Академический Документы
Профессиональный Документы
Культура Документы
N. Instefjord
FN3023, 2790023
2011
Undergraduate study in
Economics, Management,
Finance and the Social Sciences
This subject guide is for a 300 course offered as part of the University of London
International Programmes in Economics, Management, Finance and the Social Sciences.
This is equivalent to Level 6 within the Framework for Higher Education Qualifications in
England, Wales and Northern Ireland (FHEQ).
For more information about the University of London International Programmes
undergraduate study in Economics, Management, Finance and the Social Sciences, see:
www.londoninternational.ac.uk
This guide was prepared for the University of London International Programmes by:
N. Instefjord, PhD, Associate Graduate Director, Essex Business School, University of Essex.
This is one of a series of subject guides published by the University. We regret that due to
pressure of work the author is unable to enter into any correspondence relating to, or arising
from, the guide. If you have any comments on this subject guide, favourable or unfavourable,
please use the form at the back of this guide.
Contents
Contents
23 Investment management
ii
Contents
23 Investment management
iv
Chapter 1: Introduction
Chapter 1: Introduction
Finance is essentially about pricing financial assets, but in this subject
guide we will focus more on what we use pricing theory for from an
investment perspective. We will seek to apply pricing theory (among other
things) to tell us something about how to invest our money optimally in
financial assets rather than for pricing itself. We will spend some time
looking at how to protect our investments using techniques from the area
of risk management. For those who want a more thorough overview of
pricing theory, see the subject guide for course 92 Corporate finance.
23 Investment management
Learning outcomes
At the end of this course, and having completed the Essential reading and
activities, you should be able to:
list given types of financial instruments and explain how they work in
detail
contrast key characteristics of given financial instruments
briefly recall important historical trends in the innovation of markets,
trading and financial instruments
name key facts related to the historical return and risk of bond and
equity markets
relate key facts of the managed fund industry
define market microstructure and evaluate its importance to investors
explain the fundamental drivers of diversification as an investment
strategy for investors
aptly define immunisation strategies and highlight their main
applications in detail.
discuss measures of portfolio risk-adjusted performance in detail and
critically analyse the key challenges in employing them
competently indentify established risk management techniques used by
individual investors and corporations.
Syllabus
Exclusions: This course has replaced course 121 International
financial markets and may not be taken if you are taking or have
passed course 121 International financial markets. If you have
failed course 121 and wish to transfer to course 23, your fail on course
121 will count as one of the three chances you have to pass the new
course.
Prerequisites: If you are taking this course as part of a BSc degree, you
must have already taken course 24 Principles of banking and
finance (or course 94 Principles of banking for students registered
before 1 September 2005). Course 92 Corporate finance must also be
taken with or before this course.
The syllabus comprises the following topics:
1. Financial markets and instruments: money and bond markets;
equity markets; derivative markets; managed funds; margin trading;
regulation of markets.
2. History of financial markets: historical and recent financial innovation;
historical equity and bond market returns; equity premium puzzle.
2
Chapter 1: Introduction
Reading advice
At the start of each chapter in this subject guide your recommended
reading appears in two categories, Essential reading and Further reading,
to be found in both textbooks and journal articles.
Essential reading
The course uses two essential textbooks as listed below:
Bodie, Z., A. Kane and A.J. Marcus Investments. (Boston, Mass.; London:
McGraw-Hill Irwin, 2008) eighth edition [ISBN 9780071278287].
Elton, E.J., M.J. Gruber, S.J. Brown and W.N. Goetzmann Modern Portfolio
Theory and Investment Analysis. (New York; Chichester: John Wiley & Sons,
2010) eighth edition [ISBN 9780470505847].
Detailed reading references in this subject guide refer to the editions of the
set textbooks listed above. New editions of one or more of these textbooks
may have been published by the time you study this course. You can use
a more recent edition of any of the books; use the detailed chapter and
section headings and the index to identify relevant readings. Also check
the VLE regularly for updated guidance on readings.
