathematics for Economics and Business
numbers which cannot be expressed as the ratios of two integers and are
non-terminating, non-repeating decimals such as V2 = 1.414213 ...
and the value n = 3.141592 ...
To sum up, real numbers can be sub-divided into rational and
irrational numbers. Rational numbers can also be sub-divided into
fractions and integers. In elementary applications of mathematics in
economics and business studies only real numbers are needed. How-
ever, in section 3.3 below a new type of number, known as an imaginary
number, is introduced and, when combined with a real number, gives a
complex number.
The values of the real numbers are fixed or constant. This contrasts
with a variable which is anything that can take on different values. An
example of a variable is the price of apples since it can change as time
passes or take different values at any particular time in different places.
Other examples of variables in economics are profits, quantities de-
manded, the level of inventories, consumption, income, imports and
exports. Variables can be discrete, that is they can only take on a lim-
ited number of values between any given two values, or continuous.
Examples of discrete variables are the number of students in a class and
the amount of cash (measured in pounds or dollars) that is in one's
hand. Notice that, as in this latter example, discrete variables need not
be integers. Continuous variables can, in principle, take on any value
between any two given values. For example, if the weights of two bags
of apples are 2.10kg and 2.11kg, it is possible to find another bag
of apples with a weight anywhere between these two, such as
2.10419345kg.
In applying mathematics in economic and business studies it is
frequently convenient to represent variables by symbols, so that price
might be represented by p, quantity sold by q, revenue by rand total
cost by TC. This makes it easier to see how the different variables are
related.
For example, it is frequently the case that the total cost (TC) of
producing an output, q, of goods is related to the level of q. That is, TC
depends on q or
TC = f(q)
which is read 'TC is a function of q'. The symbol f, together with the
brackets, is a shorthand way of saying that the two variables TC and q
are related in some way.
The statement that one variable depends on another - in this
example that TC is related to q - may not appear to be very useful, but
Linear Equations 3
in fact it can be the starting point in many applications of economic

theory and management science to practical problems. The equation is a
statement of the behavioural relationship linking the two variables.
Another form of equation is an identity, which is a definition or state-
ment that two variables or expressions are always equivalent, such as
total revenue, TR, is price, p, times the quantity sold, q. The third form
of equation which is common in economics is the equilibrium condition
in which two or more variables are set equal to one another, as in supply
and demand analysis where, in equilibrium, the quantity demanded is
set equal to the quantity supplied.
Behavioural equations can arise either from introspection (as do
many ideas in economics and management science) or from observing
what happens in the real world. In the relationship between total cost
and the level of output, TC is called the dependent variable and q is the
single independent variable, and TC depends on, or is determined by, q.
In stating this a concept of causality is being introduced: a change in q
causes a change in TC. A further distinction which is important in
economic modelling is that between endogenous variables, whose value
is determined or explained by the model being considered, and exogen-
ous variables, whose value is determined outside the model. Whether a
variable is endogenous or exogenous is determined by the circumstances
of the model. For example, a particular fruit-grower may believe that
the price of apples is determined by the world market and so is exogen-
ous for each individual producer, while for a fruit-growers' federation,
their total production may affect the world price, making the price
endogenous.
The precise form of the relationship between total cost and output
has not been specified but it might be assumed that for each value of q
there is only one value of TC, so that the relationship is said to be single
valued. Also, it might be expected that as the level of output, q,
increases, TC increases.
Continuing with the example, to make progress some assumptions
have to be made about the form of the relationship (and they would
need to be checked in any particular practical situation). Here we might
assume:
(a) Some parts of the total cost of production are necessarily in-
curred if there is to be any production and they do not increase as
production is increased. These are known as fixed costs and include such
items as rent, rates, and wages of the labour force which within certain
limits are not affected by changes in the level of output.
(b) There are other parts of the total cost of production which
increase as the level of output increases. These are known as variable

costs and include such items as raw materials, power, and parts of the
labour force which can be employed on the process if and when required.
It is clear that in many practical situations the split between fixed
and variable costs cannot be made very clearly and that in the long run
all costs tend to be variable. However, a simple breakdown into these
two categories can prove to be useful in establishing such points as the
breakeven level of production, as we shall see later. But first, let us
consider an example in which the available information about the pro-
duction process indicates that the fixed costs amount to £100 and that
the variable costs associated with manufacturing 100 units of the product
amount to £300. We will assume that the variable costs are directly
proportional to the number of units of output and hence that the total
cost varies linearly with the level of output.
The relationship between costs and output can then be written
Total cost= fixed cost+ (variable cost per unit) x (level of output)
or TC =a+ b x q
where a and b are two constants which represent fixed cost and variable
cost per unit respectively.
1.2 Graphical representation

For a two-variable relationship a common method of presentation
of information is by means of a graph. This is a two-dimensional dia-
gram with two axes, one of which represents TC (in our example the
total cost of production) and the other q (the level of output). These
axes are generally drawn at right angles to each other and their point of
intersection, the origin, 0, is where both total costs and the level of
output are zero. By choosing suitable scales to represent different values
of the variables we can construct a graph from any given set of data.
In our example in Fig. 1.1, when the level of output is 0, the fixed
costs and hence the total costs are equal to £100, i.e. when q = 0,
TC = 100. Also, when the level of output is 100, the total costs are
£400, i.e. when q = 100, TC = 400.
Since we are assuming that total cost increases uniformly and
continuously with output we can join the two points by a straight line.
Hence TC = a + bq is an equation of a straight line, and in this case
a = fixed costs = £100 and
b = variable cost per unit of output = £3
Linear Equations 5
400
300
12
~u 200
iii
;§
100
~ ~ ~ ~~~ ~ ~ ~ ~ ~
Units of output, q
Fig. 1.1
Effectively we are determining the values of a and b by using the

information about the two points to solve the equation TC = a + bq.
This information enables us to locate two points on the graph with
the coordinates (0, 100) and (100, 400), where the two numbers in
brackets refer to the values of the variables measured along the horizon-
tal and the vertical axes respectively. In general, any point on this graph
(Fig. 1.1) is represented by the coordinates (q, TC).
For the first point: q = 0, TC = 100 and so
100 =a + b X0
a = 100
and for the second point q = 100, TC = 400 and so
400 =a+ 100b
Substituting the value a = 100 in this equation gives
400 = 100 + 100b
Subtracting 100 from each side of this equation leaves us with
300 = lOOb and so b=3
Hence TC = 100 + 3q is the equation which summarises the available

information about the production process and describes how costs vary
with output.
This equation can now be used to determine the total cost of
production at other levels of output. For example, when q = 20,
TC = 100 + 3 x 20 = £160.
If this point is included on the graph we find that it lies on the
straight line joining the two points corresponding to the initial data. We
would also find that all other points which satisfy the equation lie on the
same straight line. For this reason TC = 100 + 3q is known as a linear
equation.
In general the choice of symbols for the variables in an equation is
made by the people concerned with the presentation and analysis of the
information. Some letters tend to be used more frequently than others,
and some have fairly agreed usage for particular variables. However,
there are no exact rules to be followed, and it is important that the
principle is understood, and that emphasis is not placed on the symbols
themselves. It follows that Y = a + bX could have been used in the
example we have just considered, with Y replacing TC and X replacing
q. The form of the equation is unaltered. In the following sections a
number of different letters will be used to represent the variables in an
equation. This is not intended to confuse, but to assist in understanding
that the form of an equation does not depend upon the symbols which
represent the variables. However, in all practical cases a clear indication
should be given of what the symbols do represent.
It is possible to determine the form of a linear equation from two
pairs of values of the variables. This follows graphically because only
one straight line can be drawn through any two given points and
algebraically because there are two constants to be determined and two
pairs of values provide two equations in a and b which can, in general,
be solved.
1.3 Intercept and slope

In the equation Y = a + bX, a is often referred to as the intercept
term because it is the value at which the straight line intercepts the
vertical axis, i.e. it is the value of the function when X = 0, and b is
often referred to as the slope or gradient because it is a measure of the
inclination of the line to the horizontal axis.
We can now make use of such an equation in a number of ways. For
example we can examine the effect which would be produced by an
alteration in a and b, e.g. in fixed and variable costs, as follows:
Linear Equations 7
y y y
(a) Y =a+ bX (b) Y= a+ b'X (c) Y= a+ b"X
Fig. 1.2
1. Fixed costs remain the same but variable cost changes. In Figure 1.2 a
is the same in all three cases, but in graph (b) the variable cost is greater than in
graph (a) and in graph (c) variable cost is less than graph (a).
That is, b' is greater than b (written b' > b)
and b" is less than b (written b" < b)
The line will then rotate about the point at which it crosses the cost axis.
2. Variable cost remains the same but the fixed costs change. The
line will then move up or down parallel to itself, i.e. with the same
slope. In Fig. 1.3 b is constant in all three cases, but graph (b) shows
higher fixed costs than graph (a) and graph (c) shows lower fixed costs
than graph (a). That is,
a'> a and a"< a.
y y y
a{ ·( a"{
X X X
(a) Y=a+bX (b) Y= a'+ bX (c) Y=a"+bX
Fig. 1.3
1.4 Interpolation and extrapolation

We have seen that a relationship between two variables can be
shown graphically, or alternatively it can be represented by a mathemat-
ical expression. This equation can then be used to determine the value
of total cost at a point in between the points for which we have some
empirical data. This is known as interpolation.
It is, of course, possible to continue the straight line indefinitely in
either direction, i.e. for negative values of X and for values of X greater
than 100. This is known as extrapolation and care must be exercised if
such a procedure is attempted because the end result may not be a
sensible one. For example, values of X which are negative correspond to
negative output and have no practical meaning. We must, therefore,
eliminate such values by specifying that X must be greater than or equal
to zero.
But what about values of X greater than 100? These may or may
not be possible, and we have to be careful when we are considering
extrapolation that we are certain the fixed and variable costs will remain
the same when more than 100 units are produced. If they do not, then
we must add the condition that X cannot be greater than 100, or
alternatively, X must be less than or equal to 100.
The cost function can then be written
Y = 100 + 3X (0 ~ X ~ 100)
where the constraint, written inside the brackets, implies that the cost
function is only a valid representation of the process for outputs from
0 to 100 units.
1.5 Exercises
1. Draw the graphs of the following equations for the range X = 0 to
X = 6. What are the intercept term and the slope for each equa-
tion?
(a) Y = 3X +4 (b) Y = 3X- 4
(c) Y = 4- 3X (d) Y = 3X
(e) Y = 4
2. Establish a linear relationship for the total cost function which fits
the following two conditions:
(a) at an output of 10 units the total cost is £70, and
(b) at an output of 20 units the total cost is £120.
What do you infer about the fixed and variable cost? Sketch the
graph of the total cost function for outputs less than 30.
3. Establish a linear relationship for the total cost function from the
following data:
Linear Equations 9
(a) the variable cost is £6 per unit, and

(b) the fixed cost is £15.
What output would make the total cost £75? Sketch the graph of
the total cost function for outputs less than 20.
4. The total cost of production (Y) of a product is related to the
number of units of output (X) by the equation
Y=3+2X
(a) Sketch the graph of this line for 0 :s; X :s; 25
(b) What are the fixed cost and variable cost?
(c) What wil be the total cost of producing 15 units of output?
(d) If the total cost is 45, how many units of output will be
produced?
1.6 Simultaneous linear equations

We have seen how to obtain the equation of a straight line from two
pieces of information. These allowed two equations to be obtained
which had to be satisfied simultaneously and, by using algebra, the
values of the unknowns, a and b, were determined.
Before turning to applications in economics and business it is useful
to review some of the basic rules of algebra. While these may seem
obvious, it is still worthwile considering them because they do not always
transfer to later applications (such as with determinants, matrices and
sets). Let x, y and z represent any three numbers or be particular values
of three variables. Then, if they are non-zero, x, y and z satisfy the
following:
1. Commutative rule of addition and multiplication:
x +y =y +x and xy = yx
Thus the order in which two numbers are added or multiplied does not
matter. Notice that this does not apply to subtraction and division, since
5 - 3 = 2 while 3 - 5 = -2, and 8/2 = 4 while 2/8 = 0.25.
2. Associative rule of addition and multiplication:
(x + y) + z = x + y + z = x + (y + z)
(xy)z = xyz = x(yz)
3. Distributive rule: x(y + z) = xy + xz
or (x + y)z = xz + yz
4. Operations involving zero: If x is non-zero,

0/x = 0, x/0 = oo, 010 is indeterminate.
5. Any valid equation remains valid if we add the same expression
to or substract the same expression from both sides of the equation.
6. Any valid equation remains valid if we multiply or divide both
sides of the equation by the same non-zero expression.
These are useful for simplifying equations. For example, the ex-
pression (x + 2)(y - 1) can be expanded as
(x + 2) (y - 1) = x(y - 1) + 2(y - 1) = xy - x + 2y - 2
Also, in the equation
X - 2y = 2x - y + 3
the terms in x and y can be collected together on the left hand side by
subtracting 2x from each side, and adding y. This results in
X - 2y - 2x + y = 2x - y + 3 - 2x + y
or, -x- y = 3
and multiplying by -1, x + y = -3.
In economics and business studies many situations arise where
there are two or more equations which have to be satisfied simul-
taneously. For illustration let us return to our earlier example in which
the cost of production could be represented by a linear equation
Y =a+ bX
This relationship is obviously important but does not tell us any-

thing about the profitability of the overall operation. This can only be
assessed if the price is known at which the resulting product might be
sold along with the quantity that can be sold at this price.
Under conditions of perfect competition it is usual to assume that
the price of the article is fixed and that at this price it is possible to sell all
that can be produced. This would be the situation confronting an indi-
vidual producer whose output formed only a very small part of the total
demand. In this case his demand curve would be horizontal (i.e.
price = p) (Fig. 1.4).
The revenue function can then be represented by a linear equation
which can be written as
Linear Equations 11
Total revenue = price per unit X number of units sold

i.e. R = pX
Thus when X = 0, R = 0.
Quantity
Fig. 1.4
Since in this case the revenue function starts at zero and the cost
function at the value a, it is obvious that the two lines will intersect at a
positive level of output only if the constant p is greater than the constant b.
Assuming that this is so we would have the situation in Fig. 1.5.
Notice that we are using one set of axes for two graphs. We can do
this because the horizontal axis represents the number of units of output
and this scale is the same for both equations. The vertical axis represents cost
and revenue respectively for the two equations but these are both measured
in£ and, therefore, the same axis with the same scale can be used.
The point K where the lines intersect is often referred to as the
breakeven point because at this point the total revenue equals the total
cost and the net revenue is zero. The volume of output corresponding to
this point is equal to X 0 , which is obtained by dropping a perpendicular
from the point of intersection to the horizontal axis. If sales are below
X 0 a loss will be made and if sales are above X 0 a profit will be made.
The graphs which have been drawn show the cost and revenue
equations and from them we can determine the breakeven point.
The same result can also be obtained from the following set of
equations:
x.
Number of units
Fig. 1.5
Y =a+ bX (1)
R =pX (2)
with the added condition that at the breakeven point
R=Y (3)
These are three equations in three unknowns (R, Y and X) and are
then a complete description of the system. By using (3) we can equate
(1) and (2) to determine X 0 , the breakeven point:
R=Y
:. pX =a+ bX
This equation can be solved for X by grouping terms involving X on
one side of the equality. Subtracting bX from each side gives
pX - bX = a + bX - bX
or pX- bX =a
We can now take out the factor X from the terms on the left hand
side because it is common to both and obtain
X(p- b)= a
We then complete the algebraic manipulation by dividing both
sides of the equation by the term (p - b) to give
a
X=--
(p-b)
Linear Equations 13
None of the changes we have made has in any way altered the
relationship between the variables, although we have ended up with an
expression which enables us to calculate the breakeven point for the
process for any value of p, given that we know the values of a and b. In
descriptive terms,
. fixed cost
Breakeven pomt = pnce
. - vana
. bl e cost
The difference between the price and the variable cost at a particu-
lar level of output is often termed the contribution per unit of product,
because it is the net revenue which is produced by each extra unit sold.
The breakeven point X 0 is equal to the ratio of the fixed cost to the
contribution per unit of product. Since the producer's net revenue
(represented by N) is the difference between his revenue and his costs, it
is given by
N = R - Y = pX - a - bX
= (p- b)X- a
Example
Total cost of production is given by Y = 25 + 6X and price is fixed
at £11. What is the value of output at the breakeven point, and what is
the net revenue if 20 units of output are produced?
At the breakeven point
total cost = total revenue
that is, = pX = llX
Y
25 + 6X = llX
25 = 5X
X=5
At the breakeven point the output is 5 units. The net revenue is given by
N = pX - Y = llX - 6X - 25 = 5X - 25
When X= 20, N = 5 X 20 - 25 = 100 - 25 = 75. Hence, a net rev-
enue of £75 is obtained when the output is 20 units.
1. 7 Demand and supply

For any particular product there are, in general, a number of
potential consumers and the quantity which each consumer buys
depends, at least in part, upon the price at which the product is offered on
the market. It is safe to assume, therefore, that the quantity demanded
of any good is affected by its price and that for most goods the rela-
tionship is an inverse one. That is, the quantity demanded decreases as
the price is increased and vice versa. This can be expressed mathemat-
ically, and in the simplest case it is possible to think of it as being
represented by a straight line (Fig. 1.6).
Fig. 1.6
Although the quantity demanded is, by convention, represented

along the horizontal axis the equation is usually written algebraically in
the form
qd = f(p) =a + bp
In this example the constant b would be negative because the line
is downward sloping and the relationship is such that the quantity de-
manded qd decreases as the price is increased.
In a similar way the market, in a competitive situation, can be
considered to be supplied by a large number of independent producers.
The number of such producers and the level of their individual outputs
for any product is determined by the cost of production and the price
which the market is prepared to pay for the product. It can be assumed
that the supply of any product is affected by the price which can be
obtained for it, and, in this case, the quantity supplied increases as the
price increases (Fig. 1.7).
Linear Equations 15
Fig. 1.7
Algebraically this can be written
qs =a'+ b'p
The two equations can be drawn on the one set of axes (Fig. 1.8).
p
Po
qo q
Fig. 1.8
The point of intersection of the two lines is the point at which the
demand and supply for the product are equal and the market is in
equilibrium.
This point can be found by solving the pair of simultaneous equa-
tions:
qd =a+ bp
qs =a'+ b'p
using the information that at the equilibrium point qd = qs = q0, say.
From these three equations
a + bp0 = a' + b'p0
Subtracting a and b'p0 from each side and cancelling gives
Po(b - b') = a' - a
a'- a
Po= b- b'
and q0 =a+ b(po)
=a+b - - [a'b--a]
b'
ab- ab' + ba' - ba ba' - ab'
b- b' b- b'
These examples show one method of solving a system of simul-
taneous equations. In general, if there are as many equations as un-
knowns, then it is possible to obtain a unique value for each of the
unknowns. Cases do arise, however, when this is not possible, but we
will defer discussion of these until Chapter 2.
1.8 Exercises
1. The total cost of production (Y) of a product is related to the
number of units of output (X) by the equation
Y = 33 + 2X
(a) What is the breakeven point if the price is fixed at £13 per unit?
(b) What would the net revenue be at an output of 15 units?
(c) Sketch the total cost and revenue functions on the same graph.
Linear Equations 17
(d) Why would the producer not fix the price at £1 per unit?
2. A producer has a fixed cost of £50 and a variable cost of £5 per unit
of output if his output is less than 200 units.
(a) What is the equation of his total cost function (which is linear)?
(b) What is the breakeven point if the producer fixed his price at
£10 per unit?
(c) What is the producer's net revenue if the output is 12 units and
the price is (i) £5, (ii) £10, (iii) £15?
3. A producer is willing to supply a market with an output qs accord-
ing to the relationship
qs = 25p- 10
where p is the price.
The demand function for the product is
qd = 200- Sp
where qd is the quantity demanded at price p.
(a) What is the equilibrium price and quantity?
(b) If the producers supply function changes to
qs = 20p- 25
what is the new equilibrium position?
4. Draw the graphs of the following equations and where possible
find the values of p and q which satisfy the equations simul-
taneously.
(a) 2p + 3q = 17 (b) 6p- q =3
Sp- 4q = 8 4p + 7q=2
(c) 2p + 4q = 7 (d) 2p + q = 1
4p + 8q = 14 4p + 2q = 1
1.9 The effects of taxes on demand

In section 1. 7 we discussed the determination of the equilibrium
price and quantity when the demand and supply equations are linear.
We now continue with this model and examine the effects of taxes on
the equilibrium values. Initially we will consider a flat-rate tax oft per
unit, which is independent of the selling price and then discuss a tax
which is a percentage of the price. The former corresponds to an excise
tax such as might be imposed on a bottle of spirits, while the latter is

more like a sales or value-added tax.
Let the original demand and supply curves be
qd = 100- 5p
and qs = -20 + lOp
where p is the market price, qd the quantity demanded and qs the
quantity supplied. With no taxes the supplier receives p from each sale
and the purchaser pays p. The equilibrium is when qd = qs and the
equilibrium values are q 0 = 60, Po = 8.
Suppose a flat-rate tax of 3 per unit is imposed. The price received
by the supplier p' is now
p' = p- 3
since 3 is paid to the government in tax. The new supply function is
therefore
qs = -20 + lOp'
= - 20 + lO(p - 3)
=-50+ lOp
The effect of the tax is to change the intercept term of the supply
function from -20 to -50. The demand curve is unaffected. The new
equilibrium is when qs = qd or
-50+ 10p 1 = 100- 5p 1
or 15p 1 = 150
giving p 1 = 10 and q 1 = 50
The equilibrium has changed from Po = 8, q0 = 60 to p 1 = 10,
q 1 = 50 and so the new price is higher and the quantity is lower. This is
illustrated in Fig. 1. 9 where the original supply curve is labelled S and
the after-tax supply curve is S'. As p is on the vertical axis the supply
curve has shifted to the left. Also, notice that while the tax is 3 per unit,
the increase in price to the consumer is only 2 per unit and so the tax has
not been fully passed on to the customers. The tax revenue is 3q 1 or 150
while the supplier's income is q 1(p 1 - 3) or 350 compared with p 0 q0 or
480 originally.
The above process can also be carried out for general linear supply
and demand functions
Linear Equations 19
p
20
5'
10
d
~ ~ ~ ~~
20 40 50 60 80 100 q
Fig. 1.9
qs =a'+ b'p 1 (b' > 0)

qd =a + bp (b < 0)
and pt = p- t
The new equilibrium is now
P1 = (a' - a - b't)!(b - b')
and q1 = (a'b - ab' - bb't)!(b - b')
If the tax t is zero these equations reduce to those at the end of section
1.7 while if t > 0, since b' >0, it can be seen that p 1 >Po and q 1 < q 0 •
We now turn to the case of a percentage tax which is at a rate of
lOOr% (so that r = 0.15 for a 15% tax). Again the demand curve is
unaffected and there is a difference between the market price, p, and
the price received by the supplier, p'. Here p' = p - rp. Taking the
example used above and r = 0.15, the original supply curve is
qs = -20 + lOp
and the tax changes this to
qs = -20 + lOp'= -20 + 10(1 - r)p
= -20 + 8.5p
The effect of the percentage tax is to change the slope coefficient in the
supply curve from 10 to 8.5, a decline of 15%. The demand curve is
qd = 100- 5p
and the equilibrium is when qs = qd so that
- 20 + 8.5p 1 = 100 - 5p 1
= 120
13.5pl
or P = 8.89 and q = 55.56
1 1
The original equilibrium was Po = 8, q = 60 and so

0 the 15% tax has
increased the price by 11% and the supplier's revenue has dropped from
p 0q0 = 480 to (1 - r)p 1q 1 = 419.8. The graphs of the two supply curves
and the demand curve are given in Fig. 1.10, where again p is on the
vertical axis and q on the horizontal axis.
20
S'
10
8.89
8
d
~ ~ ~ ~
20 40 55.5 60 80 100 q
Fig. 1.10
Finally, we repeat the case of a percentage tax for the general linear
supply and demand curves. The new supply curve is
qs =a' + b'pr (b' > 0)
Linear Equations 21
the demand curve is

qd =a + bp (b < 0)
and the supplier's price is
p' = (1 - r)p
By substitution, the equilibrium price (p,) and quantity (q,) are

p, = (a - a')/{(1 - r)b' - b}
and q 1 = {ab' (1 - r) - a'b}/{(1 - r)b' - b}
As previously, if the tax rate is zero these reduce to the results given at
the end of section 1. 7. It is easily seen that the effect of a percentage tax
is to increase the equilibrium price, and less easily seen that the equilib-
rium quantity will be reduced.
1.10 Simple national income models

As a further illustration of the use of simultaneous equations in
economics and business studies we now consider some small macro-
economic models. For simplicity we will start with a two-equation
consumption-national income model, in which there is no government
sector and no external sector. The first equation is a linear consumption
function which relates aggregate consumption, C, and national income,
Y, while the second equation, an identity, defines national income as
being the sum of aggregate consumption and what we will call non-
consumption, N. Here the two equations explain the endogenous vari-
ables C and Y, and there is one exogenous variable, N, which is
determined outside the system. Initially we will also assume that N is
fixed at some value, N 0 • The equations are
C = 100 + 0.6Y
Y =C+N 0
Here the marginal propensity to consume, the proportion of an extra

unit of income which is consumed, is 0.6, and the level of consumption
when income is zero is 100. These two equations can be solved to give
the equilibrium values of C and Y in terms of N 0 • Substituting for C in
the second equation gives
Y = 100 + 0.6Y + No
and collecting the Y terms on the left hand side results in

0.4Y = 100 +No
and dividing through by 0.4 gives
Y = 250 + 2.5N 0
Substituting this into the consumption equation,

C = 100 + 0.6(250 + 2.5N0 )
and simplifying,
C = 250 + 1.5N0
In these solutions the values of Y and C depend on the value of N 0 • If N 0
increases by one unit then Y will increase by 2.5. This value, 2.5, is
called the multiplier, since it gives the effect on aggregate income of a
unit increase in non-consumption expenditure.
More generally, if the two equations are
C =a+ bY
Y =C+N 0,
the solutions are easily shown to be

C = (a + bN0 )/(1 - b)
Y = (a + N 0 )/(1 - b)
Here the multiplier, from the income equation, is 1/(1 - b). Since b is
the marginal propensity to consume, 0 < b < 1, and the multiplier is
positive. If this condition is violated strange results can occur. For
example if b = 1, the multiplier is infinite as are the equilibrium values
of C and Y, while if b is negative, both C and Yare negative.
The above two-equation model is obviously very unrealistic and we
now expand it by introducing a simple investment function. We will
assume that
I= c + dY
so that investment is a linear function of national income, with c > 0 and
0 < d < 1, making it endogenous. (A model in which investment is
affected by the rate of interest is discussed in section 1.11 below.) The
national income identity needs to be modified to take account of invest-
ment:
Linear Equations 23
Y= C +I
and the consumption function is as before:
C =a+ bY
To solve these three equations for the equilibrium levels of income,
consumption and investment, substitute for C and I in the national
income equation to give
Y = a + bY + c + dY
and rearranging,
Y = (a + c)/(1 - b - d)
We notice that for Y to be positive, which is to be expected, since both a
and c are positive, (1 - b - d) must also be positive and so
1>b+d
The solutions for C and I are easily given by substitution as
C = (a - ad + bc)/(1 - b - d)
I = (c - be + ad)/(1 - b - d)
Another obvious modification is the introduction of a government
sector which has an income from taxation (1) and expenditure (G). We
will assume that G and T are determined by the government and so are
exogenous. The national income identity is now
Y=C+I+G
The consumption function is modified to relate consumption to dispos-
able income (Y- T),
C =a+ b(Y- T)
and investment is determined as before
I= c + dY
The model consists of three equations in three endogenous variables, C,
Y and I. The solution follows the previous method of starting with the
national income identity to give
Y = a + b(Y-1) + c + dY + G
which reduces to
Y = (a -bT + c + G)/(1 - b - d)
and the solutions for the other variables are
C = {a- ad+ be+ bG- b(1-d)T}/(1- b- d)
I= (c- be+ ad- bdT + dG)/(1 - b -d)
The tax multiplier gives the effect on Y of a unit increase in T. From the
solution for Y it can be seen that if T increases by 1, Y changes by
-b/(1 - b - d). Since b and (1 - b - d) are both positive, this
multiplier is negative. That is, in this model, increasing the tax rate
reduces national income. Similarly, the government expenditure multi-
plier is the coefficient on Gin the solution for Y and so is 11(1 - b - d),
which is positive. This implies that an increase in government expendi-
ture will result in an increase in national income, if the model is correct.
We will discuss more complicated national income models in sec-
tions 1.11 and 1.15 below. However, before leaving these simple models
it is important to realise that they are not intended to represent any
particular economy, the numerical values are purely for illustrative
purposes and any policy implications are as valid only as the assump-
tions underlying the models.
1.11 IS-LM analysis

In section 1.10 a simple model of national income was developed in
which the relationships between national income, consumption, invest-
ment and taxation were discussed. Here we consider a different
approach, known as Is-LM analysis, in which equilibrium occurs in two
separate markets. These are the market for goods and the market for
money, where interest rates now have an important role. Again, we
initially ignore the government and foreign trade sectors.
At the aggregate level the goods market is in equilibrium when
income equals expenditure or Y = C +I, where Y is national income, C
is consumption and I is investment. The IS-schedule shows the different
combinations of interest rates and national income at which the goods
market is in equilibrium. Here it is assumed that investment is affected
by the interest rate, which is endogenous. For example, the investment
function might be I= 90- 60i where i is the rate of interest (so for a 5%
rate i = 0.05). The constant term in this investment function can be
interpreted as being autonomous investment- that is, the part of invest-
ment which is unrelated to interest rates and income. We will
assume that the consumption function is a simple linear one and that
Linear Equations 25
C = 10 + 0.8Y. The goods market therefore has the three equations to

determine Y, C, I and i:
Y= C +I
I = 70- 60i
C = 10 + 0.8Y
For equilibrium, the three equations are satisfied simultaneously and so
Y = C + I = 10 + 0.8Y + 70 - 60i
and simplifying gives Y = 400 - 300i which is the IS-schedule. Notice
that there is a negative relationship between Y and i.
The IS-schedule is concerned with the goods market. We now
consider the money market where, for equilibrium, money supply, M.,
equals money demand, Md. In this simple model we will assume that
money supply is exogenous and is in fact fixed at Ms = 250 and that
money demand is given by
Md = Lt + L2
where L 1 = 0.2Y is the transaction-precautionary demand for money,
which is determined by the level of national income, and L 2 = 190 - 100i
is the speculative demand for money which is negatively related to the
rate of interest. Here, equilibrium occurs in the money market when
Ms = Md
250 = 0.2Y + 190- lOOi
or Y = 300 + 500i
This is the LM-schedule which shows the different combinations of
national income and interest rates at which the money market is in
equilibrium.
For there to be equilibrium in both the goods and the money
markets, the values of Y and i must be the same in the IS- and LM-
schedules so that we need to solve the equations simultaneously. They
are
IS Y = 400- 300i
LM Y = 300 + 500i
which solve to give Y = 362.5 and i = 0.125 or 12.5%.
This model can be used to examine the effects of changes in the
assumptions about the different components. For example, suppose

there is an increase in autonomous investment and the investment
function changes from I = 70 - 60i to I = 80 - 60i, this moves the
IS-schedule to
Y = C + I = 10 + 0.8Y + 80 - 60i
or y = 450- 300i
With the LM-schedule from above the new equilibrium values are
i = 0.1875 or 18.75% andY= 393.75, so that, in this particular model,
an increase in autonomous investment results in a higher interest rate
and a higher national income. Investment was originally 62.5 and in-
creases to 68.75, so that the increase in autonomous investment is partly
off-set by the effects of the increase in the interest rate.
The model can also be extended to make it more realistic. For
example, government expenditure (G) could be included by adding G to
the national income identity and then treating G as exogenous or
endogenous. Further possible extensions include the effects of taxation,
the overseas sector and the labour market. In section 1.15 more com-
plex models will be discussed.
1.12 Exercises
1. The demand and supply equations for a particular product are
qd = 200- 4p
qs = -10 + 26p
(a) Determine the equilibrium values of p and q and the pro-
ducer's revenue they imply.
(b) A flat-rate tax of 5 per unit is imposed on each unit sold.
Determine the new equilibrium position, the tax revenue at
the equilibrium and the producer's revenue.
(c) Instead of the flat-rate tax of (b) above, a tax of 20% of the
price is impoosed on each item sold. Determine the new
equilibrium position, the tax revenue at the equilibrium and
the producer's revenue.
2. The demand and supply equations for a new product are
qd = 300- 6p
qs = -40 + 15p
Linear Equations 27
The government are undecided as to whether to impose a tax of

10% on each item sold or to impose a flat-rate tax. Show that the
price to the consumer would be the same in either case if the
flat-rate tax is approximately 1.7 per unit sold.
3. In a simple national income model, consumption (C) is given by
C = 20 + 0.7 Yd where Yd is disposable national income. The
national income identity is Y = C + N where N is non-
consumption and is exogenous. Determine the equilibrium values
of Y and C in the following cases if N = 4:
(a) There are no taxes and Yd = Y
(b) A lump-sum tax of T = 10 is imposed on income so that
disposable income is now Yd = Y - T
(c) The lump-sum tax is replaced by a tax on income at a rate of
25% so that Yd = Y - (0.25 Y)
(d) The lump-sum and the percentage tax on income are combined
so that Yd = Y - 10 - 0.25Y.
4. In a simple IS-LM model of an economy with two sectors the
goods market is described by
C = 15 + 0.8Y
I = 75 - 100i
Y= C +I
where Y is the national income, Cis consumption, I is investment
and i is the interest rate. The money market has the transaction-
precautionary demand for money given by
L1 = 0.1Y
and the speculative demand for money given by
L2 = 250 - 160i
Assume that the supply of money is fixed at Ms = 250.
(a) Determine the IS and LM schedules and hence the equilibrium
levels of Y and i.
(b) How is the equilibrium changed if autonomous investment
falls from 75 to 50?
(c) What is the effect on the equilibrium of the original model of
changing the money supply from 250 to 275?
5. The model in Question 4 can be extended to include a government

sector by changing the goods market equations to
C = 15 + 0.8Yd
I = 75- lOOi
Y=C+I+G
where Yd is disposable income, G is government expenditure and

the other symbols are as before.
(a) If Yd is defined by Yd = Y- T, where Tis taxes, and Tis given
by T = 5 + 0.2¥, and G = 300, determine the IS schedule.
(b) If the transaction-precautionary demand for money is given by
L, = 0.2¥
and the speculative demand for money by
L 2 = 750 - 260i,
and the supply of money is fixed at M. = 950, determine the LM
schedule and hence the equilibrium values of Y and i.
(c) What is the effect on the equilibrium if G increases to 350?
1.13 Determinant notation

We now consider the solution of a general set of simultaneous
equations. The general system of two equations in two unknowns is
auX 1 + ~ = b, (1)
a2 ,x1 + ~ = b2 {2)
where the as and bs are constants. Here, x, and x 2 are the unknowns and
au is the coefficient in {1) of x, and a12 is the coefficient in {1) of x 2 etc. In
general, therefore, aii refers to the coefficient in (i) of xi.
To obtain the solution for x, we eliminate x 2 from the equations by
multiplying (1) by a22 and (2) by a12 to give
auanX, + ~ = a22b,
a,2a2,x, + ~ = aub2
Subtraction now gives
Linear Equations 29
Hence, provided alla22 - a12a21 oF 0,

a22h1 - a12h2
(3)
and similarly (4)
Notice that the denominator in (3) and (4) is the same and that it is a
combination of all the a;1 values from (1) and (2). In particular if the
pattern of a;1 values from (1) and (2) is reproduced:
we see that the denominator is the cross-product of alla22 minus the

cross-product a12a21 . We now adopt a special notation and define the
determinant of the numbers a;j as
IAI =
It should be noted that the result of multiplying out or evaluating a

determinant is a number. For example, if all = 4, a12 = 2, a21 = 2, an = 3,
then
4 2
= (4 X 3) - (2 X 2) = 12 - 4 =8
2 3
or if all = 2, a12 = 0, a21 = 1, a22 = 3 then

2 0
= (2 X 3) - (0 X 1) = 6- 0 = 6
1 3
From the definition of a determinant it will be clear that
and
and hence (3) and (4) can be written in determinant notation:

b! atz att b!
I b2 azz I azt b2
and Xz =
x, ~ all atz allatz
azt azz I azt azz
The solutions to (1) and (2) are seen to be the ratios of determinants.
The denominators are the determinants of the coefficients of X 1 and x 2
whilst the numerators are essentially the same determinants with one
column, the first for X 1 and the second for x 2 , replaced by the column
made up of b 1 and b2 • This method of solving simultaneous equations is
known as Cramer's rule. For example,
3x 1 - X2 = 1
2xt + Xz = 4
att atz 3 -1
= (3 X 1) - [( -1) X 2) =3+2=5
azt an 2 1
b! atz 1 -1
= (1 X 1) - [( -1) X 4) =1+4=5
b2 azz 4 1
att b! 3 1
= (3 X 4) - (1 X 2) = 12 - 2 = 10
azt b2 2 4
Thus the solutions are
It will be clear from this that if the determinant of the coefficients of X 1

and x 2 is zero then there is no unique solution to the equations.
Linear Equations 31
For example, 2x, + 2x2 = b,

3x, + 3x2 = b2
Here
= (2 X 3) - (2 X 3) = 6 - 6 = 0
and there is no solution. This is because either the two equations are
identical and their graphs coincide, or the two equations are contradic-
tory and their graphs are parallel lines. In either case there is no unique
point of intersection. Determinants of the kind just discussed, which
have two rows and two columns, are known as two by two or (2 x 2)
determinants, and are said to be second-order determinants. It is poss-
ible to define higher-order determinants such as a (3 x 3) determinant
all al2 al3
a22 a23 a2, a23 a2, a22
a2, a22 a23 =all - a,2 + al3
a32 a33 a3, a33 a3, a32
a3, a32 a33
which is defined in terms of the three second-order determinants. These
are obtained by multiplying all by the determinant of the coefficients
which remains when the first row and first column are eliminated,
all. .a,2 .. .al3
a22 a23
a2, a22 a23 --+
a32 a33
a3, a32 a33
and similarly for the other terms. Notice that the pattern of signs is
positive on all, negative on a 12 and positive on a 13 . In this example the
determinant is said to have been expanded by the first row and it is
frequently convenient to do this. However, the value of the determinant
can also be obtained by expanding by any row or column, using the
same principles of multiplying the element from the particular row or
column chosen by the second-order determinants formed by deleting
the appropriate row and column. The rule for deciding the sign on the
coefficient is that if the coefficient for row i, column j is a;j then it is

r
multiplied by (-1 +j so that if i + j is even the coefficient is positive
while if i + j is odd the coefficient is negative.
To illustrate this, consider expanding the general (3 x 3) determi-
nant by the second column. This gives
all al2 a13

a21 a22 a23 = (-1 r +
2
al21 a21 a231 + (-1 )2+ 2a221 all a131
a31 a32 a33 a31 a33 a31 a33
+ (-1) 3• 2a321 all a13 1
a21 a23
= -al2(a2la33- a23a31) + a22(alla33- al3a31)
-a32(alla23 - a13a21)
This property of determinants is particularly useful when a determinant

includes some zero elements. For example, the following determinant
can be evaluated easily by expanding by the first column:
1 -1 2
0
0 -2
1 0 = 1
4
I-21 04 1- 0 + 0 = (1 X 4)- (0 X (-2)) = 4- 0 = 4
The possibility of expanding a determinant by any row or column

gives the useful result that if all the elements of any row (or column) are
zero then the determinant must also be zero. For example,
0 0 0
3 1 2 =011 21-013 21+013 11=0
2 4 1 4 2 2 1 2 4
Another useful property is that the value of a determinant is unchanged
if a constant multiple of any row (or column) is added to (or subtracted
from) any other row (or column). This is demonstrated by noting that
1 ~ = (2 X 1)- [(-1)(-1)) = 2- 1 = 1
1-1 2
while if the second row is added to the first,
0 1 I= (0 X 2)- [1 X (-1)] = 0 + 1 = 1
I -1 2
Another example of this property can be seen in the evaluation of

Linear Equations 33
10 15 -5
IAI = 8 -9 10
-4 3 2
Here if twice the third column is added to the first column, the result is
zero in the first row, first column position. Similarly, adding three times
the third column to the second gives another zero in the first row, and it
is then straightforward evaluating the determinant by expanding by the
first row. That is,
10 - 10 15 - 15 - 5
IAI = 8 + 20 -9 + 30 10
-4 + 4 3+6 2
-5(28 X 9) = -1260
In this example the numbers could have been simplified by using the result
that if any row or column is multiplied by a constant, the value of the
determinant is also multiplied by the constant. Thus, if a row has a common
factor, the factor can be moved outside the determinant. In the previous
example row 1 has a common factor of 5 which can be removed:
10 15 -5 5 2 3 -1
IAI = 8 -9 10 = 8 -9 10
-4 3 2 -4 3 2
and now column 1 has a common factor of 2 while column 2 has a
common factor of 3. Removing these gives
5(2)(3) 1 1 -1
IAI = 4 -3 10
-2 1 2
and this is easier to evaluate than the original determinant.
If any two rows (or columns) of a determinant are the same, the
value of the determinant is zero, since subtracting these will give a row
(or column) of zeros. For example,
1 2 3 0 0 0
1 2 3 = 1 2 3 =0
1 0 4 1 0 4
Further discussion of the properties of determinants is deferred until
Chapter 2, and this section concludes with the use of third-order deter-
minants in the solution of simultaneous equations.
The general system of three equations in three unknowns can be

written
a21X1 + azzXz + az3X3 = bz
a3lxl + a32Xz + a33x3 = b3

and by analogy with the two-equation system the solution is, using
Cramer's rule,
where
a11 a12 a 13 b 1 a12 a 13
IAI = az1 azz az3 , IA1I = bz azz az3
a3l a3z a33 b3 a3z a33
a11 b1 a 13 a11 a12 b1

IAzl = az1 bz az3 , IA31 = az1 azz bz
a3l b3 a33 a3l a3z b3
so that IA;I is the value of IAI with column i replaced by the column of
b-values. For example,
3x 1 + X2 - 2x 3 = 2
X1 + X2 + X3 = 1
Here
3 1 -2 1 1 1 1 1 1
IAI = 1 1 1 =3 - 1 + (-2)
2 2 3 2 3 2 3 2 2
= 3(3 - 2) - 1(3 - 2) - 2(2 - 2)
=3X1-1X1-2X0=2
2 1 -2 1 1 1 1 1 1
IAll = 1 1 1 =2 -1 + (-2)
3 2 3 2 3 3 3 3 2
= 2(3 - 2) - 1(3 - 3) - 2(2 - 3) = 4
Linear Equations 35
3 2 -2 1
IA21 = 1 1 1 =3 -2 + (-2)
2 3 3 3 3 2 3 2 3
= 3(3- 3)- 2(3- 2) - 2(3- 2) = -4
3 1 2 1 1 1
IA31 = 1 1 1 =3 - 1 +2
2 2 3 2 3 2 3 2 2
= 3(3- 2) - 1(3- 2) - 2(2- 2) = 2
Therefore the solution is
XI-
~
IAI -2
X
2= IAII
IAI
=- _j = -2
2
x3-
~
IAI - 2
which is easily checked by direct substitution in the original equations.
Cramer's rule is probably most useful for systems of three equa-
tions in three unknowns but it can be applied to larger systems. The
definition of a (4 x 4) determinant follows in the obvious way from that
of a (3 x 3) determinant. For larger systems the algebra involved can
become tedious and, for numerical problems, computer packages are
available which solve such systems without generating serious rounding
errors. A related approach is to use matrix methods, which are discus-
sed in Chapter 2.
1.14 Exercises
Use Cramer's rule to find the values of the variables which satisfy
the following equations.
1. 2xi+ 2x2 = 2
3xi- x 2 = 1
2. XI+ 3X 2 = 4
2xi + x2 =3
3. xi - X2 + X3 = 4
xi + X2 + 3x3 = 8
X1 + 2x2- x 3 =0
4. 2xl + 2x2 + x =3
3
2xl- 2x2 + x = -1
3
xt+ x2- 2x3 = 4

5. xt + x2- 2x3 = 0
2xl - 2x2 - 2x3 = -4
X1 + 2x2 + x =8
3
6. xt+ x2- 2x3 = 3

2xt- 3x2 + x =1
3
X1 - x2- x = 4.
3
1.15 Applications of determinants

In this section we examine some applications of the use of determi-
nants to solve simultaneous equations. The main areas of application
are in linear macroeconomic models and in the analysis of equilibria in
several markets.
We saw in section 1.10 that a simple three-equation national in-
come model, given by the national income identity
Y = C +I+ G,
a consumption function in which taxes are deducted from gross income,
C =a+ b(Y- n
and investment is a linear function of income,
I= c + dY,
can be solved for Y by substituting from the C and I equations into the Y
equation and then the solutions for C and I follow.
We now solve this system using determinants. First the equations
are written in the standard form, with the variables on the left hand side
of the equals sign and the constants on the right hand side. For compari-
son we also write out the standard equations.
Y-C-I=G
-bY+C=a-bT
Linear Equations 37
-dY +I= c
In the general case the solutions are, using Cramer's rule,
IA,I IA21 IA31

x, =TAT , X2 = lA/ , X3 = TAT
For the model the solutions are
y = IA,I c = IA21 I = IA31

IAI' IAI' IAI
where 1 -1 -1 G -1 -1
IAI = -b 1 0 'IA,I = a-bT 1 0
-d 0 1 c 0 1
1 G -1 1 -1 G
IA2I = -b a-bT 0 , IA31= -b 1 a-bT
-d c 1 -d 0 c
By adding the second and third rows to the first row, and expanding by
the first row,
1-b-d 0 0
IAI = -b ~ ~ = (1-b-d) ~ ~ = (1-b-d)
-d
Similarly, IA,I = G +a- bT + c,
1-d G+c 0
IA =
21 -b a-bT 0 = [(1-d)(a-bT)]- [(-b)(G+c)]
-d c 1
0 -1 G 1-b 0 G+a-bT
IA31 = 1-b 1 a-bT = 1-b 1 a-bT = 11-b ~ I
-d 0 c -d 0 c -d
= c(1-b) + d(G+a-bD
IA,I G + a - bT + c
Therefore, Y=-=-----
IAI (1- b- d)
c = _IA2I = [(1- d) (a- bT)] + b(G +c)
IAI (1- b- d)
IA31 c(1-b) + d(G + a-bT)

I = TAT = (1 - b - d)
These answers agree with the ones obtained at the end of section 1.10.
Our next application of determinants is to the analysis of equilibria
in several markets. When different products are competing in the same
market, the demand for each product will be affected by the prices of all
the products. For example, suppose that there are three types of apple,
A, Band C, being sold and their demand and supply equations are:
qj = 20- 2pA + 4pB + Pc ~ = -5 + 4pA
q_t = 10 + 3pA- 5p 8 + 2pc ~ = -7 + 3pB
~ = 70 + 4pA + 2p 8 - 5pc q-::: = -16 + 5pc
Here the supply depends only on the price of the particular type of apple
but the demands are interrelated. Equating q' and qd in each market, the
equilibrium prices satisfy
20 - 2p A + 4p B + PC = -5 + 4p A
10 + 3pA -5p 8 + 2pc = -7 + 3p 8
70 + 4pA- 2p 8 - 5pc = -16 + 5pc
and rearranging in the standard form:
- 6p A + 4p B + PC = -25
3pA- 8p 8 + 2pc = -17
4pA + 2p 8 - lOpe= -86
Using determinants,
-6 4 1
IAI = 3 -8 2 = -266
4 2 -10
-25 4 1 -6 -25 1
IAII = -17 -8 2 -3990, IA 21 = 3 -17 2 = -3192
-86 2 -10 4 -86 -10
-6 4 -25
IA31 = 3 -8 -17 = -4522
4 2 -86
Linear Equations 39
and so the equilibrium prices are pA = -3990/-266 = 15,

p 8 = -3192/-266 = 12 and Pc = -4522/-266 = 17. The corresponding
quantities of apples are q A = 55, q 8 = 29 and q c = 69.
1.16 Exercises
1. A simple macroeconomic model is given by
Y =C+l+G
C = 20 + 0.7Yd
I = 15 + 0.1Y
Yd = Y- T
T = 5 + 0.3Y
G = 20
(a) By eliminating G, T and Yd from the equations, reduce the
system to three equations in three unknowns and hence use
Cramer's rule to find the equilibrium values of Y, C and I.
Is there a government budget deficit?
(b) How does the equilibrium change if the government has to
balance its budget so that G = T?
2. In a simple national income model with a foreign sector the equa-
tions are
Y = C +I+G+ (X- M)
C =a+ bY
M = c + dY
T =fM
so that imports (M) depend on income and are taxed at a rate f per
unit. Assume exports (X) are exogenous and fixed at X= X 0 and that
G = T and I= 10 • Determine the equilibrium levels of Y, C, M.
3. The market for a product has three main brands and the demands
for each product are interrelated. For brand A demand (measured
in thousands) is q1 = 3- PA + p 8 and supply is q: = pA- 2, for
brand B, ~ = 8- 2p 8 + Pc and q: = p 8 - 1 and for brand C,
~ = 6 + 2pA- Pc and q<j = 2pc- 2. Find the prices and quanti-
ties which result in equilibrium for each brand.
4. Three products compete in a market and their demand and supply
functions are
qdt = 44 - 2pt + P2 + P3 qst = -10 + 3p 1

qd2 = 25 + Pt - P2 + 3p3 qs2 = -15 + 5p2
qd3 = 40 + 2pt + P2 - 3p3 qs3 = -18 + 2p3
where Pu p 2 , and p 3 are the prices. Determine the prices and
quantities which result in equilibrium in each market.
1.17 Revision exercises for Chapter 1 (without answers)

1. Sketch the graphs of the following equations for x = 0 to x = 10.
(a) y = 4x- 3 (b) y = 30- 2x (c) y = 15
(d) x =5 (e) y = 20x
2. If total cost is 100 when output is 20, and total cost is 90 when
output is 10, determine the equation of the total cost function,
assuming it is linear.
3. If the total cost (y) is related to the level of output (x) by the
equation
y=30+6x
determine the breakeven point if the price is 9 per unit. What
would the net revenue be at an output of 8 units?
4. The supply function for a product is
qs = 20p- 30
where q is the quantity and pis the price. The demand function is
qd = 170- 5p
(a) Determine the equilibrium price and quantity and sketch the
demand and supply equations.
(b) If the demand equation changes to
qd = 180 -lOp
what is the new equilibrium point?
5. Solve the following simultaneous equations:
(a) 6x + 6y- 2z = 4 (b) 2x + 2y- 3z =4
3x + 3y- z = 2 6x- 4y- 5z = 9
X+ 4y- Z = 3 x+ y+ z=2
6. If the supply and demand equations for a product are
Linear Equations 41
q' =SOp- 40
qd = 480- 15p
determine the supplier's revenue at the equilibrium point. A
flat-rate tax of 4 per unit is imposed. What effect does this have on
the supplier's revenue?
7. In the following national income model, Y is national income, C
is consumption, Yd is disposable income, I is investment, G is
government expenditure and Tis tax revenue:
Y =C+I+G
C = 10+ 0.6Yd
yd = Y- T
I = 25
T = 2 + 0.2Y
(a) If G = 25, determine the equilibrium level of national in-
come and whether there is a budget deficit.
(b) If G = T, determine the equilibrium level of national income.
(c) Does changing the tax function to
T= 2 + 0.15Y
increase national income?
8. The goods market in an economy is given by
C = 25 + 0.7Y
I= 90-40i
Y= C +I
where Cis consumption, Y is national income, I is investment and
i is the rate of interest. The money market has money supply fixed
at M' = 500 and the demand for money is given by
Md = L 1 + L 2
where the transactions demand for money is
L1 = 0.25Y
and the speculative demand for money is
L2 = 550- 80i
(a) Determine the IS and LM schedules and sketch their graphs.
(b) What are the equilibrium levels of Y and i?

(c) If money supply changes to 550, what effect that this have on
the equilibrium value of i?
9. A macroeconomic model of a closed economy is given by the
following equations:
Y =C+I+G
C =a+ bYd
Yd = Y- T
T = c + dY
I = I0
G =Go
where Y is national income, Cis consumption, I is investment, Yd
is disposable income, Tis tax revenue, G is government expendi-
ture and both I 0 and G0 are constants.
(a) What is the equilibrium level of national income?
(b) If the budget must balance (so that G0 = T) what is the new
equilibrium level of national income?
10. There are three main brands in the market for petrol in a particu-
lar city and the demand functions are all interrelated. They are:
qf = 175- 3pi + 2p2 + 3p3
~ = 20 + P1- 2pz + 2p3
~ = 40 + 2p 1 + Pz - 2p3
The supply functions are
q: = 6pi- 60
~ = 3p2- 25
~ = 3p3- 10
(a) What are the equilibrium prices and quantities?
(b) If a flat-rate tax of 2 per unit sold is imposed, what is the new
equilibrium position?
Chapter 2
Elementary Matrix Algebra
2.1 The matrix notation
A matrix is a rectangular array of elements in rows and columns.
Examples of matrices are
: l
which are rectangular arrays with 2 rows and 2 columns, 3 rows and 3
columns and 2 rows and 3 columns respectively. The order of a matrix is
the number of rows and the number of columns, so that in the above
examples the orders are (2 X 2) or (2 by 2), (3 x 3), and (2 x 3)
respectively. The elements, which may be numbers or constants or
variables, are enclosed between square brackets to signify that they
must be considered as a whole and not individually. In contrast to the
determinants of Chapter 1, a matrix does not have a single numerical
value.
A matrix is often denoted by a single letter in bold-face type and a
general matrix of order m X n is written as
Xu X12 Xln
X = X21 X22 X2n
where the subscripts identify the row and column in which the element is
located. For example
x 12 is the element in row 1, column 2
x 34 is the element in row 3, column 4
and X;i is the element in row i, column j.
43
A square matrix has an equal number of rows and columns and in the
general case has order (n X n). If a matrix has only one row it is referred
to as a row vector and similarly a matrix with only one column is referred
to as a column vector.
Examples of these are
is a square matrix of order (2 x 2)
B = [3 4 5 1] is a row vector of order (1 x 4)
C __ [ C
11
Czt
l is a column vector of order (2 x 1)
2.2 Elementary matrix operations

MATRIX EQUALITY
Two matrices are equal if, and only if, the corresponding elements
of each matrix are all equal. For example,
if
if Xn = alH
Xzt = a21•
l [. . ·~ l
and, in general
X12 X In atz aln
[X"
Xzt Xzz
~ = ay
azz
Xml Xmz ··· Xmn am! am2 •.• amn
that is, X = A, if xij = aij for all i and all j, where xij• aij represent the
elements in the ith row and jth column of the two matrices respectively.
It follows that two matrices can be equal only if they have the same
number of elements arranged in corresponding positions, i.e. they must
be of the same order. In the above case, both X and A are (m x n).
Elementary Matrix Algebra 45
SCALAR MULTIPLICATION
A scalar is a real number, in contrast to a vector or matrix. If k is a
scalar then the product of k and a matrix is obtained by multiplying each
element of the matrix by k. For example,
kxu
kX = [ kx21
kx31
3A =[ 6 12]
0 -3
ADDITION AND SUBTRACTION

If A and B are of the same order then C = A + B is found by
adding the corresponding elements in A and B, and D = A - B is found
by subtracting the corresponding elements in B from A. For example,
Ut A ~ [ i _1] 8 ~ ~ ~
[ - l
then C =A+ B = [ ~ ~ 4 ~~~ l 2~
0+2 -1+2
= [
3
;
1
xl
and D =A- B ~ (~ =~ -2~ 3-3~ xl
~0-2~ 4-1-2
MATRIX MULTIPLICATION
Two matrices A and B can be multiplied together to give AB only if
the number of columns in A is equal to the number of rows in B. For
example,
if and
then A is (3 X 2), B is (2 X 3) and the product of A and B can be found

since A has 2 columns and B has 2 rows. The product matrix C is
a11b1z + atzbzz
aztbtz + azzhzz
a3tbl2 + a3zbzz
where
and the general term is C;i = a; 1bti + a;2 b 2i.

The elements of C are obtained by multiplying row i from A by
column j from B to give c;i·
Notice that A is (3 x 2), B is (2 X 3) and C is (3 X 2)(2 x 3)
= (3 X 3).
Example 1
A is (2 X 2), B is (2 X 1) soC is (2 X 2)(2 X 1) = (2 X 1)
[2 1][3] [ 2X3+1X2] [ 6+2] [ 8]

C= 4 -1 2 = 4X3+(-1)X2 = 12-2 = 10
Example 2
A= [2 1 4],
3 2 0
Here A is (2 x 3) and B is (2 x 2) and so the product AB is not defined
since the number of columns in A does not equal the number of rows
in B. However, if
then B is (3 x 2) and the product is of order (2 X 3)(3 X 2) = (2 X 2)
C=AB=[
2 1 4
3 2 0
[i !]
0 1
=[2X2+1X1+4X0 2X3+1X4+4X1]
3X2+2X1+0X0 3X3+2X4+0X1
= [5 14]
8 17
In this case it is also possible to form the product D = BA, since B is
(3 x 2) and A is (2 x 3) and hence D is of order (3 x 2)(2 x 3)
= (3 X 3)
~ ~~~ ~ :1
2x2+3x3 2x1+3x2 2x4+3x0]
= [ 1X2+4X3 1X1+4X2 1X4+4X0
0X2+1X3 Ox1+1x2 Ox4+1x0
13 8 8]
= [ 14 9 4
3 2 0
Here D is (3 x 3) whilst C is (2 X 2). Therefore D and C cannot be
equal and we have
AB =I= BA
This is true in this particular case and in fact is usually true even when
the products AB and BA are of the same order.
Example 3
A is (2 x 2) and X is (2 x 1) so the product AX is (2 x 2)(2 x 1)

= (2 x 1) and is
It follows from this that the set of simultaneous equations

a11 X 1 + a12 X 2 = b,
az,x, + azzXz = bz
can be represented by
AX= b
where
THE IDENTITY OR UNIT MATRIX

Corresponding to the number 1 in ordinary algebra there is a matrix
I, known as the identity or unit matrix, which has the property
lA = A = AI
(compare with 1 x a = a = a X 1) and I is the square matrix
~~~
0 0 1
for order (3 x 3) and for order (n X n),
1 0 ... 0
I= 0 1 ... 0
0 0 ... 1
For example, if
then
IA=[1 0][1 2]=[1X1+0X4 1X2+0X3]

0 1 4 3 Ox1+1x4 Ox2+1x3
THE TRANSPOSE OF A MATRJX

If the rows and columns of a matrix are interchanged the new
matrix is known as the transpose of the original matrix. For example,
if
the transpose,
where the first row of A, 1 2 becomes the first column of A', and the
second row of A, 3 4 becomes the second column of A'.
If then
If A=[a b c] then
2.3
1.
form the following (if they exist)

(a) A+ B, (b) A- 2B, (c) AB, (d) AC, (e) CA, (f) AB', (g) C'A
and (h) show that AB =#:BA.
2. Given that
1 0
A= [ 1 1
0 2
j] C=[l 0 2]
form the following (if they exist):

(a) AB, (b) A'B, (c) AC, (d) AC', (e) CB.
2.4 The inverse matrix

In section 2.2 we saw that a set of simultaneous equations could be
written in matrix form. For example the equations
can be written as AX =b
where
The solution of these equations is obtained by expressing the unknowns

x,, x 2 , x 3 in terms of the known constants a 11 , aw ... , a33 , b,, b2 and b 3 •
For the matrix equation AX = b this is equivalent to transferring the A
to the other side of the equation. To do this we define a matrix A-',
known as the inverse of A, which has the property
AA-' =I= A-'A
i.e. the product of a matrix and its inverse is the unit matrix. If the
inverse of A exists then A is said to be non-singular.
Since AX=b
multiplying each side by A-', we have
A-'AX = A-'b
or
But IX = X and so
is the solution to the equation.
Before considering how to obtain A-' from a matrix A we recall that
in section 1.13 the condition for a solution to exist for a set of simul-
taneous equations was shown to be that IAI =I= 0, where IAI is the
determinant of the coefficients of the unknowns. In the case of three
equations in three unknowns,
all al2 au
IAI= az1 a2z az3
a31 a32 a33
and this was evaluated as

azz az3 az1 az3 az1 azz
IAI =a11 - a1z +au
a3z a33 a31 a33 a31 a3z
These (2 X 2) determinants are the minors of the elements of A, so that

the minors of a 11 and a 12 are respectively
and
and in general the minor of a;i is the determinant of A with row i and
column j eliminated. The cofactor (c;i) of the element a;i of the matrix A
is the minor of a;i multiplied by (- 1); + i, so that if i + j is even the
cofactor and minor are equal, and if i + j is odd, the cofactor is the
negative of the minor.
For example if
the minor of an is
and the cofactor of a22 is
c22 =(-1)2+2
52 Introductory Mathematics for Economics
For a 23 the minor is
and the cofactor is
Cz3 = (-1 )z + 3 all a 12 1-- _ all a 12

a3t a3z 1 a3t an
It follows that for A of order 3 X 3 then
The cofactor matrix is the matrix with elements cii• the cofactors of
A, and this expansion of the determinant of A is known as the Laplace
expansion. It extends in the obvious way to fourth- and higher- order
determinants.
Example I
A=
2 ... 1
[: :l
The minor of all is =4
0 4
2 ... 1
The minor of a 12 is =0
0 4
2 1
The minor of a 21 is = 1
0 .. .4
2 1
0 .. .4
is C= [ Cu
c,
C21 c22
Hl
The cofactors are C11 = 4, c12 = 0, C21 = -1, c22 = 2. The cofactor matrix
-1
4
0
Example 2
A= [ 02 41 -11
2 2 1
l
4 -1
The minor of a 11 is =4+2=6
2 1
0 -1
2 1
0 4
The minor of a 13 is = -8
2 2
1 1
The minor of a2 , is = -1
2 1
2 1
The minor of a22 is =0
2 1
2 1
The minor of a23 is =2
2 2
1 1
The minor of a3 , is = -5
4 -1
2 1
The minor of a 32 is =-2
0 -1
2 1
0 4
The cofactor matrix is
c=[ ~
-2
0 -8]
-2
-5 2 8
One more definition is required before the inverse can be obtained:

the adjoint of A is the transpose of the cofactor matrix of A.
Thus, if
C 11
adjoint (A) = [ C12
c13
The inverse of a matrix A is given by

1
A- 1 = IAI adjoint (A)
This can be shown to be true quite easily when A is (2 x 2) and with

more difficulty for higher orders.
Let then
and hence
adjoint (A) = [ a
22
-a 12 ]
-a2l au
Now
IAI =
If A- 1 is the inverse then AA- 1 = I.

A adjoint
AA- 1 = _
(A)
___:._ ____:___:_
Now
IAI
- a11 a12 + a12a11
- a21a12 + a22a11
l
=I
Hence A- 1 is the inverse of A.
Example 3
+ 4x 2 = 2
2x 1
3x + 4x = 1
1 2
Let
then c= [ 4-3]
-4 2
and IAI = -4
1
A- 1 =-adjoint
IAI
(A) = 1- [ 4 -4] = [ -1
-4 -3 2 _J
4
1
!
2
l
Check:
The solution to the equations is
X2 = 1
Example 4
X1+ X2 + X3 = 6
2xl - Xz + 2xJ = 6
x1 - x3 =-2
Here
A= ;
1
~0 -1; l and c = [ ~ -i0 -3~
3
l
and IAI = 6
A- 1
1
=\AT adjoint (A) = ~ ~1 ~1 -3~ l
Check:
-u lul~ [~ l
The solution to the equations is
X ~ ~ ~ ~! ~
[ l A 'b
and so X1 = 1, X2 = 2, X 3 = 3.
2.5 Exercises
1. Obtain the inverse of
and hence solve the equations

2xl + Xz = 4
x 1 + 3x2 = 7
2. Solve the following equations using the inverse matrix method.
(a) 2x1 + 5x2 = 2
3x1 - 4x 2 = 10
(b) 2x 1 + 2xz + X3 = 1
3x1 + X 2 + X 3 = 2
X1 + X2 + X3 =2
(c) X1 - + X3 = 3
X2
2xl - 2xz + x3 = 5
X 1 + X 2 + 2x 3 = 4
2.6 Inversion by Gaussian elimination

The inverse of a matrix can also be found by the method known as
Gaussian elimination. This is particularly suitable when the elements of
the matrix are numbers. The method involves augmenting the matrix to
be inverted by including the identity matrix next to it. Then row opera-
tions are applied to both matrices until the original matrix is trans-
formed into the unit matrix, whereupon the identity matrix will have
been transformed into the inverse matrix. The permitted row operations
are:
(1) Any row of the matrix may be divided (or multiplied) by any
non-zero constant.
(2) A multiple of any row may be subtracted from (or added to)
any other row.
For example, to find the inverse of
A= [ 2 -1
-1 1
l
write it as A I I or 2 -1 1
[
-1 1 0
The matrix A is to be transformed into the unit matrix so that the

first column becomes 1, 0 and the second column, 0, 1. This is best done
by working systematically on the first column, and then the second. Now
dividing the first row by 2 gives the desired 1 in the first-row, first-
column position:
1 -0.5
[
-1 1
By adding the first row to the second, a 0 will occur in the second-row,
first-column position:
[
1 -0.5 0.5
0 0.5 0.5
01 l
Next, add the second row to the first, to give a 0 in the first-row,
second-column position.
1 0 1
[ 0
0.5 0.5
Finally, divide the second row by 0.5 to give a 1 in the second-row,

second-column position.
1
The result is the unit matrix and so the inverse is

which can easily be checked by showing AA- 1 = I.

For a second example, to find the inverse of
-1 1
B =[ -1
3
1 0
2 2 -1
write the matrix next to the (3 x 3) unit matrix:
[_; -i ~ ~ ~ ~
2 2 -1 0 0
I
The procedure is to transform the first column into 1, 0, 0, the second
into 0, 1, 0, and the third into 0, 0, 1. Here we add twice the second row
to the first, to give a 1 in the (1,1) position, and also add twice row two
to row three to give a 0 in the (3,1) position:
1 1 1 1 2
[ -1 1 0 0 1
0 4 -1 0 2
Next, add the first row to the second to give a 0 in the (2,1) position,
giving the required 1, 0, 0 in the first column:
[g Ll i ~ ~ I
I
To change the (2,2) element to 1, divide the second row by 2:
[~
1 1
1 2 o
1 0.5 0.5 1.5 0
4 -1 0 2 1
Subtract row 2 from row 1, and subtract four times row 2 from row 3 to
give 0, 1, 0 in the second column:
[~
0 0.5
1 0.5
0 -3
0.5 0.5
0.5 1.5
-2 -4
~ l
Divide row 3 by 6, and add the new row to rows 1 and 2:
[ ~ 0
1
0 116 -116 116]
0 116 5/6 116
0 -0.5 -1/3 -2/3 116
Multiply row 3 by - 2, to give 0, 0, 1 in the third column:
[ ~ 0
1
0
0
0
1
1/6 -116 116]
116 5/6 116
2/3 4/3 -113
The result is that
B- 1 = [ 116 -116 1/6]

116 5/6 116
2/3 4/3 -113
and this can easily be checked by showing that BB- 1 = I.

The method of Gaussian elimination can also be used to solve
simultaneous linear equations, giving both the inverse of the coefficient
matrix and the values of the unknowns. For example, the equations
2x 3y z = 0
x+ y+ z=4
-x + y + 2z = 1
can be solved by writing the coefficient matrix, the unit matrix and the
column of constants thus:
2 -3 -1 1 0 0
[ 1 1 1 0 1 0
-1 1 2 0 0 1
and applying the row operations as before. If the third row is added to
the first row and then the second row, this gives a 1 in the (1, 1) position,
and a 0 in the (2, 1) position:
1 0 1
0 1 1
0 0 1 ~ l
Adding the first row to the third makes the first column 1, 0, 0, as
required, and adding the second row to the first starts changing the
second column to 0, 1, 0:
I~ 0 -1
~ j
3
~1 0~ i2 ~
Adding the third row to the second, and the new second row to the third
completes the operations on the second column:
I :
Divide row three by 9:
~ ~
0 0 9
1 1 2
1 1 3
2 1 5 !l
I~
0 4 1 1 2
1 6 1 1 3
0 1 2/9 119 5/9
Subtract four times row 3 from row 1, and subtract six times row 3 from
row 2 to give the unit matrix, the inverse and the solutions:
[~
0 0 119 5/9 -2/9
1 0 -113 113 -113
0 1 2/9 119 5/9
The solutions are x = 2, y = 1 and z = 1, which can be checked in the
original equations.
If the inverse matrix is not required row operations can still be used
to obtain a solution. The coefficients matrix is still transformed into the
unit matrix. For example, the equations
X + y = 2
2x y 1
give the coefficients matrix and column of constants
Subtracting twice row 1 from row 2 gives
Dividing row 2 by -3, and subtracting the new row 2 from row 1 results
in the unit matrix and the solutions:
Here, x = 1 andy = 1. Taking a more complicated example,

w + 2x - y + 2z = 3
2w + X + z = 3
w X + y 2
w + X + y + z = 2
The coefficients matrix and column of constants give
1 2 -1 2 3
2 1 0 1 3
1 -1 1 0 2
1 1 1 1 2
Subtracting twice row 1 from row 2, and row 1 from rows 3 and 4 gives
1 2 -1 2 3
0 -3 2 -3 -3
0 -3 2 -2 -1
0 -1 2 -1 -1
Divide row 2 by -3 and combine the new row 2 with the others to give
0, 1, 0, 0 in the second column
101/301
0 1 -2/3 1 1
0 0 0 1 2
00 4/30 0
At this stage row 4 can be divided by 4/3 and combined with the other
rows to give:
1 0 0 0 1
0 1 0 1 1
0 0 1 1 2
0 0 1 0 0
Finally, subtracting row 3 from row 4, adding the new row 4 to row 2
and row 3, and dividing the new row 4 by -1 results in
1 0 0 0 1
0 1 0 0 -1
0 0 1 0 0
0 0 0 1 2
and w = 1, x = -1, y = 0, z = 2. The answer should be checked in the
original equations.
2. 7 Exercises
1. Obtain the inverse of the following matrices by using Gaussian
elimination:
A - -[2 -3] - [ 0 2] C -- [3 -4] - [-21 -12 30]

B - D -
1 1 -1 -1 1 4 3 0 2
E = [ ~ ~ ~
-1 1 -1
F
3
= [- 1
2
~1 -1~ G = ~ ~2 ~1 i]
3
2. Solve the following equations using Gaussian elimination:
(a) 4x- 2y = 6
2x+ y=S
(b) 3x y+2z=7
X + 3y + z=3
2x + y+ z=4
(c) X + y+ z=2
4x + 3y + 2z = 7
X y- z=O
(d) 2w + x+ y+ z = 3
w x- y+ z=3
w + 2x + 3y - z=2
3w + 3x - y + 2z = 4
2.8 Linear dependence and rank
For an inverse to exist it is necessary for the original matrix to be
square. This has important implications in the solution of equations be-
cause a square matrix of coefficients necessarily means that the number of
equations is equal to the number of unknowns. If there are more equations
than unknowns then some of these must be redundant if a solution exists.
This means that some of them provide no additional information to what is
already known from the other equations. For example,
X + y 2
2x + y = 3
2x + 2y 4
In this case there are three equations in two unknowns, but the
third one is simply a multiple of the first and provides no further
information. A value for x and for y can be obtained from either the first
and second equations or the second and third equations taken together.
There is also the possibility that the extra equation is inconsistent.
For example,
X + y 2
2x + y = 3
2x + 2y 5
This set of equations has no solution because the values of x and y
obtained by solving the first and second simultaneously do not satisfy
the third. In particular, the first and third equations cannot be satisfied
simultaneously. (Graphically, they are parallel lines).
In general, with a set of simultaneous equations, when the number
of equations is equal to the number of unknowns a unique solution can
be obtained providing that no equation is either redundant or incon-
sistent with the others. In such a case the equations are said to be
linearly independent and IAI =I= 0. If they are not then there is no unique
solution since those that are redundant can be discarded leaving more
unknowns than equations.
For example, 2
2xl + 2x2 = 4
These two equations are obviously identical and are linearly dependent.
The value of x 1 can only be determined in terms of x 2 •
The rank of a matrix is defined as the number of linearly indepen-
dent rows (or columns) in the matrix. If two equations in two unknows
are linearly independent then the rank of the coefficients matrix is 2.
Therefore to check whether a set of equations is independent, the rank
of the coefficients matrix could be found. If this equals the number of
unknowns then the equations are linearly independent.
There are two ways of checking the rank of a matrix. First, for an
(n x n) square matrix, A, the rank is n if IAI =I= 0, while if IAI = 0 the
rank is n-1 if a non-zero determinant of size n-1 can be found within
A. Secondly, the row operations used in the Gaussian elimination
method can be applied to the matrix and the rank is the size of the unit
matrix that remains in the final matrix.
For example, consider the equations
X + 2y + Z = 6
2x+y+z=5
4x + 5y + 3z = 17
Using the Gaussian elimination method first, the coefficients matrix is
~
2
1
5 l
and taking twice row 1 from row 2, and four times row 1 from row three
gives
[01 -32 -11]

0 -3 -1
Subtracting row 2 from row 3 and dividing row 2 by -3,
~ lll3]
Subtracting twice row 2 from row 1
01 1/3
113]
[~ 0 0
No further reductions are possible and so the rank is 2, since the final
matrix includes a (2 x 2) identity matrix. Alternatively, the determinant
of the coefficients matrix is
1 2 1 1 2 1
IAI = 2 1 1 0 -3 -1 = -3 -1 I = o
4 5 3 0 -3 -1 1
-3 -1
and so the rank is less than 3 and the equations are not linearly indepen-
dent. Therefore no unique solution for x, y and z is possible. To see if
the rank is 2, try to find any non-zero (2 x 2) determinant within A.
Since
1 2
= 1- 4 = -3
2 1
is non-zero the rank is 2. Here the first (2 x 2) determinant selected
turned out to be non-zero. If it had been zero, another attempt at
finding a non-zero (2 x 2) determinant would have been made. While
this is not too arduous in this example, for larger matrices it can be
time-consuming and in these circumstances the Gaussian elimination
method is preferred.
For a second example consider the equations
XI + 2x2 + 2x3 7
4
X1 + X2 + 3X3 = 6
The matrix of the coefficients is
[
Subtracting row 1 from row 2 and from row 3,
: -1
2 2]
1
1 3
[001 -3-12 -12]1

I
Adding twice row 3 to row 1, and row 3 to row 2,
1 o 4
[ 0 -4 0
0 -1 1
Dividing row 2 by -4 and adding the new row 2 to row 3,
[~ ~I
0
1
0
Subtracting four times row 3 from row 1 gives the (3 x 3) unit matrix
and so the rank is 3 and the equations are linearly independent. Alterna-
tively, the value of IAI is
1 2 2 3 2 4
IAI = 1 -1 1 0 -1 0 =
1 1 3 2 1 4
and as this is non-zero the rank is 3.
Our discussion of rank has been concerned with simultaneous equa-
tions and square matrices. However, the concept of rank is more gener-
al and can be applied to matrices of any size. For example, if
1 -2 0 3 5
A= 2 1 0 2 4
0 -6 2 0 0
-2 2 1 -1 4
the maximum possible rank is 4, since there are no (5 X 5) determinants
within A. As mentioned above in the discussion of our first example,
a search could now be made for a (4 x 4) determinant which is non-
zero. However, there are 5 possible (4 x 4) determinants and it would
be tedious to evaluate a number of these. Instead, the Gaussian elimina-
tion method is used. Combining a multiple of row 1 with each of the
others gives
1 -2 0 3 5
0 5 0 -4 -6
0 -6 2 0 0
0 -2 1 5 14
Subtracting row 4 from row 1, adding row 3 to row 2, and dividing row 3
by 2,
1 0 -1 -2 -9
0 -1 2 -4 -6
0 -3 1 0 0
0 -2 1 5 14
Combining row 2 with row 3, and row 2 with row 4,
1 0 -1 -2 -9
0 -1 2 -4 -6
0 0 -5 12 18
0 0 -3 13 26
Multiplying row 2 by -1 and row 3 by -0.2,
1 0 -1 -2 -9
0 1 -2 4 6
0 0 1 -2.4 -3.6
0 0 -3 13 26
Combining multiples of row 3 with each of the other rows,
1 0 0 -4.4 -12.6
0 1 0 -0.8 -1.2
0 0 1 -2.4 -3.6
0 0 0 5.8 15.2
Dividing row 4 by 5.8 and combining multiples of the new row with the
other rows,
1 0 0 0 -1.07
0 1 0 0 0.90
0 0 1 0 2.69
0 0 0 1 2.62
and the rank is 4.
2.9 Exercises
1. Determine the rank of the following matrices using either determi-
nants or the Gaussian elimination method:
(a) (b)
[; ~ i
1 -2 -1
l (c)
[ 21 -12-1-4]
3 1 -5
I I jl
(d) 1 0 0 (e) 1 1
1 1 1 1 0
0 1 1 2 2
(f) 1
0
2
0
2
1
0
2
3 -: I I I (g) 1
1
0
0
-1 -1
1 -1 -1
1 -1 -1
2. Test whether the following equations have a unique solution.

(a) x, + 3xz - 2x3 = 3 (b) x, + 2x2 + x3 = 2
x, - 2x2 + x3 = 0 2x, Xz + 2x3 = -6
2x, - 2xz + 3x3 = 3 x, + Xz + x3 = 0
(c) x, + 3x2 - X3 + 2x4 = 2
x, Xz + 2x3 + x4 = 3
x, + Xz + X3 - 2x4 = 3
x, + Xz + 5x4 = 1
2.10 Further properties of matrices

For general matrices A, B and C, the following properties of the
transform operation can be shown to be valid
(A')' = A
(A+ B)' A'+ B'
(AB)' = B'A'
(ABC)' C'B' A'
assuming that the various matrices are conformable for addition or
multiplication. A special case occurs when
A'= A
since this requires that A must be a square matrix, because if A is
(m X n), A' is (n x m) so m = n. Now if the transpose of A equals A,
then A is a symmetric matrix. For example, if
~~ ! 11
then A = A' and A is symmetric. Another example of a symmetric

matrix is the unit matrix /. More generally, if
then for A to be symmetric, a;i = ai; for all i, j. If a symmetric matrix A

has the additional property that
A2 =A
so that multiplication by itself leaves A unchanged then A is a symmetric
idempotent matrix. The unit matrix is a symmetric idempotent matrix, as
is the matrix
0
0
0
A useful application of symmetric matrices occurs in statistics and

econometrics. If x is (n X 1) and x' = [x 1 x 2 • • • xn]
then x'x = [x 1 x 2 .. · x]
n [ x 1 = x 21 + x 22 + · · · + x n2
Xz
Xn
so that x'x is the sum of the squares of the elements of x. A weighted

sum of squares of xis obtained by forming x'Ax where A is a (n X n)
diagonal matrix (i.e. all a;i = 0 fori -:/= j). For example, if x' has three
elements,
~ [x, x, x,] [ au xi
azzXz
a33X3
l = ~ + ~ + a33xi
If A is generalised to be any symmetric matrix, expressions of the type

x' Ax are known as quadratic forms. For the (3 x 3) case,
[x, x, x,] [ allxl +

a12X1 +
al3xl +
a12x2 +
a22 X2 +
az3Xz +
a13x3
a23 X3
a33X3
l
allxi + ~ + ~ + 2a 12 X1X2 + 2a 13 X1X3
+ 2a23 X2X3
Further properties of the inverse operation are that

(AB}- 1 = B- 1A- 1
(ABC)- 1 = C- 1B- 1A- 1
(A')-1 = (A-1)'
where here A, B and C are all square matrices of the same size and are
non-singular so that their inverses exist. In some applications, it is useful
to partition a matrix into sub-matrices. This can be done in several ways
and is achieved by effectively adding vertical and horizontal lines to the
matrix. For example,
all al2 al3 al4

A= az1 azz az3 az4 = [ A11 A12]
a3l a3z a33 a34 A21 Azz
••••• 0. 0 •• 0 ••
a4l a4z a43 a44
where All is (3 x 3), A12 is (3 x 1), A21 is (1 x 3) and A22 is (1 x 1). If two
matrices are partitioned they can be added or multiplied only if they are
partitioned in a conformable way. Thus for addition, each of the corres-
ponding sub-matrices must have the same dimensions, so that if
then A + B can be formed only if A;i and B;i are the same sizes:
A +B = [ Au + Bu A12 + B12 ]
Azt + Bzt Azz + Bzz
For multiplication, the number of columns in A;i must equal the number
of rows in Bii:
AB = [AuBu + A12Bz1 AuBtz + A12B22]
AztBu + AzzBzt A21B12 + Az2B22

The inverse of a partitioned matrix can be found from the above result.
Since AA- 1 = I, B is the inverse of A if AB = I. Partitioning I so that it
conforms to the partitioning of the product AB,
and equating the elements of AB gives four equations in the unknowns

B11 , B12 , B21 and B22 :
I = AuBu + A12B21
0= A11 B12 + A12B22
0 = AztBu + AzzBzt
I = Az1B12 + AzzBzz
Solving these equations gives
Bu = (Au - At2Azz- 1Azt)- 1
B12 = -Bu (A12Azz- 1)
Bzt = -A22- 1A21Bu
B22 = A22- 1 + (A22- 1A2tBuA12A22- 1)
where it is assumed that Au and A22 are square and have an inverse, and
that the multiplied matrices are conformable.
The final topic we consider in this section is the characteristic
equation of a square matrix. For a non-singular matrix, A, the matrix
[A - rl] is called the characteristic matrix of A, where r is a scalar. For
example, if
A ::..= [ 5 3] then A - rl
3 5
= [5
3 5
3]- [ r 0
0 r
l= [ 5-r
3 5-r
3]
Setting the determinant of the characteristic matrix to zero gives the

characteristic equation. Continuing the example,
5-r 3 = (5-r)(5-r) -9 = 16 -lOr + r2 = 0
3 5-r
Here the characteristic equation is a quadratic and using the standard

formula for a quadratic equation (see section 3.3 for a full discussion),
the two values of r which satisfy the equation are r = 8 and r = 2. These
are referred to as the characteristic roots, latent roots or eigenvalues of A.
In general, for an (n x n) matrix, the characteristic equation will have
n roots.
Associated with each characteristic root, r, of A is a vector, x, say,
which satisfies
Ax = rx or (A - rl)x = 0
This vector is called the characteristic vector, latent vector or eigenvector
corresponding to this particular root. In the example, for r = 8, the
equation is
3
5-8
l[ l l
XI
X2
=[0
0
or 0 or X 1 = X2
0 or X1 = X2
and so even though there are two equations in two unknowns there is no
unique solution. At this point it is convenient to normalise the sum of
the squares of the elements of x by making this equal 1. That is,
xi+ ~ =1
and with X 1 = X 2, the solution is X 1 = 1/Y2 = x 2, and so the characteristic
vector is
x' = [11v'2 11v'2]
For the other characteristic root, r = 2, the equation is
or 3x 1 + 3x2 =0 or x 1 = -x2
3x 1 + 3x 2 = 0 or X1 = -X 2
and again normalising by putting

xi+ x; = 1
allows the solutions X 1 = 1/V2 and x 2 = -1/V2 to be found and the
characteristic vector is
x' = [11V2 -11V2]
Earlier, a matrix, A, was defined to be symmetric idempotent if
A2 =A
For any matrix the characteristic vector satisfies
Ax= rx
and if A is symmetric idempotent, multiplying both sides by A gives
A2x = rAx
=Ax
Therefore,
or r(r-l)x=O
Now since x =F 0, r(r-1) = 0 and r = 0 orr= 1. Thus, for any symmetric
idempotent matrix the characteristic roots are either zero or unity.
As an example consider
B = [ 0.5 0.5]
0.5 0.5
which is idempotent since B2 = B.

The characteristic equation is

IB- rii = 0
or 0.5-r 0.5 =0
0.5 0.5-r
and so (0.5-r) 2 - 0.25 = 0 or r 2 - r = 0

Therefore the characteristic roots are r = 1 and r = 0, as expected for a
symmetric idempotent matrix.
To obtain the characteristic vectors, taking r = 1 first,
Bx
l [l
= rx
or (B-rl)x =
-0.5 0.5] [ X 1 0
[
0.5 -0.5 X2 0
This gives - 0.5x 1 + 0.5x 2 = 0 or X 1 = X 2 , and normalising by imposing

the restriction that ~ + ~ = 1 results in X 1 = l!Y2 = X2 •
For the second root, r = 0 the characteristic equation is
[
0.5
0.5
0.5] [ X 1
0.5 X2
0
0
l
which gives X1 = -x2 and x 1 = l!Y2, x 2 = -1Y2.
2.11 Exercises
1. For A= [2
1 0
2] B = [ 3 -1
1 2 -2
1
0
l C = [ ~ ~
-1 0
l
show that (A')' =A, (A+ B)' =A' + B' and (BC)' = C'B'.
2. Determine which of the following matrices are idempotent.
0 0 0.25 0.25 0.25 0.25
~~~ 1 0
B=
0.25 0.25 0.25 0.25
c = [ ~ ~
0 0 0.25 0.25 0.25 0.25
0.25 0.25 0.25 0.25
3. If A ~ [ : -: l
and B ~ [ _: : l
show that (A')' ~ (A')'
and that (AB)- 1 = B- 1A- 1 •

