Académique Documents
Professionnel Documents
Culture Documents
Year 2009
Monetary Theory Analysis (G5078MTA)
taught by Omar Lakkis
http://www.maths.sussex.ac.uk/Staff/OL/teaching.html
o.lakkis@sussex.ac.uk
Contents
Whats this?
Disclaimer
Course synopsis
5
5
5
1
1
2
6
8
10
11
12
14
15
16
19
19
22
25
26
26
29
29
31
34
35
38
40
42
45
45
47
48
49
51
53
53
55
56
56
58
60
62
64
67
69
69
69
72
77
77
78
80
86
89
89
90
90
92
93
95
95
95
97
99
100
Appendix. Bibliography
101
Appendix. Index
103
Whats this?
This booklet bears the essential contents of the course Monetary Theory Analysis
(MTA), code G5078, as taught at the University of Sussex by Omar Lakkis in 2009.
It contains the (most of) the material covered in class as well as the exercises and
their solutions. If you are a University of Sussex maths student, or prospective student, you are allowed to use this document only after you have taken the Monetary
Theory Analysis course with me. Otherwise you must have explicit permission from
the tutor o.lakkis@sussex.ac.uk to use it.
Disclaimer
A part from a couple of places, these notes are largely the result of summarising
the reading from published sources [CZ03, MCS86, mainly]. I therefore have no
pretense at originality, except for the errors that you may find in the manuscript. If
you do find what you think is an error/typo/mistake, I will be grateful if you sent
me an email about it (subject: MTA G5078 notes typo) so that things can be
fixed for future releases of these notes.
Much of the material in the notes is copyrighted by the authors of the aforementioned
sources. If you publish or broadcast this material, or otherwise use it for purposes
not directly related to this particular course you may be acting illegaly at your own
risk.
Course synopsis
This course focuses on the basic mathematical aspects of money dynamics.
Interest rate theory and manipulations are at the base of the course, there we learn
about mathematical (and where necessary computational) methods in finance such
as cash flows, the equation of value and the methods for calculating mortgages.
We then introduce various applications of interest theory by looking at examples
where surrounding interest rates and inflation may influence investments. We
learn how to calculate return and introduce the concept of yield, for fixed and
variable income securities.
A learning outcome of the course is to know how to evaluate, analyse and disseminate a real world investment project by writing an easy to understand financial
report, which implies that the assessment comes in the form of an Essay covering
a particular aspect of the course.
For the Essay, you will need to be:
? comfortable mathematics methods to solve financial problems;
? preparted to do autonomous reading in the field;
? able to communicate mathematical ideas to an audience.
CHAPTER 1
and p integer, we write ih = i(p) , which also means i1/p = i(p) . According to this
definition, an investment of 1 for a period of length 1/p produces a return of 1+i(p) /p.
1.1.5. Example (nominal rates).
Problem. Working in time units of years, we will let the nominal rate of interest
per year on monthly transactions over the next year be 12. If 100 is invested at
time 0, how much is it worth at the end of the first month and at the end of the
third month?
Solution. We have defined: i1/12 = 12% and so the effective monthly interest rate
is i1/12 /12 = 1%.
At the end of the first month we have 100(1 + 0.01) = 101.
At the end of the third month we have 100 1.013 = 103.03
1.1.6. Example.
Problem. The nominal interest rate of interest per annum quoted in the financial
press for local authority deposits on a particular day are as follows:
Term
Nominal rate of interest (%)
1 day
11.75
1 week
11.5
1 month
11.375
1 quarter
11.25
Find the accumulation of an investment at this time of 100 for
(a) for 2 days,
(b) for 3 weeks,
(c) for 1 month,
MTA (G5078) Autumn 2009 1.2. From discrete time to continuous time3
1.2.1. Accumulation factors. Let time be measured in suitable units (e.g. years).
For t1 t2 we define A(t1 , t2 ) to be the accumulation at time t2 of an investment of
1 at time t1 for a term of t2 t1. Thus A(t1 , t2 ) is the amount which will be repaid
at time t2 in return for an investment of 1 at time t1 .
Expressing this in terms of capital, suppose our capital at time t1 was C(t1 ) and at
time t2 , C(t2 ), then the accumulation factor is given by
A(t1 , t2 ) =
C(t2 )
.
C(t1 )
(1.2.1)
Since hih (t) is the effective rate for the period of length h beginning at time t and
so, by the definition of ih (t), for all t and for all h > 0,
A(t, t + h) = 1 + hih (t)
(1.2.2)
(1.2.3)
(1.2.4)
(1.2.5)
1.2.3. Example. For integer times, 50 are invested at time 2 and the accumulated
amount at time 7 is 100. Find i5 (2) and i(2).
Solution. The accumulation factor is A(2, 7) = 100/50 = 2 and, using (1.2.3), we
have i5 (2) = (2 1)/5 = 0.2 = 20%.
1.2.4. Exercise. Calculate the equivalent effective annual rate of interest (assumed
to be a constant over the period) for the investment in the above example, assuming
that time is in years.
1.2.5. Exercise. 5000 is invested at time 0 and the proceeds at time 10 are 9000.
Calculate A(6, 10) if A(0, 9) = 1.8, A(2, 4) = 1.1, A(2, 6) = 1.32, A(4, 9) = 1.45.
1.2.6. Force of interests. We consider the case where the interest is paid continuously throughout the time period: a nominal interest rate convertible very frequently
(e.g., every second), then this fund steadily accumulates over the period as interest
is earned and added.
In the limiting case, i.e., as h 0, the amount of the fund can be considered to be
subject to a constant force causing it to grow. This leads us to the concept of a
force of interest per unit time at time t: (t). It is mathematically defined as
lim ih (t) = lim h 0+ (A(t, t + h) 1)/h = (t).
h0+
(t) is also know as the nominal rate of interest per unit time at time t convertible
momently.
(1.2.7)
(1.2.8)
which is a long product. A very useful property of exp (and of its inverse log) is
that it transforms sums into products (and viceversa). Using these properties we
may write the above as
A(t, s) = exp(log(1 + hih (t0 )) + log(1 + hih (t1 )) + + log(1 + hih (tn1 ))) .
(1.2.9)
Now, recalling Taylors formula of order 2 we know that
log(1 + ) = + (), for 1/2 < < 1/2,
(1.2.10)
(1/2, 1/2) .
(1.2.11)
, h > 0,
(1.2.12)
it follows that, for h < 1/2i , we can apply the Taylor formula and regroup terms
to get
A(t, s) = exp hih (t0 ) + + hih (tn1 ) + (hih (t0 )) + + (hih (tn1 ))
= exp
n1
X
(1.2.13)
k=0
where
(h) := (hih (t0 )) + + (hih (tn1 )).
(1.2.14)
Note that identity (1.2.13) is true for all h > 0 and h < 1/2i . Taking h 0 (and
using uniform convergence of step functions) we have
" n1
# Z
s
X
lim
hih (tk ) =
( ) d.
(1.2.15)
h0
k=0
To see what happens to the second term on the right-hand side of (1.2.13), we use
the properties of the remainders in Taylors formula (1.2.11) and the uniform bound
(1.2.12), we get
0 |(h)|
n1
X
k=0
n1
X
h = c0 i 2 (s t)h 0, as h 0.
k=0
(1.2.16)
Thus (h) 0 and, modulo minor technical points, we have shown that
Z s
A(t, s) = exp
( ) d.
t
(1.2.17)
MTA (G5078) Autumn 2009 1.2. From discrete time to continuous time5
1.2.8. Theorem (accumulation factor and force of interest relation). Given an
accumulation factor (t, s) 7 A(t, s), then
Z s
( ) d ,
(1.2.18)
A(t, s) = exp
t
for all t s.
Proof 1
Fix t R and let s be variable. Define
Z
( ) d.
(1.2.19)
We want to show that R(s) = L(s) for all s t. For starters we have
R(t) = 0 = L(t)
(t is frozen) .
(1.2.20)
Then for s t, we have (by the Chain Rule and the definition of A and ) that
d
1
d
R(s) =
A(t, s)
ds
A(t, s) ds
1
A(t, s + h) A(t, s)
=
lim
h0
A(t, s)
h
1
A(t, s)(A(s, s + h) 1)
=
lim
A(t, s) h0
h
A(s, s + h) 1
= lim
h0
h
= (s).
R0 (s) =
(1.2.21)
(1.2.22)
Thus R0 (s) = L0 (s), for all s t and L(t) = R(t), so by the integrating both sides
over [t, s] we obtain
Z s
Z s
0
L(s) = L(t) +
L = R(t) +
R0 = R(s), s t.
(1.2.23)
t
1.2.9. Example. Assume (a) (t) = , or (b) (t) = a + bt.
Find the formulas for the accumulation of 1 unit from t1 to t2 in each case.
1.2.10. Solution. Using (1.2.18) we have
(a) A(t1 , t2 ) = exp((t2 t1 )), and
(b) A(t1 , t2 ) = exp(a(t2 t1 ) + 21 b(t22 t21 )).
1.2.11. Exercise. A bank credits interest on deposit using accumulation factors
based on a variable force of interest. On 1 July 1999, a customer deposited 100, 000
pounds with the bank. On 1 July 2001, his deposit has grown to 120, 000. Assuming that the force of interest per annum was bt (1 July 1999 = 0, unit = year)
during the period, find the force of interest per annum on 1 July 2000.
1Technical
note. This proof, as presented, is a more rigorous, but perhaps less insightful,
alternative to the one in 1.2.7. Also this proof will work only for differentiable accumulation
functions A. The argument in 1.2.7 can be made perfectly rigorous for that is Lebesgueintegrable, so it may even have jumps. See [MCS86, Appendix 1] for another proof.
(1.2.24)
(1.2.25)
(1.3.2)
(1.3.3)
(1.3.4)
v = v(1) = exp().
(1.3.5)
where
1.3.3. Example.
Problem. Measuring time in years from present, suppose that (t) = 0.06(0.9)t for
all t 0. Find a simple expression for v(t) and find the discounted present value of
100 due in 3.5 years time.
Solution.
Z t
0.06(0.9t 1)
s
0.06(0.9) ds = exp
v(t) = exp
.
ln(0.9)
0
Hence the present value of 100 due in 3.5 years is
100 exp[0.06(0.93.5 1)/ ln(0.9)] = 83.89.
(1.3.6)
(1.3.8)
(1.3.9)
Following the summation principle (and taking the limit, were we to make a fully
rigorous argument) we obtain the following expression for the discounted value between 0 and t given by
Z t
v( )( ) d.
(1.3.10)
0
r=1
Note that in spite of using a discrete description for the process we are using a
continuous time description for the interest rate, i.e., in terms of the force of interest
function .
This equation may be written as:
Z tr
n
X
cr exp
=0
(1.4.2)
0
r=1
where cr = br ar is the amount of the net cash flow at time tr . (We adopt
the convention that a negative cash flow corresponds to a payment by the investor
and a positive
R cash
flow represents a payment to the investor.) Recalling that
t
v(t) = exp 0 we may rewrite (1.4.2) as
n
X
cr v(tr ) = 0.
(1.4.3)
r=1
Equation (1.4.2), which expresses algebraically the condition that, at force of interest
, the total value of the net cash flows is 0, is called the equation of value for the
force of interest implied by the transaction.
Under the constant force of interest assumption we have
1
e = 1 + i = and v(t) = v t ,
(1.4.4)
v
and the equation of value (1.4.3) may be then written as
!
n
n
X
X
tr
tr
cr (1 + i)
or
cr v
= 0.
(1.4.5)
r=1
r=1
MTA (G5078) Autumn 2009 1.5. Valuing cash flows, equation of value
and yield
9
1.4.3. Hybrid description. If the cash flows involve both discrete and continuous
transactions, we simply add the two contibutions.
1.4.4. Example.
Problem. A businessman is owed the following amounts: 1000 on 1 January 1986,
2500 on 1 January 1987 and 3000 on 1 July 1987. Assume a constant force of
interest of 0.06 per annum, find the value of these payments:
(a) on 1 January 1984,
(b) on 1 March 1985.
Solution. Let time be measured in years from 1 January 1984. The value of the
debts at that date is
1000v(2) + 2500v(3) + 3000v(3.5)
= 1000 exp(0.12) + 2500 exp(0.18) + 3000 exp(0.21) = 5406.85. (1.4.6)
The value at 1 March 1985 of the same debts is
5406.85 exp(0.06(14/12)) = 5798.89.
(1.4.7)
1.4.5. Example.
Problem. Suppose that time is measured in years and that , the force of interest
per unit time, is given by:
0.121 890
(t) = 0.076 961 +
(1.4.8)
1 + 0.5 exp(0.121 890 t)
(a) Find the single payment which, if invested at time 10, will accumulate to
30 000 at time 20.
(b) Find the accumulated amount after ten years of ten annual payments each of
1 000, the first payment being made at time 0.
Solution. (a) This is simply
30, 000 v(20) = xv(10)
Hence
1.4.6. Exercise. An investor owns a block of shares which are expected to pay a
dividend of amount D in one years time and dividends in each future year that are
100j % higher than in the previous year. Suppose that the bank deposit interest rate
is i, Show that the present value of the proceeds from this investment is D/(i j)
assuming the shares will be held indefinitely. (Hint: bank interest rate will reduce
the money value.)
10
1.5.1. A lazy son of a rich. Keep capital at C and receive interest as income.
This is achieved by withdrawing continuously money from the account as to keep
it at C. Note that there is no inflation involved in this model, and money does not
lose value in time.
Fix a time t > t0 . The interest received at times
t0 + h, t0 + 2h, . . . , t0 + nh = t,
(1.5.1)
where h = (tt0 )/n and n is a positive integer, can be calculated as follows. Interest
payable at time t0 + (j + 1)h for the period (t0 + jh, t0 + (j + 1)h) is
Chih (t0 + jh),
and the total of all payments is
C
n1
X
j=1
n1
X
Zt
hih (t0 + jh) C
j=1
(s)ds =: I(t).
t0
and the final capitals present value is Cv(T ). On the other hand the present value
is C, therefore we obtain the relation
Z T
(1.5.2)
C
(t)v(t) dt + Cv(T ) = C.
| {z }
0
{z
} pr.val. of capital
|
pr.val. of income
Z
= exp
t=0
T
1
= 1 v(T ).
Multiplying by C we recover
Z
C=C
(t)v(t) dt + Cv(T )
0
11
1.5.3. Remark (Capital gains and losses). So far, we have described the difference
between money returned at the end of the term and the cash originally invested as
interest.
In practice, this quantity may be divided into interest income and capital gains. A
capital loss is used for negative capital gain.
1.5.4. Exercise. Use buy-to-let concept to explain the concept of interest income
and capital gain.
1.6. Compound Interest Functions
In the following, we will work under the constant force of interest assumption. We
assume, thus that (t) is a constant, which we denote also by .2 We review in
this short section the main properties of compound interest with constant force of
interest.
1.6.1. Effective discount rates. The value at time s of 1 due at time s + t is
s+t
Z
exp (r) dr = exp(t),
s
Z
exp (r) dr = exp(t) = (1 + i)t
s
which is independent of s.
1.6.2. Exercise. Find all the relationship between , v, i and d.
Solution.
v=
1
= 1 d = e .
1+i
(1.6.1)
1.6.3. Remark (interest, discount and the (wrong) law of universal linearity).
The interest rate i and the discount rate d have an interesting dual relationship,
summarised by
(1 + i)(1 d) = 1.
(1.6.2)
One way to remember this relation is to think i as the percentage gain that compensates for the percentage loss d.
It should be clear to the reader by now that d 6= i. This is a common mistake (known
as the law of universal linearity) whereby, for instance, many people think that if
the FTSE goes down by 5% one day and by 5% up then it has recovered. What is
your guess?
2We
trust the reader will not be confused when reading expressions like (t s)
12
n2n1
i
h
annuity value at
various times
ane a
ne
n+1
time
sne sne
The reason why many people commit this mistake (and the reason behind the mistakes nickname) is the Neumann series expansion (which is just another name of
the Taylor expansion)
1
= 1 x + x2 x3 + .
1+x
(1.6.3)
1 d = 1 i + i2 i3 + ,
(1.6.4)
d = i i2 + i3 + .
(1.6.5)
Using x = i we obtain
and hence
For very small i, say of the order of 1%, the remainder term (nonlinear part of the
expansion) is negligible, and one may assume, for all practical purposes that d = i.
But as soon as i becomes bigger, one should be more careful.
1.6.4. Problem. A lender bases his short-term transactions on a rate of commercial
discount D, where 0 < D < 1. For 0 < t 1, in return for a repayment of X after
a period t, he will lend X(1 Dt) at the start of the period.
For such a transaction over an interval of length t, (0 < t 1) derive an expression
in terms of D and t for d, the effective rate of discount per unit time. Hence show
that, regarded as a function of t, d is increasing on the interval 0 < t 1.