Further reading
Please note that as long as you read the Essential reading you are then free
to read around the subject area in any text, paper or online resource. You
will need to support your learning by reading as widely as possible and by
thinking about how these principles apply in the real world. To help you
read extensively, you have free access to the virtual learning environment
(VLE) and University of London Online Library (see below).
23 Investment management
Books
Allen, F. and D. Gale Financial Innovation and Risk Sharing. (Cambridge, Mass.;
London: MIT Press, 1994) [ISBN 9780262011419].
Campbell, J.Y. and L.M. Viceira Strategic Asset Allocation. (New York: Oxford
University Press, 2002) [ISBN 9780198296942] Chapter 2.
Duffe, D. and K.J. Singleton Credit Risk: Pricing, Measurement and Management.
(Princeton, NJ: Princeton University Press, 2003) [ISBN 9780691090467]
Chapter 1.
Embrechts, P., C. Kluppelberg, and T. Mikosch Modelling Extremal
Events. (New York; Berlin; Heidelberg: Springer-Verlag, 1997) [ISBN
9783540609315]. Note that this book is very advanced and is not really
drawn on except for some initial observations made in the very beginning
of Chapter 8 of the guide.
Grinblatt, M. and S. Titman Financial Markets and Corporate Strategy. (Boston,
Mass.; London: McGraw-Hill Irwin, 2008) [ISBN 9780077119027]. Please
note that at the time of going to print there is a new edition of this textbook
due to be published.
Hasbrouck, J. Empirical Market Microstructure. (Oxford: Oxford University
Press, 2007) [ISBN 9780195301649]. A relatively current textbook on
market microstructure which forms the basis for the chapter on market
microstructure.
Lo, Andrew W. Hedge Funds. (Princeton, NJ: Princeton University Press, 2008)
[ISBN 9780691132945].
Pole, Andrew Statistical Arbitrage. (Hoboken, NJ: Wiley Finance, 2007)
[ISBN 9780470138441].
MacKenzie, Donald An Address in Mayfair. (London Review of Books) www.lrb.
co.uk/v30/n23/mack01.html
Stultz, R. Risk Management and Derivatives. (Mason, Ohio: Thomson SouthWestern, 2003) [ISBN 9780538861014]. This book specifically deals with
risk management.
Journals
There are a number of important journal articles that deal with investment
management - those listed here are just a few:
Elton, E.J. and M.J Gruber Modern portfolio theory: 1950 to date, Journal of
Banking and Finance 21(1112) 1997, pp.174359.
Elton, E.J., M.J. Gruber and C.R. Blake Survivorship bias and mututal fund
performance, Review of Financial Studies 9(4) 1996, pp.1097120.
Mehra, R. and E.C. Prescott The Equity Premium: A Puzzle, Journal of
Monetary Economics 15(2) 1985, pp.14561.
Sharpe, W.F. Asset Allocation: Management Style and Performance
Measurement, Journal of Portfolio Management 30(10) 1992, pp.716.
Chapter 1: Introduction
The VLE
The VLE, which complements this subject guide, has been designed to
enhance your learning experience, providing additional support and a
sense of community. It forms an important part of your study experience
with the University of London and you should access it regularly.
The VLE provides a range of resources for EMFSS courses:
Self-testing activities: Doing these allows you to test your own
understanding of subject material.
Electronic study materials: The printed materials that you receive from
the University of London are available to download, including updated
reading lists and references.
Past examination papers and Examiners commentaries: These provide
advice on how each examination question might best be answered.
A student discussion forum: This is an open space for you to discuss
interests and experiences, seek support from your peers, work
collaboratively to solve problems and discuss subject material.
Videos: There are recorded academic introductions to the subject,
interviews and debates and, for some courses, audio-visual tutorials
and conclusions.
Recorded lectures: For some courses, where appropriate, the sessions
from previous years Study Weekends have been recorded and made
available.