4. Determine the characteristic roots and vectors of
A ~ [: : l
and B ~ [: -: l
2.12 Input-output analysis
INTRODUCTION
In any economy there are a number of industries supplying con-
sumer demand. Many of these industries also supply intermediate
products which are further processed or utilised by other industries
before reaching the final consumer. For example, the glass industry
supplies finished products such as mirrors, but also supplies window
glass to the building industry to be used in the construction of houses,
and toughened glass to the motor industry to be used as windscreens for
cars.
The industrial sector of any economy, therefore, consists of a
number of interlinked units from which final consumer demand is met.
Changes in demand for any one of the final outputs affect the outputs
required from many other industries. In order to see the effects that
such changes might produce it is necessary to build a model of the
system. This was first attempted by Leontief and the theory has been
developed under the name input-output analysis.
INPUT, OUTPUT AND DEMAND

In building the model Leontief made an important assumption: that
the output of one industry which is required to satisfy the demand from
another industry is directly proportional to the final output of the latter.
This can be written as
X;i = ~
where Xi is the total output of industry j,
aii is a constant of proportionality, and is the proportion of

output of industry i that will be required by industry j in order
to produce one unit of output. This will depend upon the
technology of industry j,
xii is the output of industry i that will be required by industry j.
For example, let us assume then that the average number of units of
glass which is required in the construction of a house is equal to 2. Then
this will be the constant of proportionality and the relationship can be
expressed as
X;i = ~ = ~
where Xi is the output of the building industry in terms of the number of
house units constructed, and X;i is the output of the glass industry which
is required by the building industry for house construction. These
figures are, in fact, flows and represent the output per unit of time, e.g.
output per year.
If the technology of the industry changes and houses are con-
structed with much larger expanses of glass or the average size of a
house unit increases, then a new constant of proportionality must be
used: for example,
X;j = 3Xi
The constant of proportionality used in this example is an average

for all types of houses. It may be possible, and it would certainly be
useful, to classify the different class of dwellings by size and by type, e.g.
bungalows, semi-detached, detached, etc. and to use a separate con-
stant of proportionality for each sub-category.
An industry may also use some of its own output for further
processing or as raw material input and as this demand must be met, it
must be included as
xii = ~
In this case the constant of proportionality must be less than one,
otherwise the industry would require a greater amount of input of its
own output than it was capable of producing. (It is possible to consider
the system in terms of net outputs only and to ignore these inputs
altogether, but this is not done here.)
The total output of any industry depends upon the demands made
upon it by all other industries and also upon the demand for its products
made directly by the consumer. This latter, the final demand, is a
function of consumers' preferences, relative prices and consumers' in-
come, and also varies with time.
TABLE 2.1
Input to
Level
Industry I Industry 2 of output
Industry 1 X 11 = 200 x 12 = 400 XI= 1,600

Industry 2 x 21 = 600 X 22 = 100 X 2 = 2,700
For any economy, it is possible to collect data for the flow of goods
and services between industries and to specify the final demand for the
outputs at any given point in time. In the simple case of a two-industry
economy the information relating to the flow of goods which is required
to produce a particular level of output might be as presented in
Table 2.1.
From this data it is possible to establish the technological coef-
ficients a;j using the formula quoted earlier:
X;j
therefore a··=-
IJ x.I
From the information given in Table 2.1 it is also possible to

determine the final demand which can be supplied. This is given by the
level of output less the amounts which are required to meet the demands
made by the two industries in producing these outputs. Thus, level of
final demand which can be met by Industry 1 is equal to
1600 - (200 + 400) = 1000 units
and for Industry 2 it is equal to
2700 - (600 + 100) = 2000 units
From the information given in Table 2.2 it is possible to determine the
final demand which can be met at any other level of output within the
range over which the constants of proportionality can be assumed to
hold. For example, the information in Table 2.2 tells us that Industry 1
required 0.125 units of its own output to be used as input for each
further unit of output. Also Industry 2 requires 0.148 units of the output
of Industry 1 for each unit of its own output.
It follows that, in general, if X 1 and X 2 represent the total output of
TABLE 2.2
Technological coefficient for

Industry 1 Industry 2
Industry 1 200 = 0.125 400 = 0.148

1,600 2,700
Industry 2 600 = 0.375 _!QQ_ = 0.037

1,600 2,700
Industries 1 and 2 respectively and C1 and C2 are the final demands for
the outputs of the industries respectively we can write
0.125 xl + 0.148 x2 + cl ~ xl
and
The first equation states that the total demand for the product of
Industry 1 must be less than or equal to the total output of Industry 1.
The second equation states the same thing for Industry 2.
If we assume that the total output of each industry is just sufficient
to meet all the demands made upon it the inequalities can be replaced
by equations as follows:
o.12s X 1+ 0.148 x2 + c2 = xl
0.375 xl + o.037 x2 + c2 = x2
These equations can then be rearranged
xl - o.125xl - 0.148 x2 = cl
x2 - 0.375 xl - o.037 x2 = c2
that is (1 - 0.125) xl - 0.148 x2 = cl
and - o.375 xl + (1 - o.o37) x2 = c2
A pair of simultaneous equations such as these can be conveniently
written in matrix notation, as was shown in the earlier part of this
chapter.
[
(1 - 0.125)
-0.375
-0.148
(1 - o.037) x2 = c2
l[XI] [
C1 ]
The condition that this set of equations should have a unique

solution in terms of xl and x2 is that the determinant of the matrix of
coefficients is not equal to zero.
In this particular type of problem it is necessary to apply the further
condition that the values for X 1 and X 2 , the output of the industries,
must not be negative. This condition is satisfied if the value of the
determinant is positive. That is,
(1 - 0.125) -0.148
>0
-0.375 (1 - 0.037)
If this condition holds it is possible to use the equations to deter-

mine either
(a) the final demands which can be met when the output of each
industry is fully utilised, i.e. C1 and C2 given X 1 and X 2 , or
(b) the total outputs which are required to meet a given level of
final demand, i.e. xl and x2 given cl and c2.
For example, if the total output of the two industries is 1,600 and
2,700 units respectively, the maximum final demand can be obtained
l l l
from the following relationships:
[
0.125 0.148] [ 1600 + [ C1 = [ 1600
o.375 o.o37 2100 C2 2100
or [ell=
c2
[1-0.125 -0.148][ 1600]
- o.375 1 - o.037 2100
0.875 X 1600 - 0.148 X 2700
[
= -0.375 X 1600 + 0.963 X 2700
=[ 1400 - 400
- 600 + 2600
l l
= [ 1000
2000
therefore cl = 1,000 and c2 = 2,000 which agrees with the information

given in Table 2.1.
Let us now assume that the level of final demand for the output of
the two industries is reversed, i.e. cl = 2,000 units and c2 = 1,000 units.
Then we can calculate the levels of output which would be required
to meet these final demands from the following relationship:
I 0.875
-0.375
-0.148l!X1 ] = 12000 l
o.963 X2 1000
1 I
The inverse of the coefficient matrix is
0.963 0.148]
0. 7871 0.375 0.875
and hence !
XI]
X2
1 I
0.963 0.148][2000 l
= o. 7871 0.375 o.875 1000
12635]
= 2064
i.e. the required output of Industry 1 is 2,635 units and of Industry 2 is
2,064 units (approx.).
2.13 The input-output matrix

In general the technology matrix for a two-sector economy can be
written as
or simply as A. This matrix is known as the input-output matrix. To

satisfy final demand exactly from current output the following equality
must be true:
or AX+C=X
These equations can be rearranged in the form

or (I- A)X =C
where I is the unit matrix and (I - A) is known as the Leontief matrix.
For an economy which is divided into n sectors
au au . . aln
a2, a22 · . a2n
A=
an! an2. . ann
(1 - a11 ) -au
- a2, (1 - a22)
(I- A)=
The output of each industry which is required to satisfy a given final

demand from current production can be found by inverting this matrix.
Since (I- A)X =C

X = (I - A)- 1 C,
and given the technology matrix A, along with either C or X, the other
can be found.
The interpretation of this final equation is helped by letting
(I - At'= B so that, for example, taking row i,
X; = B; C + B; C + B;3C3 + · · · B;nCn
1 1 2 2
and so if C 1 increases by 1, and the other Ci values are unchanged, X;

will increase by B;1 • That is, Bil measures the effect that a unit increase in
final demand for the output of industry 1 has on the gross output of
industry i. But an increase in C 1 also affects the gross output of each of
the other industries, and so the total effect on the economy of an
increase of 1 in C 1 is found by adding together B11 , B21 , • • • , Bnt· Let the
result be
TABLE 2.3
Absorbing
sector Final
1 2 3 4 demand
Producing
sector
1 0.269 0.219 0.246 0.062 3,662

2 0.010 0.239 0.200 0.076 1,259
3 0.008 0.006 0.051 0.230 1,248
4 0.015 0.154 0.074 0.720 4,789
which is called the output multiplier for industry 1. For example, from
the end of section 2.12, the B matrix is
( ) I 1 [ 0.963 0.148] [ 1.223 0.188]

B = I- A - = 0.7871 0.375 0.875 = 0.476 1.112
so that M1 = 1.223 + 0.476 = 1.699
M2 = 0.188 + 1.112 = 1.300
Here industry 1 has the larger output multiplier and an increase of 1 in
the final demand for the output of industry 1 leads to an increase of
1.699 in total output.
Taking another example, Table 2.3 gives the technology matrix for
a four-sector economy.
From this matrix,
(1-0.269) -0.219 -0.246 -0.062
-0.010 (1-0.239) -0.200 -0.076
(I- A)=
-0.008 -0.006 (1-0.051) -0.230
-0.015 -0.154 -0.074 (1-0.720)
1.401 0.597 0.562 0.934
0.039 1.474 0.377 0.718
and
0.038 0.233 1.195 1.053
0.107 0.904 0.553 4.295
The output of each industry which is required to satisfy the final
demand is found by multiplying the inverse matrix by the given final
demand vector.
1.401 0.597 0.562 0.934 3662

0.039 1.474 0.377 0.718 1259
X= (I- At 1C =
0.038 0.233 1.195 1.053 1248
0.107 0.904 0.553 4.295 4789
11056
5908
6967
22789
:. X 1 = 11056, X2 = 5908, X3 = 6967, X4 = 22789
Here the output multipliers, found by adding the columns in
(I- At 1 are 1.585, 3.208, 2.687 and 7.000, showing that an increase in
the final demand for the output of industry 4 will have a large impact on
gross output.
If the number of sectors, n, is large it can be awkward obtaining the
inverse of the Leontief matrix. In these circumstances it may be conve-
nient to use the fact that
(I - A)(I + A + A2 + A3 + A4 + A5 + · · · + Am)
= I(I +A+ · · · +Am) -A(I +A+ · · · +Am)
= (I + A + · · · + Am) - (A + A2 + · · · + Am+l)
=I- Am+l
Now if the product had resulted in the unit matrix then the expression
(A+ A2 + · · · +Am) would be the inverse of I- A. But the presence of
Am+l prevents this. However, since every element of A is positive and
less than 1 - and, more importantly, the total of each column of A adds
up to less than 1 -then as m increases, Am+l tends to the null matrix.
Therefore, approximately, (I- At 1 =I+ A+ A2 + A3 + A4 + · · · +
Am and the size of the last term, Am, is a guide to how close the
approximation is.
As an example, from the technology matrix in Table 2.2,
0.125 0.148]
[
A = 0.375 0.037
I 1 [ 0.963 0.148] [ 1.223 0.188]

and (I- At = 0.7871 0.375 0.875 = 0.476 1.112
is the accurate inverse. Here, an approximation is given by

1.2129 0.1870 [
I + A + A2 + A3 + A4 = [
0.4547 1.1108
and to check how accurate this is the product of I - A and the approxi-
mate inverse is
0.9940 0.0008 [
(I- A) (I+ A+ · · · + A4 ) = [
0.0170 0.9996
which is reasonably close to I. The last term included in the approximate
inverse is
A4 = [ 0.0065 0.0031 [
0.0078 0.0047
and a closer approximation can be found by including A5 and further
terms. Here the accurate output multipliers are 1.699 and 1.300 while
the approximate ones are 1.668 and 1.298.
The conditions that must be satisfied in order that at least one set of
final demands can be met from any given set of output levels are
(a) (1 -all) - a,2
- a2, (1- a22)
>0
and
(b) all the elements on the main diagonal of the matrix, namely
(1 -all), (1 - a 22 ), ... , (1 -ann), are positive; that is,
(1 - all) > 0, ... , (1 - au) > 0, ... , (1 - ann) > 0
These are known as the Hawkins-Simon conditions.

One way of extending basic input-output analysis, which is also
useful in regional economics, is to disaggregate final demand (C) into
components such as household consumption, government spending,
exports and investment. At the same time, households, government and
imports can be included as additional sources of inputs. In Table 2.4 an
TABLE 2.4
Outputs Inputs Final

produced used by demand Total
by Ind. I Ind. 2 Households Govt Exports output
Industry 1 200 400 400 200 400 1,600
Industry 2 600 100 600 300 1,100 2,700
Payments for
Households 400 800 0 0 0 1,200
Government 100 400 0 0 0 500
Imports 300 1,000 0 0 0 1,300
Total
input 1,600 2,700 1,000 500 1,500 7,300
extended input-output table is given for a simple two-sector economy.

The inter-industry matrix is the same as in Table 2.1 but extra rows are
included to indicate payments to households (for supplying labour and
as profits on investments), to the government (taxes to meet the costs of
services) and for imports (for raw materials). Extra columns are in-
cluded to allow final demand to be split between consumption by house-
holds, spending by the government, and exports of finished goods. The
addition of the extra rows results in the total output of each of the
industries equalling the total input. In this example, the country exports
more than it imports and the government account is balanced.
The technology matrix can also be extended to include this extra
information. Thus, households provide 400 units out of the total inputs
of 1,600 for agriculture and so the technological coefficient is 400/1,600
or 0.25. The extended matrix is given in Table 2.5 and, because it is not
square, it needs to be used with care. The first two rows and columns
give the original A matrix, so that the formulae C = (I - A)X and X =
(I - A)- 1C can be used as previously. Now, at the end of section 2.12,
we saw that the effect of changing the final demands to C 1 = 2,000 and
c2 = 1,000 is to require XI = 2,635 and x2 = 2,064. The payments to
households will now be, from Table 2.5,
0.2500X 1 + 0.2963X 2 = 1270.31,
and tax payments will be
0.0625X 1 + 0.1481X 2 = 470.37,
TABLE 2.5
Inputs
Output required by
produced by Ind. 1 Ind. 2
Industry 1 0.1250 0.1481

Industry 2 0.3750 0.0370
Payments for
Households 0.2500 0.2963
Government 0.0625 0.1481
Imports 0.1875 0.3704
and the value of imports will be

0.1875X 1 + 0.3704X 2 = 1258.6
These calculations assume that the payments for households, taxes and
imports are exogenous so that there is no feedback from, say, increased
household income, to consumption by households. This is rather arti-
ficial and it might be more sensible to assume that some of the increased
income will be consumed. This can be achieved by treating the house-
hold sector as an industry rather than as a component of final demand so
that it is assumed to require inputs in fixed proportions. The (2 x 2) A
matrix of Table 2.5 now becomes (3 x 3) and has an extra column,
consisting of 400/1,200, 600/1,200, and 011,200, as well as the extra row
0.2500, 0.2963 and 0. The result is that
0.1250 0.1481 0.3333]

A* = [ 0.3750 0.0370 0.5000
0.2500 0.2963 0.0000
1.5620 0.4732 0.7572]
and (I- A*t 1 = [ 0.9585 1.5176 1.0783
0.6745 0.5680 1.5088
The new output multipliers are 3.1950, 2.5588 and 3.3443 compared
with the previous values of 1.699 and 1.300. The effect of including
households as an industry increases the output multipliers substantially.
It is possible to approach the problem slightly differently and to
consider the total available capacity of each industry. Let X; represent
the maximum output which can be obtained from industry i. Then we
can calculate the value of the vector C, (the maximum final demands for
each industry), which could be met if every industry produced at its
maximum level. This would be obtained from the relationship
C =(I- A)X
However, we often find that the result is a C-vector containing one
or more negative elements. These would correspond to a negative final
demand which is impossible and means that the economy would not
operate with all its industries at their maximum output level.
The problem now becomes one of deciding which industries should
operate at less than their maximum output, i.e. to decide where the
'slack' capacity should be. The latter can only be decided in terms of
some specified objective. For example, let C1 be the final demand for
the output of industry 1 which is to be satisfied at a price p 1 and in
general let en be the final demand for the output of industry n which is
to be satisfied at a price Pn· One objective may be to maximise the value
of final demand and this may be stated as:
Max n
Z = CtPt + CzPz + · · · + CnPn = ~ C;p;
i=l
subject to the following constraints on the system
allx1 + a12X2 + ... + alnxn + cl ~ XI
aztXt + azzXz + ... + aznXn + Cz ~ Xz
or ~
This is, in fact, a linear-programming type of problem and the
method of arriving at an optimum solution in such situations is discussed
in Chapter 8.
2.14 Exercises
1. Determine the maximum final demand which can be met in the
following situation shown in Table 2.6.
2. What final demand can be met when the level of output of Industry
1 is increased to 2,000 units in a situation which is in all other
respects identical to that given in Question 1?
3. Determine the level of output which is necessary to meet final
demands of 1,000 and 2,000 respectively when the technological
coefficients are given by the following matrices:
TABLE 2.6
Input to Level of
Industry 1 Industry 2 output
Industry 1 200 300 1,500

Industry 2 500 100 2,500
TABLE 2.7
Inputs Final
Output required by demand
produced by Agriculture Industry Households Exports
Agriculture 100 200 300 400

Industry 500 100 600 800
Households 100 800 0 0
Imports 300 900 0 0
Total input 1,000 2,000 900 1,200
(a) 0.2 0.41 (b) 0.1 0.61

[ [
0.3 0.2 0.4 0.1
4. The economy of a region can be summarised by the information in
Table 2.7.
(a) Treating households as exogenous, use the (2 x 2) technology
matrix to find the output multipliers for agriculture and indus-
try. What levels of gross output would be needed to meet a
final demand for exports of 500 for agriculture and 900 for
industry?
(b) Treating households as an endogenous sector, use the (3 x 3)
technology matrix to find the output multipliers for agriculture
and industry, and compare them with the results in part (a).
What levels of gross output would be needed to meet a final
demand for exports of 500 for agriculture and 900 for industry?
5. The summary input-output table for the USA in 1947 (in billions
of dollars) is shown in Table 2.8.
(a) Determine the technology matrix and show that the inverse of
the Leontief matrix is
TABLE 2.8
Input
Output to Final Total
from Agriculture Industry Services demand output
Agriculture 11 19 1 10 41
Industry 5 89 40 106 240
Services 5 37 37 106 185
1.409 0.192 0.062]

[ 0.372 1.753 0.476
0.286 0.367 1.351
(b) What gross outputs are needed to satisfy final demands of
15 for agriculture, 120 for industry and 130 for services?
(c) Show that if 'industry' and 'services' are combined into a single
sector the result in (b) is basically unchanged. What are the
implications of this for an economist who is interested only in
the agriculture sector?
L If ~[~ i -l] B ~[ ~ j -i l ~ C [ ! -n
form the following (if they exist): A- B, B- C, AB, BC, AB' and
show that AB -:/= BA.
2. Find the inverse of A =[ -i ~ -; ]

0 3 1
and B =[ ~~~
a 1 a
]
3. Write the following equations in matrix form and solve them by

matrix inversion:
(a) 3x- 2y = 4 (b) X+ y + Z = 3
2x+ y=5 2x + 3y- z = 2
3x + 2y + z = 3
4. Determine the rank of the matrices given in Questions 1 and 2
above.
TABLE 2.9
Input
Output to Final
from Agriculture Industry Services demand
Industry 250 200 100 150
Services 100 300 100 100
5. Use Gaussian elimination to determine the rank of the following

matrices:
1 -1 2 -1 -2 3 5
2 10 2 1 0 0 4
A= 3 11 B= -1 0 1 1 3
3 3 -1 3 0 -1 4 12
6. Check whether the following equations have a unique solution:
(a) 2x - 4y - 3z = 25 (b) 5x+ 3y + z = 12
6x - 2y + 4z = 118 x- y + 3z = 14
3x + 3y + z = 75 4x + 2y + 2z = 16
7. If x' = [1 2 2 3] and A = 2 -1 0 1
1 2 -1 0
0 1 2 -1
1 0 1 2
evaluate the quadratic form x' Ax.
8. By partitioning the matrix M after column 2 and row 2 so that it
consists of four (2 x 2) sub-matrices, check that the formulae
given in the text for the inverse of a partitioned matrix is correct.
a 0 1 0
0 a 0 1
M= 0 1 0 a
1 0 a 0
9. Determine the characteristic roots and vectors of A and B in
Question 2 above.
10. A simple economy has three industries: agriculture, industry and
services, and the input-output table is shown in Table 2.9. What
should the total output for each sector be if final demand changes
to 100 for agriculture, 200 for industry and 150 for services?
Chapter 3
Non-linear Equations
3.1 The quadratic
In Chapter 1 the relationship between two variables y and x was
assumed to be linear and could be written
y =a+ bx
Here a is the intercept term, or the value of y when x = 0, and b is the
slope or gradient, since if x increases by 1, y increases by b. For a linear
function the slope is a constant. If the equation represents a cost
function, with x as the level of output (where x > 0) andy as the total
cost, then b is the marginal cost of an extra unit of output. Here it is
assumed that output and costs vary continuously so that the equation is
valid for all positive values of x. We saw that the graph of such an
equation is a straight line.
There are many applications of algebra in economics and business
studies where the assumption of a (possibly approximate) straight-line
relationship is reasonably adequate. These include demand and supply
functions, some cost functions, consumption functions and production
functions. Also, in section 5.15 below, we will see that under very
general conditions MacLaurin's theorem allows a non-linear function to
be represented as a first approximation by a linear function.
However, there are also many instances where it is clear that the
relationship is non-linear and that a linear approximation will be mis-
leading. For example, suppose that the demand curve relating quantity
demanded, q, and price, p, is linear, so that
p =a - bq a > 0, b > 0
then revenue from selling the quantity q is
R = pq = (a - bq )q = aq - bq 2
which is a quadratic function. In this case the values of R for different
values of q are
92
Non-linear Equations 93
q = 1, R = a-
1 b
q = 2, R 2 = 2a - 4b
q = 3, R 3 = 3a - 9b
and while the change in revenue in going from q = 1 to q = 2 is (a- 3b),
that in going from q = 2 to q = 3 is (a- 5b). Here the marginal revenue
changes with each value of q, while with a linear function it would be a
constant.
The general quadratic function can be written
y = ax2 + bx + c
where a, b and c are constants, and a =I= 0. This is also known as a
second-order polynomial in x because the highest power to which x is
raised is 2. The effect of the squared term is to make the graph curved
instead of being a straight line. For example, consider the quadratic function
y=6x2+2x+3
This might be a cost function with x being the level of output andy being
the total cost associated with the value of x. In order to plot the graph of
this function several points which lie on the graph are needed. Since x is
the level of output we will assume that xis positive, that total cost varies
continuously with x and that the values of interest are for x = 0 to x = 5.
Let x = 0, then substituting into the equation gives
y = 6(0)2 + 2(0) + 3 = 3,
while x = 1 gives
y = 6(1)2 + 2(1) + 3 = 11,
and x = 2 gives
y = 6(2)2 + 2(2) + 3 = 31,
and x = 3 gives
y = 6(3) 2 + 2(3) + 3 = 63,
and x = 4 gives
y = 6(4) 2 + 2(4) + 3 = 107
and x = 5 gives
y = 6(5) 2 + 2(5) + 3 = 163
3 4
Fig. 3.1
Collecting these values into a table results in
X 0 1 2 3 4 5
y 3 11 31 63 107 163
These values are plotted in Fig. 3.1 and it can be seen that as x increases,
so does y, and the rate of increase of y - that is, the gradient - also
increases. The result is that the graph curves away from the x-axis. This
is because in this example the coefficient on x 2 is positive. If the value of

a is negative, the graph will curve towards the x-axis.
For example, let
y = 15 + lOx - x 2
the values of y for different values of x can be easily obtained by
including each term of the quadratic as follows:
X 0 1 2 3 4 5 6 7 8 9 10
constant 15 15 15 15 15 15 15 15 15 15 15
lOx 0 10 20 30 40 50 60 70 80 90 100
-x2 0 -1 -4 -9 -16 -25 -36 -49 -64 -81 -100
total = y 15 24 31 36 39 40 39 36 31 24 15
The graph is shown in Fig. 3.2. As x increases, y increases, initially

by large amounts, but these are decreasing with the result that the value
x = 5 gives the maximum value of y, and as x increases beyond 5 the
graph falls towards the x-axis. The shape of this graph is known as a
parabola and this is the typical shape of the quadratic function. In the
previous example, graphed in Fig. 3.1, since xis the level of output only
positive values of x are plotted. However, if the graph is extended in the
negative direction, a parabola is also obtained but this has a minimum
value (at x = 0) rather than a maximum. In general if the coefficient of x 2
is positive (as in Fig. 3.1) there will be a minimum, while if the coef-
ficient is negative (as in Fig. 3.2) there will be a maximum.
The equation of a straight line can be determined if two points
which lie on the line are known, since there are two unknown para-
meters, the intercept term and the slope coefficient. For a quadratic, as
there are three parameters, a, b and c, the equation can be determined
if three pairs of values of x andy are known. For example, suppose that
the following cost information is known:
Output (hundreds) X 1 2 3
Total cost (£thousands) y 14 20 28
The fact that x is measured in hundreds and y in thousands can be

ignored, as long as this is remembered when the equation is being
2 3 4 5 6 7 8 9 10 X
Fig. 3.2
interpreted. Let the function be

y = ax2 + bx + c
and again it is assumed that costs vary continously so that this equation
is valid for other values of x. Taking the three values of x for which y
values are known,
x = 1 gives 14 = a + b + c
x = 2 gives 20 = 4a + 2b + c
x = 3 gives 28 = 9a + 3b + c
Solving these by Cramer's rule (see section 1.13), let

1 1 1 1 1 1
det = 4 2 1 3 1 0 = 6-8 = -2
9 3 1 8 2 0
then, 14 1 1 14 1 1 12- 14 -2
a= 20 2 1 = 6 1 0 = =-= 1
28 3 1 14 2 0 det -2
det det
1 14 1 1 14 1 42 - 48 -6
b= 4 20 1 3 6 0 = = = 3
9 28 1 8 14 0 det -2
det det
1 1 14 1 1 14 28 - 48 = -20 = 10
c= 4 2 20 2 0 -8 =
9 3 28 6 0 -14 det -2
det det
and therefore, y = x2 + 3x + 10
This answer can easily be checked by substituting x = 1, 2 and 3
into the equation. Given any value of x, say x = 4.5 so that output is
450, the value of y, which here is y = 43.75 or £43,750, can be found.
3.2 Exercises
1. Determine the equations of the quadratic cost functions which fit
the following data. In each case sketch the graphs for the range
X= 0 tO X= 10.
(a) Output 0 2 6
Total cost 4 14 58
(b) Output 4 6 8
Total cost 26 40 58
(c) Output 4 6 8
Total cost 31 37 45
(d) Output 0 5 10
Total cost 20 24.5 30
(e) Output 0 2 4
Total cost 20 22 28
2. Plot the graphs of the following quadratic functions for x = 0 to x

= 10
(a) y = x2+ lOx + 100

(b) y = x 2 - lOx + 100
(c) y = -x2 + lOx + 100
(d) y = -x2 - lOx+ 100
3.3 The roots of a quadratic equation

The graph of the general quadratic function
y = ax 2 + bx +c
cuts the x-axis when y = 0, and for this value
0 = ax2 + bx +c
which is a quadratic equation. The values of x which satisfy this equation
are called the roots of the quadratic or the solutions of the equation.
There are basically three ways of solving this type of equation: by
plotting the graph of the quadratic functioP and finding where y = 0; by
rearranging the coefficients and finding the factors; and by using the
formula which gives the solution. While the first two ways can be useful
in some circumstances, they can also be tedious in that a great deal of
time might be spent obtaining a large number of pairs of values or trying
to find factors before a solution is found. There is also the possibility
that there are no solutions, and in this case the search for solutions is
futile. However, the third way provides a simple formula which can
easily be used and which quickly indicates whether there are any solu-
tions. It is therefore the recommended method for solving a quadratic
equation. To obtain the formula, write the general quadratic equation as
ax2 + bx + c = 0
Dividing throughout by the coefficient of x 2 and rearranging the equation
2 b c
x +-x +- = 0
a a
The terms containing x can be made into a perfect square by adding the
term (b/2a) 2 and if we also subtract this term from the left hand side, the
equation is unchanged;
[ Xz +:X+ (:ar ]+;- (;ar = 0
The terms in the square brackets are equal to (x + b/2a)2
.. ( x + :ar (:ar -;=
b2 - 4ac
4a 2
.. x + :a = ± V( bz ~ )
v'(b 2 4ac)
± ---''-------'
-
2a
Plus and minus signs occur in front of the square root sign. This is
because the square of a negative number is always a positive number,
e.g. ( -2)2 = 4. Therefore the square root of a positive number can be
either a positive or negative number, although its absolute value is
always the same.
-b v'(b 2 - 4ac)
.. x = - ± - - - - -
2a 2a
-b ± v'(b 2 - 4ac)
2a
The two roots of the quadratic equation are
-b + v' (b 2 - 4ac) -b - v' (b 2 - 4ac)
X1 = and
2a 2a
For example, if x 8x + 7 = 0
2 -
then a= 1, b = -8 and c = 7
8 ± v'(64- 28)
Therefore, x=
2
= 4 ± 0.5 v'36 =4 ±3
y
20
-10
Fig. 3.3
and so X 1 = 1 and x 2 = 7. Here there are two distinct roots, and this
corresponds to the case where the graph of
y=x2 -8x+7
cuts the x-axis in two places, when x = 1 and when x = 7. This is shown
in Fig. 3.3 where it can be seen that y has a minimum value when x = 4.
There are also two roots when the graph cuts the x-axis from below, as
when
y = -x2 + 8x- 7
y
4 X
-7
Fig. 3.4
which is shown in Fig. 3.4, where the roots are again x = 1 and x = 7,
and y has a maximum value when x = 4. The more general problem of
determining whether y has a maximum or minimum value is discussed in
detail in Chapter 5.
As a second example consider
x2 - 8x + 16 = 0
and using the formula,
8 ± v(64- 64)
x= =4±0=4
2
4 8 y
Fig. 3.5
so that there are two equal roots.

We obtain equal roots when the term under the square root sign is
equal to zero- that is, when b2 = 4ac. Here the graph of
y = x2 :- 8x + 16
does not cut the x-axis but touches it when x = 4. This is shown in
Fig. 3.5 and the minimum value of y is when x = 4.
As a third example consider
x 2 -8x+20=0
y
20
~ ~
4 8 X
Fig. 3.6
8 ± Y(64- 80)
Here, x= =4±0.5Y(-16)
2
and there is a negative number under the square root sign. This has no
real value and there are no real roots in this case. The graph of
y=x2 -8x+20
is shown in Fig. 3.6 where the value of y reaches a minimum at x = 4 so
that there is no value of x for which y = 0.
However, if we could obtain the square root of negative number,

we could obtain the roots of the equation. Notice that (4)2 = 16 and
(-4)2 = 16 so that V(-16) is not 4 or -4. We now introduce a new
number, denoted by the letter i, which is defined by z"2 = -1 or
i = V (-1). Since there is no 'real' number which when squared gives
-1, is known as an imaginary number.
The properties of i are such that it can be used as an ordinary
number bearing in mind that i 2 = -1, so that i 3 = - i and i 4 = 1.
We can now write -16 = 16 ( -1) = 16F
Hence v( -16) = v(16F) = 4i
Therefore,
4i .
x = 4 ± 2 v (- 16) = 4 ± - = 4 ± 2z
I d
2
and the roots of x 2 - 8x + 20 = 0 are x = 4 + 2i and x = 4 - 2i.
Numbers such as these, consisting of a real and an imaginary part, are
known as complex numbers. They are useful in several branches of
economics, in particular in the solution of differential and difference
equations arising from growth theory. These are discussed in Chapters 9
and 10.
The solutions of a quadratic equation can be summarised as follows:
If ax2 + bx + c = 0
-b ± V(b 2 - 4ac)
then x=
2a
and if b2 > 4ac there are two real and distinct roots
if b2 = 4ac there are two real and coincident roots
if b2 < 4ac there are two complex roots
Before leaving quadratic equations we consider some properties of

the roots. Let r 1 and r 2 be the roots of
ax2 + bx + c = 0
then both (x - r 1) and (x - r 2 ) are factors of the equation and so
(x - r 1) (x - r 2 ) = 0
This can be expanded to give

x2 - (r 1 + r 2) + r1 r 2 = 0
and comparing with
x2 + (b/a)x + (cia) = 0
we see that the sum of the roots,
r1 + r 2 = -b/a
and the product of the roots,

r1 r 2 = cia
These can provide useful information when a, b and c are general
parameters rather than known numerical values. For example, if cia is
negative then it is clear that, if they are real roots, one of r 1 and r 2 is
negative and the other is positive.
3.4 Exercises
1. Use the formula to obtain the roots of the following equations:
(a) x 2 - 7x + 12 = 0 (b) x2 +x- 2= 0
(c) 2x 2 + 7x + 3 = 0 (d) x 2 - 2x + 1 = 0
(e) X2 - 1= 0 (f) x + 1 = 0
2
(g) x 2 - 4x +5 = 0 (h) x 2 + 2x + 2 = 0
2. By finding the roots of the quadratic, plot the graphs of the
following equations over the ranges given:
(a) y = x lOx + 25 for x = 0 to x = 10
2 -
(b) y = - x + 4x - 3 for x = 0 to x = 4.
2
3.5 Other non-linear functions

In discussing the properties of the general quadratic function we
saw that the graph could be plotted by taking different values of x and
finding the associated values of y, and that the points where the graph
cuts the x-axis are the roots of the quadratic equation. Also, the graph is
a parabola which has either a maximum value or a minimum value,
depending on whether the coefficient of r is negative or positive.

Turning now to more general non-linear functions, the same approach
can be used to examine their properties.
Both the linear and the quadratic functions are special cases of the
general polynomial function
where an, an_ 1 , • • • a0 are constants and it is assumed that an is not zero.
Using this notation, for the linear equation, n = 1 and
y =ax + a
1 0
while for the quadratic n = 2 and
The graph of a polynomial function can be obtained as previously

by finding the values of y corresponding to particular values of x.
However, with complicated functions it is also useful to:
(a) find where the graph cuts they-axis and x-axis, and
(b) determine where the maximum and minimum values occur.
Here we will consider (a) and defer a discussion of (b) until Chapter 5.
The graph cuts the y-axis when x = 0 and the corresponding value
of y is easily evaluated. It is more difficult to find where the graph cuts
the x-axis since putting y = 0 results in the polynomial equation
anXn + an-!Xn-! + · · · + a 1X + a0 = 0
The roots of this are required. We saw that the roots of a quadratic
equation could be found by using a formula. There also exist formulae
for the solution of cubic equations (where n = 3) and quartic equations
(where n = 4) (see Further Reading for details of references) but these
are complicated and no formulae apply for n > 4. Instead, an alternative
method of solution is needed. Here we will consider a simple trial-and-
error method and leave a more efficient method to section 5.17.
The general approach is to guess the value of a root, substitute it
into the equation and if the value of the equation is not zero, change the
guess and repeat the process until the value becomes zero.
For example, consider the problem of plotting the graph of
= x3 - 8x2 - 35x + 150
y
for the range x = 0 to x = 12. If x = 0 then y = 150, which is where the
graph cuts they-axis, but this is not a root of the cubic equation. Trying
x = 1, y = 108 and so while this is also not a root, the value of y is
smaller so that increasing the value of x should be a move in the correct
direction- that is, towards the root. For x = 2, y = 56 and for x = 3,
y = 0 so that x = 3 is a root. Therefore (x - 3) is a factor and the
equation can be written
y = x3 - 8x2 - 35x + 150 = (x- 3)Z
where Z is an unknown quadratic expression. But we can divide both
sides of the equation by (x - 3) to give
(x 8x 35x + 150)
3 - 2 -
z = -'-----------
x-3
and this can be simplified by long division. The rules for this are quite
straightforward. First, always write any algebraic expression in the
order of descending powers of x. Second, the first term in the divisor
(that is, x in x - 3) is divided into the first term of the cubic ( x 3 here) and
the result placed above the cubic, as shown below. Then multiply the
(x- 3) by x 2 , placing the result under the cubic, and subtract this from
the equation. Next, move the rest of the equation down. This gives:
x - 31 x3 - 8x 2 - 35x + 150
x3 - 3x2
0 - 5x 2 - 35x + 150
Next, divide x into -5xl, which goes -5x times and so -5x is placed at
the top and multiplied by (x - 3) before subtracting this from the
equation:
8x 2 - 35x + 150
0 - 5x2 - 35x + 150

- 5x2 + 15x
-SOx+ 150
Finally dividing (x - 3) into -SOx + 150 gives -50 and nothing left
over. The result of the division is

x3 - 8x2 - 35x + 150 = (x - 3)(x2 - 5x - 50)
which can be checked by multiplication. Now since y = 0,
x2 - 5x- 50= 0
and solving by the formula gives x = 10 and x = -5. So far we have
established that the graph cuts the x-axis when x = 3, x = 10 and
X= -5.
Since the range of values which are of interest is x = 0 to x = 12, we
complete the following table:
X Q 1 2 3 4 5 6 7 8 9 lO 11 12
y 150 108 56 0 -54 -100 -132 -144 -130 -84 0 128 306
The graph is shown in Fig. 3.7 and has a minimum value at x = 7.

As a second example of solving a polynomial equation let y = 5x3
+ 27x2 + 17x -21 and puttingx = 0 givesy = -21. For x = 1, y = 28
and so as the sign on the value of y changes there must be at least one
root between x = 0 and x = 1. This is obvious once the graph is
considered. At x = 0 the value of y is negative while at x = 1, y is
positive, and so the graph must cross the x-axis. Here the values of y at
x = 0 and x = 1 are approximately equal (ignoring the sign), and so our
next guess at a root is x = 0. 5, which gives y = -5.125, indicating that
the root is greater than x = 0.5. Trying x = 0.6 gives y = 0 and sox= 0.6
is a root. Therefore (x - 0.6) is a factor and dividing through by this:
5x2 + 30x + 35
x- 0.6J5x 3 + 27x2 + 17x- 21
5x 3 3x2
0 30x2 + 17x- 21
30x2 - 18x
0 35x- 21
35x - 21
0
Thus 5x 3 + 27x 2 + 17x- 21 = (x- 0.6)(5x 2 + 30x + 35) and setting this
to zero results in the quadratic,
5x2 + 30x + 35 = 0
which has the roots x = -3 + v'2 and x = -3 - v'2
y
Fig. 3.7
While polynomial functions have many applications in economics

and business studies, other types of functions also occur. One of
the common ones is the rectangular hyperbola which has the general
equation
xy =c
where c is a constant which is neither zero nor infinity. If x is price and
y is the quantity sold then xy is the total revenue so that this curve
represents a constant revenue function. Thus whatever price is charged
the revenue is always the same, and so it can be referred to as an
iso-revenue or equal-revenue function. In section 5.5 we will see that
such a demand function has a constant elasticity of -1.
To plot the graph of a rectangular hyperbola suppose that c = 100

so that y = 100/x. Taking a range of values of x the following table can
be drawn up for positive values of x:
X 1 2 4 10 20 50 100 1,000 10,000

y 100 50 25 10 5 2 1 0.1 0.01
Notice that no matter how large x is, y is always positive and that as x
becomes larger and larger, y will become smaller and smaller. That is, as
x tends towards infinity, y tends towards zero. This can be written more
formally by using some new notation:
as x ~ oo, y ~ 0
which is read 'as x tends to infinity, y tends to zero'. Alternatively, we
can write:
as x ~ oo, lim (100/x) =0
Thus the 'limit' is the value to which the expression 100/x tends as
x becomes very large. In this example, the limit is zero.
A similar type of behaviour occurs as x gets closer and closer
to zero. For x = 1, y = 100, and if x = 0.5, y = 200, for x = 0.01,
y = 10,000, so that
as x ~ 0, lim (100/x) = oo
Thus, as x tends to zero, y approaches infinity.

So far only positive values of x have been considered. However, if
x is negative, since xy = 100, then y must also be negative. A table of
values can be drawn up as before:
X -1 -2 -4 -10 -20 -50 -100 -1,000 -10,000

y -100 -50 -25 -10 -5 -2 -1 -0.1 -0.01
Here it can be seen that,

as x ~ - oo, lim (100/x) = -0
so that as x takes a larger negative value, y remains negative but tends to
zero (from below). Similarly,
as x ~ -0, lim (100/x) =- oo
so that y becomes infinitely large in the negative direction.

y
100
80
60
40
20
30 20 10
10 20 30 X
-20
-40
-60
-80
-100
Fig. 3.8
The graph of xy = 100 is shown in Fig. 3.8 and consists of two

separate curves which do not join. As x tends to zero from the positive
side, y tends to infinity and the graph approaches, but does not meet,
the y-axis. Thus the graph is said to approach the y-axis asymptotically.
That is, the y-axis, which is the line x = 0, is called an asymptote of
xy = 100. Similarly, on the positive side, the x-axis, which is the line
y = 0, is an asymptote. It is also clear that when negative values of x are
considered the x-axis and they-axis are asymptotes of xy = 100.
The discussion above has been for xy = 100, but it obviously applies
for any curve xy = c where c is a non-zero constant. If c is negative the
two parts of the graph are in the other two quadrants.
Another version of the rectangular hyperbola occurs with average
cost functions when the total cost function is linear. Let the total cost
function be
TC =a+ bx
where x is the level of output and a and b are constants. We define the
average cost of production, AC, by
total cost TC
AC = - - - - - =
level of output x
a+ bx a
and here AC ---=-+b
X X
For example, suppose that the total cost function is

TC = 100 + 3x
so that the fixed cost is 100 and the marginal cost per unit is 3. Then
100
AC=-+3
X
Since here xis the level of output it is expected to be positive. Forming a

table of values:
X 1 2 4 10 20 50 100 1,000 10,000

AC 103 53 28 13 8 5 4 3.1 3.01
it can be seen that as x increases the value of AC approaches 3. That is,

asx- oo, lim (AC) =3
The graph is shown in Fig. 3. 9 and here the asymptotes are the y-axis
(x = 0) and the horizontal line y = 3. In economic terms the fixed cost is
being spread over more and more units of output and so the average
cost is approaching the marginal cost.
If the total cost function is non-linear, the behaviour of average cost
is more complex. For example, let
TC = 100 + 3x + x 2
100
then AC = - + 3 +X
X
and as x tends to infinity, A C also increases continuously. However, as

x increases, the term 100/x becomes smaller and smaller so that AC will
be dominated by (3 + x). That is, AC will approach the straight line
y = 3 + x but will not cross it and so this will be an asymptote. As
previously, they-axis will also be an asymptote since as x- 0, AC- oo.
The graph is shown in Fig. 3.10.
This section has been concerned with non-linear functions and, in
particular, how to draw their graphs. While it is possible to graph any
function by evaluating a large number of points, it is useful to find where
100
80
~ 60
u
Q)
Ol
ca
a;
>
<(
40
20
3 -------------------------------
~ ~ ~ ~ ~ ~ ~ ~ ~
0utput
Fig. 3.9
the graph crosses they-axis (where x = 0), and the x-axis (where y = 0).
For polynomial functions the points where y = 0 are the roots of an
equation and they can be found by a trial-and-error method. For other
non-linear functions, such as the hyperbola, it is important to detect the
presence of asymptotes. In Chapter 4 we discuss exponential and logar-
ithmic functions. A detailed treatment of maximum and minimum
values is presented in section 5. 7 and we return to the problem of
plotting graphs in section 5.18.
3.6 Exercises
1. Sketch the graphs of the following functions over the range of
values indicated.
(a) y == x 2 - 8x + 12 for x = 0 to x = 6
(b) y =x 3 - l3x 2 - 170x + 600 for x = 0 to x = 6
AC
100
80
60
40
20
~ ~ ~ ~ ~
20 40 60 80 100 X
Fig. 3.10
(c) y = (2x- 3)(x + l)(x- 5) forx = -2 tox = 6

(d) y = 32x - 2x3 for x = -4 to x = 4
(e) y = x 5 for x = 0 to x = 3
(f) xy = 20 for positive values of x
(g) xy = 20 + x for positive values of x.
2. In a simple production process, if L is the amount of labour input
and K is the amount of capital input, the volume of output, Q, is
given by Q = KL. If Q is assumed to be fixed then this gives the

trade-off between using L and K in producing the output Q. It is
known as an isoquant since it shows how different values of K and
L can be combined to give a particular value of Q. Sketch the
isoquants for Q 1 = 20, Q2 =50 and Q3 = 100. Notice that they do
not cross.
3. Sketch the graphs of the following total cost functions and their
average cost functions ( x is the level of output):
(a) TC = 100
(b) TC = 100 + 5x
(c) TC = 100 + 5x + x 2 •
3. 7 Breakeven point
In Chapter 1 the breakeven point was defined as that level of output
at which the total cost and the total revenue are equal. In general, a
unique value is obtained when both the cost and revenue functions are
linear. However, this is not necessarily the case when one or both of the
functions are non-linear.
Let us first consider the case in which the total revenue function is
linear, so that the price is independent of the quantity sold, and the total
cost function is quadratic. For example,
Cost function: TC = 7 + 2x + x 2
Revenue function: TR = lOx
where x is the level of output. At the breakeven point,
TC = TR and so
7 + 2x + x 2 = lOx
or,
This quadratic equation has the roots
8 ± v'(64 - 28)
x = = 1 and 7
2
and, therefore, either x = 1 or x = 7 satisfies the equation. This means
that there are two points at which total revenue equals total cost and this
can be seen to be true by sketching both graphs on the same set of axes
as shown in Fig. 3.11.
70 ----------------
7
Output
Fig. 3.11
In this case net revenue does not increase continuously with output
but has the following pattern:
x < 1 Net revenue is negative
x =1 Net revenue is zero
1<x < 7 Net revenue increases at first and then decreases again
x =7 Net revenue is zero
x >7 Net revenue is negative
It therefore has a maximum value for an output somewhere between
x = 1 and x = 7. The output at which this maximum value occurs can be
obtained using the differential calculus as we shall see later in Chapter 5.
For the present it is sufficient to note that the linear equation and
40
~ ~
0utput 4
Fig. 3.12
the quadratic equation intersect in two points and the maximum net
revenue occurs at an output somewhere between these two points.
But this is not always so. Consider the case where the following
equations apply:
Cost function TC 16 + 2x + x 2
Revenue function TR lOx
Then TC TR
when 16 + 2x + x 2 lOx
or x2 - 8x + 16 0
and there are repeated roots with just one solution, x = 4. The graphs
are shown in Fig. 3.12 where it can be seen that there is one point at
which the two lines meet. When x = 4 the straight line is a tangent to the
curve. Mathematically the two coincident solutions are x = 4. Economi-
cally, it is apparent that there is only one point at which the net revenue
is non-negative and at no point is it positive. It is therefore likely that
Output
Fig. 3.13
the company will consider either of the two following possibilities.

(a) A price increase if this is possible. The resulting revenue func-
tion would then cut the cost function in two places and if the necessary
output could be sold at the new price a positive net revenue could be
obtained.
(b) A cost reduction in the process. This would again result in the
cost and revenue functions intersecting in two points between which
there would be an area of positive net revenue.
One further case can occur when the two functions do not intersect
at all. This is illustrated graphically in Fig. 3.13
In this case there is no output for which there is a positive net
revenue. This situation would arise if the fixed costs of the plant were to
increase without a corresponding increase in the price of the product.
For example,
Cost function TC = 20 + 2x + x2
Revenue function TR lOx

Then TC TR
when 20+ 2x + x2 lOx
or r - 8x + 20 = 0
This equation has two complex roots and therefore there is no real
value of output for which a breakeven point occurs.
3.8 Simultaneous quadratic equations

We have used linear equations to represent demand and supply
functions but we are now able to extend the analysis a little further by
considering quadratic equations. For example,
demand qd = f(p) = p 2 - 8p + 15
supply qs = g(p) = 2p 2 + 3p - 3
Graphical representation of these two equations on the same set of axes
shows that there is a position where the curves intersect. This is the
point at which supply and demand are equal (see Fig. 3.14)
The equilibrium values of p and q could be obtained graphically but
we can also solve the two quadratic equations simultaneously. For
example,
qd = p2 - 8p + 15
qs = 2p + 3p
2 - 3
and in equilibrium qd = qs
:. p 2 - 8p + 15 = 2p 2 + 3p - 3
:. p 2 + llp - 18 = 0
-11 ± Y[11 2 + (4 X 18)] -11 ± Y193
:. p =
2 2
Y193 can be found by reference to tables of square roots or by the use
of a calculator. The result is
-11 ± 13.9
p=
2
and the two roots are p, = 2.9/2 = 1.45 and p 2 = -24.9/2 = -12.45.
6 /
/
/
I
5 I
I
I
14
I q = 2p 2 + 3p- 3
\
\
Q. '
~ 3 q=p 2 -8p+ 15
ct \ /
\/
"'
' ''
'' .......
......
......
...... ,
2 4 6 8 10 12 14 16,18 20 22 24
Quantity, q
........................................
-1
Fig. 3.14
Mathematically the two solutions are both correct because the

quadratic curves intersect in two places, but by the very nature of our
problem price cannot be negative, and therefore we must take the
positive root.
The equilibrium price will be 1.45 units and at this price
2.1 - 11.6 + 15 = 5.5 units are produced and sold
3.9 Net revenue
Quadratic equations often arise in economic theory from situations
where the original data can be represented by a linear equation.
Let the demand function for a product be
qD = f(p) = 1000- lOp
25,000
50
Price, p
Fig. 3.15
The total revenue, which is obtained by the sale of q items, is

obtained by multiplying price by quantity sold.
:. Total revenue = pq = p(lOOO - lOp)
= lOOOp - l0p 2
This is a quadratic equation and a graph of the function shows that
the total revenue rises at first and then after reaching some maximum
value starts to decline (Fig. 3.15).
This graph is a parabola. It is symmetrical about the line parallel to
the revenue axis through the point p = 50 as shown by the dotted line.
The maximum revenue is therefore obtained when p = 50 and this is
equal to
p(lOOO - lOp) = 50(1000 - 500) = 25000
The quantity which will be sold at this price is given by

q0 = 1000- lOp = 1000- 500 = 500
The total revenue can be obtained as a function of quantity deman-
ded from the equations
q 0 = 1000 - lOp (1)
TR = pq = lOOOp - 10p2 (2)
from (1) lOp = 1000 - q 0
:. P = 100 - foqo
Substituting this in (2), we obtain
TR = 1000(100 - toq 0 ) - 10(100 - toqvY
= 100000 - 100q 0 - 100000 + 200q 0 - ~
= 100qo - ~
This is again a quadratic equation which is a parabola and total
revenue is equal to zero when
100q 0 - ~ = 0
q 0 (100 - foq 0 ) = 0
:. either qD = 0
or 100 - foq 0 = 0
that is, qD = 1000
The maximum revenue is equal to 500 X 50 = 25000. The function
can now be represented graphically as shown in Fig. 3.16.
The company will be interested not so much in the total revenue
but in the net revenue after the costs of production etc have been met.
The line superimposed on the graph corresponds to the linear cost
function.
TC = 5000 + 15qD
The net revenue for any quantity of output q 0 is given by the
vertical distance between these two functions at that quantity. Zero net
revenue is obtained at the two breakeven points which are given by
TR = TC
25,000
"
0
·~
"
~
Cl>
:1
">
Cl>
.,"
"'0
"
0
~
"
~
'iii
0
u
Fig. 3.16
100qv - ~ = 5000 + 15qv

i.e. ~ - 85qv + 5000 = 0
85±V(852 -4x-/ox5000) 85±72.28

qD = 2 X _L 0.2
10
That is approximately 64 units and 786 units. Between these two points
is the quantity of output which yields the maximum net revenue and it
could be obtained from the equation representing the net revenue
function:
Net revenue = total revenue - total cost
= l00qv - ~ - (5000 + 15qv)
= 85qD - ~ - 5000
The maximum value of this function can be found graphically, but

is obtained more simply using the differential calculus (see Chapter 5).
3.10 Exercises
1. Find the breakeven points (if any) when the total cost (TC) and
total revenue ( TR) functions are as follows (x is the level of
output):
(a) TC = 9 + 2x + x 2, TR = 8x
(b) TC = 15 + x + x 2, TR = lOx- 3
(c) TC = 10 + 3x + 2x2 , TR = 6x.
2. A manufacturer has the following information about the costs of

production on a machine:
Output (x) 5 10 15
Total costs 20 65 160
(a) Determine the equation of the cost function, assuming it can
be represented by a quadratic expression.
(b) What is the cost of producing 20 units of output?
(c) Sketch the graph of the cost function for 0 ~ x ~ 20.
(d) The revenue function is given by
YR = lOx+ 15
Show that there are two levels of output at which total revenue
equals total costs, and illustrate this graphically.
3. The total cost function for an output of x units of a product is
TC = 250 + 4x
Draw the graph of the average cost function and show that the
average cost approaches the value of 4 asymptotically.
4. Show that the equations
xy = 10
and 3y = 32- 2x
are satisfied for two values of x and sketch the graphs of the
equations.
5. Show that the simultaneous equations
y=20+3x+x2
y=5x+b
have the two real solutions when b = 20, two real coincident
solutions when b = 19 and two complex solutions when b = 18.
Sketch the graphs of these equations to illustrate these three cases.
6. Find the equilibrium price and quantity for the demand curve
q = 250 - 4p - p 2
and the supply curve
q = 2p 2 - 3p - 40
and sketch the curves for the range
p=O to p = 10
7. Determine the equilibrium price and quantity for the following
demand and supply functions. Sketch the curves for the range
p = 0 top= 8
Demand Supply
(a) q - 25 + 3p = 0 q=2p 2 -40
(b) p2 + q2 = 32 4 = 3p- 2q
(c) pq =6 q = 3(p- 1)
8. The demand function for a product is q 0 = 200 - 2p and the total
cost function is TC = 20 + 5q 0 . Determine the net revenue func-
tion and the breakeven levels of output and price. Sketch the graph
of the net revenue function for q 0 = 0 to q 0 = 200.
3.11 Discontinuous functions

So far we have looked at those functions for which the graph is a
continuous line between the values considered. However, an equation
may be valid only between certain limits. For example, the relationship
between the total cost of manufacture and the number of units of output
of a product may be represented by an equation such as the following:
TC =a+ bx
where a is the fixed cost and b the variable cost of production which in
this case is constant for all levels of output. But it is apparent that the
volume of output cannot be increased indefinitely for a given amount of
fixed costs. For example extra investment in machinery will eventually
be needed. We must, therefore, add the condition that the equation is
valid only up to a certain quantity of output above which an increase in
/
I
I
I
I
I
....
"'0tJ
co
;§
/
I
I
I
100 200 300

Output
Fig. 3.17
fixed costs must be incurred. By a stepwise construction we might arrive
at the following set of equations.
TC =a,+ bx 0 <X ::S 100
TC = a2 + bx 100 < X ::S 200
TC = a3 + bx 200 < X ::S 300
with the condition a, < a2 < a3 •
This set of equations can be graphed on the one set of axes as in
Fig. 3.17. The graph in this case is not continuous between the limits
x = 0 and x = 300. It consists of three separate sections and at the
values x = 100 and x = 200 there are discontinuities. At these values a
very small increase in one variable, quantity, requires a large increase in
the other variable, total cost. In this particular example the discontinuity
occurs when the value of the constant a in the equation is changed.
Another example of a discontinuity occurs when there are reduc-
tions for purchasing large quantities of goods. For example, suppose
that the price per unit is 6 for quantities below 100 and that a discount of
20% is given for purchases of 100 to 499 and a discount of 30% on
Total
cost
3000
2400
1800
1200
600
~ ~ ~ ~ ~ ~ ~ ~
100 200 300 400 500 600 700
Quantity purchased
Fig. 3.18
purchases of 500 or more. Let x be the quantity purchased then the total
cost to the purchaser is
TC = 6x for 0 < x < 100
TC = 6x(0.80) for 99 < x < 500
TC = 6x(0.70) for 499 < x
The graph is shown in Fig. 3.18 and it can be seen that the slope of the
cost curve changes at x = 100 and x = 500.
3.12 Exercises
1. A firm has a discontinuous cost function and the relationship
between output (x) and total cost (TC) is given by
TC = 15 + 3x 0 :s;; X :s;; 30
TC = 40 + 3x 30 :s;; X :s;; 50
TC = 70 + 3x 50 :s;; X :s;; 100

Illustrate this graphically. What is the cost of raising output
(a) from 19 to 21 units?
(b) from 29 to 31 units?
(d) from 49 to 51 units?
2. A canner sells beans at a price of 30p per tin but offers a discount
of 15% if more than 100 tins are bought. Sketch the revenue
function. What is the total cost of 99 tins? What is the total cost of
101 tins? Noticing this difference, the canner decides to change the
discount to one of 20% but does not allow this on the first 100 tins.
What is the cost of 120 tins under each of these schemes?

1. The values of output (x) and total cost (TC) for a production
process are
X 10 20 30
TC 430 1,060 2,900
Assuming that the total cost function is quadratic, determine its

equation. What is the fixed cost? Sketch the average cost function
for x = 0 to x = 30.
2. Find the roots of the following quadratic equations:
(a) x 2 - +4=0
6x (b) 2x2 + 8x +8=0
(c) x 2 + 4x + 5 = 0 (d) 2x2 - 3x + 3 = 0
3. Draw the graphs of the following functions for x = 0 to x = 10:
(a) y = x2 - 6x +4 (b) y = x2 + 4x + 5
(c) xy = 15 (d) xy = 15 + 2x
4. The total cost function for a product is
TC = 100 + 4x + x 2
Sketch the graph for x = 0 to x = 20 and add the average cost

function. What happens to the average cost as x tends to infinity?
5. What is the breakeven value of output (x) if the total cost func-
tion is
TC = a + bx + cx2
and the price is p per unit? Sketch the net revenue function.
6. The demand function for a product is
qd = 200 - 2p - 2p2
and the supply function is
cf = 4p- 20
What are the equilibrium price and quantity? Sketch the two
functions.
7. Determine the equilibrium price and quantity if the demand and
supply equations are:
qd = 500 - 4p - p 2
cf =p2 +6p-40
8. Show that the equations xy = 30
6x+y=a
have no real solution when a = 20. For what values of a are there
two real solutions?
9. The total cost function fer a product is
TC = 150 + 3q + q2
If the price is 8 per unit, determine the net revenue function and
the breakeven quantity.
10. A machine for producing a component has an installation cost of
500 and the marginal cost per component is 20. The capacity of
the machine is 200 components per week and if output needs to
be greater than 200 another identical machine can be installed.
Write down the total cost function for producing up to 400 com-
ponents per week and sketch the average cost function.
Chapter 4
Series
4.1 Introduction
This chapter covers some basic applications of series in economics
and business studies, mainly in the areas of interest rates and finance,
but also extended to exponential and logarithmic functions. We start
with the idea of a progression. This is a number of terms arranged in a
definite order and for which there is a pattern or rule which allows
further terms to be identified. For example:
100, 101, 102, 103, 104, 0 0 0
1, 10, 100, 1000, 10000, 0 0 0
0, 100, 100, 200, 400, 0 0 0
are all progressions. In the first example 100 is the first term or initial
term and each subsequent term is obtained by adding 1 to the previous
term. In the second example, 1 is the first term and each subsequent
term is obtained by multiplying the previous term by 10. In the third
example the first term is 0, the second term 100, and each subsequent
term is obtained by adding together all the previous terms, so that
0 + 100 = 100, 0 + 100 + 100 = 200, 0 + 100 + 100 + 200 = 400,
and so on.
4.2 Arithmetic progressions

An arithmetic progression is one in which there is a constant differ-
ence between each term. The first example above is an arithmetic pro-
gression. The difference between any two (consecutive) terms is 1.
Another example is the progression
100, 90, 80, 70, 60, 0 0 0
where 100 is the first term and the difference between any two terms is
-10. In general, any arithmetic progression can be written as
130
Series 131
a, a + d, a + 2d, a + 3d, ...

where a is the first term, and dis known as the common difference. For
example, for the progression
100, 101, 102, 103, 104, ...
a= 100, d = 1
while for
100, 90, 80, 70, 60, ...
a = 100, d = -10.
Notice that in the general arithmetic progression the second term is
a + d, the third term is a + 2d, the fourth term isa + 3d, and so the
nth term is a+ (n - 1)d. For example for a= 100 and d = 1, the
twentieth term is
a + (20 - 1)d = 100 + 19 = 119
and for a = 100, and d = -10, the tenth term is
a + (10 - 1)d = 100 + 9(- 10) = 10
A useful application of arithmetic progressions is in the calculation
of simple interest.
For example, suppose £100 is invested at 5 per cent per annum
simple interest. Then the interest for the first year is £5 (from
£100 X 0.05 = £5). Similarly, the interest for any subsequent year will
be £5, and so the value of the investment is as shown in Table 4.1, since
a= 100, d = 5.
The value at the beginning of the tenth year (i.e. after 9 complete
years) is a + (10 - 1)d = 100 + 9(5) = £145, and the value of the
investment after 20 years (i.e. at the beginning of year 21) is
100 + (21 - 1)5 = £200
TABLE 4.1
Year 1 2 3 4 n
Value at beginning 100 100 + 5 105 + 5 110 + 5 100 + (n - 1)5

of year = 105 = 110 = 115
It is frequently useful to be able to obtain the sum of an arithmetic

progression to give an arithmetic series. Returning to the general pro-
gression a, a + d, a + 2d, a + 3d, ... , a + (n - 1)d, we see that the
sum of the first n terms, s"' is
sn = a + (a + d) + (a + 2d) + ... + [a + (n - 1)d]
= na + d + 2d +3d+ · · · + (n- 1)d
which can also be written (by reversing the order) as
sn = na + (n- 1)d + (n- 2)d + ... +3d+ 2d + d
Adding these two expressions together gives
2Sn = 2na + nd + nd + · · · + nd
since such terms as d + (n - 1)d =d + nd- d = nd and
3d + (n - 3)d = 3d + nd - 3d = nd
Hence,
2Sn = 2na + (n - 1)nd
as there are (n -1) terms equal to nd.
n
:. sn = 2 [2a + (n - 1)d]
Alternatively, this can be written as

[a + {a + (n - 1)d}]
sn =n 2
= (number of terms) (average of first and last terms).

For example, for the progression 100, 101, 102, ... in which
a =100, d= 1, the sum of the first n terms is
sn = ~ [200 + (n - 1)]
so that the sum of the first 3 terms is
s3 = t[2oo + {3 - 1)] = ~ = 303
which can easily be verified. The sum of the first 20 terms is

20
S20 = z-[200 + (20 - 1)] = 2190
Series 133
which is less easily verified.

Example
A man invests £100 per annum in bonds which pay 5% per annum
simple interest. What is the value of the investment at the beginning of
the 11th year? This value is the sum of the first eleven terms of the
arithmetic progression with a = 100, d = 5 and is given by
11
S11 = 2[200 + (11 - 1)5] = £1375
This is verified by looking at the value of the investment for the first
few years (Table 4.2). Thus, at the beginning of the 4th year the value is
100 + 105 + 110 + 115, which is the sum of the first 4 terms of the
arithmetic progression with a = 100, d = 5. Hence the value of the
investment at the beginning of the 11th year (i.e. when n = 11) is
S11 = ~ [200 + (11 - 1)5] = £1375.
TABLE 4.2
Year Value at beginning of year
1 100
2 100 + (100 + 5)
3 100 + (100 + 5) + (100 + 10)
4 100 + (100 + 5) + (100 + 10) + (100 + 15}
4.3 Exercises
1. Give the tenth term of each of the following progressions and also
determine the sum of the first fifteen terms.
(a) 1, 3, 5, 7, ...
(b) 500, 550, 600, 650, ...
(c) 60, 30, 0, -30, .. .
(d) 100, 98, 96, 94, .. .
2. What is the value of an investment of £200 after 5 years if simple
interest is paid at 10% per annum?
3. A man receives a salary of £8,500 per annum which increases

annually by £400. What salary will he receive after 7 years? What
will be his total income during the first six years?
4. A government issues savings bonds which sell for £10 and are
worth £20 after 15 years. What is the implied rate of simple
interest?
4.4 Geometric progressions

A geometric progression is one in which there is a constant ratio
between any two consecutive terms. For example in the progression
1, 10, 100, 1000, 10000, 0 0 0
each term is 10 times the previous term so that 100 = 10(10) and
10,000 = 10(1,000). In the progression 128, 64, 32, 16, 8, ... each term
is 0.5 times the previous term, e.g. 64 = 0.5(128), and 32 = 0.5(64). In
general, any geometric progression can be written as
a, ar, ar\ ar\ ...
where a is the first term and r is known as the common ratio. In the
above examples, a = 1, r = 10 and a = 128, r = 0.5. The nth term in a
geometric progression is arn- 1 ' so that for the progression with a = 1,
r = 10, the sixth term is 1(HY') = 100,000, while for the progression
with a = 128, r = 0.5, the fifth term is 128(0S) = 8.
One of the applications of geometric progressions is in the calcula-
tion of compound interest. Here the sum on which interest is paid
includes the interest which has been earned in previous years.
For example, if £100 is invested at 5% per annum compound
interest, then after 1 year the interest earned is £5 (from 100 x 0.05)
and the capital invested for the second year is £105. The interest earned
by this capital is not £5 but £105 X 0.05 = £5.25. The capital invested
for the third year is £105 + £5.25 = £110.25, and the interest earned is
£110.25 x 0.05 = £5.5125. This is shown in Table 4.3.
This can be expressed in general terms, as shown in Table 4.4 for an
investment of a and an interest rate of i per cent. This shows that the
progression for the capital is a, a(1 + i), a(1 + i)Z, a(1 + i)\ ... ,
which is of a geometric form with common ratio r = 1 + i.
For example, if i = 5% = 0.05, then r = 1.05 and so the nth term
in the progression if a = 100 is 100(1.05)n- 1 • The value of the invest-
ment after 8 years is 100(1.05)9 - 1 = £147.75 (note that n = 9 since we
want the ninth term in the progression). The evaluation of such quanti-
Series 135
TABLE 4.3
Beginning Capital Interest

of year during year
1 100 5
2 105 5.25
3 110.25 5.5125
4 115.7625 5.7881
TABLE 4.4
Beginning Capital Interest

of year during year
1 a ai
2 a + ai = a(1 + i) a(1 + i)i
3 a(1 + i) + a(1 + i)i = a(1 + i) 2 a(1 + i) 2i
4 a(1 + i) 2 + a(1 + i) 2i = a(1 + i) 3 a(1 + i) 3 i
ties as 1.058 is best done by calculator or by logarithms. A short section

on the use of logarithms is included at the end of this chapter (Section
4.14).
The value of a geometric series, which is the sum of a geometric
progression, is easily obtained by considering the general progression
The sum of the first n terms,

sn = a + ar + ar 2 + ar 3 + ... + arn-l
also r . sn = ar + ar 2 + ar 3 + ... + arn-l + arn
Subtracting gives Sn - rSn =a - arn; that is
or S = a(1 - rn)
n 1- r
For example, for the progression with a = 1, r = 10,

S = 1(1 - lOn)
n 1- 10
so that the sum of the first 5 terms is

S = 1(1 - 105 ) = lOS - 1 = 11111
5
1 - 10 9
and the sum of the first 8 terms is
s = 1(1 - 108) = 108 - 1 = 11111111
8
1 - 10 9
A special case arises when - 1 < r <1
a arn
Then sn =-----
1- r 1- r
and as ~ ~ and the second term will approach zero.
The sum of all the terms in the series or, as it is generally referred
to, the sum to infinity is given by
S =-a-
"" 1- r
For exarilple, if a = 128, r = 0.5
s"" = 128 = 128 = 256
1- 0.5 0.5
This means that adding together the successive terms of the pro-
gression 128, 64, 32, 16, 8, 4, 2, 1, 0.5, ... produces a sum which
becomes closer and closer to, but never becomes greater than, 256.
It is important to realise that r must satisfy - 1 < r < 1 for the sum
to infinity to be finite. For example, for the series a = 1, r = 10,
s"" = _1_ _ t.to"" = _ _!_ + 10""

1 - 10 1 - 10 9 9
which is obviously infinity.
4.5 Exercises
1. State the sixth term of each of the following progressions and
determine the sum of the first 10 terms. Evaluate the sum to
infinity if it is finite.
(a) 10, 30, 90, 270, ...
(b) 81, 27, 9, 3, ...
Series 137
(c) 2, -4, 8, -16, ...

(d) -1,024, 512, -256, 128, ...
(e) 1, 0.1, 0.01, 0.001, ...
2. What is the value of an investment of £300 at the end of 15 years if
compound interest is paid at
(a) 5% per annum? (b) 10% per annum?
3. An insurance policy costs £150 per annum for 25 years. What is the
final value of the policy if compound interest is earned at a rate of
6% per annum?
4. Government savings bonds are sold for £4 and are worth £5 after 4
years. What is the implied rate of compound interest?
4.6 Discounting
If money is able to earn interest its absolute value in future years
will be greater than its current or present value. Conversely money
which is to be spent in future years has a present value which is less than
its absolute value.
For example, if we have £100 now and we invest this at 6% per
annum compound interest, then we will have £106 at the end of one
year. If, therefore, we are required to spend £106 in one year's time we
would only need to have available at the present time £100 which we
could invest at 6%. The present value of a capital sum of £106 which is
required in one year's time is therefore equal to £100 when the interest
rate is 6% per annum.
A sum of £100 per required in one year's time then has a present
value given by
PV = 100 X 100 = 100 = 100 = £94.3

106 1.06 (1 + 6/100)
This method of calculating the present value of future sums of
money, whether payments or receipts, is known as discounting. It can be
used to cover any time span as follows:
If future requirements for capital are known with their appropriate
time pattern then
(PV) =a +-a_1_+ a2 + ... + an

c 0
(1 + i) (1 + i)2 (1 + i)n
where (PV)c =present value of capital requirements

a0 is the immediate requirement
a1 is the requirement in the first year and discounted as
though it appeared at the end of the first year.
a2 is the requirement in the second year
an is the requirement in the nth year.

If these capital requirements or cash outflows are for a given
project then the project will produce a series of receipts, or cash inflows,
again over a period of time. These can be discounted using the same
compound interest rate.
( PV) = b +-b_t_ + bz + ... + bn

R
0
(1 + i) (1 + i) 2 (1 + i)n
where (PV)R =present value of receipts
b0 is the immediate return
b1 is the return in the first year and discounted as though it
appeared at the end of the first year
b2 is the return in the second year
bn is the return in the nth year.