1.7. Annuities-certain: present values and accumulation
1.7.1. Annuity-certain, annuity-due, perpetuity. Consider a series of n payments, each of amount 1, made at time interval of 1 unit, the 1st payment is made
at time t0 + 1. We sketch the situation in Figure 1.
The value of this series of payments at time t0 (which is the present value if t0 = 0)
is denoted by ane , hence v = 1/(1 + i), and, for n 1, we get
ane = v + v 2 + . . . + v n =
1 vn
1 vn
=
.
1/v 1
i
(1.7.1)
If n = 0 (no payment ever done) then we define ane := 0. We refer the quantity
ane the present value of an immediate annuity-certain (annuity paid in arrear).
By contrast, the value of this series of payment at the time the first payment is
made is denoted by a
ne . Thus
a
ne = 1 + v + v 2 + + v n1 =
1 vn
1 vn
=
.
1v
d
The name of a
ne is called the present value of an (annuity paid in advance).
(1 + i)n 1
.
d
When n , i > 0, the limit of an annuity (certain or due) is called a perpetuity
(certain or due, respectively):
sne = (1 + i)n + (1 + i)n1 + + (1 + i) = (1 + i)n a
n] =
ane ae and a
ne a
e .
Hence we have
1 vn
1
=
n
i
i
ae = lim
and
1 vn
1
= .
n
d
d
1.7.2. Problem. A loan of 2, 400 is to be repaid by 20 equal annual instalments.
The rate of interest for the transaction is 10% per annum. Find the amount of
each annual repayment, assuming that payments are made (a) in arrear and (b) in
advance.
a
e = lim
1.7.3. Problem. On 10 Jan in each of the years 1964 to 1979 inclusive, an investor
deposited 500 in a special bank savings account. On 10 Jan 1983, the investor
withdrew his savings. Given that over the entire period the bank used an annual
interest rate of 7% for its special savings accounts, find the sum withdrawn by the
investor.
1.7.4. Problem. A borrower agrees to repay a loan of 3000 by 15 annual repayments of 500, the first repayment being due after 5 years. Find the annual yield
(i.e., the interest rate the creditor has to apply) for this transaction.
1.7.5. Exercise (deferred annuity-certain). Let m and n be non-negative integers,
draw a cash flow chart (see 1.9) for a series of n payments, each of amount 1, due
at times (m + 1), (m + 2), . . . , (m + n). Find the value of these payments at time 0
(this is known as deferred annuity(-certain) and is denoted by m |ane .
1.7.6. Exercise (deferred annuity-due). Define deferred annuity-due
ane ?
m |
14
1.8.1. The basic . Assume that a sum L is lent for a series of n yearly payments.
The r-th payment, of amount xr , being due at time r. Let the effective annual
interest rate for r-th year be ir . Then we have
L = x1 (1 + i1 )1 + x2 (1 + i1 )1 (1 + i2 )1 + + xn (1 + i1 )1 (1 + in )1 . (1.8.1)
However, it maybe useful to split the payment in terms of capital and interest
useful when tax needs to be calculated.
Let
F0 = L,
Ft loan outstanding immediately after payment at time t,
xt be the loan repaid at t,
Then the interest due at time t, is given by
it Ft1 ,
(1.8.2)
s=1
r=s+1
(1.8.3)
If i is the interest rate per unit time, a loan of amount sne is made at time 0 in
return for n repayments, each of amount 1, to be made at t = 1, . . . , n.
We split the payments into capital and interest: after the t-th payment, there remain
nt outstanding payments and the outstanding loan is then Ft = ante . The amount
of loan repaid at time t is
ft = Ft1 Ft = ant+1e ante = v nt+1 .
(1.8.4)
1.8.3. Exercise (producing an ). Make the schedule for a level annuity (amount of
loan ane ) by filling the following table appropriately.
Interest content of
Loan outstanding afPayment
Capital repaid
payment
ter payment
1
2
...
...
...
...
t
...
...
...
...
n1
n
15
1.8.4. Problem. A loan of 10, 000 pounds is to be repaid over 10 years by a level
annuity payable monthly in arrear. The amount of the monthly payment is calculated on the basis of an interest rate of 1% per month effective. Find
(a) The monthly repayment.
(b) The total capital repaid and interest paid in the first and last year respectively.
(c) After which monthly repayment does the outstanding loan become first less
than 5, 000?
(d) For which monthly repayment the capital repaid first exceeds the interest?
1.9. Cash flow
We have mentioned cash flows in the last few sections. We now give some more
ideas concerning cash flow.
1.9.1. Cash flow model terminology. The is a mathematical projection of the
payments arising from a financial transaction, e.g., a loan, a life insurance contract
or a capital project. Payments received are referred to as income and are shown
as positive cash flows. Payments made are referred to as outgo and are shown as
negative cash flows. The difference at a single point in time (income less outgo) is
called the at that point in time.
Where there is uncertainty about the amount or timing of cash flows, an actuary
can assign probabilities to both the amount and the existence of a cash flow.
1.9.2. Example (fixed interest security). An investor might buy a 10-year fixed
interest security of nominal amount 1, 000. This means that the face value of the
loan is 1, 000. The investor is unlikely to pay exactly 1, 000 for this security
but will pay a price that is acceptable to both parties. This may be higher or
lower than 1, 000. The investor will then receive a lump sum payment in 20 years
time. This lump sum is most commonly equal to the nominal amount, in this case
1, 000, but could be a pre-specified amount higher or lower than this. The investor
will also receive regular payments throughout the 20 years of, say, 50 pa. These
regular payments could be made at the end of each year or half year or at different
intervals.
1.9.3. Example (Index linked securities). The initial negative cash flow is followed
by a series of unknown positive cash flows and a single larger unknown positive cash
flow, all on specified dates. However, it is known that the amounts of the future
cash flows relate to the inflation index. Hence these cash flows are said to be known
in real terms.
1.9.4. Definition (real terms and nominal terms). Real terms means taking into
account inflation, whereas nominal means ignoring inflation.
1.9.5. Essay. Using internet, newspaper and financial journals to find information
of the following investment vehicles and analyse their cash flows respectively:
5. repayment loan, or
1. equity shares, also 2. cash on deposit;
3.
annuities;
mortgage;
known as shares or
equities in Britain 4. endowment assurance, 6. motor insurance;
or an interest-only 7. index linked bonds.
and
as
common
loan;
stock in the USA;
16
Exercise 1.1 (accumulation factor). For integer times, 50 are invested at time
2 and the accumulated amount at time 7 is 100. Find i5 (2) and i(2).
Problem 1.2 (commervial vs. effective rate). A lender bases his short-term transactions on a rate of commercial discount D, where 0 < D < 1. For 0 < t 1, in
return for a repayment of X after a period t, he will lend X(1 Dt) at the start of
the period.
For such a transaction over an interval of length t, (0 < t 1) derive an expression
in terms of D and t for d, the effective rate of discount per unit time. Hence show
that, regarded as a function of t, d is increasing on the interval 0 < t 1.
Problem 1.3 (annuities). A loan of 2, 400 is to be repaid by 20 equal annual
instalments. The rate of interest for the transaction is 10% per annum. Find the
amount of each annual repayment, assuming that payments are made (a) in arrear
and (b) in advance.
Problem 1.4 (annuities). On 10 Jan in each of the years 1964 to 1979 inclusive,
an investor deposited 500 in a special bank savings account. On 10 Jan 1983, the
investor withdrew his savings. Given that over the entire period the bank used an
annual interest rate of 7% for its special savings accounts, find the sum withdrawn
by the investor.
Problem 1.5 (annuities). A borrower agrees to repay a loan of 3000 by 15 annual
repayments of 500, the first repayment being due after 5 years. Find the annual
yield (i.e., the interest rate the creditor has to apply) for this transaction.
Exercise 1.6 (accumulation factor). 5000 are invested at time 0 and the proceeds
at time 10 are 9000. Calculate A(6, 10) if A(0, 9) = 1.8, A(2, 4) = 1.1, A(2, 6) =
1.32, A(4, 9) = 1.45.
Exercise 1.7 (force of interest). If a large pension fund with a value of 2000m is
assumed to grow steadily subject to a constant force of interest of 10% per annum,
how much interest is earned every second? (Assume that there are 365 days in a
year.)
Exercise 1.8 (force of interest). A bank credits interest on deposit using accumulation factors based on a variable force of interest. On 1 July 1999, a customer
deposited 100,000 pounds with the bank. On 1 July 2001, his deposit has grown
to 120,000. Assuming that the force of interest per annum was a + bt (1 July 1999
= 0, unit = year) during the period, find the force of interest per annum on 1 July
2000.
Exercise 1.9 (discount factor). Calculate the present value on 1 September 2002
of payments of 280 due on 1 September 2004 and 360 due on 1 March 2005.
Interest is constantly 10% p.a. effective.
Exercise 1.10 (dividend gain versus interest gain). An investor owns a block of
shares which are expected to pay a dividend of amount D in one years time and
dividends in each future year that are j = 100j% higher than in the previous year.
Suppose that an alternative investment opportunity is a bank deposit whose interest
rate is i.
Show that the present value of the proceeds from the share investment is D/(i j)
assuming the shares will be held indefinitely.
Hint. The bank interest rate will act as an inflation rate and reduce the money value
in time.
(1.12.2)
RT
(t)v(t)dt + Cv(T )
Problem 1.14. Use buy-to-let concept to explain the concept of interest income
and capital gain.
Exercise 1.15 (interest basic relations). Find all the possible relationships among
, v, i and d.
Exercise 1.16 (deferred annuity-certain). Let m and n be non-negative integers,
draw a cash flow chart (see 1.9) for a series of n payments, each of amount 1, due
at times (m + 1), (m + 2), . . . , (m + n). Find the value of these payments at time 0
(this is known as deferred annuity(-certain) and is denoted by m |ane .
Exercise 1.17 (deferred annuity-due). Define deferred annuity-due
ane ?
m |
Exercise 1.18 (continuously payable annuity). Let t be a nonnegative real number, using a constant force of interest , find the value of an annuity payable continuously between time 0 and time t, where the rate of payment per unit time is the
constant 1 (this is known as continuously payable annuities and is denoted by a
te ).
Exercise 1.19 (cash flow). Let n be a non-negative integer, draw the cash flow
chart for a series of n payments, each of amount Xj , due at times tj , where j =
1, . . . , n. Find the value of these payments at time 0 and time tn .
Find a simple formula in the particular case when Xj = j, tj = j (this is known as
an increasing annuity and is denoted by (Ia)ne .
Hint. Multiply (Ia)ne by 1/v and compare with (Ia)ne , and use the fact that v =
1/(1 + i) where i is the effective interest rate for unit time.
Exercise 1.20 (variable discount rate in discrete time). Show that Ft is the value
at time t of the outstanding payment:
Ft = (1 + it+1 )1 xt+1 + (1 + it+1 )1 (1 + it+2 )1 xt+2
+ . . . + (1 + it+1 )1 (1 + it+1 )1 (1 + in )1 xn
(1.20.1)
Problem 1.21 (cash flow). A loan of 10, 000 pounds is to be repaid over 10 years
by a level annuity payable monthly in arrear. The amount of the monthly payment
is calculated on the basis of an interest rate of 1% per month effective. Find
(a) The monthly repayment.
(b) The total capital repaid and interest paid in the first and last year respectively.
18
(c) After which monthly repayment does the outstanding loan become first less
than 5, 000?
(d) For which monthly repayment the capital repaid first exceeds the interest?
CHAPTER 2
(2.1.1)
20
1/2
0
0
3/4
0
0
1
0
1
owed ammount
= (1 + ) +
| {z }
last payment
(1 + i)1/4
| {z }
(1 + i)1/2
| {z }
(1 + i)3/4
| {z }
(2.1.2)
A quick calculation using the geometric-telescopic identity,
(1 + r + + rn )(1 r) = (1 rn+1 ),
(2.1.3)
(2.1.4)
= (1 + i)1/4 1.
(2.1.5)
i=
and thus
A good name for is the (effective) interest rate payable quarterly (or fourthly).
One could be also interested in the (total) nominal interest rate payable quarterly
which is given by 4.
2.1.3. Definition (rate of interest convertible fractionally). A borrower, who is lent
1 at time 0 for repayment at time 1, agrees to pay the capital 1 at time 1 and to
spread the payment of the interest on his loan in p equal instalments over the time
interval [0, 1] (e.g., if 1 time unit stands for a year and we want to repay monthly
then p = 12). Supposing that the money is borrowed at an effective rate i over the
period [0, 1], which is compounded p-thly, how much interest should he pay at each
installment?
Denote by the amount of interest to be payed at the end of each instalment and
i(p) to be the total amount of interest of such a payment.1
So, the borower pays the lender the amount at times 1/p, 2/p, 3/p, . . . , 1. The
interest rate has the same value as each of the interest payment.
Let us express i(p) in terms of i. Since i(p) is the total interest paid, the p-thly
interest rate is constant, then each interest payment is of amount = i(p) /p. On
the other hand, the effective rate being i, the borrower owes 1 + i at time 1, (1 for
the capital repayment and i as interest). Each repayments at then end of each
interval [(t 1 p)/p, (t p)/p), with a fixed t = 1, . . . , p, bears a contribution of
(1 + i)(tp)/p
(2.1.6)
that this notation clashes with the one introduced in 1.1.4 of Chapter 1. We should be
using another notation, but this is the one that is found in most textbooks and so we stick with
it, with a warning.
p
X
i(p)
t=1
21
(1 + i)(pt)/p = i.
Excluding the trivial case where i = 0 = i(p) , and changing index t 7 p t, we have
i=
p1
X
(1 + i)1/p
t
t=0
i(p) (1 + i) 1
,
p (1 + i)1/p 1
i(p) = p (1 + i)1/p 1
(2.1.7)
p
i(p)
1.
i= 1+
p
(2.1.8)
Likewise we define d(p) to be the total amount of interest, payable in equal instalments at the start of each of the p subintervals, i.e., at times 0, 1/p, 2/p, . . . , (p1)/p.
Also here, the interest rate is taken to be constant with respect to the payment interval. It is thus a consequence of the definition of d(p) that
n
X
d(p)
(1 d)(t1)/p = d
p
t=1
or, if d 6= 0,
d(p) 1 (1 d)
=d
p 1 (1 d)1/p
Hence
d(p) = p 1 (1 d)1/p
and
p
d(p)
1
=1d
p
(2.1.9)
(2.1.10)
It is customary to refer to i(p) and d(p) as nominal rates of interest and discount
convertible p-thly.
2.1.4. Continuous time repayment. The continuous repayment option (c), with
a constant force of interest cannot be expressed in terms of a table; but we can still
think of the borrower, in order to earn 1 at time 0, to have to pay 1 at time 1 and
to spread the interest repayment by paying a constant infinitesimal amount dt at
each time t between 0 and 1. Thus the payment of dt at time t will provide an
interest of (1 + i)1t dt up to time 1. The total repayment is thus given by
Z 1
1+
(1 + i)1t
dt
| {z }
| {z }
0
interest factor from time t infinitesimal payment at t
(2.1.11)
Z 1
i
=1+
exp(log(1 + i)t) dt = 1 +
.
log(1 + i)
0
This amount must match the one from repayment option (a), hence
1+i=1+
i
,
log(1 + i)
(2.1.12)
(2.1.13)
22
i
d(p)
1
4
12
52
365
0.105171 0.101260 0.100418 0.100096 0.100014
0.095163 0.098760 0.099584 0.099904 0.099986
2.1.7. Example. Given that i = 0.08, find the values of i(12) , d(4) , and .
Solution.
d(4)
i(12) = 12 (1 + i)1/12 1 = 0.077208,
= 4 1 (1 d)1/4 = 4 1 (1 + i)1/4 = 0.076225,
= log(1 + i) = 0.076961.
2.1.8. Problem. Suppose that l and m are positive integers. Express i(m) in terms
of l, m, and d(l) . Hence find i(12) when d(4) = 0.057847.
Hint. Use to link the two quantities.
2.2. Annuity payable fractionally
We have learned about annuities in 1.7 of Chapter 1. The purpose of this section
is to extend the ways of payment of an annuity from one yearly instalment to p
p-thly instalments per year (assuming the year to be the period of the annuity).
A typical example where this is important are mortgages on a house purchase, where
the interest rate is given in a yearly percentage, but the payments occur monthly
(p = 12).
23
a
ne =
(p)
a =
(p) ne
1v n
i(p)
24
(p)
2.2.7. Annuity. Similarly, we define sne and sne to be the accumulated amounts
of the corresponding p-thly immediate annuity and annuity-due respectively. Thus
i
i
(p)
(p)
sne = (1 + i)n ane = (1 + i)n (p) ane = (p) sne
i
i
and
i
i
(p)
(p)
sne = (1 + i)n a
ne = (1 + i)n (p) ane = (p) sne .
d
d
The above proportional arguments may be applied to other varying series of payments. Consider, for example, an annuity payable annually in arrear for n years, the
payment in the tth year being xt . The present value of this annuity can be obtained
as
n
X
a=
xt v t .
t=1
Consider now a second annuity, also payable for n years with the payment in the
t-th year, again of amount xt being made in p equal instalments in arrear over that
year. If a(p) denotes the present value of this second annuity, by replacing the p
payments for year t (each of amount xt /p) by a single equivalent payment at the
end of the year of amount xt i/i(p) , we immediately obtain
a(p) = a
i
i(p)
(p)
(p)
m |ane
(p)
(p)
ane
m |
(p)
and a
ne .