For the project to be profitable on a strictly financial basis we must
have the condition that
that is, the present value of the receipts from the project must be greater
than the present value of the payments.
The above is known as the present-value method for assessing
investment projects and the net present value (NPV) is defined as the
difference between the discounted values of receipts and payments.
NPV = (PV)R - (PV)c
The value used for i in the discounting process obviously has a
considerable influence on the net present value. It is known as the cost
of capital and its value is often difficult to establish accurately in many
situations. Because of this, preference is often given to a method which
allows the cost of capital to be a variable and determines that value
Series 139
TABLE4.5
Table of Discount Factors

This table shows the present value of 1 discounted for different numbers of
years and at different rates of discount
Rate of discount
(%) 8 10 12 14 16
Year
0 1.000 1.000 1.000 1.000 1.000

1 0.926 0.909 0.893 0.877 0.862
2 0.857 0.826 0.797 0.769 0.743
3 0.794 0.751 0.712 0.675 0.641
4 0.735 0.683 0.636 0.592 0.552
5 0.681 0.621 0.567 0.519 0.476
6 0.630 0.564 0.507 0.456 0.410
7 0.583 0.513 0.452 0.400 0.354
8 0.540 0.467 0.404 0.351 0.305
9 0.500 0.424 0.361 0.308 0.263
10 0.463 0.386 0.322 0.270 0.227
15 0.315 0.239 0.183 0.140 0.108
20 0.215 0.149 0.104 0.073 0.051
which equates the present value of receipts to the present value of

payments. That is,
al an bl bn
or ao +(1 + i) + ... + (1 + it = bo + (1 + i) + ... + (1 + it
Since a0 - an and b0 - bn are assumed to be known, this equation is a

polynomial in i and can be solved by the quadratic formula (when n = 2)
or by Newton's method (see section 5.17). Because i is a rate of interest
it must be positive and normally we expect only one positive value to
result. This value is known as the internal rate of return (IRR) and the
higher this is the more profitable the project will be.
Both of the above methods (i.e. NPV and IRR) require that each
year's payments and receipts are discounted by the appropriate factor
which is determined by the time period in which these payments and
receipts arise. Values of 1/(1 +it, known as discount factors, are given
in Table 4.5. For example, the discount factor at 10% per annum for
5 years is 1/(1 + 0.10) 5 = 0.621.
4. 7 Exercises
1. (a) If £300 is invested for 10 years and is then worth £800, what is
the implied rate of compound interest?
(b) What is the present value of the £800 at 16% per annum
compound interest?
2. What is the present value of £1,000 payable after 10 years if the
rate of discounting is
(a) 14%? (b) 8%?
3. (a) How much should be invested at 10% per annum compound
interest to give £250 after 5 years?
(b) What is the value of this investment after 3 years?
4. Two projects are available to a company and the estimated returns
are:
End of year 1 2 3
Project A 100 200 300
Project B 150 300 100
Which project has the greater present value if the discounting rate
is 10%?
5. What is the internal rate of return from a project which has the
following costs and receipts?
End of year 1 2 3
Costs 120 120 100
Receipts 100 110 160
4.8 Annuities and sinking funds

We now apply some of these ideas to a number of related practical
problems. First we consider annuities, where for a capital payment now
an annual income is received for a number of years in the future. The
two questions to be answered are: if I pay the capital amount p now
what annual income, A, will I receive for n years, and if I want an
annual income of A for n years, how much will it cost now? Further
applications of the same points arise with perpetual bonds and chief
rents. Next, we examine mortgages, where a loan of p now is to be
repaid at a rate of A per year, and again the questions concern p (given
the value of A) or A (given the value of p). With both annuities and
Series 141
mortgages the capital amount is required immediately and the income

or repayments are in the future. The situation is reversed with sinking
funds, pension funds and saving schemes. For these, a constant sum A is
added to the fund each year and after n years the total value V is
returned with interest. Again the questions concern the annual pay-
ment, A, and the total value after n years, V.
In the following, two simplifying assumptions will be made. First, it
will be assumed that the interest rate in the future will be a known
constant. In practice, interest rates fluctuate and future rates are un-
known. If in fact we do have some information about future rates, this
can be taken into account at the expense of making the various formulae
more complicated. If not, we can interpret the assumed rate of interest
as being an average of future values. The second assumption is that the
payments and receipts occur once a year. In practice, many are made
quarterly or monthly. This can be taken account of by adjusting the rate
of interest so that an interest rate of 8% per annum, when applied
monthly, becomes 8/12 or 0.67% per month. This is discussed further in
section 4.10 below.
An annuity is a constant annual income which can be bought for
cash. For example, an annual income of £10 for 20 years may be
purchased for £120. The present value of an annuity of £A payable for
n years is given by
A A A A
P = 1 + i + (1 + iY + (1 + i) 3 + ... + (1 + it
~
1+ i 1+ i
1
(1 + i)2
+···+ 1
(1 + i)n-l
]
The term in the brackets is the sum of a geometric series with a = 1

and r = 11(1 + i) and so
p = 1 ~ i [ 11 ~ r;] = ~ [ 1 - (1 ~ i)"]
For example, the present value of an annuity of £10 per annum for
20 years if the rate of interest is 7% is
10 [ 1 ]
p = 0.07 1- (1.07)20 = £105.94
A special case occurs when the annuity continues indefinitely, since

A A A A/(1 + i) A
P = 1 + i + (1 + iY + (1 + i) 3 + ... = 1 - 11(1 + i) - -
For example, an annuity of £20 per annum for ever has a present
value of
20
p = - = £400
0.05
if the rate of interest is 5% per annum, while if the interest rate is 10%
per annum the value would be 20/0.10 or £200.
Exactly the same situation arises with perpetual government bonds.
In the UK there are some special government securities known as 21f2%
Consols (or Consolidated Stock) which were originally sold for £100
each in return for a promise to pay the owner the fixed amount of £2.50
per annum for ever. The current value of Consols depends on the rate of
interest and, if the rate of interest i!> 9%, is given by
2.50 2.50
P =-=-=£27
i 0.09 .
78
Another application of the same concept is the idea of a chief rent.

This is an annual charge paid by some home owners in the UK to the
original owner of land which has been used for housing. The landowner
accepts an annual payment for the land instead of a lump sum. For
example, a plot of land may have an annual chief rent of £25. The
present value to the land owner, assuming that the first payment is due
one year from now and the interest rate is 8% is
A 25
P =-=-=3125
i 0.08 .
Notice that if there is a payment due immediately then the value is
A A A(1 + i)
P =A + --+ ...
1+i 1 -{11(1 + i)}
instead of Ali. For the chief rent example the value would be 337.5
which is £25 more than in the previous case because of the extra
payment of this amount.
With mortgages, an amount p is borrowed now, usually to pay for a
house, and is to be repaid by n equal annual payments of A. Let Mr be
the mortgage debt at the end of year t. Then, if p is borrowed and the
Series 143
interest rate is i, the amount owed after 1 year is

M 1 = p(1 + i) - A
since the first term covers the capital borrowed and the interest charge
and the repayment of A is made towards the debt. In the same way,
M 2 = M 1(1 + i) -A
since M 1 is owed at the end of year 1. But we can express M 1 in terms of
p and so, using the expression for Ml>
M 2 = {p(1 + i) - A}(1 + i) - A
= p(1 + iY -A {(1 + i) + 1}
Similarly,
M 3 = M 2 (1+ i) - A
= [p(1 + i) 2 - A {(1 + i) + 1} ](1 + i) - A
= p(1 + i) 3 - A{(1 + iY + (1 + i) + 1}
In general,
M, = p(1 + iY- A{(1 + iy- 1 + (1 + iy- 2 + · · · + (1 + i) + 1}
The final term can be simplified since it is the sum of t terms of a
geometric progression with a common ratio of (1 + i) and so
1(1-(1 + iY
(1 + iY- 1 + (1 + iY- 2 + · · · + (1 + i) + 1 = .
1 - (1 + z)
(1 + iY- 1
Therefore,
A{(l + i) 1 - 1}
M, = p(1 + iY - .
z
But if the mortgage is to be paid off over n years, Mn must be zero and
therefore,
A{(1 + i)n - 1}
0 = p(l +it- .
l
This gives an equation for pin terms of A, i and n,

A{(1 + i)n- 1}
p=
(1 + wi
or for A in terms of p, i and n
i(1 + it p
A= (1 +it- 1
For example, if p = £100,000, the interest rate is 10% and n = 25 years,
0.1(1 + 0.1) 25 (100000)
A= (1 + 0.1) 25 - 1
0.1(10.8347)(100000)
= 9 _8347 = £11016.8 per annum
As a second example, suppose a borrower is able to repay £7000 per
annum. What amount can be borrowed over 20 years if the interest rate
is 12%? Here,
A{(1 +it -1} 7000{(1.12)2° -1}
p=
(1 + i)" i 1.1220(0.12)
7000(8.6463)
= £52286
(9 .6463)(0.12)
and so £52,286 could be paid back over 20 years at a rate of £7,000 per
annum.
A sinking fund is a fund set up to meet some financial commitment,
to which a constant sum is added each year. For example, how much
should be invested each year if the rate of interest is 5% per annum to
give a capital sum of £600 in 10 years?
Let A be the amount invested. After 1 year, the value of the
investment will be A + A(1 + i) =A + A(l.05). After 2 years, the value
will be A + A(l.05) + A(l.05)2 and in 10 years the value will be
A + A(l.05) + A(l.05) 2 + · · · + A(1.05)9
This has to be equal to £600.
. . 600 = A(1 + 1.05 + 1.052 + · · · + 1.059 )
A[1 - (1.05) 10]
1 - 1.05
Series 145
since the expression in brackets is a geometric series.

600(1 - 1.05)
:. A = 1 - (1.05)1o = £47.7
That is, if a sum of £47.7 is invested each year it will accumulate to

£600 in 10 years at an interest rate of 5% per year.
The same principle applies to life insurance (savings) policies
where, after making a regular payment for a number of years, the
policy-holder receives a payout. If it is assumed that the whole of the
annual premium (A) is invested (so that any administration charges and
payments for life cover are ignored) then it is simply a regular savings
scheme as operated by some banks and pension funds. If the interest
rate is i then the value in year n is
V = A + A(1 + i) + A(1 + i)1 + · · · + A(1 + i)n-l
A{1 - (1 + it} A{(1 + i)n- 1}
1 - (1 + i)
For example, if A = £100, i = 8% and n = 25 then V = £7,310.59. As
previously, if A is unknown it can be determined given the values of V, i
and n, so that, for example, if a policy is required which will be worth
£5,000 after 10 years when the interest rate is 11%, the annual premium
would be
iV 0.11(5000)
A= (1 +it- 1 = 1.1110-1 =£299.01
4.9 Exercises
1. What is the present value of an annuity of £100 per annum which is
to be paid 20 times, commencing one year from now, if the interest
rate is 8%?
2. What is the present value of a chief rent of £25 if the rate of interest
is 4%?
3. A company buys a machine for £2,000 and estimates that its life
will be 15 years. How much should be paid annually into a sinking
fund to buy a replacement for the machine in 15 years time if the
replacement will cost £2,500 and the rate of interest is 9%?
4. An ex-pupil decides to donate to his school a sum of money to
provide an annual prize of £60 for the next ten years.
(a) How much should be donated if the rate of interest is 6%?

(b) How much should be donated if the prize is to be paid inde-
finitely?
5. A couple borrow £70,000 to buy a house. What would their annual
repayments be if the interest rate was 12% and the repayment
period was 20 years? They decide this repayment rate is too high,
and ask for the period of the loan to be extended to 30 years. What
will the revised repayment rate be?
6. An insurance company sells policies for which the premium is £300
per year. If the interest rate is 9% what is the value of a policy after
(a) 10 years (b) 20 years?
7. A firm introduces a pension fund for its employees in which it
makes annual payments of £250 for each employee and interest is
added at 8% per annum. How much would an employee's fund be
worth if someone was a member for (a) 25 years or (b) 35 years?
4.10 Interest paid continuously

In our discussion of interest payments we have assumed that in-
terest has been paid annually. However, there are many cases where
this is not so. For example, some bank accounts and some government
bonds pay interest twice a year, while others pay interest quarterly and
generally credit card companies charge interest on the balance outstand-
ing each month. The effect of these more-frequent payments is that the
money earns interest for a shorter time before the capital increases by
the amount of interest. For example, suppose that the interest rate is
6% per annum compound and £100 is invested. If interest is paid twice
a year then after the first six months the interest earned is a half of 6%
or £3 since the £100 has been invested for half of the year. Effectively,
the interest rate is halved because the investment period is halved. The
capital at the beginning of the second six months is £103 and so the
interest earned during the second half year is 3 per cent of £103 or £3.09,
making the value after one year £106.09 which is £100(1 + 0.03) 2 • Notice
that even though the interest rate is 6% the interest received is at the
rate of 6.09% because of the twice-yearly payment.
Now suppose for simplicity that the interest rate is 100% per annum
compound and that £1 is invested. If interest is paid just once a year the
value at the end of a year is £1 + £1 = £2. If interest is paid twice a year
the value after a year is £1(1 + 0.50) 2 = £2.25. In the same way, if we
assume that the interest is paid at the end of each quarter, the capital
Series 147
value at the end of the first quarter is £(1 + 0.25), at the end of the
second quarter £1.25(1 + 0.25), at the end of the third quarter £1.25 2
(1 + 0.25) and at the end of the year £1.25 3 (1 + 0.25) = 1.25 4 = 2.44.
In each of these cases the value of £1 at the end of the year is given
by
(1 + 1/n)n
where n is the number of times the interest is paid during the year.
That is, n = 1, ( 1 + ~ r = £2
n = 2, ( 1 + ~) 2
= 1.52 = £2.25
n = 4, ( 1 + ~ r = 1.254 = £2.44
We can draw up a table of values of (1 + 1/n)n for different values

of n.
n 1 2 4 8 10 100 1,000 10,000
(1 + lint 2.00 2.25 2.44 2.56 2.59 2.70 2.717 2.718
As the value of n increases the value of the investment becomes

larger, but never exceeds £2.719. The limit to which the value tends as n
tends to infinity (ie becomes indefinitely large) is defined as e,
that is, e =lim ( 1 +

,._."" n
2.)n
and e is approximately 2. 7183. This is an irrational number. Its value
cannot be determined exactly no matter how many decimal places of
working are used. Also, it cannot be represented by the ratio of two
whole numbers.
By allowing n to approach infinity interest is being added to the
investment more and more frequently and can be regarded as being
added continuously.
Returning to the more general case of £1 invested at x% per annum
compound and the interest being paid n times a year, the value of the
investment at the end of the year is given by
( 1 +: r
To find the limit of this as n approaches infinity it is useful to let 1/k
= x!n to that n = kx. This means that
and letting n approach infinity means that, since x is unchanged, k must

approach infinity. The term in the inner brackets approaches the value e
so that the whole expression tends to ex. Thus ex is the value of £1 after
1 year when interest is paid at x% per annum and is compounded
continuously. Before discussing how ex might be evaluated we need to
examine the binomial theorem.
4.11 The binomial theorem

An expression of the form (a+ xt can generally be expanded as a
series by applying the normal laws of algebra. For example,
(a + x)2 = (a + x)(a + x)
(a + x) 3 = (a + x)(a + x)(a + x) = a3 + 3a2x + 3ax2 + x 3
The algebra becomes a little tedious as the power of the function is
increased but it is possible to obtain the expansion in the general case
from the binomial theorem. This can be stated as follows:
(a + xt = ncoan + nclan-lx + nc2an-2x2 + ... + nc,an-rxr
+ ... + ncnxn
where nco= 1
ncl = n
n(n - 1) n(n - 1)
nc2 = = ----
1 X 2 2!
n(n - 1)(n - 2) n(n - 1)(n - 2)
nc
3
= 1X 2X 3
= ----
3!
---
Series 149
and in general
1)(n - 2) ... (n - r + 1) n!
nc, = n(n - r!
= --:-:---:-
r!(n- r)!
r! is known as factorial r or r factorial and is the product of all the
integer values from 1 to r. For example, 4! = 1 x 2 x 3 x 4. The
exception to this is 0!, which is defined to be equal to 1, so that
n! n! n!
ncn = n!(n - n)!
=--=-=
n!O! n!
1
The coefficients of the terms in a binomial expansion form a sym-

metrical pattern, as can be seen in Fig. 4.1, known as Pascal's triangle.
Each number is the sum of the two numbers on the row above which are
closest to it. Thus, the coefficient 10 is the sum of the two numbers 4 and
6 above it.
2
3 3
4" /6 4
5 10 10 5
Fig. 4.1
The third line of the triangle gives the coefficients of (a + x)Z:

(a + x)Z = la 2 + 2ax + lx 2
The fourth line gives the coefficients of (a + x) 3 :
(a + x)3 = la3 + Ja2x + J_ax2 + lx3
The fifth line gives the coefficients of (a+ xt:
(a + x) 4 = la 4 + _1a 3x + Qa2x2 + _1ax3 + lx4
and so on.
The expansion can be divided throughout by the constant a with the
result
= an [ 1 + n (--;;x) + n(n-1)(x)
2! --;; + ...
2
+
n(n - 1) · · · (n - r + 1) ( x
-a
( x )n]
)r +···+-
r! a
The expansion in square brackets is particularly interesting when

n = -1 since
( 1 +: rl = 1+ (-1)(: ~ r
+· .. + (-1)(-2) · · · ~ + ...
1.2.3 · · · r a
This series contains an infinite number of terms and the coefficient
of every term is equal to unity, the signs of these coefficients being
alternately positive and negative. That is,
( 1+ ~ r = 1- ( ~ + ( ~ r-(~ r + ... + (-1Y( ~ r + ...

When the numerical value of x/a is less than 1 the terms become
smaller and smaller as x!a is raised to successively higher powers and
eventually tend to the value zero. That is,
as for
I: I< 1
In this particular case the sum of the series tends or converges to a finite
value as more and more terms are considered. Note the condition lx!al
< 1. This means the numerical value, or modulus, of xla must be less
than 1. In general it can be shown that a series converges if successive
terms tend to zero and also alternate in sign.
It can quite easily be seen that there is an infinite series which
converges to a finite value for (1 - x/a)- 1 when the numerical value of xla
is less than one, since this is a geometric series with a common ratio
between -1 and + 1:
1 +(-1)(-::_)+(-1)(-2)(-::_)2 +···
a 1.2 a
1 +(:)+(:r +··+(:r +···

In this case all the terms are positive.
Series 151
4.12 The exponential series

The binomial theorem can be used to evaluate e. We know that e is
defined by
e = lim ( 1 + 2_ )n
n-Xl n
and so using the binomial theorem to expand (1 + lint, i.e. with a= 1,

and x = lin,
n n(n - 1) 1 n(n - 1)(n - 2)

1 +-+ -+ +·. ·+-
n 2! n2 3! n3 nn
1+ 1+( : )( ~ 2!
+(: ~ ~ ;! +
Allowing n to tend to infinity, all the ratios such as (n - 1)/n,
(n - 2)/n, ... tend to the value 1 and hence
e = lim ( 1 + 2_ ) n = 1 + 1 + -1- + -1- + -1- + ...

~ n 2! 3! 4!
where the rth term in the expansion is ll(r - 1)!.
This expansion, known as the series expansion of e can be used to
evaluate e to any number of decimal places. For example to obtain e
correct to 4 decimal places we continue summing the terms of the
expansion until a term is less than 0.0001, as in Table 4.6.
The approximate value of e from the first 9 terms of the expansion
is 2.71828.
Similarly, ex can be expanded as a power series to give
x2 x3 X4 x'
ex = 1 + X + - +- +- + · · · +- + · · ·
2! 3! 4! r!
This expansion is known as the exponential series. It can be used to
evaluate ex to any degree of accuracy. Alternatively, values of ex are
included in standard books of logarithm tables and are available at the
press of a button on many pocket calculators.
TABLE 4.6
Number
of terms Last Contribution
summed term toe
1 1 1.00000
2 1 1.00000
3 1 0.50000
2!
4 1 0.16667
3!
5 1 0.04167
4!
6 1 0.00833
5!
7 1 0.00139
6!
8 1 0.00020
7!
9 1 0.00002
8!
Total 2.71828
To illustrate the use of the series expansion of ex suppose that

x = 0.05 so that the continuous interest rate is 5%. Then substituting in
the series expansion gives
0.05 2 0.05 3 0.05 4
= 1 + 005
.
+--+--+--+
2! 3! 4!
...
1.05 + 0.00125 + 0.0000208 + 0.0000002 + ...
= 1.05127
correct to five places after the decimal point. Thus if interest is paid
continuously at 5% per annum the value of p after one year is 1.05127p.
The value after t years is given by pe0 ·051 so that the value of £100 after
seven years is
Series 153
y
r = 1.0
r = 0.5
-2
Fig. 4.2
100eo.3s
0.35 2 0.35 3 0.35 4

100 { 1 + 0.35 ~ ~ ~ + ... }
= 100 {1.35 + 0.06125 + 0.0071458 + 0.0006252 + ... }
100 {1.419021} = £141.90
The general expression for the value of p after t years when the
continuous rate of growth is r% per annum is
vt =pert
and the graph is shown in Fig. 4.2 for p = 1 and r taking the values 0.01,
0.1, 0.5 and 1.0. As r increases the graph becomes steeper.
While the concept of continuous growth has been introduced
through continuous compound interest this is not a useful practical
application since financial institutions add interest at discrete time inter-
vals. However, there are many variables relevant in economics and
business studies which can be assumed to vary continuously. An obvious
one is the total population of a country, which changes as a result of
births, deaths and migration throughout the year. For example, if a city
has a population of 1.5 million now and is growing at a rate of 3% per
annum continuously, the population after t years will be
v, = 1. 5eo.o3r
so that after 10 years,
0.3 2 0.3 3 0.3 4

= 15
.
{ 1 + 0 3 +--+--+-- + ... }
. 2! 3! 4!
1.5 {1.3 + 0.045 + 0.0045 + 0.0003375 + ... }
2.025 million, and after 20 years
v20 = 1.5e06 = 2.733 million.
Technical progress occurs as a result of the diffusion throughout the
economy of knowledge about inventions and new processes of produc-
tion, and results in the growth of national income. This growth is
continuous and the previous formula applies. Thus if V, is the level of
national income in year t and r is the rate of technical progress then
V, = pe"
For example, suppose the index of national income is 100 and technical
progress occurs at a continuous rate of 2% per annum, then after 8 years
national income will be
V8 = 100e 016 = 117.35.
In section 4.6 the concept of discounting for discrete annual pay-
ments was discussed. If the discounting occurs continuously at the rate
r% per annum the discounted value is
so that £2,000 due 10 years from now with the discounting rate 8% per
annum has a present value of
V 10 = 2000e-0 8 = £898.66
Another application is to rates of decay - that is, negative growth
rates. For example, suppose a steel manufacturer has 1,000 tons of steel
in stock and this is deteriorating at a continuous rate of 4% per annum.
Series 155
The remaining stock after t years is given by

V, = 1000e-{) 04'
and after five years, V 5 = 818.73 tons.
Finally, we saw that the value of an amount p after t years when
interest is at the rate x and is compounded m times a year is
vlt = p(1 + x/m)mt
while for interest compounded continuously at the rate r% per annum,
and so it is always possible to find a continuous interest rate which is

equivalent to a discrete one. For example, if the discrete interest rate is
10%, compounded four times a year, the value after 5 years is
vl5 = p(1 + 0.10/4Y0
and the equivalent continuous interest value is
v25 = pe5r
For these to be equal, 1.02SZ0 = e5 '
This can be solved by taking natural logarithms (see section 4.14 for a
discussion),
or r 4(0.02469) = 0.0988
so that the continuous rate of 9.88% is equivalent to 10% compounded
four times a year.
4.13 Exercises
1. £150 can be invested in one of three ways. It can receive 6% per
annum simple interest, 5% per annum compound interest, or 4%
per annum compound interest with the interest compounded twice
a year. Which is the most profitable investment over a period of
(a) five years (b) ten years?
2. What is the value of £50 after four years if it is invested at 6% per
annum compound interest with the interest compounded three
times a year?
3. Use the binomial theorem to expand
r
(a) (1 + xY (b) (2- xY
(c) ( 3 + ~
4. Use the exponential series to obtain approximately the values of
(a) e0 · 1 (b) eo.s
(c) If and (d) e- 1
and compare your results with the values given in standard tables
of e or with the values given by a pocket calculator.
5. A city has a population of 2 million which is growing at the
continuous rate of 4% per annum. What is the projected popula-
tion of the city after (a) 5 years (b) 25 years? The government
decides to introduce policies to try to reduce the growth rate to 2%
per annum. What difference will this make to the population
projections?
6. Technical progress has increased the national income of a country
by an average of 3% per annum. The index of national income is
currently 125.0. What will it be in 30 years from now?
7. In a particular year 150,000 new firms are created. It is known that
the failure rate is 15% per annum. How many of the firms will
survive (a) 4 years, (b) 20 years?
8. What is the continuous rate of interest which is equivalent to 6%
per annum compounded twice yearly?
4.14 Logarithms
We know that
100 X 1000 = 100000

and since 100 = 102 , 1,000 = 103 and 100,000 = lOS then
102 X 103 = 105 •
That is, the product of 100 and 1,000 can be obtained by writing
these numbers as powers of 10 and then adding these powers together.
These powers are known as logarithms and are referred to as logarithms
to the base 10 because they are powers of 10.
Series 157
Therefore, log 10 100 = 2

log 10 1000 = 3
log 10 100000 5
or, more simply log 100 = 2
log 1000 3
log 100000 5
Now 10 = 101 and so log 10 =1
0.1 = 10- and so log 0.1 = -1
1
1 = 10° since 10 X 0.1 = 1 = 10 1 X 10-1 = 100

and so log 1 = 0.
Logarithms of any number in the range 1-10 can be found by using
logarithm tables.
For example, log 2 = 0.3010
log 5 = 0.6990
log 7.5 = 0.8751
log 3.001 = 0.4772
log 1.055 = 0.0232
Logarithms of numbers outside the range 1-10 can be found by
expressing the number as the product of a number within this range and
ten raised to some power. For example, 500 = 5 x 102 •
Now since log 5 = 0.6990, 5 = 100·6990
that is, log 500 = 2.6990,
or log 500 = log 5 + log 100 = 0.6990 + 2 = 2.6990
Similarly 0.003001 = 3.001 x 10-3 and so log (0.003001) = log (3.001)
+ log (l0- 3 ) = 0.4772 + (-3) = 3.4772.
Notice that we write this as 3.4772 and not as -3 + 0.4772
= -2.5228. The reason for this is that it is useful for manipulation if the
part of the logarithm after the decimal point is always positive.
The general rules governing logarithms are:
1. log ab = log a + log b
We have already used this rule above. For example, a = 5,

b = 100, ab = 500.
log 500 = log 5 + log 100 = 2.6990.
2. log ( ~ = log a - log b
For example, a = 5, b = 2,
log (t) = log 5 - log 2 = 0.6990 - 0.3010 = 0.3980
that is, log 2.5 = 0.3980
3. log an = n log a
This follows directly from Rule 1.
For example, log (a 2) = log (a . a) = log a + log a = 2 log a
Similarly, log (5 3 ) = 3 log 5 = 3(0.6990) = 2.0970.
These rules can be used to reduce the arithmetic required in our
calculations.
Example 1
Evaluate 450/1.056
By Rule 1, log 450 =log (100 x 4.5) = log 100 + log 4.5
= 2 + 0.6532 = 2.6532
By Rule 3, log (1.05 6) = 6 log (1.05) = 6(0.0212)
= 0.1272
By Rule 2, log ( -450) = log 450 - log (1.05 6)

1.056
= 2.6532 - 0.1272 = 2.5260
We now need to know the number for which 2.5260 is the logar-
ithm. We can find this either by looking in the body of the logarithm
tables for 0.5260, or by using tables of antilogarithms which give the
required number directly.
antilog 0.5260 = 3.357
and therefore, antilog 2.5260 = 102 x 3.357 = 335.7; that is,
450
= 335.7
Series 159
Example 2
Evaluate 50(1.1 10)/3250.
We first of all evaluate 1.1 10 , then 50(l.P0 ) and finally divide by 3250.
log (1.1 10) = 10 log 1.1 = 10(0.0414) = 0.4140
log 50= log (10 x 5) = log 10 + log 5 = 1.6990
. . log [(50(1.1 10)] =log (1.1 10) +log 50= 0.4140 + 1.6990
= 2.1130
log 3250 =log (1000 x 3.25) = 3.5119
50(1.1 10)
. . log 3250 =log [(50(1.1 )]
10 - log (3250)
= 2.1130 - 3.5119
= - 1.3989 = - 2 + 0.6011 = 2.6011
50(1.1 10) ) -
Hence ( = antilog 2.6011
3250
= 10-2 antilog (0.6011)
= 10- 2 (3.991)
= 0.03991
Throughout this section all the logarithms used have been to the
base 10. This is generally convenient in numerical work. However, in
Chapter 5 we are going to see that for most theoretical work it is more
convenient to use logarithms to the base e. Such logarithms are known
as natural, hyperbolic or Naperian logarithms and are written as log.,x or
In x. The rules of logarithms stated above apply to logarithms with any
base. To change the base to which logarithms are measured an ad-
ditional rule is used:
4. For a positive real number x, logPx = logPq ~
Since we are mainly interested in logarithms to the base 10 and the

base e we have the special case
log.,x = loge10 log 10 x
or
which allows conversion from one base to the other.
In discussing logarithms we have not mentioned where the tables

referred to come from. In section 5.15 we will use Taylor's theorem to
show that, for -1 < x < 1,
loge (1 + x) =X-
r r X4
2 + 3-4 + 5 + · · ·
r
which is an infinite series with successive terms becoming smaller so that
the sum converges rather than tending to infinity. As it stands, this
expansion is not particularly useful because of the restriction that
- 1 < x < 1 and we need to be able to find the logarithm of any number.
Now if we replace x by -x each time it occurs in the above expansion
then,
xz x3 x4 xs
loge (1 - x) = -x - 2 - 3 - 4 - 5 - · · ·
where again -1 < x < 1. Therefore by subtraction and using rule 2 for
logarithms above,
loge(1 + x) - ~ - x) = + x)/(1 - x)}
~
x3 xs
= 2{ X + 3 + 5 + ... }
since all the even powers of x cancel out. Therefore if we wish to find the
natural logarithm of a number, a, say, we find x such that
1+x a-1
a = -- and solving, x = --
1-x a+1
For example, to use the series expansion to evaluate ~ put a = 10 so
that x = 9/11 or 0.81818. Then,
logelO = 2{0.81818+ 0.18257 + 0.07333 + 0.03506 + 0.01826
+ 0.01000 + 0.00566 + 0.00329 + 0.00194 + 0.00116
+ 0.00070 + ... }
= 2{1.15015} = 2.3003
Here the calculations are to five places after the decimal point and extra
terms were added until the last one was below 0.001, so that the result
should be accurate to at least two places after the decimal point. A
calculator gives the precise value as 2.3026 and so the result is close. In
this example the convergence of the series expansion has been slow.
As another example, to find ~ we have
X = (3 - 1)/(3 + 1) = 0.5
Series 161
0.53 0.5 5 0.5 7 0.5 9

and ~ =2{ 0.5 + - 3-+-5-+7+9+ · · ·}
= 2{0.5 + 0.041667 + 0.00625 + 0.00111 + 0.00022 + ... }
= 2{0.54924} = 1.09849
using the first five terms. A check with a calculator shows that the
accurate value is loge3 = 1.0986.
Using these results we can obtain
log 103 = ~ = 1.0986/2.3026 = 0.4771
which is the value given in logarithm tables.
The use of natural logarithms helps in some problems concerning
continuous growth. For example, suppose that the national income, Y,
is growing at a continuous rate of 4% per annum. How long will it take
for Y to double?
Here we have Y 1 = Y 0 e0 ·00
where Yo is the current value of Y and Y 1 is the final value of Y. We
require the value of t for which Y 1/Y0 = 2. Therefore,
2 = eo.04t
Taking natural logarithms,
0.69315 = 0.04t
since ~ = 1 and so t = 17.3 years.
Finally, since in many applications in economics natural logarithms
are used, the graph of y = logex is shown in Fig. 4.3. As x increases,
y increases but at a decreasing rate.
4.15 Exercises
1. Use the series expansion of ~ x)/(1 + x)} to obtain loge0.1
and ~ Hence obtain log100.1 and log 10 100 and compare the
values with those in logarithm tables.
2. The population of a region is 2.4 million and it is growing at a
continuous rate of 5% per annum. How many years will it be
before the population reaches (a) 3 million (b) 5 million?
3. Last year 100,000 new firms were created in a particular country. If
new firms fail at a continuous rate of 12% per annum, how long
will it be before there are only 50,000 firms?
4. Pareto's law of distribution of income states that the number of
y
~
10 X
Fig. 4.3
individuals n, from a population of size N, whose income exceeds

the value x is given by n = N/x 15 • The population of the UK is
approximately 55,000,000. According to this law, how many indi-
viduals are expected to have incomes above (a) £10,000
(b) £100,000?

1. For the following progressions, identify the type of series, give
the expression for the nth term, obtain the sum of the first 20
terms and, if it is finite, the sum to infinity:
Series 163
(a) 100, 105, 110, 115, 120, ...

(b) 100, 99, 98, 97, 96, ...
(c) 1, 2, 4, 8, 16, ...
(d) 256, 128, 64, 32, 16, ...
(e) 100, -90, 81, -72.9, 65.61, ...
2. The sum of £1,000 can be invested for 15 years with
(a) simple interest at the rate of 10% per annum
(b) compound interest at the rate of 8% per annum, added once
each year
(c) compound interest at the rate of 6% per annum, added every
six months
(d) continuously added interest at the rate of 5% per annum.
Which of these gives the highest return?
3. A firm can borrow money at 15% per annum compound interest
and has the opportunity of investing in a project which has the
following costs and receipts:
End of year 1 2 3 4
Costs 50,000 20,000 5,000 5,000

Receipts 10,000 25,000 55,000 20,000
Is it worthwhile investing in the project? What is the internal rate

of return of the project?
4. A lady decides to pay £100 into a fund each year on her grand-
daughter's birthday (starting at age 5). If compound interest is
received at 7% per annum, how much would the granddaughter
receive if given the fund on her twenty-first birthday?
5. What would be the annual payments on a housing loan of £75,000
payable back over 25 years if interest is charged at 11% per
annum? If repayments were made at the rate of £6,000 per
annum, when would the debt be paid off?
6. A firm has a computer which is expected to cost $10,000 to
replace in 8 years' time. If the interest rate is 9%, how much
should be saved each year in order to be able to replace the
computer?
7. Use the binomial theorem to expand (2 + 3x) 9 and (x + l!x) 10 and
hence evaluate 2.39 and 4.25 10 •
8. Use the first six terms of the exponential series to evaluate appro-
ximately e1.s, e-u and e- 4 and compare the results with the values
given by a calculator.
9. In 1971 the index of retail prices in the UK was 100. In 1985 the
value was 463.3. What is the implied rate of inflation, assuming
(a) it is compounded annually and (b) it occurs continuously?
10. Use the series expansion of log.,{(1 - x)/(1 + x)} to obtain log.,S
and loge10.
Chapter 5
Differential Calculus
5.1 Introduction
In Chapters 1 and 3 we saw that an equation relating two variables
could be represented by a graph. For example, if y is the total cost of
production and x the quantity produced then the equation
y = 100 + 3x
can be represented by a straight line, whilst
y = 100 + 2x + Jr x 2
can be represented by a curve. The graphs of these equations allow the

cost to be determined for permissible levels of production.
But the equations also provide additional information which is very
important. They tell us the change in the costs which must be incurred
for any given change in production level. This can be determined
directly from the graphs by considering a change of x from x, to x 2 •
In the case of the linear cost function the cost changes from y, to y2
(Fig 5.1). The average cost of each extra unit of production between the
values x, and x 2 is given by the quotient
Y2- y,
This is the slope of the line and it does not depend upon the value of
x, or x 2 , because the slope is constant for all values of x for a linear
function.
This is not so for the quadratic function as can be seen from Fig 5.2.
It is clear that with this curve the slope between P, and P2 depends on
where P, and P2 are, and that even if P, is held constant, the slope will
depend on the position of P2 • If the cost changes from y 3 to y 4 when the
quantity produced changes from x, to x 2 , then the average cost of each
extra unit is
165
y
Y2 -----------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
x, x> X
Fig. 5.1
y4- y3
x2- xl
This is the slope of the straight line joining P 1 and P2 and is not that of
the curve. We therefore define the slope at a point on a curve as being a
tangent to the curve at that point, where a tangent is a line which touches
but does not cross the curve. The tangent at P 1 is shown in Fig 5.3, which
is an enlargement of Fig 5.2 with P 1 and P2 close together.
To indicate that only small changes in x andy are being represented
we use the symbols c5x and c5y, which are not c5 times x and c5 times y. The
slope of the straight line joining P 1 and P2 is
by
c5x
But this is not the slope of the curve at P 1 (the tangent at P 1). However,
if c5x becomes smaller then the slope of the straight line P 1P2 becomes
very close to the slope of the tangent at P 1 •
Differential Calculus 167
y
Y4 --------------
x, x2 X
Fig. 5.2
. I
____ I
YI,1
r-- bx --1
Fig. 5.3
As <5x ~ 0, we define limit (<5y/<5x) as the derivative of y with respect

to x, and we denote it by the symbol dyldx. That is,
dy= h. m -
- (by)
dx 6x--+O <5x
Fig. 5.4
The process of finding dy!dx is known as differentiation and dyldx is

sometimes called the differential coefficient of y with respect to x, rather
than the derivative. We now discuss some methods of differentiation
which will enable us to determine the slope of a function at any point.
In the case of the linear equation y = 4x + 2, if x increases to
(x + ox), y increases to (y + by) and if the two points with coordinates
(x, y) and (x + ox, y + by) lie on the line (Fig 5.4) we must have
y = 4x + 2 (1)
y + by = 4(x + ox) + 2 (2)
Subtracting (1) from (2), we obtain
(y + oy) - y = 4(x + ox) + 2- (4x + 2)
oy = 4x + 4 ox + 2 - 4x - 2
=4 ox
.. by =4
ox
and dy = lim ( by ) =4
dx lix--+0bx
This is the expected result, because we know that the slope is equal
to 4 for all positions along the line. That is, it is independent of the value
of x.
In the case of a quadratic equation the same procedure can be
adopted.
For example if
y=x2 +2x+1 (3)
then y + by = (x + bx )2 + 2(x + bx) +1 (4)
Subtracting (3) from (4)

y + by - y = (x + bxY + 2(x + bx) + 1 - (x2 + 2x + 1)
= x2 + 2x bx + (bx) 2 + 2x + 2 bx + 1 - x2 - 2x - 1
= 2x bx + (bx) 2 + 2 bx
.. by=(2x + 2) bx + (bx) 2
~~ = (2x + 2) + bx
Now as bx decreases in size the second term on the right-hand side

of this equation becomes of less and less significance, and in the limit
dy
-=lim
dx
-
llx--+0
(by)
bx
= 2x + 2
This means that the slope of the curve at any point is a function of
the point itself. For example,
when x = 1 dy =2 +2 =4
dx
when x = 2 dy =4+2=6
dx
5.2 Some general rules

The above procedure can be used to develop general rules for
obtaining derivatives. For example, in the case of the general quadratic
y =a+ bx + cx2
(y + by)= a + b(x + bx) + c(x + bx) 2
(y + by) - y =a+ bx + b bx + cr + 2cx bx + c(bxf- (a + bx + cr)
.. by= b bx + 2cx bx + c(bx) 2
~~
.. ~
that is, if y = a + bx + ex\ then dy/dx = b + 2cx.
The derivative is in fact the sum of the derivatives of the three
terms taken separately: that is,
if y =a+ bx + cx2
then dy d d d
dx = dx (a) + dx (bx) + dx (cx2)
and in general the derivative of any polynomial in x is equal to the sum
of the derivatives of each term considered separately.
It can be seen that the derivative of the constant a is zero, that of
the term bx is b and that of cx 2 becomes 2cx. In general the derivative of
x" is equal to nx"- 1 , which can be proved as follows:
If y =ax",
y + by= a(x + bx)"
= a(x" + "C 1x"- 1 bx + · · · + "Cn(bx)")
from the binomial expansion (section 4.11).
y + by - y = a[x" + "C1x,._ 1 bx + · · · + "Cn(bx")] - ax"
or by = a["C1x"- 1 bx + "C2 x"- 2 bx2 + · · ·]
Hence by= a"C x"-1 + ~ bx + ...
bx 1
·· dy =lim by = anx"- 1 +0
dx c5x--+o bx
So if y =ax", dyldx = anx"- 1 •
Example 1
If y = 4x6 , n = 6, a = 4, so
Example 2
If y = x-Z, n = -2, a= 1, so
-dy
dx
= (- 2)x- 3 = --2
x
3
THE PRODUCf RULE

If y = uv, where u and v are both functions of x, then the derivative
dy = v du + u dv
dx dx dx
This is obtained by the same method as used above. Since u and v are
functions of x, a change in the value of x to x + ox changes both u and v.
Let the new values be u + ou and v + ov. The value of y also changes,
and we let this change be oy.
Hence, y=uv
and y + oy = (u + ou)(v + ov)
or y + oy = uv + v ou + u ov + ou ov
so that oy = v ou + u ov + ou ov
and oy = v ou + u ov + ou ov
ox Ox ox ox
Now as ~ lim oy - dy lim ou = du

ox - dx ox dx
ov dv
lim - = - lim ov ~
ox dx
Taking limits of the expression gives
dy = v du + u dv + du (O) = v du + u dv
dx dx dx dx dx dx
An example of the use of this is now given.

y = (x + 2)(x2 + 3)
Let u = x + 2 and v = x2 +3
then du =1 and dv = 2x
dx dx
so that : = (x2 + 3)(1) + (x + 2)(2x)

= x2 + 3 + 2r + 4x = 3x2 + 4x + 3
THE QuoTIENT RULE

If y = u/v where both u and v are functions of x, then
dy v(du/dx) - u(dv/dx)
dx v2
This can be shown by allowing x to change to x + Dx, and letting the
corresponding changes in u, v and y be Du, Dv and tJy. Then
u
y=-
v
u + Du
y + tJy=---
v + Dv
Subtracting, we have
u + Du u vu + v Du - uv - u Dv
{Jy = -- = - - - - - - - -
v + Dv v (v + Dv )v
v Du - u Dv
v2 +vDv
Dividing by Dx, we obtain
{Jy _ v(tJu/tJx) - u(tJv/tJx)
Dx - v2 + v Dv
{Jy dy Du du Dv dv
As Dx--+ 0, v Dv--+ 0.
Dx--+ dx' Dx--+ dx' Dx--+ dx'
dy = v(du!dx) - u(dvldx)
Hence
dx v2
Example 3
x+2
y=--
2x
Let u = x + 2 and v = 2x
du = 1 dv = 2
dx dx
dy _ (2x)(1) - (x + 2)(2)
dx- (2x) 2
2x- 2x- 4 -4 -1
= -=-
4x2 4x2 XZ
THE FUNCTION OF A FUNCTION OR CHAIN RULE

This rule states that if y is a function of u, and u is a function of x,
then
~
The proof of this rule is as follows
Let y = f(u) and u = g(x)
then y + Dy =f(u + Du) and u + Du = g(x + Dx)
where Du is assumed to be non-zero.
by _f(u + Du) - f(u)
Hence
Dx- Dx
= (f(u + Du) - f(u) )Du

Du Dx
and
~ ( ~~ ) = ~ = ( ~~ ) ( : )
Example 4
y = (1 + xt
In this case, if u = 1 + x, then we have y = u4 and u = 1 + x, so that y is

a function of u and u is a function of x.
Now du =1
dx
dy dy du
- = - - = (4u 3 )(1) = 4(1 + x) 3
dx du dx
LOGARITHMS
Functions involving logarithms can be differentiated using some of
the rules of section 4.14. Initially we will not specify the base to which
the logarithm is taken. To differentiate y = log x, we allow x to increase
by ox and let the corresponding increase in y be oy so that
y + oy = log (x + ox)
Subtracting, we obtain
y + oy - y =log (x + ox) - log (x)
or oy = log (x : ox) = log ( 1 + ~)

Now we divide by bx, and also multiply the term on the right hand side
by (x!x) or by 1. The latter leaves the value of the right hand side
unchanged.
~~ = ~ · ~ log ( 1 + ~ ) = ( ;x )~ log ( 1 + ~ )
= ~ [ log ( 1 + ~ r6x ]
Let us consider the behaviour of the expression in the outer brackets. It
is convenient to let n = x/ox. Since x is fixed, as ox- 0, n - oo
Now ~ = ~ [ log ( 1 + ~
Taking the limit as ox -o gives
r]
dy = .!. log [ lim ( 1 + .!.
dx X n---+
00 n
)n ]
But the limiting value of this expression is defined as e, and so
dy =.!.log (e)
dx x
If we now specify that we are using logarithms to the base e, since

log (e) = 1 we have
dy - 1 = lo& (x)
dx- x when y
In general, from using the function of a function rule,

if y = lo& (u)
dy _ 1 du
dx- udx
since dy = .! .
du u
Example 5
y = loge (2x + 3)
u = 2x + 3, duldx =2
Hence dy - 1 (2) - 2
dx (2x + 3) 2x + 3
We will assume for the rest of this book that whenever we are
concerned with differentiation, logarithms use the base e. If this were
not the case, a factor logn (e) would need to be included, where n is the
base of the logarithms.
ExPONENTIAL FuNCTIONS
These can be differentiated most easily by converting to logarithms.
For example, y = ~
Taking logarithms, we obtain
log y = log (ex) = X log e = X
The derivative of log y is
~ (log y) = ! ( dy )
dx y dx
This arises from the function of a function rule, since y is a function of x,
and log y is a function of y.
Hence, .! dy 1
ydx
or dy = y
dx
This shows that if y = eX, dyldx = ex which is a rather surprising result.
The same result is obtained by differentiating the series expansion of ex
(section 4.12) term by term.
Other exponential functions can be differentiated by the same
method. In particular,
if y = eax
log y = ax log e = ax
Differentiating, we have
1 dy
--=a
y dx
Hence dy = ay = ae=
dx
Example 6
We have a = 4, and so
TRIGONOMETRIC FUNCTIONS
A number of these are differentiated from first principles in Appen-
dix A, Section A.5 and we summarise the results in Table 5.1.
5.3 Exercises
Find dy!dx for the following functions:
l.y=3x-4
2. y = x 2 - 3x + 3
3. y = 3x4 - x 3 + x 2 + 25x
x4 x3 1
4. y = 6x3 - 4 + T - x2
5. y = (2x + 3x
2 - 1) ~- 3x + 2)
TABLE 5.1
Summary table of derivatives
dy dy
y y
dx dx
1 du
a 0 log u
udx
axn anxn-! eax aeax
du dv
uv v- + u- sin ax a cos ax
dx dx
-u v (du/dx) - u (dvldx)
v cos ax -a sin ax
v2
dy du a
f(u) tan ax
du dx cos 2ax
a is a constant, u and v are function of x.
6. y = (2x + 4) (3x + ~
4x 2 + 4
7. y
(x + 3x + 2)
2
8. y = (2x + 3) 5
9. y = log (2x + 3)
10. y = x log (2x 2 + 3x - 5)
11. y = 3e2x + 4e-Jx + x 2ex
12. y = sin 2x + 3 cos Sx - tan 3x.
5.4 Elasticity of demand

The quantity demanded of some goods is much more sensitive to
changes in price than is the quantity demanded of others and this
difference is clearly very important. A measure of sensitivity to price

change can be obtained by expressing the percentage change in the
quantity demanded of a good in terms of the percentage change in price.
For very small changes in price this ratio is the price elasticity of demand:
E = (dq!q) X 100 = dq p_
D (dpfp) X 100 dp q
If the demand curve is known in the form q = f(p), then it is a
simple matter to differentiate and to substitute the values of dq I dp in the
above equation. For example, if
q = f(p) = 100 - p - p2
~ = - 1 - 2p = - ( 1 + 2p)
Ev = - (1 + 2p) !!_
q
If the present price is £5 then the quantity demanded is
q = 100 - 5 - 25 = 70
and ED= - (1 + 10) ~
= -0.8 approx.
This means that the demand decreases by approximately 0.8% for a
1% increase in price or, conversely, increases by 0.8% for a 1% de-
crease in price. It follows that, in this case, a decrease in price results in
a decrease in total revenue because
total revenue = price x quantity = pq
In this case the demand is said to be inelastic at a price of £5.
If, however, the present price is £6,
then q = 100 - 6 - 36 = 58
and Ev= -(1 + 12) 568
= - 1.3 approx.
Thus a 1% decrease in price results in a 1.3% increase in the
quantity demanded and if this is supplied it results in an increase in the
total revenue. The demand is then said to be elastic at a price of £6.
When the elasticity of demand is unity, a rise or fall in price results

in the same percentage decrease or increase in quantity demanded and,
therefore, leaves the total revenue the same.
Other elasticities are of importance in economics and these are
determined in a similar way to the above, e.g. we define
. . . percentage change in quantity demanded

mcome elastzctty of demand = . .
percentage change m mcome
which can be expressed mathematically as
(dq/q) X 100 dq I
or
(dJ/J) X 100 dlq
To determine this elasticity it is necessary to know the relationship
between the quantity demanded of a good and the income, that is,
q = f(I)
It is, therefore, possible to determine other forms of elasticity if the
relationship between the variables is known.
One special case which is of interest in economics occurs when the
elasticity is a constant. For a demand curve, we know that if total
revenue is unchanged when the price changes then the price elasticity is
-1. But for total revenue to be a constant we have pq = c, where cis a
constant. Therefore,
q = c/p = cp-!
d ( 1) - 2
and dq = (-1)cp- 2 so that E = _!!_ p = - cp P = - 1
dp v dp q cp-t
More generally, if the demand curve is qpn = c then the price elasticity is
- n. This can be shown using the same method as above, and alterna-
tively by taking logarithms (to the base e) to give
log q + n log p = log c
or, log q = log c - n log p
Differentiating with respect top,
_!_ dq = 0 - !!._
q dp p
and hence E =dq !!_= -n

D dq q
This method also indicates that if q is raised to a power, for example,
qmpn = c then taking logarithms gives
m log q + n log p = log c
and the elasticity is - n/m.
5.5 Marginal analysis

We have, in the differential calculus, a method for determining the
slope of a function. In terms of the total cost curve example, the slope at
any point tells us the marginal or extra cost incurred if the volume of
production is increased by a very small amount or, conversely, the cost
saved if the volume of production is decreased by a very small amount.
The same principles also apply in other uses of marginal concepts,
such as marginal revenue (obtained from a total revenue function), the
marginal productivity of labour (obtained from a production function),
and the marginal propensity to consume (obtained from a consumption
function).
For the linear cost function
dy
y = 100 + 3x -=3
dx
The marginal cost is equal to 3 and is independent of the volume of
production.
This can be checked by calculating the cost at any two levels of
production which are one unit apart. For example,
when X= 3, y = 100 + 9 = 109
X= 4, y = 100 + 12 = 112
or when X= 15, y = 100 + 45 = 145
X = 16, y = 100 + 48 = 148
Both these sets of costs differ by 3 units.
For the quadratic function
y = 100 + 2x + fox 2
dy
= 2 + kx
dx
In this case the marginal cost is a function of the output. For

example,
when X= 50, dy = 2 + 10 = 12
dx
dy
and when X= 100, - =2+20=22
dx
Therefore, the marginal cost is positive and increasing as output in-
creases.
However, if, as is often the case, more efficient use can be made of
the variable factors of production as the level of output is increased,
then it is reasonable to expect the marginal cost to be positive but to
decrease with output. To satisfy this requirement the derivative could be
of the form
dy
- = a - bx
dx
where a and b are positive constants. Then as xis increased the marginal
cost decreases.
Applying the same ideas to total revenue (TR), and using the fact
that TR = pq where p and q are related through the demand function,
marginal revenue is found by differentiating TR with respect to q.
Notice that the demand function allows p to be replaced by a function of
q so that TR is expressed solely in terms of q. It is also possible to use
the demand function to replace q by a function of p, so that TR depends
on price, but when marginal revenue is referred to it is normally in the
context of a change in output rather than price. To illustrate the deter-
mination of marginal revenue, suppose the demand function is
p = 250- 2q
then TR pq = (250 - q)q = 250q - q 2
dTR
and so MR = - = 250 - 2q
dq
Therefore, as q increases, MR falls and if q > 125 MR is negative so that
as output increases above this level, total revenue declines.
It should be clear that marginal revenue and elasticity of demand
are related. Since TR = pq then differentiating gives
dTR dp (
dq=p+qdq=p 1 +pq dp)
dq
But marginal revenue is given by
MR = dTR
dq
and elasticity of demand
p dq
ED=--
q dq
and hence
MR =p ( 1 + ;J
When Ev = -1, MR = 0, that is, revenue is constant for changes in q.
When ED = oo, MR = p, that is, when the firm is a competitor in the
product market and the demand is perfectly elastic, the marginal rev-
enue equals price. ·
Bringing together revenue and costs allows net revenue (NR) or
profits (.n), to be defined by
NR = n = TR- TC
Continuing with the previous example, and assuming the cost function
TC = 100 + 3q + 0.5q 2
then NR = (250q - q 2) - (100 + 3q + 0.5q 2)
= 247q - 1.5q2 - 100
The marginal NR or marginal profit is
dNR
--= 247- 3q
dq
and this is positive when q < 82.33.
A production function relates the level of output from a production
process to the level of input. Let q be the level of output and L be the
number of units of labour used; then, if other inputs such as capital
services and energy are ignored the production function is q = f(L). As
previously, the marginal product is found by differentiating q with
respect to L. For example, suppose that the production function is

q = 4U- 20L
then the marginal product of labour is
dq = 8L- 20
dL
and in this example the marginal product is positive when L > 2.5.
Our final application of marginal analysis is with the consumption
function which relates consumption expenditure to the level of income.
For example, a simple consumption function which is commonly used in
basic economics is
C =a+ bY
where a and b are constants. The marginal propensity to consume
(mpc), which is the amount out of a unit of extra income which is spent
on consumption, is given by
dC
mpc=- = b
dY
and here this is a constant. The average propensity to consume (ape) is
CIY and here,
a+ bY a
ape= =- +b
y y
Therefore, for this linear consumption function, if a is positive the ape is

greater than the mpc.
In this section we have illustrated some of the many uses of mar-
ginal analysis in economics. Perhaps the most important application,
however, is the determination of maximum and minimum values and
this is the next topic we consider.
5.6 Exercises
1. Find the marginal-cost and average-cost function from the follow-
ing total-cost functions:
(a) TC = 4q 3 + 2q2 - 25q
(b) TC = (q 3 - 3q)(16 + 5q)
(c) TC = 25 + 6qe 2q
(d) TC = (3 log q + 5)(q2 - q)

2. Find the elasticity of demand when p = 10 for the following
equations:
(a) p + 2q = 50
(b) qp 2 = 400
(c) q- 35 = -3p
3. Determine the marginal-revenue function associated with each of
the following demand curves (p =price, q =quantity demanded):
(a) p = 200 - 3q (b) pq = 100 (c) p = 250 - e4q
4. A firm has a cost function TC = 2q2 + 20q + 300 and its demand
function is 4p + q = 260. Determine (a) the marginal-cost function
(b) the marginal-revenue function (c) the marginal-profit function
(d) the value of q for which marginal revenue equals marginal cost.
5. The demand function for a product is p = 177 - q2 and the cost
function is TC = 30 + 2q + 2q2 • Determine the profit function
and hence the marginal-profit function. For what level of output is
marginal profit zero?
6. If a consumption function is given by C = a + bY + c Y2 determine
the average propensity to consume and the marginal propensity to
consume.
5. 7 Maxima and minima

A total-cost function of the form
y = 100 + 2x + toX2
tells us the total cost of manufacture in terms of the total output. The
average cost per unit can be obtained by dividing throughout by the
quantity produced:
total cost 100 + 2x + ~
average cost per unit = - - - - =
quantity X
100 1
=- + 2 +loX
X
This is an equation showing that the average cost is very high when only
a small number of units are made. Initially this value decreases with an
Total cost
Average cost
Fig. 5.5
increase in output, but eventually the increasing marginal costs cause

the average cost per unit to rise. There is, therefore, some value of
output for which the average cost per unit is a minimum.
A rough sketch of the graphs of the functions shows up the situation
more clearly (Fig 5.5).
Then for
dAC <0 (i.e. the slope is negative)

dx
X> Xmin ~ >0 (i.e. the slope is positive)
X= Xmin
dAC =O
dx
Therefore, the value of output for which the average cost per unit is
a minimum can be found by the differential calculus.
100 I
AC = - + 2 +!oX
X
= 10ox- 1 + 2 + -f-o.x
d(AC) -100 +
-1 00x- + w =
2 I _!_
2 10
dx X
d(AC) =O 100
x2 +fo=O
when
dx
that is,
and or X = v'(1000) = 31.623

This is the value of output for which the average cost per unit is a
minimum. For an output level of 31 or 32 units,
100 31
X = 31 and AC = - + 2 +- = 8.3258
31 10
X= 32 and AC = 100 + 2 + 32 = 8.3250

32 10
If it were possible to obtain an output of 31.623 units then the
average cost per unit would be
AC = ____!22._ + 2 + 31.623
31.623 10
= 3.1623 + 2 + 3.1623 = 8.3246
The shape of the curve is, therefore, fairly flat about the minimum
point and for most practical purposes the difference between the aver-
age cost per unit at these levels of output would be considered to be
insignificant.
It is interesting to note that at the minimum the term 100/x is equal
in value to the term x/10. The former is equal to the share of the fixed
costs which must be carried by each unit of output and, therefore,
decreases with output. The latter corresponds to that part of the mar-
ginal cost function which depends on the quantity produced. One term,
therefore, decreases with output and the other increases, and the mini-
mum average cost occurs when the two terms are equal. At this point
the average cost per unit is equal to the marginal cost.

Total cost, y = 100 + 2x + f-oX 2
marginal cost, dy = 2 + tx
dx
For an output of 31.623 units
marginal cost = 2 + ! x 31.623 = 8.3246
which is the value of the minimum average of cost per unit.

This follows from the fact that for values of output less than xmin the
average cost is falling and this can only happen if the cost of producing
each extra unit is less than average cost of producing all previous units;
that is, when
0 <X< Xmin• MC<AC
When the average cost is increasing, this is due to the fact that the
cost of producing each extra unit is greater than the average cost of
producing all previous units; that is, when
X> Xmin• MC>AC
Then at the minimum the two must be equal; that is, when
X= Xmin• MC=AC
In this example the minimum value was determined by setting the
derivative equal to zero. For a curve which reaches a maximum value
the same sort of arguments apply with the result that:
for x < Xmax the slope is positive
x > Xmax the slope is negative
x = Xmax the slope is zero
and at the maximum value the derivative is zero. It is therefore necess-
ary to decide whether a point for which the derivative is zero corre-
sponds to a maximum or a minimum value. For the average cost function
the graph shows that there is a minimum value. However, it is not
always easy to sketch the graph and it is not necessary since the differen-
tial calculus can be used to distinguish between values corresponding to
maximum and minimum positions.
To do this it is simplest to differentiate the function a second time
Fig. 5.6
following the same procedure as was used to obtain the first derivative
dy I dx. This can best be illustrated by reference to two examples.
In the case of the linear function
y = 100 + 3x, dy = 3
dx
This is constant for all values of x and a graph of dy I dx againt xis as
shown in Fig. 5.6.
The slope of this line is zero at all points. The derivative of dy I dx is
therefore zero. If we write y' for dy I dx, then
dy dy'
y' = - = 3 and - =0
dx dx
The derivative of y' is then the second derivative of y and can be
written in the alternative forms
- - -
dy' - - -
d (dy) - -
d 2-
y
dx dx dx dx 2
If we let y" =d 2 y I dx 2 , then the third derivative can be formed in a
similar way:
and in general the nth derivative is

dny
dxn
In the case of the quadratic equation
y =a+ bx + cx 2
dy
then =b+2cx
dx
d2y
- =2c
dx2
and d 3y I dx 3 and all higher-order derivatives are equal to zero.
In general a polynomial of degree n, i.e. an equation in which the
highest power of the variable is n, has n derivatives not equal to zero
and the (n + l)th is always equal to zero.
For the purposes of the analysis we can restrict ourselves to the first
two derivatives because they are usually sufficient to determine whether
a function has a maximum or minimum value and to distinguish between
the two.
For the quadratic
y = a + bx + cx 2
the slope at any point of the curve is given by
dy
-=b+2cx
dx
This will be positive or negative depending upon the values of b, c
and x. Let us consider the two cases of the general quadratic equation
over the whole range of possible values of x:
(a) when cis positive and b is negative
(b) when c is negative and b is positive
The general shape of these function is shown in Fig 5. 7.
Fig 5.7(a) has a minimum point, i.e. for x = xmin the function has a
value which is lower than the value for all other values of x.
Fig 5.7(b) has a maximum point, i.e. for x = Xmax the function has a
value which is higher than the value for all other values of x.
y y
X min X X
(I) (b)
Fig. 5.7
The slope of these curves depends upon the value of x and the sign
of the slope depends on the sign of the constant c. This can be seen more
clearly if the graph of dy I dx against x is drawn (Fig 5. 8).
The slope of the curve at a minimum and a maximum point is zero
and therefore the graph of dy I dx crosses the axis of x at this point. But
in Fig 5.7 (a) for
X< Xmim
dy < 0
dx
X> Xmim
dy > 0
dx
dy dy
dx dx
(a) (b)
Fig. 5.8
The slope of the graph of dy I dx against xis therefore positive. In Fig 5. 7

(b), for
dy > 0
dx
dy < 0
dx
Therefore, the slope of dy I dx against x is negative.
The slope of these two graphs can, of course, be determined by
differentiating the linear function
dy
-=b+2cx
dx
with the result
The graph of this is illustrated in Fig 5. 9 for both signs of the constant c.
d2y
dx 2
~
~ ~
(a) (b)
Fig. 5.9
It follows that a quadratic function has a maximum or minimum at

the point where
dy
- =0
dx
and that this point is a maximum if d 2 y I dx 2 is negative and a minimum if

d 2y/dx 2 is positive.
The use of these rules is demonstrated in the following examples:
Example I
y = 100 + 2x - -fux 2
dy
= 2- !x
dx
d2y
and = ~
This is negative and the function has a maximum at the point where
dy 1
- = 2- 5x =0
dx
that is, where x = 10 and the maximum value of y is given by y = 100
+ 20- ~ = 110.
Example 2
Given that total cost y = 100 + 2x + -fuxZ, find the value of x which
minimises the average cost, and hence find the minimum value.
100
= AC = -
X
Average cost +2+-
X 10
d(AC) 100
xz + +o = - 1oox- + +o
2
dx
There is a maximum or a minimum when d(AC)Idx = 0, that is, when

-100
--+-fu=O
xz
or x = 31.623 (taking the positive root)
At this value of output
d 2 (A C) 200 200
=-=----
dx2 x3 (31.623) 3
Xmax Xmin X Xo X
(a) (b)
Fig. 5.10
This is positive and therefore the value of x = 31.623 gives the minimum
value of the average cost function
AC = ___!_QQ._ + 2 + 31. 623 = 8.3246

31.623 10
Example 3
The cubic equation
y =a+ bx + cx 2 + dx 3
can have a number of forms depending upon the sign and size of the
coefficients of the variable x. The graphs in Fig 5.10 are examples of two
of the forms which it might take.
Fig 5.10 (a) has no true maximum or minimum value in terms of our
previous definition because y increases indefinitely as x increases and y
decreases indefinitely as x decreases. The values of the function corre-
sponding to the value Xmax and xmin are, however, called local maximum
and local minimum. If we apply the graphical reasoning to this function
that was applied to the quadratic function it is easily seen that these two
points fulfil the conditions required for maxima and minima.
Fig 5.10 (b) has no local maximum or minimum values but has a
value x 0 for which the function has a slope of zero. This, therefore,
satisfies the condition dyldx = 0. But it gives the value d 2yldx 2 = 0,
which does not satisfy the second-order conditions for a maximum or

minimum. It is, in fact, called a point of inflexion, which means that
d 2y I dx 2 = 0 there but dy I dx does not change sign as x increases through x 0 •
From these results it is possible to define a set of rules.
(1) A function has a stationary value, which can be either a max-
imum, a minimum or a point of inflexion at the point where
dy =0
dx
(2) (a) if d 2yl dx 2 is negative at that point, the stationary value is a
maximum
(b) if d 2y I dx 2 is positive at that point the stationary value is a
minimum
(c) if d 2yldx 2 is zero the stationary value may be a maximum,
minimum or point of inflexion and the curve should be
sketched to examine it.
The condition (1) is frequently referred to as the necessary condi-
tion for a stationary value. That is, unless this is satisfied neither a
maximum, nor minimum nor point of inflexion can occur. This condition
is not, however, sufficient for a maximum, since it is also satisfied at a
minimum or point of inflexion. Conditions (1) and (2) together make up
the necessary and sufficient conditions for a maximum, minimum or
point of inflexion since they include all the possibilities. If a stationary
value occurs, it satisfies these conditions. If a value of x occurs which
satisfies these conditions it is a stationary value.
5.8 Cost and revenue analysis

The differential calculus enables us to determine the stationary
values of a function and to decide whether these values are maxima or
minima. This important result is of frequent use in economics as the
following example shows:
Example
Let the total cost function be
c = 1()() + 2q + foq 2
and let the demand function be

p=20-!q
Then the total revenue TR = pq. Substituting the above expression for p
in the revenue function gives
TR = pq = (20- ~
= 20q- ~
This is a quadratic equation and it can be sketched roughly by

determining a few points on the curve.
1. TR = 0 when 20q - !q 2 = 0
that is, when q(20 - ~ = 0
q = 0 or 100
2. d(TR) = 20 - ~
dq
This equals zero when
~ = 20, q =50
and at this point
TR = 20 X 50 - ~ X 50 X 50 = 500
To check whether this is a maximum or a minimum it is necessary to
obtain the second-order derivative
d 2 (TR)
dq2
This is negative, and therefore the maximum total revenue occurs
at an output of 50 units.
The revenue function is shown in Fig 5.11, along with the cost
function. The vertical distance between the two curves indicates the net
revenue, NR, or profit, defined by
net revenue = total revenue - total costs
or NR = TR- C
This is positive over the range q = q 1 to q = q 2 • It is not, however,
necessarily at its maximum at the output where total revenue is at its
maximum, because costs are increasing with output. To determine the
output where net revenue is a maximum use can be made of the
differential calculus:
c
500
Cost
function
q, 50 100 q
Fig. 5.11
NR = TR- C
= (20q - kq 2) - (100 + 2q + foq 2)
18q - -foq 2 - 100
. . d(NR) = 18 _ ~
dq
This is zero when
~ = 18 or q = 30
At this output
NR = 18 X 30 - fo X 30 X 30 - 100 = 170
To check whether this is a maximum or a minimum we differentiate
a second time with the result
This is negative and, therefore, this net revenue is a maximum. The

maximum total revenue occurs when q = 50. At this output
NR = 18 X 50 - fo X 50 X 50 - 100 = 50
This is considerably less than at the reduced output of 30. Finally,
d(TR) = 20 _ ~ and
dC
-=2+!q
dq dq
Therefore at an ouput of 30
Marginal revenue = d(TR) = 20 - t . 30 = 8

dq
dC 1
Marginal cost = - = 2 + 5 . 30 = 8
dq
This confirms that maximum net revenue is obtained when
marginal revenue = marginal cost
and for q < 30, MR>MC
for q > 30, MR<MC
5.9 Exercises
1. Determine the stationary values of the following functions:
(a) y = 3x2 - 120x + 30
(b) y = 16 - 8x - x 2
(c)y=x4
(d) y = (x - 5) 3
2. The total cost of producing an output q is given by
c= 500 + 4q + !q 2
Obtain the value of q which minimises the average cost.
3. If the demand function for a monopolist's product is q + 2p = 10,
determine the price and output required to maximise the total
revenue. What is the elasticity of demand at this price?
4. The total cost function for a product is TC = q3 + 20q 2 + 20 and
the demand function is q + p = 240. Find the price and level of
output required to maximise a monopolist's net revenue.
5. Compare the levels of output which (a) maximise total revenue (b)
minimise average costs and (c) maximise profit for a monopolist
faced with a demand function 2q + 4p = 200 and the total cost
function TC = 256 + 2q + 2q 2 • Sketch the total revenue, average
cost and profit functions.
5.10 Profit maximisation in several markets

A common problem in economics is where a monopolist trades in
more than one market. For example, a firm may face different demand
curves in the home and overseas markets, or a company such as a
national railway network may be able to charge different prices to
different groups of customers (e.g. season-ticket holders, students, pen-
sioners, groups, same-day returns). In these circumstances, two types of
profit maximisation can occur: with price discrimination and without it.
If the firm has all the relevant information it will be able to decide
whether price discrimination is worthwhile.
For example, suppose that a national electricity supplier has two
types of customer - domestic (d) and industrial (i) - and that the
demand functions are
5Qd + pd = 1000
P; + 2.5Q; = 10000
Here the demand from industrial users is higher but declines more
steeply as the price increases. The cost of producing electricity is the
same for both groups of customers and is given by
TC = 100 + 20Q + O.lQ 2 where Q = Qd + Q;
Initially we will assume that there is price discrimination and that profits
are to be maximised. For the domestic market total revenue is
TRd = PdQd = (1000 - 5Qd)Qd
from using the demand function, where we have chosen to obtain TRd in
terms of Qd because TC is in terms of Q. It is also possible to use the
demand functions to eliminate the quantities and to operate on prices.
The total cost is
TC = 100 + 20(Qd + Q;) + 0.1(Qd + Q;) 2
where Q; is supplied to industrial customers. From the point of view of
the market for domestic users, Q; is exogenous and can be regarded as a
constant. The profit function for the domestic market is

nd = TRd - TC
= (1000- 5Qd)Qd- [100 + 20(Qd + Q;) + 0.1(Qd + Q;)l]
= (980- 0.2Q;)Qd- ~ 100- 20Q;- 0.1Qf
Profit is maximised when the derivative with respect to Qd is zero,
treating Q; as a constant. Therefore,
d;;rd
dQd = 980 - 0.2Q; - 10.2Qd
and setting this to zero gives
Qd = 96.0784 - 0.0196Q;
That is, the value of Q in the domestic market which maximises profits
depends on the value of Q in the industrial market. This is to be
expected when the cost function is non-linear since it is the total of Qd
and Q; which determines TC. We note that the second-order derivative
equals - 10.2 and so the stationary value is a maximum.
In the industrial market the total revenue is
TR; = P;Q; = (10000 - 2.5Q;)Q;
and profits are given by
n; = TR; - TC
= (10000- 2.5Q;)Q;- [100 + 20(Qd + Q;) + 0.1(Qd + Q;) 2 ]
= (998o- o.2Qd)Q;- 2.6m- 100- 2oQd- ~
d;;r
and - ' = 9980 - 0.2Qd - 5.2 Q;
dQ;
Setting this to zero gives
Q; = 1919.2307 - 0.0385Qd
Again the second-order derivative indicates a maximum value, and the
value of Q; depends on the value of Qd. We now have two simultaneous
equations resulting from maximising profits in each market:
Qd = 96.0784 - 0.0196Q;
Q; = 1919.2307 - 0.0385Qd
and solving these gives Qd = 58.5056 and Qi = 1916.9782. The prices

are Pd = 707.474 and Pi = 5207.556 from the demand functions. The
value of Q is 1975.48 giving TC =429863.29. The total revenues are
TRd = 41391.19 and TRi = 9982771.3 so that the overall profit is
TRd + TRi - TC = 9594299
Next we assume that there is no discrimination so that prices must
be the same in both markets. Let P be the common price. The demand
functions are
5Qd + P = 1000 or Qd = 200 - 0.2P
P + 2.5Qi = 10000 or Qi = 4000 - 0.4P
and so the total demand is
Q = Qd + Qi = 4200 - 0.6P
and the total revenue function is
TR = PQ = (7000 - 1.667Q)Q
Profit is
;r = TR - TC = (7000- 1.667Q)Q- (100 + 20Q + 0.1Q 2 )
= 6980Q - 1.767Q2 - 100
dn
so that dQ = 6980 - 3.534Q
and setting this to zero gives Q = 1975.1 and Jr = 6892996 which is a

maximum. Notice that the value of Q is approximately the same with
and without price discrimination and that profits are higher with price
discrimination.
5.11 Maximising tax revenue

In section 1. 9 the effects of taxes on the supply of goods was
discussed. We now consider the problem faced by the government of
choosing the level of a tax so as to maximise revenue. Initially the case
of a flat-rate tax oft per unit is considered. In the absence of a tax the
demand and supply functions are
demand: qd = f(p)
supply: qs = g(p)
while with the flat-rate tax they are

demand: qd = f(p)
supply: qs = g(p')
where p'=p-t
is the price received by the supplier. The total tax revenue is
T = tq
where q is the equilibrium quantity and T is to be maximised with
respect to q.
For example, let the demand and supply functions be
demand: qd = 194 - 3p
supply: qs = 2p' - 6
and p' =p- t
The equilibrium is when qs = qd so that
194 - 3p = 2(p - t) - 6
or 200 + 2t = 5p
Since T is expressed in terms of q the value of p can be substituted into
the demand equation to give the equilibrium q as
q = 194 - 3(200 + 2t)/5 = 74 - l.2t
and hence T = tq = 74t - 1.2t 2
dT
To maximise T, - = 74 - 2.4t
dt
and setting this to zero, t = 30.83. For this value, q = 37, p = 52.33
and the price received by the supplier is p' = 21.5. The second-order
derivative is - 2.4 and so the stationary point is a maximum. The
maximum tax revenue is T = 1140.7. It is easily seen that if t = 0 the
equilibrium is p = 40 and q = 74, showing that the flat-rate tax has a big
effect on the value of q.
For a percentage tax the approach is similar except that a tax of
100r% changes the supply curve in the example to
supply: qs = 2pr - 6
where pr = p(l - r)
is the price received by the supplier. The equilibrium is
when qd = 194 - 3p = 2p(l - r) - 6
or p = 200/(5 - 2r)
The corresponding q is q = 194 - 600/(5 - 2r)
Tax revenue is given by
T = rpq = 200r [ 194 _ 600 ]

(5 - 2r) (5 - 2r)
74000r - 77600r 2
=------
(5- 2r)l
To maximise T, we require the value of r for which dT!dr is zero.
Treating the expression for T as a quotient,
_dT = ~~ ~ -_2:__.!__r)
dr (5- 2rt
Setting the numerator to zero and simplifying,
(5 - 2r)(185 - 388r) + (740r - 776r2) = 0
and the solution is r = 0.5892. That this is the maximum can be shown
by checking the second-order derivative. The maximum value of T is
1140.9 which occurs when q = 37 and p = 52.33. The price received by
the supplier is pr = 21.5.
The results with these two types of taxes, the flat-rate and the
percentage tax, are the same. This is to be expected since for any
percentage tax there is an equivalent flat-rate tax. The tax per unit is rp
for the percentage tax and t for the flat-rate tax, and the tax revenue
functions which are maximised are T = rpq and T = tq respectively.
5.12 Inventory models

One of the costs of a manufacturing business is that of holding
stocks or inventories. For example, suppose a firm has an annual de-
mand for N units of a product and that the demand is evenly spread
throughout the year. One way of meeting this demand would be to
produce N units at the beginning of the year and use them up during the
year. This would result in only one lot of set-up costs, but relatively high
storage or inventory costs. Another way would be to produce N/12 units
Inventory
level
Time
Fig. 5.12
each month. This would reduce inventory costs, since the maximum
inventory is now N/12 instead of N, but would result in twelve times as
high set-up costs.
The problem is to decide how to minimise the total annual costs and
so balance the benefits of long production runs against the costs of
holding inventories. For simplicity, we will assume that production is
instantaneous and that no shortages are allowed. Let the set-up costs be
S per batch and Q be the number of items produced in each batch. The
total annual production is N so the number of batches per year is NIQ
and the annual cost of setting up production is SN/Q. Since production
is assumed to be instantaneous, the number of units in stock immedi-
ately after production has ceased is Q, and the stock then runs down
until it reaches zero and the next lot of production starts. This implies
that the average level of inventory is Q/2 (see Fig. 5.12) and if the
annual cost of having an item in inventory is £i per unit then the total
annual inventory costs are iQ/2. The total of the set-up costs and
inventory costs is
C = SN + iQ
Q 2
and the problem is to choose Q so as to minimise C, treating i, Sand N

as constants. Differentiating,
dC SN
-= --
Q2
+
dQ 2
and for a minimum this is zero so that Q = Y(2SN/i). To check that this
is a minimum,
d 2C 2SN
--=--
dQ2 Q3
and since S, Nand Q are all expected to be positive we have a minimum.