2.2.11. Example.
Problem. An investor wishes to purchase a level annuity of 120 per annum
payable quarterly in arrear for five years. Find the purchase price, given that it
is calculated on the basis of an interest rate of 12% per annum:
(a) Effective
(b) Convertible half-yearly
(c) Convertible quarterly
(d) Convertible monthly.
Solution. (a) The value is
(4)
a5e
i(4)
= 451.583.
(2.2.1)
25
(b) Since the rate of interest is a nominal one convertible half-yearly, we take the
half-year as our unit of time and 6% as our rate of interest. The annuity is payable
twice per half-year for ten half-years at the rate of 60 per half-year. Hence its
value is
i
(2)
60a10e at 6% = 60 (2) a10e = 448.134
i
(c) We take the quarter-year as the unit of time and 3% as the rate of interest.
The value is thus
30a20e at 3% = 446.324.
(d) We take the month as the unit of time and 1% as the rate of interest. The
annuity payments of amount 30, at the end of every third month can be replaced
by a series of equivalent monthly payments, each of 30/s3e (at 1%). The value is thus
30
a60e at 1% = 445.084.
s3e
2.3. Annuities payable at time r > 1
In Section 2.2 we showed how, by replacing a series of payments to be received by
an equivalent series of payments of equal value, we could immediately write down
an expression for the value of a fractional annuity.
2.3.1. Extension to higher integers. The same technique, i.e., the use of equivalent payments of the same value, may be used to value a series of payments of
constant amount payable at intervals of time length r, where r is some integer
greater than 1.
For example, suppose that k and r are integers greater than 1 and consider a series
of payments, each of amount X,due at times r, 2r, 3r, . . . , kr. What is the value of
this series at time 0 on the basis of an interest rate i per unit time?
Let us replace the first payment of X due at time r by a series of r payments, each
of amount Y , due at times 1, 2, . . . , r, where Y is chosen to make these r equivalent
payments of the same total value as the single payment they replace. This means
that Y and X are related by Y sre = X at rate i.
Similarly each payment of amount X can be replaced by r equivalent payments of
amount Y of the same value. Thus the series of payments of X, due every rth time
interval, has the same value as a series of kr payments, each of Y = X/sre due at
unit time intervals.
Hence the value of the annuity equals
akre
Y akre = X
(2.3.1)
sre
at rate i.
By analogy with the p-thly payments we may extend the definition of annuity payable
p-thly in advance to fractional values of p = 1/r, r integer, by replacing n with kr
and X by r
rane
(1/r)
.
(2.3.2)
ane :=
sre
2.3.2. Exercise. Prove this result by using the first principles (i.e., by summing
the appropriate geometric progression).
26
1v n
i(p)
(p)
(2.4.1)
27
(p)
(p)
m |ane
(p)
(p)
ane
m |
(p)
and a
ne .
CHAPTER 3
30
DN DN DN
0
1
C = RN = redemption price
DN DN DN coupon
payments
time
n2n1
n+1
A = P N = purchase price
(a) above par, or at a premium, if a bond of 100 nominal is worth more than
100 (if nothing else is specified, this is meant to happen at redemption)
(b) below par, or at a discount, if a bond of 100 nominal is worth less than
100,
(c) at par, if a bond of 100 nominal is worth 100.
Price per unit: P is the (purchase) price per unit nominal. In practise, if nothing else is specified, stocks are quoted per 100 nominal.
Price: A = N P is the price of purchase of the holding.
Redemption price: C = N R is the cash received at redemption.
3.1.3. Ordinary shares or equities. These securities are issued by commercial
companies. The holders are entitles to all the net profits after deduction of payment
of interests on loans and fixed interest stocks.
If a share is bought ex-dividend, then the seller collects the dividend. If it is bought
cum dividend, then the buyer receives the dividend. This is different from fixed
interest securities where interest is shared according to the number of days owned
by the seller and buyer.
In most countries, taxation laws require a levy on the income from equities (or
securities in general). This is known as the dividend tax (even for ex-dividend
shares). We will denote the percentage of this share by and be careful not to
confuse it with the time variable t.
3.1.4. Example (Compound interest valuation of ordinary shares).
Problem. A pension fund, which is subject to 20% dividend tax but no capital
gains tax, has a large holding of a single UK company share with current market
value of 14 700 000. The current rate of dividend payment per year is 620 000p
(payable at the end of the year). The pension fund wishes to value its holding, at
an effective interest rate of 6% per annum effective, on the assumption that
(i) both dividend income from the share and its market value will increase at the
rate of 2% per annum,
(ii) the shares will be sold in 30 years time.
What value should the pension fund place on the shares?
Solution. Let us start by identifying the terms. We have = 0.2 (the tax rate),
the payment effected each year is Xk , based on the initial payment of X0 = 620 000
per year which must be multiplied by 1.02 for each additional year, i.e., Xk =
Xk1 (1.02) = X0 (1.02)k . Accounting for an effective interest rate of 0.06, we get
31
that the discount rate is v = 1/1.06. The number of years is n = 30. So, the capital
value at redemption time is C = 14 700 000 (1.02)n .
Thus, using the Equation of Value, the (present) value placed on the holding must
be
(1 )X0 (1.02)v + (1.02)2 v 2 + + (1.02)v n + Cv n
1 (1.02)n v n
= (1 )X0
(1.02)v + Cv n
1 (1.02)v
30
(1 (1.02/1.06)30 )1.02
1.02
= 620 000 0.8
+ 14 700 000
1.06 1
1.06
= 13 295 152.
(3.1.1)
(3.2.1)
32
A = N D/p +
| {z }
paym. 1.9.45
N Dane
| {z }
+ N
Rv n}
| {z
redemption
(3.2.2)
(2)
1/6
3+
(2)
6a22e
+ 100v
22
(3.2.3)
where i is the value we are after. To obtain i we must solve the nonlinear equation
F (i ) 117 = 0. This can be achieved with the help of a computational package
(or, less efficiently, by interpolation and trial-and-error). We obtain
i = 0.049437.
(3.2.4)
r
The following series of printouts is an example of Octave (or Matlab
) files that
2
solve this problem for us.
2The
r
only difference between Octave and Matlab
, in this case, is the endfunction keyword
r
which you should comment for Matlab use.
function F = equity1fun ( i )
v = vfromi ( i ) ;
F = v .^(1/6) .*(3+6* acpfromip (i ,22 ,2) +100* v .^22) -117;
endfunction
function ip = ipthly (i , p )
ip = p .*((1+ i ) .^(1/ p ) -1) ;
endfunction
function a = acpfromip (i ,n , p )
a = (1 - vfromi ( i ) .^ n ) ./ ipthly (i , p ) ;
endfunction
33
34
function v = vfromi ( i )
v = 1./(1+ i ) ;
endfunction
3.2.6. Example.
Problem. A newly issued stock bears interest at 7 21 % per annum, payable annually
in arrear, and is redeemable at par in 20 years time. Find the net yield per annum
to an investor, liable to income tax at 33 13 %, who buys a quantity of this stock at
80% of the nominal price.
Solution. We have coupon rate D = 0.075, price paid per unit nominal P = 0.8,
redemption price per unit nominal R = 1, rate of income tax = 13 , and term to
redemption n = 20. The equation of value is
P = D(1 )ane + Rv n at rate i,
i.e.,
0.8 = 0.05a20e + v 20 .
By bisection-interpolation (or using a Newton method on a computer), we find that
i = 6.8686% up to 4 digits.
3.3. Perpetuity
3.3.1. Why use perpetuity? If a security is undated, i.e., if it has no final redemption date, or if it runs for a very long time, it is regarded and valued as a .
(A practical example of such securities are loans on land purchase in Great Britain,
where the land property is sometimes leased for 999 years. These numbers are as
high as infinity for all practical purposes and a perpetuity calculation will work.)
3.3.2. How to use a perpetuity? Assume that the first interest payment is due
at time t years (t not necessarily integer) from the present, and that interest is at
rate D per annum per unit nominal, payable p times per annum. The price P per
unit nominal to give a net yield of i per annum, and liable to income tax at rate ,
may be found from the equation
D(1 )v t
at rate i.
d(p)
This equation is obtained by taking n in the relation
(p)
P = D(1 )v t a
e =
(p)
(p)
35
(3.3.1)
(3.3.2)
=Cv n + (1 )gCane
(3.4.2)
1 vn
i(p)
n
Recalling that K = Cv and rearranging terms we obtain
(1 )g
A=K+
(C K).
(3.4.3)
i(p)
Note that K is the present value of the capital repayment, and(C K) g(1 )/i(p)
is the present value of the net interest payments.
Relation (3.4.3) is known as Makehams formula and is of great importance.
=Cv n + (1 )gC
36
37
Since 66636.60/75000 = 0.8885, this price may be quoted as 88.85 for 100
nominal.
(b) Choose six months as the basic unit of time. The required yield per unit time
is 5%. Thus i = 0.05. Note now that interest is paid twice per time unit, so in the
notation above p = 2. Also, per time unit the amount of interest payable is 4% of
the outstanding loan, so now we have g = 0.04. The capital repayments occur at
times 10, 12, 14, ..., 38, so
5000
K=
(a40e a10e ) at 5% = 25377.27.
a2e
Hence the value of the entire loan is
0.04
25377.27 +
(75000 25377.27) = 65565.63
0.05(2)
or 87.42 for 100 nominal.
3.4.5. Example.
Problem. In relation to the loan described in Problem 3.4.4, find the price to be
paid on the issue date by a purchaser of the entire loan who is liable to income tax
at the rate of 40% and wishes to realise a net yield of 7% per annum effective.
Solution. The capital payments have value
K = 5000(a19e a4e ) at 7% = 34741.92.
Hence the price to provide a net yield of 7% per annum effective is
0.08(1 0.4)
(75000 34741.92) = 63061.89,
0.07(4)
i.e., after dividing the result by 100C, 84.08 for 100 nominal.
34741.92 +
3.4.6. Example.
Problem. A loan of nominal amount 100 000 is redeemable at 105% in four equal
instalments at the end of 5, 10, 15, and 20 years. The loan bears interest at the rate
of 5% p.a. payable half-yearly.
An investor, liable to income tax at the rate of 30%, purchased the entire loan on
the issue date at a price to obtain a net yield of 8% pa effective. What price did he
pay?
Solution. Note that the total indebtedness, C, is 100000 1.05 = 105000.
Each year the total interest payable is 5% of the outstanding nominal loan, so
that the interest payable each year is g times the outstanding indebtedness, where
g = 0.05/1.05.
Choose one year as the unit of time, then i = 0.08 and at the issue date the capital
payments have value
K = 25000 1.05(v 5 + v 10 + v 15 + v 20 ) at 8% = 43931.12.
Using the value of g described above, we obtain the price paid by the investor as
43, 931.12 +
0.05 (1 0.3)
(105, 000 43, 931.12)
1.05 0.08(2)
0.05 0.7
(105, 000 43, 931.12) = 69, 875.92
1.05 0.07846
which is equivalent to about 70 for 100 nominal.
= 43, 931.12 +
38
i) :=(1 )DN a + Cv n
A = A(n,
ne
(p)
(p)
=(1 )Gane + C 1 i(p) ane
(3.5.1)
(p)
at rate i. The first two relations are obtained from the definition of the price A
and Makehams formula, respectively. The last relation can be obtained by general
reasoning: if the net annual rate of interest per unit indebtedness were i(p) , the value
of the loan would be C. In fact the net annual rate of interest per unit indebtedness
is (1 )g, so the second term in the right-hand side of the last equation is the value
of net interest in excess of the rate i(p) .
The following are immediate conse3.5.2. Behaviour of the price function A.
quences of equations in (3.5.1):
i) = C, i.e., the function A(,
i)
(a) If i(p) = (1 )g, then, for any value of n, A(n,
is a constant equal to the capital at redemption C.
i) is increasing.
(b) If i(p) < (1 )g, then, the function A(,
i) is decreasing.
(c) If i(p) > (1 )g, then, the function A(,
) is decreasing.
(d) For any fixed n, the function A(n,
3.5.3. Term to redemption variation and comparison. Consider now two
loans, each of which is as described in the first paragraph of this section except that
the first loan is redeemable after n1 years and the second loan after n2 years, where
MTA (G5078) Autumn 2009 3.5. The effect of the term to redemption
on the yield
39
n1 < n2 . Suppose that an investor, liable to income tax at a fixed rate, wishes to
purchase one of the loans for a price A. Then
(a) If A < C (where C = N R), the investor will obtain a higher net yield by
purchasing the first loan (i.e., the loan which is repaid earlier).
(b) If A > C, the investor will obtain a higher net yield by purchasing the second
loan (i.e., the loan which is repaid later).
(c) If A = C, the net yield will be the same for either loan.
3.5.4. Exercise. Give an intuitive explanation to the above conclusion.
3.5.5. Partial redemption by instalments. When an investor purchases part
of a loan redeemable by instalments, the yield he will obtain depends on the actual
date (or dates) at which his holding is chosen for redemption. In relation to such
a loan, issued in bonds of equal nominal amount, suppose that a nominal amount
Nr is redeemable at time nr t = 1, . . . , k where n1 < n2 < < nk . Suppose that
(for each bond) the purchase and redemption prices per unit nominal are P and R
respectively.
Consider a purchaser of one bond, subject to income tax at rate , and let the net
yield per annum which he will obtain if his bond is redeemed at time n, be denoted
by ir , r = 1, . . . , k.
If the bonds redeemed at anyone time are drawn by lot, the probability of obtaining
a particular yield ir is equal to
Nr
pr = Pk
r=1
Nr
and the expected value of the yield, in the probabilistic sense, is therefore
Pk
k
X
Nr ir
pr ir = Pr=1
.
i =
k
r=1 Nr
r=1
It should be noted that this quantity is not in general equal to the net yield i on
the whole loan, but in most practical cases i and i will be quite close to each other.
It is also clear that both i and i will lie somewhere between i1 and ik .
3.5.6. Problem (partial redemption). A loan of nominal amount 80, 000 is redeemable at 105% in four equal instalments at the end of 5, 10, 15, and 20 years.
The loan bears interest at the rate of 10% p.a. payable half-yearly.
Suppose that an investor, who is subject to income tax at 30%, purchases one bond
of 100 nominal on the issue date for 95.82. Find the net yield per annum he will
obtain, assuming redemption after 5, 10, 15, and 20 years, and plot these net yields
on a graph. Find also the probability that the net yield will exceed 9% p.a.. Find
the expected value of his yield and show that it is not identical with the net yield
which he would obtain if he purchased the entire issue at the same price.
3.5.7. Problem (partial redemption). A loan of nominal amount 100 000 is to
be issued in bonds of nominal amount 100 bearing interest of 6% per annum
payable half-yearly in arrear. The loan will be repaid over 20 years, 50 bonds being
redeemed at the end of each year at a price of 120 per 100 nominal (also denoted
by 120%). The bonds redeemed in anyone year will be drawn by lot (so the buyer
of one particular bond ignores its redemption time but knows the probabilities of it
being redeemed in a certain year). The issue price of the loan is 94.32%.
An investor, liable to income tax at the rate of 25% is considering the purchase of
all or part of the loan.
40
(a) Show that if he purchases the entire loan, his net annual yield on the transaction
will be 7%.
(b) Show that if he purchases only one bond his net annual yield could be as high
as 32.36% or as low as 5.61% and find the probability that he will achieve a net
annual yield of (i) at least 8% and (ii) between 6% and 8%.
3.5.8. Exercise. A loan of nominal amount 1000 is to be issued in bonds of
nominal amount 10 bearing interest of 4% per annum payable quarterly in arrear.
The loan will be repaid over 10 years, 10 bonds being redeemed at the end of each
year at a price of 110 per 100 nominal. The bonds redeemed in anyone year will
be drawn by lot. The issue price of the loan is 90%.
An investor, liable to income tax at the rate of 20% is considering the purchase of
all or part of the loan.
(a) Calculate his net annual yield if he purchases the entire loan at the outset.
(b) If he purchases only one bond, find the probability that he will achieve a net
annual yield of (i) at least 6% and (ii) between 6% and 10%.
3.6. Real returns, inflation and index-linked stocks
3.6.1. Price indexes and inflation. Inflation may be defined as a fall in the
purchasing power of money. It is usually measured with reference to an index
representing the cost of certain goods and (perhaps) services. For example, in the UK
the index used most frequently is the Retail Prices Index (RPI), which is calculated
monthly by the Central Statistical Office.