The result is that each batch of production should be Q = V(2SN/i)
items and these should be produced NIQ times a year. This value of Q is
referred to as the economic order quantity.
To take a particular example, suppose that the annual demand is
for N = 10000 units, the set-up costs are S = 1000 and the inventory
costs are i = 5 per unit per annum. Then Q = 2000 and there are
N/Q = 5 production runs per year, with a total cost of C = 10000.
Notice that when Q = 2000 the total inventory cost is 5000 and so is the
total set-up cost. Next we will see what happens when the annual
demand changes. Suppose the new value is N = 2500, a quarter of the
previous value, and the costs are unchanged. Then
Q = v(2.1000.2500/5) = 1000
which is half of the previous value. Therefore the value of the economic
order quantity, Q, is not very sensitive to changes in demand. From the
formula for Q it can be seen that the value of Q is proportional to the
square root of Sand N, and inversely proportional to the square root of
i. We will return to a discussion of sensitivity analysis in section 5.14
below.
The various assumptions made above can be relaxed. For example,
if the assumption of instantaneous production is dropped and instead it
is assumed that production is at the rate of R units per year, the situation
is as in Fig. 5.13. If production occurs for the period ti (measured as a
fraction of a year) then the number of units produced is Q and the
demand during this period is tiN. Notice that ti = Q/R. The resulting
maximum inventory at the end of ti is Q - tiN or Q(1 - N/R). This now
declines at a rate N for a period t2 until the inventory becomes zero and
production starts again. The average level of the inventory is again half
Inventory
level
Q- t,N
Time
t, - -... ~ ~--- t,
Fig. 5.13
of the maximum (because additions and withdrawals from the inventory
take place at a steady rate) so that the total annual inventory cost is
IC = 0.5iQ(l - NIR)
The total annual cost, C, is /C plus the set-up cost and so
C = 0.5iQ(l - N!R) + SN!Q
This is to be minimised, treating everything except Q as a constant.
~= 0.5i(l - NJR) - SN/Q 2
Setting this to zero and solving,

Q2 = 2SNR
i(R- N)
It is easily shown that the second-order condition for a minimum is
satisfied.
For example, suppose that S = 1000, N = 10000 and i = 5 as pre-

viously but now the production rate is R = 20000 per year. Then
Q = 2828.4 items per batch and there are N!Q = 3.54 production runs
per year.
5.13 Exercises
1. A car firm sells in three markets: to its own workers, where
demand is 0.1P 1 + Q1 = 50, to the domestic market, where de-
mand is 0.2P2 + Q2 = 20000, and to the overseas market, where
demand is 0.4P 3 + Q3 = 25000. The total cost function is
TC = 1000 + 15(Q 1 + Q 2 + Q 3 ). The firm wishes to maximise
profits. Should it follow a policy of price discrimination?
2. A company operates in two markets with the following demand
functions:
P2 = c - eq2
and the total cost function is TC = f + g(q 1 + q2 ).Show that if
the firm maximises profits the quantity sold is the same with or
without price discrimination.
3. The demand for a product is given by p + 2q = 250 and the supply
by p - 4q = 100.
(a) If a flat-rate tax is imposed on each unit sold what is the
maximum possible tax revenue?
(b) Show that if a percentage tax is imposed on each unit sold the
same maximum possible tax revenue is obtained.
4. If the market for a product is given by
demand: p = 110 - 3q 2
supply: p = 10 + q2
determine the equilibrium values of p and q,
(a) in the absence of any taxes
(b) if a flat-rate tax, chosen to maximise tax revenue, is imposed
on each item sold.
5. An electronics company is under contract to supply 20,000 radios
per annum at a uniform rate to a chain-store. The annual inventory
cost per radio is £5 and the set-up cost for a production run is
£2,000. Assuming production is instantaneous and shortages are
not permitted, how many radios should be produced in each run so

as to minimise the total annual cost?
6. If the circumstances in Question 5 change so that the rate of
production is 50,000 per annum, what effect does this have on the
size of the optimal production run?
7. A manufacturer of business stationery has agreed to supply, at a
constant rate, 3,000 boxes of A4 headed notepaper each year to a
local company. The cost of storing a box is £3 per year and the cost
of setting up the printer is £500. Once set up, production is at a
constant rate of 9,000 boxes per year. If no shortages are permit-
ted, how many boxes should be produced in each run in order to
minimise the total annual cost?
5.14 Differentials
The derivative dyldx represents the rate of change of y with respect
to x for very small changes in x. It is defined as the ratio
where Dy and Dx are small but finite changes in y and x. The relationship
between these is shown in Fig. 5.14
In fact, Dy = dy + E
Dx dx
or Dy = ( : ) ox + E ox
where ~ as Q moves close to P. When this occurs, ~ ~
and f ~ so that
dy = (:)dx
We define dy as the differential of y and dx the differential of x. It

follows that
dy = differential of y = lim ( Dy)

dx differential of x .sx-o Dx
Using this notation it is possible to calculate the change which will occur
Fig. 5.14
in y for a small change in x when y is a function of x. For example if

y = 3x 2 + 2x + 1, then dyldx = 6x + 2 and dy = (6x + 2) dx which
means that the change in y which occurs due to a small change in x
depends upon the value of x at which the change takes place.
If x = 10 then dy = ((6 X 10) + 2)] dx = 62 dx
If x = 1 then dy = (6 + 2) dx = 8 dx
Therefore, the change in y resulting from a small change in x becomes
larger as x increases in value.
Notice that the value given by the differential is an approximation
which is close for small values of dx. Thus in the example, when x = 1,
y = 6 and if the change in xis dx = 0.1, the approximate change in y is
dy = 0.8 while the accurate value of y for x = 1.1 is 6.83 so that the
actual change in y is 0.83.
Differentials can be used in what is known as sensitivity analysis, in
which the importance of small changes in values of constants or vari-
ables is of interest. For example, in the simple inventory model discus-
sed in section 5.12 above, the economic order quantity, Q, is given by
Q= ~
where S is the set-up cost, i is the inventory cost and N is the annual
demand. Suppose that we have S = 1000, N = 10000, i = 5 so that
Q = 2000 and we wish to see how sensitive the value of Q is to these
initial values. To differentiate this expression it is convenient to take
logarithms (to the base e) so that
log Q = 0.5log2 + 0.5logS + 0.5logN - 0.5logi
and then
1 dQ 0.5 so dQ = 0. 5Q dS
--=-
Q dS S s
This gives the approximate change in Q resulting from a change (or
error) of dS in S. For example, suppose that dS = 100, then dQ = 100
so that an error of 100 in the set-up cost changes the optimal batch size
by 100, approx. Similarly,
dQ = 0. 5QdN and
N
and if dN = 100, dQ = 10 approx., while if di = 1, dQ = -200 approx.
Here it is clear that errors in S and i can have a big effect on the value
of Q.
5.15 Taylor's and MacLaurin's theorems

These are two theorems which have many important applications in
advanced economic theory. Here they will be used to obtain some
important series expansions. Taylor's theorem states that, for a general
function f,
x2 xn
f(a+x) = f(a)+xf'(a) + -f"(a) + · · · +-J<n>(a) + · · ·
2! n!
if the series converges, where
f(a) is the value of the function f(a + x) when x = 0,
f'(a) is the value of the f'(a + x) when x = 0,
t<n>(a) is the value of the nth derivative of f(a + x) when x =0
and provided the derivatives are finite and continuous. This theorem
allows any general function to be approximated by a polynomial. If the
terms in the expansion decrease quickly, the polynomial may be of low
order, such as a quadratic or linear one.
To show that Taylor's theorem is valid suppose that

f(x) = 3 + 2x + X2
which has f'(x) = 2 + 2x

f"(x) = 2
f"'(x) = 0
and we let a = 1 so that
/(1 + x) = /(1) + xf'(1) + ~ f"(1)
Further terms can be ignored since they are zero. By putting x = 1 we
get /(1) = 6, f'(1) = 4, f"(1) = 2 and so
2x2
/(1 + X) = 6 + 4x + Z = 6 + 4x + r
which is the same as f(x) with x replaced by 1 + x.
As an illustration of the use of Taylor's theorem let us determine
the series expansion of log(1 + x), where logarithms to the base e are
used.
Let f(a + x) = log (1 + x)
Then f(a) = log (1 + 0) = 0
1
f'(a + x) = - - and f'(a) = 1
1 +X
-1
f"(a + x) = and f"(a) = -1
(1 + xY
2'
f"'(a + x) = · and f"'(a) = 2!
(1 + x) 3
-31
J<4>(a + x) = · and J<4>(a) = -3!
(1 + xt
so that
x2 x3 x4
log (1 + x) = 0 + x --+-(2!) +- (-3!) + · · ·
2! 3! 4!
x2 X3
=x - - + - - - + - + ...
~ r
2 3 4 5
which is an infinite series. It can be shown that this series converges

provided - 1 < x < + 1.
MacLaurin's theorem can be regarded as the special case of
Taylor's theorem which occurs when a = 0. It states
x2 x"
f(x) = f(O) + xf'(O) +- f"(O) +· · ·+ - J<">(O) +· · ·
2! n!
if the series converges. The derivatives are evaluated at x = 0.
Example 1
f(x) =sin x f(O) = sin 0 = 0
f'(x) =cos x f'(O) = cos 0 = 1
f"(x) = -sin x f"(O) =-sin 0 = 0
f"'(x) = -cos x f"'(O) =-cos 0 = -1
This process can be repeated indefinitely and an infinite series of
terms is obtained of which all the odd terms are zero and the even terms
are alternately+ 1 and -1. Applying MacLaurin's theorem, we obtain
x2 x3
f(x) = sin x = 0 + x(1) +- (0) + - (-1) + · · ·
2! 3!
x3 xs x7
=X--+---+ ...
3! 5! 7!
Sin x is known as an odd function of x because its expansion contains
only odd powers of x.
MacLaurin's theorem can also be used to expand the functions cos x
and ex
Example 2
f(x) =cos x f(O) = cos 0 = 1
f'(x) = -sin x f'(O) =-sin 0 = 0
f"(x) = -cos x f"(O) = -cos 0 = - 1
f"'(x) =sin x f"'(O) = sin 0 = 0
Extending this process further and substituting the results in Mac-
Laurin's expansion gives
x2 x4 x6
cosx=l--+---+···
2! 4! 6!
Cos x is known as an even function of x because its expansion contains

only even powers of x.
Example 3
f(x) =ex f(O) =e 0 =1
f'(x) =ex f'(O) = e0 =1
f"(x) =ex f"(O) = e 0 =1
f"'(x) =ex f"'(O) = e 0 =1
which can be continued to give
xz x3
ex = 1 + X +- + - + · · ·
2! 3!
5.16 Euler relations

These arise from the three series which were obtained using Mac-
Laurin's theorem. Since
xz x3 x4
ex = 1 + X +- + - + - + • • ·
2! 3! 4!
if the index is multiplied by the imaginary number i (see section 3.3) and
the function expanded as an infinite series, we obtain
ix . (ixY (ix) 3 (ixt
e = 1 + (1x) + - + - +- + ···
2! 3! 4!
and remembering that i = v'(- 1), i 2 = - 1, i 3 = - i, i 4 = 1, etc, we
have
. r ix 3 X4
eu: = 1 + ix - - - - + - +···
2! 3! 4!
This can be separated into real and imaginary parts as follows:
eix=( ~ +::-··· ~ ~ ···)

=cos x + i sin x.
By similar reasoning it can be shown that
e-ix = cos x - i sin x.
These relationships are important and will be found extremely

useful in the treatment of certain differential equations (see section 9.5).
5.17 Newton's method

It is possible to use the differential calculus to determine the real
roots of any polynomial in x. This is particularly useful in the case of
cubic and higher-order equations when the roots are not integer values.
For example let the cubic equation be
f(x) = x 3 + 2x + x2 - 5 =0
We can determine a very approximate value for a root to this
equation by substituting various integer values for the variable. For
example,
let x = 0 then f(x) = 0 + 0 + 0- 5 = -5
X= 1 f(x) = 1 + 2 + 1-5 = -1
x=2 f(x) = 8 + 8 + 2- 5 = 13
With this information we can see that the function changes sign
between the values x = 1 and x = 2. It follows that at least one root of
the cubic equation lies between these two values. One of these values is
taken as a starting point from which it is possible to proceed in a series
of steps to the actual solution. The procedure can be stated, in general,
as follows.
Let a be one of the two consecutive integer values between which
the function changes sign. Then the true root of the equation is a + h,
where h is an unknown whose absolute value lies between 0 and 1. If a is
the lower of the two integers then h is positive whereas if a is the higher
integer h is negative.
Taylor's theorem states
h2
f(a +h) = f(a) + hf'(a) + 2! f"(a) + · · ·
If h is small then approximately

f(a + h) = f(a) + hf'(a)
Now for this to be a root, f(a + h) = 0 and hence
- f(a) (1)
h, = f'(a)
Obtain an approximate
root, a
The root
is a
No
Form a new value

f(a)
new a = a- f(a)
Fig. 5.15
is close to h. Thus a first approximation to the root a + h is given by

a + h 1 where a is a known value and h 1 is - f(a)/f'(a).
This suggests an iterative (or repeated) strategy for obtaining a root
of an equation:
1. Obtain an approximate root, a.
2. Evaluate f(a) and iff( a) is close enough to zero to be acceptable,
stop the procedure. Otherwise continue.
3. Obtain a better approximation to the root by using
f(a)
a + h 1 = a - f'(a) and go back to step 2.
This procedure is summarised in the flow diagram (see Fig. 5.15)

where, starting at the top with the first step of obtaining an approximate
root, we follow the arrow to the second step, represented by the dia-
mond, indicating that a decision is needed. If the answer to 'isf(a) = 0?'
is yes the process stops. If the answer is no, we move to the lowest box
(for step 3) and form a new approximation, the new value of a, and
move to step 2.
In the example we are considering
f(x) = x + 2x2 + x -
3 5 and a =1
.. f(a) = /(1) = 1 + 2 + 1- 5 = -1
d
Also f'(x) = dx {f(x)} = 3x + 4x +
2 1
:. f'(a) = /(1) = 3 + 4 + 1 = 8.
Substituting these values in (1)
h = - /( 1) = - (- 1) =.! = 0.125
I /'(1) 8 8
At this stage in the procedure a better approximation to the root of
the cubic equation is given by (a + h,), that is (1 + 0.125) = 1.125. Let
us call this value a,). Then we can approach even closer to the true root
by letting this be denoted by (a 1 + h2 ) where
f(a, + h 2 ) = f(a,) + hJ'(a 1 ) = 0
or h = -f(a,)
2 f'(a,)
As before f(x) = x 3 + 2x2 + x - 5
.. f(a,) = /(1.125) = (1.125Y + 2(1.125)2 + (1.125)- 5 = 0.0801
and f'(aJ = /'(1.125) = 3(1.125)2 + 4(1.125) + 1 = 9.2969
:. h = - /(1.125) = - 0.0801 = -0.0086
/'(1.125)
2
9.2969
A closer approximation to the root of the cubic equation is given by
(a, + h 2 ) = (1.125 - 0.0086) = 1.1164, for which /(1.1164) = 0.0005.
This procedure can be repeated an indefinite number of times and
successive estimates of the root of the equation will oscillate about the
true value with ever decreasing amplitude. It can be carried out on a
computer because of its repetitive stepwise approach. The roots of any
polynomial in x for which the first and second derivatives exist (i.e. for
any continuous function of x) can be obtained by this method.
As another example let
f(x) = xlogx - 4 = 0
To find an approximate root, since logarithms are not defined for
negative numbers, we try positive values of x. When x = 1, /(1) = -4,
forx = 2,/(2) = -2.6andforx = 3,/(3) = -0.70. Thereforethereis
a root near x = 3 and this can be used as the starting value for Newton's
method. Since
f(x) = xlogx- 4
f'(x) = logx + xlx = logx + 1
When a= 3, f(a) = -0.70
f'(a) = log3 + 1 = 2.0986
and so h =- f(a) = 0 ·70 = 0.3336
1
f'(a) 2.0986
The new approximation to the root is 3 + h 1 = 3.3336
and /(3.3336) = 0.0138
/'(3.3336) = 2.2041
ivin h = - f(a) = -0.0138 = -0.00626.

g g 2 f'(a) 2.2041
The new approximation to the root is 3.3336 - 0.00626 = 3.32734 for
which /(3.32734) = 0.00004. This is close to zero and so we will take
3.32734 as the root. A closer approximation could be found by repeating
the process.
In our discussion of Newton's method we have used the first two
terms of Taylor's theorem and assumed that the other terms are zero.
Another approach is to include the third term so that, approximately,
f(a +h) = f(a) + hf' (a) + hz f" (a)

2
where a is the first approximate root and h is the adjustment to improve
it. Setting this to zero and rearranging gives the solution for h, ~ as
h =_ f(a)
(2)
1
f'(a) + 0.5hd"(a)
so that the new approximate root is (a + h 1 ). Notice that comparing (2)
with (1), there is an extra term in the denominator, and this includes h 1 •
Since h 1 is unknown we can take the approximate value given by (1),
and substituting this into (2) in the denominator the following formula
results
1 -f'(a) f"(a)
-=--+-- (3)
hi f(a) 2f'(a)
To demonstrate the effect of taking account of the third term, in the

previous example
f(x) = xlogx- 4
f'(x) logx + 1
f"(x) llx
and so taking a = 3 as the first approximate root,
f(3) = -0.70
f'(3) 2.0986
f"(3) 0.3333
1 -2.0986 0.3333
and so + = 3.0774
hi -0.70 2(2.0986)
Therefore, from using (3), hi = 0.3249, the new approximation to the
root is 3.3249 and /(3.3249) = -0.0053 compared with /(3.3336)
= 0.0138 from using (1). Using (3) has resulted in a better approximate
root after one iteration. Whether the extra complication of using (3) is
worthwhile depends on how easy it is to evaluate f"(x).
5.18 Plotting graphs

The use of differential calculus to determine the stationary values of
a function can make the plotting of graphs easier. For any function the
important characteristics are first, where the graph cuts axes and
secondly, where the stationary values are and whether they are local
maxima, local minima or points of inflexion. If y is a function of x, the
graph cuts the x-axis when y = 0 and putting this value into the equation
results in an equation in the one unknown, x. If this equation is a
quadratic it can be solved as shown in Chapter 3. For higher-order
polynomials and other functions finding the solutions can be difficult and
Newton's method may be needed.
For example, to sketch the graph of
y = 2x2 + 3x - 14
the points where it cuts the x-axis are when y = 0 and, solving the
quadratic, x = 2 and x = -3.5. It cuts they-axis when x = 0 and
y = -14. The stationary values are when the derivative is zero or when
dy
- = 4x + 3 = 0 sox= -0.75
dx
The second-order derivative is 4 and so x = -0.75 is a local minimum.
There are no other stationary values and so all that is needed is a few
points to give the general shape of the graph. The particular values
already found are used (i.e. x = 2, -3.5, 0, -0.75) plus a few other
values to give a wider picture. Here we include two extra values, for
x = -5 and x = 5:
X -5 -3.5 -0.75 0 2 5
y 21 0 -15.125 -14 0 51
The graph is plotted in Fig. 5.16.

As a further example consider plotting the graph of
y = 2x 3 + 0.2x2 - 19.3x + 15.75
When x = 0, y = 15.75. When y = 0,
f(x) = 2x 3 + 0.2x2 - 19.3x + 15.75 = 0
To solve this we try to find an approximate root. When x = 1, f(x)
= -1.35 so that the change of sign between f(O) = 15.75 and f(1)
indicates there is at least one root between x = 0 and x = 1, and the
values of f(x) indicate the root is near x = 1. Taking x = 1 as an
approximate root and using Newton's method,
f'(x) = 6x2 + 0.4x - 19.3
and f'(1) = -12.9
/(1) 1.35
so ---=
h = -0.1047
f'(1) 12.9
The new approximate root is 1-0.1047 = 0.8953 and /(0.8953)
= 0.0663. Using Newton's method again,f'(0.8953) = - 14.1325 and so
h = -0.0663 = 0.0046
-14.1325
The new approximate root is 0.8999 and /(0.8999) = 0.0014. Now
!'(0.8999) = - 14.0811 and soh = 0.0001. The new approximate root is
y
20
I
10
5 X
-10
-20
Fig. 5.16
0.9 andf(0.9) = 0.0. The factor (x- 0.9) can now be removed from the
cubic equation by long division:
+ 2x - 17.5
2x 2
x- 0.9J2x3 + 0.2x2 - 19.3x + 15.75
2x3 - 1.8xz
2.0x 2 - 19.3x
2.0x 2 - 1.8x
- 17.5x + 15.75
- 17.5x + 15.75
0
Therefore, f(x) = (x- 0.9)(2x 2 + 2x- 17.5) = 0
The roots of the quadratic are 2.5 and -3.5.
The stationary points are when the derivative is zero. Here,
dy
- = 6x + 0.4x
2 - 19.3
dx
y
50
40
30
-4 -2 4 X
-10
-20
-30
Fig. 5.17
and setting this to zero gives x = 1.76 and x = -1.83. The second-order
derivative is
and sox= 1.76 is a local minimum and x = -1.83 is a local maximum.

The table of values is:
X -3.5 -1.83 0 0.9 1.76 2.5

y 0 39.48 15.75 0 -6.69 0
The graph is shown in Fig. 5.17.
5.19 Exercises
1. Use Taylor's theorem to expand (1 + x) 4 and hence show that
1S = 5.0625.
2. Use MacLaurin's theorem to expand (1 + x)- 1 as a power series.
3. Use Newton's method to obtain a value of x which satisfies the
following equations using the given approximate root as a starting

point.
(a) r + 4.55x =0
- 8.70
Approximate root x = 1.5
(b) _x3 - 2.34x2 + 2x - 4.68 = 0
Approximate root x = 2.25
(c) 2_x3 - 8.2x 3x + 12.3 = 0
2 -
Approximate root x = 4
4. What rate of interest makes the present values of the following two
projects equal?
Return at end of year 1 2 3 4

Project A 70 80 100 100
Project B 60 90 80 130
(Hint: If i is the rate of interest and R = 1 + i then the cubic

equation in R can be solved using R = 1 as an approximate root.)
5. Plot the graphs of the following functions:

(a) y = r + x + 12
(b) y = x3 - x2
(c) y = _x3 - 3x2 - 24x + 26
(d) y = .r

1. Differentiate
(a) y = 6x4 - 3x2 + (9/x3 )
(b) y = (3x - 4)(5r - 2x + 20)
(c) y = (2x3 + x- 1 + 7)(4x2 - x 3 )
(d) y = (x 4 + 3x)/(7x2 + 4x + 4)
(e) y = (3r - x 3 + 5) 3
(f) y = log (4x 5 + 3x3 - 2x)
(g) y = ~ - 3e-3x
(h) y = e4x log (x2 - 6x4 )
(i) y = e3x sin 2x - e-Jx cos 2x.
2. For the following demand equations, where p = price,
q = quantity, determine the marginal revenue functions and the
elasticity of demand when p = 5:
(a) 3p = 200 - Sq
(b) p 2 = 350 - 13q
(c) p = 500 - ~
3. Find the stationary values of

(a) y = 4x 2 - 16x + 25
(b) y = (2x - 5) 3
(c) y =x 6•
4. A monopolist has a demand function p = 400 - Sq and the total

cost function is TC = 50 + 6q + q 2 • Determine the level of output
( q) which gives
(a) the maximum value of total revenue
(b) the minimum value of average costs
(c) the maximum value of profit.
Compare the values of profit for each of these.
5. A whisky producer can sell on the home market, for which the
demand function is ph = 1000 - 4qh, and on the overseas market,
for which the demand function is 2p0 = 1500 - 9qo. The total cost
function is TC = 300 + 4qh + 5q 0 • Should the producer charge
different prices in the two markets?
6. The demand function for a product is p = a - bqd and the supply
function is p = c + eif, where a, b, c, e are all positive. Show that
imposing a flat-rate tax oft per unit sold reduces the equilibrium
quantity and also reduces the equilibrium price. What value of t
maximises the tax revenue?
7. A company produces components on a machine with set-up costs
of £500 and the cost of storing the components is £4 per year.
Production is at the constant rate of 12,000 per year. The only
customer purchases 6,000 components at a uniform rate each
year. How many components should be produced in each run in

order to minimise the total annual cost?
8. Use MacLaurin's theorem to expand (1 - xt 1 as a power series.
9. Use Newton's method to find a root of 4x4 - 17x2 + 0.2 = 0
starting with x = 0 as an approximate root.
10. Plot the graphs of
(a) y = 2x3 - 6x2 + 15
(b) y = x 4 - 8x2 - 5
(c) y = x 3 + 3x2 + 6x - 2.
Chapter 6
Integral Calculus
6.1 Introduction
If x is the level of output and TC is the total cost, then given a cost
curve
TC = 3x2 + 20x + 13
by differentiation the marginal cost curve is
dTC
MC = - = 6x + 20
dx
That is, by the process of differentiation the total cost function becomes
the marginal cost function. If we are now told that a marginal cost
function is
MC = 6x + 20
and asked to determine the total cost function we must reverse the process
of differentiation, that is, we must integrate the marginal cost equation.
From our knowledge of differential calculus we might guess that
TC = 3x + 20x
2
and we should also realise that there may have been a constant term in
the TC function which would have disappeared on differentiation.
However, we cannot determine the value of this constant, if indeed
there was one, from the marginal cost function alone. For example
TC = 3x2 + 20x, TC = 3x2 + 20x + 13, TC = 3x2 + 20x +50
each have the same marginal cost function
MC = 6x + 20
We therefore write the total cost function associated with this marginal
cost function as
224
Integral Calculus 225
TC = 3x2 + 20x + C
where Cis an arbitrary constant, the constant of integration, whose value
can be determined only if additional information is available. This
additional information may be in the form of an initial condition, which
corresponds to the starting value of x, such as when x = 0 TC = 20,
resulting in C = 20, or a boundary condition, which states the maximum
value of x, such as when x = 100 TC = 33,000, resulting in C =1 ,000, or
a particular value such as that TC = 100 when x = 3 so that C = 13.
The notation commonly adopted is
J(6x + 20) dx = 3x2 + 20x + C

where the f is the integral sign, the process is known as integration and
the result is called the indefinite integral of 6x + 20. The term dx is
included to show that x is the variable being considered.
Other examples of applications in economics and business of in-
tegration as the reverse process to differentiation follow directly from
Chapter 5. Thus, for example, given a marginal-revenue function and an
initial condition, the total-revenue function can be found. Also, if the
expression for the marginal propensity to consume is known, then by
integrating, the corresponding consumption function can be found and
it will include a constant of integration.
By making use of the fact that integration can be thought of as the
reverse of differentiation it is possible to find, by trial and error, the integral
of many common functions. However, this may not always be easy because
a function may be altered in appearance after differentiation by grouping
together like terms or by cancelling out of common factors.
It is because of this that integration is generally more difficult than
differentiation and requires a good deal of practice and a bit of intuition
to become proficient. We have at our disposal, however, standard
methods for finding the integral of most common functions, and a
number of examples are considered here to illustrate these methods. We
defer some more difficult cases to Chapter 9 where differential equations
are considered.
6.2 Exercises
1. Derive the total cost function for processes in which the
(a) marginal cost is £1 per unit and the fixed cost is £600
(b) marginal cost is £1 per unit and the cost of production of 250
units is £300
(c) marginal cost is given by

MC = 2x +3
and the total cost is 100 when x = 5.
2. For the consumption function C = a + bY + c P determine the
marginal propensity to consume (mpc) function. Hence, if mpc
= 0.1 + 0.02Yand when Y = 0, C = 0, determine the values of a, b
and c.
3. If capital formation occurs continuously as time passes then the
rate of change of capital stock, with respect to time, is net invest-
ment. Suppose that net investment is given by
dK
J(t) = - = a + 2bt + 3ct 2
dt
Determine the equation for the capital stock if when t = 0, K = 0.
6.3 Techniques of integration

In this section the integrals of the functions discussed in Chapter 5
are to be determined mainly by reversing the process of differentiation.
More difficult forms are deferred until section 6.14.
(a) If y = axn, dy!dx = anxn-t

Hence,
axn+t
f axn dx =n-+-1+ C
which is true except in the case of n = -1 where the denominator would
become zero. This exception is considered under (b) below.
Examples
x4
1. a= 1, n = 3, f x 3 dx =-+C
4
2xl.5
n = 0.5, 2x0.5 dx = - + C
2. a= 2, f 1.5
4x- 1
n =- 2, 4x- 2 dx = - + C
3. a= 4, f -1
(b) If y = log x, dyldx = llx
Hence,
J -X1 dx = log x +C
which is the special case referred to above. More generally, if

dy 1du
y = log u, dx = -;; dx
ldu
Hence, J-;;dx = log u + C
Thus if the numerator is the derivative of the denominator the integral is
the logarithm of the denominator.
Examples
1. U =X,
du = 1
dx
and J-1
X
dx = log x +C
du 2dx
2. u = 2x + 3, -=2
dx
and f- -=
2x + 3
log (2x + 3) + C
3. u = x 3 - 4x + 1, du = 3xz- 4
dx
3x 2 - 4
f x - 4x + 1
3
dx = log ( x 3 - 4x + 1) + C
4. In some cases the expressions to be integrated need to be modified

slightly to obtain the standard form. For example,
xz + 2x
J x3 + 3xz + 5 dx
Here, u = x 3 + 3x2 + 5 and

du
- = 3x 2 + 6x = 3(x2 + 2x)
dx
Therefore write
x 2 +2x 1( 33x +6x)

2
X 3 + 3x2 + 5 = 3 X + 3x2 + 5
Hence
= -J
x 2 + 2x 1 3x2 + 6x
Jx 3 + 3x2 + 5
dx
3 X3 + 3x2 + 5
dx =~ log ( x 3 + 3x2 + 5) + C
(c) If y = aebx, dyldx = abebx
Hence
Examples
1. a = 3, b = 2, J 3ezx dx = -3ezx2- + C
2. a = 4, b = - 1, J4e-x dx = -4e-x + C
(d) If y = sin x, dyldx = cos x
Hence Jcos x dx = sin x + C
Similarly Jsin x dx = - cos x + C
The more general versions of these and also the integrals of other
trigonometric functions are deferred to Appendix A, Section A.6.
6.4 Exercises
1. Obtain the total cost functions from the following marginal cost
functions:
(a) MC = 2x + 3 with TC =50 when x = 5
(b) MC = x 2 + 2x + 4 with TC = 450 when x = 9.
2. Integrate the following with respect to x:
(a) 3x- 4 (b) 2x 2 - 4x +3
(c) 3x5 --
4
+ 2x ~ ~
xz x3 x
3 4x + 1
(e)-- (f)
3x- 5 2x 2 +x-4
(h) 4x + 6
x 2 +3x+l
(i) 2e3x + e-x (j) ~ 4e3x
(k) 2 sin x - 4 cos x
3. If marginal revenue is given by MR = a - 2bq where a and bare

positive constants, determine the total revenue curve if TR = 0
when q = 0. Hence determine the demand function.
4. The elasticity of demand for a certain product is ed = -0.3plq.
What is the demand curve if when p = 10, q = 100?
5. If the marginal propensity to consume is mpc = 0.5 + y-t - y- 2
and when Y = 1, C = 1, what is the consumption function?
6.5 The calculation of areas

The area under a continuous function can generally be determined
by using the integral calculus.
Let AB be a section of the curve y = f(x) between the values
x = X 1 , and x = X 2 , and let Z be ABX2 X 1 , the area which is required
(Fig. 6.1 ). In this illustration f(x) is taken to be steadily increasing: the
treatment in other cases is similar.
X, x, X
Fig. 6.1
Let us consider two points P 1 and P 2 on the curve which are a small
distance apart and from which perpendiculars are dropped to the x-axis.
If the heights of these perpendiculars are y and y + f>y respectively and
their horizontal distance apart is f>x, then it is possible to estimate the
area between these two lines, the curve, and the x-axis. This area, which
we call {Jz, is shaded in the diagram and it must be greater than the area
of the rectangle y f>x and less than the rectangle (y + f>y) f>x
that is yf>x < {Jz < (y + f>y)f>x
{Jz
or y <-<y + {Jy
f>x
When f>x .- 0, {Jy .- 0 and f>zlf>x .- dzldx, so that
dz
- = y = f(x) (1)
dx
and the rate of change of the area depends on the value of x. The total
area ABX2 X 1 is made up of an unfinitely large number of these small
areas such as dz. Its value is therefore given by the sum of all such areas
which are located between the values x = ~ and x = X 2 • This is usually
written as
Z = J
~ (dz)
dx dx = J f(x)dx
X2
(2)
X=Xl Xl
where the elongated Sis used as a summation sign.

This is known as a definite integral (or Riemann integral) and it has
a single numerical value associated with it. In this example the value
corresponds to the area under the curve. The definite integral is
obtained via the indefinite integral in the following way:
Step 1. Obtain the indefinite integral of the function by one of the
methods discussed previously.
Step 2 Substitute the value
0 X = xl in the indefinite integral.
Step 3 Substitute the value
0 X = x2 in the indefinite integral.
Step 4. Subtract the numerical value obtained in Step 2 from the
numerical value obtained in Step 3 and the result is the value of the
definite integral of the function between the limits X = XI and X = x2.
a---- -- X
Fig. 6.2
A derivation of these rules is rather difficult but a justification is

provided by the following simple example.
Example 1
Let us consider the linear function y = x. This is a straight line which
passes through the origin and bisects the angle between the x- and the
y-axis (Fig. 6.2).
Any point on this line has the same value for y as it does for x, and
P, with co-ordinates (a, a), can be considered as a typical point.
The area which is bounded by the line y = x, the ordinate x = a and
the x-axis is triangular in shape and is shaded in the diagram. Its area is
calculated very easily by simple geometry and is equal to ! a2 • Let us
compare this with the result that is obtained via the integral calculus.
Step 1. Determine the indefinite integral of the function with re-
spect to x.
I=Jydx
= Jx dx (by substitution y = x)
= ~ x2 + C
Step 2. Substitute the value x = 0 in the indefinite integral (x = 0 is

the ordinate which corresponds to the left-hand side of the required
area). The result is that
I,= 0 + C = C
Step 3. Substitute the value x = a in the indefinite integral ( x = a is

the ordinate which corresponds to the right-hand side of the required
area). Then
Step 4. Subtract I, from I2

z = I2- J,
ia2 +C-C
ia2•
This is an identical result to that obtained by simple geometry and it

is obvious now that the definite integral will never contain a constant of
integration. This is always eliminated during Step 4 of the above
method.
It is not really necessary to use the integral calculus to determine
the areas in simple cases such as those where the answer is obvious from
the geometry of the system. It can be of great use, however, in many
other cases, particularly where the function is non-linear as the follow-
ing example shows.
Example 2
For a company using a process where the marginal cost of produc-
tion is a function of the level of output, decreasing at first and then
increasing when the output is above a certain level, the following might
apply:
MC = 10 - q + zt q 2 , 0 < q ~ 20
This can be presented graphically as shown in Fig. 6.3.
The area under the marginal cost curve between the values q = 10
and q = 20 is equal to the extra cost which must be incurred when
production is increased from 10 units to 20 units of output.
0 10
Output, q
Fig. 6.3
Extra cost f (MC) dq
20
J (10- q + -f.oq 2 )dq

10
= {10(20)-! (20) 2 +6h (20Y}- {10(10)- h10)2 + 6h (10) 3 }
200- 200 + 400 - 100 +50- 50 = 200

3 3 3
6.6 Exercises
1. Calculate the area between the axis of x, the straight line given by
the following equations and the two values of x.
(a) y = 2x x, =0 x2 = 3
(b) y = 2x x, =3 x2 = 6
(c) y = 2x x, =0 x2 = 6
(d) y = 5x x, = 6 X2 =8
(e) y=2+3x x, = 0 x 2 = 3.
2. The marginal cost curve for a product is MC = 100 - 4q + q 2 •
What is the total cost of increasing output from q = 20 to q = 30?
3. If marginal revenue is given by MR = 200 - 6q, what extra total
revenue is obtained by increasing sales (q) from 15 to 20?
4. When interest at a rate r is compounded continuously, the present
value, P, of the amount, S, due n years from now is P = se-rn. The
present value of an annual income of S per year for n years is given
by
Determine the present value of 1,000 per year payable 10 times if

the interest rate is 5% compounded continuously.
6. 7 Absolute area
When the graph of f(x) cuts the x-axis, areas below the x-axis are
taken to be negative. For some applications this is the correct inter-
pretation. For others, however, the absolute area under the curve is
required, i.e. the deviations of the curve from the x-axis are to be
summed. When this happens, the direct evaluation of the integral gives
an incorrect solution. This can be seen by reference to the function y = x
and by considering the area shaded in Fig. 6.4.
The shaded area between the ordinates x = -a and x = +a is
obviously equal to! a2 + ! a\ that is, a2 • Using the integral calculus as
before would lead to
X=+a x=+a
Z= f
x=-a
ydx= J x dx = [ ~
x=-a
X2 ] ~ = ! a2 - ~ a2 =0
This is true if we consider the area above the x-axis to be positive and
the area below the x-axis to be negative. But if the total shaded area is
required irrespective of whether it is above or below the axis, then the
two sections must be calculated separately and their individual values
added together. The dividing point is that at which the function crosses
y y = x
Fig. 6.4
the x-axis. (Functions of higher order than the simple linear one may
cross the axis in a number of points and in such cases it may be necessary
to consider all the areas separately.)
In the example with the linear function y = x the dividing point is at
x = 0 and the area is found as follows:
Z1 =
X=O
f X dx = u X 2
] ~~ = - i a2
=n
x=-a
Zz = XT X dx x 2] ~~= ia 2
X=O
The area is the sum of the absolute values of Z 1 and Z 2 , i.e. the sum
of their numerical values when the negative sign associated with Z 1 is
ignored:
6.8 Exercises
1. Sketch the following curves and calculate the absolute area be-
tween the curves, the axis of x and the given values X 1 and x2 •
(a) y = 2x
(b) y = 2 + 3x xi=- 3 +2 x2 =
(c) y = x2 xi= -2 x 2 = + 3
(d) y=1+x+x 2 xi=- 4 X 2 = + 1
2. In measuring the accuracy of forecasts, negative and positive
errors do not necessarily cancel out and it is the absolute error
which matters. If forecasts are made for 10 periods into the future
and the forecast error at time t is given by E(t) = 55 - 16t + f,
determine the total absolute error by evaluating the absolute area
between E(t) and the t-axis from t = 0 to t = 10.
6.9 Consumer's surplus and producer's surplus

If the demand curve for a product is downward sloping, there will
be some consumers who are willing to pay more than the current market
price in order to obtain the product. The gap between the market price
and the price the consumer is willing to pay is a measure of the con-
sumer's surplus satisfaction and is known as consumer's surplus. Since
the demand function is continuous the total consumer's surplus is mea-
sured by the area ~ shown in Fig. 6.5, where Pi and Xi are the
market price and quantity. This area can be evaluated by integration as
the area ABXiO minus the area PiBXP.
For example, if the demand function is
p = 45- 2x- x2
and the market price is Pi= 10 and the quantity is Xi= 5, then the area
A BXi 0 is given by
x=5
I= f
x=O
pdx= J (45 -
X=5
X=O
2x - x 2 ) dx
= [45x- x 2 -
x3] s
3 0
= ( 225 - 25 125) -
-3 (0) = 158.33
The area P 1 BX10 is P 1X 1 = 50 and therefore the consumer's surplus is
measured as 158.33 - 50 = 108.33.
The concept of producer's surplus is rather similar to that of con-
sumer's surplus but relates to the supply curve.
Since the supply curve is assumed to be upward sloping, for any
Fig. 6.5
given market price, P,, say, some producers would have been willing to
supply a lower quantity at a lower price. The vertical gap between the
market price, P,, and the supply curve (see Fig. 6.6) is a measure of the
producer's extra satisfaction and is known as producer's surplus. It is
evaluated by finding the total area which is above the supply curve and
below the horizontal line P = P,. For the price P, and quantity X, the
shaded area is given by
x,
I = P,X, - f 0
P dx
For example, suppose that the supply curve is
p = 20 + 2x + 0.3x 2
and the producer's surplus is required for a price p = 70 and quantity

x = 10, then
10
I = 700 - J 0
(20 + 2x + 0.3x2 ) dx
700 - [20x + x 2 + 0.1x3g0
= 700 - (200 + 100 + 100) = 300
x, X
Fig. 6.6
6.10 Areas between curves

We now consider two problems where the area of interest is in
between two curves. The first concerns profit, given by the area between
a marginal revenue function and a marginal cost function. The second
relates to depreciating assets and the decision to scrap a capital good.
In discussing profit (or net revenue) maximisation in section 5.8 the
starting point was the net revenue function, defined as being total
revenue minus total costs. Net revenue was maximised when marginal
revenue equalled marginal cost and, given the maximising value of
output, the maximum value of net revenue could be obtained. Here we
start with the marginal cost (MC) and marginal revenue (MR) functions.
The difference between these can be defined as being the marginal profit
(MP) function so that
MP= MR- MC
If the level of output is q 1 then the total value of profits is
ql
profit = f0 (MR - MC) dq
For example, suppose that the marginal revenue function is

MR =50- 2q
and the marginal cost function is
MC = 10 +q
For an output of q = 10 the value of profit is
10
Profit = J (50 -
0
2q - 10 - q) dq
to
= f 0
(40 - 3q) dq = [40q - 1.5q2W
= (400 - 150) - (0) = 250

The situation is represented graphically as the shaded area in Fig. 6.7.
Clearly, at q = 10 MR > MC and so it pays to increase output. Profit is
maximised when MR = MC and here this is when q = 13.33, giving a
profit of 266.67.
50
MR or MC
10
10 13.3
q 25
Fig. 6.7
Before leaving this approach to measuring profit it is worth pointing

out that the use of the MR and MC curves means that the fixed costs of
production (if any) are ignored. This is because in using marginal
analysis we are concerned with short-run behaviour and therefore fixed
costs, which are defined as those costs which cannot be changed easily,
are not relevant.
Turning now to the second application of areas between two
curves, we consider problems with depreciating assets and the decision
of when to scrap a capital good. Suppose that a company instals a
machine which dispenses hot drinks. The revenue from this is MR(t)
which is a function of time. Notice that this is the marginal revenue at
time t and not the total revenue from time 0 to t. As the machine gets
older it is more prone to being out of order so that revenue falls as time
passes. Also, the cost of servicing the machine, MC(t), rises as time
passes, since parts wear out and need to be replaced. Initially we will
assume that the machine has no scrap value. The profit from having the
machine is the difference between MR(t) and MC(t) and is the area
between the two curves. Let the marginal revenue function be
MR(t) = 200 - 2t 2
and the marginal cost function be
MC(t) = 25 + 2t + t 2,
where t is in years. These are shown in Fig. 6.8. The profit from using
the machine for 4 years is
4
Profit = f 0
(MR(t) - MC(t)) dt
4
= s (200 -
0
2t 2 - 25 - 2t - f)dt
4
= s (175 -
0
2t- 3t 2 )dt
= [175t- t 2 - ~
= (700 - 16 - 64) - (0) = 620
The machine will become unprofitable when MR(t) MC(t) (since

beyond this point MR < MC so that profit is lower) or
200 - 2t 2 25 + 2t + t2
or 0 3t 2 + 2t- 175
MR or MC
MR
200
25
4 8
Fig. 6.8
and sot= 7.31 years. The maximum total profit is earned if the machine
is scrapped after 7.31 years and this is
[175t- ( 2 - ~ = 835.20
If the machine were scrapped after 5 years the loss of revenue compared
with scrapping after 7.31 years is the area from t = 5 tot= 7.31 or
[175t- ( 2 - ~ = 835.20- 725 = 110.20
Now suppose that the machine has a scrap value given by
S(t) = 1000- 119t
This changes the situation because the value of the project of installing
the machine now has two components: the profit earned and the scrap
value. If the machine is scrapped at time T the total revenue is
R(T) = s: (MR(t) - MC(t) )dt + S(T)
175T- P - P + 1000- 119T

= 56T - P - P + 1000
To maximise R with respect to T we proceed as in section 5.8:
dR =56- 2T- 3T2

dT
and setting this to zero gives T = 4, which is a maximum since the
second-order derivative is negative. At T = 4 the value of R is 1144.
6.11 Exercises
1. Determine the value of the consumer's surplus if the demand curve
is p = 100 - 6q and the market price is 40.
2. The demand curve for a product is p = 60 - q - q 2 and the supply
function is p = 10 + 4q. Determine the equilibrium price and
quantity and the value of the consumer's surplus and the pro-
ducer's surplus at this equilibrium. What happens to the values of
consumer's surplus and producer's surplus if a tax of 2 per unit is
imposed on the supply of the product?
3. The marginal revenue function for a particular product is MR
= 200 - 3q and marginal cost is given by MC = 2q. What is the
value of profit if q = 30? What is the maximum value of profit?
4. A marketing manager expects that, as a result of an advertising
campaign which is about to start in month 0, sales in month t will
be to the value of S(t) = 25 + 30t 2 - 4t 3 and the cost of the
campaign in month t will be C(t) = 15 + St. What is the total net
revenue expected to be in the first 4 months? The campaign is
planned to last 6 months. Is it worthwhile extending it to 8 months?
5. A new machine is installed which is expected to produce a mar-
ginal revenue at timet of MR = 600- lOt- 3f and the marginal
cost function is MC = 50 + 1St. Determine the economic life of the
machine and the expected total net revenue over its life. If the
machine has a scrap value given by S(t) = 5000 - 300t, when
should the machine be scrapped?
6.12 Numerical methods of integration

So far we have assumed that it is possible to evaluate the definite
integral directly. However, in some cases it may prove to be difficult to
find the integral of the function in question, while in other cases the
function may not be known and the problem is one of finding the area
(x, y,)
y
(x, y,)
a b X
Fig. 6.9
under a curve defined by a table of values. We will consider two

procedures which give approximate, numerical solutions to these prob-
lems. Each assumes that the function is smooth and has no discon-
tinuities in the range of interest. The first is known as Simpson's rule,
and relies on obtaining a table of values of the function, while the
second uses Taylor's theorem which relies on being able to evaluate
derivatives of the function.
Using Simpson's rule involves first splitting the area to be evaluated
into an even number, n, of parts with equal width. The area under each
pair of parts is then evaluated approximately by using a parabolic arc.
Suppose that the problem is to obtain the area under f(x) between
x = a and x = b, and that the distance between a and b is split into
n parts, each of width c = (b - a)!n. Here n is an even number.
Referring to Fig. 6.9, where for simplicity it is assumed that the whole of
the area of interest is above the x-axis , the consecutive x-values are x 0 ,
~ X2 , • • • , xn, with the first corresponding to a and the last to b. The
area under the parabola passing through the three points (x 0 , y0 ), (x 1 , y 1)
and (x 2 , y 2 ) is given by
area(l) = c(y0 + 4y 1 + Y2 )13
This is an approximation to the area under the curve but it will be a close
approximation if c is small. The corresponding area under the parabola
passing through the three points (x 2, y 2 ), (x 3, y 3 ) and (x 4, y 4) is given by
area(2) = c(y 2 + 4y 3 + y4)/3,
and similarly, up to the final area, through the three points (xn_ 2, Yn- 2),
(xn_ 1 , Yn- 1) and (xn, Yn) which is given by
area(n/2) = c(Yn-2 + 4Yn-l + Yn)/3
The total area is the sum of all these and, noting that the last term in
each area expression is the first term in the next one, Simpson's rule
gives the approximate integration as
Lf(x)dx = c(yo + 4yl + 2y2 + 4y3 + 2y4 + ... + 2Yn-2 + 4Yn-l + Yn)/3,
b
where cis the constant difference between any two consecutive x values.
The coefficients on the odd values of y are 4 while those on the even
values are 2, with the exceptions of the first and last.
To illustrate Simpson's rule we will use it to evaluate approximately
the area under the curve y = x- 1 between the values x = 1 and x = 4 and
compare it with the exact value.
First the values of n and c need to be chosen. As stated above,
n should be even, and the larger the value the more accurate the
approximation will be. Since the area between x = 1 and x = 4 is
required, and 4- 1 = 3, which is not even, we might choose n = 6 with c
= 0.5. Next we evaluate y = x- for x = 1, 1.5, 2, ... , 3.5, 4. This gives
1
the following set of values, where we work to four places after the
decimal point and include the value of the subscript, i on y:
0 1 2 3 4 5 6
X 1 1.5 2 2.5 3 3.5 4
y 1 0.6667 0.5000 0.4000 0.3333 0.2857 0.2500
The approximate area is therefore,

J; X- dx = c(y
1 + 4yl + 2y2 + 4y3 + 2y4 + 4y5 + y6 )13
0
= 0.5(1 + 2.6668 + 1 + 1.6 + 0.6666 + 1.1428 + 0.2500)/3

= 0.5(8.3262)/3 = 1.3877
Direct integration gives the exact value as

r x- 1dx = [log x]i = log4 - log1 = 1.3863
The approximate value is close to the accurate value here, and a more
accurate approximation could have been obtained by taking a smaller
value of c.
As a second example suppose that we have the following data on
marginal cost (MC) and level of output (q) and we require the total cost
of increasing output from q = 10 to q = 50,
q 10 20 30 40 50
MC 15 16 20 30 50
Here n = 4 (since the area is split into 4 parts) and c = 10 so Simpson's

rule gives the approximate area as
I 50
10
MCdq = c(MC + 4MC + 2MC + 4MC + MC4)/3
0 1 2 3
= 10(15 + 64 + 40 + 120 + 50)/3

= 10(289)/3 = 963.3
In this example it is not possible to evaluate the integral directly since
the functional form is not given.
The second method of numerical integration which we consider
uses Taylor's theorem (see section 5.15). This states that
x2 xn
f(a + x) = f(a) + xf'(a) + 2 ! f"(a) + · · · + n! J<n> (a) + · · ·
where the primes indicate derivatives. It is assumed that this series

converges and if it does not, this method should not be used. Now since
we wish to integrate f(x) rather than f(a + x) it is convenient to let
z = a + x so that x = z - a and substituting these into the series
expansion gives
(z - a) 2 (z - a)n
f(z) = f(a) + (z - a)f'(a) + 2! f"(a)+ · · · + n! f<n> (a)+· · ·
and replacing z by x results in the value of f(x) as

(x- a)2 (x- aY
f(x) = f(a) + (x - a)f'(a) + f"(a)+ · · · + f<n> (a)+· · ·
2! n!
To integrate the left hand side with respect to x we just integrate the
right hand side, noting that we have a polynomial in x, which is simple to
integrate. The terms f(a), f'(a), f"(a), ... are all constants and, be-
cause differentiation is generally straightforward, are easily obtained.
Of course, if the function is a polynomial then the Taylor's series
expansion will give the same polynomial and so no benefit arises from
using this method. The value of a, around which the series expansion is
taken, is usually chosen to be in the middle of the interval over which
integration is taking place. The general statement is
(x- aY
I f(x)dx =I
c
b
c
b
{f(a) + (x - a)f'(a) + 2 ,. f"(a) +···
(x- aY
+ f(n) (a)+ · · · }dx
n!
(x af (x a) 3
= [f(a)x + 2! f'(a) + 3! f"(a) + ···~
To illustrate the use of Taylor's theorem we will take the first
example from p. 244 above. We require the area under the curve y = x- 1
between x = 1 and x = 4. Taking the first three derivatives we have
f(x) = x- 1 ,f'(x) = -x-2 ,f"(x) = 2x-3 ,f'" (x) = -6x- 4
Choosing a = 2.5, being halfway between the limits of 1 and 4,
f(2.5) = 0.4, f' (2.5) = - 0.16, f"(2.5) = 0.128, !"'(2.5) = -0.1536
and so the integral of Taylor's expansion gives
(x - a)2 (x - a)3
I 4
1
f(x)dx = [f(a)x + 2! f'(a) + 3! f"(a) + · · · ~
(x - 2.5) 2 (x - 2.5) 3
= [0.4x + 2! (- 0.16) + 3! (0.128) + ·· · ~
(1.6 - 0.18 + 0.072 - ... ) - (0.4 - 0.18 - 0.072 - ... )
(1.492) - (0.148) = 1.344
The accurate value is 1.3863 and so the Taylor's expansion value is
reasonably close. Notice that the third terms in the expansion are
relatively large. A closer approximation would result from taking
account of extra terms in the expansion.
As a second example of the use of Taylor's theorem we evaluate the

area between y = (1 - x 2 )- 1 between x = -0.5 and x = 0.5. The first
four derivatives are
f'(x) = 2x(l - x 2 )-Z, f"(x) = (2 + 4x 2 - 6x4 )(1 - x 2 )- 4
f"'(x) = 24x(1 - x )(1 - x 2 )- 5
4
f""(x) = (24 + 216x 2 - 120x4 - 120x6 )(1 - x 2 )- 6

Since the limits are -0.5 and +0.5 we choose a = 0 so that
f(O) = 1, f' (0) = 0, f"(O) = 2,/"'(0) = 0, f""(O) = 24
and the integral of the series expansion is
r::f(x)dx = [f(a)x +
(x
;! a) 2
f'(a) +
(x
;! a) 3
f"(a)+ .. ·~
x3 xs
[X + -3 + -5 . . .)o.s
-0.5
(0.5 + 0.0417 + 0.00625 + .. ·)

- (-0.5- 0.0417- 0.00625- .. ·)
= 0.54795 + 0.54795 = 1.0959
This is the approximate value of the area between x = -0.5 and x = 0.5,
the x-axis and the curve y = ( 1 - x 2 )- 1 • The accurate value is, using the
r::
method of partial fractions explained in section 6.14,
(1 - X 2 )- 1dx = 0.5[1og{(1 + x)/(1 - ~

= 0.5(1og3 - log0.3333)
0.5(1.0986 - { -1.0986})
1.0986
Here the Taylor's theorem approximation is close to the accurate value
and a more accurate approximation would be obtained by taking more
terms in the series expansion.
6.13 Exercises
1. Evaluate the area between the x-axis, the curve y = 4x 3 and the
values x = 1 and x = 3 by (a) exact integration and (b) Simpson's
rule.
2. Use Simpson's rule to obtain the total cost of expanding output

from q = 10 to q = 14 given the following marginal cost data:
q 10 11 12 13 14
MC 3 4 9 10 15
3. Obtain the approximate value of the integral J: x dx using Simp-

4
son's rule with (a) n = 2 and (b) n = 8 and compare these with the
exact value.
4. To see how important the choice of the value of 'a' is in integration
by use of Taylor's series, use the working in the text (p. 244) to
integrate y = r 1 between x = 1 and x = 4 with (a) a = 1 and (b)
a = 3, and then (c) compare with the results for a = 2.5 and the
J:
exact value.
5. Evaluate exdx by (a) exact integration (b) Simpson's rule with
n = 4 and (c) Taylor's series using the first three derivatives.
6.14 More difficult integration
INTEGRATION BY pARTS
Integration by parts can be thought of as the reverse of differentia-
tion of a product. We know that if u and v are functions of x then
d dv du
-(uv) = u-+ v -
dx dx dx
Integrating both sides of this, we have
d dv du
dx J
-(uv)dx = J + Jv-dx
u-dx
dx dx
uv = Ju dv + Jv du
:. Ju dv = uv - Jv du
Thus to find the integral of a function such as xex we can let x = u and
ex dx = dv.
Then v = Jdv = Jex dx = ex' and du = dx
:. Jxex dx
ex(x- 1) + c
We can check this by differentiation:
~ [ex(x - 1) + C] = ex(x - 1) + ex = xex

This method is of great importance in integration but care must be
taken in deciding which part of the function should be represented by u
and which by v. It is usually better to represent by dv that part which can
be integrated most easily. If both parts are simple to integrate then
represent by u that part which is reduced to a constant value by dif-
ferentiation in the shortest possible time. There is, however, no hard
and fast rule about this selection. The following examples have been
chosen to show the method of selection in a few typical cases.
Example 1
Jx2ex dx
Let and ex dx = dv
Then f x2ex dx = x2ex - f ex 2x dx
= X2ex - 2 f exx dx
The second part of this must be integrated by the same method
Let x=u and
Then Jxex dx
xex- f ex dx
xex- ex+ c
= ex(x- 1) + c
Then taking the two parts together, we have
f x2ex dx = X2ex - 2ex(x - 1) + c
= ex(x 2 - 2x + 2) + c
It is not necessary to represent the terms by u and v in the order in
which they appear in the function as the second part of Example 1 and
Example 2 show.
Example 2
Jx ~ dx
Let loge x = u and x dx = dv
Then !X dx = du and h =v+C

2
.. Jx loge x dx = J ~ x )x dx
~ ~ - J ~ ( ~ dx
= !x2 lo& x - ~ Jx dx
~ X2 loge X - !x +2 C
~ x - !) +C
It may be necessary to use more than two stages to complete the
integration, but when this happens we should be reasonably certain that
we are applying the appropriate technique. In general, this method
reduces the part to be integrated to a simpler form at each stage.
THE METHOD OF p ARTJAL FRACTIONS

In attempting to integrate some functions it is convenient to express
the function as a series of partial fractions. For example, to integrate
(a 2 - x2 )- 1 notice that
a2 - x2 = (a + x)(a - x)
and hence
1 1 A B
---= =--+--
a2 - x 2 (a + x)(a - x) a + x a- x
where A and B are as yet unknown constants. Their values are deter-
mined as follows:
1 A B A(a - x) + B(a + x)
-=---=- =- - + -
a2-x2 a+x
- = ----::---::----
a-x a 2 -x2
Since the denominators are equal then
1 = A(a - x)+ B(a + x)
(Aa + Ba) + x(B - A)
This can be true only if the coefficients of x in this equation are equal, so
that
O=B-A
and also the constant terms are equal so that
1 = Aa + Ba
From these two equations the values of A and B which satisfy the
identity can be found and the result is
1 1
A= 2a' B =-
2a
The original problem can now be restated as
1 112a ]
f (a 2 - x 2)
dx=f [ (a112a
+ x) + (a - x) dx
=-J
1 [ 1 1
2a (a + x)dx+J(a - x) dx
]
1
= 2a [loge (a + x) - loge (a - x)] + C
The negative sign in front of the second term is due to the fact that
the denominator of the second fraction has a negative sign prefixing the
variable x. Now the difference between the logarithms of two numbers
can be written as the logarithm of the quotient of the numbers:
f (a-x
1
/X =21-a loge (a+x)
2 2 -- + C
a-x
The method of partial fractions can be extended to cover cases

where the denominator is of higher degree than the second. It may be
possible to split the denominator into several factors but this is likely to
result in a factor which is a quadratic. In this case we let the numerator
of the fraction containing this factor be linear in x. For example,
1 dx:: [ A Bx + C ]
f (x + 2)(x + 3x + 1)
2 f (x + 2) + (x 2 + 3x + 1) dx
where 1 ::: A(x2 + 3x + 1) + (Bx + C)(x + 2)
= Ax2 + 3Ax + A + Bx2 + 2Bx + Cx + 2C

= x 2 (A + B) + x(3A + 2B + C) + (A + 2C)
Then equating the coefficients of x 2 , of x and of the constant term
on each side of the identity we obtain three equations in three un-
knowns.
A+B 0 (1)
3A + 2B + C = 0 (2)
A+ 2C = 1 (3)
Solving these equations gives
A = -1, B = 1, C= 1
1 -1 x+1
:. =--+-----
(x + 2)(x + 3x + 1) x + 2 x + 3x + 1
2 2
The first term on the right hand side is integrated easily:

-1
Jx + 2 dx = -log..(x + 2)
The second term on the right hand side requires further considera-
tion. It can be written as
x+1 ~
x + 3x + 1
2 x + 3x + 1
2
!(2x + 3 - 1)
x 2 +3x+1
1[ x
=2 2
2x+3 1 ]
+ 3x + 1 - x 2 + 3x + 1
Again, the first term on the right hand side can be integrated easily
because the numerator is the derivative of the denominator.
J21( x2 2x+3 )
+ 3x + 1 dx = ! log.. (x2 + 3x + 1)
The only term remaining is

-1
Jx2 +3x+l dx
Since - (x 2 + 3x + 1) = (D - (x + ~ Y and, from above,

1 1 (a+x) + C
Ja2- x2 dx =-loge--
2a a-x
1 2 ( ~ ~
c
then
f s4- (X+ 23Y dx = 2V5loge ~ ~
) +
Collecting the three parts together gives

1
J (x + 2)(x 2+ 3x + 1)dx = -loge(x + 2) + !lo&(x2 + 3x + 1)
1 ~ ~
+ 2v'5loge ~ - X - ~ + C
The result could be further simplified but this is left to the reader.
The method may appear rather complicated but it serves to illustrate the
way in which the integral of more complex functions can often be
obtained by breaking down the function in successive stages into a series
of terms for which the integral is known.
The major problem is to recognise the most suitable way in which
to break the function down in order to arrive at a standard form. Care
must then be taken in the application of such techniques as partial
fractions to ensure that the function which is finally integrated is in fact
identical in all respects with the original function.
INTEGRATION BY SUBSTITUTION
This can be illustrated by a simple example where the result can be
recognised at a glance.
Example 3
Let
du
Then -=2x
dx
and we treat this as du = 2x dx
Substituting these values in the original expression, we have
2x 1
fr + 2 dx = f-; du = loge u + c
Then substituting back the value (x 2 + 2) for u we have
2x
fx 2 + 2 dx = loge{x2 + 2) + C
This result was, of course, obvious. A more realistic example is the

following.
Example 4
J(4x- 3) 6 dx
This could be obtained by use of the binomial expansion. However,
let u = 4x- 3 so that du = 4 dx
u6 u7
Then J (4x- 3) 6 dx = J4 du = 28 + C
(4x- 3f
=
28
+ c
Other examples of integration by substitution are provided in
Appendix A, sections A.6 and A. 7, where trigonometric substitutions
are used. Some standard integrals considered in this chapter are given in
Table 6.1.
6.15 Exercises
Integrate the following expressions with respect to x.
1. (5x 2 + 4)e3x
3 X
3 · (x 4.------
+ 1)(x- 1) (x - 2)(x 2 - 3)
5. (4x + 3) 10 6. loge X

1. Integrate the following functions with respect to x:
(a) y = 3x 2 - 6x + 4
(b) y = 2x4- 3x-3
TABLE 6.1
Some standard integrals
(The constants of integration are omitted.)
Function Integral Conditions

axn+J
axn n=l=-1
n+1
1
log., x
X
1 du
log., u u is a function of x
udx
a ebx
a ebx
b
COS X sin x
sin x -cosx
u dv uv- Jv du u, v are functions of x
(c) y = 3x/(x2 + 5)
(d) y = (3x2 - 6)/(x3 - 6x)
(e) y = 3ex - 2e- 4x
(f) y = sin x - cos x + tan x.
2. If marginal revenue is given by MR = 50 - 2q where q is quantity
sold, and total revenue is zero when q is zero, determine the
equation of the total revenue curve. What is the elasticity of
demand when q = 25?
3. Evaluate the area between x = 1 and x = 8 if
(a) y = 8x- 3
(b) y = x4 + 2._x-l
(c) y = 2e-o.4x
(d) y = 2x/(2x2 + 4).
4. For a product the marginal revenue is given by MR = 170 - 3q
where q is the level of sales. What is the extra revenue gained by
increasing sales from 40 to 50 units?
5. Assuming that the marginal cost curve is linear, and that when q
(output) is 5, marginal cost is 140, while q is 10, marginal cost is
130, find the total cost of increasing output from 50 to 60.
6. Sketch the following curves and hence evaluate the absolute areas
between the curves, the x-axis and the given values of x:
(a) y = x 2 - 1 between x = -2 and x = 2
(b) y = x 3 between x = -3 and x = 1
(c) y = x2 - 5x + 6 between x = 0 and x = 6.
7. The demand curve for a product is p = 200 - 3q - 2q 2 and the
supply curve is p = 4 + 11q. At the equilibrium price determine
the elasticity of demand, the consumer's surplus and the pro-
ducer's surplus. Sketch the curves and indicate the consumer's
surplus and the producer's surplus.
8. If the marginal revenue function for a product is MR = 300 - 4q
and marginal cost is given by MC = 6q + 10, what is the max-
imum value of profit (ignoring fixed costs)?
9. Evaluate the integral of y = ll(x- 1) between x = 1 and x = 5 by
(a) direct integration
(b) using Simpson's rule with n = 4
(c) using Taylor's series with four terms and compare the results.
10. Obtain the integrals of the following functions:
(a) y = xe-2x
(b) y = log(2x- 4)
(c) y = 3x/(x + 2)(x2 - 1)
(d) y = (x 2 - 1)'5
(e) y = x sin x
(f) y = cos 3x.
Chapter 7
Partial Differentiation
7.1 Functions of more than one variable
So far we have been concerned with the relationship between pairs
of variables, such as between y and x in calculus and between price and
quantity in demand analysis. In doing this it has been necessary to make
the common assumption that ceteris paribus, or that everything else is
unchanged. However, if our methods are to be useful in the real world,
we need to relax this assumption and take account of other variables
which are known to be important. For example, the demand for a good
depends not only on its own price but also on other prices, on con-
sumers' incomes, on the amount of advertising and on other factors
which are generally ignored such as the weather, the location of shops
and the availability of credit. Thus the demand function for a particular
brand of ice cream might be of the form
q = f(po, p,, Pz, Y, A, W, ... )
where q is the quantity demanded, p 0 is the price of the brand, p, is the
price of a competing brand, p 2 is a measure of the general price level, Y
is the level of consumers' income, A is expenditure on advertising, W is
an index of the weather, and so on. We need to become familiar with
methods which allow us to handle such a situation.
These methods are a logical extension of the differential calculus,
but to avoid confusion a new and slightly different notation is required.
To introduce this notation and to illustrate its use, let us consider a
three-variable model which it is possible to represent in diagrammatic
form.
Let x and y be two independent variables and z the dependent
variable. Then z = f(x, y) and if we represent x andy on axes at right
angles to each other in the horizontal plane we can imagine z as being on
an axis in the vertical plane and, therefore, at right angles to both x and
y. Any variation in x or y then causes a change in z and the point
257
Fig. 7.1
representing z moves, in fact, in such a way as to trace out a surface

above the plane xy rather like a hill or series of hills on the earth's
surface.
The point Pin Fig. 7.1 has a height above the xy-plane equal to z,
and therefore its z-coordinate is given by z = ~ To determine uniquely
the location of P in the three-dimensional space it is also necessary to
know the distance of the point from the yz-plane and from the xz-plane.
These distances are x = X1 and y = y1 respectively and the point P can
then be specified uniquely in relation to the three axes by the co-
ordinates (x1, ~ z1)·
Because z = f(x, y) the point P moves over the surface ABCD as
the values of x andy change, and we are interested in measuring the rate
of change of z with respect to changes in the other variables. In fact,
what is required is some value similar to the derivative, which can be
used in the three-variable case to measure rates of change.
Partial Differentiation 259
Fig. 7.2
This gives us a basis on which to work. If one of the variables, say y,

is considered to be fixed, then this is equivalent to taking a section
through the surface parallel to the x-axis as shown in Fig. 7 .2. Any
change in z on this section is then due entirely to the change in x.
Any point P on this section is, therefore, similar to the point in a
two-dimensional model. A tangent can be drawn to the surface at P
parallel to the x-axis and the slope of this line is a measure of the rate of
change of z with respect to x when y is kept constant. It is therefore a
form of derivative, but because it is the rate of change with respect to
only one variable, the other variable being held constant, it is called a
partial derivative, and to prevent confusion with dz I dx it is denoted by
iJzliJx or Zx.
In a similar manner it is possible to take a section through the
surface for a constant value of x and determine the rate of change of z as
y changes. This ratio is written a z I iJ y or zy and is a measure of the slope
of the section shown in Fig. 7.3.
It is not easy to obtain the value of the derivative from a diagram
and, as might be expected, it is not necessary because with only minor
modification, the differential calculus can be used to obtain the result
Fig. 7.3
very simply. The modification is necessary because zx refers to the

partial derivative of z with respect to the variable x and implies that all
other variables in the function must be considered as constant.
For example, if
z
az
= - = 2x
X dX
This follows because y 2 is considered to be a constant and the derivative
of a constant is zero. Similarly,
z = - = 2y
az
y ay
because x 2 is considered as a constant when the partial derivative with
respect to y is required.
The same procedure applies if the expression is more complicated.

For example,
z = x 2 + 3xy + y 2
To obtain the partial derivatives of z with respect to x only, we treat
y as though it were a constant wherever it appears.
oz
z =-=2x+3y
X OX
Similarly z
oz
= - = 3x + 2y
y oy
The second-order partial derivatives are defined in a similar way to
second-order derivatives, i.e. they are the result of differentiating the
first-order derivatives. We define
Here there are four second-order derivatives, Zxx and zyy which corre-
spond to d 2yldx 2 and also the two 'mixed' derivatives Zxy and Zyx- The
notation used indicates the order in which the variables occur, i.e. Zyx is
found by differentiating z with respect to x and then with respect to y.
For all values of x andy for which Zxy and Zyx are continuous:
and the order of differentiation does not matter.

For example z = x 2 + 3xy + 2y 3 - Sx - 4y + 20
oz
z =-=2x+3y-5
X OX
dZ
zy = - = 3x + 6y 2 - 4
oy
a 2
z XX
=-
OX (z) X
=
and hence
The above definitions of first- and second-order partial derivatives can
easily be generalised to cover the cases of more variables and higher-
order derivatives.
One application of partial derivatives is in testing for functional
dependence. Suppose that w = f(x, y) and z = g(x, y) so that the
first-order partial derivatives are wx, wy, Zx and zy; then, using determi-
nants (see section 1.13), if the special determinant,
IJI =
Zx Zy
is zero, wand z are linearly or non-linearly related. Here IJI is called a

Jacobian determinant or simply a Jacobian. For example, consider
W =X+ 2y
z = X2 + 4xy + 4y 2
then the first-order partial derivatives are
Wx = 1, Wy = 2, Zx = 2x + 4y, Zy = 4x + 8y
and
1 2
IJI = = 4x + 8y - (4x + 8y) = 0
2x + 4y 4x + 8y
Since the Jacobian is zero, w and z are related and it can be seen that
z = w2 •
7.2 Marginal analysis

This is an extension of the two-dimensional model and is best
explained by the use of an example.
Let us assume a function in which the level of output is related to
the quantity of inputs x and y in the following way:
q = f(x, y) = SOx - x 2 + 60y - 2y2
then iJq =50- 2x

iJx
iJq
- =60- 4y
ay
These partial derivatives denote the marginal productivity of factor
x and factory respectively, and in this example they are a function of the
amount already in use. That is, if we are at present using 10 units of each
factor then
oq = so - 2 x to = 30
ox
iJq = 60 - 4 X 10 = 20
iJy
and the marginal productivity of x is 30 whilst that of y is 20.
An example which is more useful in economics is the Cobb-
Douglas production function
q = ALaKP
where q = output
K = capital input
L = labour input
A, a and {3 are constants. The marginal product of labour is
oq aALaKP aq
qL = iJL = aALa-IK/3 = L L
and the marginal product of capital is

oq {3AUKP {3q
q = - = {3ALaKf3-t = =-
K aK K K
It can easily be seen that these marginal products are positive if a > 0
and {3 > 0, since q, K and L are all positive. Now the average product of
labour is q I L and the average product of capital is q I K and so the above
equations can be rearranged to give
qL marginal product of labour
a=--=
qI L average product of labour
{3 = ..!l.!:_ = marginal product of capital
and
qI K average product of capital
It is also easily seen that if the wage is qL and the cost of capital is qK, so
that each factor is paid its marginal product, then the total cost is
qL L + qK K = aq + {3q = (a + {3)q
This will equal output, q, if a + {3 = 1, which occurs when there are
constant returns to scale (see section 7.10).
The way in which these marginal products are changing can be seen
from the second-order derivatives
It is usually a requirement of production functions that the marginal

products of capital and labour should, at least for low values of inputs,
be positive, and as the values of the inputs increase the marginal
products decline. That is, qK > 0, qL > 0, and qKK < 0, qLL < 0. This
occurs with the Cobb-Douglas production function if 1 > a > 0 and
1 > {3 > 0.
7.3 Elasticity of demand

If the quantity demanded of a good (A) is determined by the price
of that good and also the price of a substitute (B) then the demand
function can be written as
qD = f(pA, PB)
Here there are two important elasticities. The first concerns the change
in the quantity of A demanded when the price of A changes and the
price of B is held constant. This can be written in partial derivatives as
or
and is referred to as the partial elasticity of demand for good A with

respect to the price of good A, or the own-price elasticity of demand for
good A. It is expected to be negative for normal goods.
For example, suppose that
qD = f(pA, PB) = 100 - 3pA + PB
then ()qD = -3
dpA
PA
partial elasticity of demand =- 3
If the present prices are PA = 10, p 8 = 20

qD = 100 - 3(10) + 20 = 90
10
partial elasticity of demand = - 3 · - = -!
90
This can be interpreted as meaning that the quantity demanded of
A falls by !% if its price is increased by 1% and the price of B remains
unchanged.
The second elasticity gives the reaction of the quantity of good A
demanded to a change in the price of good B. Using partial derivatives
this is
or
and is called the partial elasticity of demand for good A with respect to
the price of good B, or the cross-elasticity of demand for A with respect to
B. If A and B are complementary goods, this cross-elasticity will be
negative, while if they are substitutes it will be positive. The higher the
value of the cross-elasticity the stronger is the degree of substitutability
or complementarity of A and B.
In the example used previously, when p 8 = 20,
iJqD
- = 1 and qv = 90
iJpB
The cross-elasticity of demand = 1 (20) = 0.22
90
Therefore the demand for A increases by 0.22% if its price remains

unchanged and the price of B increases by 1%. Here A and B are
substitutes.
This analysis extends naturally to the case of more than two inde-
pendent variables. The partial derivative with respect to any one of the
variables is obtained by differentiating the function with respect to this
variable and treating all other variables as constants.
For example, suppose that
aqD
then - - = a2
ap2
aqD
--=a
ap4 4
As a further example consider the demand function

q =A pf pf1 P
where q is the quantity of a good demanded, p, is the price of the good,
p 2 is the price of a competing good and Y is the value of consumers'
income. Here it is convenient to take logarithms (to the base e), giving
log q = log A + a log p, + f3 log p 2 + r log Y
The partial derivatives are obtained as above by differentiating with
respect to each variable in turn and treating the other variables as
constants. The function of a function rule is used to differentiate log q
(see section 5.2). The partial derivative with respect top, is
1 aq a p, aq
--=-ora=--
q ap, p, q, ap,
and thus the own-price elasticity of demand is a. Similarly, the cross-
elasticity is f3 and the income elasticity is r. Since all of these are
unrelated to q and the other variables, it is clear that the particular
function under consideration is a constant elasticity demand function.
7.4 Exercises
1. Find a z I ax and a z I a y for the following functions:
(a) z = xy + x 2y + xi
(b) z = 3x2y + 4xy 2 + 6xy
(c) z = x sin y + y cos x

2 2
(d) z = (x + y)e<x+y)
(e) z = log (x + y2 2)
x+y
(f) z = --
x-y
x + sin y
(g) z = ---'-
y +sin x
2. For the production function
q = 2(K - 30) 2 + 3(L - 20)2 + 2(K- 30) (L - 20)
show that the marginal products of capital (K) and labour (L) are
equal when K = 50 and L = 30.
3. Let the demand for apples be given by
q = 240 - ~ + 6po - PaPo
where q is the quantity of apples demanded at price Pa• and Po is
the price of oranges. Evaluate the partial elasticity of demand for
apples when Pa = 5 and p 0 = 4 with respect to
(a) the price of apples
(b) the price of oranges.
4. For a production function Q = f(K, L) the partial elasticity of
output with respect to the labour input is defined as ( aQ I aL)
(L! Q). Show that for the Cobb-Douglas function Q = AL aKf3 this
partial elasticity is a.
5. In section 1.13 we saw that the condition for the linear equations
to have a unique solution (and so be independent) was that IAI =I= 0.

That is, if IAI = 0 the equations are dependent. Show that this
condition is the same as the requirement that the Jacobian is zero.
7.5 Differentials
In the two-variable model, y = f(x), we obtained a value for the
magnitude of the change in y corresponding to a small but finite change
in x by using the expression

dy
dy = -dx
dx
This procedure can be extended when there are more variables in the
model. In fact, for z = f(x, y) the small change in z is equal to the small
change in x multiplied by the rate of change of z with respect to x,
keeping y constant, plus the small change in y multiplied by the rate of
change of z with respect to y keeping x constant. This is conveniently
written as
dZ dZ
dz = - dx +- dy
dX dy
This can be extended to any number of variables. For example,

suppose z = f(p, q, r, s, ... )
dZ dZ dZ dZ
then dz = - dp +- dq +- dr + - ds + · · ·
dp dq dr ds
This expression is known as the total differential of z.
This can be illustrated by reference to the production function
relating the quantity of wheat produced to the quantity of labour (x) and
land (y) which is employed.
q = 40x - X2 + 60y - 2y2
dq
-=40-2x
dX
dq
-=60- 4y
ay
dq dq
dq=-dx + -dy
dX dy
= (40- 2x) dx + (60- 4y) dy
This equation can be used to obtain an approximate value for the
increased production which is possible when both labour and land are
increased together. The smaller the increases in the variables dx and dy
that are considered, the more accurate the answer will be. However, an
approximate answer can be obtained when the supply of each factor is

increased by 1 unit. If the present quantity in use is given by x = 10
units, y = 10 units, then the approximate increase will be given by
dq = [40 - 2(10)] + [60 - 4(10)]
=20+20
= 40
The accurate value of the increase can be obtained by substituting
x = 11, y = 11 in the production function to give q = 737 and substract-
ing the value of q when x = 10, y = 10. The accurate increase is 37 units.
Differentials are also useful with indifference curves. If the utility
function is
U = f(qH q2)
where q, and q2 are the quantities of two goods consumed, then a given
level of utility, u0 , can be achieved with different combinations of q, and
q 2 • The expression
U0 = f(q,, q2)
gives the indifference curve and the total differential of u0 is
at
du 0 = - dq, +-
at dq2
iJq, iJq2
Since u0 is a constant (i.e. any movement is along the indifference curve)
then du 0 = 0 and
0 = f, dq, + !2 dq2
-dq 2 = f, =marginal utility of good 1 = MRS
or
dq, f 2 marginal utility of good' 2
This ratio is the marginal rate of substitution (MRS) or rate of
technical substitution of q, for q 2 and is the negative of the slope of the
indifference curve at any point. It gives the number of units of good 2
which must be given up in exchange for an extra unit of good 1 while
keeping utility constant. For example, if MRS = 2 then we have
dq2 = -2dq,
and so increasing the consumption of good 1 by one unit requires a
reduction of two units in the consumption of good 2.
The same concept can be applied to production functions where an

isoquant is the locus of all technically efficient methods of producing a
given level of output. If output (Q) depends on the level of capital
services (K) and labour services (L) then
Q = f(K, L)
and an isoquant is
Qo = f(K, L)
Moving along the isoquant, so that dQ 0 = 0,
dQ 0 = QK iJK + QL iJL =0
and solving, -° K
iJL
= QL
QK
= MRS
which is the marginal rate of substitution of labour for capital.