Real investment returns, as opposed to the money (or cash) returns we have so far
considered, take into account changes in the value of money, as measured by the
RPI or another such index. It is possible for all calculations relating to discounted
cash flow, yields on investments, etc. to be carried out using units of real purchasing
power rather than units of ordinary currency.
A price index is a number that quantifies the amount of currency needed to purchase
one unit of average goods. This number is relative to some fixed time t0 where we
consider the index to be 100% (i.e., 1 unit of good costs 1 unit of currency). Having
fixed t0 , let us define Q(t) to be price of one unit of goods at time t.
For example, the following table describes the RPI
Calendar year 1980 1981 1982 1983
Value of RPI for January 245.3 277.3 310.6 325.9
Annual inflation, calculated over a period of 12 months, at time t is defined as the
rate (expressed in percentile) in the price index increase over a year, i.e., that number
j(t) such that
Q(t)(1 + j) = Q(t 1),
(3.6.1)
for each fixed t and when the time unit is a year. So in relation to the example we
have that the annual inflation in Jan 1981, with respect to 12 months earlier, is of
j=
Q(1)
277.3
1=
1 = 0.1304 = 13.04%.
Q(0)
245.3
(3.6.2)
As an exercise calculate the annual inflation in 1982 and 1983, with respect to 12
months earlier.
Note the relative nature of the price index Q: changing the unit of goods will
change the values of Q, but not their ratios. For example, an equally good table as
the one above is given by
41
(3.6.5)
(3.6.7)
This is the equation of value for the transaction, measured in units of purchasing
power at a particular time tk . It means that the values of Q can be taken as relative
values against any arbitrarily fixed reference purchasing power (namely, Q(tk )).
3.6.4. Problem (accounting for inflation). On 16 January 1980 a bank lent 25, 000
to a businessman. The loan was repayable three years later, and interest was payable
annually in arrear at 10% per annum. Ignoring taxation and assuming that the RPI
for any month relates to the middle of that month, find the real annual rate of
42
return, or yield, on this transaction. Values of the RPI for the relevant months are
summarised in the following table.
Calendar year 1980 1981 1982 1983
Value of RPI for January 245.3 277.3 310.6 325.9
Exercises and problems on Application of interest rate
Problem 3.1 (fixed income investment: priceyield relation). A loan of nominal
amount 550 000 is to be issued bearing interest of 10% pa payable half-yearly. At
the end of each year part of the loan will be redeemed at 105%, The nominal amount
redeemed at the end of the first year will be 10 000 and each year thereafter the
nominal amount redeemed will increase by 10 000 until the loan is finally repaid.
The issue price of the loan is 90 for 100 nominal.
Find the net effective annual yield to an investor liable to income tax at 40%, who
purchases the entire loan on the issue date.
Problem 3.2 (fixed income investment: priceyield relation). 5 years ago a loan
was issued bearing interest payable annually in arrear at the rate of 8% per annum.
The terms of issue provided that the loan would be repaid by a level annuity of
1 000 over 25 years.
An annuity payment has just been made and an investor is considering the purchase
of the remaining instalments. The investor will be liable to income tax at the rate
of 40% on the interest content (according to the original loan schedule) of each
payment. What price should the investor pay to obtain a net yield of 10% per
annum effective?
Problem 3.3 (Makehams formula). A loan of nominal amount 1, 000, 000 is to
be issued bearing interest of 10% p.a. payable half-yearly. At the end of each year
part of the loan will be redeemed at 105%, The nominal amount redeemed at the
end of the first year will be 1, 000 and each year thereafter the nominal amount
redeemed will increase by 10, 000 until the loan is finally repaid. The issue price
of the loan is 98.80 for 100 nominal.
Find the net effective annual yield to an investor liable to income tax at 40%, who
purchases the entire loan on the issue date.
Problem 3.4 (partial redemption). A loan of nominal amount 80, 000 is redeemable at 105% in four equal instalments at the end of 5, 10, 15, and 20 years.
The loan bears interest at the rate of 10% p.a. payable half-yearly.
Suppose that an investor, who is subject to income tax at 30%, purchases one bond
of 100 nominal on the issue date for 95.82. Find the net yield per annum he will
obtain, assuming redemption after 5, 10, 15, and 20 years, and plot these net yields
on a graph. Find also the probability that the net yield will exceed 9% p.a.. Find
the expected value of his yield and show that it is not identical with the net yield
which he would obtain if he purchased the entire issue at the same price.
Problem 3.5 (partial redemption). A loan of nominal amount 100 000 is to
be issued in bonds of nominal amount 100 bearing interest of 6% per annum
payable half-yearly in arrear. The loan will be repaid over 20 years, 50 bonds being
redeemed at the end of each year at a price of 120 per 100 nominal (also denoted
by 120%). The bonds redeemed in anyone year will be drawn by lot (so the buyer
of one particular bond ignores its redemption time but knows the probabilities of it
being redeemed in a certain year). The issue price of the loan is 94.32%.
An investor, liable to income tax at the rate of 25% is considering the purchase of
all or part of the loan.
43
(a) Show that if he purchases the entire loan, his net annual yield on the transaction
will be 7%.
(b) Show that if he purchases only one bond his net annual yield could be as high
as 32.36% or as low as 5.61% and find the probability that he will achieve a net
annual yield of (i) at least 8% and (ii) between 6% and 8%.
Problem 3.6. A loan of nominal amount 1000 is to be issued in bonds of nominal
amount 10 bearing interest of 4% per annum payable quarterly in arrear. The loan
will be repaid over 10 years, 10 bonds being redeemed at the end of each year at a
price of 110 per 100 nominal. The bonds redeemed in anyone year will be drawn
by lot. The issue price of the loan is 90%.
An investor, liable to income tax at the rate of 20% is considering the purchase of
all or part of the loan.
(a) Calculate his net annual yield if he purchases the entire loan at the outset.
(b) If he purchases only one bond, find the probability that he will achieve a net
annual yield of (i) at least 6% and (ii) between 6% and 10%.
Problem 3.7 (accounting for inflation). On 16 January 1980 a bank lent 25, 000
to a businessman. The loan was repayable three years later, and interest was payable
annually in arrear at 10% per annum. Ignoring taxation and assuming that the RPI
for any month relates to the middle of that month, find the real annual rate of
return, or yield, on this transaction. Values of the RPI for the relevant months are
summarised in the following table.
Calendar year 1980 1981 1982 1983
Value of RPI for January 245.3 277.3 310.6 325.9
Problem 3.8 (inflation). On 1 January 1970 a bank lent 1, 000 to a businessman.
The loan was repayable three years later, and interest was payable annually in arrear
at 5% per annum. Ignoring taxation and assuming that the RPI for any month
relates to the middle of that month, find the real annual rate of return, or yield,
on this transaction. Values of the RPI for the relevant months are reported in the
following table.
Calendar year 1970 1971 1972 1973
Value of RPI for January 225.3 257.3 290.6 305.9
CHAPTER 4
46
time 0
S(0) = certain
B(0) = certain
time 1
(
S u with chance p
S(1) =
S d with chance (1 p)
B(1) = certain
assets: a risk-free one, which we call B (for bond) and a risky one which we call S
(for stock).
4.1.3. A two-instant model. To get to a reasonable model, we need to add time.
As we have learned in previous chapters, time can be modelled in two ways: continuous and discrete. The simplest form of discrete time can be modelled with a sequence
of equally spaced time instants, usually normalised to 1 time-unit, i.e., 0, 1, 2, . . . , N
where N is some final time for the model. Such models entail the knowledge (or the
pretension thereof) of the asset prices at all instants. By knowledge we mean a
stochastic type of knowledge for the random variables, which means knowledge of
the chances rather than the outcomes.
To make things the simplest possible we will use only 2 times, i.e., take N = 1
above. So we have an initial time 0 and a final time 1. Think of these as the
beginning of the year and the end of the year. This is the so-called two-instant
(or one-step) model, and, after combining it with a two-asset model it becomes
the two-by-two market model, also called simple market model. Finally, we will
consider the uncertain asset (i.e., the stock) to have a binomial price at the end
time; this leads to the so-called one-step binomial market model. We can summarise
the situation in Table 1.
4.1.4. Definition (return on an asset). An important financial quantity (maybe
the most important one) is the return on a given asset over a given time interval.
In our simple market model we have two returns:
B(1) B(0)
S(1) S(0)
KB :=
and KS :=
.
(4.1.1)
B(0)
S(0)
Note that while KB is certain, KS is uncertain (i.e., genuinely random).
4.1.5. Example (simple market). Consider the following simple market
asset time 0
time 1
(
125 with chance 45%
S
S(0) = 100 S(1) =
95
with chance 55%
B
B(0) = 100 B(1) = 110
The return on B is easy,
110 100
KB :=
= 10%,
(4.1.2)
100
whereas the return on S takes just a bit more math (and we use vector notation to
compress the repetitive calculations)
125
100
95
25%
45%
KS :=
=
with chance
.
(4.1.3)
5%
55%
100
47
(4.2.2)
This is, of course, equivalent to 0 < 1 p < 1, for those who have a glass half-empty
view of life.
4.2.3. Definition (portfolio). A portfolio is a certain quantity of the possible assets that we (or the investor, if you prefer) holds. Mathematically a portfolio is
|
arranged in a (column) vector, say w = x, y , where x is the quantity of S and y
is the quantity of B held (portfolio means wallet in rennaissance Italian, whence
the symbol w).
4.2.4. Hypothesis (Divisiblity, liquidity and position of assets). We assume that
assets can be held and traded in arbitrary quantities. Mathematically, this means
x, y (and any variation thereof) belongs to R. Negative values of x or y mean that we
(investor) are short on (also known as in a short position), i.e., we owe to someone,
the asset S or B, respectively, while postive values mean that we are long on (also
known as in a long position), in the sense that we actually possess, the asset S or
B, respectively.
4.2.5. Definition (value of a portfolio and market matrix).
| Suppose we hold a
|
portfolio w = x, y in a one-step binomial market S , B , then the value of w
at time t {0, 1} is defined as
x
V (t) := xS(t) + yB(t) = S(t), B(t)
= [M ]t w,
(4.2.3)
y
where we use the market vector [M ]t , as being the t-th row of the market matrix
S(0) B(0)
M=
,
(4.2.4)
S(1) B(1)
which is a random matrix because the entry S(1) is random.
4.2.6. Hypothesis (admissible porfolio and solvency). Given a simple market with
matrix M and a portfolio w, we say that w is an admissible portfolio for M if and
only if
V (t) = [M ]t w > 0
1Our
(4.2.5)
48
If we have an admissible portfolio, we say that we are solvent. This means that we
are able to close any short position in the portfolio (by using the long ones). If one
holds a portfolio for which V (t) 0 for some t and positive probability, we say that
the investor is insolvent or bankrupt.
4.2.7. Exercise. Consider the simple market
asset time 0
time 1
(
52 with chance p
S
S(0) = 50 S(1) =
48 with chance 1 p
B
B(0) = 100 B(1) = 110
.
(1) Calculate the returns KB and KS .
|
(2) Consider the portfolio w = 20, 10 , and calculate its value, as an algebraic
expression, at the generic time t.
(3) We define the (total) return on the portfolio w as
V (1) V (0)
KV :=
.
(4.2.6)
V (0)
Calculate KV .
(4) Is it true that
KV = xKS + yKB ?
(4.2.7)
(5) Show that
S(0)
.
(4.2.8)
KV = 0 KS + (1 0 )KB , where 0 = x
V (0)
(6) Deduce that (4.2.7) is generally wrong, but the following is true
min {KS , KB } KV max {KS , KB } .
(4.2.9)
49
(4.3.2)
50
what. In other words the bond in Exercise 4.3.6 is dearer2 than the stock. Indeed,
we will now prove this in a more general setting.
4.4.2. Theorem (intermediate bond-price property). Consider a binomial simple
market that operates under the blanket assumptions 4.2 and the NAP 4.3.7. If
B(0) = S(0)3 then
S d < B(1) < S u .
(4.4.1)
Proof Without loss of generality let us assume that B(0) = 100, hence S(0) = 100.
Suppose that B(1) S d , then the buy-cheap-sell-expensive and the fact that B(1)
is cheap in the future, makes it expensive. Let us go short on B then for starters:
borrow 1 of B at time 0 (i.e., set y = 1), and use the 100 to buy S (i.e., set
x = 1). With this portfolio we have V (0) = 0. Furthermore at time t = 1 we obtain
u
u
S B(1)
S B(1)
1
0
V (1) = d
= d
,
(4.4.2)
0
S B(1) 1
S B(1)
[]:
Check!
thus V (1) 0 surely and V (1) > 0 with some positive probability, which amounts
to an arbitrage opportunity.
If we suppose B(1) S u then it is enough to invert the roles, x = 1 and y = 1 to
find an arbitrage opportunity.[] Therefore a no-arbitrage assumption implies that
S d < B(1) and B(1) < S u .
4.4.3. Corollary (intermediate bond-return property). Consider a binomial simple
market that operates under the blanket assumptions 4.2 and the NAP 4.3.7..
Then, using the returns defined in 4.1.4, we have
min KS < KB < max KS .
(4.4.3)
S d S(0)
S u S(0)
and max KS = KS u :=
.
S(0)
S(0)
(4.4.4)
Proof left as exercise Extend the Intermediate Bond-Price Theorem 4.4.2 to the
case where B(0) 6= S(0) and then take the returns.
Hint. Normalise, apply the theorem and then denormalise back.
4.4.4. Risk and return (yield). Consider the binomial simple market given by
asset time 0
time 1
(
100 with chance 80%
S
S(0) = 80 S(1) =
60
with chance 20%
B
B(0) = 100 B(1) = 110
and suppose you have 10 000 to be invested and you decide to go with a portfolio
of
x = 50 of S and y = 60 of B.
(4.4.5)
(Check that V (0) = 10 000.) Then we have
100 110 50
11600
80%
V (1) =
=
with chance
.
(4.4.6)
60 110 60
9600
20%
2We
51
16%
= 80% 20%
= (12.8 0.8)% = 12%.
4%
(4.4.7)
(4.4.8)
This give us the expected return, but it does not tell us how much we are risking.
In fact, though small, the probability of making a loss on S is not negligible. The
risk can be quantified by using the variance of the return KV or, more appropriately
in terms of units, the variances square root (also known as standard deviation). We
define the risk V associated to our portfolio as follows
p
V := var KV .
(4.4.9)
In our context, we evaluate
var KV = E (KV E KV )2 = E[KV ]2 (E KV )2
= KV2 u P u + KV2 d P d (0.12)2 = 64 104
and hence the risk associated with this portfolio is
(4.4.10)
(4.4.11)
52
Exercise 4.3. Consider two assets, one risky and one risk-free, denoted S and
A respectively. Suppose their prices are
asset time 0
time 1
(
S u with chance p
S
S(0) = 34 S(1) =
S d with chance 1 p
B
B(0) = 100 B(1) = 112
Is it possible to find an arbitrage opportunity if the forward price of stocks is F =
38.60 with delivery date 1?
Exercise 4.4. Let the bond B and stock S prices be
asset time 0
time 1
(
120 with chance p
S
S(0) = 100 S(1) =
80
with chance 1 p
B
B(0) = 100 B(1) = 110
Price (i.e., find C(0) for) a call option with exercise time 1 and
(a) strike price F = 90,
(b) strike price F = 110.
CHAPTER 5
(5.1.2)
for s = 1, . . . , S.
5.1.2. Portfolio and realized
| return table. The investor chooses a represented
by the vector n = n1 , . . . , nN , where ni is the number of units (shares) asset held
for i = 1, . . . , N . The initial values of the assets are the components of a vector v,
and the is n| v = W0 .
The (1+ rate of return) on the i-th asset in the state s is Zsi = Ysi /vi . The state
space table of return is
Z = Y D 1
(5.1.3)
v
where D v is the diagonal matrix with i-th element vi and A1 indicates the inverse
matrix of a generic square matrix A. We assume that vi 6= 0 to ensure that D v is
invertible.
The i-th column of Z represents the (vector of) S possible returns on the asset i,
corresponding to all possible states ranging from 1 to S. The row vector z s | , which
will denote the s-th row of Z has N components, each of which represents the return
on each of the N assets, in a given state s = 1, . . . , S. The matrix Z is often referred
to as an .
e a random vector whose possible realizations are the (transposed)
We denote by z
rows of Z, z s . When we say we mean that each state s = 1, . . . , S has a certain
probability (which may be known or unknown to the investor) of occurring. In
e, denoted by zi , is a scalar-valued
particular, this means that each component of z
random variable defined on the {1, . . . , S}.
54
In the return space, investments are characterized by a vector , called the , where
each component is given by
i = ni vi .
(5.1.4)
If the investment involves a total net amount, called the total commitment,
|
W0 = n v =
N
X
i = C> 0,
(5.1.5)
i=1
1 3
(5.1.6)
Z = 2 1 .
3 2
|
An investment commitment of = 1, 3 has state-by-state outcomes of
|
Z = 10, 5, 9
(5.1.7)
|
and an original commitment of 4. The corresponding portfolio is w = 1/4, 3/4
|
with returns Zw = 2.5, 1.25, 2.25 .
5.1.6. Riskless portfolios and linear programming. To gain some insight into
how an arbitrage portfolio may be financed, we define a riskless portfolio to be a
portfolio w with the same return in any state, i.e., w satisfies
Zw = R1 and 1| w = 1,
(5.1.8)
for some constant R, called the riskless return. The constant R 1 defines the
risk-free rate.
We are assuming here that the return R is a constant with respect to s = 1, . . . , S,
but in general this may not be true. Indeed, for a given w it may be that the vector
Zw had non-constant entries.
1Make
sure you distinguish the latin letter double you w from the greek letter omega .
55
This observation prompts us to introduce the following returns which will be useful
later:
|
R := max
min Zw = max
min z s w ,
(5.1.9)
|
|
1 w=1
1 w=1 s=1,...,S
and
|
R := min
max Zw = max
min z s w
|
|
1 w=1
1 w=1 s=1,...,S
(5.1.10)
we may well get R 6= R. The minimax and maximin problems to be solved to obtain
these returns are known as linear programming problems.
5.1.7. Remark. In reality, the Arbitrage Principle in economics can be phrased as
follows:
There is never an opportunity to make a risk-free profit that gives a greater
return than that provided by the interest from a bank deposit.
A simple common-sense argument that supports this principle is that if there
were an opportunity to make a risk-free profit greater than that provided by a bank
deposit, then no one would be depositing in the bank. This would force the bank
to raise the interest rate (for example by investing in that risk-free profit scheme)
and the situation would end up being balanced. In other words, if the market is out
of equilibrium (because of an arbitrage opportunity) then it will be driven to the
equilibrium. This is called the arbitrage-free condition under which economists and
financial theorists operate.
In real life, arbitrageurs make profit by borrowing bank money to invest, they take on
risks to get better return. So they may well lose the money as well as making more.
The bottom line is that a successful arbitrageur must rely on extra-information (e.g.,
insider trading) to perform real arbitrage. Without such infomation, the probability
of winning or loosing is 50% each.
5.1.8. Exercise (can you spot arbitrage?). For each of the following, state whether
or not there is an arbitrage opportunity:
(a) An individual needs change and is offering you a 10 pound note in return for 9
pounds change.
(b) England football team is playing Croatia tonight, a friend says that he will pay
you 20 pounds if England loses and if you support Croatia during the game.
(c) Mr A offers to purchase a car for 12000 pounds, Mr B offers to buy the same
car for 15000 pounds.
(d) You and your friend are betting on the weather tomorrow. If it does not rain,
your friend will pay you 2 pounds, if it rains, you pay your friend 2 pounds.
(no, gamble 6= arbitrage)
5.2. Redundant Assets
5.2.1. Duplicable portfolios. Consider a given return table Z. Let w1 denote a
specific portfolio for this table. This portfolio is duplicable if and only if there exists
w2 6= w1 such that Zw2 = Zw1 .
Note that if there exists a pair of duplicable portfolios w1 6= w2 , then any vector w
is duplicable. Indeed, for x := w + (w1 w2 ), we have
Zx = Zw + Zw1 Zw2 = Zw.
(5.2.1)
What really matters thus, when identifying redundant assets, is their difference
w1 w2 , which constitutes an arbitrage portfolio. This prompts the following.
56
redundant?
5.2.4. Problem
(redudant assets and linear algebra). What can be said of the
Z
matrix | if one of the assets in Z is redundant?
1
5.3. Contingent Claims and Derivative Assets
A contingent claim or derivative assets end of term payoff can be precisely determined by the payoff on one or more of the other assets.
Common types of derivative assets are put options and call options, warrants, rights
offerings, futures and forward contracts, and convertible or more exotic types of
bonds. Let us define some of these.
5.3.1. Definition (forward contracts, options, call and put). A forward contract is
the obligation to purchase or sell, at the end of the contracts period (also known as
expiration time), a set number of units (shares) of a particular primitive asset at a
price specified at the beginning of the period. This is a one-period forward contract
or option.
It is important to distinguish the buyer/seller of the contract and buyer/seller of
the shares. This is not the same, as we explain next.
Call option: known briefly as call.
The buyer of the option has the right, but not the obligation, to buy an
agreed quantity of units (shares) from the seller of the option at a certain time
(the expiration date or period) for a certain price (the strike price). The seller
(also known as writer) is obligated to sell the commodity or financial instrument
should the buyer so decide. The buyer pays a fee (called a premium) to acquire
this right. The strike price is sometimes referred to as exercise price.
Put option: known briefly as put.
The put allows the buyer the right but not the obligation to sell an agreed
quantity of units (shares) to the writer of the option at a certain time for
a certain price (the strike price). The writer (seller of the option) has the
obligation to purchase the underlying asset at that strike price, if the buyer
exercises the option.
5.3.2. Essay. Options come in different styles. There are European style options
where the expiration time (period) is fixed, and there are American style options
where the expiration time can be decided by the buyer of the option. Take a Library
(or Wikipedia) tour and explore the various different flavours. Write a report.
5.4. Insurable portfolios and states
An insurable portfolio, with respect to a given economy Z, is a portfolio that pays
off only in a particular state.
57
5.4.1. Definition (insurable state). The state s is insurable if there exists a vector
s such that
(
6= 0 if = s,
|
z s
(5.4.1)
= 0 if 6= s.
Moreover, by rescaling s appropriately, we may assume
z s | s = 1.
(5.4.2)
|
In matrix parlance, this means that Z s = es where es := 0, . . . , 0, 1, 0, . . . , 0 ,
that is the s-th column of the S S identity matrix.
In this case the scalar 1| s is the cost per currency unit (say ) of the insurance
policy against the occurrence of state s and s is the corresponding insurable portfolio.
5.4.2.
Theorem
(characterization of insurable states). For a given economy Z =
|
z 1 . . . z S and a state s = 1, . . . , S, the following are equivalent.
(i) State s is insurable.
(ii) The vector of asset returns in this state, z s is linearly independent of the vectors
of assets returns on in the other states, z , 6= s.
(iii) Denoting by es the s-th column of the identity matrix we have
(5.4.3)
rank Z es = rank Z.
Proof
[(i) (ii)] Suppose s is insurable, we want to show that z s cannot be expressed as
a linear combination of the other state vectors, z s.
Suppose, by contradiction that z s is linearly dependent on the z s , i.e.,
X
z ,
(5.4.4)
zs =
6=s
z | .
(5.4.5)
6=s
(5.4.6)
(5.4.7)
and thus
1 = zs|s =
s z | s = 0,
6=s
(5.4.8)
where the span of a matrix is understood as the span of its columns, namely,
span Z | = span {z : 6= s} , and span Z | = span {z : = 1, . . . , S} . (5.4.9)
Introduce now an orthonormal basis of the space span Z | ,
{l span Z | : l = 1, . . . , L}
(5.4.10)
58
It follows that
2
z s s = |z s |
L
X
(z s | l )2 > 0
(5.4.12)
l=1
L
X
l l
(5.4.13)
l=1
z s =
L
X
l l z s
l=1
L
X
l z s | l = 0.
(5.4.14)
l=1
3 2
Z = 1 2 ,
(5.4.15)
2 4
|
|
the portfolio 2, 1 gives a return 4, 0, 0 , which means that 1 is an insurable
state corresponding to the insurable portfolio
1 := 1/2, 1/4 .
(5.4.16)
So insurance against state 1 is possible at cost 1/4 per unit. The returns in states
2 and 3 are collinear (i.e., z 2 linearly depends on z 3 and viceversa), so insurance is
not possible for either.
5.5. Dominance and arbitrage
5.5.1. Definition (dominant portfolio). We say that portfolio w1 dominates w2 if
and only of
Zw1 > Zw2 ,
(5.5.1)
where given two vectors x and y we define
x > y xi yi , for all i = 1, . . . , N and xi > yi , for at least one i . (5.5.2)
Thus the > may be a non-strict inequality for some components but it must be
strict on one component at least. We denote this relation by w1 w2 .
Since the dominated portfolio never outperforms the dominating one, investor should
prefer to hold the dominating one.
5.5.2. Proposition (dominance and unbounded portfolios). If a dominated portfolio exists, then there is no unbounded optimal portfolio.
59
(5.5.4)
(5.5.6)
(5.5.7)
1
0
Z = 1 1
(5.5.11)
1
1
60
1
Z = 1 2 .
(5.5.12)
1 + 2
For 1 = 1 and 2 = 1 we see that
1| = 0
(5.5.13)
while
1
0
Z = 0 > 0 .
(5.5.14)
0
0
This creates an arbitrage opportunity of the first type. As an exercise show that
there is no arbitrage of second type for Z.
5.5.7. Problem. Show that if there exists a positive investment portfolio, say w >
0, with semi-positive return, that is, Zw 0, then an arbitrage opportunity of the
second type guarantees the existence of an arbitrage opportunity of the first type.
5.6. Pricing under NAP
As we have remarked, investment equilibrium can only be achieved in the absence
of arbitrage. This is why a proper pricing theory works (i.e., for it to have realistic
relevance) only under the No-Arbitrage Principle (NAP). The goal of this section is
to understand how to price a given economy.
5.6.1. Definition (pricing). Given an economy in the form of a table Z, a vector
p RS is said to be a pricing vector supporting the economy Z if and only if
Z | p = 1,
(5.6.1)
p| Z = 1 | .
(5.6.2)
which is equivalent to
5.6.2. Theorem (characterisation of pricing vectors). There exists a non-negative
pricing vector, p RS , supporting an economy Z if and only is there are no arbitrage
opportunities of the second type for Z.
Proof Consider the linear programming problem of finding 0 such that
1| 0 = m = min {1| : Z 0} .
(5.6.3)
61
5.6.3. Example. This theorem does not prohibit the existence of pricing vectors
with some negative prices. For example, for the economy
1 3
Z = 2 1
(5.6.5)
3 2
the pricing support equations are
p1 + 2p2 + 3p3 = 1
3p1 + p2 + 2p3 = 1
(5.6.6)
p1 = 0.2(1 )
p2 = 0.2 1.4
p3 = .
(5.6.7)
with solution
|
This problem has thus a positive solution p = 0.18, 0.26, 0.1 . Hence no arbitrage
|
opportunities of 2nd type available. Nevertheless, the pricing vector 0.1, 0.3, 0.5
also supports the economy.
5.6.4. Example. Consider the economy
3 2
Z = 1 2
2 4
(5.6.8)
(5.6.9)
5
2
Z=
5 2
(5.6.10)
5p1 5p2 = 1
2p1 2p2 = 1
(5.6.11)
rank Z 6= rank Z | 1
(5.6.12)
By comparing
(or by trying to solve it), this system is seen to have no solution. Hence arbitrage
opportunities of second type exist for this economy.
5.6.6. Exercise (arbitrage via pricing). Derive pricing equations for the economy
5 2
Z = 3 1
(5.6.13)
2 1
and find out if there are any arbitrage opportunities of first or second type.
62
5.6.7. Exercise. Find out whether there are arbitrage opportunities (first or second
type) in an economy Z such that its transposed matrix is
2 1 3
|
Z =
.
(5.6.14)
1 1 3
5.6.8. Theorem. There exists a positive pricing vector p which supports an economy Z if and only if there are no arbitrage opportunities of first nor second type.
Proof Necessity Assume is an arbitrage opportunity of the first type. Then
Z 0 and p > 0 imply
p| Z > 0.
(5.6.15)
Since p is a pricing support vector, p| Z = 1| . and we get
1| > 0
(5.6.16)
5.6.9. Remark. The positivity of the pricing vector assures the absence of arbitrage
and the linearity guarantees the absence of any monopoly power in the financial
market.
5.7. Riskless Issues
From Section 5.1.6 we know that an economy Z has a riskless asset if there exists a
portfolio w1 with 1| w1 = 1, and Zw1 = R1.
5.7.1. Proposition. If there is a riskless asset or portfolio with return R, then in
the absence of arbitrage, the sum of the state prices of any valid supporting pricing
vector is equal to 1/R.
Proof By assumption, there exists a portfolio w1 with
1| w1 = 1,
Zw1 = R1.
(5.7.1)
Then for all valid supporting price vectors p, recalling that p| Z = 1| , we have
R
S
X
(5.7.2)
s=1
which implies
S
X
ps =
s=1
1
.
R
(5.7.3)
More generally, if there is no way to construct a riskless portfolio, then we have the
following result.
5.7.2. Proposition. The sum of the state prices is bounded below by 1/R and
above by 1/R where R and R are defined by
R := max
min Zw
|
(5.7.4)
R := min
max Zw.
|
(5.7.5)
1 w
and
1 w
63
Proof By (5.7.5) there exists a portfolio w1 such that R = max Zw1 and thus
Zw1 R1.
For any princing vector p we have p| Z = 1| and thus
p| 1R p| Zw1 = 1| w1 = 1,
from which
S
X
ps
s=1
1
.
R
(5.7.6)
(5.7.7)
(5.7.8)
(5.7.9)
and thus
R1 Zw2 .
By the same argument as above it follows that
p| (1R) p| Zw2 = 1| w2 = 1,
from which
S
X
s=1
ps
1
.
R
(5.7.10)
(5.7.11)
(5.7.12)
1 3
Z = 2 1 .
3 2
(5.7.13)
(5.7.14)
3p1 + p2 + 2p3 = 1.
(5.7.15)
and
Solve these equations to get
p1 = 1/7 + (1/7)p2
p3 = 2/7 (5/7)p2
(5.7.16)
All the values of p2 between 0 and 2/5 give valid pricing vectors.
The sum of the state prices takes on all values between 3/5 and 3/7 which are the
reciprocals of R and R for this Z.
5.7.4. The single price law of markets. Two investments with the same payoff in every state must have the same current value. That is, if n1 and n2 are two
vectors of asset holdings and Y n1 = Y n2 , then v | n1 = v | n2 or, equivalently,
Y m = 0 v | m = 0.
(5.7.17)
64
65
total 7000, but the shares are, according to the market price, really worth 6900,
so Investor I makes a profit of 100 20 = 80. Investor II, by contrast, is the writer
of the contract and is obliged to buy made a loss of 100 20 = 80 in this situation.
5.8.4. Example. Forward price of a security without income or cost. Consider the
following two portfolios:
Portfolio A: At t = 0, enter a forward contract to buy one unit of an asset S,
with forward price K maturing at time T . Simultaneously invest an amount
KeT in the risk-free investment the price of the portfolio is KeT .
Portfolio B: At t = 0, buy one unit of the asset, at the current price S(0) the
price of the portfolio is S(0).
At time t = T , the price of the portfolio A is: receiving K from the risk-free
investment, pay K to buy one unit of Sthe payout of the portfolio is one unit of
S share.
The price of the portfolio B is one unit of S share. If we follow the principle of no
arbitrage, the two portfolio has the same payout, hence the initial price is the same,
KeT = S(0), which implies K = S(0)eT . This gives us a guide on the price of
the value of K.
5.8.5. Exercise. A three-year forward contract exists in a zero-coupon corporate
bond with a current price per 100 nominal of 50. The yield available on threeyear government securities is 6% p.a. effective. Calculate the forward price.
In the above problem, what if the three-year forward contract is replaced by sixmonth forward contract ?
(Note: a zero-coupon bond is a security that people purchase at time t = 0 for a
specified lump sum at some specified future date.)
5.8.6. Example (Forward price of a security with fixed cash income). Consider
the following problem: at a time t1 where 0 < t1 < T , the security underlying the
forward contract provides a fixed amount c to the holder.
Portfolio A: Enter a forward contract to buy one unit of S, with forward price
K, maturing at T , simultaneously invest an amount KeT + ce(t1) in the
risk-free investmentthe price at t = 0 is hence KeT + ce(t1) the payout
at t = T is hence K + ce(T t1) K + one unit of asset with price
S(T ) = ce(T t1) + one unit of asset with price S(T ).
(5.8.1)
Portfolio B: Buy one unit of the asset, at the current price S(0). At time t1,
invest c in the risk-free investment the price at t = 0 is hence S(0) the
payout at t = T is hence ce(T t1) + one unit of asset with price S(T )
Since yield from A and B are the same, no arbitrage assumption gives
KeT + ce(t1) = S(0),
(5.8.2)
(5.8.3)
and thus
5.8.7. Example (forward price for a security with known dividend yield). The
dividend yield for an equity is defined to be:
Dividend per share
Dividend yield =
.
(5.8.4)
Price per share
Let D be the known dividend yield per annum. Assume that dividends are received
continuously, and are immediately reinvested in the security of S. (Note: The actual
amount of dividend varies with the price).