The marginal rate of substitution has the disadvantage that its size
depends on the units of measurement of K and L. An alternative is to
define the elasticity of substitution ( o) as the percentage change in the
capital-labour ratio as a proportion of the percentage change in the
marginal rate of substitution. Therefore,
percentage change in Kl L
o=
percentage change in MRS
d(KI L)I(KI L)
d(MRS)!(MRS)
To illustrate these concepts we will consider the Cobb-Douglas
production function (see section 7.2)
Q = AUKfl
for which the marginal products of labour and capital are
Q = aQ
L L
The marginal rate of substitution of labour for capital is
MRS= QL = aQ! L = a K
QK f3Q/K f3 L
Now the elasticity of substitution is
d(KI L)I(KI L)
a=
d(MRS)I(MRS)
d(KI L) (MRS)
d(MRS) KIL
and rearranging the expression for MRS,
Kl L =MRS ({3/a)
so that, differentiating Kl L with respect to MRS gives
d(KI L) f3
=
d(MRS) a
f3 MRS f3 (aK/f3L)
and hence a=---=- =1
a KIL a (K/L)
Therefore the elasticity of substitution for a Cobb-Douglas production
function is 1.
This raises the question of what kind of function has a constant
elasticity of substitution ( CES) which differs from 1. The CES produc-
tion function is defined as
Q = r[(1 - ~ + DL ~ ~ ~
where r, D, cp and Q are parameters. Here ds the efficiency parameter,

so that the larger the value of r the greater the level of output for given
K and L, t5 is related to the marginal products of K and L, Q is the
returns to scale parameter which will be discussed in section 7.10 below,
and we will show that cp is related to the elasticity of substitution. To
simplify the algebra we will let r = 1 and Q = 1 so that
Q = ((1 - ~ + t5L ~ ~
The marginal products are obtained using the 'function of a function'

rule:
QL = (- 1/c!J)b(- c!J)L -4>- 1 [(1 - b)K-4> + DL -4>]-(1l4>)- 1
t5 Q1+4>
L1+4>
QK = (-llc!J)(1 -D) ( ~ [(1 - ~ + DL -4>]-< 114>>- 1
(1 - t5)Q 1+4>
~
and the marginal rate of substitution is
MRS= QL = D
QK (1 -D)
and rearranging,
Kl L = [(1 - ~ ~
The elasticity of substitution is

d(K!L) MRS
a=
d(MRS) KIL
and since d(KI L) = _1_ (1 - D) [(1 - ~

d(MRS) 1 + fP D
1 (1 - D)[(1 - ~ ~ MRS
we have a=--
1 + fP D [(1 - ~ ~
and simplifying this expression gives the result

1
a=--
1 + rp
Thus the elasticity of substitution is constant and is related to fP.
The ideas of differentials can also be used in IS-LM models. Sup-
pose that the IS equation is
Y = C(Y) + I(r) + G
where Y is national income, Cis aggregate consumption, G is govern-
ment expenditure and r is the rate of interest. We assume that the
partial derivatives satisfy
0 < Cy < 1 I,< 0
The LM equation is M/P = L(r, Y)
where M is the quantity of money and Pis the price level and L, < 0, Ly
> 0. Here Y and rare endogenous and G and Pare assumed to be fixed.
To determine the effect on Y of an increase in M we find the total
differentials of the IS and LM curves. Taking these in turn:
dY= CydY+ I4r + dG

p -
dM M
p dP = L4r + LydY
Setting dG = dP = 0 and rearranging gives

(1 - Cy)dY- I,dr = 0
Ly dY + L,dr = dMIP
These can be solved by Cramer's rule (see section 1.13) or by substitu-
tion. Since we are interested in dY/dM we solve the first equation for dr
and substitute into the second to give
Ly dY + L,(l - Cy)dY/1, = dMI P
or dY[Ly + L,(l - Cy)!I,] = dM! P
dY IJP
and so
dM Lyl, + L,(l - Cy)
This gives the effect on Y of a change in M. Now we know that I,< 0, Ly
> 0, L, < 0 and 0 < Cy < 1. The numerator is therefore negative (since
P > 0). The product Lyl, is also negative, as is L,(l - Cy). Therefore
the ratio must be positive and so an increase in M will result in an
increase in Y in this model.
7.6 Total derivatives

Many situations arise where the independent variables can all be
expressed as a function of a single variable such as time. For example,
z = x2 + y2
with x = e1 cost, y = e1 sin t
What is required is the change in z which is brought about by the
change in t as the latter affects both x and y. This can be obtained by
substituting the function of t for x and y in the main equation and then
differentiating this expression with respect to t. The same result can be
obtained in a simpler way by making use of the results of section 7.5.
()z iJz
dz = ax dx + ay dy
Dividing through this expression by dt

dz = az dx + az dy
dt ax dt ay dt
This is similar to the function of a function or chain rule discussed in
section 5 .2.
For example in the case where z = x 2 + y 2 and x = e' cos t, y = e1 sin

t, we can form the derivatives
: = e1 cost - e sin t = e(cos t - sin t)
'Z = e1 cos t + e1 sin t = e( cos t + sin t)

dz = iJz dx + iJz dy
dt iJx dt iJy dt
= 2xe(cos t - sin t) + 2ye(cos t + sin t)
This can be rearranged by gathering together terms containing cos t
and terms containing sin t with the result
~~ = ~~ (2 cos2 t - 2 sin t cos t + 2 cos t sin t + 2 sin2 t) = ~

using result (4) of Appendix A.l.
7. 7 Exercises
1. The quantity of output of a product depends on the labour (L) and
capital (C) inputs in such a way that
q = L 3 - 3L + 4LC - C 2
If L = 10, C = 5, what is the approximate increase in q for an
increase of 1 in both Land C? What is the approximate increase in
q if L alone increases by 1?
2. The demand for apples (a) is related to the price of oranges (o) by
q = 240 - ~ + 6po - PaPo
What is the approximate change in the demand for apples if the
price of apples increased from Pa = 5 to Pa = 6 and the price of
oranges changes from Po = 4 to Po = 5?
3. Find dx/dt, dyldt and dzldt for the following functions:
(a) Z = X 2 - y 2 , X = e COSt, y = e1 sin t
(b) z = log., (x + y), x = t e,

2 y=sint
2
(c) z = ~ , x = e1, y = loge t
(d) z = x 2 + xy + y 2 , x = t + t, y = t - ~
4. The Cobb-Douglas production function can be generalised by
allowing technical progress to occur as time passes. This means
that new methods of production are used resulting in a higher level

of output for given levels of K and L. If these new processes are
introduced gradually their effect can be represented by a variable t,
known as a time trend, and if the rate of technical progress is r %
per annum then output at time t is given by
Q1 = ~ ~
Show that (a) the marginal rate of substitution is still ~ and
(b) the elasticity of substitution is still unity.
5. For the model Y= C +I+ G
C=a +bY
I= c- er
MIP=fY- gr
where a, b, c, e, f and g > 0, Y = income,
C = consumption, I= investment, r = interest rate,
M = quantity of money and P = price level, and the variables
Y, C, I and M are endogenous:
(a) What is the effect on Y of a change in r (assuming that G is
fixed)?
(b) What is the effect on M of a change in G (assuming that r and
P are fixed)?
7.8 Implicit differentiation

The functions we have met have all been of the form Y = f(x) or
Q = f( K, L) which are explicit functions, with the dependent variable on
the left hand side. In contrast,
x3 + 2xzy = 3xyz + y3
is an implicit function in which there is a complicated relationship
between y and x. With implicit functions it is not easy to see how the
derivative can be found. However, we know that if z = f(x, y) then,
allowing x to change by ~ and y by ~ the total change in z is
approximately,
dz iJz
~ = ~ + ~
dx iJy
Dividing this throughout by ~ we have approximately
dZ = dZ ~ + dz ~
dX dX ~ iJy ~
and as bx ~ this becomes

dz = az + az dy
dx ax ay dx
If we have a relation in which z = 0 (identically), then dzldx also
equals 0.
This fact can be used as follows:
Suppose z = x 3 + 2x 2y - 3xy2 - y3 = 0
then dz =O
dx
az
- = 3x 2 + 4xy - 3y 2
ax
az
ay = 2xz - 6xy - 3yz
Substituting these values in the total derivative

dz = az + az dy
dx ax ay dx
results in
0 = (3x 2 + 4xy - 3y2 ) + (2x2 - 6xy - 3y 2 ) dy
dx
and rearranging these terms, we have
dy _ (3x 2 + 4xy - 3y 2 )
dx - (2x 2 - 6xy - 3y2 )
The above method can be generalised and used to determine the
derivative of any implicit function of two variables. The function is first
rearranged so that all the terms are on one side of the equality sign, i.e.
it is made equal to 0. Then dzldx is equal to zero and the following
holds:
0 = az + az dy
ax ay dx
dy = _ az1ax
dx az;ay
This is known as the implicit function rule.
For example, suppose
then z = y2 - 3xy - x 2 = 0
i)z
i)y = 2y- 3x
iJz
dX = - 3y - 2x = - (3y + 2x)
dy- 3y + 2x
and
dx- 2y- 3x
An alternative approach is to differentiate each term with respect to
x, remembering that the function of a function rule must be used on
such terms as y 2 and xy. For example, suppose
y 2 = 3xy + x2
Differentiating term by term gives
2y dy = 3x dy + 3y + 2x
dx dx
that is : (2y - 3x) = 3y + 2x
dy- 3y + 2x
or
dx- 2y- 3x
Similarly:
Differentiating gives
0 = 3x2 + 2x2 dy + 4xy - 6xy dy - 3y2 - 3y2 dy
dx dx dx
2 2 2
= 3x + 4xy - 3y + dy (2x - 6xy - 3y2)
dx
2 2
dy _ 3y - 3x - 4xy
dx - 2x2 - 6xy - 3y2
7.9 Exercises
1. Determine dy/dx if
(a) y 3 - 4x2y 2 + 6x2y + 4x = 0
(b) x y
2 2+ e - eY = 0
(c) 3 ~ + y) + 4x2 - y 2 = 0
(d) sin (x + y) - 3xy = l0xy 2
2. What is the elasticity of demand when p = 2 if

500 - p 2
q2 = + pq
2+p
3. The relationship between output (Q), capital input (K) and labour
input (L) for a particular process is
Q2 = SKL - 3Q + U + ~
Determine the marginal products of capital and labour, the mar-

ginal rate of substitution and the elasticity of substitution.
7.10 Homogeneous functions

A function is said to be homogeneous of degree k if
f(tx, ty) = tkf(x, y)
for all values oft, where t and k are constants.
An example of a homogeneous function of degree one is
z = f(x, y) = 2x + 3y
Replacing x by tx and y by ty gives
f(tx, ty) = 2(tx) + 3(ty) = t(2x + 3y)
= tf(x, y) = tz
The function
z = f(x, y) = X2 + 2xy + 3y2
is homogeneous of degree two. This can be shown by the method above
since
f(tx, ty) = (tx)l + 2(tx)(ty) + 3(ty)l
= t 2(x 2 + 2xy + 3y2 ) = t 2z
It can also be seen by noticing that if we take each additive term in the
expression for z and add the powers of its factors, we find that their sum
is 2. By this method it is easy to see that
z = f(x, y) =2 + 2x + 3y
is not homogeneous, since the powers of x and y in the first term are 0,
and the total power in each of the other terms is 1. Also,
f(tx, ty) = 2 + 2tx + 3ty =I= tz
A simple example of a homogeneous function of degree one occurs

in the production function
q = 3L + SK
where q is the level of output resulting from inputs of L units of labour
and K units of capital. If both L and K are increased by a factor t the
new level of output is
q' = 3(tL) + S(tK) = tq
and so doubling the inputs of capital and labour (that is, t = 2) doubles
the amount of output. This is a case of constant returns to scale.
If the degree of homogeneity k is less than one there are decreasing
returns to scale while if k is greater than one there are increasing returns
to scale.
For the Cobb-Douglas production function
q = ALaK{j
where A, a and {3 are constants, if the inputs are changed by a factor t
the new output is
q' = A (tL )a (tK)fj = A ta+fj La Kfj
= ta+fj (q)
Therefore the degree of homogeneity is a + ~ and if a + ~ = 1 there are
constant returns to scale. For a + ~ > 1 there are increasing returns to
scale and for a + ~ < 1 there are decreasing returns to scale.
As we saw in section 7.2 above, if there are constant returns to scale
and each factor is paid its marginal product then the total cost is the
value of output since
qLL + qKK =q
The idea of homogeneous functions has a useful economic applica-
tion in the context of money illusion. Consider the simple demand
function
q = f(p, Pc• Y)
where q is the quantity of a good demanded, pis its price, Pc is the price
of a competing product and Y is income. Money illusion occurs if
behaviour changes as a result of a change in the value of currency. For
example, if all prices and incomes were multiplied by 100, so that they
were quoted in pence instead of pounds (or cents instead of dollars), as
relative prices and real incomes are unchanged, there should be no change
in behaviour, otherwise there is money illusion. Now suppose all prices and
incomes change by a factor t, then the new quantity demanded is
q1 = f(tp, tpc, tY) = tk f(p, Pco Y)
where k is the degree of homogeneity of the demand function. For no
money illusion, so that q = ~ then k = 0 and the demand function must
be homogeneous of degree zero. An example of such a function is
q = b(pl P)a (p) P)fJ (YIP)"'
where P is the general price level. Here demand depends on real prices
(i.e. prices relative to the general price level) and real income. Alterna-
tively, consider
q = b pa JJ': Y'"
and let all prices and incomes change by a factor t, then
qJ = ta+{J+r b pa JJ': yr = ta+{J+T q
and the degree of homogeneity is a+f3+r so that the condition for no
money illusion is that a+/3+-r: = 0.
7.11 Euler's theorem

For the homogeneous function z = f(x, y) Euler's theorem states
that
XZx + yzy = kz
where k is the degree of homogeneity and zx and Zy are the partial
derivatives of z.
For example, if
z = x2 + 3y2 + 2xy
then Zx = 2x + 2y, Zy = 6y + 2x and k=2
Therefore, XZx + yzy = x(2x + 2y) + y(6y + 2x)
= 2r + 2xy + 6y2 + 2xy
= 2(x2 + 3y2 + 2xy) = 2z
For a homogeneous production function, where z is the level of output
and x, yare measure of inputs, Zx and zy are the marginal products of x
and y and so Euler's theorem shows that

(quantity of x) (marginal product of x)
+ (quantity of y) (marginal product of y) = k (total output)
Another application of Euler's theorem occurs in supply theory. If
the quantity of a product supplied (q) depends on the price of the
product p, and the price of a competing product r, then assuming the
function is homogeneous of degree k, Euler's theorem states that
p iJq + r iJq = kq
ap dr
Dividing by q gives
.1!.... iJq + _!_ iJq = k
q iJp q iJr
that is, the sum of the own price and cross elasticities of supply equals k.
For example, if the supply function is
q = 6p 2 - 2r 2 - rp
iJq = l2p - r, iJq = - 4r - p
dp ar
Own-price elasticity is
p_ iJq = p_ (l2p - r)
q dp q
Cross-elasticity is
r iJq r
- - = - (- 4r - p)
q iJr q
Sum is
12p2 - rp (- 4r 2 ) - rp l2p 2 - 4r 2 - 2rp 2q
---=-----=- + = =- =2
q q q q
which is the degree of homogeneity.
For the Cobb-Douglas function
q = ALaKfJ
qL = aALa-lKfJ = aq/L
and qK = {JALa](P-l = {Jq/K

Hence, by Euler's theorem,
LqL + KqK = L(aq/L) + K({Jq!K) = (a+ {J)q
or L qL + K qK =
(a + {3) (a + {3) q
The total product, q, is distributed between labour and capital in the
proportions
qL!(a + {3) : qJ(a + {3) or qL : qK
and so the distribution is in proportion to the marginal products of
labour and capital.
7.12 Exercises
1. Show that the following functions are homogeneous and state the
degree of homogeneity:
(a) f(x, y) = x3 + 3y 3 - x 2y
xy + y2
(b) f(x, y) = 4x3
(c) f(x, y, z) = 2xyz + x 2y - y 2z

2. Verify Euler's theorem for the function in l(a), l(b) above.
3. The demand for butter depends upon the price of butter (p) and
the price of a substitute (r) according to the function
q = 10r 2 - 2p 2 + 4rp
Show that the sum of the own-price and cross-elasticities is equal to
the degree of homogeneity.
4. Show that the production function q = l.lU 6 K 0 ·2 has decreasing
returns to scale and that q = 1.1U·6 K 0 · 5 has increasing returns to
scale.
5. The demand for a product is q = p 1 Yl p; where p 1 is the price of the
product, p 2 is the price of a competing product and Y is income.
Determine the three partial elasticities, the degree of homogeneity
of the function and verify Euler's theorem.
6. A production function is said to be homothetic if it is homogeneous
and the slope of an isoquant ( -qL/qK) is the same at the point (U,
K 0 ) and (kU, kK 0 ) where k > 0. Show that the Cobb-Douglas
function q = AUKf3 is homothetic.
7.13 Maxima and minima

In the three-dimensional model the function z has a maximum
value at the 'top of a hill' and a minimum value at the 'bottom of a
valley'. These are the points which are important in economics. A third
case exists when z is at a maximum, say, with respect to variation along
the x-axis, but is at a minimum with respect to variation along they-axis.
This is called a saddle point since it corresponds to the shape of a saddle:
viewed from the side of a horse the saddle is U-shaped, whereas from
the front of a horse the saddle is arch-shaped.
To determine the position of these three points it is first necessary
to obtain the partial derivative of the function with respect to each
variable in turn. A maximum, minimum or saddle point can only exist at
a point where each of these partial derivatives is equal to zero. This is a
necessary condition, e.g. if z = f(x, y), then for a maximum, minimum
or saddle point
iJz = 0 and az = 0
ax ay
The solution of these equations simultaneously determines the values of
x and y, at which one of these states exists.
To decide whether a particular function has a maximum, a mini-
mum or a saddle point, the second-order partial derivatives must then
be considered in the light of the following rules:
1. The function has a maximum or a minimum point if ZxxZyy is
greater than (zxyY· If Zxx and Zyy are both negative there is a
maximum and if they are both positive there is a minimum.
2. If ZxxZyy is less than (ZxyY then there is a saddle point of Zxx and
zyy are both equal to zero.
These rules can be summarised as follows:
Maximum ZxxZyy > (zxyY, Zxx < 0, Zyy <0
Minimum ZxxZyy > (zxyY, Zxx > 0, Zyy >0
Saddle point ZxxZyy < (Zxy)Z, Zxx = Zyy = 0
3. If ZxxZyy equals (zxyY then the point may be a maximum, a
minimum, a saddle point or some other type of turning point and
requires further investigation.
These conditions can also be written in determinant form (see
section 1.13) since
which is positive for a maximum or minimum and negative for a saddle

point. This type of determinant, made up of second-order partial deriva-
tives, is known as a Hessian determinant or simply as a Hessian, and is
particularly useful when more general functions are considered, as in
section 7.15 below.
It can be seen from the above that there are a number of similarities
with the conditions for stationary values in the two variable model.
Thus, a stationary value can only exist at a point where all the first-order
derivatives are equal to zero. For a saddle point the second-order
derivatives must also be equal to zero.
If both Zxx and Zyy are negative (cf d 2yldx 2 < 0) there is a maximum
and if both Zxx and Zyy are positive (cf d 2yldx2 > 0) there is a minimum.
The procedure for finding maximum and minimum values is, there-
fore, very simple as the following example shows:
Suppose that z = x2+ 2xy + 2y2 - 5x - 4y
then Zx= 2x + 2y- 5
and Zy= 2x + 4y- 4
Equating these partial derivatives to zero produces two simul-

taneous equations in the variables x andy:
2x + 2y = 5
2x + 4y = 4
Subtracting the equations gives
2y = -1 or y = -2I and hence x=3
There is, therefore, a stationary point when x = 3, y = -i.
To determine whether this is a maximum, a minimum or a saddle
point, the second-order partial derivatives must be considered:
Zxx = 2, Zyy = 4, Zxy = 2 = Zyx
and
ZxxZyy > (Zxy) 2
The point is a maximum or a minimum and as Zxx > 0, Zyy > 0 it is a
minimum.
The function has a minimum value at the point
X= 3, y = -2
I
and this value is given by

z = 32 + (2)(3)( -i) + 2( -i)2 - 5(3) - 4( -!)
=9 - 3 + i - 15 + 2 = - 6 i
As an economic application, consider the problem of a producer
who uses L units of labour and K units of capital and has the production
function
q = 12L + 5K - 0.2K 2 - 0.5V
and the cost of labour is 8 per unit and that of capital is 4 per unit.
Assuming there is pure competition and the price of the product is 4
then the profit is
n = 4q - 8L - 4K
= 48L + 20K - 0.8K 2 - 2V - 8L - 4K
To maximise profit we find where lrx = 0 = nL and check the second-
order conditions. Now,
lrx = 20 - 1.6K - 4
JrL = 48 - 4L - 8
and setting these to zero gives K = 10 and L = 10. Here,
and so profits are maximised when K = 10 and L = 10, and the

maximum is n = 280, q = 100.
7.14 Exercises
1. Find Zx, Zy, Zxx' Zxy' Zyy in the following cases:
(a) z = x2 - xy + y2 (b) z = (x + y) 3
2. Find the maximum, minimum or saddle-point values (if any) of the

following functions:
(a) z = i + 2xy + X3 (b) z = 3x + 2y - 3x2 - 4y 2
3. Total cost, z, depends on the output x of product A and the output
y of product B.
What is the minimum value of total cost for the following func-
tions:
(a) z = 50 + x 2 + y 2 - 4x - y
(b) z = 200 + x 3 - 3y + y 2 - 2x
(c) z = 250 + x3 + y 2 - 27x - Sy
4. For the production function
q = 36K + 120L - ~ - 4U - 10 + 2KL
determine the maximum value of q.
5. A monopolist produces two goods with demand functions
Pi= 40- qi and
and the joint cost function is
TC = qi + 3qi q2 + q;
Determine the maximum value of profits.
6. A bus company runs services during the peak period, for which the
demand function is pi = 105 - 25qi, and also in the off-peak period
for which the demand function is p 2 = 50 - 2q 2 • The cost function
is TC = 10 + Sqi + 2q 2 • Determine (a) the maximum value of
profits if there is price discrimination, (b) the maximum value of
profits if there is no price discrimination (and so pi = p 2 = p, say).
7.15 Maxima and minima for functions of more than two

variables
An extension of this procedure is needed when z is a function of
more than two variables because here there are more partial derivatives
to deal with.
The first-order conditions for both maxima and minima require all
the first partial derivatives of the function to be equal to zero. To
distinguish between the maximum and minimum positions we must
consider the second-order partial derivatives. This becomes a little more
complicated as the number of variables increases and to help make the
conditions clear they are stated in determinant notation as explained in
section 1.13.
Consider the case where z = f(xi, X 2 , x3 , • • • , xn)·
The first-order conditions for a maximum or a minimum are that
where
These equations are solved simultaneously to find positions at which the
conditions are satisfied. The second-order partial derivatives Zw Z 12 , Z 13
etc. are then calculated and the following determinants formed:
Zu Z12 Z13
Z12 Zzz Zz3

Z13 ZzJ Z33
The numerical values of these determinants are considered as fol-

lows
1. If all the determinants are positive there is a minimum.
2. If determinants of even order (i.e. 2 x 2, 4 x 4 etc.) are posi-
tive but those of odd order (i.e. 1 x 1, 3 x 3 etc.) are negative
there is a maximum.
3. If any determinants of even order are negative there is a saddle
point.
As mentioned in section 7.13, these determinants, made up of
second-order partial derivatives, are referred to as Hessian determi-
nants or Hessians.
To illustrate the use of these suppose that in a production function
output, z, depends on the level of inputs Xu X 2 , and x 3 so that
z = f(x 1 , X 2 , X 3) = 100x 1 - 2xi + 50x 2 - x; + 200x 3 - 2x;
Output is to be maximised. The first-order derivatives are
Z1 = 100- 4x 1
Z2 = 50- 2x2
Z3 = 200- 4x3
If the function is to have a maximum or a minimum then this must
occur when
that is
The second-order partial derivatives are then calculated:
Z 11 = -4 z 12 = 0 z 13 =0
Zzz = -2 ZzJ = 0 Z33 = -4
The Hessian determinants are

-4 0 -4 0 0
-4 0 -2 0
0 -2 0 0 -4
that is -4, + 8 and -32
These values, alternately negative, positive, negative, satisfy condi-
tion 2, and the function therefore has a maximum when the variables
have the values x 1 = 25, X 2 = 25, x 3 =50 and the value of this maximum
is given by
z = 100(25)- 2(25) 2 + 50(25)- (25) 2 + 200(50)- 2(50)2 = 6875
As a further example, suppose the total cost (TC) function for a
firm producing three goods depends on the outputs, x, y and z of these
goods so that
TC = 1000 + 3x 2 + 2y 2 + 2z 2 - 2xy - 40z - 20x
and assume that total cost is to be minimised. The first-order derivatives
are
TCx = 6x - 2y - 20
TCy = 4y- 2x
TCz = 4z- 40
and setting these to zero (the necessary conditions for a minimum) gives
x = 4, y = 2 and z = 10.
The second-order derivatives are
TCxx = 6, TCxy = -2, TCxz = 0, TCyy = 4, TCyz = 0, TCzz = 4
and letting x, y and z correspond to x 0 x2 and X3 in the previous
formulae, the Hessians are
Z11 Z 12 1 = I TCxx TCxy I = I 6 ~ = 24- 4 = 20

I Zu z22 TCxy TCyy -2 4
z11 Z 12 Z 13 6 -2 0
Z12 Zzz Zz3 -2 4 0 = 4(24 -4 ) = 80
Zu Zz3 Z33 0 0 4
Since each of these is positive we have a minimum when x = 4, y = 2

and z = 10 and TC = 760.
7.16 Constrained maxima and minima

The situation frequently arises when a function has a definite max-
imum which is not attainable because of some restriction on the value
which the variables can assume. This can happen in the case of a
production function where it may be impossible to reach the level of
output which would yield the maximum revenue due to an insufficient
supply of the factors of production. In these cases it is important to
determine the maximum which can actually be achieved. This can be
done using the method of Lagrangian multipliers.
Let us consider the function
z = 100x 1 - 2xi + 60x 2 - x;
To find the maximum value of this without restriction the first-order
partial derivatives are obtained and equated to zero.
Z1 = 100- 4x 1 =0
Z2 = 60 - 2x 2 =0
The second-order partial derivatives are

z 11 = -4
and
The function, therefore, has a maximum value when x 1 = 25,
X2 = 30 and z = 100(25) - 2(25Y + 60(30) - (30Y = 2150.
If there is a limited amount of capital available to buy resources
then a constraint is applied to the system. Let this be represented by
+ X 2 = 40
X1
One method of maximising z subject to this constraint is to substi-

tute directly for x 1 , using X 1 = 40 - X 2 , into z so that
Z = 100(40- X 2 ) - 2(40- X 2Y + 60X 2 - ~
= 4000- 100x 2 - 3200 + 160x2 - ~ + 60x 2 - ~
= 800 + 120x 2 - ~
This function can be maximised as in section 5. 7 by
.!!!_ = 120 - 6x and .!!!_= 0 when X2 = 20

dx 2
2
dx2
d2z
-6
~
therefore x2 = 20 gives a maximum and
z = 800 + 120(20) - 3(20)2 = 2000
is the maximum value of z.
In this particular case the method of substituting for X 1 from the
constraint into the function for z has worked out well. However, in
general such a substitution can be awkward to carry out and also does
not produce as much information as an alternative method, using Lag-
rangian multipliers, which is now considered.
This involves forming a new expression
z' = z +AU
where u = (40 - x 1 - x 2 ) and). is a variable known as the Lagrangian
multiplier. It follows, that by optimising z' we are effectively optimising
z and at the same time satisfying the constraint condition. The expres-
sion for z' has three unknowns, xH x 2 and A, and the procedure is to find
the stationary values by setting the first-order partial derivatives to zero.
Here
z' = 100x 1 - 2xf + 60x 2 - ~ + A(40- X 1 - X2 )
z; = 100 - 4x 1 - A = 0
~ = 60 - 2x 2 - A = 0
~ = 40 - X1 - X2 = 0
The points at which the function has a maximum or minimum value
are obtained by solving this set of simulataneous equations to give x 1 = 20,
x 2 = 20, A = 20. To check that this is a maximum value it is necessary
to know the values of the second-order partial derivatives of the original
function and the first-order partial derivatives of the constraint equa-
tion. Again we denote partial derivatives with respect to x 1 by the
subscript 1, and partial derivatives with respect to x2 by the subscript 2.
There is a maximum value if
and the minimum value if
provided u is linear in both x 1 and X 2 •

These conditions can be written in the form of a special Hessian
determinant which has U 1 and u2 on the borders:
zll z12 ul > 0 for a maximum
Z12 Z22 U2 < 0 for a minimum
u1 u2 0 = 0 for inconclusive
In the example
= -4
Z 11
u = 1
1 u2 =1
Hence 2u 1U 2Z 12 =0
and
and so we have a maximum value when X1 = 20 and X2 20. This
maximum value is
z = 100(20) - 2(20) 2 + 60(20) - (20)2 = 2000
The maximum output which could be obtained without restriction
on X 1 and X 2 was 2,150. Thus, in general, the consequence of placing a
constraint upon any of the variables is to reduce the maximum value
attainable. An exception would be the case where the two maxima were
equal but under these circumstances the constraint is not really effec-
tive.
The economic interpretation to be placed on the value of A. (20
here) is of interest. It is the marginal product of resources, i.e. it is a
measure of the increase in z which would be produced by an increase in
the level of the constraint by one unit. For example, if x 1 + x 2 can be
increased to 41 the function to be maximised is changed to
z" = 100x1 - ~ + 60x2 - x; + ..\(41- X1- X2)
and in this case it would be found that the maximum value of z" is
greater than z' by approximately 20 units (see Section 7.18, Exercise 1).
As a further example suppose that an individual has a utility func-
tion,
u(x 1 , x 2 ) = log(x 1x 2 )
which depends on the quantities consumed of the two goods x 1 and X 2 • In
the absence of any constraint, utility is maximised when X 1 and x2 are
infinite. If the total income is M and the prices are p 1 and p 2 then,
ignoring the possibility of debt, expenditure is constrained to be equal to
income. That is, the budget constraint is
M = P1X1 + PzXz
The problem is to determine the values of x 1 and x2 which maximise
utility, subject to the budget constraint. That is, maximise
z' = log(x 1x 2 ) + A(M- p 1X 1 - p 2 X 2 )
Notice that we have written the constraint as M - p 1x 1 - p 2 x2 when it
could have been written as p 1x 1 + p 2 x 2 - M. This has no effect on the
optimisation procedure but changes the sign on A. Keeping to our
previous notation, the partial derivatives are
z; = (1/xl) + A(-pl)
~ = (1/x 2) + A(-p 2)
and setting to zero and solving these gives X1 = M12p 1 , x2 = Ml2p 2 ,

A= 2/M
Letting u = M - p 1X 1 - p 2 X 2 , U 1 = -p 1, U 2 = -p 2 and the second-
order derivatives are, with z =log (x 1x2 ), Z 11 = -llxi, Z 22 = -1/x;, Z 12 = 0.
The bordered Hessian determinant is
zll Z12 U1 -llx7 0 ~
Z12 Zzz Uz 0 -llx; -pz
U1 U2 0 -pi -p2 0
which is positive and so we have a maximum, given by
u(X 1 , X2 ) = log(M 14p p
2
1 2)
We also note that the budget constraint is satisfied at the maximum.
7.17 A linear expenditure system

The linear expenditure system was proposed by R. Stone and arises
from maximising a special form of utility function subject to a budget
constraint. Consider a simple model with two goods and let x and y be
the quantities of each purchased by a consumer. The utility derived

from the goods depends on x and y and we will take the particular form
of the utility function to be
U(x, y) = (x- a)a (y- b)fl
where it is assumed that x > a > 0 andy > b > 0. Notice that this utility
function is similar to the Cobb-Douglas production function discussed
above. Here, a and bare the minimum quantities that will be consumed
and actual consumption will be greater than the minimum. We assume
that all the income (M) is spent. Below we will show that a and {3 are the
marginal budget shares which indicate by how much expenditure on the
product will increase if income increases by one unit. Since changes in
expenditure equal changes in income, a+ {3 = 1. Also, a> 0 and {3 > 0.
Taking logarithms to the base e,
log U = a log(x- a) + {3 log(y- b)
and the budget constraint is
M =PIX+ P2Y
To maximise utility subject to the budget constraint we maximise
F =a log(x- a) + {3 log(y- b) + J..(M- pix- p 2y)
where ). is a Lagrangian multiplier. The partial derivatives are
iJF a
- = - - - J..pi
iJx x- a
iJF = _{3_- J..p

ay y- b 2
aF
iii = M -PIX - P2Y
The first-order condition for a maximum is that these are all zero.
Solving for ). gives
a {3
PI(x- a)= P2(y- b)
or a P2(y- b) {3pi(x- a)
and so PIX = {ap2(y- b) + a{3pi}/{3
Substituting in 0 = M- PIX- P2Y
gives 0 = fJM- ap 2(y- b)- a{Jpi- fJP2Y

and so P2Y = fJM - a{Jpi + bap2
using a + f3 = 1. Finally, replacing a by 1 - f3 gives
P2Y = bp2 + fJ{M- api- bp2}
Now p 2 y is the total expenditure on the second good and since b is the
minimum quantity of y that is consumed, bp2 is called the subsistence
expenditure on the second good. The final term is the supernumerary
expenditure on the second good, being {3, the marginal budget share,
times the income remaining after the subsistence expenditure. Notice
that the expenditure on the second good is a linear function of p 1 , p 2 and
M, which explains why it is known as a linear expenditure system. The
corresponding result for the first good is
P1X = api + a{M- api- bp2}
and a is the marginal budget share.
This linear expenditure function can be transformed into a demand
function by dividing by p 1 to give
x =a+ a{M- api- bp2}lp1
The own-price elasticity of demand is from
ax
-;--- = a{(- ap 1)- [M- ap 1- bp2]}1pi
up!
= - a[ M - bp2]1pi
E = pJJx =- a[M- bp2]
and so,
xapi pix
This is negative since a> 0, M > bp 2 and p 1x > 0. The (absolute)
elasticity is less than 1 if
and since, from above,

p 1X = api + a{M- ap1- bp2}
the condition becomes
0 < ap 1 + a(- ap 1) = ap1(1 -a),
which is satisfied if a < 1. Therefore the elasticity is less than 1 if
0 < a < 1, and similarly, for the second good, 0 < f3 < 1.
7.18 Exercises
1. Find the maximum value of the function
z = 100x 1 - ~ + 60x2 ~
subject to the constraint X 1 + X 2 = 41.

2. Find the maximum or minimum value of
z = 3x2 + y 2 - 12x - 14y + 150
Find the restricted maximum or minimum value of z when there is
a constraint requiring 2x + y = 9.
3. Find the maximum or minimum value of the following functions
subject to the given constraints:
(a) z = x 2
(b) z = 18 -
(c) z
r-
+ y 2 with 5x + 4y
i
= 3xY with x + 3y = 18
with 2x +
= 40
3y = 12
4. Let the production function for a firm be given by

x = 20/ + 40c - 2F - ~
and the cost of labour (l) and of capital (c) be 4 and 5 units
respectively.
Find the maximum value of x if the total cost is equal to 28
units.
5. The utility function of a consumer is U = 6xy where x andy are the
quantities consumed of two goods. The price of x is 5 per unit and.
that of y is 10 per unit. If total expenditure is limited to 100, what is
the maximum value of U?

1. Determine Zx, Zy, Zxx• Zyy• and Zxy for
(a) z = 2xy- X 2 - y2
(b) z = ye-x + xe-Y
(c) z = log (2x + 3y)
(d) z = x!(x2 - y2)
(e) z = x sin y.
2. The demand for a product is given by
qd = 500- 3p + lOpe- p 2 + 0.2Y

where p is the price of the product, pc is the price of a competing
product and Y is income. Determine the own-price, competing
price and income partial elasticities if p = 10, pc = 12 and
Y = 100. Would revenue increase if p was reduced to 8?
3. For the production function Q = 3K 2 + 4U- 15KL determine
the marginal rate of substitution of capital (K) for labour (L) and
the elasticity of substitution.
4. If z = xy - x 2 - y 2 and x = e 1, y = log t then determine dzldt.
5. Determine dyldx if (a) y 4 + x 4 - 6xy = 3x2y 3 (b) xeY = yex.
6. Check whether the following are homogenous functions and, if
they are, verify Euler's theorem.
(a) z = x 2 y- y2x
(b) z = xe-Y + ye-x
(c) z = (x 2 - y2 )/ 6xy
7. Determine the maximum, minimum or saddle-point values (if
any) of the following functions:
(a) z = x2 + y 2 - 6xy
(b) z = 200 - x 3 - 2y 3
(c) z = x2 - 4x- xy + 2y2
8. The production function for a product is
q = 6K + 8L - 0.4K 2 - 0.2U
and labour (L) costs 3 per unit while capital (K) costs 2 per unit.
The product sells for 15 per unit. What level of output will
maximise profits?
9. Find the maximum value of z = x 2y 2 subject to the constraint that
x + y = 16. If the constraint changes to x + y = 17 what is the
change in the maximum value of z?
10. The utility of a consumer is given by U = x2 + y2 • If the price of
x is 3, the price of y is 5 and the income is 100, what is the
maximum value of U? What is the effect on the maximum of
increasing income to 101?
Chapter 8
Linear Programming
8.1 Introduction
We saw in Chapter 7 how it is possible to find the maximum of a
given function when there are constraints on the values which some or
all of the variables can assume. To do this we made use of the differen-
tial calculus and the method of Lagrangian multipliers. This is, however,
only applicable when the constraint is in the form of an equality, that is,
one which must be exactly met.
There are many situations where the restriction is not so precise
and in particular where the constraint is expressed in the form of an
inequality. This might be the case, for example, where a company is
concerned with maximising the profitability of its operations, subject to
the limitations imposed upon it by the availability of production ca-
pacity, whether in the form of machine time, skilled labour or financial
resources. This type of optimisation problem cannot generally be handled
with the help of the differential calculus, but there are mathematical
tools available which can provide a solution in a simple way. To illus-
trate this we take the case where all the interrelationships between the
variables and all the constraints can be expressed in linear form and
linear programming can be used. The following example explains the
method, and illustrates the principles by means of a graphical approach.
This might be used in simple situations, but these are unlikely to occur
in practice and so a mathematical method is explained which has more
general applicability and can be computerised easily.
8.2 A product-mix problem

A small company has two machines U and V which are both
required to produce two products A and B. The problem is one of
deciding how much of each product should be manufactured in order to
provide the maximum contribution to the profitability of the company.
To obtain a solution it is necessary to have information about the
297
TABLE 8.1
No.
of hours Total
required no.
per unit of hours
of product available
A B
Machine U 1 2 30
Machine V 2 1 30
Contribution
per unit of product 2 3
capacity of each machine, the amount of time which is required on each

machine by each product, and the contribution which each product
makes to profitability. Such a situation might lead to the information
provided in Table 8.1.
The contribution per unit of product is equal to the selling price less
the variable cost of production, and in order to use the linear-
programming approach for solving this product-mix problem it is necess-
ary to assume that this remains constant over the levels of output which
are under consideration.
If we let x 1 and x 2 be the number of units of products A and B
respectively which the company should manufacture, the relevant in-
formation can be neatly summarised in algebraic form as follows:
Maximise 2x 1 + 3x2 = z (1)
subject to the constraints
xt + 2xz :S::: 30 (2)
2x 1 + X 2 :S::: 30 (3)
X1 ~ 0 (4)
X2 ~ 0 (5)
The first relationship is an equation which represents the total
contribution which the company can expect from the two products, i.e.
it is the sum of the contribution per unit multiplied by the number of
units which will be manufactured of each product. This is called the
Linear Programming 299
objective function and the purpose of the exercise is to obtain the values
of X 1 and x 2 for which this is a maximum.
The second relationship is an inequality which states that the
number of hours used making x 1 units of product A on machine U plus
the number of hours used making x 2 units of product B on machine U
must not exceed the maximum number of machine hours available. The
third relationship is again an inequality and describes the constraint on
machine V in a similar manner. Inequalities (4) and (5) merely state that
the output of either product cannot be negative. These are obvious
practical constraints on the system.
Any values of the variable X 1 and x 2 which satisfy these inequalities
are said to constitute a feasible solution to the problem. What is required
is a method for determining from all the possible feasible solutions that
one which yields the maximum value of z as specified in the objective
function. This gives the values of the variables which produce the
maximum profit and is known as the optimal solution.
8.3 A graphical approach

The situation being considered can be analysed using a graphical
method because there are just two variables. The two axes represent the
variables X 1 and x 2 • The lines corresponding to the equations X 1 + 2x2
= 30, 2x1 + x 2 = 30 are then drawn on the graph as shown in Fig. 8.1.
These equations correspond to the upper limit of machine capacity
and therefore any combinations of A and B which it is possible to
manufacture must lie on the side of the lines closest to the origin. To
satisfy both constraints they must, in fact, lie within the area which is
doubly cross-hatched in the diagram. This is known as the feasible
region. We now superimpose on this graph the profit line 2x 1 + 3x 2 = z
for profits of 24 and 60 units, that is 2x 1 + 3x2 = 24 and 60 respectively.
These lines are parallel to each other because the slope of the line
z = 2x 1 + 3x2 is independent of the value of z. They are shown as broken
lines in Fig. 8.2.
The further away from the origin this line is taken the higher the
value of z, but the upper limit which is allowable is determined by the
constraints on the system. It is immediately apparent that no point on
the line z = 60 lies within the feasible region, and therefore this value of
z is unattainable. In fact, the highest value z which satisfies the con-
straints is reached when the profit line passes through one of the ex-
treme points of the feasible region. This will always be so if the con-
straints are linear except in the special case where the profit function is
x,
30
30 x,
Fig. 8.1
x,
30
10
' z = 24
......
......
......
5 ......
......
......
...........
......
Fig. 8.2
parallel to one of the constraints, and in this case, the problem has a
multitude of solutions corresponding to the points on the constraint line.
To obtain the solution graphically we move the line with slope - ~
outwards until it reaches the limit of the feasible region, and this point
determines the values of x 1 and X 2 corresponding to the outputs of A and
B. In this case it is at the point where the two constraint equations
intersect, and hence the optimum combination of products to manufac-
ture can be easily determined by the usual method of solution:
2x 1 + x 2 = 30 (6)
X1 + 2x 2 = 30 (7)
Making use of determinants, we have
30 1
I 30 2 60-30
- - - = 10
XI= I 2 1 4-1
1 2
121 30 I
30 60-30
X = - - - = 10
4- 1
2
I i ;I
The result is that 10 units should be manufactured of both products
A and B. It follows that the contribution will equal
10 X 2 + 10 X 3 = 50
It is easy to check that the constraints on machine capacity are not
violated by substituting in the original equations.
XI+ 2x2 = 30
2x 1 + X2 = 30
These equations are exactly satisfied and therefore both machines
are fully utilised, i.e. there is no spare or slack capacity.
8.4 The simplex method

The simplex method employs an iterative procedure which pro-
gresses in a series of steps ultimately leading to the optimal solutions.
The rules of operation are described below, using the same example for
illustrative purposes.
Step 1. Convert the inequalities (2) and (3) into equations by the
insertion of new variables x 3 and x 4 •
X1 + 2x2 + x 3 = 30 (8)
2xl + X2 + X 4 = 30 (9)
The only qualification which must be added here is that the new
variables must only take on positive values:
This is necessary because they represent the difference between the

time which is utilised on each machine and the total time available on
the machines. They are therefore known as slack variables and cannot,
by definition, be negative.
Step 2. Obtain a first feasible solution to the system. The insertion
of the slack variables proves useful here because an immediately ob-
vious solution is given by
A unique solution to two equations in four unknowns can only be

obtained if two of the unknowns are given defined values. For con-
venience they are assigned the value zero in this exercise.
This solution corresponds to zero output and slack or spare capacity
on both machines and also produces zero profit. It corresponds to the
origin in the graphical method and hence it is on the edge of the feasible
region. The reason it is chosen is that it is an easily found solution. The
non-zero variables x 3 and X 4 are taken as a starting point for the iterative
procedure and are known as the basic elements and form a basis in a
tableau which is drawn up in Table 8.2.
The elements in the tableau are obtained from the initial data for
the system. They correspond to the coefficients of equations (8) and (9)
and of the objective function equated to zero.
TABLE 8.2
xi x2 x3 x4 p
x3 1 2 1 0 30
x4 2 1 0 1 30
z -2 -3 0 0 0
Equation (8) X1 + 2x2 + x 3 + Ox4 = 30

Equation (9) 2x 1 + X2 + Ox3 + X4 = 30
Objective function z - 2x 1 - 3x2 + Ox3 + Ox4 = 0
In the tableau the basic elements X 3 and x 4 are written to the left of
the equation in which they appear.
The first feasible solution is obtained by letting x 1 = 0 and x 2 = 0
with the result that
(from (8))
(from (9))
and the profit, z1 = 0 (from the objective function)
It is obvious that this solution can be improved upon and that any
improvements can only come from the manufacture of either A or B.
This could be shown in the tableau by either x 1 or x 2 appearing in the
basis with the result that either x 3 or x 4 must become zero and leave the
basis.
The method consists of
(a) determining the variable to enter the basis, known as the
entering variable, and,
(b) determining the variable to leave the basis, known as the
departing variable, and
(c) calculating the profit which results from the interchange.
One such change is made at each stage in the iterative procedure
and if the rules are followed carefully the method leads progressively to
the optimal solution and indicates clearly when this optimum is reached.
Step 3. Select the entering variable by considering the elements in
the lowest row of the tableau. These elements correspond to the
marginal profit which would be obtained by the manufacture of any
products other than those at present in the basis. In the first tableau the
elements are - 2 and - 3, which correspond to the contribution per unit
from products A and B respectively. It is possible to introduce only one
new variable into the basis at a time and it seems logical to select that
variable which offers the largest marginal profit, in this case x 2 , or
product B. Thus we have
Rule 1. The entering variable is chosen as that variable which has
the largest negative element in the bottom row of the tableau.
Step 4. Select the departing variable. Before doing this, it is im-

portant to remember that the variable leaving the basis becomes zero
and the two non-basic zero variables determine the values of the basic
variables. For example,
if X3 =0 and X 1 =0
then 2x 2 = 30 (from (8))
and X2 + X4 = 30 (from (9))
That is X2 = 15 X4 = 15
But if and
then 2x 2 + x 3 = 30 (from (8))
X2 = 30 (from (9))
That is X2 =30 x 3 =-30
This second result is not feasible because the values of the variables
in a linear programming problem must all be positve and x 3 = - 30
violates this constraint.
It follows therefore that the departing variable can only be x 3 •
In the case of large problems, it would be a laborious task to check
each variable independently to see whether the resulting interchange
violates the non-negativity constraints. It is not, in fact, necessary
because the departing variable can be found by considering the ratio
p;laii for each basic variable. Pi is the element in the last column of the
row in which the basic variable is written and aii is the element in the
column of the entering variable and the row of the basic variable.
The calculation for the present example is as follows:
Pi
xi Xz x3 x4 p
aii
30
~
1 1 0
D
x3
2
30
x4 2 0 1 0
1
z -2 -3 0 0 0
t
entering variable
In order to ensure that all the variables will have positive values at
each state of the procedure, it is necessary to specify
Rule 2. The departing variable is that one for which the ratio pJaii
has the smallest positive value.
The element at the intersection of the column containing the
entering variable and the row containing the departing variable is
known as the pivot and plays an important part in the following step.
In this example the entering variable is x 2 , the departing variable is
x 3 and the pivot element is 2.
xi x2 x3 x4 p
x3 1
0 1 0 30 ~ departing variable
x4 2 1 0 1 30
z -2 -3 0 0 0
i
entering variable
We now require to form the new set of equations from which we

can determine the value of x 2 when x 1 and x 3 have the value zero. This
will tell us how many units of product A to manufacture and how many
units of spare capacity there will be. This information, along with the
profit which will be produced by this course of action, can be obtained in
the following way.
Step 5. Rewrite the tableau with the entering variable replacing the
departing variable in the basis and form the new matrix of coefficients as
follows:
(a) Divide all the elements in the pivot row by the pivot element.
(b) Add or subtract such multiples of the element in the pivot row
from corresponding elements of the other rows to reduce all
other elements in the pivot column to zero.
In this example all the elements in the pivot row are divided by two
to form a new row. Multiples of this new row are added or subtracted
from the other rows as follows:
(1) The elements of the new row are subtracted from the corre-
sponding elements of row two of the old tableau to form row

two of the new tableau.
(2) Three times the elements of the new row are added to the
corresponding elements of row three of the old tableau to form
row three of the new tableau.
XI Xz x3 X4 p
I I
Xz 2 1 2 0 15
x4 (2- !) (1 - 1) (0- !) (1- 0) (30- 15)
z (-2 + ~ (-3 + 3) (0 + ~ (0 + 0) (0 + 45)
This can be rewritten as
XI Xz x3 x4 p
I
Xz 2 1 I
2 0 15
3 I
x4 2 0 -2 1 15
I 3
z -2 0 2 0 45
This tableau corresponds to the solution

X 1 = 0, X2 = 15, x 3 = 0, X4 = 15, z = 45
that is, the company could manufacture 15 units of B and as each unit of
B contributes 3 units to profits, the total contribution would amount to
45 units.
It can also be noted that this product mix uses all the available
capacity of machine U (slack variable x 3 = 0) but leaves spare capacity
on machine V (slack variable X 4 = 15). Check: 15 units of product B
require 15 x 2 = 30 hours on machine U but only 15 x 1 = 15 hours on
machine V.
The second tableau gives a solution which is an improvement on
that given by the first tableau. But is it the optimal solution? The answer
is no, and this is indicated by the presence in the bottom row of the
tableau of a negative element. These elements correspond to the mar-
ginal profit which could be obtained by the introduction into the basis of
a non-basic element. Therefore, in this example, profit is being foregone
by not having x 1 in the basis, i.e. by not manufacturing product A. This
t
profit is equal to unit for each unit of x 1 introduced. It should be noted
here that this marginal profit is no longer equal to the direct contribu-
tion per unit from A because the introduction of this product necessarily
means that we must reduce the quantity of product B which is produced
in order not to violate the capacity constraint on machine U. (There was
no spare capacity on this machine with the first product mix.) The
negative figures in the bottom row of the tableau therefore correspond
to the marginal profit which is obtained after the product mix has been
changed by introducing product A and adjusting the level of manufac-
ture of the other product (or products) to satisfy the constraints on the
system.
The variable with the largest negative element in the bottom row of
the tableau is selected for the entering variable (in this case x 1 which has
the only negative element). The departing variable is found by consider-
ing the ratio p;la;i and the interchange carried out as before.
Steps 3 and 4 repeated
P;
Xt Xz x3 x4 p
a;i
I
Xz 2 1 I
2 0 15 15/i
x4 OJ 0 _I
2 1 15 ~ - departing variable
_I 3
z 2 0 2 0 45
i
entering variable
Step 5 repeated
X! Xz x3 x4 p
Xz d- !) (1- 0) <! + ~ (0- b (15 - 5)
2
X! 1 0 I
3 3 10
z (-! + !) (0 + 0) ~ i) (0 + 1) (45 + 5)
which is rewritten as
XI x2 x3 x4 p
2 I
x2 0 1 3 3 10
I ~
XI 1 0 3 3 10
4 I
z 0 0 3 3 50
This tableau corresponds to the solution

XI= 10, X2 = 10, x 3 = 0, X4 = 0, z =50
that is, the company would manufacture 10 units of A, 10 units of B, and
the total contribution from this combination of output would be 50. The
last iteration has again improved the profit of the company and at the
same time has arrived at a combination which utilises all the available
capacity of both the scarce resources (slack variables x 3 and x 4 are both
equal to zero). It can also be seen that none of the elements in the
bottom row of the tableau are negative and this indicates that no
increase in profit can be obtained by a change in the product mix. The
final tableau therefore provides the optimal solution to the problem.
8.5 A summary of the simplex method

To explain and carry out the iterative procedure at one and the
same time may give the impression that it is lengthy and difficult. To
show that this is not so, the necessary stages are repeated here with all
the detail that would be required to carry out the calculations in a
practical case. The problem itself has been altered slightly to one in
which the company has more capacity available on machine U than on
machine V. The data for the problem are as shown in Table 8.3.
Mathematically this can be stated as follows.
Maximise
subject to the constraints
XI+ 2x2 ~ 40
2xl + X2 ~ 30
The slack variables, x 3 and x 4 are inserted in the inequalities with the
qualification that x 3 :;;:: 0, x 4 :;;:: 0
TABLE 8.3
No.
of hours Total
required no.
per unit of hours
A B
Machine U 1 2 40
Machine V 2 1 30
Contribution
X1 + 2x 2 + X = 40
3
2xl + X 2 + X 4 = 30
The procedure is as shown in Tableaux 1-3. The optimal solution is
to make ~ units of A and ~ units of B; the total contribution from this
combination would be 63j. This solution assumes that it is possible to
manufacture fractional parts of a product and in many cases this is
realistic because it is the average output per week which is being
discussed, and this is likely to include fractional units. When such a
situation is not permissible it is possible to continue the procedure
further and use the method of integer programming which will produce
a solution in which the variables are restricted to whole units.
Tableau 1
P;
XI x2 x3 x4 p
a;i
x3 1 w 1 0 40 40/2 +- departing variable

x4 2 1 0 1 30 30/1
z -2 -3 0 0 0
r
entering variable
Tableau 2
Pi
XI Xz x3 x4 p
a;i
1 1
Xz 2 1 2 0 20 20/!
x4 []] 0
_1
2 1 10 ~ ~ departing variable
_1 3
z 2 0 2 0 60
i
entering variable
Tableau 3
XI Xz x3 x4 p
2 _1 50
Xz 0 1 3 3 3
_1 2 20
XI 1 0 3 3 3
4 1 190
z 0 0 3 3 3
8.6 Incremental values

In the problem under discussion we have assumed that the capacity
of the machines U and Vis fixed and cannot be increased. However, the
capacity might be increased by, say, buying new machines or introduc-
ing overtime working on the present machines. This raises the question
of whether it is worthwhile increasing the capacity of machine U or
machine V, or of both of them. There are two ways of deciding on this.
The first is by reworking the problem with new constraints while the
second uses information from the final simplex tableau.
Taking the first of these, since it helps in understanding what is
happening: in the original problem, if the total number of hours avail-
able on machine U, is increased from 30 to 31, (2) of section 8.2
becomes
and together with (3),

2x 1 + X2 = 30
the graphical method gives the new solution as x 1 = 29/3 and x 2 = 32/3.
The contribution to profit is z = 51.33 compared with the previous value
of 50. Increasing capacity on machine U by 1 hour will therefore
increase profits by 1.33. Similarly, if the total hours available on ma-
chine Vare increased from 30 to 31, with machine U being limited to 30
hours, the new solution is x 1 = 32/3 and x 2 = 29/3 giving z = 50.33, an
increase of 0.33 on the previous value. The value of one extra hour of
time on machine U is thus 1.33 while that of one extra hour on machine
Vis 0.33. These are called the incremental worth, opportunity cost, dual
price or shadow price of machine time, since they indicate how much it is
worth paying for the extra time. That is, if one hour of time on machine
U can be bought for less than 1.33 it is worth buying since using it will
increase profits by 1.33. Here it is clear that one hour of time on
machine U is more valuable than one hour of time on machine V.
Turning now to the second way of checking whether it is worth-
while increasing capacity on the machines, we can use some of the
information which is obtained as a by-product of the simplex method.
This appears in the final tableau from the simplex method, as shown as
in tableau 4.
Tableau 4
X1 x2 x3 x4 p
2 _I
x2 0 1 3 3 10
-1 2
X1 1 0 3 3 10
4 I
z 0 0 3 3 50
The bottom row contains zero elements in the columns correspond-

ing to the variables in the basis, that is, x 1 and x 2 , but positive elements
in the columns corresponding to the non-basic variables, that is, x 3 and
X 4 • These positive elements correspond to the incremental value of one
hour of machine time on U and V respectively. This means that one
extra hour of available machine time on U would increase the total
contribution by 1 to ~ and one extra machine hour on V would
mcrease
· · by 31 to 50 3·
It 1
In many cases, of course, extra capacity cannot be added in single

units, e.g. hours, but must be added in fairly large steps. This would
occur where the purchase of an extra machine was contemplated with a
resulting increase of capacity of, say, 30 hours. Occasionally the struc-
ture of the problem allows the increased contribution to be determined
directly from the incremental value. In the second example, an increase
in capacity of 10 hours in machine U resulted in an increased contribu-
tion of 13j units, and this is exactly equal to 10 X t i.e. the extra
contribution was directly proportional to the number of additional units
of capacity. This is not generally the case, and therefore it is better to
re-run the linear programme if the effect of changes in any of the
constraints is required.
An interesting point which arises from the solution of the linear
programme is, as economists might expect, that at the optimum the
marginal revenue equals the marginal cost for all those products which
the company should produce.
In the example, the incremental value of machine capacity of U and
V is respectively 1 and j , and this must therefore be the value to the
company of time on these machines at the margin. By multiplying these
values by the hours which a unit of A and B uses on the machines we will
obtain a value for the marginal cost of the two products.
Product A: marginal cost = 1 x 1+ 2 x j = ~ = 2
The contribution per unit of A, which is in fact its marginal
revenue, is also equal to 2.
Product B: marginal cost = 2 x 1+ 1 x ~ = ~ = 3
This is equal to the contribution per unit or marginal revenue from B. It
can be quite easily shown that the marginal revenue equals the marginal
cost for all products that appear at a positive level in the optimal
solution. Those products which do not appear in the final solution have
a marginal cost greater than their marginal revenue, which is a result
that would be expected on theoretical grounds.
8. 7 The general problem

The linear programming problem discussed above may be written
more generally as
Max
subject to
a 21 x, + a 22 X 2 :::::: b2
with ~ ~
This can be extended to the case of n variables:

Max
subject to aux 1 + a 12 X2 + · · · a,nXn:::::: b,
a 21 X 1 + a 22 X 2 + · · · a2nxn:::::: b 2
with x 1 ~ 0, x 2 ~ 0, · · · x n ~ 0. Notice that the number of constraints is

m and this need not be the same as n, the number of variables.
Examples of applications of this are
(a) maximising profits (z) from selling xi units of output each
giving a profit ci, subject to restrictions on the capacity (bi) of
the machines used in production;
(b) maximising the return on a portfolio where xi is the quantity
held of security of type i, subject to restrictions on the spread of
the securities;
(c) maximising utility (z) from consuming xi units of good i subject
to an income constraint and limits to the time that can be used
consuming the goods;
(d) maximising revenue (z) from selling xi units of good i subject to
restrictions on the expenditure on advertising each product in
different media (bi);
(e) maximising the consumption of vitamins (x) subject to the total
cost and the way the vitamins are combined in different foods.
These more general problems can be tedious to solve, particularly if
n is large and nowadays computer packages (such as LINDO/PC) are
available to obtain the solution.
A rather different complication occurs when the problem is one of
minimisation subject to constraints, rather than maximisation. This
occurs when the objective is to minimise costs, subject to achieving
minimum values of the variables. For example, minimise the cost of a
particular diet with the requirement that a basic level of each type of
vitamin is consumed or, given a set of production targets, how different
machines can be used to produce the output at a minimum cost. The
solution procedure using the simplex method is basically the same and
so is not discussed here. However, it turns out that every linear-
programming maximisation problem (generally referred to as the pri-
mal) has an equivalent minimisation problem (known as the dual).
Returning to our two-variable maximisation problem we have:
Max Z = C 1X 1 + C2X 2
subject to au x I + a 12x 2 :s:; b I
a21X1 + a22X2 :S: b2
with x 1 ;;?:0,x 2 ;;?:0
The equivalent minimisation problem can be written
Min
subject to
with u 1 ;;?: 0, u 2 ;;?: 0. The variables u 1 and u 2 are the dual variables. To
help in the interpretation of these we recall that in the maximisation
example z is the total profit, c 1 and c 2 are the profit per unit of x, and x 2 ,
and b 1 and b 2 are the total numbers of hours available on each machine.
The problem is to choose the values of x 1 and x 2 which maximise total
profit while satisfying the constraints; that is, in the numerical example,
Max z=2x 1 +3x 2
subject to x 1 +2x 2 :S:30
2xl+ X 2 :S:30
with
Putting the numbers into the dual we have
Min z' = 30u 1 + 30u 2
subject to U1 + 2u 2 ;::?; 2
2u 1 + U 2 ;::?; 3
with
Now u 1 is associated with the first constraint and is in fact the oppor-
tunity cost or shadow price of one hour of time on machine U for the
decision maker. We saw in the previous section that tableau 4 includes
these shadow prices and if the first constraint is changed marginally, say
from 30 to 31 hours, the maximum value of total profit will increase by

U1 units. Similarly, Uz is the shadow price of one hour of time on
machine V. If the cost of the unit of resource is less than the shadow
price it is worthwhile buying more of it. However, remember that the
shadow price applies only at the margin and not necessarily if there is a
large change in the value of a constraint.
The importance of the dual is that the optimal value of the objec-
tive function of the primal always equals the optimal value of the
objective function of the dual. Therefore, if the solution to the primal
problem is known, the solution to the dual can be found. More generally,
the primal problem with n variables and m constraints can be written:
Max
subject to
with
The corresponding dual problem has m variables and n constraints
and is
Min
subject to
a 1nU 1 + a 211U 2 + · · · + amnum ~ Cn
with U1 ~ 0, U2 ~ 0, · · · , Um ~ 0
If n < m it is usually easier to solve the primal problem.
8.8 Exercises
1. Use the simplex method to establish the product mix which will
provide the maximum contribution in the example shown in Table
8.4. Calculate the additional contribution which will be obtained
if one additional hour of capacity became available on either
machine U or machine V and show that this is equal to the
TABLE 8.4
No.
of hours Total
required no.
per unit of hours
A B
Machine U 2 3 25
Machine V 4 1 35
Contribution 9 7
per unit of product
incremental value of machine capacity derived from the final

tableau of the simplex procedure.
2. Using the basic data provided in Exercise 1, determine the number
of units of A and B which should be manufactured if either of the
following conditions applies.
(a) the total number of available hours of machine Vis increased
to (i) 40 and (ii) 45 in successive stages;
(b) the contribution per unit of A is reduced to 4;
(c) the number of hours required by product B on machine U is
reduced to 2.
3. A farmer can grow two types of crop, A and B. The profit per unit
of A is 3 and that for B is 2. Each unit of A requires one unit of
labour, two units of land and three units of machine time, while for
B the requirements are four of labour, one of land and one of
machine time. There are 60 units of labour, 40 units of land and 50
units of machine time available. (a) What should the farmer
produce to maximise profits? (b) Is it worthwhile renting 10 extra
units of land at a total cost of 5?
4. Suppose that for a healthy diet an adult has a daily requirement of
at least 10 units of energy and 15 units of protein and the following
foods are available:
Bread Cheese Meat
Price per unit 2 4 5

Energy per unit 1 2 1
Protein per unit 1 3 6
(a) Set up the linear-programming problem for minimising the

total cost of satisfying the daily requirements.
(b) Form the dual of the minimisation problem and obtain the
optimal solution.

1. Use the graphical method to determine the maximum
value of z=l0x 1 +5x 2
subject to 5x 1 + 5x 2 ~ 21
~
and
2. Use the simplex method to answer Question 1. What is the shadow
price of the first constraint?
3. Use the graphical method to determine the maximum
value of z = 2xl + x2
subject to ~
2xl + X2 ~ 40
and x 1 ~ 0, x 2 ~ 0, and comment on the solution.
4. A publisher produces hardback and paperback versions of a book.
To produce one hardback copy requires 5 units of labour and 3 of
machine time, and to produce one paperback copy requires 3 units
of labour and 2 of machine time. The profit from one hardback is 5
and that from one paperback is 2. There is a limit of 40 units of
labour and 30 of machine time. What combination of books should
be produced in order to maximise profits? Is it worthwhile buying
extra labour at a cost of 2 per unit?
5. Write down and solve the dual of the following problem:
min z=2x 1 +3x 2 + X3
subject to 2xl + Xz + X3 ~ 4
X 1 + 2x 2 + 4x 3 ~ 6
with
6. A workshop can make chairs, tables and desks. The profit is 5 from
each chair, 15 from each table and 20 from each desk. A chair
needs 3 units of labour, 2 of machine time and 1 of polishing. A
table needs 4 units of labour, 4 of machine time and 3 of polishing.

A desk needs 8 units of labour, 6 of machine time and 6 of
polishing. Labour is limited to 35 units, machine time to 30 units
and polishing to 30 units. What should be produced? What are the
opportunity costs of labour, machine time and polishing?
7. Determine the maximum value of
subject to
x 1 +3x 2 +3x 3 + x 4 :;;;50
3x 1 + X 2 + 2x 3 + 2x 4 :::;; 60
X 1 + X 2 + 4x 3 + 3x 4 :::;; 60
2x 1 + 2x 2 + X 3 + X 4 :::;; 50
with X;;;:: 0
Chapter 9
Differential Equations
9.1 Introduction
In Chapter 6 we saw that equations of the form
dy = f(x)
dx
could be solved by integration. Such an equation is known as a differen-
tial equation since it expresses the relationship between the derivative of
y with respect to x and the value of x itself. It contains only the
first-order derivative and is, therefore, known as a first-order differential
equation and because this term is only raised to the power one it is a
first-degree equation.
( dy I dx )2 = 6x + 2 is a differential equation of first-order and
second-degree because it contains only a first-order derivative and this is
raised to the power two.
d 2y I dx 2 = 6 is a differential equation of second-order and first-
degree because it contains a second-order derivative which is raised to
the power one.
and
are both differential equations of second-order and first-degree because

the highest-order derivative that is present in each equation is the
second and this is raised to the power one in both cases.
In general, a differential equation contains one or more derivatives,
all or some of which may be raised to a power other than zero.
Differential equations are classified into groups by
1. The order. This is the same as that of the highest-order deriva-
tive which is present in the equation.
319
2. The degree. This is determined by the power to which the

highest-order derivative is raised.
The general differential equation
is one of the nth order and the xth degree, even if () is greater than x.
The equations which occur mostly frequently are those of first-and
second-order, both usually being of the first degree. The problem is to
find a solution which satisfies the following conditions.
1. It must be free of terms containing derivatives.
2. It must satisfy the differential equation.
This can be illustrated by reference to the example
d2y
-=6
dx2
Integration of this second-order equation leads to the first-order equation
dy = 6x +A
dx
which on further integration leads to
y = 3x 2 +Ax+ B
This is the solution of the second-order equation since it does not
contain derivatives and it satisfies the original equation (as can be
checked by differentiation). In obtaining the solution the result
dy
Jdxdx=y
was used and since

Jldy = y
we are effectively treating dy I dx as the ratio of two differentials. This
separation of dy and dx is particularly useful in the solution of certain
differential equations, as will be seen below.
We now consider some of the standard techniques of solution of
first-and second-order differential equations.
Differential Equations 321
9.2 First-order differential equations

The general form of first-order and first-degree differential equa-
tions is
dy
dx = F(x, y)
We will consider various special cases before looking at a general

method of solution of this equation.
SEPARABLE VARIABLES
An equation with separable variables is of the form
dy = ~
dx !2(y)
or, using differentials
~ = fiy)dy
where / 1 ( x) contains only terms in x, and / 2 (y) contains only terms in y.

Integrating gives
f/1(x)dx = J!2 (y)dy
For example, if
dy X
dx y
then y dy = x dx.
The solution can be found by integrating both sides of this equation
J y dy = J X dx
or y2 - x 2 = 2C = constant =A
A check can be made that this solution does satisfy the differential
equation by differentiating, in the following manner. (For discussion see
section 7.8.)
Let
then oz = -2x oz
-=2y
ox ay
dy iJzliJx (- 2x) x
and -=---=---=-
dx iJz!iJy 2y y
the solution does therefore, satisfy the equation. The arbitrary constant,
A, arises because the process is simply one of integration. To determine
the value of this constant requires some further information about the
values of x and y. This is frequently in the form of initial conditions,
where the starting values are given (such as revenue is zero when sales
are zero), or boundary conditions, where the maximum or minimum
values are stated (such as x is positive or y has a maximum of 100). In
the example we assume that the initial condition is
y=2 when X= 1
then or A= 3
and the unique solution to the differential equation
dy X
given y = 2 when x = 1
dx y
is
Example I
What is the general form of the demand equation which has a constant
elasticity of - 1?
Let q be the quantity demanded at price p. Then elasticity, e, is
defined by
p dq
e=--
q dp
Fore = -1, we require
I!. dq = -1 or dq = - dp
q dp q p
Integrating, we have
loge (q) =-loge (p) + lo& (A)
where A is a constant.
and or pq =A
DIFFERENTIAL EQUATIONS WITH HoMOGENEous CoEFFICIENTS

The above method breaks down if the variables are not separable
but if the differential equation
/ 1 (x, y) dx + / 2 (X, y) dy = 0
has homogeneous coefficients it can be converted to the separable-

variables type by a suitable change of variable. (For a discussion of
homogeneity see section 7 .10.)
For example, if
dy X+ y
-=--
dx x
then (X + y) dx - X dy =0
In this case the variables are not separable but the coefficients of dx
and dy are homogeneous of degree one. This can be converted into the
separable-variables type by the substitutions
y = vx and dy = v dx + x dv
After carrying out the substitution we have
(x + vx) dx - x(v dx + x dv) = 0
which can be rearranged into
x(l - v) dx + x(v dx - x dv)= 0
x[(l - v) dx + (v dx - x dv)]= 0
If this is to be valid for all values of x then
( 1 - v) dx + (v dx - x dv) = 0
that is dx - v dx + v dx - x dv = 0
or dx- x dv =0
This is now in the separable-variables form and can be treated as
such
dx =x dv
dx
-=dv
X
and hence loge X =v + c
To obtain the solution in terms of the variables x and y it is

necessary to substitute for v in this equation.
v = 1'._
X
loge ~ C or y = x(loge X - c)
X
This is, therefore, the solution of the linear differential equation

dy X+ y
-=--
dx X
The result can be checked by differentiation.

Example 2
The total cost, y, of producing x units of output of a product is related to
the marginal cost of production by the equation
dy y- 2x
-=---
dx 2y- X
What is the total cost function? ·

The equation can be rewritten as
(2y- x) dy + (2x- y) dx = 0
The coefficients of this equation are homogeneous of degree 1, and so
we substitute y = vx and dy = v dx + x dv
(2vx - x)(v dx +x dv) + (2x - vx) dx =0
(2v 2x - xv + 2x - vx) dx + (2vx 2 - x 2 ) dv =0
Dividing through by x gives
(2v 2 - 2v + 2) dx + x(2v - 1) dv =0
dx 2v- 1
that is - + dv =0
x 2v 2 - 2v +2
Integrating gives
loge x + 1-lo& (2v 2 - 2v + 2) = ~A
or x(2v 2 - 2v + 2)112 =A
Substituting v = yIx and squaring gives
~
x2 x
or
This is the relationship between total cost, y, and output, x. The value of
A can be found if some extra information on x and y is available.
Suppose that the initial condition is that when x = 0, y = 100 so that the
fixed cost is 100. Substituting in the solution gives A 2 = 20,000.
DIFFERENTIAL EouATIONS WITH NoN-HoMoGENEous CoEFFICIENTS

These are of the form
(a 1x + b 1 y + c,) dy + (a 2x + b 2y + C2 ) dx =0
The coefficients are not homogeneous but they can be made so by
the substitutions
a,x + b 1 y + c1 = Y (1)
a2x + b2y + C2 = X (2)
This results in
X dx + Y dy =0
dy X
or -= -- (3)
dx y
It is now necesary to differentiate (1) and (2)
dY
dx
= a
1
+b dy
1dx
= a
'
+ b ( - X)
1 y (4)
dX
-=a+b-=
dx 2 2dx a2 +
dy
b(-X)
2 y (5)
Dividing (4) by (5), we obtain

dY/dx dY a,+ b,(-XIY)
--=-=....:....-...!...!...._--.!..
dX/dx dX a2 + b2 ( - XIY)
dY a,Y- b 1X
dX a2 Y- bzX
or (a,Y- b,X) dX- (a 2 Y- b2 X) dY =0

This equation is homogeneous in X and Y and it is possible to
obtain a solution as above. The exercise is left to the reader to complete
as the procedure is no different from that previously demonstrated
(section 9.4, Exercise 1).
Example 3
(x + 2y + 1) dx + (2x + y + 2) dy =0
Let X+ 2y + 1 =X (1a)
2x+y+2=Y (2a)
dy X
and -= (3a)
dx y
Differentiating (1a) and (2a), we have

dX = 1 + 2dy = 1 _ 2X = Y - 2X (4a)
dx dx Y Y
dY = 2 + dy = 2 _ X= 2Y- X (Sa)
dx dx Y Y
Dividing (Sa) by (4a) gives
dY/dx dY (2¥- X)/Y 2¥- X
dX/dx dX (Y - 2X)/Y Y - 2X
(2 Y - X) dX - ( Y - 2X) dY =0
This is homogeneous in X and Y, so let
Y= VX and dY= VdX+ XdV
then (2VX - X) dX - (VX - 2X)(V dX + X dV) =0
X[(2V- 1) dX- (V- 2)(V dX +X dV)] =0
(2V- 1) dX- (V- 2)(V dX +X dV) = 0
2V dX - dX - V 2 dX - VX dV + 2V dX + 2X dV = 0
(2V - 1 - V 2 + 2V) dX- (VX- 2X) dV = 0
(- V 2 + 4V- 1) dX- X(V- 2) dV = 0
This is in the separable-variables form and can be written
dX_( V-2 )dv

X - V 2 + 4V- 1
The integral of the left hand side is simply ~ X. The integral of
the right hand side is obtained as follows:
S- V V+- 4v2 - dV =- f !(2V - 4) dV

2 1 V 2 - 4v + 1
- _1 f 2V- 4 dV
- 2 V2 - 4V + 1
= - ! loge (V 2 - 4V + 1) + C
Combining these two results
loge X= -! ~ (V 2 - 4V + 1) + C
It is more convenient in examples such as this to let the constant
assume a logarithmic form, i.e. let C = ~ A. From the properties of
logarithms (see section 4.14) it is possible to write this as
loge X= ~ A -loge Y(V 2 - 4V + 1)
A
= ~ Y(V 2 - 4V + 1)
X= A
Y(V 2 - 4V + 1)
or X 2 (V 2 - 4V + 1) =A 2 =constant= B
The result must be expressed in terms of the variables x and y and
this is done in two stages
1. Replace V by YIX
then X 2 ( -Y 4Y + 1)
2
- - = B
X2 X
that is Y 2 - 4YX + X 2 = B
2. Replace X by (x + 2y + 1) and Y by (2x + y + 2)
then (2x + y + 2Y - 4(x + 2y + 1)(2x + y + 2)
+ (x + 2y + 1Y = B
that is,
4x2 + y 2 + 4 + 4xy + 4y + 8x- 8x2 - 4xy- 8x- 16xy- 8y2 - 16y

- 8x - 4y - 8 + x 2 + 4y 2 + 1 + 4xy + 4y + 2x = B
or - 3x2 - 3y2 - 3 - 12xy - 12y - 6x = B
- 3(x2 + y 2 + 4xy + 4y + 2x + 1) = B
x2 + y 2 + 4xy + 4y + 2x = - 3B - 1 = constant
:. the solution to the differential equation
(x + 2y + 1) dx + (2x + y + 2) dy =0
is given by
x2 + y 2 + 4xy + 4y + 2x = C
EXAcr DIFFERENTIAL EouATIONs

An exact differential equation involving x and y is one which
may immediately (i.e. without multiplying through by any factor) be
expressed in terms of derivatives of functions of x and y, where these
functions do not themselves involve derivatives.
3x 2y + x3 dy = 0
dx
is an exact differential equation because it is formed by differentiating
the equation x 3y = A.
It should be possible to find the solution to an exact differential
equation quickly if it is recognised as such. This is not always obvious at
first sight but the following rule always holds.
p dx + Q dy =0
is an exact differential equation if
iJP iJQ
-=-
iJy iJx
For example, consider the equation
3x2y + X3 dy =0
dx
(3x 2y) dx + (x 3 ) dy =0
and
These two partial derivatives are equal and therefore the equation
is exact. The solution in this case is fairly obvious when the values of
P and Q are considered.
and the solution is x 3y = C.

When the solution is not so obvious it can be found by using the
following relationships:
If z = f(x, y) = 0
Then
az
dz = -dx + -dy = 0
az
ax ay
(See section 7.5)
To reverse this procedure and find z it is necessary to compare the
coefficients in the equation with the partial derivatives
az and
az
ax ay
For example, in the exact equation
(2x + 3y) dx + (3x- 2y) dy = 0
dZ
- = 2x + 3y (1)
dX
and
az
- = 3x- 2y (2)
ay
It is now possible to suggest the function from which these two
partial derivatives were formed.
From (1), z must contain the terms x 2 + 3xy + C
and from (2), z must contain the terms 3xy - y 2 +C
It must be concluded that the function will include all these terms
without double counting any that are common to both. This leads to the
solution
z = x2 + 3xy - y 2 + C = 0
and it is easily checked that this is the solution to the differential
equation.
This differential equation has homogeneous coefficients and could
have been solved as such with the same result but with a good deal more
effort. It is, therefore, extremely useful to be able to recognise an exact
differential equation when it occurs and to apply the above method to
obtain a solution.
A GENERAL METHOD FOR FIRST-ORDER LINEAR EQUATIONS

It is generally possible to make a non-exact differential equation
into an exact one by multiplying by an integrating factor. For example,
let dy =X+ y
dx
dy
then - - y =x
dx
The left hand side of the equation is not exact and it is difficult to
integrate as it stands. But if all the terms are multiplied by e-x the
equation becomes
The left hand side is now exact and is the derivative of e-xy and the
right hand side of the equation can be integrated by parts (see section
6.14).
Jxe-x dx = x( -e-x) - J(-e-x) dx
= - xe-x - e-x +C = e-x(- x - 1) + C
This is equal to the integral of the left hand side, i.e. the exact part,
of the equation
Dividing throughout by e-x, we have
Y = -X - 1+ ~ = -X - 1 + Cex
e
therefore the solution to the differential equation
dy =X+ y is y = Cex- X- 1
dx
This can be checked by differentiation:
dy = cex- 1 and Cex =X+ y + 1

dx
:=x+y+1-l=x+y
In general, the integrating factor is equal to e8 where (J = J P dx for

all differential equations of the form
dy + Py = Q
dx
where P and Q are functions of x only, including the case where they are
constants. This method is useful where both P and the product of the
integrating factor and Q are reasonably easy to integrate.
Example 4
dy + 4y = 2x
dx
In this case P=4 and JP dx = 4x
Therefore the integrating factor is e4x, and multiplying by this gives
e4x dy + 4e4xy = 2xe4x

dx
The left hand side is now the derivative of ye4X, and integrating the
right hand side gives
X 1
or Y = - - - + Ae-4x
2 8
9.3 Applications of first-order ditferential equations

We now examine a number of applications of first-order differential
equations in economics and business.
THE DoMAR DEBT MonEL

Suppose that the national debt, D, varies as time (t) passes and its
rate of change is related to the level of national income such that
dD
-=ay (a> 0)
dt
where y varies with t. Also, suppose that y is increasing at a constant
rate b, so that
dy = b (b > 0)
dt
Solving for y, y = bt + A
where A is a constant of integration. Therefore,
dD
- = abt + aA
dt
and integrating, D = 0.5abt 2 + aAt + B
where B is a constant of integration. The ratio of debt to national
income is given by
D = 0.5abt 2 + aAt +B = O.Sat + 0.5aAt + B
y bt +A bt +A
As t ~ oo, 0.5at ~ oo and so Dly ~ oo. That is, with this model, the
ratio of debt to national income increases without limit as time passes.
One unrealistic aspect of this model is the assumption that y in-
creases at a constant rate. Suppose instead that y increases at a constant
percentage rate,
dy
-=by (b > 0)
dt
Now the solution for y is from
dy = b dt or log y = bt + log C
y
andy = C ebr where Cis a constant.
The relationship for the change in debt is now

dD
- = aCebt
dt
and integrating gives
aC
D=-eb1 +E
b
where E is a constant. The ratio of debt to national income is now
D = (alb)C ebt + E = (alb) + (EIC)e-bt

y C ebt
As ~ oo, Dly ~ which is a positive constant. That is, the debt
becomes proportional to the level of national income.
CosT FuNCTIONS
The costs of many activities vary as time passes. For example, the
running costs of a car generally increase as the car gets older and at the
same time the value of a car declines. Suppose that the relationship
between running costs (C) and resale value (V) is given by
dC a
dt = v (a> 0)
where a is a constant. Also, suppose that the value changes according to
dV = -bV (b > 0)
dt
where b is a constant. There are two obvious initial conditions here:
when t = 0 the running costs, C, are zero and the resale value, V, is the
purchase price ofthe car, V 0 , say. Thus, when t = 0, C = 0 and V = V 0 •
Solving the second equation:
dV
- = -b dt or log V = - bt + log A
v
where A is a constant. Rearranging gives
V = Ae-br
The initial condition t = 0, V = V 0 gives V 0 = A since e0 = 1 so that the
the solution for the value is

V = Voe-bt
Substituting into the cost equation,
dC a
dt
= =
Integrating,
aebt
C=-+B
V0b
where B is a constant. Using the initial condition t = 0 and C = 0,
a
0=-+B or
V0b
The solution is
a (ebr - 1)
c= V:b
0
MARKET EQUILIBRIUM
In some models of demand and supply a dynamic term is included
to allow for adjustments of behaviour. For example the market for a
product might be described by
demand: q = 150 - 6p + 5 ~
supply: q = 4p - 50 + 20 dt
Here both demand and supply are higher if prices are rising. In equilib-
rium:
150 - 6p + 10 dp = 4p - 50 + 20 dp
dt dt
or - 10 dp = lOp - 200
dt
-10 dp
Rearranging:
lOp - 200 = dt
Integrating: -log (lOp - 200) = t- log A

or lOp - 200 = Ae-t
where A is a constant. The value of A can be determined from an initial
condition. Suppose that when t = 0, p = 21 then A= 10 and the time
path of prices is given by
lOp = 200 + we-t or p = 20 + e-t
W ALRASIAN ADJUSTMENT
In the study of markets, the equilibrium price is that for which the
quantity oemanded ( qd) and the quantity supplied (if) are equal. The
analysis is static and time has no role to play. However, the basic laws of
demand and supply state that if demand exceeds supply for any good
there is a tendency for the price to rise, while if supply exceeds demand
the price will fall. These effects can be modelled by the Walrasian
adjustment mechanism
dp = a(qd- lf) (a> 0)

dt
where a is a constant, in which the change in price is positive if qd > lf.
Here it is assumed that prices do not adjust instantaneously and that the
rate of adjustment is larger the greater the difference between lf and qd.
Now suppose that the demand and supply curves are
qd = 60- 3p
qs = -15 + 2p
and that a = 0.4, then substituting into the differential equation gives
dp = 0.4(60 - 3p + 15 - 2p) = 0.4(75 - 5p) = 30 - 2p

dt
While this is a non-homogeneous equation the solution can still be
found as previously. The equation can be written
dp
30- 2p = dt
and integrating,
-0.5 log (30- 2p) = t +A
This can be rearranged as

30 - 2p = Be- 21 or p = 15 - 0.5Be-21
Notice that as t ~ oo, p ~ 15 which is the equilibrium price. In this case
the price converges to the equilibrium price. However, it is easily seen
that if, say, a= -0.4 then the solution is
p = 15 + 0.5Be 21
and as t ~ oo, p ~ oo (if B > 0) so that equilibrium is not achieved.
CoNTINuousLY CoMPOUNDED INTEREST

The continuously compounded interest model of sectiorr 4.10 can
be obtained by using differential equations. Let S be the amount in a
savings account and suppose that initially S = 100. If interest is com-
pounded continuously at 5% per annum then the change inS is 0.05 of
S or
~~ = 0.05 s
This can be solved by writing it as
sdS = 0.05 dt
and integrating to give
logS= 0.05t + log A
or S = Aeo.ost
The initial condition is when t = 0, S = 100 and so A = 100 and the
solution is that the value in the account at time t is
9.4 Exercises
1. Find the solution to the equation
(a, Y - b, X) d X - (a 2 Y - b 2 X) d Y = 0.
2. What is the general form of the demand equation which has an
elasticity of - n?
3. The total cost of production, y, and the level of output x are

related to the marginal cost of production by the equation
dy 24x 2 - Y2
dx xy
What is the total cost function if x = 2 when y = 4?