66
Start with one unit of S at time t = 0, the accumulated holding at time T would be
eDT units of the security. Consider the following two portfolios:
Portfolio A: Enter a forward contract to buy one unit of S, with forward price
K, maturing at time T ; simultaneously invest an amount KeT in the risk-free
investment - At t = 0, the price of Portfolio A is KeT .
Portfolio B: Buy eDT units of S, at the current price S(0). Reinvest dividend
income in S immediately after it is received At t = 0, the price of Portfolio
B is S(0)eDT .
At t = T , the net portfolio A is: 1 unit of S.
At t = T , the net portfolio B is: 1 unit of S.
Using the no-arbitrage assumption the prices must also be the same, that is,
K eT = S(0)eDT
(5.8.5)
thus
K = S(0) e(D)T .
(5.8.6)
The following is a concrete numerical case: Assume that the dividend yield of a
portfolio of shares with current price of 1 000 is 4% p.a. Calculate the forward
price of a one-year forward contract, based on the portfolio, assuming dividends are
received continuously and the risk-free rate of interest is 4.602 8% p.a. effective.
4.602 8% p.a. effective is equivalent to a force of interest = 4.5%. Therefore
K = 1000 e0.0450.04 = 1005.012521.
(5.8.7)
5.8.8. Exercise. The current share price of a stock is 100. Dividends are paid
continuously and the current dividend is 2 per annum. Calculate the forward price
of a five-year contract on the asset if the risk-free force of interest is 5% per annum
and the dividend yield remains constant.
5.8.9. Exercise. Deduce a formula for the forward price, K, for an equity forward
contract in T years time (T is an integer). Assume a constant dividend yield D,
and that dividends are received in the middle of each year and are immediately
reinvested.
5.8.10. Definition (portfolio hedging). Hedging is a general term which describes
the use of financial instruments (including stocks, bonds, forward contracts and
more complex financial contracts such as options) to reduce or eliminate a future
risk of loss.
5.8.11. Exercise (hedging). In the above example, is it possible to calculate K if
the dividend is fixed amount of cash provided
1. The dividend is reinvested in shares of S.
2. The dividend is reinvested in risk-free investment.
5.8.12. Example. An investor agrees to sell an asset with no income in 2 years
time. The current price of the asset is 100 and the risk-free force of interest is 10%.
We assume that the investor borrows and invests money at the risk-free rate.
The forward price must be 100e2x0.1 = 122.1402758. This is the amount that the
investor will receive at time 2 for selling the asset.
In order the sell the asset, the investor first needs to buy the asset.
Suppose the investor chooses to buy the asset now and so the accumulated profit at
time 2 is 0.
If the price of the asset at time 2 turns out to be 110, then the investor would have
made a profit if she had waited and purchased the asset at time 2. The accumulated
profit would have been 122.1402758 110.
67
If the price of the asset at time 2 turns out to be 130, then the investor would have
made a lose if she had waited and purchased the asset at time 2. The accumulated
lose would have been 130 122.1402758.
This is called a static hedge since the hedge portfolio, which consists of the asset
to be sold plus the borrowed risk-free investment, does not change over the term of
the contract. For more complex financial instruments, such as options, the hedge
portfolio is more complex, and requires (in principle) continuous rebalancing to
maintain. This is called a dynamic hedge.
5.8.13. Exercise. Assume that there is no interest or dividend income and consider a forward contract agreed at time t = 0, with a forward price K, for buying
one unit of S. The maturity date of the contract is time T . Find the value of the
contract at an arbitrary intermediate time t. Construct a numerical example to
illustrate the point.
Hint. These values can be found using the no arbitrage assumption and similar
techniques to those used to find the forward price.
Exercises and problems on Arbitrage and portfolios
Exercise 5.1 (short selling). Find the definition of short sale and some application
of such activity.
Exercise 5.2 (can you spot arbitrage?). For each of the following, state whether
or not there is an arbitrage opportunity:
(a) An individual needs change and is offering you a 10 pound note in return for 9
pounds change.
(b) England football team is playing Croatia tonight, a friend says that he will pay
you 20 pounds if England loses and if you support Croatia during the game.
(c) Mr A offers to purchase a car for 12000 pounds, Mr B offers to buy the same
car for 15000 pounds.
(d) You and your friend are betting on the weather tomorrow. If it does not rain,
your friend will pay you 2 pounds, if it rains, you pay your friend 2 pounds.
(no, gamble 6= arbitrage)
Exercise 5.3. Are any of the assets in the economy
1 1 2
Z=
1 2 3
(5.3.1)
redundant?
Z
Problem 5.4. What can be said of the matrix | if one of the assets in Z is
1
redundant?
Problem 5.5. Show that if there exists a positive investment portfolio, say w > 0,
with semi-positive return, that is, Zw 0, then an arbitrage opportunity of the
second type guarantees the existence of an arbitrage opportunity of the first type.
Exercise 5.6 (arbitrage detection via pricing). Derive pricing equations for the
economy
5 2
Z = 3 1
(5.6.1)
2 1
and find out if there are any arbitrage opportunities of first or second type.
68
Exercise 5.7 (arbitrage detection via pricing). Find out whether there are arbitrage opportunities (first or second type) in an economy Z such that its transposed
matrix is
2 1 3
|
Z =
.
(5.7.1)
1 1 3
Exercise 5.8. A three-year forward contract exists in a zero-coupon corporate
bond with a current price per 100 nominal of 50. The yield available on threeyear government securities is 6% pa effective. Calculate the forward price.
In the above problem, what if the three-year forward contract is replaced by sixmonth forward contract?
(Note: a zero-coupon bond is a security that people purchase at time t = 0 for a
specified lump sum at some specified future date.)
Exercise 5.9. The current share price of a stock is 100. Dividends are paid
continuously and the current dividend is 2 per annum. Calculate the forward
price of a five-year contract on the asset if the risk-free force of interest is 5% per
annum and the dividend yield remains constant.
Exercise 5.10. Deduce a formula for the forward price, K, for an equity forward
contract in T years time (T is an integer). Assume a constant dividend yield D,
and that dividends are received in the middle of each year and are immediately
reinvested.
Exercise 5.11 (hedging). In the above example, is it possible to calculate K if
the dividend is fixed amount of cash provided
1. The dividend is reinvested in shares of S.
2. The dividend is reinvested in risk-free investment.
Exercise 5.12. Assume that there is no interest or dividend income and consider
a forward contract agreed at time t = 0, with a forward price K, for buying one unit
of S. The maturity date of the contract is time T . Find the value of the contract at
an arbitrary intermediate time t. Construct a numerical example to illustrate the
point.
Hint. These values can be found using the no arbitrage assumption and similar
techniques to those used to find the forward price.
APPENDIX A
A review of optimisation
1.1. Homotheticity and homogeneity
1.1.1. Definition (homogeneous functions). A function F : Rd R is said to be
homogeneous of degree k, for some given k R, at the point x0 if for all 6= 0 we
have
F ((x x0 )) = k F (x x0 ).
(1.1.1)
(1.1.2)
Hint. Differentiate both sides of (1.1.1) and use the chain rule.
This is a consequence of the following more precise statement.
1.1.3. Theorem (Euler). If F is homogeneous of degree k, then
x Rd ,
x| F (x) = kF (x),
(1.1.3)
where
|
x F (x) =
d
X
i=1
xi
F
(x).
xi
(1.1.4)
(1.2.1)
70
(resp. 0) .
(1.2.3)
(1.2.5)
and calculate the critical points of L, i.e., as those pairs (x , ) that solve the
following constrained EulerLagrange equations (also known as first-order optimality
conditions)
L(x, ) = 0,
(1.2.6)
which can be more explicitly written as a (usually coupled) system of (usually nonlinear) equations:
f (x) g(x) = 0,
g(x) = a.
(1.2.7)
(Note that the gradient of g, g is the transpose of the Jacobian matrix Dg.)
The following second-order optimality condition, makes the first-order optimality
condition sufficient: let x be a critical point of f constrained to g, define the
bordered Hessian at x as being the matrix given by
2
D f (x ) g(x )
H :=
.
(1.2.8)
Dg(x )
0
If H is positive definite, then the critical point x is a local minimum of f constrained to g = a.
1.2.3. Exercise (constrained maximisation). Adapt 1.2.2 for a maximisation
problem.
1.2.4. Exercise (simplifying the constraints constant). Explain the sentence without loss of generality we may consider a = 0 in 1.2.2.
1.2.5. Exercise (gradient calculus). Recall that the gradient of a field g is the
transpose of its Jacobian (i.e., the derivative) and show that for constant with
respect to x we have
x [| g(x)] = g(x).
(1.2.9)
1The
H11 . . . H1d
2
(1.2.2)
i = 1, . . . , m.
(1.2.11)
Solution (KarushKuhnTucker (KKT) method). Introduce the Lagrange multiplier , the KarushKuhnTucker multiplier and the KKT function
M (x, , ) := f (x) + | g(x) + | h(x).
(1.2.12)
(1.2.13)
(1.2.14)
(1.2.15)
with active j-th constraint if j > 0 and inactive j-th constraint if j = 0 for
j = 1, . . . , m;
(iv) and the complementary slackness condition
| h(x ) = 0.
(1.2.16)
Note that the Lagrange multiplier method is a submethod of KKT. Indeed, the
Lagrange multiplier method is the case where h 0 and where (iii)(iv), as well as
the second equation in (1.2.14) are simply dropped.
Conditions (i)(iv) are only necessary. Second order conditions are possible, but
stating them here is beyond our scope. The interested reader might want to check
some textbook on the matter [NW99, e.g.].
1.2.7. Exercise (KKT for maximisation problems). Consider a maximisation problem max f (x) with the same constraints as in (1.2.10) and show that (i) (iv) are
necessary, but with the KKT function defined as
M (x, , ) := f (x) + | g(x) | h(x).
(1.2.17)
Hint. Rewrite the maximisation problem for f (x) as a minimisation problem for
f(x) = f (x) and apply KKT.
1.2.8. Example. Let f : Rd R be a generic function. The necessary first-order
optimality conditions for maximising f (x) subject to the constraint x x0 (which
can be writen as h(x) := x0 x 0), can be obtained by taking h(x) = x0 x
and g(x) 0. Hence we need to solve
f (x) + = 0,
(1.2.18)
x0 x 0, 0 and | (x0 x) = 0.
(1.2.19)
(1.2.20)
This is equivalent to
72
1.2.9. Exercise. Find the points of the form (x, y) that realize min(x2 + y 2 + xy)
and max(x2 + y 2 + xy), subject to x2 + y 2 8.
Solution. Let f (x, y) := x2 + y 2 + xy, then
2x + y
f (x, y) =
.
2y + x
(1.2.21)
Furthermore, we introduce the KKT functions (with + for the minimum and for
the maximum)
M (x, y, ) := f (x, y) (x2 + y 2 8),
(1.2.22)
and derive:
(i) the stationarity condition
0
2x + y
2x
2(1 )x + y
=
=
0
2y + x
2y
x + 2(1 )y
which happens to be linear in (and easily solved) for (x, y) as follows
(
(0, 0)
for 6= 1/2,
(x, y) =
(t, t), t R for = 1/2;
(1.2.23)
(1.2.24)
(1.2.25)
0,
(1.2.26)
, R and x, y Rd (1.3.1)
MTA (G5078) Autumn 2009 1.3. Linear programming (LP) and duality
73
Figure 1. Geometric representation of the LP problem in Example 1.3.2.
|P +
lues c
rget va
a
t
4
4
3
1
(0,4)
(0,6)
c
(3/2,3)
(3,0)
set P
feasible
(6,0)
or, more simply put, all the involved functionals (f ) and constraints (g and h) are
of the form
k(x) = Kx + k0 for x0 Rd ,
(1.3.2)
for some fixed matrix K and vector k0 .
We used LP in Chapter 5. It is widely used in Economics and Business Management.
In this section we define is the linear optimisation problem (abusively called also
LP) which the LP method can solve and then explain how the LP method works.
1.3.1. Problem (linear programming (LP)). Given vectors c Rd and b Rk , a
matrix A Rkd and a scalar R, find x Rd such that
c| x + = max {c| x + : Ax b} .
(1.3.3)
6
2
1
2
3
12
1
d = 2, k = 4, A =
1 0 , b = 0 , c = 1/2 , = 1.
0 1
0
An example is given in Figure 1.
1.3.3. Analysis of the LP Problem 1.3.1.
Without loss of generality we will
assume that = 0 and drop it. Let P := x Rd : Ax b be the feasible set,
which is a polyhedron in Rd . Finding the maximum above can be seen as shifting
the hyper-plane orthogonal to the vector c, as long as it contains points in P .
Suppose the maximum is finite, say its value is , and attained by the element
x P . Let J be the subset of [1 : k] for which the inequalities in Ax b are
actual equalities
[A]j x = bj j J
(1.3.4)
Here [M ]j indicates the j-th row of a matrix M . Geometric intuition implies that
J is not empty because a linear function cannot realise its maximum in an internal
point of P , unless it is constant. The set J is known in jargon as the set of active
constraints.
74
Elementary geometry implies that = c| x is a nonnegative-coefficient linear combination of [A]j x = bj , for j J , i.e., there exist j 0, for j J , such that
P
= J j bj . Posing j := 0 when j 6 J , we obtain a vector Rk such that
|
0 and c x = = b =
k
X
j [A]j x = | Ax .
(1.3.5)
j=1
(Note that although Ax need not equal b, in the above equation we do have an
equality.) It follows that
A | c x .
(1.3.6)
Say c = 1 a1 + . . . + k ak and = 1 1 + . . . + k k , with 1 , . . . , k 0.
Since 1 1 + . . . + k k = b, we also know that 0 and A = c from above. This
implies
max {cx|Ax b} = = 1 1 + . . . + k k = b min{yb|y 0, yA = c}.
The inequality
max {cx|Ax b} min {yb|y 0, yA = c}
is trivial. Hence the equality
max {cx|Ax b} = min {yb|y 0, yA = c}.
Second, lets introduce complementary slackness: Let A be a matrix, let b be a
column vector, and let c be a row vector. Consider the LP duality equation
max {cx : Ax b} = min {yb : y 0 and yA = c} .
(1.3.7)
Assume that both optima are finite, and let x0 and y0 be feasible solutions, then the
following are equivalent:
(i) x0 and y0 are optimum solutions in (1.3.7);
(ii) cx0 = y0 b
(iii) if a component of y0 is positive, the corresponding inequality in Ax b is
satisfied by x0 with equality, i.e. y0 (b Ax0 ) = 0.
Condition (iii) is called complementary slackness. The equivalence of (i) and (ii)
follows directly from the Duality Theorem. The equivalence of (ii) and (iii) follows
from the having that cx0 = y0 Ax0 y0 b and y0 Ax0 = y0 b are true if and only (iii)
is true.
The primal-dual iteration method for LP was established by Dantzig, Ford and
Fulkerson [DFF56]. The idea is as follows: starting with a dual feasible solution y,
the method searches for a primal feasible solution x satisfying
y(b Ax) = 0.
(1.3.8)
If such a solution has been found, x and y are optimum primal and dual solutions.
If no such primal solution, the method prescribes a modification of the dual feasible
solution y, after which we iterate the procedure (start anew).
Suppose we want to solve the LP problem (where c = (1 , 2 , . . . , n ) is a given
vector)
min{cx|x 0, Ax = b}
(1.3.9)
where A is a m n matrix, b is a m-vector and c is an Rn vector. The dual LP
problem is
max{yb|yA c}
(1.3.10)
MTA (G5078) Autumn 2009 1.3. Linear programming (LP) and duality
75
1.3.4. Primal-dual iteration. Suppose we have a feasible solution y0 for the dual
problem (1.3.10). Let A0 be the submatrix of A consisting of those columns aj of A
for which y0 aj = j . To find a feasible primal solution for which the complementary
slackness condition holds, solve the restricted linear program
min{|x0 , 0, A0 x0 + b = b} = max{yb|yA0 0; yb 1}
(1.3.11)
x00 ,
(1.3.12)
APPENDIX B
(2.1.1)
(2.1.2)
F 3 ;
A, B F A B S
F;
{Ai : i N} F i Ai F ;
A F r A F.
Such a collection goes by the somewhat unlucky name of -algebra, or -field. Less
frequently we talk about sum-algebra/field, or you may want to strike the middle
ground we could use sigmalgebra.1
1Part
78
P = 0 and P = 1;
A, B F and A B P A P B;
P (A
P A + P B P (A B);
S B) = P
P i=1 Ai
i=1 P Ai .
It follows that 0 P (A) 1 for any A F . Other interesting facts like the
Inclusion-Exclusion Principle follow. The concept of indepedent event is maybe the
deepest in Probability Theory and you should check some good reference on the
topic [JP03, e.g.].2
2.2. Random variables
Given a probability measure (, F , P ). A random variable (also known as random
number) is a function X,
X : R,
(2.2.1)
such that the counterimage of any open interval in R is an element of F . In symbols
we write this as
{ : a < X() < b} := {a < X < b} F ,
(2.2.2)
for all a, b R.