4. Determine the consumption function for which the marginal prop-
ensity to consume is
de
- = a - bx
dx
where c is consumption and x is income.
5. With a Walrasian adjustment process, if the quantities supplied
and demanded differ then the rate of change of prices is given by
dp = kz
dt
where tis time, z is excess demand and p is the deviation of price

from its equilibrium level. Solve the differential equation for p
when
(a) z = ap (b) z = ap + bp 2
6. The Harrod-Domar growth model can be formulated as follows:
S = aY (1)
I= b dY (2)
dt
I= S (3)
Where S is savings, Y is income and I is investment. Substitute

from (1) and (2) into (3) and solve the differential equation for Y.
7. Solve the following equations
(a) X dy =X- y (b) dy = y + 1

dx dx 2y + x
(c) ~~ = x(1 + x) - 2y (d) dy

dx
+ 4y = x3
8. If supply and demand are given by

if = -a + bp and qd = e - fp + dp (a, b, e,f > 0)
dt
assuming that the market clears at every point in time determine
the time path of p.
9.5 Second-order homogeneous differential equations

The general form of linear second-order differential equation with
constant coefficients is
d2 d
a ___I + b _l + cy = f(x)
dx 2 dx
where a, b and c are constants. When f(x) is equal to zero the ~ is
said to be homogeneous and one solution of the homogeneous equation
d2 d
a2+b_l+cy=O
dx 2 dx
is of the form y = emx where m is a constant. To determine m let y = emx.
Then dy =me=
dx
Substituting these values in the homogeneous equation, we obtain
am 2 emx+ bmemx + ce"" = 0
emx(am 2 + bm + c) = 0
(am 2 + bm + c) = 0
This quadratic in m is called the characteristic or auxiliary equation
and the roots can be found by the usual method (see section 3.3).
Therefore
- b ± V(b 2 - 4ac)
m=
2a
- b + V(b 2 - 4ac) - b - V(b 2 - 4ac)

or
2a 2a
This suggests two possible alternatives for the solution to the dif-
ferential equation:
and y = e"'zX
But the solution to a second-order differential equation must con-

tain two constants of integration. This is achieved by combining the two
solutions in the form
where k 1 and k 2 are the two arbitrary constants. This value of y satisfies
the differential equation as can be shown by substitution.
Example 1
d 2y 5 dy
---+6y=O
dx 2 dx
Then if y=fF
m 2 fF - Sme= + 6£F = 0
e=(m 2 - 5m + 6) = 0 or m2 - 5m + 6 = 0
:. m 1 = 2 and m2 = 3 and the solution is
Y = kte2x + k2e3x
Check: dy = 2kte2x + 3k2e3x

dx
Hence
and this is equal to 0.

The term emx will always be ignored after the function has been
differentiated and it is not necessary to include it once we are familiar
with its purpose. If it is ignored the above procedure can be convenient-
ly summarised in the following set of rules.
1. From the differential equation form a new equation in which
d 2y/dx 2 is replaced by m 2 , dyldx is replaced by m and y is
replaced by unity:
am 2 + bm + c = 0
2. Find the roots of this quadratic equation, m 1 and m 2 •

3. Form the solution to the differential equation
Y = kte"'tx + kzemzX
The values of the constants k 1 and k 2 can be found if two sets of data
are available about the system.
Returning to the particular differential equation discussed above,
let us assume that the initial conditions are when x = 0, y = 1 and when
x = 0, dy/dx = 4. This information can be used to find the values of k 1
and k 2 •
For y = k 1e2 x + k 2 e3x
dy
dx
= 2k t e2 x + 3k ze3 x
Substituting the initial conditions into these equations
1 =ke +ke =k +
1
0
2
0
1 k2
4 = 2k 1e0 + 3k2e0 = 2k 1 + 3k2
The solution of this pair of simultaneous equations is
k1 = -1 k2 =2
The solution to the differential equation
d 2y 5 dy
---+6y=O
dx 2 dx
with X =0 y = 1; X= 0, dyldx =4
is given by
The above method must be modified when the roots of the auxiliary
equation are equal, i.e. m 1 = m 2 = m. The solution is given by
y = (kl + kzX)enx
If this modification is not applied the result contains only one
constant of integration and this could not form the complete solution to
a second-order differential equation:
Y = k enx + k emx = (k + k )enx = kemx
1 2 1 2
The basic result does, however, hold when the roots of the auxiliary
equation are complex. For example,

m1 =a+ bi m 2 =a- bi
In this case the solution is given by
y = k 1e(a+bi)x + k 2e(a-bi)x
This can be expressed in a more convenient form if the following
expansions are used (see section 5.16):
eix = cos x + i sin x e-ix = cos x - i sin x
Replacing x by bx gives
ebix = cos bx + i sin bx e-bix = cos bx - i sin bx
By algebraic manipulation
y = k[ e<a+bi)x + k2 e<a-bi)x
becomes y = k[ eax ebix + k2 eax e-bix = eax[k[ ebix + k2 e-bix]
= ~ [k 1(cos bx + i sin bx) + k 2 (cos bx - i sin bx)]
and by collecting together the terms in cos bx and sin bx
y =~ + k 2 ) cos bx + i(k 1 - k 2 ) sin bx]
= eax[ k3 cos bx + k. sin bx]
where and
An even more useful form can be obtained by considering a right-
angled triangle with sides equal to k 3, k 4 and V(k; + k;) as shown in
Fig. 9 .1. Let the angle indicated be f.
k3 k.
Then cos f = - - - - - and sin f = -----
V(k; + kD V(k; + kD
These results can be used if the solution to the differential equation
is written in a slightly different form.
y = eax[k3 cos bx + k. sin bx]
= eax ~v1(k; + k;) [ cos bx ~ 1 k3 + sin bx ~ 1 k 4 ]
v (k; + kD v (k; + k;)

= eax + kD[cos bx cos t + sin bx sin t]
V(k;
= eax A[cos bx cos f + sin bx sin f]
k3
Fig. 9.1
where A = v'(ki + kD and is a constant.

Using the trigonometrical identity
cos (bx - t) = cos bx cos f + sin bx sin f
(see Appendix A, section A.2) the solution can now be written in the
form
y = Aeax COS (bx - f)
where A and f are the two necessary constants with values
and
The values of both can be obtained from the initial data about the
system. Thus the solution to a second-order homogeneous differential
equation whose auxiliary equation has complex roots involves trigo-
nometrical functions.
Example 2
The relationship between national income, y, and time, x, is given
by
2d d
- 3 --.l
2
(a) + 2y = 0
dx 2 dx
(b) cz - i + 10 25y = 0
d2 d
(c) ___l + 6 __1. + lOy = 0
dx 2 dx
In each case, obtain y as a function of x if x = 0 when y = 0 and dyldx = 2
when x = 0.
d 2y dy -
(a) dx2 - 3 dx + 2y - 0.
Let y = f!"X. Then
dy = memx and
dx
Substituting in the equation gives
emx(m 2 - 3m + 2) =0
+ 3 ± v'(9- 8)
and so m = = 2 or 1.
2
y = Ae2x +Be
The initial conditions are x = 0, y = 0 and x = 0, dyldx = 2. Substituting

gives two equations in A and B:
0 =A+ B
and since dy/dx = 2Ae2x + BeX, 2 = 2A + B. Therefore, solving for A
and B gives A = - B and so A = 2, B = -2.
(b) dzy- lOdy + 25y = 0

dx 2 dx
Let y = emx. Then
dy = memx and
dx
Substituting in the equation gives
f!"'X(m 2 - 10m + 25) = 0
10 ± Y(lOO - 100)
and so m= =5
2
y =(A + Bx)e 5x
dy =(A + Bx)Se 5x + Be 5x = (SA + B + SBx)e 5x

dx
When X= 0, y=O and A= 0
When X= 0, dyldx = 2, so 2 =SA+ B
A = 0 and B = 2, and the solution is y = 2xe 5x
d2 d
(c) _2. + 6 __l + 1Oy = 0
d.x2 dx
Let y = e=. Then
dy = memx and
dx
Substituting into the equation gives
f!=(m 2 + 6m + 10) = 0
- 6 ± V(36- 40)
and so m= -3±i
2
We know that if m =a ± bi then
y = Aeax cos (bx - f)
and so since m = -3 ± i, a = - 3 and b = 1.
y = Ae-3x cos (x - f)
The initial conditions are x = 0, y = 0 which gives
0 =A cos(- f) (1)
and X= 0,
Now dy = - 3Ae-3x cos (x - f) - Ae-Jx sin (x - f)

dx
2 = - 3A cos (- f) - A sin (- f) (2)
From (1) either A = 0 or cos (-f) = 0, that is f = n/2
From (2): 2 = - A[3(0) + sin (- n/2)) = -A( -1)
A=+ 2, f = n/2 and y = 2e-3x cos (x - n/2)
9.6 Exercises
Solve the following differential equations:
d 2y dy
1. dx2 - dx - l2y = 0
d 2y dy -
2. dxz - 3 dx + 2y - 0
d 2y dy -
3. dx2 - 6 dx + 9y - 0
d 2y dy -
4. dxz - 2 dx +y- 0
d2y -
5. dx 2 - l6y - 0
d2y -
6. dx2 + 16y - 0
d 2y dy -
7. dx 2 + 2 dx + lOy - 0
9. 7 Second-order non-homogeneous differential equations

The non-homogeneous second-order differential equation with con-
stant coefficients has the general form
d2 d
a ____I + b ....1.. + cy = f(x)
dx 2 dx
where f(x) is either a constant or a function of x only, and a, b, care
constants.
It can be shown that the solution to this type of equation is the sum of
two parts which are obtained independently. These are called the com-
plementary function (CF) and the particular integral (PI). The com-
plementary function is simply the solution of the homogeneous part of the
equation formed by lettingf(x) = 0 and this can be obtained by the method
of section 9.5. The particular integral is any solution which satisfies the
differential equation and its form depends upon the form of f(x).
(a) f(x) is a constant. For example,
d 2y - 5 dy + 6 = 18
dx 2 dx y
If we let y = K = a constant then
dy =0
dx '
Substituting these values in the differential equation
0- 0 + 6K = 18 or K=3
Thus the solution y = 3 is a particular integral because it satisfies the
differential equation.
The complementary function was obtained in section 9.S when the
solution to
d2y - s dy + 6y = 0
dx 2 dx
was shown to bey = k 1e2 x + k 2 e3x. Therefore the complete solution to
the differential equation
d 2y - s dy + 6y = 18
dx 2 dx
is the sum of the complementary function and the particular integral and
is given by
y = klezx + kze3x +3
This can be verified by differentiation and substitution in the usual
way. Any initial conditions are applied to the complete solution to give
the values of the arbitrary constants.
(b) f(x) is linear in x. For example,
d 2y - S dy + 6 = l8x
dx 2 dx y
Let us try a solution of the form y = K x + K Then
1 2•
dy- d 2y
dx - KH - 2
dx
=0
Substituting these values in the differential equation
0 - SK 1 + 6(K 1x + KJ = 18x
or 6K 1x - SK 1 + 6K = 18x
2
For this equation to be satisfied it must be true for all values of x

6K 1 = 18
TABLE 9.1
Particular
f(x) integral
ax 2 + bx + c K 1X 2 + K 2X + K 3
ex K tex
sin x K1 sin x + K2 cos x
and - 5K 1 + 6K2 =0
and K 2 -- 2~
The particular integral is then
= 3x + 2
5
y
and the complete solution to the differential equation
2
2d - d
5 ...l + 6y = 18x
dx 2 dx
is
From the above results it would seem reasonable to expect the

particular integral to assume the form that f(x) has in the differential
equation and this is indeed so in simple cases. Table 9.1 gives a few
examples. The reason that the last of these functions contains both sine
and cosine terms is that the differentiation of one leads to the other and
therefore a combination of two terms is used.
For example
Let the particular integral be represented by

y = K1 sin x + K2 cos x
then ~~ = K1 cos x- K 2 sin x
d2y .
dx 2 = - K1 sm x - K2 cos x
Substituting these values in the differential equation, we obtain
(-K 1 sin x- K2 cos x)- 5(K1 cos x- K2 sin x)
+ 6(K1 sin x + K2 cos x) = 18 sin x
Equating the coefficients of the sine and cosine terms

sin x(- K 1 + SK2 + 6K1) = 18 sin x
cos x(- K 2 - SK1 + 6K2 ) = 0
that is SKI+ SK2 = 18
-SK1 + SK2 =0
From which and
:. The solution to the differential equation
d 2y dy
- - S- + 6y = 18 sin x
dx 2 dx
is y = ~ + k 2 e3x + ~ x + cos x)
In some cases the particular integral may not be easy to find using
the above procedure.
For example, difficulty is experienced in the case where the term in
y is not present in the left-hand side of the equation and f(x) is a
constant. For example,
d 2y dy
- - 3-= 16
dx 2 dx
The auxiliary equation is
m2 - 3m= 0
that is m(m- 3) =0
and
and
For the particular integral let y = K. Then
dy =0 d2y =0
dx ' dx 2
and substitution of these values leads to 0 - 0 = 16, which is a
contradiction.
In such a case we can try y = Kx. Then
dy
-=K
dx '
and on substitution 0- 3K= 16

and K __
-
l§
3
and PI= ~
The solution to the differential equation
d 2y dy
- - 3 - = 16
dx 2 dx
is
In general, the particular integral can be found by this trial and
error method. It is also possible, and sometimes better, to find it by the
method of operators. This is not, however, discussed here because very
little advantage is gained by its use in simple cases. The reader should
consult a specialised text for a discussion of the method and its applica-
tion to more difficult problems.
9.8 Exercises
1. Solve the following differential equations with the given initial
conditions:
d 2y dy
( a ) - - 10- + 16y = 2x with x = 0, y = -5/64 and
dx 2 dx
dy
X= 0, -= 1
dx
d2y
(b) - - l6y =ex with x = 0, y = 14/15 and x = 0,
dx 2
dy = 29/15
dx
d 2y dy
(c) - - 2 - + y = e 2 x with x = 0, y = 2 and x = 1,
dx 2 dx
y = e2
d 2y dy
(d)-+2-+5y=x with x=O, y=-2/25 and x=O,
dx 2 dx
dy = 6/5
dx
d 2y dy
(e) ---=2cosx with x=O, y=1 and x=O,
dx 2 dx
dyldx =2
2. The relationship between output (x), and total cost (y) for a firm is
given by
d 2y dy
- +- - 6y = 18 X 2
dx 2 dx
where dy I dx is the marginal cost, and d 2y I dx 2 is the rate of change
of marginal cost. Given that y = 0 and dy I dx = 1 when x = 0,
obtain the total cost as a function of output only.
9.9 Graphical presentation

The solution of a differential equation expresses one of the vari-
ables as a function of the other. In many economic problems the
independent variable is time and it is of particular interest to consider
the variations which occur in the dependent variable over a period of
time. A convenient way of presenting this information is by a graph with
time (t) on the horizontal axis. It is impossible to discuss all the types of
differential equations, but a few examples are given here to indicate the
method to be used in obtaining a graph of the solution in any particular
case.
LINEAR FIRST-ORDER DIFFERENTIAL EQUATIONS

The solution to an equation of this form is
1. a polynomial in t, no general form exists. The graph can be
obtained by using the methods discussed in section 5.18.
2. an exponential function y = ke" 1• In this case various possibilities
exist:
(a) k is positive and a is positive (Fig 9.2) e" 1 continuously
increases with time and at an ever-increasing rate. Therefore
the function becomes very large and is said to be explosive.
The value of k determines the starting position, i.e. the
value of the function at time t = 0.
(b) k is positive and a is negative (Fig 9.3). This function con-
tinuously decreases with time and is said to be damped.
l
k
!L------
Fig. 9.2
Fig. 9.3
(c) k is negative and a is positive.

(d) k is negative and a is negative.
The last two function are the mirror images in the time axis of the
functions which are sketched in Figs. 9.2 and 9.3.
LINEAR SECOND-ORDER DIFFERENTIAL EQUATIONS WITH CONSTANT

COEFFICIENTS
The general equation
d 2y dy
a- + b- + cy = f(t)
dt 2 dt
has a solution of the form
y = k 1e"' 1' + k 2e"'Z' + PI
where PI depends upon the form of f(t).
1. The complementary function is the sum of two exponential
terms and the graph will depend upon the values of m 1 and m2 ,
the roots of the auxiliary equation.
(a) m 1 > 0 and
The function will increase continuously from the value y = k 1
+ k2 at timet= 0 and will, for any value oft, be equal to the sum
of the two separate functions
y = k e"'
1
11 and y = k e"'Z
2
1
If m1 is very much greater than m2 (written m, ~ m2 ) then

k,e"''' ~ k 2e"'z' as t becomes very large and the function approxi-
mates toy = k,e"' 1'.
(b) m, < 0, and
The function decreases continuously over time from the value
(k, + k2 ) and approaches the value zero.
(c) m, > 0, m2 < 0, k1 >0 and k2 >0
The term y k2e"'2' decreases continuously as in (b) above
because m2 is negative and the function tends towards y = k 1e"' 1'
as t increases.
The same reasoning can be applied to other possible com-
Fig. 9.4
binations of the constants in the complementary function and

the shape of the function determined.
(d) Complex roots
The solution is of the form
y = Ae'" cos (kt - f)
and this can be thought of as two parts. Ae'" has a form similar to
one of those discussed above, and cos (kt - f) is a function
which can take on all values between - 1 and + 1 but has no
values outside these limits. If A > 0, a > 0, the solution oscil-
lates about the time axis with extreme values Ae"1 and - Ae"1
(Fig. 9.4).
The amplitude of the oscillations increases at each success-
ive cycle and the function is explosive. The constant E" helps to
determine the starting point in the cycle because when t = 0
y = A cos (- f) = A cos f
If f = 0 the function has an initial value which is deter-

mined by the value of A. For all other values of f the initial
value of the function is less than A with the extreme case of cos f
= - 1 when the function starts at -A.
f is said to determine the phase, or position, within the
cycle at which the function starts.
2. The effect of the particular integral is to modify the function in a
manner which is determined by the form of the integral.
(a) PI is a constant
The shape of the function is unaltered but its value at time
t = 0 is changed to the value of the constant. That is, when
t = 0, Y = k 1 + k 2 + C.
(b) PI is a polynomial in t
This can have a large effect upon the shape of the curve
particularly for small values of t, but has little effect for
larger values of t, on functions which contain exponential
terms which are explosive.
(c) PI is a trigonometric function
This will add a periodic oscillation to the function, the
amplitude and period of which will be determined by the
trigonometric function.
9.10 Exercises
Sketch the graphs of the solutions to Exercises 9.8, Questions 1
and 2.
9.11 An inflation-unemployment model

A popular way of analysing the interactions between inflation,
unemployment and monetary policy is based on the expectations-
augmented Phillips curve. Here we will examine the model in con-
tinuous time using differential equations. The equivalent model using
discrete time will be discussed in section 10 .11.
The original Phillips curve can be written
w =a- bU (a, b > 0)
where w is the rate of change of money wages and U is the level of
unemployment. This implies that the higher is unemployment then the
lower will be wage changes. The equation is usually modified in two
ways. First, w is replaced by w - pe, where pe is expected inflation, so
that it is expected real wages which are related to unemployment.

Secondly, the rate of inflation, p, is related to w by
p=w-c (c > 0)
where c is the rate of productivity increase (assumed to be constant).
Together these result in the expectations-augmented Phillips curve
being written as
p = a - bU - c + pe (a, b, c > 0) (1)
To explain inflation expectations a simple adaptive expectations equa-
tion can be used:
(1 ~ k > 0) (2)
This states that the change in the expectation is a proportion, k, of the

error in expectations in the current period. Finally, we assume that the
change in unemployment is related to the rate of growth of real money,
or
dU
- = -e(m- p) (e > 0) (3)
dt
where m is the nominal rate of growth of money and p is the rate of
inflation. We will also assume that m is a constant.
The three equations (1)-(3) explain p, pe and U and can be com-
bined to give differential equations in any of these variables. Initially we
will reduce the model to a differential equation in p.
First we differentiate (1) with respect to time, t, to give
dp = _ b dU + dpe
(4)
dt dt dt
since a and c are constants. Next we substitute from (2) and (3) to
remove the derivatives of U and pe:
dp
- = be(m - p) + k(p - pe) = bern+ (k- be)p - kpe
dt
Since this includes pe we differentiate it to give
d 2p dp dpe
- = (k - be) - - k- (5)
dt 2 dt dt
Now we can use (4) to eliminate the dpeldt term

d 2p = (k - be) -dp - k ( -dp
- + bdU)
-
dt 2 dt dt dt
dp dU
=-be-- b k -
dt dt
Finally, using (3),
d 2p dp
- = - b e - + bek(m- p)
dt 2 dt
which can be written:
d 2p dp
- + be - + bekp = bekm (6)
dt 2 dt
This is a second-order non-homogeneous differential equation. The
particular integral is found by trying p = Z, say, a constant, so that
dp d 2p
- = 0 and-= 0 and bekZ = bekm giving Z = m
dt dt 2
Therefore in this model the equilibrium rate of inflation equals the rate
of growth of nominal money.
The complementary function has the auxiliary equation, from put-
ting p = ef1 ,
f 2 + be f + bek = 0
and this has roots which depend on b, e and k. They will be real if
~ 4bek or, since be> 0, ~ 4k. Taking an example, suppose that
k = 1 so that in (2) expectations adjust immediately to the actual rate of
inflation, the rate of monetary expansion, m = 4 and also be = 5. The
auxiliary equation is
f 2 +Sf+ 5 = 0
and the roots are f 1 = - 1.382 and f 2 = -3.618. The negative signs
indicate that the solution will converge towards the equilibrium value.
The full solution is
p = Ae-1.3s2r + Be-3.6181 + 4
If we assume the initial conditions are t = 0, p = 5, and t = 1, p = 6,
then
t= 0, p = A + B + 4 = 5
t = 1, p = 0.2511A + 0.0268B + 4 = 6
giving A = 8.8, B = -7.8 with the result that

p = 8.8e-1.3821- 7.8e-3.6181 + 4
As t increases, p rapidly approaches the equilibrium value of 4. For

example, when t = 5, p = 4.009.
To see the effect of changing k, which reflects the speed of adjust-
ment of expectations to actual inflation, suppose that k = 0.4 and
be = 5. The auxiliary equation is now
[ 2 +Sf+ 2 =0
which gives f, = -0.4384 and / 2 = -4.5615. Again the negative signs
indicate that the solution will converge on the equilibrium value. The
full solution is
p = Ae-0.43841 + Be-4.56151 + 4
and with the same initial conditions as above,
t = 0, p = A + B + 4 = 5
t = 1, p = 0.6451A + 0.01048 + 4 = 6
resulting in A = 3.1347 and B = -2.1347. The solution is
p = 3.1347e-o43841 - 2.1347e-4.56151 + 4
In this case for t = 5, p = 4.35, for t = 10, p = 4.04 and for t = 15,
p = 4.004 so that the model takes longer to reach equilibrium.
Finally, suppose that be = 3 and k = 1 making the auxiliary
equation
[ 2 + 3f + 3 = 0
with the roots - 1.5 ± 0.866i. From section 9.5 the solution will be of
the form
p = Aea 1 cos(bt- E") +4
or p = Ae- 151 cos(0.866t- E")
The initial conditions give
t = 0, p =A cos(- E") + 4 =5
t = 1, p = A (0.2231) cos (0.866 - E") + 4 =6
Since cos (- f) = cos (f) the first gives A = 1/cos (f) and substituting
into the second and simplifying,
2 = 0.2231 cos(0.866 - f)
cos( f)
Now from (9) in Appendix A, section A.2,
cos(x - y) = cos x cosy + sin x sin y
and so using this,
2 = 0.2231 [cos (0.866) cos f + sin (0.866) sin f]

cos f
where the angles are measured in radians. Since 0.866 radians is 49.6°,
= 0.2231 [0.6481 + 0.7615 tan f]
2
or tan f = 10.92 and so f = 1.48 radians. Therefore, A = 1/cos f =
11.0288. The full solution is
p = 11.0288 e-ur cos(0.866t- 1.48) + 4
The cosine term results in a cyclical pattern in the behaviour of p
but convergence towards the equilibrium value of 4 is rapid (fort = 5,
p = 3.99).
The expectations-augmented Phillips curve model can also be
analysed as a differential equation in either unemployment or in ex-
pected prices. The equations of the model are
p =a - bU - c + pe (a, b, c > 0) (1)
dpe
dt= k(p- pe) (1 ~ k > 0) (2)
dU
- = -e(m- p) (e > 0) (3)
dt
To obtain the differential equation in U, first notice that putting (1) into
(2) gives
dpe
- = k(a - bU - c) (7)
dt
and putting (1) into (3),
dU
- em + ea - eb U - ec + epe
dt
Differentiating results in
d 2U dU dpe
= -eb- + e-
dt2 dt dt
and using (7),
d2 U dU
-eb- + ek(a- bU- c)
dt 2 dt
This can be written as
d 2U dU
- + be - + bekU = (a - c)ek (8)
dt 2 dt
which has the same coefficients as (6) above, the differential equation in
p, except for the constant term. The behaviour of the solutions for U is
therefore the same as those for p, apart from the equilibrium solution
which depends on the constant term. This is to be expected since there is
no reason for the equilibrium level of unemployment (which is likely to
be affected by labour market institutions and demographic factors) and
the equilibrium level of inflation (which will depend on productivity and
the rate of increase in nominal money) to be identical. We also note that
the rate of monetary expansion, m, does not occur in (8) and so
changing m would not affect unemployment. The equilibrium solution
of (8), found by putting U = Z, say, so that dU/dt = 0 and d 2UI
dt 2 = 0, is U = (a - c)/ b which is also independent of m. This is
known as the natural rate of unemployment and is unrelated to the rate
of inflation. Thus, in the long run, there is no trade-off between unem-
ployment and inflation in this model.
Finally, to get the differential equation in pe, differentiate ( 1),
dp dU dpe
-=-b-+-
dt dt dt
and using (3),
dp dpe
dt = be (m - p) + dt
Differentiating (2),
kdp kdp"
dt dt
and substituting for dp/dt,
d2pe
- = kbe(m- p)
dt 2
and then for p from (2),
d2pe dp"
--
2
= kbem - kbep" - be
dt dt
This can be written
d2pe dp"
- - + be - + bekp" = bekm
dt 2 dt
and again the coefficients on the left hand side are the same as in (6) and
(8). The equilibrium value of p", from puttingp" = Z, a constant, is p" = m
and so expected inflation equals the rate of monetary expansion.
9.12 Exercises
1. In section 9.11 the original Phillips curve was modified to give real
wages, (w- p"), as a function of unemployment. If there is money
illusion and workers do not get full compensation for inflation, the
expectations-augmented Phillips curve can be written
p =a- bU- c + fp" (a, b, c,f> 0) (1)
where in the special case f = 1 we have the previous formulation.
Using the other two equations of the model,
dp"
- = k(p- p") (k > 0) (2)
dt
dU
- =- e(m- p) (e > 0) (3)
dt
obtain the differential equation in p and compare it with (6) of
section 9.11.
2. The market for a product is given by
dp d p 2
cf = ap - b +c- +f - (a, b, c, f > 0)
dt dt 2
qd = g- hp (g, h > 0)
where the quantity supplied is affected by both the first- and
second-order derivatives of price. Assuming
q' = qd,
(a) derive the condition for the time path of prices to be cyclical,
and
(b) if a= 10, b = 200, c = 8,/= 1, g = 1600 and h = 5, and the
initial conditions are t = 0, p = 130 and t = 0, dpl dt = -2,
determine the solution for p.

1. Solve the following differential equations:
dy dy
(a) - = 6x 2 (b)-= 4xy
dx dx
dy x + 2y dy x 2 - y2
(c) dx = (d) - =
2x + y dx xy
(e) (2x- y + 3) dx = (y - x - l)dy
(f) (x 2 + y 2 ) dx = (x- y)dy
(g) y dx + x dy =0
(h) (x + 2) dy = (4y + x 2) dx
(i) x 2 dy = (3x 2 - 4 - 3y) dx
2. If the marginal cost for an output of q is given by
MC = 25 + 3q2
determine the total cost function if TC = 1500 when q = 10.
3. The elasticity of demand for a product varies with the price (p)
and quantity demanded ( q) and is - 2p I q. Determine the demand
function if, when q = 30, p = 10.
4. An economist finds that the marginal propensity to consume is
given by MPC = 0.5 + 0.2log Ywhere Yis income. What is the
equation of the consumption function if C = 50 when Y = 100?
5. Solve the following differential equations with the given initial

conditions:
~ ~ ~
(a) - + - + 2y = 2x + 3 x = 0, y = 0, - =0
dx 2 dx dx
~ ~
(b)-- 2y =3 X= 0, y = 0, - =0
d,x2 dx
dy
X = 0, y = 0, - =0
dx
d 2y 4dy dy
( d ) - - - + 5y =1 X= 0, y = 0, - =0
d,x2 dx dx
6. For a particular economy the path of national income is given by
d 2Y 2dY
-2- - + 2Y= 2
dt dt
The initial conditions t = 0, Y = 100 and dY!dt = 5. Solve the
equation and plot the graph of Y against t. What happens to Y as
~ oo?
7. If the inflation rate in an economy is constant, what is happening
to prices?
8. In a particular country inflation is increasing according to
d 2P
-=2t
dt 2
where P = prices and t = time. If when t = 0, P = 100 and dP I dt
= 0, plot the graph of prices against time.
9. For a production process the marginal cost, MC, satisfies
dMC
--=2+q
dq
where q is output. Determine the total cost function if when
q = 0, total cost = 1000 and MC = 5. Sketch the graph of
total cost against q.
10. The market for a product is given by
q
d
= 500- p -dp
-
dt
At the equilibrium, what is the time path of prices?
Chapter 10
Difference Equations
10.1 Introduction
In Chapter 9 we saw that differential equations express the rela-
tionship between two variables (e.g. x andy) and also the rate of change
of one variable with respect to the other, (i.e. dy!dx). We know that this
rate of change (or derivative) relates to variables which change con-
tinuously. In this chapter we will be considering variables which change
in discrete steps and in particular variables which are measured at or
over certain periods of time. Examples of such variables are the level of
stocks at the end of a month, the national income earned during a
calendar year, the daily closing price of shares on the stock exchange,
and the value of a bank deposit at the end of a year.
For example, if £C is invested in a bank at an interest rate of i per
cent per year compounded annually, then the value of this investment
after one year' cl> is given by
cl = (1 + i)C
and the value after two years, C2 , is
Cz = (1 + i)Ct = (1 + iYC
In general, the value of the investment at the end of year t is
Ct = (1 + i)IC = ate where a= 1+i
Now Ct = (1 + i)Ct-1 = aCt-!

where a relates the value of the investment after t years to its value after
(t- 1) years. This expression is known as a difference equation.
It is a first-order difference equation because the time periods
involved are only one period apart. The value in any time period can be
determined if the value in the preceding time period is known.
364
Difference Equations 365
Thus C, = aC,_1
Ct+, = aC, = a 2 C,_,
C,+ 2 = aCt+, = a3 C,_,
In particular, if Co = C, then C, = a1C.
This equation is said to be the solution to the difference equation
C, = aC,_, since it enables us to determine the value of the investment at
any time period directly from its value initially.
For example, if C = £100 and the rate of interest is 5 per cent per
annum then a = 1 + 0.05 = 1.05 and C, = (1.05)' x 100 so that after 20
years, c20 = 100(1.05)20 = 265.33.
In general, the necessary requirements for a solution to a difference
equation are as follows.
1. It expresses the value of the variable in time period t in terms of
the given data, i.e. the value of the variable in the base period. It
can therefore be used to determine the value of the variable in
any time period directly without calculating all the intermediate
values as would be necessary if the original difference equation
were used.
2. It satisfies the original difference equation.
For the example used above, the second condition can be checked as
follows:
Let C, = a1C0 be the solution to the difference equation C, = aC,_,;
then ct+l = a1+ 1Co is obtained by substituting t + 1 for t whenever it
occurs in the solution. Now
Ct+ 1 = aCt = a(a'C)
0 = a1+'C0
and so C, = a'C0 satisfies the original difference equation.
10.2 Compound interest and the addition of capital at

yearly intervals
It is possible to construct a difference equation to represent the
situation in which the capital sum has interest added to it annually along
with a further injection of capital. For example, an initial capital sum C
with interest i per cent compounded annually has a further sum of A
added at the end of each year.
Then with a = 1 + i as above,
Co= C
C1 = C0 (1 + i) +A = aC0 +A
C2 = (aC0 + A)(1 + i) + A = a2 C0 + aA + A
C3 = (a 2 C0 + aA + A)(1 + i) + A = a3 C0 + a2A + aA + A
and in general,
Ct -- a1C0 + A(a + ar-z + · · · + 1)
1- 1
The first term on the right hand side of the equation is identical to
that in the previous example where A was not added at the end of each
year. The second term is a geometric progression the sum of which is
quite easily obtained (see section 4.4) and the result is
C, = a1C0 + A ( ~~ : )
This is a solution because

(a) it enables us to determine the value of the capital sum at any
time period from a knowledge of C, a and A, the three given
quantities;
(b) it satisfies the difference equation
C,+t = C,(1 + i) + A
= aC, +A
For example, let £100 be invested at an interest rate of 10% per year
compounded annually and let the capital added each year be £20.
Then C0 = 100
cl = 100(1 + o.1o) + 20
cz = 100(1 + o.10y + 20(1 + 0.10) + 20
The value after any time period can be determined by considering

the solution
C, = atCo + A ( aa -
1
-1)1
where a = 1 + i = 1 + 0.10 = 1.1
(1.1'-1)
then c, = 1.1 (100) + 20
t
1.1- 1
= 100(1.1)' + 20 ( ~ 1)
= 100(1.1)' + 200(1.1'- 1)
After 5 years, t = 5 and
C5 = 100(1.1Y + 200[(1.1) 5 - 1]
= 100(1.61) + 200(1.61- 1) = 161 + 200(0.61)
= 161 + 122 = 283
The solution in this case is made up of two terms, the second of
which arises from the presence of the additional term in the difference
equation.
There are a number of other applications of this simple difference
equation model, including any process such as an insurance policy, a
bank loan, a house mortgage or a sinking fund where regular payments
are made and the rate of interest can be assumed to be approximately
constant. The solution can always be obtained from first principles, as
above, but this is not necessary since, by classifying difference equations
in the same way as that used in Chapter 9 for differential equations,
some general solution methods can be found.
10.3 First-order linear difference equations

HOMOGENEOUS
These can be written in general
Yt+, = A.Y,
where A. is a constant and the solution is given by
Y, = k(A.)'
k is a constant whose value is determined by the initial conditions. For
example, if Y, = C when t = 0, then C = kA0 = k and the solution is
Y, = C(A.)'. As with differential equations, the solution to every first-
order difference equation contains one arbitrary constant whose value
can be determined only from other information about the system.
Notice that, fork > 0, as t increases, Y will tend to zero if -1 <A< 1,

while if A > 1, Y will tend to infinity.
NoN-HOMOGENEous
These can be written in general as
Y,+, = AY, + f(t)
or Y,+,- AY, = f(t)
The solution to such an equation is obtained in two parts. The first
is the solution of the homogeneous form Y,+, = AY, and this is obtained
as above. It is known as the transient solution. The second part is a
particular solution to the difference equation and its form depends upon
the form off( t). It is known as the equilibrium solution. The complete
solution is the sum of these two parts.
When f(t) = K, where K is a constant, the equilibrium solution is,
from section 10.2:
y = K1- A'
I 1- A
The complete solution to the difference equation
Y,+, -).Y, = K
1- A' K
is Y = Y ().)' + K - - = A).' + - -
' o 1-A 1-).
where A is a constant.
The behaviour of Y, as t changes is determined by the values of A,
K and A. The most interesting of these is A. Assuming that A > 0 and
K > 0, then if -1 <). < 1, as t increases, ).'- 0 and Y1 converges on the
value Kl( 1 -).). The difference equation is said to be stable. If). > 1 then
Y 1 will 'explode' and tend to infinity, while if ). < -1 Y 1 will become
larger and larger but will oscillate between positive and negative values.
First-order difference equations in whichf(t) is not constant will be
considered along with second-order equations in which the same condi-
tion holds.
10.4 The cobweb model

In section 1. 7, we discussed the situation in which the quantity
demanded and supplied of a commodity are equated under conditions of
perfect competition. Thus,

qD = f(p) demand function
qs = J(p) supply function
and
But in many cases the response to changes in demand is not instan-
taneous. Thus a price increase does not immediately result in an in-
creased supply. This is particularly so of such things as agricultural
products, the supply of which cannot be altered to any great extent in a
short time. The rate of change in the quantity which can be supplied is
determined by the time cycle of growth or production of the good. Once
a crop is planted its size is governed by weather conditions etc., and an
increase in price during the growth and harvesting period has little effect
on this year's supply. Similarly industries which are working at full
capacity cannot increase the supply without putting in hand a capital-
investment programme which is likely to take some time to implement.
This problem can, of course, be overcome to some extent by allowing
goods to be transferred into and out of inventory, i.e. by having a buffer
stock in which fluctuations can take place instantaneously. If, initially,
this possibility is ignored it seems reasonable to assume that a change in
the price which can be obtained for a good has a delayed effect upon the
quantity of the good supplied.
The simplest case we consider is one in which the delay, or lag is for
a single time period and applies to the supply function only. This can be
written as follows:
q, = fv(p,) demand function
q, = !s<Pr-I) supply function
This model states that the price in any time period determines the
quantity supplied in the subsequent time period and this latter quantity
when considered in relation to the demand function determines the
price in that period. This price then determines the quantity supplied in
the next period and so on.
If the price in two successive periods is different, then the quantity
supplied in these two periods is different, and it is not difficult to
visualise a situation where the price and the supply continue to change
from period to period. If this occurs it is of interest to see whether there
is an equilibrium position for the system, and to ask how the price and
p
Supply function
Fig. 10.1
quantity supplied change over a period of time. The situation can be

represented graphically and the time path of the system determined.
An example where the demand and supply functions are both linear
is illustrated in Fig. 10.1. Given that the system starts with a price p 1 in
the initial time period, the supply function enables us to determine the
quantity q2 which will be supplied at this price in the next time period.
But if q2 is supplied in the period 2 it can demand a price p 2 which is
considerably higher than that in the preceding period. This figure can be
found on the graph by moving vertically from the supply function to the
demand function.
This high price results in an increase in the quantity supplied in the
following time period. In this case it is equal to q3 • This figure can be
found on the graph by moving horizontally from the demand function to
the supply function.
From this point onwards, the procedure is to repeat the above
steps, i.e. moving alternately from the supply to the demand function in
vertical steps and from the demand to the supply function in horizontal
steps. The result is that the changes in price and quantity between
successive time periods become smaller and smaller and the system
approaches an equilibrium value which is located at the intersection of
the demand and supply functions.
p,
P•
Po
---------
~ ~
- - - - 1 - - f---

Ps
Pl ~ ~~ --1---
p, --------- ---f---
q 2 3 4 5 6
Fig. 10.2
q 2 34 56 7
Fig. 10.3
It is interesting to show the changes in price which occur by trans-

ferring the information to a graph where time is represented on the
horizontal axis (Fig. 10.2). The price fluctuates between time periods,
the size of the fluctuations decreasing with time. The system tends to an
equilibrium value and is, therefore, said to be stable or converging.
Fig. 10.3 illustrates a situation in which the size of the fluctuations
increases with time. This system is said to be unstable, diverging or
explosive, and if it remained unchanged (which is highly unlikely) it
would oscillate violently and the price of the good would fluctuate wildly
from year to year.
A third situation is possible in which the price changes between
successive time periods but alternates in such a way that it assumes
either one or other of only two values, i.e. the system oscillates
(Fig. 10.4).
p
q 2 34 5678
Fig. 10.4
Any linear system conforms to one of these three types. The

particular type is determined by the slopes of the demand and supply
functions and it is possible to decide into which category a system falls
by following the above procedure, i.e. by constructing the time path of
price from the demand and supply functions. But this is not really
necessary in the linear case, because the same information can be
obtained from the equations of these functions.
For example, let
qt = fo(Pt) = at + btPt
and qt = fs(Pt-1) = az + bzPt-t
Then if we assume that the quantity supplied equals the quantity
demanded in any time period,
that is
This is a first-order difference equation
btPt = hzPt-t + az - at
bz ] az - at
or Pt = [ b; Pt-t + -b-1 -
The solution to the homogeneous form
b2 b2)
= k (b;
1
Pt= j;Pt-t is given by Pt
1
(see section 10.3) and it can be shown that the particular solution is
(a2 - a1)1b 1 a2 a
=b - -1 -
p=
t 1 - (b 2/b 1) b 1 - 2
The complete solution is therefore,
p, =k [: 2
I
]' + : 2I =~~ 2
where k is an arbitrary constant, the value of which is determined from

the initial conditions.
The particular solution is a constant and has no influence on the
changes which occur in the time path of price. These are determined by
the solution to the homogeneous form. The manner in which the price
changes between successive time periods is determined by the term
bzfb 1 , where b 1 is the slope of the demand curve and is a function of the
elasticity of demand which can be assumed, in general, to be negative.
Similarly b 2 is a function of the elasticity of supply which is positive. The
term bzfb 1 must, therefore, be negative, (b/b 1 ) 2 is positive and (b/b 1) 3 is
negative and so on. Its value is alternately positive and negative in each
successive time period. It follows from the fact that the particular
solution is constant that the function will oscillate about the value
p, = (a 2 - a 1)/(b 1 - b2 ) and the size of the oscillation depends upon the
ratio b2 /b 1 • Three distinct cases arise.
1. lb2/b 1 l < 1, or lbzl < lb.l, that is the modulus or absolute value of
b 2 is less than that of b 1 • In this case (b 2 /b 1 ) ' decreases in value as t
increases and the size of the oscillations decreases over time. This
corresponds to the stable system and the equilibrium price is (a 2 - a 1)/
(b 1 - b 2 ).
2. lbzib.l = 1, that is lbzl = lbJ In this case (b/b 1)' has an absolute
value of unity but is alternately positive and negative as t changes. This
corresponds to the situation in which the price takes on only two values
and
3. lb/b.l > 1, that is lbzl > lbJ

In this case (b 2/b 1)' increases in absolute value as t increases, chang-
ing sign between successive time periods. This corresponds to the un-
stable system because the size of the oscillations becomes larger and
larger.
It is apparent, from this analysis, that the stability or otherwise of
any demand-supply system depends upon the relative slopes of the two
functions. In most practical situations these functions are not linear and
therefore the slopes will be functions of price. In such cases it is possible
to visualise a system which is unstable over one price range but stable
for changes which are outside this price range. This can be demon-
strated by the graphical approach and the result confirmed algebraically.
10.5 The Harrod-Domar growth model

This assumes that
1. The amount of resources saved by the community during any
time period is a function only of the income received by the
community during that period.
That is S, = aY,
where S, represents savings in time period t
Y, represents income in time period t
a is a positive constant which is less than 1
2. The amount of investment which occurs in any time period will
be a function of the increase of income in that period over the
income of the previous time period.
That is
where g is a constant which is less than 1.
3. The amount which will be available for investment in any time
period will be equal to the amount saved in the same time period.
That is I, = S,
Using these three assumptions it is possible to construct a model of
the system as follows.
S, = aY, (1)
I, = g(Y,- Y,_ 1) (2)
I,= S, (3)
If (1) and (2) are substituted into (3) a first-order linear difference
equation is formed.
g(Y,- Y,_ 1) = aY,
Y,(g- a) = gY,_ 1
Y, = (-g
g-a
)Y,_l
The solution to this is
y = k(-g-)t = k).t
t g- a
where k is equal to the value of income in time period t = 0
Since Y0 = k (-g
g-a
) 0
= k
then y
t
= y
o
(-g)t
g _a
From assumptions 1 and 2, a and g must both be positive and therefore

the sign of Adepends on whether g is greater than or less than a. If g = a
then A is infinitely large and the model breaks down.
10.6 A consumption model

A further example of a simple economic model is provided by the
equations
Ct = aYt-! +B (1)
Yt = ct +It (2)
where Ct = consumption expenditure
Yt = national income
It = investment expenditure
If It is constant in each year - for example, It = F- then substituting
from (2) into (1) gives
Ct = aCt-! + (aF + B)
which is a first-order non-homogeneous difference equation. The solu-
tion of the homogeneous part is
ct =kat
and the equilibrium solution is
C = (aF +B)
t 1- a
so that the complete solution is
Ct = kd ~ _+aB)
10.7 Exercises
1. Solve the following difference equations and sketch the time path
of Y, fort= 0 tot= 5.
(a) Y,+ 1 = 1.5 Y, + 3 with Y0 = 2
(b) Y,+ 1 = 0.9 Y, + 2 with Y0 = 3
(c) Yr+ 1 = Y, with Y0 = 6
(d) Yr+ 1 = 1.1 Y, + 4 with Y0 = 1
2. A savings club collects £15 from each of its members at the begin-
ning of each year. Assuming that the savings increase at 5% per
annum compound interest, express the relationship between the
values of the savings at the beginning of year t + 1 and year t in
difference-equation form and hence show that the value of the
savings after 20 years is £535.8.
demand: qr = 4- Pr
supply: qr = 1 + 1.5 Pt-1
with p 0 = 1 lead to an 'exploding' cobweb model.
demand: qt = 4- Pr
supply: qt = 1 + 0.5 Pr- 1
with Po = 1lead to a stable cobweb model. What is the equilibrium
price?
demand: qr = 8-3 Pr
supply: qr = 3 Pr-1
with p 0 = 2 lead to an oscillating cobweb model.
6. Solve the Harrod-Domar growth model (see section 10.5) when
a = 0.9, g = 0.3 and Y0 = 100.
7. A similar model to the cobweb arises when there are static supply
and demand equations but the new price is affected by the level of
inventory (the difference between q' and qd) in the previous
period. For example,
demand: q1 = 150 - 0.6 Pr
supply: q: = 0.4 p, - 50
pricing: Pt+t = p,- 0.3(q:- q1)
Find the solution for p, and comment on whether it converges.
8. Obtain the solution of the following simple accelerator model:
Y, = C, +I,
I, = a(Y,- Y,_ 1) (a > 0)
C, = b¥,_1 +C (b, C > 0)
What is the condition for it to be stable?
10.8 Samuelson's multiplier-accelerator model

This assumes that
1. Consumer demand in any time period is a function of consumer
income in the previous time period.
That is C, = c¥1_ 1
where c is a constant which is less than one and is the marginal
propensity to consume.
2. Investment in any time period is a function of the difference in
income between the two previous time periods.
I, = b(Yt-t - Y,_ 2)
where b is the acceleration coefficient, which is a constant.
3. The national output, or income, Y,, in any time period is made
up of the estimated consumer demand and the estimated invest-
ment demand in the same period.
Y, = C, +I,
Based on these assumptions we have the model
C, = c¥1_ 1 (1)
I, = b(Y,_I- Y,_2) (2)
Y, = C, +I, (3)
Substituting (1) and (2) into (3) leads to
Y, = c¥,_1 + b(Y,_ 1 - Y,_2 )
= (c + b)Y,_t- b¥,_2
This is a second-order difference equation because it relates the

value of income in time period t to the value of income in the two
immediately preceding time periods. It is possible to determine the
value of income in any time period if we know the value of income in
any two adjacent periods and the value of the constants band c.
For example, let us suppose that
b = 0.6 c = 0.9 Y 0 = 100 Y1 = 120
Then Y2 = (c + b)Y1 - bY0 = 1.5(120)- 0.6(100)
= 180 - 60 = 120
y3 = 1.5(120) - 0.6(120) = 180- 72 = 108
The value of income in each successive time period can be calcu-
lated by continuing the above procedure. But this is not necessary,
because it is possible to obtain a solution to a second-order difference
equation and to use this solution to determine Yt in any time period as
we did in the first-order case.
For convenience the equations and their methods of solution are
again discussed under appropriate headings.
10.9 Second-order linear difference equations

HoMOGENEous
These can be written in general as aYt+z + b Yt+ 1 + cYt = 0 and the
solution is of the form Yt = )..' where ).. is a constant whose value is as yet
undetermined.
If Yt = A.t, then Yt+ 1 = A.t+ 1 and Yt+z = A.t+ 2 • Substituting these values
in the difference equation, we obtain
aA.t+ 2 + b)..t+t + cA.t = 0
A.t(aA.2 + b).. + c) = 0
Therefore
).. = - b + V (b2 - 4ac)
1 2a
- b- Y(b - 4ac) 2
Az = ---'-----'-
2a
and the solution is of the form
Yt = ~ + kzA.;
where k 1 and k 2 are arbitrary constants.
The equation a)..Z + b). + c = 0 is known as the characteristic

equation and is easily formed by substituting A2 for Yt+ 2 , Afor Yt+ 1 and 1
for Y, in the difference equation. When the roots of the characteristic
equation are equal, that is A1 = A2 = A, the solution is
Y, = (k 1 + k 2 t)A1
This contains two arbitrary constants, which is a necessary require-
ment for it to be the solution to a second-order difference equation
(compare differential equations, section 9.5).
When the roots of the characteristic equation are complex, that is
A1 = a + ib and A2 = a - ib, the solution is
Y, = k 1(a + ib)' + k 2(a- ib)'
The solution is not very useful in this form because it is not possible
to analyse the future time path of Y from a combination of real and
imaginary numbers. But the solution can be transformed into a more
suitable form by making the following substitutions.
Let a=rcose and b = r sine
Then
= r 2 ( COS2 e + sin2 e) = r 2
or r = v'(a 2 + b2 )
and e = cos- a or sm--b
1 -
r
• I
r
These can be combined to give
tan e = b/a or e = tan- 1 (b!a)
Here e is measured in radians (see Appendix A, section A.3). Now we
know from the Euler relations (see section 5.16 that for all values of e,
= cos e + i sin
ei8 e
t
Now (ei 9 = eina = cos ne + i sin ne.
It can be shown, therefore, that
(cos e + i sin e)n = cos ne + i sin ne
This is usually referred to as De Moivre's theorem and it holds for all
values of e.
Making use of the substitutions a = r cos e and b = r sin e
(a + ib) = r cos e + ir sin e = r(cos e + i sin e)
(a + ibY = [r(cos () + i sin 8)]'

= r 1(cos () + i sin 8Y
= r 1( cos t8 + i sin t8)
Similarly, from the properties of complex numbers it can be shown
that (a- ib) 1 = r 1(cos t8- i sin t8)
The solution to the difference equation which has complex roots to
the characteristic equation can now be expressed in terms of trig-
onometrical functions:
Y, = k (a + ib Y + k (a- ib Y
1 2
becomes Y, = k1[r 1( cos t8 + i sin t8)] + k2 [r 1( cos t8 - i sin t8)]

The right hand side of this equation can be arranged in a more
convenient form by collecting together the sine and cosine terms:
Y, = r [(k + k cos t8 + i(k
1
1 2) 1 - k 2 ) sin t8]
= r (k cos t8 + k sin t8)
1
3 4
where k 3 = k 1 + k 2 , k 4 = i(k 1 - k 2 ), which are both constants.

This result can be compared to the solution obtained for a second-
order differential equation.
y = ff'X(k 3 cos bx + k 4 sin bx)
It was shown in Chapter 9 that it is possible, and often more useful,
to express this in the form y = Ae= cos (bx- f). In a similar way it can
be shown, although the mathematics are not discussed here, that the
solution to a second-order difference equation can be expressed in the
alternative form
Y, = Ar 1 cos (t8 - f)
where A and f are the two arbitrary constants, whose values are deter-
mined in a particular case from two pieces of information which are
available about the system. The behaviour of the solution depends
mainly on the value of r, since a cosine function is always cyclical. If
r > 1 or r < -1 then Y will tend to oscillate between ± oo. Otherwise Y
will tend to zero.
NoN-HoMOGENEous
This has the general form
aYt+2 + bYt+, + cYr = f(t)
It can be shown that the solution to this type of equation is the sum
of two parts which must be obtained independently. These are usually
called the transient function and the equilibrium, or particular solution.
The transient function is simply the solution of the homogeneous
part of the equation formed by lettingf(t) = 0 and this can be obtained
by the method of the previous section.
The particular solution is any solution which satisfies the difference
equation and its form depends upon the form of f(t).
1. f(t) is a constant. For example, if aYt+ 2 + bYt+l + cYr = K, we
try a constant as the particular solution, that is Yr = Z for all t and it
follows that Yt+z = Yt+, = Yr = Z.
Substituting these values in the difference equation, we have
aZ + bZ + cZ =K
That is Z(a + b + c) =K or Z =a +~+ c
It will be found that this value of Z satisfies the difference equation

and is therefore a particular solution for all cases except those in which
a+b+c=O
In such cases we must try the solution
Yr = Zt
so that Yt+, = Z(t + 1) = Zt + Z
Yt+ 2 = Z(t + 2) = Zt + 2Z
Notice that Zt is the product of Z and t. Substituting these values in the
difference equation gives
aZ(t + 2) + bZ (t + 1) + cZt = K
or aZt + 2aZ + bZt + bZ + cZt = K
Z(a + b + c)t + Z(b + 2a) = K
If this equation is to be valid for all values of t, it is necessary that
both of the following conditions hold:
Z(a + b + c) =0 and Z(b + 2a) =K
The second equation gives a value for Z:

K Kt
Z = - - and hence Yc = Zt = - -
b+2a b+2a
provided that (b + 2a) is not equal to zero.
2. f(t) is not a constant
(a) f(t) is a linear function. For example,
aYt+z + bYt+t + cYr = K 1 + K 2t
The particular solution is of the form
Yr = Zo + Z 1t
so that Yt+t = Zo + Z (t + 1)
1
and Yt+z = Zo + Z 1(t + 2)

These values are substituted in the difference equation:
a[Zo + Z 1(t + 2)] + b[Zo + Z (t + 1)] + c[Zo +
1 Z 1t] = K 1 + Kzt
This can be rearranged with the constant terms and the terms in t
collected separately:
(aZ0 + 2aZ + bZ + bZ + cZ0 ) + (aZ + bZ + cZ )t
1 0 1 1 1 1
that is (a + b + c)Z0 + (2a + b)Z + (a + b + c)Z t 1 1
If this equation is to be true for all values oft, the following two
conditions must hold:
(a + b + c)Z0 + + b)Z1 =
(2a K1 (1)
(a + b + c)Zt = Kz (2)
Kz
From (2) zl = ----
a+b+c
and from (1) and (2)
K1 - (2a + b)Z1
Zo = - - - - - -
a+b+c
Kt (2a +b) Kz
=
a+b+c (a + b + c) (a + b + c)
Kt (2a + b)K2
=
a+b+c (a+ b + c) 2
The particular solution can now be expressed in terms of the known

constants a, b, c, K 1 and K 2 •
¥1 Z0 +Z t
1
K1 (2a + b )K2 K2t

= - + --"'---
a+b +c (a + b + c) 2 a+b +c
= 1 [ K1 - K 2 (2a+b - t )]
a+b+c a+b+c
(b) f(t) is of the form K 1 • For example,
a¥1+ 2 + b¥1+ 1 + c¥1 =K 1

Y, = ZK 1
so that yt+l = ZKt+l

and ¥,+2 = ZKt+2
These values are substituted in the difference equations
aZK 1+2 + bZK 1+1 + cZK 1 = K 1
K (aK 2 + bK + c)Z = K
1 1
1
Z=------
aK2 + bK + c
and the particular solution is
Kr
= ZK = - -
Y --- 1
1aK 2 + bK + c
provided that aK 2 + bK + cis not equal to zero.
If this condition is not satisfied then we must try a particular
solution of the form
¥1 = ZtK 1
If this results in a particular solution with a denominator equal to
zero then it is necessary to try the substitution
Yr = Zt2Kt
This procedure can be continued until a solution is obtained.
(c) f(t) is a trigonometrical function. For example,

aYt+ 2 + bYt+ 1 + cYt = K sin t
Yt = Zo sin t
+ Z 1 cos t
so that Yt+ = Zo sin (t + 1) + Z 1 cos (t + 1)
1
and Yt+2 Z 0 sin (t + 2) + Z 1 cos (t + 2)

The values of Z 0 and Z 1 can be found by substituting these values
in the difference equation and by making use of the following trig-
onometrical identities.
sin (A + B) = sin A cos B + cos A sin B
cos (A + B) = cos A cos B - sin A sin B
The exercise, which is rather more lengthy than the previous ones, is left
for the reader to attempt.
10.10 A consumption-investment model

The following example shows how the above methods are applied
in the analysis of the time path of national income Yt given the assump-
tions that the consumption and investment functions are
ct = o.s Yt-1 + 0.19 Yt-2 (1)
It = 10 + 50 (1.05)1 (2)
Yt = ct +It (3)
Y0 = 100 and Y1 = 120 (4)
A second-order difference equation is obtained by substituting (1)
and (2) into (3).
yt = 0.5 yt-1 + 0.19 yt-2 + 10 + 50(1.05)1
or yt- 0.5 yt-1 - 0.19 yt-2 = 10 + 50(1.05)t
The solution to the reduced form is obtained from the characteristic
equation
,F - 0.5A. - 0.19 =0
the roots of which are

0.5 ± V[(0.5)2 + 4 X 0.19] 0.5 ± 1
A.= =---
2 2
A1 = 0.75 A2 = -0.25
and the transient solution is
Y, = k 1(0. 75)' + k2 (- 0.25)'
The particular solution can be found by applying the methods of
section 10.9 to the two terms on the right hand side in turn. For
Y,- 0.5 Y,_ 1 - 0.19 Y,_ 2 = 10
the particular solution is obtained by trying Y, = z to give
y = { 1 } 10 = ~ = 32.26
t 1 - 0.5 - 0.19 0.31
For the equation
Y,- 0.5 Y,_ 1 - 0.19 Y,_ 2 = 50(1.05)'
the particular solution is obtained by trying Y, = z(l.OS)' to give
y =50 { 1.052 } 1.05 t

I 1.052 - 0.5 (1.05) - 0.19 ( )
Hence
Y, = 50 {-l.OSZ
-} (1.05)' = 142.26(1.05)'
0.3875
The solution can then be written as
Y, = k 1(0. 75)' + k2 (- 0.25)' + 32.26 + 142.26(1.05)'
and the value of the constants, k 1 and k 2 , found from the initial
conditions.
Yo= 100
100 = kl + k2 + 32.26 + 142.26 (1)
yl = 120
120 = 0.75kl - 0.25k2 + 32.26 + 142.26(1.05) (2)
32.261---------------------
Fig. 10.5
From (1)
kl + k2 = 100- 174.52 = -74.52
and from (2)
0.15kl - 0.25k2 = 120 - 32.26 - 149.37 = -61.63
These two equations can be solved simultaneously with the result
k1 = -80.26 and k 2 = 5.74
The complete solution to the difference equation
Y, - 0.5 Y,_ 1 - 0.19Y,_ 2 = 10 + 50(1.05)'
and the initial conditions Y 0 = 100, Y 1 = 120 is given by
Y, = - 80.26(0.75)' + 5.74( -0.25)' + 32.26 + 142.26(1.05)'
It is easy to check that this solution satisfies the initial conditions,
but not as easy to show that it satisfies the difference equation.
The general shape of the time path of national income can be
obtained by considering the function in terms of the four constituent
parts. The graphs of these can be sketched as shown in Fig. 10.5, where
the term 142.26(1.05)' becomes predominant with time.
10.11 An inflation-unemployment model

In section 9.11 a continuous time model of the interactions between
inflation, unemployment and monetary policy was examined based on
the expectations-augmented Phillips curve. Here we consider the dis-
crete time version of the same model. Adding time subscripts, the
expectations-augmented Phillips curve can be written as
p, = a - bU, - c + ~ (a, b, c, > 0) (1)
and the simple adaptive expectations equation is
~ - ~ = k(p,- ~ (k > 0) (2)
This states that the change in the expectation is a proportion, k, of the
error in expectation in the current period, or alternatively, next period's
expected inflation is a weighted average of this period's actual and
expected inflation, since
~ = kp, + (1 - ~
Finally, we assume that the change in unemployment is related to the
rate of growth of real money, or
Ut+i - U, = - e(m - Pt+t) (e > 0) (3)
where m is the nominal rate of growth of money and p is the rate of
inflation. We will also assume that m is a constant.
The three equations (1}-(3) explain p, pe and U. Moving (1) for-
ward one period gives
Pt+t = a - bU,+i - c + ~
and subtracting (1) from this results in
Pt+t - Pt = - b( Ut+t - U,) + ~ - ~
Substituting from (2) and (3) into this gives

Pt+t - Pt = be(m - Pt+i) + k(p, - pn
or (1 + be)p,+ = (1 + k)p, + bem -
1 ~ (4)
This includes pe which can be obtained from (1),
~
and substituting this into (4),

(1 + be)pt+ 1 = p, + bem + ka - bkU, - ck
Finally to eliminate U we first-difference this equation

(1 + be)(Pt+1 - p,) = (p, - Pt-1) - bk(U, - U,_1)
where the difference in m is zero, since m (the growth of nominal
money) is assumed to be a constant. Next, we use (3) to eliminate U,
(1 + be)(Pt+1 - p,) = (p, - Pt-1) + bek(m - p,)
Rearranging,
(1 + be)pt+ 1 - (1 + be + 1 - bek)p, + Pt- 1 = bekm
This is a second-order non-homogeneous difference equation which can
be put into the more usual form by shifting it forward one period and
simplifying the notation:
Ap,+2 + Bpt+1 + p, = C (5)
where A = (1 + be), B = - (2 + be[1 - k]), C = bekm. The
characteristic equation is
A).? + B). + 1 = 0
and, since the roots depend on the values of the parameters, it is difficult
to analyse the behaviour in the general case. Before looking at some
examples we notice that the particular or equilibrium solution, found by
putting Pt+2 = Pt+t = p, = p is
c
p= A+B+1=m
and so the equilibrium rate of inflation is m, the rate of monetary
expansion. This exists for any (non-infinite) values of the parameters.
Taking some examples, first suppose that k = 1, so that, from (2),
~ = p,, and the expected rate of inflation next period is this period's
actual rate. Here (5) reduces to
(1 + be)pt+2 - 2p,+ 1 + p, = bem
for which the characteristic equation is
(1 + be)).2 - 2). +1= 0
and the roots are real if
4 > 4 (1 +be)
which requires either b < 0 or e < 0. Since these are assumed to be

positive then when k = 1 the difference equation has complex roots and
the solution is expressed in terms of trigonometrical functions (as in
section 10.9 above). The time path of inflation (p) will therefore follow
a cycle and the value will fluctuate.
For example, if be = 5 the characteristic equation is
6)}-2)..+1=0
and the roots are).. = (1 ±V5i)/6
so that a = 1/6, b = V5/6 and r = Y(a 2 + b2 ) = 0.4082. Also,
(} = tan- 1 (bla) = tan- 1 V5 = 65.9° = 1.15 radians.
Therefore, Pt = Ert cos (tO- c) = E(0.4082t) cos (1.15t- c) where E
and c are constants. The equilibrium solution, from above, is p = m and
so the full solution is
Pt = E(0.4082t) cos (1.15t- c) + m
For simplicity we will assume that m = 0. The values of E and c are
given by the initial conditions. Let these be t = 0, p 0 = 0 and t = 1,
Pt = 1. The first gives
0 = Ecos (-c)
and so either E = 0 (which is ignored because it gives Pt = m for all t) or
c = ;r/2 = 1.5708.
The second gives
1 = E(0.4082) cos (1.15 - c)
and so E = 11[0.4082 cos (- 0.4208)]
Since 0.4208 radians= 24.11° then E = 2.684.
The full solution is
Pt = 2.684(0.4082t) cos (1.15t- 1.5708)
The first few values are Po = 0, p 1 = 1, p 2 = 0.33, p 3 = -0.06,
p 4 = -0.07, Ps = -0.03 and so there is a rapid cyclical convergence
towards the'equilibrium value of zero.
Next, suppose that k = 0.4, so that expected inflation gradually
adjusts to actual inflation, and also that be = 5. The characteristic
equation of (5) is now
6)..2 - 5). +1=0
and the roots are A1 = 0.5 and A2 = 0.33. Therefore,

Pt = F(OS) + G(0.33t)
is the transient solution. As t ~ oo, this will tend to zero and so inflation
will converge on its equilibrium value.
Finally, suppose that k = 0.4 and be = 2 so that the characteristic
equation of (5) is
3A. 2 - 3.2A. + 1= 0
and the roots are complex. As in the first case above the time path of
inflation will follow a cycle.
The expectations-augmented Phillips curve model can also be
analysed as a difference equation in unemployment or expected infla-
tion. For unemployment, the equations of the model are
Pt = a - bUt - c + ~ (a, b, c > 0) (1)
P:+t - ~ = k(pt - ~ (k > 0) (2)
U,+t - Ut = - e(m - Pt+t) (e > 0) (3)
From (1), ~ = p, - a + bUt + c
Substituting for pe in (2),
(Pt+t - a + bUt+t + c) - (Pt - a + bUt + c)
kpt -k(pt- a+ bUt+ c)
or Pt+t- p, bUt - bUt+t + ak - bkUt - ck (6)
Now differencing (3) and then using (6),
Ut+t - 2Ut + U,_ 1 = e(pt+t - Pt)
= e[b(1 - k)Ut - bUI+ 1 + ak - ck]
or (1 + be) ut+l - [2 + be(1 - k)] ut + ut-l = (a - c)ek
This can be shifted forward one period to give
AUI+Z + BUI+l + ut = C' (7)
where A = (1 + be), B =- [2 + be(1 - k)] and C' = (a - c)ek
Comparing this equation with (5) above we see that the coefficients
on the left hand side of the difference equation are the same in each case
but the constant terms differ. Thus the solutions to the homogeneous
equations, that is the transient solutions, will be the same- so that if pis
cyclical then U will also be cyclical - but because the constant terms
differ the equilibrium solutions will differ. This is to be expected since
there is no reason for the equilibrium level of unemployment (which is
likely to be affected by labour market institutions and demographic
factors) and the equilibrium level of inflation (which will depend on
productivity and the rate of increase in nominal money) to be identical.
We also note that the rate of monetary expansion, m, does not occur in
(7) and so changing m would not affect unemployment. Finally, the
equilibrium solution of (7), found by putting U1 = U1+ 1 = Ut+ 2 = Z, say,
is U = (a - c)/b which is also independent of m. This is known as the
natural rate of unemployment and is unrelated to the rate of inflation.
Thus, in the long run, there is no trade-off between unemployment and
inflation in this model.
A difference equation in pecan also be found. Differencing (1),
Pt- P1-1 = -b( ut- ul_,) + ~ p'f-,
= be(m - Pt) + ~ P1-,
from using (3). Now (2) can be written
kpt = P1+, - (1 - ~
and so substituting for p in the previous equation and simplifying gives

(1 + ~ [2 + be(l- k)] ~ + ~ = bekm
which has the same coefficients as (5) and (7), and the constant term is
the same as in (5). Therefore in equilibrium, pe = m and the behaviour
is as above.
With all mathematical models it is important to realise that the
properties follow from the assumptions and that changing the assump-
tions, in general, changes the behaviour of the model. The challenge is
to find simple models which can make a useful contribution to our
understanding of the real world.
10.12 Exercises
1. Show that Samuelson's multiplier-accelerator model leads to a
characteristic equation with
(a) complex roots when c = 0.9, b = 0.5
(b) repeated roots when c = 0.96, b = 0.64
(c) real roots when c = 0.9, b = 0.3
2. Solve the following difference equations and sketch the time path
of Y, fort= 0 tot= 5 for (a)-(e).
(a) Yr+z- Y,+l- 2Y, = 3 with Yo= 2 yl = 2
(b) Yt+z- 3Yt+ 1 + 2Y, = 4 with Y0 = 1 yl = 2
(c) Yr+z- 4Yr+ 1 + 4Y, = 3' with Y0 = 3 yl = 4
(d) Yr+z- 3Y,+ 1 + 2Y, = 3(2') with Y0 = 1 yl = 5
(e) 4Y,+ 2 + 4Yt+ + Y, = 2 + 3t
1 with yl = 1 y2 = 2
(f) Yr+z + 4Yt+ + 5Y, = 4
1
(g) yt+2 - 4 yt+l + 8Y, = t.
3. For the equations

Y, = C, +I, (1)
C, = aY, + f3Y1-1 + {J (2)
I, = y(Y,_t - Y,_z) (3)
show that substitution from (2) and (3) into (1) produces a second-
order difference equation which, if {J = 0, is similar to the equation
from Samuelson's multiplier-accelerator model (Section 10.8).
Solve the equation when
(a) a = 0.6, f3 = 0.2, y = 0.3 and {J = 1
(b) a = 0.5, f3 = 0.2, y = 0.2 and {J = 1
4. In section 10.11 the original Phillips curve was modified to give real
wages, ( w - pe), as a function of unemployment. If there is money
illusion and workers do not get full compensation for inflation, the
expectations-augmented Phillips curve can be written
p, = a - bU,- c + ~ (a, b, c, f > 0) (1)
where in the special case f = 1 we have the previous formulation.
Using the other two equations of the model,
P:+t- ~ = k(p,- ~ (k > 0) (2)
Ut+ 1 - U, = -e(m - Pr+ 1) (e > 0) (3)
obtain the difference equation in p and compare it with (5) of
section 10 .11.
5. One variation on the multiplier-accelerator model of section 10.8 is

obtained by introducing government expenditure, G, which might
be determined by the previous period's income and the previous
period's government expenditure. Suppose that the equations of
the model are
Y, = C, + I, + G,
C, = aY,_ 1
I, = b(Y, - Y,_1)
G, = cY,_1 + dG,_l
Obtain the difference equation in Y and show that (a) if d = 0 it
reduces to a first-order equation, and (b) if a= 0.6, b = 2, c = 0.1
and d = 0.5 and the initial conditions are Y0 = 1, Y1 = 1, the value
of Y will eventually approach infinity.

1. Solve the following difference equations and plot the time path of
Y, for t = 0 to t = 5:
(a) Y, = 0.5 Y,_ 1 + 150 with Y0 = 100
(b) Y, = 1.3 Y,_1 - 75 with Yo = 100
(c) Y, = Y,_ + 1 with
1 Y0 = 1.
2. Examine the behaviour of the following equations as t increases:
demand : q, = 100- ap,
supply : q, = bp,- 3
if (a) a = 2, b = 3 (b) a = 2, b = 2 (c) a = 3, b = 2.
3. The value of sales, S, of a product is found to satisfy
S, = 0.8 S,_l + SA,
where A is the expenditure on advertising. The company deter-
mines its advertising budget according to the rule
A,= 0.151-1
If 50 = 50, determine the value of sales for t = 1 to t = 5.
4. Solve the following difference equations and explain how the time
path of Y, will vary as t tends to infinity:
(a) Y,+ 2 + 2Y,+ 1 = 4 with Y0 = 1,

(b) Yt+ 2 + 2Y,+ 1 - 0.5Y, = 5 with Yo= 5, Y 1 = 6,
(c) Y,+ 2 - 2Y,+ 1 + 0.5Y, = 3 with Y0 = 3, Y 1 = 6,
(d) Yt+ 2 - 4Y,+ 1 + 6Y, = t with Y0 = 1, Y, = 2,
(e) Y,+ 2 - 2Y,+ 1 + Y, = 5 + 2t with Y0 = 2, Y 1 = 3.
5. Examine the behaviour of the equilibrium solution of the following
supply and demand equations:
supply : q, = 6p,- 5 + 4(p,_l - Pt-2)
demand: q, = 500- 2p,.
6. Given the simple national-income model
Y, = C, + I, + G,
C, = 0.7Y,_, +2
I, = 2(Y,_,- Y,_2)
G, = 0.2Y,_,
and the initial conditions Y0 = 100, Y, = 102, examine the time
path of Y, for t = 2 to t = 6.
7. A university gives resources (R) to each department according to
the formula
R, = 0.5(R,_ 1 + R,_2) + 0.2S,
where S measures new student registrations. It is also known that
S, = 1.2S,_1 with S1 = 10. By first obtaining the solution for S,
determine how R moves given the initial conditions that R 0 = 0,
R 1 = 10.
Appendix A
Trigonometric Functions
c
a
Fig. A.l
A.l Definitions
Let us consider a right-angled triangle ABC (Fig. A.l). If the side
BC is of length a, AC of length b, AB of length c, andx is the angle BAC
then we define
sine x = sin x = f!c (1)
.
cosme x = cos x = -bc (2)
tangent x = tan x = ~ (3)
Values of these functions can be obtained from tables.

For a right-angled triangle,
a2+b2=CZ
395
396 Appendix A
or, dividing by c?,
That is
(sin xY + (cos xY = 1
or sin 2 x + cos2 x = 1 (4)
Therefore, sin 2 x = 1 - cos 2 x
cos2 x = 1 - sin 2 x
a ale sin x
Notice that tanx = - = - = - -
b b!c cosx
From (4), by dividing by cos2 x,
sin 2 x cos 2 x 1
--+-- = -2-
2
COS2 x cos x cos x
or, tan 2 x + 1 = sec2 x (5)
where sec x = secant x = 1/cos x
Two other trigonometric functions are the cosecant and cotangent
defined by
1
cosecant x = cosec x = - -
sin x
1 COS X
and cotangent x = cotan x = - - = -.--
tan X SID X
A.2 Compound angles

For any two angles x andy the following relationships can be shown
to be true:
sin (x + y) sin x cos y + cos x sin y (6)
cos (x + y) = cos x cos y - sin x sin y (7)
sin (x - y) sin x cosy- cos x sin y (8)
cos (x - y) cos x cosy + sin x sin y (9)
Trigonometric Functions 397
tanx+tany
tan (x + y) (10)
1-tanxtany
tan x- tan y
tan (x- y) (11)
1 + tan x tan y
Using these relationships, we have
(a) when x = y, from (6)
sin (x + x) = sin 2x = 2 sin x cos x (12)
(b) when x = y, from (7),
cos 2x = cos x cos x - sin x sin x
= 1 - 2 sin 2 x = 2 cos 2 x - 1 (13)

since cos 2 x + sin 2 x = 1. Hence,
2 sin 2 x = 1 - cos 2x
• 2 1-cos2x (14)
or sm x = - - - -
2
2 1 + cos 2x (15)
and COS X = - - - -
2
(c) When x = y, from (10)
2 tan x
tan 2x = (16)
1- tan 2 x
(d) sin x + sin y = 2 sin ( x : y ) cos ( x ; Y ) (17)
sm . y = 2 cos (x+y).
. x- sm - 2- sm (x-y)
- 2- (18)
COS X + COS y = 2 COS (X : y) COS ( X ; y) (19)
. (x+y).
cosx-cosy=- 2 sm - 2- sm (x-y)
- 2- (20)
398 Appendix A
The formulae (17) to (20) can be verified by using (6) to (9) to expand
the right-hand side of each statement.
A.3 Degrees and radians

In elementary mathematical work it is convenient to measure
angles in degrees. In more advanced work, however, it is more con-
venient to use another measure of angles, known as radians. A radian
is defined as the angle at the centre of a circle which contains an arc
length equal to the radius of the circle (see Fig. A.2).
1 radian
Fig. A.2
If r is the radius of a circle, the circumference is of length 2nr. The

angle at the centre of a circle is 360°, and so
360° = 2nr = 2lr radians

r
or n radians = 180°
and n/2 radians = 90°

where n = 3.14159. Now consider the relationship between an angle
measured as Do and R radians. Since 360° = 2lr radians,
D R
=
360 2n
and so D = 360 R and R =2nD

2n 360
These convert radians to degrees and degrees to radians. For example,
if D = 30° R = 2lr( 30) = 0.5236 radians

' 360
Also, if R = 1.30 radians, D = 360 (1. 30) = 74.48°

2Jr
A.4 General angles

From Fig. A.3 and the definitions sin x = ale, cos x = blc,
tan x = alb we can see that varying x causes the point P to mark out a
quadrant of a circle from A to B. At the same time the values of a and b
vary.
8
Fig. A.3
If we allow x to become very small and approach zero, a ap-

proaches zero, and b approaches c, hence sin 0 = 0, cos 0 = 1, tan 0 = 0.
Similarly if we allow x to approach 90°, a approaches c and b
approaches zero. Hence,
sin 90° = sin (nl2) = 1
cos 90° = cos (tr12) = 0
tan 90° = tan (trl2) = oo
Let us now increase x to 135° or 3nl4 radians (Fig. A.4). We know that,
for example, using (6),
sin 135° = sin (90 + 45t = sin 90° cos 45° + cos 90° sin 45°
= cos 45° = 0.7071,
400 Appendix A
b A
Fig. A.4
that is, we can obtain sin x even when x is greater than 90° or n/2
radians. Similarly it can be shown that
cos 135° =-sin 45° = -0.7071
tan 135° = -tan 45° = -1.0000
Notice that cos 135° and tan 135° are both negative. This is because
b a
cosx =- tan x = b
c
and b is measured in the negative direction from 0, and so is a negative
number. Both a and c are positive numbers.
Fig. A.5 shows the angle x = 225°. In this case both a and b are
negative numbers so that sin x and cos x are negative, but tan x is
positive. It can be shown that
sin 225° = -0.7071 cos 225° = -0.7071 tan 225° = 1. 000
Similarly, the anglex can be increased to, say, 315°, and in this case
a is negative and both b and c are positive, so that sin x and tan x are
negative and cos x is positive. These results can be summarised as in
Fig A.6 to indicate which of sine, cosine and tangent are positive.
So far we have considered only angles of up to 2Jr radians (360°),
Fig. A.S
Sin All
Tan Cos
Fig. A.6
but it is possible to allow an angle to exceed this by rotating the point P

(Fig. A.5) around the circle. In this way, an angle of, say, 2.rr + x has the
same sine, cosine, and tangent as x has.
We know that
sin 0 =0 cos 0 =1 tan 0 =0
sin (.rr/2) =1 cos (.rr/2) =0 cos (.rr/2) =00
402 Appendix A
Fig. A.7
By evaluating sin x, cos x, and tan x for other values of x we can draw
the graphs of these functions (Fig. A.7).
The graph of cos x is identical to the graph of sin x moved n/2 to
the left. The graph of tan x has asymptotes at n/2, 3n/2, Sn/2, ... and so
it consists of a series of sections.
A.5 Differentiation
If y =sin x and we allow x to increase by a small amount ox, and let
the corresponding increase in y be oy, then
y + oy = sin (x + ox)
Subtracting, we have
Oy = sin (x + OX) - sin X
Using (6), we obtain
Oy = sin XCOS OX+ COS X sin OX- sin X
= sin x (cos ox - 1) + cos x sin ox
Hence,
Oy
-=
.
smx
(COS OX- 1) +-----
X sin OX
COS
ox ox ox
Now by MacLaurin's theorem (section 5.15)
x2 x4 x6
COS X = 1 -- + - - - + · · ·
2! 4! 6!
. x3 x 5 X1
and sm x = x-- +- - - + · · ·
3! 5! 7!
That is, cos ox = 1- (6x)2 + (ox)4- (ox)6 + ...

2! 4! 6!
cos ox- 1 ox (ox) 3 (ox) 5
so that ----= --+-----+ ...
Ox 2! 4! 6!
and lim(cos ox
/lx--.0 OX
-1) = 0
Similarly,
and l1 (sin -ox)

. m - =1
/lx-+0OX
Therefore, lim (oy) = dy = sin x (0) + cos x (1) = cos x

Ox
llx-+0 dx
Hence, if
404 Appendix A
y = sin x dy =COS X
dx
The same method can be used to show that if y =cos x, dyldx =-sin x.
If y =tan x = (sin x)/(cos x), then, using the rule for differentiating
quotients, we have
dy cos x (cos x)- sin x (-sin x)
dx (cos x) 2
= cos 2 x + sin 2 x = _1_ = sec 2 x

cos 2 x cos 2 x
Similarly it can be shown that if

1 dy -1
y = cotan x = - - - = -cosec2 x = - -
tan x dx sin 2 x
1 dy -cos x -cosec x
y =cosec x =-.- -=---=----
smx dx sin 2 x tan x
1 dy sin x
y= secx = - - - = - - = tan x sec x
cos X dx COS2 X
The above results can be generalised for the angle mx (where m is a

constant) by use of the function of a function rule.
For example, if y = sin mx, let () = mx so that d()!dx = m.
Then y = sin () dy = cos ()

d()
dy dy d()
- = - - = m cos mx
dx d() dx
The derivatives of the trigonometric functions are listed in Table A.l.
A.6 Integration
The integrals of some trigonometric functions can be deduced from
Table A.l. For example,
Jcos x dx = sin x + c
and Jsin x dx =-cos x + c
TABLE A.l
Function Derivative Function Derivative
sin x COS X sin mx m cos mx
COS X -sin x cos mx -m sin mx
tan x sec2 x tan mx m sec2 mx
cotan x -cosec2 x cotan mx -m cosec 2 mx
sec x tan x sec x sec mx m tan mx sec mx
-cosec x -m cosec mx
cosec x cosec mx
tan x tan mx
Now J tan x dx = J sin x dx __ -J -sin x dx

COS X COS X
=-log (cos x) + c
The integrals of other trigonometric functions are more difficult but
some can be obtained by making the substitutions
. 2 tan (x/2) 1 - tan 2 (x/2)

smx = COS X=
1 + tan 2 (x/2) 1 + tan 2 (x/2)
or, writing t = tan (x/2),
. 2t 1- (2
smx=-- cosx = - -
1 + (2 1 + t2
These can be shown to be true by using the formulae for compound
angles:
2 tan (x/2) 2[sin (x/2)]/[cos (x/2)]

1 + tan 2 (x/2) 1 + lsin 2 (x/2) ]/[cos2 (x/2)]
2 sin (x/2)
cos (x/2)[cos 2 (x/2) + sin 2 (x/2)]/cos 2 (x/2)
2 sin (x/2) cos (x/2)
= 1 = sin x, using (12) and (4)
406 Appendix A
Similarly,
1- tan 2 (x/2) = 1- [sin (x/2)]/[cos2 (x/2)]
2
1 + tan (x/2)
2 1 + [sin (x/2)]/[cos2 (x/2)]
2
cos 2 (x/2) - sin2 (x/2)

= = cosx
cos2 (x/2) + sin2 (x/2)
using (4) and (13). These can be used to integrate some functions which
cannot be integrated by other methods.
For example I _..!!:!...._
COS X
Let t = tan (x/2)

1- t 2 . 2t
Then cosx = - - smx=--
1 + t2 1 + (2
dx=_l:_!!!_
1 + t2
Now I dx I (1 + t 2) 2 dt I 2dt
COS X = (1 - t 2) (1 + t 2) = 1- t2
1 1
= J1+t dt + J1 - t dt
= log (1 + t)- log (1- t) + C

= log ( 1 + t) + C
1- t
1 + tan (x/2))
= Iog ( +C
1- tan (x/2)
Some standard integrals are given in Table A.2 (the constants of
integration are omitted).
A. 7 Inverse functions
If y = sin x then x = sin-1 y is known as the inverse sine function, or,
x is an angle whose sine is y. Similarly, x = cos-1 y and x = tan-1 y are the
inverse cosine and tangent functions.
These are useful in integration when trigonometric substitutions
can be made. For example,
TABLEA.2
Function Integral Function Integral
sin x -cosx sin ax -a1 cos ax

COS X sin x cos ax a1 sin ax
-1
tan x -log (cos x) tan ax alog (cos ax)
cotan x log (sin x) cotan ax a1 log (sin ax)
f V(a 1 2 - X2 )
dx
Substitute x=asinO
dx =a cos 8 dO
so that f 1 dx = f a cos 8 dO = f a cos e dO

v (a 2 - r) v (a 2 - a2 sin2 8) a cos e
= f1 de= e + c
but if X = a sin 8, 8 = sin- xfa, hence1
J V (a2-
dx
x2)
= sin- 1 (xla) + C
In the same way,
J(a dx+ r) = .!tan- 1 ~ +C

2 a a
where the appropriate substitution is X = a tan 8.
Appendix B
Set Theory
B .1 Introduction
A set is the name given to a group or collection of distinct objects.
Each member of a set is called an element. For example the set A might
be defined as consisting of the three integers 1, 2 and 3, so that
A= {1, 2, 3},
where curly brackets or braces are used around the elements, while the
set B might consist of four integers
B = {1, 3, 5, 7}.
To indicate that an element is a member of a set the special symbol E,
which is read as 'belongs to' or 'is an element of', is used. Here 1 E A
and 3 E B. When an element does not belong to a set the symbol fE. is
used so that 4 fE. A and 2 fE. B.
The special set which has no elements is called the null set or
referred to as an empty set and is represented by 0.
In these examples the sets A and B have been defined by enumerat-
ing all of the elements. This is possible when the number of elements in
the set is small but in other cases it is more convenient to include a
mathematical description, as in
P ={pIP> o}
where the set P is defined as containing those values of p for which p is
greater than zero, and here p might be the price of a good. Here the
convention of using capital letters for the set and lower case letters for
the elements of the set is adopted. Another example is
Q = {q 13 < q < 30}
where Q is those values of q which are greater than 3 and less than 30.
408
Set Theory 409
More generally, the name R is commonly given to the set of all real
numbers,
R = { x I x is a real number}
and the statement 'x is a real number' can be written x E R.
Another useful concept is the idea of the universal set, which is all
the elements under discussion, and therefore varies with the context.
For example, in discussing the properties of the set A the corresponding
universal set, S, might be defined in one of the following ways
s = {1, 2, 3, 4, 5, 6, 7, 8}
or s = {-1, 0, 1, 2, 3, 9}
or S = {x I xis a real number}
B.2 Combining sets
We now consider relationships between sets. First, two sets are
equal only if they contain identical elements. Thus if
A= {1, 2, 3},
and E = {2, 3, 1}
then E = A, even though the order in which the elements are written
differs. Conversely, if at least one element does not occur in both sets
the sets are not equal. For example, if
B = {1, 3, 5, 7}
A =I= B since 2 E A and 2 fE. B. In the special case where all the elements
of one set, M say, are also elements in a larger set, N say, then M is said
to be a sub-set of N, and this is written M C Nand is read 'M is a sub-set
of N' or 'M is contained in N'. An example of a sub-set is where
M = {Monday, Tuesday}
N = {x I xis a day of the week}
Another example is where
A = {1, 2, 3},
and s = {1, 2, 3, 4, 5, 6, 7, 8, 9}
so that A is a sub-set of S. Notice that every set is a sub-set of its
corresponding universal set.
410 Appendix B
Next, we can combine together two sets to form their union (indi-
cated by U), which includes those elements belonging to either or both
sets. For example,
A UB = {1, 2, 3} U {1, 3, 5, 7}
= {1, 2, 3, 5, 7}.
More formally, A U B = {x I x E A or x E B}
In the same way we can define the intersection (indicated by n) of two
sets as being only those elements which occur in both sets. Thus,
A n B = {1, 2, 3} n {1, 3, 5, 7}
= {1, 3}
The formal definition is
A n B = {x I x E A and x E B}
As another example consider
F = {a, b, c, -1, 0, 6}
and G = {c, d, 0, 1, 6},
then F U G = {a, b, c, d, -1, 0, 1, 6}
and F n G = {c, 0, 6}.
An interesting special case occurs when two sets have no elements in
common, and so are said to be disjoint, and their union is the universal
set. For example, consider
A= {1, 2, 3},
p = {4, 6, 7, 9}
and s = {1, 2, 3, 4, 6, 7, 9}
Since S=AUP
then P is called the complement of A, written A. Thus the union of a set
and its complement is the universal set.
B.3 Venn diagrams

The properties of combinations of sets can be illustrated by the use
of 'Venn diagrams'. In these, the universal set is represented by a
Set Theory 411
Fig. B.l
Fig. B.2
rectangle, and sub-sets of this by circles. Thus Fig. B.l shows the set A,
whose elements are within the circle and its complement, A, whose
elements are outside the circle but within the rectangle. The union of A
and B is shown in Fig. B.2 as the shaded area, which includes all the
elements in A or in Borin both. In Fig. B.3 the intersection of A and B
is shown as the shaded area and here it includes only those elements in
both A and B.
The use of Venn diagrams allows the following statements to be
verified for any sets A, Band C.
412 Appendix B
Fig. B.3
(a) Commutative law of unions and intersections:

AUB=BUA
AnB=BnA
By considering Fig. B.2 and Fig. B.3 it can be seen that
these are true and therefore the order of the sets does not
matter for unions and intersections. This corresponds to addi-
tion and multiplication in algebra where a + b = b + a and
ab = ba. Notice that in algebra the commutative law does not
apply for substraction and division.
(b) Associative law of unions and intersections:
~
These are illustrated in Fig. B.4 and Fig. B.S and also corre-
spond to addition and multiplication in algebra.
(c) Distributive law of unions and intersections:
A U (B nq = (A U B) n (A qU
An (B U q =(An B) U (An q
These are shown in Fig. B.6 and Fig. B.7.
Set Theory 413
Fig. B.4
Fig. B.S
The laws of unions and intersections of sets extend to the

special sets mentioned earlier, namely the universal set, the
null set and the complement of any set.
414 Appendix B
Fig. B.6
Fig. B.7
B.4 Relations and functions

The concept of a set is essentially one-dimensional since only one
variable is considered at a time. We now extend the ideas to combina-
tions of sets. For example, suppose we observe the price, p, of a product
Set Theory 415
and the corresponding quantity demanded, q, where both p and q are

real numbers and so are elements of the real number set R. That is,
p E R and q E R. The sets of prices, P, and quantities, Q, are of course
related since for each particular price there is an associated quantity and
these two sets can be combined into an ordered set, say,
M = {(p, q) I pER, q E R}
Here the outer curly brackets or braces define the ordered set M, while
the round brackets contain the elements of M, which are the sets P and
Q. Therefore M consists of ordered pairs of prices and quantities and so
while two sets, A = {1, 2, 3} and E = {2, 3, 1} are identical since they
contain the same elements, the two ordered pairs from the ordered set
M, (1, 5) and (5, 1) are different because the first is p = 1 and q = 5,
while the second is p = 5 and q = 1.
Any ordered pair of values, say (p, q) or more generally (x, y)
define what is called a relation between the variables. That is, a given
value of xis associated with one or more values of y. For example, the
relation defined by the ordered set {(x, y) I y = x + 2} includes the
ordered pairs (0,2), ( -1,1) and (100,102). Graphically, this relation can
be represented by a straight line. Another example is the relation
defined by the ordered set {(x, y) I y < 2x} which includes the ordered
pairs (3,0), (3,1) and (1,0). In this example it is clear that the relation
between y and x is not unique since for a given value of x many different
values of y can occur. The graphical representation consists of the area
below the straight line y = 2x.
When there is just one value of y corresponding to each value of x
the relation defines y as a function of x or y = f(x). That is, a function
relating y and x is a set of ordered pairs with the property that for each
value of x there is a unique value of y. Alternatively, the way in which y
and x are related is said to be single valued. Notice that it is possible for
a function to give the same y value for different values of x and an
obvious example is y = x 2 where x = - 1 and x = 1 both give y = 1. In
the converse case, for example if y2 = x, for each value of x there are
two values of y and here we have a relation that is not a function.
The rule defined by a function, such as y = x 2 , can also be inter-
preted as a mapping or transformation, since the value of x is mapped or
transformed into the value of y. The notation
f: X-> y
indicates the mapping from x toy with/representing the particular rule,
y = x 2 here. Also, y is called the value of the function and x the
416 Appendix B
argument. Returning to the terminology of section 1.1, the variable xis

the independent variable and y is the dependent variable. Two other
terms are used in the literature: the domain of a function is the set of all
permissible x values while the range is the set of all the values resulting
from the mapping and so is the set of y values. For example suppose
y=5x-5
and x lies between 2 and 20. Then the domain is the set {X 12 <X< 20}
and the range is the set {Y I 5 < Y < 95} .
Finally, it is sometimes convenient to find the inverse relationship
for a particular relation. This can generally be found but may not be a
function. For example, if
y=5x-5
the inverse relationship, which is a function, is
X= 0.2y + 1,
while if
the inverse relationship is
x= Yy
which has two values of x for each y and so is not a function.
Appendix C
Answers to Exercises
Exercises 1.5
1. (a) Y = 3x + 4; intercept is 4, slope is 3
(b) Y = 3x - 4; intercept is -4, slope is 3
(c) Y = 4 - 3x; intercept is 4, slope is -3
(d) Y= 3x; intercept is 0, slope is 3
(e) Y= 4; intercept is 4, slope is 0.
2. Let TC = a + bq
then from (a) 70 = a + lOb
and (b) 120 = a + 20b
Solving these equations simultaneously gives the values a = 20,
b=5
TC = 20 + 5q
Fixed cost = 20, variable cost = 5.
3. TC = 15 + 6q
Total cost is £75 for an output of 10 units.
4. (b) Fixed cost is 3, variable cost is 2 (c) 33 (d) 21.
Exercises 1.8
1. y = 33 + 2x
(a) At the breakeven point Y = total revenue = 13x
13x = 33 + 2x
x=3
(b) Net revenue = pX - Y = 13(15) - 33 - 2(15) = 132
417
418 Appendix C
(d) Because the slope is 2, i.e. the variable cost is £2 per unit.
2. (a) TC = 50 + 5q
(b) At breakeven point, 50 + 5q = 10q so that q = 10
(c) Net revenue = 12p - 110
(i) p = 5, net revenue = - £50
(ii) p = 10, net revenue = £10
(iii) p = 15, net revenue = £70.
3. (a) At equilibrium q s = q d
:. 25p - 10 = 200 - 5p
:. p = 7 and q = 165
(b) 20p - 25 = 200 - 5p
:. p = 9 and q = 155.
4. (a) p = 4, q = 3
(b) p = 4, q = 0
(c) There is no unique solution (only one line)
(d) The equations are inconsistent (parallel lines).
Exercises 1.12
1. (a) In equilibrium, qd = qs or 200- 4p = -10 + 26p. Rearranging
gives 210 = 30p or p = 7, hence q = 172 and pq = 1204.
(b) Here p 1 = p -5 so that qs = -10 + 26(p- 5) = 26p- 140.
Equating to demand gives 26p -140 = 200 - 4p so 30p = 340
and p = 11.3, q = 154.7. Tax revenue = 5q = 773.3 and the
producer's revenue is p 1q = 974.6.
(c) Here p' = 0.8p so that qs = -10 + 26(0.8p) = -10 + 20.8p.
Equating to demand gives -10 + 20.8p = 200- 4p so 24.8p =
210 and p = 8.47, q = 166.1.
Tax revenue = 0.2(8.47)(166.1) = 281.4
Producer's revenue= p'q = 0.8(8.47)(166.1) = 1125.5.
2. Let the flat-rate tax be t per unit so that qs = -40 + 15(p - t).
Equating to demand, - 40 + 15(p - t) = 300 - 6p or p = (340 +
Answer to Exercises 419
15t)/21 and if t = 1.7, p = 17.4.

For a 10% tax, qs = -40 + 15(0.9)p = -40 + 13.5p
In equilibrium, 300- 6p = -40 + 13.5p or p = 340/19.5 = 17.4
3. (a) Y = 20 + 0.7Y + 4 = 24 + 0.7Y or 0.3Y = 24 soY= 80 and
c = 20 + 0.7(80) = 76
(b) Y = 20 + 0.7(Y - 10) + 4 = 17 + 0.7Y or 0.3Y = 17 so
Y = 56.7 and C = 20 + 0.7(56.7- 10) = 52.7
(c) Y = 20 + 0.7(0.75)Y + 4 = 24 + 0.525Y or 0.475Y = 24 so
Y = 50.5 and C = 20 + 0.7(0.75)50.5 = 46.5
(d) Y = 20 + 0.7(0.75Y- 10) + 4 = 17 + 0.525Yor 0.475Y = 17
so Y = 35.8 and C = 20 + 0.7[(0.75)(35.8) - 10] = 31.8.
4. (a) IS: Y = C + I = 15 + 0.8Y + 75 -100i or Y = 450 - 500i
LM: 250 = 250- 160i + 0.1 Y or Y = 1600i
Solution is i = 450/2100 = 0.214 oe 21.4% and Y = 342.9
(b) IS: Y = 65 + 0.8Y -100i or Y = 325 - 500i
LM: Y = 1600i. Solution is i = 325/2100 = 0.155 andY= 248.
(c) IS: Y = 450 - 500i
LM: 275 = 250 - 160i + 0.1 Y or Y = 250 + 1600i.
Solution is i = 200/2100 = 0.095 and Y = 402.
5. (a) IS: Y = C + I + G = 15 + 0.8Yd + 75 - lOOi + 300
and Yd = 0.8Y - 5 so Y = 386 + 0.64Y - 100i and this
simplifies to Y = 1072.2 - 277.8i.
(b) LM: 950 = 0.2Y + 750- 260i or Y = 1000 + 1300i
Equilibrium requires 1072.2 - 277 .8i = 1000 + 1300i or
i = 72.2/1577.8 = 0.046 and Y = 1059.8.
(c) IS: Y = 440 + 0.64Y- 4- 100i or Y = 1211.1 - 277.8i
LM: Y = 1000 + 1300i
Equilibrium has i = 211.111577.8 = 0.134 andY= 1173.9.
Here both Y and i are higher than previously.
Exercises 1.14
1. 2xl + 2x2 = 2
3x 1 - X 2 = 1
420 Appendix C
2 2
1 -1 (-2 -2) -4
..
x, ~ I
I
2
3
2
-1
=
(-2 -6)
= - =;z1
-8
2 2
I 3 1 2-6 -4
and = - - = - =;z1
x, ~ 2 2 -8 -8
3 -1
2. x, = 1, X2 = 1.
3. x, - x 2+ x3 = 4
x, + X2 + 3x3 = 8
X 1 + 2x2 - X3 = 0
4 -1 1 1 3 8 3 8 1
8 1 3 4 - (-1) +1
0 2 -1 2 -1 0 -1 0 2
.. x, =
1 -1 1 1 3 1 3 1 1
1 1 3 1 - (-1) +1
1 2 -1 2 -1 1 -1 1 2
4(-7) + (-8) + (16) -20

= =-=2
(-7) + (-4) + (1) -10
1 4 1
1 8 3
1 0 -1 -8 + 16 -8
x2 = =0
1 -1 1 -10
1 1 3
1 2 -1
1 -1 4
1 1 8
1 2 0 - 16- 8 + 4 -20
x3 = =-=2
1 -1 1 -10 -10
1 1 3
1 2 -1
4. x,= 1, x2 = 1, x3 = -1.
5. x, = 2, x2 = 2, x3 = 2.
6. x, = -1, x2 = -2, X3 = -3.
Exercises 1.16
1. (a) Y = C + I + G = C + I + 20
C = 20 + 0.7(Y- T) = 20 + 0.7(Y- 5- 0.3Y)
= 16.5 + 0.49Y
I= 15 + 0.1Y
Arranging in the order Y, C I: Y-C-I=20
-0.49Y + C = 16.5
-0.1 Y +I = 15
Using the notation of the text,
1 -1 -1 0.9 -1 0
IAI = -0.49 1 0 -0.49 1 0
-0.1 0 1 -0.1 0 1
= 0.9 - 0.49 = 0.41
20 -1 -1 35 -1 0
IA,I = 16.5 1 0 = 16.5 1 0
15 0 1 15 0 1
= 35 + 16.5 = 51.5
422 Appendix C
1 20 -1 0.9 35 0
-0.49 16.5 0 = -0.49 16.5 0
-0.1 15 1 -0.1 15 1
= 14.85 + 17.15 = 32
1 -1 20 0.51 0 36.5
-0.49 1 16.5 -0.49 1 16.5
-0.1 0 15 -0.1 0 15
= 7.65 + 3.65 = 11.3
51.5 32 11.3
Hence, Y = - = 125.6, C = - = 78.0 and I = - = 27.6
0.41 0.41 0.41
At this equilibrium, G = 20 and T = 5 + 0.3(125.6) = 42.7 and so there

is a surplus of 22.7.
1. (b) If G = T = 5 + 0.3 Y the national income identity becomes
Y = C +I+ 5 + 0.3¥ or 0.7¥- C- I= 5. The other two
equations are unchanged: -0.49¥ + C = 16.5
-0.1 Y + I = 15
0.7 -1 -1 0.6 -1 0
Here, IAI = -0.49 1 0 = -0.49 1 0
-0.1 0 1 -0.1 0 1
= 0.6- 0.49 = 0.11
5 -1 -1 20 -1 0
IA,I = 16.5 1 0 = 16.5 1 0
15 0 1 15 0 1
= 20 + 16.5 = 36.5
0.7 5 -1 0.6 20 0
IA21 = -0.49 16.5 0 -0.49 16.5 0
-0.1 15 1 -0.1 15 1
= 9.9 + 9.8 = 19.7

0.7 -1 5 0.21 0 21.5

-0.49 1 16.5 = -0.49 1 16.5
-0.1 0 15 -0.1 0 15
= 3.15 + 2.15 = 5.3

36.5 19.7 5.3
Hence Y = - = 331.8 C =- = 179.1 I = - = 48.2
' 0.11 ' 0.11 ' 0.11
and T = 5 + 0.3(331.8) = 104.6 = G- higher than in (a) above, as are
Y, C and I.
2. Y =C+ I + G + (X - M) =C+ I0 + fM + X - M
or Y - C - (f - 1)M = I0 + X
-bY+C=a
-dY +M = c
Using the notation from the text,
1 -1 -f + 1 1-b 0 -f + 1
IAI = -b 1 0 = -b 1 0
-d 0 1 -d 0 1
= 1- b- df +d
Io +X -1 -f + 1 I0 + X + a 0 -f + 1
IA,I = a 1 0 a 1 0
c 0 1 -c 0 1
= Io + X + a + cf - c
1 I 0 +X -J + 1
IA2I = -b a 0
-d c 1
= (-f + 1)(ad- be)+ (a+ bi0 + bX)
1 -1 I 0 +X 1-b 0 a+ I0 +X
IA31 = -b 1 a -b 1 a
-d 0 c -d 0 c
424 Appendix C
= (1 - b)c + d(a + / 0 +X)
10 + X + a + cf - c
Therefore, Y = '
1- b- df + d
C = (- f + 1)(ad - be) + (a + b/0 + bX)

1- b- df+ d
(1 - b)c + d(a + /0 +X)

and M = --'----'--------'----....:.._
1- b- df + d
3. For A, 3- PA + p 8 = PA -2 or-2pA + p 8 = -5
ForB,8-2p 8 +pc=Ps-1 or-3ps+Pc =-9
For C, 6 +2pA- Pc = 2pc -2 or 2pA -3pc = -8
-2 1 0 0 1 0
0 -3 1 -6 -3 1
2 0 -3 2 0 -3
= -(18 - 2) = -16
-5 1 0 0 1 0
~ = -9 -3 1 = -24 -3 1
-8 0 -3 -8 0 -3
= -(72 + 8) = -80
-2 -5 0 0 -13 -3
IA21 = 0 -9 1 = 0 -9 1
2 -8 -3 2 -8 -3
= 2(- 13 - 27) = - 80
-2 1 -5 0 1 -13
IA31 = 0 -3 -9 = 0 -3 -9
2 0 -8 2 0 -8
= 2(-9 -39) = -96
rr = 3, qB = 4, qC = 10
4. Equating supply and demand in each market,
44 - 2pl + P2 + P3 = - 10 + 3pl or 54 = 5pl - P2 - p3
25 + P1 - P2 + 3p3 = - 15 + 5p2 or 40 = -p1 + 6p2 - 3p3
40 + 2pl + P2- 3p3 = -18 + 2p3 or 58= -2pl - P2 + 5p3
5 -1 -1 0 -1 0
IAI = -1 6 -3 29 6 -9
-2 -1 5 -7 -1 6
= (29 X 6) - (7 X 9) = 111
54 -1 -1 -4 0 -6
IA1i = 40 6 -3 = 40 6 -3
58 -1 5 58 -1 5
= -4(30 - 3) - 6 (- 40 - 348) = 2220
5 54 -1 0 254 -16
IA21 = -1 40 -3 -1 40 -3
-2 58 5 0 -22 11
= 2794 - 352 = 2442
5 -1 54 7 0 -4
IA31 = -1 6 40 = -13 0 388
-2 -1 58 -2 -1 58
= 2716 - 52 = 2664
426 Appendix C
2220 2442 2664

Therefore, p 1 = i l l = 20, P2 =ill= 22, p3 =ill= 24
q! = 50, q2 = 95, q3 = 30.
Exercises 2.3
(a) A + B = [ 2 + 2 4 + 0
1 + (-1) 3 + 1
l = [ 4 4]
0 4
(b) A- 2B = [
2 - (2
1 - (2 X
X 2) 4 - (2
1) 3 - (2
X 0) ]-- [ -2 4
1)
3 1
l
l l
- X
2 4][ 2 0 [(2x2) + (4x-1) (2x0) + (4x1)

(c) AB =[ =
1 3 -1 1 (1x2) + (3x-1) (lxO) + (3x1)
4-4 0+4] [ 0 4]
[
= 2- 3 0 + 3 = -1 3
(d)AC=[ 2 4][3]=[(2x3)+(4x1)]=[10]
1 3 1 (1 X 3) + (3 X 1) 6
(2 X 2) (2 X 1) (2 X 1)
_ [ 3] [ 2 4] It is not possible to form the product

(e) CA- matrix CA
1 1 3
(2 X 1) (2 X 2)
(f) B =[ 2 0
-1 1
l
:l[:-: l
=[ (2
(1
X
X
2) + (4
2) + (3
X
X
0)
0)
(2
(1
X
X
-1) + (4
-1) + (3
X
X
1)
1)
l
(g) C ~ [ : l C' ~ [3 1]
C'A =[3 1] [ 21 43]
(1 X 2) (2 X 2)
=((3 X 2) + (1 X 1) (3 X 4) + (1 X 3)) = (7 15)
(1 X 2)
(h) BA = [ 2 0
-1 1
l[ 2 4]
1 3
=
[
(2x2) + (Ox1)
( -1x2) + (1x1) ( -lx4) + (1x3)
(2x4) + (Ox3) l[
=
4 8]
-1 -1
This is not equal to AB (obtained earlier in (c)).

428 Appendix C
2.
~
0
1
2
j] B~ ~ -n c= [1 0 2]
(a) AB ~ [i
0
1 q[L:j
= [ (1x2)
(1x2)
2 -1 2 2
+ (Ox1) + (2x2) (1x1) + (Ox-1) + (2x2)

+ (1x1) + (3x2) (1x1) + (1x-1) + (3x2)
(Ox2) + (2x1) + (-1x2) (Ox1) + (2x-1) + (-1x2)
l
=[ 0~ -4~ l
(b)A'B = [ ~ ~ ~ i2 ~2
2 3 -1
l[ l
=
(3 X 3) (3 X 2)
(1x2) + (1x1) + (Ox2) (1x1) + (1x-1) + (Ox2)
[ (Ox2) + (1x1) + (2x2) (Ox1) + (1x-1) + (2x2)
(2x2) + (3x1) + (-1x2) (2x1) + (3x-1) + (-1x2)
l
=[ 5~ -3~ l
(3 X 2)
(c) A C does not exist
(3 X 3) (1 X 3)
(d) AC' = [ ~~
0 2 -1
; l[~ l
2
(1 1) + (0 0) + (2 X 2)
HJl
X X
= [ (1 X 1) + (1 X 0) + (3 X 2)
(0 X 1) + (2 X 0) + (-1 X 2)
(e) CB ~ 0 2] [ H]
(1 X 3) (3 X 2)
=[(lx2) + (Ox1) + (2x2) (1x1) + (Ox-1) + (2x2)]
=[6 5]
(1 X 2)
Exercises 2.5
~
3 1
.. C =[ - ] and IAI = 5
-1 2
:. A-1 = ~ [ l l
3 -1
-1 2
= [ _: -:
5 5
The solution to the equations is given by
X = A- 1b = [ l[4] ~ ~~
~ 7
= [ aX
~
4)
X
+l
4) + ~
~X
X
7)
7)
=[
.". X1 = 1, X2 =2
2. (a) X1 = 58/23, X2 = -14/23
(b) X1 = 0, X2 = -1, X3 =3
(c) X1 = 2, X2 = 0, X3 = 1
430 Appendix C
Exercises 2. 7
1. Including the unit matrix with A gives
2 -3 1
[
1 1 0
[
1 -4
1 1
1 -1
0 1
l
Subtract row 2 from row 1 to give 1 in the (1,1) position,
Substract row 1 from row 2 to give 0 in the (2,1) position,
[
1 -4
0 5
1 -1
-1 2
l
Divide row 2 by 5, to give a 1 in the (2,2) position, and add four
times the new row 2, to give a 0 in the (1,2) position,
[:
0 0.2 0.6]
1 -0.2 0.4
Hence, A-'= [ 0.2 0.6 rod A-' ~I
-0.2 0.4
:l
For 8, [ 0 2 1
-1 -1 0
To get 1 in the (1,1) position, subtract row 2 from row 1,
Add row 1 to row 2

[_: -1
3 1
0 -: l
[
1
0
3
2
1
1 -: l
Subtract 1.5 times row 2 from row 1, divide row 2 by 2

0
1
-0.5 -1
0.5 0
l
Hence, B- 1 = [ -0.5 -1 land B- 1 B = I
0.5 0
For C, -4
4
1 0
0 1
l
Divide row 1 by 3 and subtract the new row 1 from row 2
-4/3 113
16/3 -113
Add 0.25 times row 2 to row 1, and multiply row 2 by 3/16
114
[ I 0 114 l
0 1 -1116 3/16
Hence, c- 1 =
[ 114
114 rod c-• C ~
-1116 3/16
ForD, [ I -1 0
-2 2 3
3 0 2
1
0
0
0
1
0 n
Add twice row 1 to row 2, and subtract three times row 1 from row 3,
1 0
~
[ I -1 0
0 0 3 2 1
0 3 2 -3 0 l
Adding one-third of row three to row 1 and row 2
'
1/3]
[~
0 2/3 0 0
1 1113 1 1 113
3 2 -3 0 1
432 Appendix C
Subtract three times row 2 from row 3
~ ~ ~~0
-6 -3
l
Divide row 3 by - 9 and combine with row 1 and row 2
1 0 0 -4/9 -2/9
113]
[ 0
0
1 0
0 1
-13/9 -2/9
2/3 1/3
1/3
0
Hence, o- 1 =
[ -4/9 -2/9 1/3 ]"nd n-• D ~
-13/9 -2/9 1/3
2/3 113 0
For, E,
u 0 1
3 -2
1 -1
1
0
0
0
1
0 ~l
Divide row 1 by 2, subtract twice the new row 1 from row 2, and
add the new row 1 to row 3,
0 0.5
3 -3
1 -0.5
0.5
-1
0.5
0
1
0
0
0
1
l
Divide row 2 by 3 and subtract the new row 2 from row 3,
[~
0
0
1
0.5
-1
0.5
0.5 0 0
-113 1/3 0
516 -1/3 1
l
Subtract row 3 from row 1, add twice row 3 to row 2, and multiply
row 3 by 2,
[~
0
1
0
0
0
1
-1/3 113 -1
4/3 -1/3 2
5/3 -2/3 2
l
Hence, E-• ~ [ -113 113 -1 ]"nd E-• E ~
4/3 -113 2
5/3 -2/3 2
For F, 3 -2 3
[ -1 0 3 01 01 00 l
2 1 -1 0 0 1
Divide row 1 by 3, add the new row 1 to row 2 and subtract twice
the new row 1 from row 3,
1 -2/3
[ 0 -2/3 4
0
1
7/3 -3
113
113
-2/3
01
0
00
1
l
Subtract row 2 from row 1, multiply row 2 by -3/2, and subtract
l
7/3 times the new row 2 from row 3
[~ 0
~ ~
0 11
0 -1 0
-0.5 -1.5 0
0.5 3.5 1
Divide row 3 by 11, add three times the new row 3 to row 1, and
add six times the new row 3 to row 2
1
[ 0
0
0
1
0
0
0
1
3/22 -1122 3/11
-5/22 9/22 6/11
1122 7/22 1111
l
Hence, F- 1 = [ 3/22 -1122 3/11 land F- 1 F = I
-5/22 9/22 6/11
1122 7/22 1111
For G, -1 0 2
[ 1 2 1
2 1 3
Add row 1 to row 2, add twice row 1 to row 3, and multiply row 1
by -1,
1 0 -2
[ 0 2 3
0 1 7
-1 0 0
1 1 0
2 0 1
l
Divide row 2 by 2 and subtract the new row 2 from row 3
-2
1.5
5.5
-1
0.5
1.5 -0.5
g_ 5 ~ l
434 Appendix C
Divide row 3 by 5.5, add twice the new row 3 to row 1 and subtract
1.5 times the new row 3 from row 2
1 0 0
[ 0 1 0
0 0 1
-5/11
1111
3/11
-2/11
7/11
-1111
4/11
-3111
2/11
l
Hence G- 1 = -5/11 -2/11 4/11 land G- 1 G = I
[ 1111 7/11 -3/11
3/11 -1111 2/11
2. As explained in the text it is not necessary to calculate the inverse
matrix in order to get the solution but it is included in these
answers for completeness.
(a) 4 -2 1 0
[
2 1 0 1
Divide row 1 by 4 and subtract twice the new row 1 from row 2
[ I -0.5
0 2
0.25
-0.5
0
1
~ l
Divide row 2 by 2 and add 0.5 times the new row 2 to row 1
(b)
[ : 0
1
0.125 0.25
-0.25 0.5
[ 3 -1 2
: rnd so x ~ 2, y ~ I
1 0 0
1 3 1
2 1 1
0
0
1
0
0
1 !]
Divide row 1 by 3, subtract the new row 1 from row 2, and subtract
twice the new row 1 from row 3
1 -113
[ 0 10/3
0 5/3
2/3
113
-113
1/3
-113
-2/3
0
1
0
0
0
1
7/3
2/3
-2/3
l
Multiply row 2 by 3/10, add one-third of the new row 2 to row 1,
and subtract 5/3 times the new row 2 from row 3.
[ ~ 0 00 . 71
1
0.3 0.1 0
-0.1 0.3 0
0 -0.5 -0.5 -0.5 1
2.4]
0.2
-1
Multiply row 3 by -2, subtract 0.7 times the new row 3 from row 1,
[~
0 0
1 0
0 1
-0.2 0.2 0.2
1 1 -2
~ l
and subtract 0.1 times the new row 3 from row 2
-0.4 -0.6 1.4 and x ~ I, y ~ 0, z ~ 2
1 1 0 0
!l
(c)
[ I I
4 3 2 0 1 0
1 -1 -1 0 0 1
Subtract four times row 1 from row 2 and subtract row 1 from row 3,
1 1 1
[ 0 -1 -2
0 -2 -2
1 0 0
-4 1 0
:....1 0 1
-i
-2
l
Add row 2 to row 1, multiply row 2 by -1, add twice the new row 2
to row 3,
0 -1 -3 1 0
[~ 1 2
0 2
4 -1 0
7 -2 1 il
Subtract row 3 from row 2, add half of row 3 to row 1, and multiply
il
row 3 by 0.5,
0 0 0.5 0 0.5 and x ~ I, y ~ I, z ~ 0
[~ 1 0
0 1
-3 1 -1
3.5 -1 0.5
(d) 2 1 1 1 1 0 0 0
~l
1 -1 -1 1 0 1 0 0
1 2 3 -1 0 0 1 0
3 3 -1 2 0 0 0 1
Divide row 1 by 2, subtract the new row 1 from row 2 and from row
3, and subtract three times the new row 1 from row 4
436 Appendix C
0.5 0.5 0.5 0.5 0 0 0

1.5]
[~
-1.5 -1.5 0.5 -0.5 1 0 0 1.5
1.5 2.5 -1.5 -0.5 0 1 0 0.5
1.5 -2.5 0.5 -1.5 0 0 1 -0.5
Add row 2 to row 3 and row 4, divide row 2 by - 1.5 and subtract
half of the new row 2 from row 1
[~
0 0 2/3
1 1 -1/3
0 1 -1
0 -4 1
1/3
1/3
-1
-2
1/3
-2/3
1
1
0 0
0 0
1 0
0 1
Subtract row 3 from row 2, add four times row 3 to row 4,
~ l
0 0 2/3 113 1/3 0 0 2
[~
1 0 2/3 4/3 -5/3 -1 0 -3
0 1 -1 -1 1 1 0 2
0 0 -3 -6 5 4 1 9
Divide row 4 by- 3, add the new row 4 to row 3, subtract 2/3 times
the new row 4 from row 1 and from row 2
0 0 0 -1 13/9 8/9 2/9
4rdw =4
[~
1 0 0 0 -5/9 -119 2/9 -1 X = -1
0 1 0 1 -2/3 -113 -113 -1 y = -1
0 0 1 2 -5/3 -4/3 -1/3 -3 z = -3
Exercises 2.9
1. (a) llere, li - ~ I= 2 - (- 1) = 3 and so rank = 2
By Gaussian elimination,
[i ~~~ ~
Subtract row 2 from row 1 and subtract the new row 1 from
row 2
[ 1 -21 1 -1]
0 3 -1 2
Divide row 2 by 3 and add twice the new row 2 to row 1; this
will give the unit matrix and so the rank is 2.
Answers to Exercises 437
(b) 1 1 3
2 1 1
=11
-2 -111-121 -111
1 -2 -1
+312 11=1+3-9=-5
1 -1
and so the rank is 3. By Gaussian elimination,
~1 -2~ -1~ 0~ 0~ ~1
Subtract twice row 1 from row 2 and row 1 from row 3
1 1 3 1
[ 0 -1 -5 -2 1 0
0 0]
0 -3 -4 -1 0 1
Add row 2 to row 1 and subtract three times row 2 from row 3,
and multiply row 2 by - 1
1 0 -2
[0 1
0 0
5
11
-1
2
5
1 0
-1 0
-3 1
l
It can be seen that further operations will result in the unit
matrix and so the rank is 3.
(c) The determinant can be simplified by adding twice column 2 to
column 1, subtracting column 2 from column 3 and expanding
by row 2
1 2 -4 5 2 -6
2 -1 -1 0 - 1 0 = ~ - 61 = 0 so the
3 1 -5 5 1 - 6 5 - 6 rank < 3.
Next, try to find a non-zero (2 x 2) determinant. Here,
121 21 = - 1 - 4 = - 5 and so the rank is 2.

-1
[~
2 -4
-1 -1
1 -5
438 Appendix C
Subtract twice row 1 from row 2 and subtract three times row 1
l
from row 3
1 2 -4
[ 0 -5 7 -21 01 0
0
0 -5 7 -3 0 1
Subtracting row 2 from row 3 will give 0, 0, 0 on row 3 and
further row operations will give a (2 X 2) unit matrix and so
the rank is 2.
(d) Expanding the determinant by the first row,
1 0 0
1 1 1 = I ~ ~ I= 0 and rank < 3
0 1 1
Searching for a non-zero (2 x 2) determinant, I~ ~ I= 1
and the rank = 2
l
1 0 0
[ 1 1 1 1 0 0
0 1 0
0 1 1 0 0 1
Subtract row 1 from row 2 and subtract the new row 2 from
row 3,
1 0 0 1 0 0] and the rank is 2.
[ 0 1 1 -1 1 0
0 0 0 1 -1 1
(e) The determinant can be simplified by subtracting column 1
from column 2 and from column 3, and then expanding by
row 1
1 1 1 1 0 0
1 0 2 = 1 -1 1 = 3 and rank = 3
2 2 -1 2 0 -3
~l
1 0
u
1 1
0 2 0 1
2 -1 0 0
Subtract row 1 from row 2 and subtract twice row 1 from
l
row 3
1 1 1
[ 0 -1 1 0 0
1 -1 1 0
0 0 -3 -2 0 1
It can be seen that further row operations will result in the unit
matrix and so the rank is 3.
(f) Here the matrix is (3 x 4) and so the maximum rank is 3.
Taking the first 3 columns, the determinant is
1 0 0
0 2 2
2 1 2 =I 2 2
1 2
I =2 and rank =3
[~ ~
0 0 -1 1 0 0
2 2 1 0 1 0
1 3 1 0 0 1 l
Subtract twice row 1 from row 3 and divide row 2 by 2
1 0 0 -1 1 0 0 0]
[0 1 1 0.5 0 0.5 0 0
0 1 3 3 -2 0 1 0
Subtracting row 2 from row 3 will give 0 0 2 on row 3 and so
a unit matrix can be formed and the rank is 3.
(g) The maximum rank is 3 and taking the first three columns,
adding column 1 to column 3
1 0 -1 1 0 0
1 1 -1 1 1 0
0 1 -1 0 1 -1
=i 11 0
-1
I=- 1 so rank =3
u
0 -1 -1 1 0 0 0
1 -1
1 -1
-1
-1
0 1 0 0
0 0 1 0 l
440 Appendix C
Subtracting row 1 from row 2 and the new row 2 from row 3
1 0 - 1 - 1 1 0 0 0]
[0 1 0 0 -1 1 0 0
0 0 -1 -1 1 -1 1 0
and further row operations will give a (3 x 3) unit matrix and
so the rank is 3.
2. In each case there is a unique solution if the determinant has a
non-zero value.
(a) 1 3 -2 1 3 -2
1 -2 1 = 0 -5 3 =i-5 31=-11
2 -2 3 0 -8 7 -8 7
where row 1 was subtracted from row 2 and twice row 1 from row
3, and there is a unique solution.
(b) 1 2 1 5 2 5
2 -1 2 0 -1 0
1 1 1 3 1 3
where twice column 2 was added to column 1 and column 3, and as
the determinant is zero there is no unique solution.
(c) 1 3 -1 2 1 2 -1 -3
1 -1 2 1 = 1 -2 2 -4
1 1 1 -2 1 0 1 -7
1 1 0 5 1 0 0 0
2 -1 -3
-2 2 -4
0 1 -7
where column 1 was subtracted from column 2 and five times
column 1 was subtracted from column 4. Adding row 2 to row 1
gives a determinant with two identical rows and so it has a value
zero. There is no unique solution.
Exercises 2.11
1. A=[2 1
1 0
2 1
1 0
2 1
r ~~ i ~ i] =A
A+B=[21 1 2]+[3 -1 1]=[5 0 3]

0 1 2 -2 0 3 -2 1
and (BC)' = [ - 6 - 6]
-1 -2
Also, C'B' = [- 62 -1][ 3 2]=[-6 -6]= (BC)'.
-2
1 0 ~ ~ -1
2. Require A2 = A for A to be idempotent
~ ][ ~ ~ H~
0 0 0
A'= [ ~ 1
0
1
0
1
0
~ ] = A - idempotent
0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25

Bz= 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25
0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25
0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25
0.25 0.25 0.25 0.25
0.25 0.25 0.25 0.25 = B - idempotent
0.25 0.25 0.25 0.25
0.25 0.25 0.25 0.25
C2 = [ a 0 ] [ a 0 ] = [ a2 0 ] = C only if a = 0 or a = 1.
Oa Oa Oa 2
3.A'=[ 2 1]and(A't 1 =[0.2 -0.2]

-3 1 0.6 0.4
Now A-I= [ 0.2 0.6] and so (A-T = [0.2 -0.2] = (A't 1
-0.2 0.4 0.6 0.4
AB=[2 -3][ 2 2]=[7 1]and(ABt 1 =1_[ 3 -1]

1 1 -1 1 1 3 20 -1 7
442 Appendix C
Also, B- 1 = [ 0.25 -0.5]

0.25 0.5
and so B-tA- 1 =_1_[ 3 -1]=(ABt 1 •
20 -1 7
4. Since A = [ ~ j ] then lA - rii = 3- r 1 I
1
1 3- r
= (3 - r) (3 - r) - 1
and setting this to zero gives 9 - 6r + r 2 - 1 = 0. Solving gives
r = 4 and r = 2.
For r = 4 the characteristic vector is from [A - 4I]x = 0,
Or [ 3 - 4 1 ] [X1] =[ 0 ] and -X 1 + X 2 : 0 SO X 1 = X 2
1 3 - 4 X2 0 X1 - X2 - 0.
Normalising by setting xi +xi = 1 gives x 1 = 11\12 = X2 • For
r= 2 the characteristic vector is from [A - 2I]x = 0,
Or [ 3 - 2 1 ] [ X 1 ] = [ 0 ] and X1 + X 2 = 0 SO X 1 = -X 2
1 3 - 2 X2 0 X1 + X 2 = 0.
Normalising by setting xi + x; = 1 gives x 1 = 11\12 and
X2= -11\12.
Since B = [~ - ~ then IB - rii = 12 ~ r 1- ~rI
= (2 - r) (1 - r)
and setting this to zero gives r = 2 and r = 1.
For r = 2 the characteristic vector is from [B - 2I]x = 0,
or [ 2 - 2 - 1 ] [x1] =[ 0 ] and - x 2 =0
0 1- 2 X2 0 -x 2 = 0.
Normalising by setting xi +xi = 1 gives x 1 = 1, X 2 = 0.
For r = 1 the characteristic vector is from [B - ll]x = 0,
Or [2 - 1 - 1 ] [X1] =[0 ] and X1 - X2 = 0 SO X 1 = X 2
0 1- 1 X2 0
Normalising by setting x 12 + x} = 1 gives x 1 = 11\12 and x 2 =
11\12.
Exercises 2.14
1. Level of final demand which can be met by Industry 1 is
1500 - (200 + 300) = 1000, and by Industry 2 is
2500 - (500 + 100) = 1900.
2. The matrix of technological coefficients is
[~500
1500
100
2500
~ fsI
3
is]
25
I
and the new situation is as shown in the table.
Input to Level of
Industry 1 Industry 2 output
2 3
Industry 1 J5X 2,000 zsx 2,500 2,000
I I
Industry 2 3 X 2,000 zsx 2,500 2,500
Therefore final demand which can be met by Industry 1 is equal to

2,000 - (fs x 2,000 + ~ x 2,500) = 1,4331 and by Industry 2 to
2,500 - ~ X 2,000 + fs X 2,500) = ~
3. (a)
[ 1-0.2 ~
-0.3 1 - 0.2 x2 2000
which can be solved using the inverse matrix to give X 1 = 3077
and X 2 = 3654 approx.
(b)
[ 0.9 ~
-0.4 0.9 x2 2000
X 1 = 3684 and X 2 = 3860 approx.
4. (a) HereA=[100/1000 200/2000]=[0.1 0.10]
500/1000 100/2000 0.5 0.05
Therefore, I- A= [ 0.9 -0.10] and
-0.5 0.95
(I- A)- 1 =[1.180 0.124]
0.621 1.118
444 Appendix C
by row operations. The multipliers are 1.180 + 0.621 = 1.801

and 0.124 + 1.118 = 1.242.
If the final demand for exports becomes 500 for agricul-
ture and 900 for industry, the total final demands are 800 and
1500 respectively. Therefore,
X=(I-A)- 1 C=[1.180 0.124][ 800]=[1130.0]
0.621 1.118 1500 2173.8
instead of 1000 and 2000.
(b) Here 0.1 0.10 300/900
0.1 0.40 0/900
l
[0.1 0.10 0.33]
A = [ 0.5 0.05 6001900 = 0.5 0.05 0.67
0.1 0.40 0.00
Therefore, I- A= [ 0.9 -0.10 -0.33]
-0.5 0.95 -0.67
-0.1 -0.40 1.00
and (I - At 1 can be found
0.9
[ -0.5
-0.1
-0.10
0.95
-0.40
by row operations:
-0.33 1 0 0
-0.67 0 1 0
1.00 0 0 1
l
Divide row 1 by 0.9 and combine the new row 1 with rows 2
and 3:
1 -0.111 -0.370 1.111 0 0]
[0 0.895 -0.855 0.556 1 0
0 -0.411 0.963 0.111 0 1
Divide row 2 by 0.895 and combine the new row 2 with rows 1
and 3:
1 0 -0.476 1.180 0.124 0
[ 0 1 -0.955 0.621 1.117 0
0 0 0.570 0.366 0.459 1
l
Divide row 3 by 0.570 and combine the new row 3 with row 1
and row 2:
1 0 0 1.486 0.507 0.835]
[ 0 1 0 1.234 1.886 1.675
0 0 1 0.642 0.805 1.754
(Check that (I - A) (I- At 1 = I, apart from rounding errors)
The multipliers, from adding the columns, are 3.362, 3.198

and 4.264, compared to 1.801 and 1.242 previously. If the final
demand for exports becomes 500 for agriculture and 900 for
industry, these are the total final demands (since households
are an industry). Therefore,
X= (I- At'C =[ 1.486 0.507 0.835][500] [1199.3]

1.234 1.886 1.675 900 = 2314.4
0.642 0.805 1.754 0 1045.5
instead of the original values of 1000, 2000, 900 and the values
in (a) of 1130 and 2173 (and 900 for households).
5. (a) The technology matrix is obtained by dividing each column by
the gross output of the sector:
11/41 19/240 11185] [ 0.268 0.079 0.005]
A = [ 5/41 89/240 40/185 = 0.122 0.371 0.216
5/41 37/240 37/185 0.122 0.154 0.200
0.732 -0.079 -0.005]
Hence, I - A = [ - 0.122 0.629 - 0.216
-0.122 -0.154 0.800
and (I - A) multiplied by the given matrix is approximately I.
(b) Here, X= (I- At 1
C =[ 1.409 0.192 0.062] [ 15]
0.372 1. 753 0.476 120
0.286 0.367 1.351 130
52.2]
= [ 277.8
224.0
(c) Combining the industry and services sectors gives:
Input to Final Total

Output from Agriculture Other demand output
Other 10 203 212 425
The technology matrix is A = [ 11141 20/425]

10/41 203/425
= [ 0.268 0.047]
0.244 0.478
446 Appendix C
and so I-A=[ 0.732 -0.047]

-0.244 0.522
By row operations (I - At = [ 1.408
I 0.127]
0.658 1.975
Hence, X= [1.408 0.127] [ 15] =[ 52.9]
0.658 1.975 250 503.6
These values compare with 52.2 and 277.8 + 224.0 = 501.8
and so, apart from rounding errors (which could be reduced by
working to more places after the decimal point in both A and
(I - A)- 1), the values for agriculture and the aggregate are the
same. Thus, for an economist interested only in agriculture it
does not matter whether we have one or two other sectors and,
by implication, the results for agriculture would be the same if
there were many other sectors.
Also, the multipliers for agriculture are 2.066 in both
cases, while for industry the multiplier is 2.312 and for ser-
vices, 1.889, and for the combined sector 2.102, the average of
the two values.
Exercises 3.2
1. (a) Let x = output and TC = total cost, and since the cost func-
tion is quadratic, let TC = a + bx + cx 2 •
Using the three pairs of values given:
4 =a (1)
14 = a + 2b + 4c (2)
58 = a + 6b + 36c (3)
Substituting from (1) into (2) and (3) and simplifying:
10 = 2b + 4c (4)
54 = 6b + 36c (5)
Subtract three times (4) from (5)
24 = 24c or c = 1
Putting a = 4 and c = 1 in (2) gives b = 3 and so
TC = 4 + 3x + x 2
X 0 2 4 6 8 10
TC 4 14 32 58 92 134
(b) TC = 10 + 2x + 0.5x 2
X 0 2 4 6 8 10
TC 10 16 26 40 58 80
(c) TC = 25 + 0.5x + 0.25x 2
X 0 2 4 6 8 10
TC 25 27 31 37 45 55
(d) TC = 20 + 0.8x + 0.02x 2

X 0 1 3 5 7 9 10
TC 20 20.82 22.58 24.50 26.58 28.82 30.00
(e) TC = 20 + 0.5x 2
X 0 2 4 6 8 10
TC 20 22 28 38 52 70
2. X 0 1 2 3 4 5 6 7 8 9 10
(a) y 100 111 124 139 156 175 196 219 244 271 300
(b) y 100 91 84 79 76 75 76 79 84 91 100
(c) y 100 109 116 121 124 125 124 121 116 109 100
(d) y 100 89 76 61 44 25 4 -19 -44 -71 -100
Exercises 3.4
1. Formula is x = {-b ± v'(b 2 - 4ac)}/2a
(a) a= 1, b = -7, c = 12 sox= 4 or 3
448 Appendix C
(b) a = 1, b = 1, c = - 2 so x = 1 or - 2
(c) a= 2, b = 7, c = 3 sox= -0.5 or -3
(d) a= 1, b = -2, c = 1 sox= 1 (repeated root)
(e) a = 1, b = 0, c = -1 sox = 1 or -1
(f) a = 1, b = 0, c = 1 sox = i or - i
(g) a = 1, b = - 4, c = 5 so x = 2 + i or 2 - i
(h) a = 1, b = 2, c = 2 sox = -1 + i or -1 - i.
2. (a) y = 0 when x 2 - lOx + 25 = 0 or x = 5. Drawing up a table
of values:
X 0 3 5 7 10
y 25 4 0 4 25
(b) y = 0 when - x 2 + 4x - 3 = 0 or x = 1 or x = 3:
X 0 1 2 3 4
y -3 0 1 0 -3
Exercises 3.6
1. (a) y = 0 when x = 2 or 6:
X 0 1 2 3 4 5 6
y 12 5 0 -3 -4 -3 0
(b)
X 0 1 2 3 4 5 6
y 600 418 216 0 -224 -450 -672
Here x = 3 is a root and the cubic= (x- 3) (x 2 - lOx - 200).

The roots of this quadratic are x = 20 and x = - 10, which are
outside the range of interest and can be ignored.
(c) Setting y = 0 gives x = 1.5, x = -1 and x = 5 so these are
included in the table of values:
X -2 -1 0 1 1.5 2 3 4 5 6
y -49 0 15 8 0 -9 -24 -25 0 63
(d) Setting y = 0 gives x = 0, and so 32 - 2x 2 = 0 resulting in

x = 4 or -4.
X -4 -3 -2 -1 0 1 2 3 4
y 0 -42 -48 -30 0 30 48 42 0
(e) When y = 0, x = 0:
X 0 1 2 3
y 0 1 32 243
(f) y = 20/x so as x ~ 0, y ~ oo
X 1 5 10 20 100
y 20 4 2 1 0.2
(g) y = (20/x) + 1. ~ ~ oo, ~ ~ 1:
X 1 5 10 20 100
y 21 5 3 2 1.2
2. Q = KL so that K = Q/L
L 10 20 50 80 100
Qt/L 2 1 0.4 0.25 0.2
QiL 5 2.5 1 0.625 0.5
QiL 10 5 2 1.25 1
3. AC = TC/x. Here x > 0 and negative values can be ignored:
X 1 10 20 50 100 200
(a) AC 100 10 5 2 1 0.5
(b) AC 105 15 10 7 6 5.5
(c) AC 106 25 30 57 106 205.5
450 Appendix C
Exercises 3.10
1. The breakeven point has TC = TR.
(a) x 2 - 6x + 9 = 0 has x = 3 as the repeated root and this is the
breakeven point
(b) x 2 - 9x + 18 = 0 has x = 6 and x = 3 as the breakeven

points
(c) 2x 2 - 3x + 10 = 0 has no real roots and so there is no
breakeven point.
2. (a) TC = 25 - 6x + x 2
(b) 305
(d) total revenue equals total cost when
lOx + 15 = 25 - 6x + x 2
that is when x 2 - l6x + 10 = 0
The roots of this quadratic are given by
16 ± Y(l6 2 - 4 X 10)
X =
2
= !(16 ± V(256 - 40)) = !(16 ± V (216))
= 8 ± 7.35 = 15.35 or 0.65.
These are the two levels of output at which total revenue
equals total cost.
4. The equations are satisfied by the values x = 1 and x = 15.
5. y = 20 + 3x + X 2
y = 5x + b
20
+ 3x + x 2 = 5x + b
or x2 - 2x + (20- b) = 0
2 ± V[4- 4(20- b)]
X =
2
= 1 ± V[ 1 - (20 - b)]
= 1 ± V'(b- 19)
when b = 20 x = 1 ± 1 =2 or 0 (2 real solutions)
b = 19 x = 1 (the solutions are coincident)

b = 18 x = 1 ± Y( -1) (2 complex solutions)
6. 2p 2 - 3p - 40 = 250 - 4p - p 2
3p 2 +p - 290 = 0
-1 ± Y(l + 12 X 290)
p
6
-1 ± \13481 -1 ±59
6 6
58 -60
or
6 6
In this example we take the positive root, p = 9§, and at this value
q = ~
7. (a) p = 5 q = 10 (b) p = 4 q=4
(c) p = 2 q =3
8. Net revenue NR = TR - TC and TR = pq. Since TC is expressed
in terms of q rearrange demand to give p = (200 - qD)/2 so
NR = qD (200 - qD)/2 - 20 - 5qD.
At the breakeven points, NR is zero when qD = 0.21 or 189.8.
The corresponding prices are 99.9 and 5.1:
0 10 20 50 100 150 200

-20 880 1680 3480 4480 2980 - 1020
(maximum NR is when qD is 95 and NR = 4492.5)
Exercises 3.12
1. (a) When x = 19, TC = 72 and when x = 21, TC = 78 and extra
cost is 78 - 72 = 6
(b) When x = 29, TC = 102 and when x = 31, TC = 133 and
extra cost is 133 - 102 = 31
(c) When x = 49, TC = 187 and when x =51, TC = 223 and
extra cost is 223 - 187 = 36.
2. Let x = number of cans sold
Revenue = 30x for 0 < x ::::; 100 and
= 30 (0.85)x for x > 100
452 Appendix C
If x = 99, cost = 30 (99) = 2970p

If x = 101, cost = 30 (0.85) 101 = 2575.5- cheaper than for 99!
New system: Revenue = 30x for 0 < x < 101
Revenue = 100 (30) + (x - 100) (0.80) 30 for x > 100
If x = 120, old revenue = 30 (0.85) 120 = 3060,
new revenue = 100 (30) + (20) (0.80) 30 = 3480.
Exercises 4.3
1. (a) 19, 225 (b) 950, 12,750
(c) -210, -2,250 (d) 82, 1,290
2. £300 3. £11,300, £67,900
4. Arithmetic progression with 20 = a + (n - 1)d = 10 + 15d. Hence
d =~ rate of interest is ~ = 0.067 or 6.7 per cent.
Exercises 4.5
10(1 - 3 10)
1. (a) 2,430; S10 = = 295240
-2
1
(b) 3; S10 =
81[1 - Grol = 121.5 approxtmately
.
1
1- 3
81
Soo = I = 243/2 = ~
1 - (3)
2[1 - (- 2) 10]
(c) -64;5 10 = =-682
3
(d) 32· s = -1024[1 - ( ~

= - 682 approximately
' 10 I
1- (-2)
-1024 2
Soo = I = -6823
1 - ( -2)
1 [1 - (0.1) 10] .
(e) 0.00001; S10 = 1 _ 0. 1 = 1.111 approxtmately
1
Soo = - 10
1 - 0.1 - 9
2. (a) 300(1.05) 15 = 623.7 (b) 300(1.10f 5 = 1253.2
150(1 - 1.0625) = 8229.7

(1 - 1.06)
4. 4(1 + r) = 5 4
. . (1 + rt = 1.25
1 + r = 1.06 approximately
and implied rate of compound interest is 6 per cent.
Exercises 4.7
1. (a) Present value of £800 received in 10 years time is given by
£800/(1 + i)' 0 ; this is equal to £300 .
.". (1 + i) 10 = 800/300 = ~
We therefore require the discount factors which make
1 = ~ = 0.375
(1 + i)'o 8
From Table 4.5 it can be seen that 0.386 is the factor used to
discount sums of money received 10 years hence when the
interest rate is equal to 10 per cent. Therefore the implied rate
of compound interest is slightly greater than 10 per cent.
(b) The present value of the £800 when the interest rate is 16 per
cent is given by £800 x 0.227 = £181.6
2. (a) £1,000 x 0.270 = £270 (b) £1,000 X 0.463 = £463
3. (a) £250 x 0.621 = £155.25
(b) £155.25/0.751 = £206.72
4. Present value of A = 100 x 0.909 + 200 x 0.826
+ 300 X 0.751 = 481.4
Present value of B = 150 x 0.909 + 300 x 0.826
+ 100 X 0.751 = 459.3
:. Project A has the greater present value if the discount rate is 10
per cent.
5. Let r = internal rate of return; then
454 Appendix C
(100 - 120) + (110 - 120) + (160 - 100) = 0

(1 + r) (1 + r) 2 (1 + rY
- 20(1 + r)2 - 10(1 + r) + 60 = 0
(1 + r) 3
2(1 + r) 2 + (1 + r) - 6 = 0
2 + 4r + 2r 2 + 1 + r - 6 = 0
2r 2 + 5r- 3 = 0
r = -5 ± Y(25 + 24t -5 ± 7
4 4
Taking the positive root, r = 0.5, or 50 per cent.
Exercises 4.9
1 PV 100 [1
= 0.08 -- 1 -] = 981.
. 1.0820
2. PV = Ali = 25/0.04 = 625.
3. A = 2500 (1 - 1.09)/(1 - 1.09'5 ) = 85.15.
4 · (a) PV ~ [1 -1.06
0.06
-1 -] = 441.6
10
(b) PV = 60/0.06 = 1000 .

. (1
+ ')n
5. For a mortgage, A =l l p
(1 +it- 1
where A = annual payment, i = interest rate, p = amount bor-
rowed.
0.12 (1.12) 20 70000
Here A = (l.l2)2o _ 1 = 9371.51 per year
With n = 30, A = 8690.05.
6. Value, V = A{(1 +it- 1}/i = 300{1.09'0 - 1}/0.09 = 4557.88
With n = 20, V = 15348.04.
7. Using the same formula as in Question 6, V = 250 {1.08n- 1}/
0.08 and (a) V = 18276.49 (b) V = 43079.20.
Exercises 4.13
1. (a) Simple interest of 6% per annum produces £45
Compound interest of 5% produces £41
Compound interest of 4% paid twice per
year produces £ 33
:. most profitable investment is that returning simple interest
at 6%.
(b) £90, £94, £73; most profitable investment is that returning
compound interest of 5% per annum.
2. 50(1.02) 12 = 63.4
3. (a) 1 + 5x + 10x2 + 10x3 + 5x4 + x 5
(b) 8 - 12x + 6x2 - x3
(c) 81 + 108 ( ~ + 54 ( ~ r+ 12 ( ~ r ~r +(
(d) 64 - 192x2 + 240x4 - 160x6 + 60x 8 - 12x10 + x12
(0.1)2 (0.1)3
4. (a) e01 1 + 0.1 +-
2- + -
6 -+ · · ·
1 + 0.1 + 0.005 + 0.00017 + ...

1.10517 approx. (from tables e0 · 1 = 1.1052)
(b) 1+05 + (0.5)2 + (0.5)3 + ...

. 2 6
1.6458 approx. (from tables e0 · 5 = 1.6487)
If we include the next term in the expansion we obtain a closer
approximation with e0 · 5 = 1.6484.
(c) ~ = 1 +2+~+~+~+ ~ + ~ + ···
7.35556 approx. (from tables e 2 = 7.3891)
. . in this case we need to include more terms to obtain a
good approximation.
(d) e-1
0.368 approx. (from tables e- 1 = 0.3679).
5. Let P, be the population in millions after t years so that P, = 2e0 ·04'.
456 Appendix C
When t = 5, P5 = 2e0 ·2 = 2(1.2214) = 2.44 million. When t = 25,

P 25 = 2et.o = 2(2.7183) = 5.44 million.
If growth rate is 2%, P, = 2e0 ·021 and P5 = 2e0 · 1 = 2(1.1052)
= 2.21 instead of2.44, while P25 = 2e0 ·5 = 2(1.6487) = 3.30 instead
of 5.44.
6. Let Y, be the value of national income in after t years so that
Y, = 125e0 ·031 • When t = 30, Y 30 = 125e0 ·9 = 125(2.4596) = 307.5.
7. Let S, be the number of firms surviving in year t then S,
= 15(}()()0e-O lSI and for ( = 4, S4 = 150000e-0·6 = 150000(0.5488) =
82320. fort= 20, S20 = 150000e-3 ·0 = 150000(0.04979) = 7469.
8. Let r be the required continuous rate of interest, then e' = (1
+ 0.0612Y = l.OY = 1.0609. Taking natural logarithms, r = 0.0591
or 5.91%.
Exercises 4.15
(1 + x) x3 r
1. These use log.,(l _ x) = 2{ x + )+S+ · · ·}
and x = (a- 1)/(a + 1). Also, log10 x = ~ ~ For a= 0.1,
x = - 0. 9/1.1 = - 0.8181818 so all the terms in the series are
negative. The series is 2{- 0.8181818 - 0.1825695 - 0.0733296
- 0.0350631 - 0.0182560 - 0.0099990 - 0.0056638 - 0.0032859
- 0.0019409 - ... } = 2{ -1.1482896} = -2.2965792 compared
with -2.302585 in the tables. Using loge10 = 2.3025851,
log 100.1 = -2.2965792/2.3025851 = -0.9974 while the accurate
value is - 1.0.
For a = 100, x = 991101 = 0.980198 and the series is
2{0.980198 + 0.3139209 + 0.1809669 + 0.1241934 + 0.0928072
+ 0.0729557 + 0.0593111 + 0.0493874 + 0.0418684 + ... }
= 2{1.915609} = 3.831218. The accurate value is 4.60517 and the
error is because of the slow convergence of the series which occurs
because x is close to 1. log10100 = 3.831218/2.3025851 = 1.6639
instead of the accurate value of 2.0.
2. Let P, be the population in year t. Here, P, = 2.4e0 ·051 •
(a) Require solution of 3 = 2.4e0 ·051 or 1.25 = e0 ·051 • Taking natural
logarithms, log.,l.25 = 0.05t or 0.22314 = 0.05t so t = 4.46
years.
(b) Require solution of 5 = 2.4e0 · 051 or 2.0833 = e0 ·051 • Taking
natural logarithms, 0.73397 = 0.05t sot= 14.68 years.
3. Let S, be the number of firms surviving in year t. Here, S,

= 100000e-O 12'. Require the solution of 50000 = ~· or 0.5
= e- 0121 • Taking natural logarithms, -0.69315 = -0.12tsot= 5.78
years. But this is with this year taking the value 1, so it will be in
4.78 years from now.
4. Pareto's law: n = N/xu. Here N = 55000000
(a) x = 10000 son = 55000000/1000015
Taking logarithms to the base 10, log 10n = 7.74036 -
{1.5(4.00)} = 7.74036-6 = 1.74036 and antilog is 55.0 and 55
are expected to have incomes over £10,000. This is rather low
and implies that the power to which x is raised should be
smaller than 1.5.
(b) x = 100000 so n = 5500000011000001.5
Taking logarithms to the base 10, log1on = 7.74036- {1.5(5)}
= 0.24036 and antilog is 1.7. Therefore 2 are expected to have
incomes over £100 000. Again this suggests the power to which
x is raised should be lower.
Exercises 5.3
1. y = 3x - 4, dyldx =3
2. y = x2 - 3x + 3, dyldx = 2x - 3
3. 12x3 - 3x + 2x + 25
2
5. (1/x - 3x + 2)(4x + 3) + (2x 2 + 3x -1)( -1/x 2 - 3)

6. (3x + 2/x 2 )(2) + (2x + 4)(3 - 4/x 3 )
(x2 + 3x + 2)(8x) - (4x2 + 4)(2x + 3)
7· (x2 + 3x + 2)2
1
8. 5(2x + 3)4 • 2 9. 2x + 3 . 2
x(4x + 3)
10. log(2x2 + 3x- 5) + 2x 2 + 3x _ 5
3
12. 2 cos 2x - 15 sin 5x - 2 3X
COS
458 Appendix C
Exercises 5.6
1. (a) TC = 4q + 2q
3 2 - 25q
d(TC)
Marginal cost (MC) = dq = 12q2 + 4q- 25
TC
Average cost (A C)=-= 4q 2 + 2q- 25
q
(b) MC = (q 3 - + (16 + 5q)(3q 2
3q)5 - 3)
AC = (q 2 - 3)(16 + Sq)
(c) MC = e2q(6) + 12qe2q
AC = 25 + 6e 2q
q
(d) MC= (q 2 - q) ~ + (3log q + 5)(2q- 1)

AC = (3 log q + S)(q- 1)
2. Elasticity of demand = ddq .f!_
p q
(a) p + 2q = 50 :. dq!dp = ~
and whenp = 10, q = 20
:. elasticity of demand = - (!}(!8) = -!
(b) -2 (c) -6
3. Marginal revenue (MR) is found by expressing total revenue (TR)
in terms of q and finding d(TR)Idq.
(a) TR = pq = 200q - 3q 2 d(TR)Idq = 200 - 6q
(b) TR = pq = 100 d(TR)Idq = 0
(c) TR = pq = 250q - qe4q d(TR)Idq = 250 - e4q(1 + 4q).
4. (a) Marginal cost, MC = dTC/dq = 4q + 20
(b) Total revenue, TR = pq = 0.25(260 - q)q
Marginal revenue, MR = d(TR)Idq = 65 - O.Sq
(c) Profit or net revenue, NR = TR - TC
= (65q - Oo25q 2 ) - (2q 2 + 20q + 300)

= 45q - 2025q 2 - 300
Marginal profits = d(NR)/dq = 45 - 4o5q
(d) MR = MC when 65- Oo5q = 4q + 20 or 45 = 4o5q so q = 100
50 Profit or net revenue is NR = TR - TC = pq - TC
= 177q- q3 - 30- 2q- 2q 2 = 175q- q3 - 30- 2q2
Marginal profit is d(NR)Idq = 175- 3q2 - 4q and this is zero when
3q2 + 4q- 175 = 0 or q = 7 or -8033 and so the answer is q = 70
60 Average propensity to consume:
APC = C/Y = a/Y + b + cY
Marginal propensity to consume: dC/dY = b + 2cYo
Exercises 5.9
1. (a) y = 3x 2 - 120x + 30
dy
-=6x- 120
dx
dy
For a stationary value dx = 0
so 6x - 120 = 0 and x = 20
dzy 0 0 00
dx2 = 6 whtch IS positive
: 0 x = 20 is a minimum value andy= -1170

(b) y = 16 - Bx - x 2
dy d 2y
-=-8-2x -=-2
dx dx 2
dyldx = 0 when x = -4 and there is a maximum value at this
point with y = 32
(c)y=x 4
dy
-=4x 3
dx
460 Appendix C
There is a stationary value when 4x 3 = 0, that is x = 0. This is

in fact a minimum, withy = 0, and the curve is symmetrical
about the y-axis. (Draw the curve for x = - 3 to x = + 3.)
(d) y = (x - 5)3
dy
- = 3(x- 5)2
dx
There is a stationary value when
3(x - 5) 2 = 0 that is when x = 5
at this point d 2yldx 2 is also equal to 0 and it is necessary to
sketch the curve in order to show it is a point of inflexion.
2. c= 500 + 4q + 1q 2
500 q
. . average cost = - - + 4 + -
q 2
d(AC). -500 1
--=--+:z and
dq q2
There is a stationary point when 500/q2 = 1
that is q2 = 1000 and q = ± 31.6
There is a minimum when q = 31.6 because d 2AC/dq 2 > 0 for this
value.
3. Demand function is given by q + 2p = 10
Total revenue = pq = p(10 - 2p)
.. d(TR) = 10 _ 4p d 2 (TR) = _4
dp ' dp 2
This has a stationary value when p = ¥= ~
The second-order derivative is negative and there is therefore a
maximum when p = ~
q=l0-2p=10-5=5
. .
Elasttctty of demand = - - =
dq p <-z)<n = - 1
dp q 5
4. Net revenue = pq- TC = q(240- q)- (q 3 + 20q 2 + 20)

d(NR)
.. - - = 240- 2q- 3q 2 - 40q = -3q 2 - 42q + 240
dq
This has a stationary value when q 2 + 14q - 80 = 0, that is when
q = 4.36 or - 18.35
d 2 (NR)
----:---::--
dq2
= - 6q - 42
This is positive when q = - 18.35 and negative when q = 4.36.