In practice, you need not worry whether a given function X defined on is a random
variable: you may safely assume it is.3
2.2.1. Expectation. Given a random variable, its expectation is its integral against
the probability measure P over :
Z
E[X] = E X :=
X()P d.
(2.2.3)
On non-discrete spaces a whole theory of integration (or measure) must be developped, which is beyond this brief summary of probability. We will assume that
we talk about a probability space we actually mean a triple (, F , P ), not only which
is just a set, devoid of structure, without F and P . Often, though, F and are understated and
all that is mentioned is the probability measure P . In fact, given P we may obtain F as Dom P
and as the union of the collection F , so mentioning the triple is redundant.
3Although functions that are not measurable are shown to exist, using the Axiom of Choice,
you need to work quite hard to find one in practical applications.
2When
79
(2.2.5)
and the expectation (in practice thought as the average price) is given by
Z
EX =
P d = 1.
(2.2.6)
The reason behind the last equality (beyond the intuitively clear fact that the
average of 1 always, must be 1) is that the integral, being the area under the
graph, can be, in this particular situation calculated as
Z
P d = 1 P = 1 1 = 1.
(2.2.7)
(2.2.8)
(2.2.9)
then
E X = E[c1A ] = c E 1A = c
Z
P d = cP A.
(2.2.10)
(2.2.11)
where A and B are two events in F , and a, b R is a simple function. The linear
combination of two random variables is a random variable. Thus, simple functions
are random variables and
E X = E[a1A + b1B ] = aP A + bP B.
(2.2.12)
finance (, F , P ) is supposed to model all of the universe, and how eventss occurence
would affect the uncertain prices in stock markets, so describing would be mad. Of course,
events such as the birth of a star, or the fall of a comet onto a distant black hole, wont affect
financial markets on Earth as much as other events, such as the closure of the straight of Hormuz,
or the fall of a comet on Earth. In practice, one is not interested in describing , but simply uses
it to calculate quantities of interest, e.g., the expectation of the price of a stock.
4In
80
(d) Suppose a stock is priced at 3 at time 0 and that its price at time 1 is given by
the random number
(
5 with probability 30%,
S=
(2.2.13)
2 with probability 70%.
This means that S is a random variable of the form a1A + b1B with a = 5, A =
{S = 5}, b = 2, b = {S = 2}. We can thus calculate the stocks expected price as
E S = 5 0.3 + 2 0.7 = 2.9 .
(2.2.14)
Random variables taking only 2 possible value, such as S in this example, are called
binomial.
2.2.3. Proposition (CauchyBunyakovskiiSchwarz inequality). If X and Y are
two random variables on a space (, F , P ) then
E[XY ]2 E X 2 E Y 2 .
(2.2.15)
Note that standard operator precedence in notation means E[XY ]2 = (E[XY ])2 in
the above result.
2.2.4. Variance, standard deviation and covariance. Suppose X is a random
variable on the probability space (, F , P ). Its expected value E X, when finite,
is also known as the average. C.F. Gau, in proving his celebrated Law of Large
Numbers realised the importance of the variance of X, which he defined to be
var X := E (X E X)2 = E[X 2 ] E[X]2 = E X 2 E[X]2 .
(2.2.16)
The second and third identities are a result of simple algebra and properties of E.
Many authors use the notation = X for E X and the variance is thus expressed
as
var X = E[X 2 ] 2 .
(2.2.17)
The standard deviation of the random variable X, often denoted by X (or just
if the only random variable being discussed is X), is the square root of its variance,
i.e.,
1/2
X := var X = E[X 2 ] 2
.
(2.2.18)
If Y is another random variable then the covariance of the pair (X, Y ) is defined as
cov[X, Y ] := E[(X E X)(Y E Y )] .
(2.2.19)
(2.3.1)
81
a < b R.
(2.3.3)
Also, we have
PX [a, b] = FX (b) FX (a) and PX (a, b) = FX (b) FX (a),
(2.3.4)
As we shall see next, somebut not all!probability measures on R can be represented using a combination of a density function and a pointwise mass distribution.
2.3.3. Lebesgues Decomposition and RadonNikodym Theorems. 5 Although the following results are valid for general measurable spaces, we concentrate
on the sum-algebra B of Borel/Lebesgue measurable sets in R.
Denote by L the Lebesgue Measure on R, and L( dx) simply by dx. The Lebesgue
Decomposition Theorem says that any measure, call it M , on R can be decomposed
into two measures: a regular part, M reg , and a singular part, M sng such that
M (A) = M reg (A) + M sng (A)
L(A) = 0 M
reg
(A) = 0 (M
reg
A B,
82
(2.3.8)
(2.3.9)
ha | gi sup g =: kgk .
R
(As an exercise you should check the above properties.)
For this reason, the duality pairing ha | gi is often denoted as an integral:
Z
a (x)g(x) dx := ha | gi = g(a).
(2.3.10)
R
We stress that the notation above is not an integral, but merely a notation that
looks like and often (but not always!) behaves like an integral. It is possible to
intergrate a against a continuous function g also on a subset A of R, by using As
indicator function. In particular we may define
(
Z
g(a), if a b,
a (x)g(x) dx := a | 1(,b] g =
(2.3.11)
0,
if a > b.
(,b]
While the first equality is a definition, the second equality, which follows from the
properties of a and 1(,b] , is left as an exercise to check. Again, we warn against
the abuse of the integral notation. For instance, you should convince yourself that
in general
Z
Z
a (x)g(x) dx 6=
(,a]
a (x)g(x) dx.
(2.3.12)
(,a)
83
where J is a countable index set (e.g., some subset of Z), {xj : j J} R and
f j > 0.8
In particular we have that
Z x
X
P {X x} = FX (x) =
fXreg (y) dy +
f j x j ,
(2.3.16)
xj x
for any x R. Using the abuse of notation, whereby we integrate the xj s, this
may also be written as
Z
FX (x) =
fX (y) dy.
(2.3.17)
(,x]
2.3.7. Remark (existence of density function not always possible). Not all random
variables X have a piecewise differentiable distribution function cdf X . An example
is given by the Inverted Devils Staircase, originating in Georg Cantors work, and
described in Example 2.3.9.
2.3.8. Example (distribution and density of a binomial). Consider the binomial
random variable given by the price of a stock
(
5 with probability 70%,
S=
(2.3.18)
2 with probability 30%.
The cumulative distribution of S is easily described as follows
for x < 2,
0
cdf S (x) = 0.3 for 2 x < 5,
1
for 5 x.
8Note
(2.3.19)
that in spite of its name the density may in general fail to be a classical function. Indeed,
it is a function if and only if its singular part is 0.
84
0
if 2 and 5 6 B,
0.3 if 2 B and 5 6 B,
(2.3.20)
lawS B =
0.7 if 2 6 B and 5 B,
1
if 2 and 5 B.
The density of S is not a classical function, it is the sum of two impulses, namely
pdf S = 0.32 + 0.75 .
(2.3.21)
In this case the regular part of S is 0, and the expression above consists of the
singular part.
2.3.9. Example (inverted Devils Staircase). 9 Consider to be the space of infinite
sequences of independent fair coin tosses, where each is a sequence of symbols
in {H, T } (head or tail), i.e.,
= ((1), (2), . . . , (k), . . .) and (k) = H or T
k N,
(2.3.22)
with the probability measure P induced by the elementary binomial toss probability p on {H, T } where p {H} = 1/2 for heads and p {T } = 1/2 for tails. For
example, the probability of the event where the first toss is a head, the 5th a tail,
and the 6th or 7th toss is a head, can be calculated as follows
P { : (1) = H and (5) = T and ((6) = H or (T ) = H)}
(2.3.23)
113
3
= p {H} p {T } P {(6) = H or (T ) = H} =
= .
224
16
The space (, P ) constructed here is a common model probability space found in
many standard textbooks [Bil95, JP03, e.g.].
Consider playing the following game now. A chocolate bar C0 , of length 2 metres
is divided into three pieces, A1 , B1 and C1 , of equal length. You toss the coin then
bar B1 goes to the Bank, bar C1 is saved for the next round and bar A1 goes to you
if the outcome is T , otherwise (if the toss produces H) A1 goes to the bank. The
next step is played with C1 divided into three pieces of equal length, A2 , B2 , C2 ,
you toss the coin, B2 to the bank, C2 kept for next toss, and A2 to you if T or to
Bank if H. Consider now the random variable X that gives the length of chocolate
obtained after infinitely many tosses.
To model this game, introduce the basic random variable V : {H, T } R, where
V (H) := 0, V (T ) := 1, and let X : R be the random variable defined as follows
X
V ((k))
X() := 2
.
(2.3.24)
k
3
k=1
For example we have
2
X(T, T, T, T . . .) = 1, X(T, H, H, H . . .) = ,
3
1
2
X(H, T, T, T . . .) = , and X(H, T, T, T . . .) = .
3
9
Proposition. Let K N. For each integer of the form
j := 2
K1
X
(2.3.25)
(2.3.26)
k=1
9You
may safely skip this example if you do not know Measure and Integration Theory.
85
we have that
P
j
X K
3
=
K
X
!
k 2k + dl/2e
k=1
1
.
2K
(2.3.27)
Proof The complete proof of this result is not too hard and left as an exercise.
As an example take K = 3, then we may produce the following schedule
(0, 0)
(0, 1)
(1, 0)
(1, 1)
(1 , 2 )
l 0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3
j 0 1 2 3 6 7 8 9 18 19 20 21 24 25 26 27
P {X j/33 } 0 18 18 14 14 83 38 12 12 58 58 34 34 87 78 1
(2.3.28)
It follows that the cumulative distribution function, FX , of X is the so-called Devils
Staircase, which is a well-known fractal structure. The function FX is continuous (it
has no jumps), it is differentiable almost everywhere, with derivative 0, but it has
infinitely many points where it is differentiable only on one side. So its generalised
derivative is not a function and neither a countable sum of mass points, but a diffuse
singular measure concentrated on a set, denoted [0, 1), and known as the Cantor
set.
An approximation of the Devils Staircase can be plotted as follows
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
The Cantor set can be shown to consist of all those real numbers between 0 and 1
with a base-3 expansion that has no 1s (for those numbers that admit two different
expansions such as 1/3 = 0.1 = 0.0222 . . . , it is enough to have one expansion
with no 1s). A diagonal argument shows then that the Cantor set is uncountable.
Furthermore, it is possible to show that is closed and has Lebesgue measure 0.
Thus the distribution measure PX of X, defined by the elementary blocks
PX (a, b] := P {a < X b} = FX (b) FX (a),
(2.3.29)
is a singular measure with respect to the Lebesgue measure and the RadonNikodym
Theorem does not apply.
86
Using the terminology introduced in this section, if fX denotes the generalised density function of X, then the decomposition fX = fXreg + fXsng satisfies
Z
reg
sng
(x) dFX (x),
(2.3.30)
fX = 0 and hfX | i =
where the last integral is the RiemannStieltjes integral. The fact that fXreg = 0
means that the only contribution to the generalised probability density function in
this case comes from its singular part. For more details, you could see the book by
Stroock [Str99].
2.3.10. Expectation rule. Laws, c.d.f. and p.d.f. are useful tools. One the main
purposes for them is to get rid of . This means, for example, that by using the
law of a random variable on , integration on can be replaced by integration on
R (this is some kind of change-of-variable formula). This is an advantage, especially
in practical applications of probability, because generally the space is very hard
to model and manipulate, whereas R is (as you know) quite an easy place to work
in.
Proposition (expectation rule). If X is a real-valued random variable, with PX =
lawX and FX = cdf X . Let g be some function whose domain includes the set X().
Then g(X) is a random variable too and
Z
Z
E[g(X)] = g(x)PX ( dx) = g(x) dFX (x),
(2.3.31)
R
R
where the last integral is understood as a Riemann-Stieltjes integral.
2.3.11. Densitymass only distributions. Most of what we describe next works
for general random variables. Nonetheless, unless otherwise stated, we will consider
from now on only random variables which have a generalised probability distribution function consisting of the sum of a density function and a (possibly infinite)
countable sum of modulated Dirac masses. That is, given a random variable X we
assume that
X
fX (= pdf X ) = f reg +
f i xi
(2.3.32)
iJ
for some function f reg and a point mass distribution fi R+ for all i J, a finite
or countably infinite set. In particular, if is a smooth function, then we have
Z
Z
X
(x) dFX (x) = (x)f reg (x) dx +
f i (xi ).
(2.3.33)
iJ
(2.3.34)
R
R
iJ
provided the integral exists. If a = , the mean of X, then the moments are called
central moments and are conventionally written without the prime. For any other
a, the moments are non-central moments. The most common alternative is a = 0.
2.4.2. Characteristic function. Let X be a random variable and let F := cdf X
and (possibly) f = pdf X . The characteristic function of X, denoted by chf X , X , or
simply , whenever X is clear from the context, is the Fourier transform of dF ,
namely,
Z
Z
i tx
(t) := E[exp(i tX)] = e dF (x) = (possibly) ei tx f (x) dx, for t R. (2.4.4)
R
R
2
Here i is the unit imaginary number such that i = 1 (or i = 1 if you prefer).
The characteristic funtion and the n-th moment of f (x) about the origin are related
by the relation,
0n (0) = in (n) (0).
(2.4.5)
In the sequel we shall give some examples of characteristic functions and illustrate
their use to find moments. For a full development of the subject and more examples,
however, you should consult a standard textbook [JP03, Ch.13].
2.4.3. Exercise. Verify the characteristicmoment relation (2.4.5).
Hint. By the Lebesgue Dominated Convergence Theorem, you may take the
under the integral.
d
dt
sign
(2.4.6)
(2.4.7)
(2.4.8)
you dont know what P -almost all means just think all.
(2.4.9)
88
2.4.5. Example (characterstic function of a binomial). Consider the binomial random variable S : R such that
(
a with chance p,
S=
(2.4.10)
b with chance 1 p,
for some p (0, 1).
As seen in Example 2.3.8 we have
pdf S = pa + (1 p)b .
(2.4.11)
(2.4.13)
(2.4.14)
so E S = ap + b(1 p), as expected. The second derivative yields the second moment
d d
0
2 (0) =
(t)
= pa2 + (1 p)b2 ,
(2.4.15)
dt dt
t=0
and thus, with = E S, the variance can be computed as follows:
var S = E S 2 (E S)2 = 00 (0) 2
= pa2 + (1 p)b2 (ap + b(1 p))2 = p(1 p) a2 + b2 2p(1 p)ab
= p(1 p)(b a)2 .
(2.4.16)
For instance, if we had a = 0 and b = 1 we would have the standard deviation
p
(2.4.17)
S := p(1 p).
2.4.6. Characterisation Theorem. The characteristic functions of a distribution are always defined (because probability measures sum to 1 and the RiemannLebesgue Lemma guarantees convergence of the integral in (2.4.4)) and uniquely determine the distribution, in the following sense, suppose X1 and X2 are two random
variables, and let i and Fi indicate their characteristic and cumulative distribution
functions, respectively, for i = 1, 2. Then, 1 = 2 if and only if F1 = F2 , i.e., X1
and X2 are identically distributed. (Note that this does not mean that X1 and X2
are equal, nor equal P -almost everywhere.)
For a proof of uniqueness we refer to Jacod & Protter [JP03, Ch.14].
2.4.7. Moment generating function and Laplace transform. The characteristic function of a random variable is the Fourier transform of its probability
distribution measure. Similarly, one may consider the Laplace transform,
Z
Z
rx
L(r) = e
dF (x) = (possibly) erx f (x) dx,
(2.4.18)
R
R
provided the integral makes sense, which is the case when the random variable X
(for which F = cdf X and f = pdf X ) is positive.
t > 0.
(2.4.21)
(2.5.4)
2.5.1. Definition (standard normal deviates). If X is a normally distributed random variable with mean = 0 and variance 2 = 1, it is said to be a standard normal
deviate. The density and distribution functions of a standard normal deviate are
often denoted by n(x) and N (x).
2.6. Limit Theorems
2.6.1. Weak law of large numbers. Let (Xi )iN be a sequence of independent
random variables11 on a probability space (, F , P ), which are identically distributed, i.e., there is one distribution F such that cdf Xi = F for all i N. Denote
by the, then for any constant > 0,
)
( n
1 X
Xi < = 1.
lim P
n
n
i=1
11Independence
is the single most important probability concept, but it is not covered in this
appendix. For a proper discussion you should consult one of the recommended texts.