:. net revenue is maximised when q = 4.36 and p = 235.6.
5. (a) Total revenue, TR = pq = (50 - 0.5q)q = 50q - 0.5q 2
Maximum is when d(TR)Idq =50- q = 0 so q =50, provided
that the second-order derivative is negative. Here d 2 (TR)Idq 2
= - 1 so q = 50 maximises total revenue.
(b) Average cost, AC = TC/q = 256/q + 2 + 2q. Stationary values
occur when d(AC)Idq = - 256/q 2 + 2 = 0 so q = 11.3 or
-11.3 The second-order condition for a minimum is that
d 2 (AC)Idq 2 is positive. Since d 2 (AC)Idq 2 = 512/q 3 which is
positive when q = 11.3, this is the value of q which minimises
average cost.
(c) Profit or net revenue is NR = TR- TC = 50q- 0.5q 2 - (256
+ 2q + 2q 2) = 48q - 2.5q 2 - 256 and the stationary value is
when d(NR)Idq = 48 - 5q = 0 or q = 9.6. Here d 2 (NR)Idq 2
= -5 and so q = 9.6 maximises profit.
Notice that to maximise total revenue, q = 50, to mini-
mise average costs, q = 11.3 and to maximise profits, q = 9.6.
It is the last of these that the monopolist should adopt.
Exercises 5.13
1. With price discrimination the firm will maximise profits in each
market.
For its own workers, profit or net revenue is NR 1 = total
revenue - total costs = P1 Q 1 - TC = (500 - 10Q1 )Q 1 - 1000
- 15Q 1 - 15Q2 - 15Q3 and the maximum is when d(NR 1 )/dQ 1
= 500 - 20Q 1 - 15 = 0 or Q 1 = 24.25. That this maximises profit is
shown by d 2 (NR 1 )1dQi = -20. Here P 1 = 257.5.
462 Appendix C
For the domestic market, NR 2 = (100000 - 5Q2 )Q 2 - 1000

- 15Q 1 - 15Q2 - 15Q3 and d(NR 2 )1dQ2 = 100000 - 10Q2 - 15
which is zero when Q2 = 9998.5. The second-order derivative is
negative and so this maximises profit. Here P2 = 50007.5.
For the overseas market, NR 3 = (62500 - 2.5Q3)Q3 - 1000
- 15Q 1 - 15Q2 - 15Q3 and d(NR 3 )1dQ3 = 62500- 5Q 3 - 15 which
is zero when Q3 = 12497. The second-order derivative is negative
and so this maximises profit. Here P3 = 31257.5.
The total profit under price discrimination is P 1 Q1 + P 2 Q2
+ P3 Q3 - TC = 6244.375 + 499999988.8 + 390624977.5 - (1000
+ 15(24.25 + 9998.8 + 12497) = 890292414.4. Notice the wide
divergence of the prices in these three markets and that the total
sales, Q = Q 1 + Q2 + Q3 = 22519.75.
With no price discrimination all the prices are P, say, and the
three demand functions are combined by adding Q 1 + Q 2 + Q3
=(50- 0.1P) + (20000- 0.2P) + (25000- 0.4P) = 45050- 0.7P
= Q. The profit or net revenue is given by
NR = PQ- TC = (64357.14- 1.4286Q)Q- 1000- 15Q.
The stationary value occurs when d(NR)/dQ = 0 or 64357
2.8572Q - 15 = 0 so Q = 22519.25 and the second-order deriva-
tive is negative, indicating a maximum value. Here P = 32186.00
and NR = 724465791.8 which is lower than the price-
discrimination one of 890292414.4. Therefore it pays the monopol-
ist to follow a policy of price discrimination. Also, notice that the
value of Q is (apart from rounding errors) the same in both cases
with linear demand functions. See question 2 below for a demon-
stration of this.
2. With price discrimination, maximising profits in the first market
requires the maximisation of NR 1 = p 1q1 - TC = (a - bq 1 )q 1 - f
- g(q 1 + q2) which occurs when d(NR 1 )1dq 1 = 0 or a - 2bq 1 - g =
0 so q 1 = (a - g)/2b. The second-order derivative is - 2b which is
negative if b > 0 (which requires the demand curve to be down-
ward sloping) and so NR 1 is maximised. From the demand function
p 1 = (a + g)/2.
For the second market, NR 2 = (c - eq 2 )q2 - f- g(q! + q2 )
and when d(NR 2 )1dq2 = 0 then c - 2eq 2 - g = 0 and q2 = (c -
g)/2e. The second-order condition proves that this gives a max-
imum of NR 2 • Here p 2 = (c + g)/2.
The quantity sold with price discrimination is
(a - g) (c - g) ae - eg + be - bg
q = q, + qz = 2b + 2e = 2be
With a common price p, the total demand is

(a - p) (c - p) (ae + be) - (b + e)p
q = q, + qz = b + e = be or
(ae + be)- beq (ae + be)

p = (b +e) A - Bq where A = (b + e) ,
B = bel(b + e) and NR = pq - TC = (A - Bq)q - f- gq

Differentiating gives
~ ~
dq =A - 2Bq - g which is zero when q = 28
The second-order derivative is -28 which is negative (since b

and e are positive) so that the stationary value is a maximum. Here
p =(A + g)/2.
Therefore the quantity sold without price discrimination will
be
(A - g) ae + be - bg - eg
q
2B 2be
which is the same value as with price discrimination.
3. (a) With a flat-rate tax, demand is p + 2q = 250
supply is (p - t) - 4q = 100
and tax revenue is T = tq where tis the flat-rate tax. Equilib-
rium is when q = 0.5(250- p) = 0.25(p- t- 100) or 500- 2p
= p- t- 100 sop= (600 + t)/3 and q = 25- (t/6). Hence, T
= tq = 25t - (t 2/6) and so dT/dt = 25 - t/3 and this is zero
when t = 75 and q = 12.5, p = 225. This is a maximum since
the second-order derivative, d 2 Tidt 2 = -1/3 is negative. The
maximum tax revenue is T = tq = 12.5(75) = 937.5.
(b) With tax of r% the supply function is p(1 - r) - 4q = 100 and
demand is p + 2q = 250. In equilibrium, q = 0.5(250 - p) =
0.25[p(1 - r) - 100] or 500 - 2p = p - rp - 100 so that
p = 600/(3- r), q = 125- [300/(3- r)] = (75- 125r)/(3- r).
The tax revenue is
464 Appendix C
r(600)[75 - 125r) 600[75r - 125r 2]

T = rpq = (3 - r )(3 - r) = (3 - r )2
So,
dT 600{ (3 - r )2 (75 - 250r)- (75r - 125r 2 )(2)(- 1)(3 - r)}
dr (3-rt
Setting to zero: (3 - r )(75 - 250r) = - 2(75r - 125r 2) or
225 - 75r - 750r + 250r 2 = 250r 2 - 150r or 225 = 675r and
r = 113. To check for a maximum the second-order derivative
is required. Now, simplifying,
dT 600{225 - 75r - 750r + 250r 2 + 150r - 250r 2 }
dr (3 - r) 3
600{225- 675r} 135000{1- 3r}
(3 - r) 3 (3 - r) 3
d 2T 135000{(3 - r)3(- 3) - (1 - 3r)3( -1)(3 - r)Z
When r = 113, the sign of this is negative and so the maximum

is when r = 1/3, p = 225, q = 12.5 and the tax revenue is
T = rpq = 937.5. This is the same as for the flat-rate tax case
in (a).
4. (a) In equilibrium, 110 - 3q 2 = 10 + q 2 or 100 = 4q 2 and q = 5
(taking the positive root), p = 35.
(b) With a flat-rate tax oft per unit supply is now p - t = 10 + q2
and demand is p = 110 - 3q 2 • The new equilibrium is when
10 + t + q2 = 110- 3q 2 or q2 = 25- (t/4) and p = 110- 3q 2 =
35 + 3(t/4). Tax revenue is T = tq = t(25 - (t/4) )05 and so
dT t(0.5)(- 0.25)
dt = (25 - (t/4) )05 + (25 - (t/4) )05
Setting to zero, (25 - (t/4)) = 0.125t and t = 66.67. For this

value q = 2.89 and p = 85. To check for a maximum, the
second-order derivative is needed. It is convenient here to use
q 2 = 25 - (t/4)
and to note that differentiating this gives
dq -1
2q dt =4
dq -1 I
so that -=-=q
dt 8q
dT 0.125t
Since -=q---
dt q
d 2T dq {q - tq'} -1 1 t
-=-- =-----
dt2 dt 8q 2 8q 8q 64q 3
which is negative. The maximum is when t = 66.67 and T = tq
= 192.68.
5. Here N = 20000 = annual demand, S = 2000 =set-up cost Q =
batch size (to be determined), i = 5 = inventory cost per unit per
year. The number of batches per year is N/Q. The total annual
cost, Cis given by the sum of total set-up costs times number of
batches and the inventory cost
SN iQ 40000000 5Q dC - 40000000 5
C =-+-= +-and-= +-
Q 2 Q 2 dQ Q2 2
Setting to zero gives Q = 4000 and the second-order derivative is
positive, indicating this is a minimum. The total annual cost is
£20,000.
6. Let R = 50000 be the rate of production. As before, i = 5, N =
20000, S = 2000. The maximum inventory is now Q(l - NIR) (see
the text, p. 204) and the total annual cost is
N SN 40000000
C = 0.5iQ(l- R) +Q= 1.5Q + Q
dC 40000000
and dQ = 1.5 - Q2 which is zero when Q = 5164.
The second-order derivative is positive and so Q = 5164 give the

production run size which minimises the total annual cost. The
total cost is 15492, which is below the value in question 5 since
there is a saving in inventory costs when production is spread over
time, rather than being instantaneous.
7. Here, annual demand, N = 3000 boxes, inventory cost i = 3,
set-up cost S = 500 and the annual production rate is R = 9000
boxes. The average inventory size is 0.5Q(l - NIR) = Q/3 and so
the total annual cost is
C = iQ/3 + SN/Q = Q + 1500000/Q
466 Appendix C
dC 1500000
-=1-
dQ
Q2 and this is zero when Q = 12240 7 0
As the second-order derivative is positive this is the minimising
value and the total cost is 2449060
Exercises 5.19
10 Taylor's theorem states
xz
f(a + x) = f(a) + xf'(a) + 2 !f"(a) + o o o
Let f(a + x) (1 + xt
f(a) 1
f'(a + x) 4(1 + x) 3 and f'(a) = 4
f"(a + x) = 12(1 + xY and f"(a) = 12
f'"(a + x) = 24(1 + x) and f"'(a) = 24
rca+ x) = 24 and r(a) = 24
X2 x' X4
00 (1 + xt = 1 + x(4) + T (12) + 6 (24) + 24 (24)
= 1 + 4x + 6x 2 + 4x 3 + x 4
and 1S = (1 + ~ = 1+ 4 ~ 6 ~ 4 X k+ ft
= 1 + 2 + 1.5 + 005 + 000625 = 500625
20 MacLaurin's theorem states
xz
f(x) = f(O) + xf'(O) + 2! f"(O) + o o o
Let f(x) = (1
+ x}-1
then f'(x) = - (1 + x)-2
f"(x) = 2(1 + x)-3
f"'(x) - 6(1 + x)- 4
Therefore (1 + x)- 1 = 1 - x + x2 - x3 + 000

3. (a) f(x) = X + 4.55x

2 - 8.70
f'(x) = 2(x) + 4.55
Using as a starting point a = 1.5
/(1.5) = 2.25 + 6.825 - 8.70 = 0.375
/'(1.5) = 3.0 + 4.55 = 7.55
f(a) 0.375
·· h1 = - f'(a) = - 7 .55 = -0.0497
.. new approximate root is 1.5 - 0.05 = 1.45 for which

/(1.45) = 2.1 + 6.6- 8.7 = 0
This result shows that 1.45 is a good approximation to one of
the values of x which satisfies the equation.
(b) f(x) = X 3 - 2.34x 2 + 2x - 4.68
f'(x) = 3x 2 - 4.68x + 2
Using as a starting point a = 2.25
f(a) = 11.39 + 4.50- 4.68 = -0.64
- 11.85
f'(a) = 15.19- 10.53 + 2 = 6.66
f(a) 0.64
· · h 1 = - f'(a) = 6 .66 = 0.096
.. new approximate root is 2.25 + 0.096 = 2.346

for which /(2.346) = 0.045. A closer approximation can be
obtained by repeating this process.
(c) x = 4.1 for whichf(4.1) = 0.
70 80 100 100
4. Pvl =If+ R2 +w+RI
60 90 80 130
PV2=R+ R2 +w+RI
these are equal when PV1 - PV2 = 0 that is

10 10 20 30
R- R2 + R3 - R4 = 0
468 Appendix C
or R3 - R2 + 2R - 3 =0
f(R) = + 2R - 3
R3 - R2
f'(R) =
3R 2 - 2R + 2
Using Newton's method with a = 1 as a starting point, we obtain
f(a) 1-1+2-3=-1
f'(a) 3-2+2=3
f(a) 1
- f'(a) = 3 = 0.33
. . approximate root is 1 + 0.33 = 1.33; using this as a new

starting point, we have
/(1.33) = 0.2437
/'(1.33) = 4.6467
/(1.33) 0.2437
hz = - /'(1.33) = - 4.6467 = -0.052
Try R = 1.33 - 0.052 = 1.278 as the next root: /(1.278) = 0.0101
and so the rate of interest is approximately 27.8 per cent. A more
accurate value can be obtained by repeating the process.
5. In each case the places where the function cuts the axes are
required and also where the stationary values are.
(a) y = x 2 + x + 12. This cuts they-axis when x = 0, y = 12, and
the x-axis when 0 = x 2 + x + 12 or
-1 v'(1- 48)
x = 2 so there are no real roots
The stationary values are when

dy -1
- = 2x + 1 = 0 or x = -
dx 2
d2y
dx2- 2
and so there is a local minimum when x = -0.5, y = 11.75.

As extra values to help to position the graph, x = -3, y = 18,
x = 3, y = 24, and the table of values is:
X -3 -0.5 0 3
y 18 11.75 12 24
min
The graph is shown in Fig. C.1.

y
-3.0 -0.5 3.0 X
Fig. C.1
(b) y = x 3 - x 2 • This hasx = 0, y = 0, and also since y = x 2 (x- 1),

when x = 1, y = 0
dy
Here,
dx
The stationary values are when 3x 2 - 2x = 0 and sox = 0 or
x = 2/3. The second-order derivative is negative for x = 0
(a local maximum withy= 0) and positive for x = 2/3 (a local
minimum withy = -0.148). For extra values take x = -1,
y = - 2 and x = 2, y = 4. The table of values is:
X -1 0 2/3 1 2
y -2 0 -0.148 0 4
max min
470 Appendix C

y
-2 -1 2 X
-2
-4
Fig. C.2
(c) y = x 3 - 3x 2 - 24x + 26. This cuts they-axis when x = 0,

y = 26, and the x-axis when y = 0 = x 3 - 3x 2 - 24x + 26. To
find an approximate root, since the cubic has some terms with
negative signs, positive values of x are worth trying. For x = 1,
y = 0 so that (x - 1) is a factor and dividing into the cubic:
x2 - 2x - 26
x- 1lx 3 -3x 2 - 24x + 26
- 2x 2- 24x
-2x2+ 2x
-26x+26
-26x+26
and soy= (x- 1)(x 2 - 2x- 26) and if y = 0, x 2 - 2x- 26 = 0
so x = 6.2 or x = - 4.2
dy d 2y
Here - = 3x 2 - 6x- 24 and-= 6x- 6
' dx dx 2
The stationary values are when 3x 2 - 6x - 24 = 0 and sox = 4

and x = - 2. When x = 4 the second-order derivative
is positive and so the local minimum is y = - 54 and when
x = - 2 the second-order derivative is negative and so the local
maximum is y = 54,
The table of values is:
X -4.2 -2 0 1 4 6.2
y 0 54 26 0 -54 0
max min
75
50
-6 -2 4 X
-25
-50
-75
Fig. C.3
(d) y = x 5 • This goes through the origin (x = 0, y = 0) and does

not cut the axes anywhere else. In attempting to find stationary
points, we get
dy d 2y
- = 5x 4 and-= 20x 3
dx dx 2
and when
472 Appendix C
dy
- = 0 5x 4 = 0 so x = 0
dx '
but for this value the second-order derivative is also zero. This
means that the general rules do not give a clear indication of
whether at x = 0 there is a maximum, a minimum or a point of
inflexion. We therefore work out a table of values centred on
x = 0 to obtain the general shape of the curve:
X -3 -2 -1 0 1 2 3
y -243 -32 -1 0 1 32 243
The graph is shown in Fig. C.4 and it can be seen that there is a
point of inflexion at x = 0.
y
30
15
-2 -1 X
-15
-30
Fig. C.4
Exercises 6.2
dTC
1. (a) dq = 1 TC= q + C
when q = 0, TC= 600

:. TC = 600 + q
(b) TC =50+ q
(c) TC = x 2 + 3x + C
100 = 25 + 15 + c
:. c = 60 and TC = x 2 + 3x + 60.
2. Given C =a+ bY+ c¥ 2 then mpc = ~~ = b + 2cY
Since mpc = 0.1 + 0.02¥, b = 0.1 and c = 0.01
The consumption function is C =a+ 0.1Y + 0.01¥ 2
When Y = 0, C = 0 and so a= 0, giving C = 0.1Y + 0.01¥ 2
dK
3. Here I(t) = dt = a + 2bt + 3ct 2
and K = JI(t)dt = at+ bt 2 + ct 3 + d

Since K = 0 when t = 0, d = 0 and K = at + bt 2 + ct 3
Exercises 6.4
1. (a) TC = x + 3x +
2 10
(b) TC = ~ x 3 + x 2 + 4x + 90
2. (a) ~ x 2 - 4x + C (b) ~ x 3 - 2x 2 + 3x + C
(c) !x 6 + 4/x + x 2 + C (d) -llx 2 - 4logx + C
(e) log (3x - 5) + C (f) log (2x 2 + x - 4) + C
(g) i log (x 4 - 2x 2 + 2) + C (h) 2 log (x 2 + 3x + 1)
(i) ~ e 3x - e-x + c (j) 2ex - ~ e 3x + c
(k) - 2 cos x - 4 sin x
3. Given MR =a- 2bq, TR = JMR dq = aq- bq 2 + c
When q = 0, TR = 0 so that c = 0 and TR = aq- bq 2
Now TR = pq and sop= TR!q =a- bq is the demand function.
p pdq dq
4. Given ed = -0.3- and since ed =-then-= -0.3
q qdp dp
474 Appendix C
Integrating, q = - 0.3p + c
When p = 10, q = 100 = -30 + c and so c = 130
The demand curve is q = 130 - 0.3p
5. Given mpc = 0.5 + y- 1 - Y-2 = dC!dY
Integrating, C = 0.5Y + logY + y-1 + A
When Y = 1, C = 1 = 0.5 + 0 + 1 +A so A = -0.5 and the
consumption function is C = 0.5Y + logY+ y- 1 - 0.5.
Exercises 6.6
J y dx = J2x dx =
x2 ~ 3 3
1. (a) ~ =9
Xt = 0 0
(b) ~ = 36 - 9 = 27
(c) [x 2 ]g = 36
(d) ~ ~ = ~ [64 - 36] = 70
(e) [2x + ~ ~ =6+ ¥= 19!
2. Since MC = 100 - 4q + q\ TC = f2030 MCdq
= [100q - 2q 2 + ~
= (3000 - 1800 + 9000) - (2000 - 800 + 2666.7)

= 10200 - 3866.7 = 6333.3.
3. TR =f 20
15
MR dq = [200q - 3q 2 m
= (4000 - 1200) - (3000 - 675) = 475.
4. pn = ~ 1000e- 005 t dt = 1000[e- 0 ·05t/(- o.o5m0
= -20000{e- 05 - 1} = -20000{0.6065- 1} = 7870.
Exercises 6.8
1. (a) y = 2x crosses the x-axis at x = 0.
:. area = 2 J2x dx =
1
~ =2
0
(b) y =2+ 3x crosses the x-axis at x =-~

-V3 2
0 0 area =- J (2 + 3x)dx + J (2 + 3x)dx
-3 -V3
= _ [2x + J2 x2]-v1
-3
+ [2x + J2 x2]z-V3
=- [-1 + ~ c-6 + 13o5)J + [4 + 6- ~ + m

= 8o16 + 10067 = 18083
3
(c) Area = Jx 2 dx = Bx 3]:2 = ~ + 9 = ~ (as the curve
-2 does not cross the x-axis)
(d) y = 1 + x + x 2 does not cross the x-axis
+I
area = J (1 + x + x )dx 2
-4
= [x + ~x2 + ! ~ = 115/6
20 E(t) = 55 - 16t + t 2 This cuts the t-axis when E(t) = 0 or t = 5 or
o
11. The required area is evaluated in two parts: from t = 0 tot= 5

and from t = 5 to t = 10
J: E(t)dt = [SSt - 8t 2 + Oo3333t 3 ]g
= (275 - 200 + 41.67) - (0) = 116067
J:oE(t)dt = [55t - 8t 2 + ~
= (550 - 800 + 333033) - (116067) = - 33o33
Total absolute error =116067 + 33o33 = 1500
Exercises 6.11
1. When p = 40, q = 100 Consumer's surplus = ~ pdq - pq
= ~ (100 - 6q)dq - 400 = [100q - ~ - 400
= (1000 - 300) - (0) - 400 = 300
20 Equilibrium is when 60- q- q 2 = 10 + 4q so q = 5 or q = -100

For q = 5, p = 30:
area under demand curve is J: pdq = [60q - Oo5q 2 - Oo3333q 3 ]g
= (300 - 1205 - 41.67) - (0) = 245083
consumer's surplus = 245083 - 150 = 95083
476 Appendix C
area under supply curve = J; (10 + 4q)dq = [lOq + 2q ]g = 100 2
producer's surplus = pq - 100 = 150 - 100 = 50

With a tax of 2 per unit supply is now p - 2 = 10 + 4q or
p = 12 + 4q (see section 5.11). The new equilibrium is when 12 +
4q = 60 - q - q 2 or q = 4.865, p = 31.468.
area under demand curve is
[60q - 0.5q 2 - 0.3333q 3]ri· 865 = (291.9 - 11.834 - 38.382)
= 241.684
and the consumer's surplus= 241.684- (4.865)(31.468) = 88.59,
which is smaller than the previous value of 95.83
area under the supply curve is now s:· 865
(12 + q)dq = [12q
+ 0.5q 2 ]ri· 865 = (58.38 + 11.834) = 70.21
producer's surplus is pq - 70.21 = 153.09 - 70.21 = 82.88, an
increase on the previous value.
30
3. MR = 200 - 3q and MC = 2q so profit =

30
f0
(MR-MC)dq =
f 0
(200 - 5q)dq = [200q - ~ = (6000 - 2250)- 0 = 3750
Maximum profit is when MR = MC or 200- 3q = 2q and q = 40
(assuming that the second-order condition is satisfied, which can
easily be checked)
40
f (MR-MC)dq = [200q- ~ = 4000.

J (S(t) -
0 4
4. S(t)- C(t) = 10- 5t + 30t 2 - 4t 3 • Total net revenue=

4 0
C(t)) dt = f (10- 5t + 30t

0
2 - 4t 3 )dt =[lOt- 2.5t 2 + 10t3 - ~=
(40 - 40 + 640 - 256) - (0) = 384
If the campaign is extended from 6 to 8 months the extra net
revenue is
8
f 6
(S(t) - C(t))dt = [lOt- 2.5t 2 + 10t 3 - t 4 ]: = (944) - (834)
= 110
and so it is worthwhile. But this is not optimal since S(7) - C(7) =
73 > 0 while S(8) - C(8) = - 158 < 0. That is, the optimal length
is the value oft for which S(t) = C(t), and is approx. 7.4 months.
5. MR = MCwhen 600- lOt- 3t 2 =50+ 15tor 3t 2 + 25t- 550 = 0
so t = 10 years. Total net revenue is

10 10
f 0
(MR - MC)dt = f 0
(550 - 25t- 3t 2 )dt = [550t- 12.5t2 -
~ = (5500 - 1250 - 1000)- (0) = 3250

T
Total return at time Tis R(T) = J0

(MR - MC)dt + S(T)
= 550T- 12.5P - P + 5000- 300T
= 250T- 12.5P- P + 5000
dR
= 250- 25T- 3P which is zero when T = 5.87
dT
The maximum value is R(5.87) = 5834.53.
Exercises 6.13
3
1. (a) Exact: J 4x1

3 dx = [x4n = 81 - 1 = 80
(b) Simpson's rule: Taking n = 4 and c = 0.5 the table of values is:
X 1 1.5 2 2.5 3
y 4 13.5 32 62.5 108
The approximate area = c[y 0 + 4y 1 + 2y2 + 4y3 + y4]/3

= 0.5 [4 + 54 + 64 + 250 + 108]/3 = 80.
14
2. Here n = 4 and c = (b- a)/n = 1. Total cost = J MCdq 10
:::::: c[MC0 + 4MC1 + 2MC2 + 4MC3 + MC4]/3

:::::: [3 + 16 + 18 + 40 + 15]/3 = 30.67.
3. (a) n = 2, c = (b - a)/n = 4. The table of values is:
X 0 4 8
y 0 256 4096
area""' c[y 0 + 4y 1 + y2]3 = 4[0 + 1024 + 4096]/3 = 6826.7.

(b) n = 8, c = 1. The table of values is:
X 0 12 3 4 56 7 8
y 0 1 16 81 256 625 1296 2401 4096
478 Appendix C
area= c[yo + 4yt + 2y2 + 4y3 + 2y 4 + 4y 5 + 2y6 + 4y 1 + Ys]/3

= [0 + 4 + 32 + 324 + 512 + 2500 + 2592 + 9604 +
4096]/3 = 6554.7.
8
(c) Exact: f 0
X 4 dx = ~ = 0.2(32768) = 6553.6
Here Simpson's rule with n = 8 is close to the exact value.
4
4. Require f I
x- 1dx. Here f(x) = x-t, f'(x) = -x- 2 , f"(x) =2x-\
f'"(x) = -6x- 4 and the integral of Taylor's series is:
4 (x a) 2 P(a) (x a) 3
f I
f(x)dx = [f(a)x + -
2!
1 + - f"(a) +
3!
· · ·t
I
(a) For a= 1,/(1) = 1,/'(1) = -1,/"(1) = 2,/"'(1) = -6

4 (X - 1)2 (X - 1)3 (X - 1)4
f I
f(x)dx = [x-
2
+
3
-
4
+ · · ·t1
= (4 - 4.5 + 9 - 20.25) - (1)
= -12.75
This poor result occurs because the Taylor's series for a = 1 does
not converge.
(b) For a = 3, /(3) = 0.3333, /'(3) = -0.1111,
!"4 (3) = 0.0741, /"'(3) = -0.0741
f I
f(x) = [0.3333x-0.0556 (x- 3) 2 + 0.0124 (x- 3) 3
-0.0031 (x- 3t + · · · ~
= (1.333 - 0.0556 + 0.0124 - 0.0031)
- (0.3333 - 0.2224 - 0.0992 - 0.0496)
= 1.287 - (- 0.0379) = 1.3249.
(c) From section 6.12, the value when a = 2.5 is 1.344 and the
exact value is 1.3863. For a = 1 the result -12.75 is very
inaccurate, while with a = 3, 1.3249 is worse than when a =
2.5. This suggests that the value of a is important and should
be in the middle of the range of x-values. Also, with Taylor's
series it may be necessary to check whether the expansion
converges and include more than just the first three deriva-
tives.
4
5. (a) Exact: f exdx =

0
~ ~ = (e4) - (e0) = 54.598 - 1 = 53.598.
(b) Simpson's rule with n = 4: With c = 1 the table of values is:
X 0 1 2 3 4
y 1 2.718 7.389 20.086 54.598
area = c[y 0 + 4y 1 + 2y 2 + 4y3 + y4]/3

= [1 + 10.872 + 14.778 + 80.344 + 54.598]/3 = 53.864.
(c) By Taylor's series: f(x) = ex= f'(x) = f"(x) = f"'(x) etc.
Taking a= 2,/(2) = 7.389 = ['(2) = ["(2) = f"'(2) etc. and this
can be taken outside the brackets:
4 (x a) 2 (x aY
J f(x)dx
0
= [f(a)x + -
2!
f'(a) + -
3!
f"(a) + · · · ~
(x - 2y (x - 2y (x - 2y (x - 2) 5 4
= 7.389 [x + 2! + 3! + 4! + 5! ]o
= 7.389 {(4 + 2 + 1.333 + 0.6667 + 0.2667)

- (2 - 1.3333 + 0.6667 - 0.2667)}
= 7.389 (8.2667 - 1.0667) = 53.2
Here Simpson's rule gives a better approximation than does
the Taylor's series result, but adding extra terms to the
Taylor's series expansion will improve its accuracy.
Exercises 6.15
1. Let (5r + 4) = u and e3xdx = dv
Then J (5x 2 + 4)e3xdx = (5x 2 + 4)je3x- J-

10xe1x
3- dx
and J 10xe1xdx = 10x(je x) - J-10e3x

3- dx
3
and J 10e3xdx = 10 {j e1x)

480 Appendix C
5x2 + 4 lOx 10]

.. f (sr + 4) ~ = ~[ 3 - 9 + 27 + c.
2. Let log x = u and x 2 dx = dv
Then fr log X dx = f log X X 2 dx
= log x(Jx3 ) I -
Jl3r-1 dx
X
~ X 3 log X - ~ f r dx
= HX 3 log X - ~ r] + c
= ~ x 3 [log x - ~ + C.
3 A B
3. Let ----=--+--
(x + 1)(x - 1) (x + 1) (x - 1)
Then 3 = A(x - 1) + B(x + 1)
and 3 =-A+ B
O=A+B
:. ~ B=~
3 3
3
:. J(x + l)(x - 1) dx
= f (x -2+ 1) dx + f (x-2 1) dx
= ~ log (x + 1) + ~ log (x- 1) + C
Hlog (x - 1) - log (x + 1)] + C
3 (x- 1)
= -log--+ C.
2
(x + 1)
x A Bx + C
4. Let ----- = -- + --
(x- 2)(r- 3) (x- 2) (x 2 - 3)
then x = A(x2 - + (Bx + C)(x - 2)
3)
= Ax2 - 3A + Bx2 + Cx - 2Bx - 2C
A+ B = 0
-2B + C = 1
-3A- 2C = 0
and A = 2, B = - 2, C= - 3
··I (x - X
2)(x2 - 3)
dx=J 2
(x - 2)
dx-J2x+3dx
(x 2 - 3)
1
= 2 log (x- 2) -log (r- 3)- 3 Jr- 3 dx
=
lo [(x- 2)2] + _3_lo ( v'3 +
g (x 2 - 3) 2v'3 g v'3 - x
+D x)
This can be further simplified if necessary.
5. Let 4x + 3 = u; then du = 4 dx
:. J(4x + 3) 10 dx = J-
u10
4
u 11
du = - + C =
44
( 4x + 3) 11
44
+C.
6. Let log x = u and dx = dv

Then Jlog xdx = (log x)x- Jx-1 dxX
=X log X- X+ C
=X (log X - 1) + C.
Exercises 7.4
(Jz
1. (a) - = y + 2xy + y2
(Jx
(Jz
-=x+x2 +2xy
(Jy
(Jz
ax = 6xy + 4y + 6y
(b) - 2
az 2
-=3x +8xy+6x
ay
az
(c) - = 2x sin y - y2 sin x
ax
az
- = x 2 cos y + 2y cos x
ay
482 Appendix C
az
(d) - = (x + y)ex+y + ex+y
ax
iJz
- = (x + y)ex+y + ~
ay
az 2x
(e)-=--
ax X + y 2
2
az 2y
-=--
ay x2 + y2
az (x-y)-(x+y) -2y
(f) ax= (x - YY = (x- Y) 2
az (x- y) - (x + y)(- 1) 2x
ax (x- YY (x- YY
(y +_ sinx)-
(g) -az = ....;;._ (x + siny) cosx
__;__ __;__ __;__
ax (y + sin x) 2
az (y + sinx) cosy- (x + siny)
ay (y +sin xY
2. Marginal product of capital is
aq = 4(K - 30) + 2(L - 20)

aK
= 80 + 20 = 100 when K = 50, L = 30
Marginal product of labour is
aq = 6(L - 20) + 2(K- 30) = 60 + 40 = 100

aL
when K =50 and L = 30.
3. (a) Partial elasticity of demand for apples with respect to price of
apples is
aq P. = (- 2pa- Po) P. = 14(5)
---=
70
apaq q 219 219
when Pa = 5, Po = 4 and q = 219.
(b) Partial elasticity of demand for apples with respect to price of

oranges is
iJq Po (6- p.)Po 4
--=
dp 0 q q 219
when Pa = 5, p 0 = 4 and q = 219.
iJQ aQ iJQL
4. Since Q = AUKfl t h e n - = - and s o - - = a.
iJL L iJLQ
5. Let w = a 11 x 1 + ~ and z = a21 X 1 + ~ so that the first-order
partial derivatives are
/J/ = = /A/ as required.
Exercises 7. 7
1. dq = aq dL + aq de
aL ac
= (3£2- 3 + 4C)dL + (4L- 2C)dC
= 317(1) + 30(1) = 347 when L = 10 and C = 5
If dC = 0 and dL = 1, dq = 317.
aq aq
2. dq = -dpa + - dp 0
dpa dpo
= (-2pa- Po) dpa + (6- Pa) dpo
= -14(1) + 1(1) = -13.
dz azdx azdy
3. - = - - + - -
dt ax dt ay dt
dx dy
(a) - = e'(-sin t) + e' cost, - = e' cos t + e' sin t
dt dt
dz
- = 2x(e')(cos t- sin t) + (-2y)(e')(cos t +sin t)
dt
484 Appendix C
dx dy
(b) dt = 2te' + t 2 e', -=cost
dt
dz = (-1-)(2t + t 2 )e' + ( -1-) cost

dt X+ y X+ y
dx dy 1
(c) - = e', - =-
dt dt t
-dz = 2
(2xyex )e' + (ex)2(1)
-
dt t
dx 1 dy 1
(d) - = 1 - _, - = 1 + -2
dt t 2 dt (
~~ = (2x+ y)( 1- t12)+ (x + 2y)( 1 + t12).

4. (a) Since Q, = ~
then QL = ~ = aQ!L and QK = f3Q!K
so MRS = QL!QK = aK!f3L which is unaffected by r.
(b) From above, (K!L) = ({3/a)(MRS) and so the elasticity of
substitution is
d(K!L)I(K!L) ({3/a) MRS
a = = = 1 as before
d(MRS)/MRS KIL
Therefore technical progress (of the form given in the question)
does not affect the elasticity of substitution with the Cobb-Douglas
production function.
5. Since C =a+ bY, Cy =band as I= c- er, Ir = -e
(a) Taking the total derivative of Y = C + I + G,
dY = Cy dY + Ir dr + dG
= b dY - e dr + dG (1)
Since G is fixed, dG = 0 and (1) can be written
dY -e
(1 - b)dY = - e dr and so- = which is negative.
dr (1 -b)
(b) Taking the total derivative of Ml P = fY - gr gives
dM dP
-
p
- M-
p2
= f dY - g dr
(2)
Now P and rare fixed and so dr = 0 = dP. Hence (1)
dM
becomes (1- b)dY = dG and (2) is-p = fdY
dM fP dY
fP
Therefore, -= =--·
dG (1-b)dY (1-b)
Exercises 7.9
dy (iJz!iJx)
1. - = - - - -
dx (iJzliJy)
dy (-8xy + 12y + 4)
2
(a) - = - ...;____.;.__ __.;.__
dx 3y 2 - 8x 2y + 6x2
dy (2xy 2 + e')
(b) dx =- 2x2y - &'
dy [(3/(x + y)] + 8x
(c) - =- .::..c...__;__-=--.:...;:.___
dx [(3/(x
+ y)]- 2y
dy [cos (x + y)- 3y- 10y2 ]
(d)-=- .
dx cos (x + y)- 3x- 20xy
2. When p = 2, q = 12.18.
By differentiating term by term
dq (2 + p)(- 2p)- (500- p 2 ) dq
2q dp = (2 + p y + p dp + q
dq 1 [ - 4p - 500 - p 2 ]
Hence-= +q
dp (2q - p) (2 + p y
and elasticity when p = 2 and q = 12.18 is
!!_dq =- 2-(- 1 -) [ - 512 + 12.18] = -0.15.

qdp 12.18 22.36 16
486 Appendix C
3. Differentiating Q2 = BKL - 3Q + V + 2K 2 with respect to

L: 2QQL = BK - 3QL + 2L
and QL = (8K + 2L)/(2Q +3)
Similarly for K, 2QQK = BL - 3Qx + 4K
and Qx = (8L + 4K)/(2Q + 3)
The marginal rate of substitution is
MRS = _QL = (BK + 2L) = _4_(K_IL_)_+_1 (1)
QK (8L + 4K) 2(KIL) + 4
For the elasticity of substitution (a), we need d(K/L)/d(MRS),and
so rearranging ( 1),
[2(K/L) + 4] MRS= 4(K/L) + 1
Differentiating term by term, with respect to MRS,
[2(K/L) + 4] + MRS[2d(KIL)Id(MRS)] = 4 d(KIL)/d(MRS)
. d(K/L) 2(KIL) + 4
and solvmg = -----
' d(MRS) 4 - 2MRS
d(K/L) MRS [(KIL) + 2] MRS
Therefore, a= =------
d(MRS) KIL [2- MRS] KIL
[(K/L) + 2][4(K/L) + 1]
= 7(KIL)
Exercises 7.12
1. Replace x andy by tx and ty and show f(tx, ty) = tkf(x, y) where k
is the degree of homogeneity: (a) k = 3; (b) k = -1; (c) k = 3.
2. Euler's theorem: xzx + yzy = kz
For (a), Zx = 3x2 - 2xy, Zy = 9y2 - x 2
hence x(3x 2 - 2xy) + y (9y 2 - x 2 ) = 3f(x, y)
( 4 x 3 )y - 12x2 ( xy + y2) x+2y
For (b), Zx = (4 rY ' zy =--
4x 3
hence xzx + yzy = - f(x, y).
3. Own-price elasticity is
poq P
- - = - (4r - 4p)
qop q
Amwers to Exercises 487
Cross-elasticity is
rCJq r
- - =- (20r + 4p)
qCJr q
Sum is 2, the degree of homogeneity.
4. Replacing L and K by tL and tK, the first gives t 0 · 8q and the second
gives tuq as required.
5. The partial derivatives of q = p 2Yipi are
CJq Y CJq -2p2Y dq P2
- = ·~
CJp2 ~ CJp, Pi CJY Pi
and so the elasticities are
p 2CJq p 2Y p,CJq -p,2p 2Y YCJq Yp 2
-!I- = - =2
1'- - = - - -3 = - 2 -~- = - = 1
qup 2 qp 1 !I
q up 1 q p1 qu qp 21
Notice that the sum of the elasticities is zero
The degree of homogeneity is found by multiplying each variable
tpztY
by t to give - - = t 0 q and so is zero
(tp,y
Euler's theorem states that, for q = ~ p2 , Y)
pJ, + Pzfz + Yfy = kq
where k is the degree of homogeneity (zero here) and f, and f2
indicate partial derivatives with respect to p 1 and p 2
Substituting in,
(-2p2Y) + + Y (pz) = - 2q + +q=0
p, 3 P2 (Y) 2
q
p, p,2 p,
as required.
6. For q = ALaKfJ, qL = aq!L and qK = {3q!K so that the slope of an
isoquant at (U, K 0 ) is- qL/qK = - aK 0/{3L 0 , and at (kL 0 , kK 0 ) it is
- ~ · = - aK 0 /L 0{3. Therefore this Cobb-Douglas produc-
tion function is homothetic.
Exercises 7.14
1. (a) z = x 2 - xy + y 2
Zx = 2x- y Zy -x + 2y
488 Appendix C
Zxx =2 Zyy = 2
Zxy = -1
(b) z = (x + y) 3
Zx = 3(x + Y) 2 zy = 3(x + y) 2
Zxx = 6(x + y) Zyy = 6(x + y).
Zxy = 6(x + y)
2. Find when Zx = 0 ·= Zy and check for
Maximum: ZxxZyy > (ZxyY, Zxx < 0, Zyy <0
Minimum: ZxxZyy > (ZxyY, Zxx > 0, Zyy >0
Saddle point: ZxxZyy < (zxyY, Zxx = 0 = Zyy
(a) Zx = 0= zy when x = 0, y = 0 and x = ~ y = ~

Zxx = 6x, Zyy = 2, Zxy = 2
For x = 0, y = 0, tests fail - further investigation needed.
2 2 • . • 4
For x =-3' y = --3' mtmmum w1th z =- -
27·
(b) zx = 0 = zy when x =! andy=!. Maximum value is z = 1.
3. Method is as in Problem 2 above.
(a) Zx = 0 = Zy when x = 2 andy = 0.5.
Zxy = 0, Zxx = 2, Zyy = 2 so minimum value of z = 45.75.
(b) Zx =0 = Zy when X = ± v'(0.67), y = 1.5.
Minimum is z = 196.7 when x = v'(0.67), y = 1.5
(c)
Zx = 0 = Zy when x = 3, y = 4 and x = -3, y = 4.
Minimum is z = 180 when x = 3, y = 4.
4. Here q = 36K + 120L- 3K 2 -4£2- 10 + 2KL
and so qL = 120- 8L + 2K
qK = 36 - 6K + 2L
and setting these to zero and solving, K = 12, L = 18. Also, qKK =
-6, qLL = -8, qKL = 2 so that the conditions for a maximum are
satisfied. The maximum is q = 1286.
5. Profits, :Jr = total revenue - total costs
= p.q. + p2q2 - (qi + 3q.q2 + q;)
and substituting from the demand functions for p 1 and p 2 ,
:Jr = 40q. - qi + 60q2 - 2q; - qi - 3q.q2 - q;
= 40ql - 2qi + 60q2 - ~ - 3qlq2

The first-order derivatives are
1r1 = 40 - 4q 1 - 3q2 and lrz = 60 - 6qz - 3qt
Setting these to zero gives q 1 = 4, q 2 = 8 and the second-order
derivatives are Jr11 = -4, Jr22 = -6, 1r12 = -3 and so we have a
maximum value of profits of 320.
6. (a) With price discrimination, profits,
1r = Ptqt + Pzqz - TC
= 105ql - 25qi + 50q2 - ~ - 10 - 5qi - 2q2
= 100ql - 25qi + 48q2 - ~ - 10
The first-order derivatives are 1r1 = 100 - 50q 1 and 1r2 = 48 - 4q2
and setting these to zero gives q 1 = 2, q2 = 12
Here 1r11 = -50, 1r22 = -4, 1r12 = 0 and so we have a maximum for
which 1r =378. The prices are p 1 = 55 in the peak period and p 2 =
26 in the off-peak period.
(b) With no price discrimination the demand functions are p = 105
- 25q 1 and p = 50 - 2q2. From these we obtain q 1 = 4.2 -
0.04p and q2 = 25 - 0.5p
Profits are given by 1r = p(q 1 + q2) - TC and so it is convenient
to substitute for q 1 and q2 in TC to get everything in terms of p:
TC = 10 + 5(4.2 - 0.04p) + 2(25 - 0.5p) and so
1r = p(29.2- 0.54p) - 10- 21 + 0.2p- 50+ p
= 30.4p - 0.54p 2 - 81
To maximise this we set d1rldp = 0, giving 0 = 30.4 - 1.08p or
p = 28.15 and check that d 21rldp2 = -1.08 is negative as required.
With p = 28.15 the profits are 346.9 and q 1 = 3.074, q2 = 10.93.
With price discrimination, 1r = 378, q 1 = 2, p 1 = 55, q2 = 12 and
P2 = 26.
Exercises 7.18
1. Using a Lagrangian multiplier, as in section 7.16,
z' = 100x 1 - 2xi + 60x2 - ~ + A.(x1 + X2 - 41)
Setting z; = 0, ~ = 0, ~ = 0 gives X 1 = 20.33
x 2 = 20.67 which are the maximising values, z = 2019.3 (compare
z = 2000 when the constraint was X 1 + x 2 , = 40, and A. = - 20).
490 Appendix C
Alternatively, substitute X 1 = 41 - x 2 in the expression for z and

maximise as in section 5.7.
2. For the unrestricted case, Zx = 0 = zy gives x = 2, y = 7 and z = 89
as the minimum value (see section 7.13 for method).
For the restricted case, using a Lagrangian multiplier, the first-
order partial derivatives are zero when x = 10/7, y = 43/7, and the
minimum value of z is 635/7.
3. (a) Using a Lagrangian multiplier shows that the first-order partial
derivatives are zero when x = 200/41, y = 160/41, and the
minimum value of z is 1600/41.
(b) Using a Lagrangian multiplier shows that the first-order partial
derivatives are zero when x = 48/17, y = 36117, and the
maximum value of z is 162/17.
(c) Using a Lagrangian multiplier shows that the first-order partial
derivatives set equal to zero give
0 = 6xy 2 + A., 0 = 6x 2y + 3A., 0 = x + 3y - 18
These are satisfied by three pairs of values:
x = 0, y = 6, minimum value z =0
x = 18, y = 0, minimum value z =0
x = 9, y = 3, maximum value z = 2187
4. The constraint is 4/ + 5c = 28. Using a Lagrangian multiplier, we

find that the first-order partial derivatives are zero when I = 93/49
and c = 200/49. The maximum value of x is 144.0.
5. The budget constraint is 100 = 5x + lOy and so let z' = 6xy +
A.(lOO- 5x - lOy). The partial derivatives are z: = 6y + SA., ~ =
6x + lOA., ~ = 100- 5x- lOy and setting these to zero gives x =
2y so that x = 10, y = 5 and A = 6. To check whether this is a
maximum, the Hessian is, with z = 6xy and u = -5x- lOy+ 100,
Zxx Zxy Ux 0 6 -5
Zyx Zyy Uy 6 0 -10 = 600
Ux Uy 0 -5-10 0
and as this is positive, x = 10, y = 5 gives a maximum with U
= 6xy = 300.
Exercises 8.8
1. The situation shown in Table 8.4 can be expressed as
2x, + 3x2 ~ 25
4x 1 + X2 ~ 35
where x, and x 2 are the numbers of units of product A and B which
are manufactured. x, and X 2 must both therefore be ~ 0
The contribution is given by 9x, + 7x 2 •
TABLE 8.4
No.
of hours Total
required no.
per unit of hours
A B
Machine U 2 3 25
Machine V 4 1 35
Contribution
Slack variables x 3 and x 4 are inserted in the equations in order to

obtain a first feasible solution. The Simplex tableau is as shown in
Tableau 1.
Tableau 1
x, Xz x3 x4 p p;la;i
x3 2 3 1 0 25 25/2
1 departing
x4 4 0 1 35 35/4 ~
z -9 -7 0 0 0
t
entering
variable
492 Appendix C
Using the method of row and column operations successive

tableaux are obtained as shown in Tableaux 2 and 3.
The final tableau shows that the maximum contribution is
obtained by manufacturing 8 units of product A and 3 units of
product B, in which case z = 93.
Check 8 X 9 + 3 X 7 = 72 + 21 = 93.
Tableau 2
XI Xz x3 x4 p pla;j
5 I 15 departing
x3 0 2 1 -2 2 3 -variable
I I 35
XI 4 0 4 4 35
19 9 315
z 0 4 0 4 T
i
entering
variable
Tableau 3
XI Xz x3 x4 p
2
Xz 0 1 5
I
5 3
I 3
XI 0 -TO TO 8
19 26
z 0 0 TO 20 93
The value of increasing capacity on machine U by one unit is

19/10 and that on machine V is 26/20. These values are obtained
directly from the bottom line of the final tableau. They can be
verified by using the simplex method with the constraints adjusted
by one unit of capacity for machines U and V separately.
2. (a) (i) A = 9i, B = 2. (ii) A = 11, B = 1
(b) A = 0, B = ~ (c)A = 7!, B = 5
3. (a) Let x 1 = number of units of A produced
x 2 = number of units of B produced
then the problem is to max z = 3x 1 + 2x 2
subject to X1 + 4x 2 ~ 60 (labour)
2x 1 + X 2 ~ 40 (land)
3x 1 + X 2 ~ 50 (machines)
and X 1 ~ 0, x 2 ~ 0, x 3 ~ 0
Introducing three slack variables, X 3 , X4 and x 5 the con-
straints are
X1 + 4x 2 + X 3 = 60
2xl + X2 + X4 = 40
3x 1 + X2 + X 5 = 50
The first tableau is:
XI x2 x3 x4 Xs p
x3 1 4 1 0 0 60
x4 2 1 0 1 0 40
Xs 3 1 0 0 1 50
z -3 -2 0 0 0 0
The first feasible solution is x 3 = 60, X 4 = 40, x 5 = 50

The entering variable is x 1 and the departing variable x 5
Dividing row 3 by 3 and combining the new row 3 with the
other rows gives:
0 1113 1 0 -113 130/3

0 113 0 1 -2/3 20/3
1 113 0 0 1/3 50/3
z 0 -1 0 0 1 50
The entering variable is x 2 and the departing variable x 3

Dividing row 1 by 1113 and combining the new row with the
other rows gives:
494 Appendix C
xi x2 x3 x4 Xs p
x2 0 1 3/11 0 -1/11 130/11

x4 0 0 -1/11 1 -21/33 90/33
xi 1 0 -1/11 0 12/33 140/11
z 0 0 3/11 0 10/11 680/11
Since all the coefficients on the bottom row are positive the
solution is X 1 = 140/11, X 2 = 130/11 and z = 680/11.
(b) The slack variable for land is x 4 and from the bottom row of
the final tableau above the shadow price is zero. Therefore
there is some land not being used (check the land constraint
2x 1 + x2 :::: 40). It is not worthwhile renting extra land since it
would not increase profits.
4. (a) Let u 1 , u2 and u3 be the number of units of bread, cheese and
meat consumed. The total cost is z = 2u 1 + 4u 1 + 5u3 which is
to be minimised subject to the constraints:
energy: u 1 + 2u 2 + u 3 ;a, 10
protein: U1 + 3u2 + 6u3 ;a, 15
and U 1 ;a, 0, U2 ;a, 0, u3 ;a, 0.
(b) The dual problem is to max z' = 10x1 + 15x2 subject to
2x 1 + 3x2 :::: 4
xi + 6x2:::;; 5
and X 1 ~ 0, X 2 ~ 0. Introducing three slack variables x 3 , X4
and x 5 , the first tableau is
xi X2 x3 x4 Xs p
x3 1 1 1 0 0 2
x4 2 3 0 1 0 4
Xs 1 6 0 0 1 5
z' -10 -15 0 0 0 0
The first feasible solution is X 3 = 2, X 4 = 4, x 5 = 5

The entering variable is x 2 and the departing variable is x 5
Dividing row 3 by 6 and combining the new row with the other
rows:
x, Xz x3 x4 Xs p
x3 5/6 0 1 0 -1/6 7/6

x4 3/2 0 0 1 -112 3/2
Xz 1/6 1 0 0 1/6 516
z' -15/2 0 0 0 5/2 12.5
The entering variable is x, and the departing variable is X 4

Dividing the second row by 3/2 and combining the new row
with the other rows:
Xz p
0 0 1 -5/9 1/9 113

1 0 0 2/3 -1/3 1
0 1 0 -1/9 2/9 2/3
z' 0 0 0 5 0 20
The coefficients on the bottom row are positive and so the

solution is x, = 1, x 2 = 2/3 and z' = 20. The only positive
shadow price is for x4 , the slack variable associated with the
price of cheese, and it would be advantageous to buy more
cheese if the price is less than 5.
Exercises 9.4
1. Let Y = VX, so that dY = V dX + X dV
Substituting and rearranging gives
(a,V- b, - a2 V 2 + b2V)dX = (a 2VX- b7)dV
dX (a 2V- b2 )dV
or -=
X a,V- b, - a2 VZ + b2V
which can be integrated given the values of a" a2 , b, and b 2 •
496 Appendix C
2. Let q be the quantity demanded at price p. Then we require

pdq
--= -n
qdp
dq dp
or -= -n-
q p
Integrating, we obtain
log q = - n log p +A' or q = Ap-n
Thus qp" = A is the general form required.
3. Put y = vx, dy = v dx + x dv; then, on re-arranging,
v dv dx
--
24- 2v 2 x
or - i log (24 - 2v 2 ) = log x + A'
This becomes
A = (24 - 2v 2 )x4 = 24x 4 - 2x2 y 2
and using x = 2, y = 4 gives A = 256.
bx 2
4. Separable variables: c = ax - - +A .
2
dp dp
5. (a) - = kap or - = ka dt
dt p
Integrating gives
log p = kat + A' or p = Aekac
dp , = k dt
(b) - = k(ap + bp-) or dp
dt p(a + bp)
Using partial fractions (see Section 6.14), we have
dp b dp =k dt
ap a(a + bp)
1 1
or -log p- -log (a+ bp) = kt + A'
a a
P = Ae"k'
(a+ bp) ·
6. Substituting gives
dY dY
b- = aY or b - =a dt
y
dt
Hence, b log Y = at + A' or Yb =Ae"'.
7. (a) x dy = (x- y)dx
Put y = vx, dy = v dx + x dv, then (-1 + 2v)dx = - x dv
dx dv
or -+--=0
x 2v- 1
so log x + ! log (2v - 1) = A' or 2xy - x 2 = A.
(b) Let X= y + 1 andY= 2y + x; then
dY dY/dx 2(dyldx) + 1 2(X/Y) + 1
dX dX/dx dyldx XIY
and X dY = (2X + Y)dX
Let Y = VX, dY = V dX +X dV
Hence 2 dX = X dV and 2 log X = V + A
2y +X)
Thus 2 log (y + 1) = ( - - + A·
y + 1
(c) Rewriting as
dy
- + 2y = x(l + x)
dx
we have e2x as an integrating factor and the integral of the left
hand side is ye2x. The right hand side is
Je2x (x + x2)d.x = (x +2x )e2x - J2(1

e2x2
+ 2x)dx
(x + x 2)e2x (1 + 2x)e2x e2x

= - + J-(2)dx
2 4 4
498 Appendix C
Hence the solution is
dy
(d) If -+ 4y = x 3
dx
then e4x is an integrating factor and integrating gives
3x 3 ]2
Jx3e4x dx = e4x [x4 -M 3x
3
ye4x = + 32- 128 +A.

8. Since the market clears, if = qd and so
dp dp
-a+ bp = e- fp + - o r - = (b + f)p -(a+ e)
dt dt
or dp = dt
(b + f)p - (a + e)
For convenience, multiply each side by (b + f) and integrating
gives
log [(b + f)p - (a + e)] = (b + f)t + log A
or (b + f)p - (a + e) = Ae1b+J>•
and so p = [(a+ e)+ Ae 1b+n•]f(b +f)
Exercises 9.6
In each case put y = e= etc. and form the reduced equation am 2 +
bm +c. The roots are m 1 and m 2
1. m 1 = 4, m 2 = -3, y = k 1e4k + k 2e-3x
2. m 1 = 1, m 2 = 2, y = ~ + k 2e'
3. m 1 = 3, m 2 = 3, Y = (kl + ~
4. m 1 = 1, m2 = 1, y = (k! + kzX)e'
5. m 1 = 4, m 2 = -4, Y = k!e4x + kze-4x
6. m 1 = 4i, m 2 = -4i, y = k 3 cos 4x + k4 sin 4x
= A cos (4x - E)
7. m 1 = -1 + 3i, m2 = -1 - 3i, y = e-x (k3 cos 3x + k 4 sin 3x)

= Ae-x cos (3x - E)
Exercises 9.8
1. (a) Complementary function: m 1 = 8, m2 = 2
y = klesx + k2e2x
Particular integral: try y = K1 + ~
then y = 5/64 + x/8
Solution is
5 X
Y = k esx + k 2e2x + - + -
I 64 8
The initial conditions give k 1 = 0.1979, k2 = -0.3542
(b) Complementary function [from Exercise 9.5, Question 5],
y = kle4x + k2e-4x
Particular integral: try y = aeX, then y = - ex/15.
Solution is
ex
Y = -15
- + k e4x + k e-4x
I 2
The initial conditions give k, = t k2 = !.

(c) Complementary function [from Exercise 9.5, Question 4],
Y = (k1 + k 2 x)ex
Particular integral: try y = ae2 x; then y = e2 x
Solution is y = (k 1 + k 2 x)ex + e2x
and the initial conditions give k 1 = 1, k2 = -1.
(d) Complementary function: m 1 = -1 + 2i, m2 = -1 - 2i:
y = Ae-x COS (2x - f)
Particular integral: try y = ax + b; then y =:!:- 2

5 25
Solution is y = :!: - 2 + Ae-x cos (2x - f)
5 25
The initial conditions give A cos (- f) = 0
1 = - 2A sin (- f)
500 Appendix C
and hence f = - ~ and A = - !.

(e) Complementary function: m1 = 0, m = 1 2
y =A+ Bex
Particular integral: try y = a cos x + b sin x; then
y = -cos x - sin x.
Solution is y = A + Bex - cos x - sin x
Initial conditions give A = -1, B = 3.
2. Solve homogeneous part by putting y = tt= etc; then m 1 = 2,
m2 = -3 and
y = Acx + Be-3x
Particular integral: try y = a + bx + cx2 ; then
a= -1/6,b= -1,c= -3.
Solution is y = Acx + Be-3x- {t) - x- 3r
Initial conditions give A = -h\, B = fs.
Exercises 9.12
1. Differentiating (1),
(4)
and using (2) and (3),
dp = be(m- p) + fk(p- p")

dt
Differentiating:
dzp = -be dp + fk dp - fk dp"

dt 2 dt dt dt
and using (4) to eliminate dp"ldt,
dzp =-be dp + fk dp- k{dp + b df/l

dt 2 dt dt dt ii f
Using (3) to remove dU/dt and simplifying,
d 2p dp
- + (be + k(1 - [])- + bekp = bekm
dt 2 dt
which reduces to (6) in the text when f = 1.

d2 d
2. (a) lf = qdgivesf_E + cJ!. +(a+ h)p = b + g
dt 2 dt
For the homogeneous equation, putting p = e"'' gives
fm 2 + em + (a + h) = 0 and the roots are complex, giving a
cyclical solution, if c2 - 4(a + h) f < 0.
(b) Putting the values into the equation,
d2 d
_p_ + 8J!. + 15p = 1800
dt 2 dt
and the auxiliary equation is m 2 + 8m + 15 = 0 which has the
roots - 3 and - 5. The particular integral is from putting
p = K and K = 1800/15 = 120 and so the solution is
p = Ae- 3' + Be- 5' + 120. The condition t = 0, p = 130 gives
130 =A + B + 120 or A + B = 10. The condition t = 0,
dp/dt = -2 gives -2 = - 3A - 5B. Solving, A = 24,
B = -14 and p = 24e- 3' - 14e- 5 ' + 120. As t increases the
value of p quickly approaches 120.
Exercises 10.7
1. The general solution of Y,+, = AY, + K is
K
y =AA'+--
' 1- A
where A is a constant
(a) A= 1.5, K = 3, Y, = A(1.5') - 6
Initial condition gives A = 8
0 2 3 4 5
2 6 12 21 34.5 54.75
(b) A= 0.9, K = 2, Y, = A(0.9') + 20

Initial condition gives A = - 17
0 1 2 3 4 5
3 4.7 6.23 7.61 8.85 9.96
502 Appendix C
(c) A= 1, K = 0, Y, =A.
0 1 2 3 4 5
Y, 6 6 6 6 6 6
(d) A= 1.1, K = 4, Y, = A(l.1') - 40

0 1 2 3 4 5
Y, 1 5.1 9.61 14.57 20.03 26.03
2. Let Y, be the value of the savings after t years; then Yo = 15,

Y 1 = 1.05(15) + 15 and in general Y,+ 1 = 1.05 Y, + 15. The solu-
tion is Y1 = A(l.OS)' - 300 and Y0 = 15 gives A = 315.
y21 = 535.8.
3. The difference equation is p, = -l.5p 1_ 1 + 3 and the solution is
p, = - ~ 1.5)' + 1.2, which is 'exploding' since 1.5 > 1.
4. The difference equation is p 1 = 0.5p,_ 1 + 3 with the solution
p, = - (- 0.5)' + 2, which is stable since 0.5 < 1. The equilibrium
price is 2.
5. The difference equation is p, = - p,_ 1 + 8/3 with the solution
p 1 = [2 (- 1)' + 4]/3 which oscillates since I-ll = 1.
6. The difference equation is 0.9Y, = 0.3Y,- 0.3¥,_ 1 or
Y 1 = -0.5¥,_ 1 , and the solution is Y, = 100(-0.5)'.
7. Substituting into the pricing equation:
Pt+l = Pt - 0.3(0.4p, - 50 - 150 + 0.6p,)
= 0.7p, + 60
The solution is p, = A(0.7') + 60/(1 - 0.7) = A(0.7') + 200
As t ~ oo, p converges on the value 200.
8. Substituting into the first equation:
Y, = b Y,_ 1 + c + a(Y, - Y,_ 1)
or _ (b- a) y
Yt - _c_
t-1 +
(1 - a) (1 -a)
For stability, the coefficient on Y,_ 1 has to lie between 1 and -1 or
-1 < (b - a)/(1 - a) < 1.
Exercises 10.12
1. Difference equation is Y, = (c + b) Y,_ 1 - bY,_ 2 and putting
Y, = )./gives the characteristic equation A2 = (c + b) A - b
(a) Roots are A1 = 0.7 + 0.1i, A2 = 0.7 - O.li
(b) Roots are A1 = 0.8, A2 = 0.8
(c) Roots are A1 = 0.845, A2 = 0.355.
2. Solve homogeneous part by putting Y, = A, and obtaining the roots
A1 and A2 of the characteristic equation. The equilibrium or particu-
lar solution is found by trying expressions of the same form as f(t).
(a) A1 = 2, A2 = -1. Try Y, = Z as particular solution then
Y, = -1.5. Solution is Y, =A(21) + B( -1Y - 1.5 and initial
values give A = 7/3, B = 7/6.
(b) A1 = 2, A2 = 1. Try Y, = Z as particular solution- fails. Try
Y, = Zt; Y,+ 1 = Zt + Z, Yt+2 = Zt + 2Z, then Z = -4. Solu-
tion is Y, = A(21) + B - 4t and initial conditions give A = 5,
B = -4.
(c) A1 = 2, A2 = 2. Try Y, = Z(3 1); then Z = 1. Solution is
Y, = (k 1 + k 2 t)(2 1) + 31 and initial conditions give k 1 = 2,
k2 = -1.5.
(d) A1 = 2, A2 = 1. Try Y, = Zt(21); then Z = + 1.5. Solution is
Y, = (A(21) + B + 1.5t(21) and initial conditions give A = 1,
B = 0.
(e) A1 = -0.5, A2 = -0.5. Try Y, = Z 0 + Z 1t; then Z 0 = ~
Z 1 = 1. Solution is Y, = (k 1 + k 2t)(- 0.5y - ~ + (t/3) and ini-
tial conditions give k 1 = -88/9, k 2 = 8.
(f) A1 = -2 + i,A2 = -2- ihencea = -2,b = l,r = V5and
cos () = - 2/V 5 so () = -26° 34' = -0.4636 radians. Try
Y, = Z; then Z = 0.4 and the solution is
Y, = A(V 5Y cos (- 0.4636t - f) + 0.4
504 Appendix C
(g) A1 = 2 + 2i, A2 = 2- 2i, hence a= 2, b = -2, r = v'8 and

cos (} = 2/V8 = 0.7071, so (} = n/4. Try Y, = k 1 + k 2t and
k 1 = 2/25, k 2 = !. The solution is Y, = A(v'8)' cos
[n(t/4) - f] + (2/25) + t/5).
3. Equation is (1 - a) Y, = (/3 + y) Y,_l - yY,_2 + c5
(a) A1 = 0.625 + 0.599i, A2 = 0.625 - 0.599i
hence a= 0.625, b = 0.599, r = 0.8657, (} = 0.7642 radians.
Try Y, = Z and Z = 5.
Solution is Y, = A(0.8657)' cos (0.7642t - f) + 5.
(b) A1 = 0.4 + 0.49i, ~ = 0.4 - 0.49i; hence a = 0.4, b = 0.49,
r = 0.6324, (} = 0.886 radians. Try Y, = Z and Z = 3.33. The
solution is Y, = A(0.6324)' cos (0.886t - f) + 3.33.
4. The method is to use all three equations and to eliminate U and pe.
Differencing ( 1),
p,- Pt-1 = - b(U, - U,_l) + ~ - P1-l)
Substituting from (3),
p, - Pt-1 - be(m - p,) = ~ - P1-l)
Using (2), p, - Pt-1 - be(m - p,) = fk(pt-t - P1-l)
and since, from (1), ~ = p, - a + bU, + c
p, - Pt-1 - be(m - p,) = fkPt-1 - k(p,_l - a + bU,_I + c)
Simplifying: (1 + be)p, + (k - kf - 1)pt-t - bern - ak + ck
= -bkU,_ 1
Differencing again to eliminate U,
(1 + be)(p,- p,_ 1) + (k- kf- 1)(Pt-t- p,_ 2) = - bk( U,_ 1 - U,_ 2)
Using (3) and simplifying,
(1 + be)p, - (kf- k + 2 + be - bke)p,_l + (1 - k + kf)p,_2
= bkem
Comparing this with the result in the text:
(1 + be)pt+ 1 - (2 + be[1 - k])p, + p,_ 1 = bkem
it can be seen that they are the same iff = 1, but if this is not the
case then the dynamics are affected.
5. Substituting in the national income identity,

Y 1 = a¥1_ 1 + b(Y1 - Y1_ 1) + c¥
1_ 1 + dG 1_ 1
or (1 - b)Y1 = (a - b + c)YI-t + dG1_,

From the national income identity, G 1 = Y 1 - C, - 11 and so
G1_ 1 = ¥ 1_ 1 - C1_ 1 - 11_ 1 = ¥ 1_ 1 - a¥1_ 2 - b¥1_ 1 + b¥1_ 2 and
so (1 - b)Y1 = (a - b + c)Y1_ 1 + d(1 - b)¥1_ 1 + d(b - a)Y1_ 2 or
(1 - b)Y1 = (a- b + c + d- bd)¥1_ 1 - (b- a)d¥1_ 2 which is the
required equation.
(a) If d = 0 the equation is (1 - b)Y1 = (a - b + c)Y1_ 1 as re-
quired.
(b) With a = 0.6, b = 2, c = 0.1 and d = 0.5 the equation is
- Y 1 = -1.8¥1_ 1 - 0.7¥1_ 2 or Y 1 - 1.8¥1_ 1 - 0.7¥1_ 2 = 0.
The roots of the characteristic equation are -0.33 and 2.13
and so the general solution is Y1 =A(- 0.33Y + B(2.13y. The
initial conditions Y0 = 1 and Y, = 1 give 1 = A + B and 1 =
-0.33A + 2.13B. The solution is A = 0.45 and B = 0.55 so
that Y1 = 0.45(- 0.33Y + 0.55(2.13Y. As t increases the first
term will tend to zero and the second term approach infinity.
Further Reading
This book is intended as an introduction to the use of mathematics in
economics and business. As a supplement to the material covered here,
and for those who like to use a work-book with many worked problems
to accompany a text, the following excellent book is recommended:
A. J. Mabbett, Work Out Mathematics for Economists (Basingstoke:
Macmillan, 1986).
Once the basics have been mastered, the student should be able to
benefit from the many more advanced books which are available for
those wishing to study the subject seriously. Two recommended
mathematics books for economists are:
A. C. Chiang, Fundamental Methods of Mathematical Economics (New
York: McGraw-Hill, 1984) 3rd edn.
R. C. Read, A Mathematical Background for Economists and Social
Scientists (Englewood Cliffs, N.J.: Prentice-Hall, 1972).
There are a large number of books which use mathematics as a
means of developing economic theory. The following are suitable for
intermediate level students:
A. Koutsoyiannis, Modern Microeconomics (Basingstoke: Macmillan,
1984).
J. M. Henderson and R. E. Quandt, Microeconomic Theory (New
York: McGraw-Hill, 1980) 3rd edn.
R. Dornbusch and S. Fischer, Macroeconomics (New York: McGraw-
Hill, 1987) 4th edn.
W. M. Scarth, Macroeconomics (Toronto: Harcourt Brace Jovanovich,
1988).
One of the major applications of mathematics in economics is in the
area of economic modelling and forecasting. Introductions to these
topics are given in:
506
Further Reading 507
K. Holden, D. A. Peel and J. L. Thompson, Economic Forecasting: an

introduction (Cambridge: Cambridge University Press, 1990).
R. S. Pindyck and D. L. Rubinfeld, Econometric Models and Economic
Forecasts (New York: McGraw-Hill, 1990) 3rd edn.
Applications of quantitative methods in marketing and marketing
research are the subject of many books, and two recent ones are:
G. Oliver, Marketing Today (New York: Prentice-Hall, 1990) 3rd edn.
C. McDaniel and R. Gates, Contemporary Marketing Research (St Paul,
Minnesota: West Publishing, 1991).
For details of the formulae for the solution of cubic equations, see:
J. Parry Lewis, An Introduction to Mathematics for Students of Econ-
omics (Basingstoke: Macmillan, 1969) 2nd edn.
For both cubic and quartic equations, see:
F. Gerrish, Pure Mathematics, vol. 2 (Cambridge: Cambridge Univer-
sity Press, 1960).
Further applications of mathematics to finance, and to the use of
mortality tables, are contained in:
F. Ayres, Theory and Problems of Finance (New York: Schaum, 1963).
For a detailed treatment of the properties of matrices, see:
G. Hadley, Linear Algebra (Reading, Mass: Addison-Wesley, 1961).
For the application of matrices in econometrics, see:
J. Johnston, Econometric Methods (New York: McGraw-Hill, 1984) 3rd
edn.
Introductions to the uses of input-output analysis in regional econ-
omics are provided by:
H. Armstrong and J. Taylor, Regional Economics and Policy (Dedding-
ton: Philip Allan, 1985).
P. N. Balchin and G. H. Bull, Regional and Urban Economics (London:
Harper & Row, 1987).
A more general treatment of input-output analysis is given by:
R. O'Connor and E. W. Henry, Input-Output Analysis and its Applica-
tions (London: Griffen, 1975).
508 Further Reading
The inventory models discussed in Chapter 5 assume a known

demand. The extension of the model to probabilistic demand is covered
in books on operations research, for example:
D. R. Anderson, D. J. Sweeney and T. A. Williams, An Introduction to
Management Science (St Paul, Minnesota: West Publishing, 1991).
This also discusses linear programming in detail as well as other applica-
tions of mathematics to management problems. A general treatment of
difference and differential equations is provided by:
W. J. Baumol, Economic Dynamics (New York: Macmillan, 1959) 2nd
edn.
Finally, since any list of books inevitably dates, the student is
advised to look at the current journals in economics and business to see
the latest applications. A start could be made with The American
Economic Review, The Economic Journal, Management Science, The
Journal of Marketing and The Journal of Marketing Research.
Index
Absolute areas, 234--6 Coordinates, 5
Adjoint, 54 Cosine, see Trigonometric functions
Annuities, 140-2 Cramer's rule, 30, 34-5, 97
Areas, calculation of, 229-36 Cubic equations, 106, 507
Arithmetic progression, 130-4
Arithmetic series, 132-4 Demand, 13-21, 177-80, 198-202,
Associative rule, 9, 412 236-8, 264--6, 322, 334--6, 368-74
Asymptotes, 111-13 De Moivre's theorem, 379
Auxiliary equation, 338 Dependent variable, 3, 416
Average cost, 111-15, 184 Derivative, 167
Average propensity to consume, 183 Determinants, 28-36, 51-7, 79-80,
287, 301
Binomial theorem, 148-50, 170 Difference equations
Boundary conditions, 322 first-order, 364-78
Breakeven point, 11, 115-19 second-order, 378-91
Differential equations
CES production function, 271 first-order, 319-38, 350-2
Chain rule, 173 second-order, 319, 338-54
Characteristic equation, 338, 379 Differentials, 207-9, 267-73
Characteristic roots and vectors, 73 Differentiation, 165-223, 257-96
Chief rent, 142 exponential functions, 175--ti
Cobb-Douglas production function, functions of one variable, 165-223
263-4, 270-1, 274, 279, 281, 282 functions of more than one
Cobweb model, 368-74, 376 variable, 257-96
Cofactor, 51-4 logarithmic functions, 174-5
Commutative rule, 9, 412 trigonometric functions, 176, 403-4
Complementary function, 345 Discontinuous functions, 125-7
Complex numbers, 2, 104, 341, 379 Discounting, 137-9, 154
Compound interest, 134-5, 146-8, Distributive rule, 9, 412
336, 365-7 Domar debt model, 332-3
Constrained maxima and minima, 88, Dual price, 311
289-318 Dual problem, 314-5
Consumer's surplus, 236-8
Consumption-investment models, 23, e, 147, 151, 159
24--6, 384--6 Economic order quantity, 204
Consumption model, 375 Eigenvalues, see Characteristic roots
Contribution per unit, 13, 298 Elasticity, 177-80, 182, 264--6, 281,
Convergence of a geometric series, 322
150 Elasticity of substitution, 270
509
510 Index
Endogenous variable, 3 Infinity, 10, 110

Equations Inflexion, point of, 194
inconsistent, 64 Initial conditions, 322
linear, 1-42, 48, 50 Input-output analysis, 76--91
linear dependence, 64-9 Integers, 1, 408
non-linear, 92-129 Integrating factor, 330-1
quadratic, 92-125 Integration, 224-55
redundant, 64 by parts, 248--50
roots, 98--105, 213--17 by substitution, 253-4
roots of quadratic, 98--105 of trigonometric functions, 404-6
simultaneous linear, 9-42, 48, see also Differential equations
50-1, 60-9 Intercept, 6, 92
simultaneous quadratic, 119-25 Internal rate of return, 139
Equilibrium condition, 3 Interpolation, 8
Equilibrium solution, 368, 381 Inventory models, 202-6, 208--9
Euler relations, 212, 341 Inverse trigonometric functions,
Euler's theorem, 280-2 406--7
Exogenous variable, 3 Irrational numbers, 1-2, 147
Exponential series, 151, 212 IS/LM model, 24-8, 41, 272-3
Extrapolation, 8 Isoquant, 270
Iterative procedure, 214, 301
Factorial, 149
Feasible solution, 299 Jacobian, 262
Fixed costs, 3
Function of a function rule, 173--5 Lagrangian multipliers, 289-94
Functions, 2, 415 Laplace expansion, 52
Latent roots, see Characteristic roots
Gaussian elimination, 57 Leontief, 76
Geometric progression, 134-6 Life insurance policies, 145
Geometric series, 135-6 Limit of a function, 110
Gradient, 6, 165 Linear equation, 6
Graphs, 4 Linear expenditure system, 292-4
Linear programming, 88, 297-318
Harrod-Domar growth model, LM schedule, 25
374-5, 376 Logarithms, 156-61
Hawkins-Simon conditions, 85
Hessian, 284, 287, 291 MacLaurin's theorem, 209-12, 403
Homogeneous forms, 323--5, 338, 367 Marginal analysis, 180-3, 238-42,
Homogeneous functions, 278--80 263-4
Homothetic functions, 282 Marginal propensity to consume, 21,
Hyperbolic logarithms, 159 183
Marginal rate of substitution, 269-72
Identity, 3 Matrices, 43--91
Imaginary numbers, 2, 104 addition, 45
Implicit differentiation, 275-7 cofactor, 51-2
Incremental worth, 311 equality, 44
Independent variable, 3, 416 identity, 48--9
Indifference curves, 269 idempotent, 70
Index 511
inverse, ~ Quadratic equation, 98

multiplication, 45-8 Quadratic form, 70
partition, 71 Quadratic function, 92
rank, 65 Quartic equation, 106, 507
square, 44
symmetric, 69 Radians, 398
transpose, 49 Rank of a matrix, 65
unit, 48--9 Rational numbers, 1
Maximum and minimum values, Rectangular hyperbola, 109-13
100--1, 184--94, 282-94 Returns to scale, 279
Minors of a matrix, 51-4 Riemann integral, 230
Modulus, 150, 373 Row operations, 57-63
Money illusion, 279
Mortgages, 142-4
Multiplier, 22, 24, 83 Saddle point, 283-4, 287-9
Samuelson's multiplier-accelerator
Napierian logarithms, 159 model, 377-8
National income models, 21-8, 36-42 Scalars, 45
Natural logarithms, 159 Sensitivity analysis, 208
Natural rate of unemployment, 359, Set theory, 408--16
391 Shadow price, 311, 314
Necessary and sufficient conditions, Simplex method, 301, 315
194 Simpson's rule, 243-5
Net present value, 138 Sine, see Trigonometric functions
Net revenue, 122-4, 182, 195 Sinking fund, 144
Newton's method, 213-20 Slope, 6, 92, 165-6
Numbers, 1 Stone, R., 292
Supply, 13-6, 281, 368--74
Objective function, 299
Opportunity cost, 311, 314 Tangent to a curve, 117, 166
Optimal solution, 299 Taxes and demand, 17-21, 200--2
Taylor's theorem, 209-11, 213, 216,
Parabola, 95, 121 245
Partial differentiation, 257-94 Total derivatives, 273-4
Partial fractions, 250--3 Transient solution, 368, 381
Particular integral, 345-9 Trigonometric functions, 17fr-7,
Particular solution, 368, 381 211-3, 273-4, 341-2, 347-8,
Pascal's triangle, 149 379-80, 384, 395-407
Perfect competition, 10
Phase, 354 Utility function, 269, 292, 293
Phillips curve, 354--5, 387
Point of inflexion, 194
Polynomial, 93, 106, 189 Variable, 2
Present value, 137-9, 154 Variable costs, 4
Producer's surplus, 237 Vector, 44
Production function, 263, 270--2, 279 Venn diagrams, 410--14
Product-mix problem, 297-9
Progression, 130 Walrasian adjustment, 335

Introductory Mathematics For Economics and Business by K Holden

Uploaded by

Copyright:

Available Formats

Introductory Mathematics For Economics and Business by K Holden

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Introductory Mathematics For Economics and Business by K Holden

Uploaded by

Copyright:

Available Formats

Introductory Mathematics for Economics and Business

K. Holden and A. W. Pearson

Softcover reprint of the hardcover 2nd edition 1992 978-0-333-57649-6

All rights reserved. No reproduction, copy or transmission of

No paragraph of this publication may be reproduced, copied or

Any person who does any unauthorised act in relation to this

First published 1975 by

Published as Introductory Mathematics for Economists

This edition published 1992 by

ISBN 978-0-333-57650-2 ISBN 978-1-349-22357-2 (eBook)

A catalogue record for this book is available

2 Elementary Matrix Algebra 43

2.12 Input-output analysis 76

5 Differential Calculus 165

5.2 Some general rules 169

6 Integral Calculus 224

7 Partial Differentiation 257

7.2 Marginal analysis 263

8 Linear Programming 297

9 Differential Equations 319

9.9 Graphical presentation 350

10 Difference Equations 364

Appendix A Trigonometric functions 395

Appendix B Set Theory 408

Appendix C Answers to Exercises 417

need trigonometric functions. Detailed numerical answers to the exer-

in fact it can be the starting point in many applications of economic

increase as the level of output increases. These are known as variable

1.2 Graphical representation

Effectively we are determining the values of a and b by using the

Hence TC = 100 + 3q is the equation which summarises the available

1.3 Intercept and slope

(a) Y =a+ bX (b) Y= a+ b'X (c) Y= a+ b"X

1.4 Interpolation and extrapolation

(a) the variable cost is £6 per unit, and

1.6 Simultaneous linear equations

4. Operations involving zero: If x is non-zero,

This relationship is obviously important but does not tell us any-

Total revenue = price per unit X number of units sold

1. 7 Demand and supply

Although the quantity demanded is, by convention, represented

1.9 The effects of taxes on demand

tax such as might be imposed on a bottle of spirits, while the latter is

qs =a'+ b'p 1 (b' > 0)

The original equilibrium was Po = 8, q = 60 and so

the demand curve is

By substitution, the equilibrium price (p,) and quantity (q,) are

1.10 Simple national income models

Here the marginal propensity to consume, the proportion of an extra

and collecting the Y terms on the left hand side results in

Substituting this into the consumption equation,

the solutions are easily shown to be

1.11 IS-LM analysis

C = 10 + 0.8Y. The goods market therefore has the three equations to

assumptions about the different components. For example, suppose

The government are undecided as to whether to impose a tax of

5. The model in Question 4 can be extended to include a government

where Yd is disposable income, G is government expenditure and

1.13 Determinant notation