90
"
n
n 1X
lim P a <
(2.7.1)
xi < b = N (b) N (a)
n
n i=1
2.8. Examples of random variables
2.8.1. Bivariate Normal Variables. Let x1 andx2 be bivariate normal random
variables with means i variance i2 and covariance 12 . The weighted sum w1 x1 +
w2 x2 is normal with mean and variance
= w1 1 + w2 2 , 2 = w12 i2 + 2w1 w2 12 + w22 22 ,
For such x1 andx2 and for a differentiable function h(x), cov[x1 , h(x2 )] = E[h0 (x2 )]12
This property can be proved as follows. If x1 andx2 are bivariate normals, then from
our understanding of regression relationships we may write
x1 = a + bx2 + e,
where b = 12 /22 and e is independent of x2 . Therefore,
cov[x1 , h(x2 )] = cov[a + bx2 + e, h(x2 )]
= b cov[x2 , h(x2 )]
= bE[(x2 2 )h(x2 )]
Z
(2.8.1)
=b
(2.8.2)
b22
h(x2 ) df (x2 )
b22
(2.8.3)
91
upon integrating by parts. Then if h(x2 ) = o(exp(22 x2 )), the first term vanishes at
both limits, and the remaining term is just E[h0 (x2 )]12 .
2.8.2. Lognormal variables. If x is normally distributed, then z = ex is said to
be lognormal. The lognormal density function is
(ln z )2
f (z) = ( 2z)1 exp(
)
2 2
(2.8.4)
n2 2
n = exp n +
2
2
z = exp +
2
var(z) = exp 2 + 2 exp 2 1
(2.8.5)
ln a
2
+ .
zf (z) dz = exp( + )N
2
(2.8.6)
x D,
(2.8.7)
92
Then
Z
E[G(X)] =
G(x)f (x) dx
Z
Z
0
(x) dx
= G(X)
f (x)dx + G (X)
(x X)f
Z
1
2 f (x)dx
+
G00 (x )(x X)
2
Z
1
2 f (x)dx
= G(X) +
G00 (x )(x X)
2
< G(X),
(2.9.1)
Here the n are independent and identically distributed. The parameter , which
is the expected change per period, is called the drift. Another stochastic process is
the autoregressive process
Xn = (1 a) + aXn1 + n
(2.9.2)
where again the error terms are independent and identically distributed.
2.9.2. Definition (Markov process). A Markov process is a stochastic process for
which everything that we know about its future is summarized by its current state.
In terms of the distributions of Xn we have
fni (Xn ; x0 , . . . , xi ) = fni (Xn ; xi )
(2.9.3)
(2.9.6)
E[Xn+1 |Y0 , . . . , Yn ] = xn .
(2.9.7)
and
The martingales property can also be stated as Xn+1 Xn is a fair game with respect
to {Yi }.
2.9.4. Proposition. A time series of updated conditional expectations is always
a martingale in this generalized sense (provided that the expectations are finite).
That is, let X be a random variable, let {Yi } be any stochastic process. Then
xn = E[X|y0 , . . . , yn ] are the realizations of a martingale.
93
(2.9.8)
But Xn+1 is a function of y0 , . . . , yn+1 so the inner expectation is just the realization
xn+1 which is also, by definition
E[X|y0 , . . . , yn+1 ],
therefore
E[E[Xn+1
= E[X|y0 , . . . , yn ] = xn
This is the martingale property.
(2.9.9)
(2.9.10)
|y0 , . . . , yn ] = xn
(2.9.11)
|Y0 , . . . , Yn ]] = E[Xn ]
(2.9.12)
(2.10.2)
and minimize S(a, b) with respect to a and b. Using the usual rules of finding
minimum, we get
s(x, y)
a
= y b
x and b =
,
(2.10.3)
S(x, x)
where
n
n
X
1X
1X
S(x, y) =
(
x xi )(
y yi ), x =
xi and y =
yi
(2.10.4)
n
n
i
i=1
i=1
94
The random variables are assumed to satisfy var(ei ) = 2 , after substituting the
expressions of a
and b into (2.10.2), we obtain
S(a, b) = S(y, y)
S(x, y)2
S(x, x)
(2.10.5)
(2.10.6)
(2.10.7)
where e is the error term, assumed to be N (0, 2 ) distributed. Again, we fix the
parameters a, b and c by minimising the squares sum function
S(a, b) = (y1 a bx11 cx12 )2 + (y2 a bx21 cx22 )2 + . . . + (yn a bxn1 cxn2 )2
(2.10.8)
APPENDIX C
96
possibility of short-term changes in the set of properties sold from month to month
(for example, shifts in the regional complexion of the market or a change towards
more large properties being sold) giving a misleading impression of the change in
the price of a typical house.
Analyses of house prices are based on simple arithmetic average prices.
3.2.3. Data. The Halifax House Price Indices are derived from information on the
following house characteristics:
? Purchase price.
? Location (region).
? Type of property: house, sub-classified according to whether detached, semidetached or terraced, bungalow, flat.
? Age of the property.
? Tenure: freehold, leasehold, feudal.
? Number of rooms: habitable rooms, bedrooms, living-rooms, bathrooms.
? Number of separate toilets.
? Central heating: none, full, partial.
? Number of garages and garage spaces.
? Garden.
? Land area if greater than one acre.
? Road charge liability.
Although one hundred per cent coverage of all house purchase transactions financed
by the Halifax is obtained, those transactions that do not constitute a fully consistent
body of data for the purpose of house price analysis are excluded from the Indices.
These exclusions primarily cover property sales that are not for private occupation
and those that are likely to have been sold at prices which may not represent free
or normal market prices, for example, most council house sales, sales to sitting
tenants, etc. Only mortgages to finance house purchase are included; remortgages
and further advances are excluded.
The data refers to mortgage transactions at the time they are approved rather
than when they are completed. Whilst this may cover some cases which may never
proceed to completion, it has the important advantage that the price information
is more up-to-date as an indicator of price movements and is on a more consistent
time-base than completions data (such as the ODPM Index) given the variable time
lags between approval and completion.
The monthly indices cover transactions during the full calendar month and the
regional quarterly indices cover transactions over the entire quarter. Properties over
1 million have been included since December 2002 to reflect the increasing number
of this hitherto tiny market segment.
3.2.4. Seasonality. House prices are seasonal with prices varying during the course
of the year irrespective of the underlying trend in price movements. For example,
prices tend to be higher in the spring and summer months when more people are
looking to buy. We therefore produce seasonally adjusted series to remove this
effect and to allow us to concentrate on the underlying trend in house prices. These
seasonal factors are updated monthly.
The Halifax House Price Index is prepared from information that we believe is
collated with care, but we do not make any statement as to its accuracy or completeness. We reserve the right to vary our methodology and to edit or discontinue
the indices at any time for regulatory or other reasons. Persons seeking to place
97
reliance on the indices for their own or third party commercial purposes do so at
their own risk.
Full Technical Details are available free on request - contact the HBOS Economics
Help-line on 01422 333558.
3.3. Nationwide House price Index
There are several methods that could be used to calculate the trend in house prices,
ranging from a simple average of purchase price to a statistical method of averaging.
Then there is the matter of making sure that the different mixture of properties
sold in each month does not give a false impression of the actual change in house
prices. The next few sections explain the way we do this as well as providing some
background to the Nationwide house price series and the current methodology that
we employ to calculate average house prices.
3.3.1. Background to Nationwide House Price Information. Nationwide
Building Society has a long history of recording and analysing house price data and
has published average house price information since 1952. The following provides a
short chronology of publish series and developments in Nationwides methodology
of calculating average house prices:
1952: annual publication of house price data;
1974: quarterly data is published for the first time;
1989: development of new house price methodology (a statistical regression
technique was introduced under guidance of Fleming and Nellis Loughborough University and Cranfield Institute of Technology);
1993: the house price system was further improved following publication of the
Census 1991 data, and frequency for UK series increased to monthly.
The monthly figure measures the mix adjusted average house price for all houses in
the UK. Every quarter the Nationwide also publishes a more detailed breakdown of
house prices. These include both UK and 13 regional estimates for:
5 types of house: detached, semi-detached, terraced, flats and bungalows;
2 types of buyer: first time buyer and former owner occupiers;
3 property ages: new, modern and older.
This makes a total of 154 separate series, all of which are published in the Nationwides Quarterly Review.
3.3.2. Data source. All house price information is derived using Nationwide mortgage data. This data is extracted monthly for mortgages that are at the approvals
stage and after the corresponding building survey has been completed.
3.3.3. Data processing (also known as cleaning). Nationwide house price series utilise only owner occupied property information. In addition, properties that
are not typical and may distort the series are also removed from the data set. Therefore, the following criteria is used to select which properties to include:
? house purchasesremortgages and further advances are excluded;
? owner occupied properties;
? purchase prices below 1 million;
? properties sold at true market pricesright to buy sales at discounted price are
excluded;
? floor size has to be within specified limits for a give type of property, e.g., a
detached house has to have at least 400 square foot [sic] floor area.
98
3.3.4. Sample Size. The number of cases that are used to calculate the average
price for a given month will depend on the volume of monthly mortgage activity and
out of these the cases that meet the criteria in the cleaning process. The monthly
sample size will therefore vary from month to month. Nationwide has sufficient
sample size to produce a representative house price series. N.B. Net lending figures
quoted at our half yearly and annual results are not a guide to our sample size.
Sample size is based on the number of new loans we write i.e. the amount of gross
lending for house purchase(remortgage cases are excluded).
The Nationwide Building Society is the 5th largest mortgage lender in the UK. Our
share of the gross house purchase market has averaged 9% over the last 3 years
(reaching 11% during our financial year 2000/01. This allows us to be confident
that the series based on Nationwide mortgage data is representative of the whole
house market.
The quarterly UK series for all houses uses 3 months of data and hence a much
larger sample than at the month. The samples sizes for the other quarterly series
will depend on what it is they are measuring, for example the series for first time
buyers only considers properties being brought by first time buyers and hence this
will have a smaller sample size than that used for the whole of the UK. It is for this
reason that detailed breakdown of house prices are produced quarterly.
3.3.5. Mix Adjustment Process. The purpose of mix adjustment is to simply
isolate pure prices changes. The simple example below illustrates how the changes in
the mixture of properties sold each month could give a misleading picture of what is
actually happening to house prices. The set of properties sold from month to month
will vary by location and design etc. and some adjustment is necessary to make
sure all of these do not give a false impression of the actual changes to house prices.
A mix-adjusted or standardised index is not affected by such changes because the
relative weight given to each characteristic of a property in the mix (or basket,
to use an analogy with retail prices) is fixed from one period to the next.
99
The simple average of both kinds of properties will be influenced by the proportion
of each property sold. In periods 3 and 4 the simple average shows a decrease,
whereas the actual prices of both increased!
The mix adjusted average uses a consistent measure of the proportion of each type
of property and is able to better reflect the true change in prices.
Time Period P1 P2 P3 P4 P5
% Flats
50 30 70 30 70
% Detached
50 70 30 70 30
The mix-adjusted price represents the price for an average or typical house. This
should not be confused with the average price of all houses. The latter is usually
higher because even though there are fewer more expensive houses sold, their price
is such that they bias the simple average to be greater than the price of the typical
house.
3.4. Calculating the price of a typical house
The price of a property will depend on the characteristics of the property. These
characteristics could include physical properties of the house, like its design, but
other aspects such as the type of neighbourhood the house is located in will also
contribute to the price someone is willing to pay. Using mortgage data, the Nationwide house price system can relate all the observed combinations of these factors and
relate them to the price of which the house was sold for. From this, the model can
estimate how much on average a house would cost given a set values for these characteristics, in particular a set of characteristics that describes the typical house.
This typical house does not physically exist, it is an average house across all the
characteristics that the model uses. This method is repeated on data sets at different points in time and changes in the price of this typical house reflect only the
price changes over the same time periods, and not the mixture of properties sold in
the current or pervious periods.
3.4.1. Factors that affect the price of a house. The following are the items
that are used to describe the characteristics of a property. There is no set order that
these contribute most to the price of the house, although UK location, the type of
neighbourhood and house size are consistently the three most important followed
by the design of the house.
Geographic location: Type of neighbourhood. The Nationwide index uses an
established demographic system that classifies areas in the UK into 54 categories
based on the type of people that live there, two examples include retirement
and council areas.
floor size: Property design (detached house, semi-detached house, terraced house,
bungalow, flat, etc.)
tenure: freehold/leasehold/feudal, except for flats, which are nearly all leasehold;
number of bathrooms: 1 or more than 1;
type of central heating: full, part or none;
type of garage: single garage, double garage or none;
number of bedrooms: 1,2,3,4 or above 4;
new/old: property is either new or not.
100
3.4.2. Seasonal Adjustment. House prices are slightly seasonal - that is, prices
are higher at certain times of year irrespective of the overall trend. This tends to
be in spring and summer, when more buyers are in the market and hence sellers
do not need to discount prices so heavily, in order to achieve a sale. The effect on
prices over the year is of the order of 2%; however this is much smaller than the
change in volume of property transactions. The seasonal effect is estimated twice a
year using established statistical methods.
For the monthly house price index where changes can be as little as 0.1%, seasonal
factors are important. The Nationwide therefore produce a seasonally adjusted series
for UK house prices which seeks to remove this effect so that the overall trend in
prices is more readily apparent.
Seasonal adjustment shows that June is generally the strongest month for house
prices (raw prices are 1.3% above their SA level) and January is the weakest (raw
prices are 1.9% below their SA level).
3.5. Pooled Property Fund Indices
Details of this index can be found on http://www.ipdindex.co.uk
Published by HSBC and The Association of Property Unit Trusts
The Pooled Property Fund Indices are jointly published by the Association and
HSBC. The data is compiled and calculated by IPD.
The index is designed to bring greater transparency to the PUT and managed pension fund industry; provide fuller and more timely information to both existing and
potential investors (to a target of 10 working days); allow investors to compare
performance of individual vehicles with the rest of the Pooled Property sector.
3.5.1. Important Information.
1. Following a consultation exercise the decision has been made to change the calculation basis of the Pooled Property Fund Indices to a net asset value basis,
from an offer price basis.
2. The new Indices no longer contains an All Property Unit Trusts Index - this
has been replaced by an All Balanced Funds Index.
It splits the 29 participating Funds into three categories:
balanced funds: holding a wide mix of property assets,
specialist vehicles: focusing on a particular type of property,
managed property funds: managed mainly by insurance companies.
Their performance is measured by NAV to NAV unit returns. These are compiled
from quarterly unit price and distribution records, supplied to IPD by individual
fund managers.
One of the unique advantages of the new service is that it enables investors to judge
the unit price performance of a fund in the context of gearing percentages and
independently verified IPD information relating to portfolio asset allocation.
3.5.2. Problem. Discuss the differences in the three property related indices.
Bibliography
[Bil95]
Patrick Billingsley. Probability and measure. Wiley Series in Probability and Mathematical Statistics. John Wiley & Sons Inc., New York, third edition, 1995. ISBN 0-471-007102. xiv+593 pp. A Wiley-Interscience Publication.
[CZ03] Marek Capi
nski and Tomasz Zastawniak. Mathematics for finance. Springer Undergraduate Mathematics Series. Springer-Verlag London Ltd., London, 2003. ISBN 1-85233330-8. x+310 pp. An introduction to financial engineering.
[DFF56] G. B. Dantzig, L. R. Ford, Jr., and D. R. Fulkerson. A primal-dual algorithm for linear
programs. In Linear inequalities and related systems, Annals of Mathematics Studies, no.
38, pages 171181. Princeton University Press, Princeton, N. J., 1956.
[Eat]
John W. Eaton. GNU Octave. URL http://www.gnu.org/software/octave/.
[Fle77] Wendell Fleming. Functions of several variables. Springer-Verlag, New York, second edition, 1977. xi+411 pp. Undergraduate Texts in Mathematics.
[JP03]
Jean Jacod and Philip Protter. Probability essentials. Universitext. Springer-Verlag,
Berlin, second edition, 2003. ISBN 3-540-43871-8. x+254 pp.
[LL01] Elliott H. Lieb and Michael Loss. Analysis, volume 14 of Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, second edition, 2001. ISBN
0-8218-2783-9. xxii+346 pp.
[MCS86] J. J. Mc Cutcheon and W. F. Scott. An Introduction to the Mathematics of Finance.
ButterworthHeinemann, 1986. ISBN 0-7506-0092-6.
[NW99] Jorge Nocedal and Stephen J. Wright. Numerical optimization. Springer Series in Operations Research. Springer-Verlag, New York, 1999. ISBN 0-387-98793-2. xxii+636 pp.
URL http://dx.doi.org/10.1007/b98874.
[Str99] Daniel W. Stroock. A concise introduction to the theory of integration. Birkhauser Boston
Inc., Boston, MA, third edition, 1999. ISBN 0-8176-4073-8. xiv+253 pp.
Index
NAP, 48
net cash flow, 15
No-Arbitrage Principle, 48
one-step binomial, 46
perpetuity, 34
portfolio, 47, 53, 54
portfolio weights, 54
price index, 40
purchasing power, 40
random number, 45, 100
random variable, 45, 100
random vector, 53
realized return, 53
redeem
a bond, 45
redemption
bond, 45
return, 46
cash, 40
money, 40
on an asset, 46
real, 40
risk, 51
risk-free
asset, 45
risky
asset, 45
sample space, 53
security, 45
short on, 47
short position, 47
state, 53
stock, 45
time
discrete, 46