Study skills: Expert advice on preparing for examinations and
developing your digital literacy skills.
Feedback forms.
Some of these resources are available for certain courses only, but we
are expanding our provision all the time and you should check the VLE
regularly for updates.
Examination structure
Important: the information and advice given here are based on the
examination structure used at the time this guide was written. Please
note that subject guides may be used for several years. Because of this
we strongly advise you to always check both the current Regulations for
relevant information about the examination, and the virtual learning
5
23 Investment management
Essential reading
Bodie, Z., A. Kane and A.J. Marcus Investments. (Boston, Mass.; London:
McGraw-Hill Irwin, 2008) Chapters 1, 2, 3, 4, 14, 20, 22 and 23.
Elton, E.J., M.J. Gruber, S.J. Brown and W.N. Goetzmann Modern Portfolio
Theory and Investment Analysis. (New York; Chichester: John Wiley & Sons,
2010) Chapters 2 and 3.
Further reading
Grinblatt, M. and S. Titman Financial Markets and Corporate Strategy. (Boston,
Mass.; London: McGraw-Hill Irwin, 2008) Chapters 1, 2 and 3.
Guide to readings
This chapter is an introductory chapter and contains a great deal of
background readings from both the Essential and the Further readings.
The Essential reading for the general material in this chapter is contained
in Bodie, Kane and Marcus Chapters 1 through to 4. Here you can read the
material relatively quickly as there are few technical details to remember.
Some of the more technical material in this chapter is covered in Bodie,
Kane and Marcus Chapters 14, 20, 22 and 23. Here you can be more
selective in your reading, but you may also have to read the material more
carefully so that you are sure you understand it properly.
Introduction
Financial assets are distinct from real assets in that they do not generate
a productive cash flow that is what real assets do. Examples of real
assets are: a block of flats that can be let to provide the owner with future
rental income; the rights to manufacture and sell a particular product
generating future sales revenue; or a particular piece of computer software
that generates future sales and registration income. Examples of financial
assets are: a loan that is used to fund the acquisition of the block of
7
23 Investment management
flats and whose payments are financed by the rental income; or equity
capital used to fund the research and development costs for the consumer
software products mentioned above. The equity holders benefit in terms of
future dividends or capital gains that are generated from the future sales
income of the products. From the point of view of cash flow generation,
therefore, financial assets do not have much of a role to play. Financial
assets are not neutral in the sense that they transform the cash flow of real
assets for the holder. For instance, the loan generates a relatively stable
income even though the underlying cash flow is risky. Also, the loan might
have enabled the investor to raise suffcient funds for investment in the
first place. We trade financial assets, therefore, to repackage or transform
the cash flow of real assets, either through time or across states of nature.
Financial assets also do another important job they enable us to separate
the functions of ownership and control of real assets. As a rule, real assets
do not just passively generate a cash flow they need to be managed.
A company owns a collection of real assets. The job of managing these
is highly specialised and it is necessary that it is done by a professional.
This individual may or may not be the owner of the real assets, so it
makes sense for the company to keep the ownership and control functions
separate. This can be done by issuing equity with claims on the real assets
of the company the owners of the companys equity then become the
owners of the company so that the company can hire a professional
manager to manage its pool of real assets.
Who uses financial markets? There are three key sectors:
The household sector you and me who need to invest for retirement
income or mortgages for house acquisitions and various insurance
products, for instance.
The business sector consisting of firms that need to issue financial
claims on their future cash flow to finance current investments which
need to manage the risk of their business through derivatives trading
and insurance products.
The government sector that has a need to finance public expenditure.
This sector is special as it sometimes also intervenes in financial
markets to provide a public policy objective for instance, influence
the interest rate to manage inflation and additionally by acting as a
regulator of the activity in financial markets.
On the other hand, financial markets are not the only way these sectors
are served by financial instruments. Financial intermediaries also provide
services. These are companies such as banks and investment houses which
can lend money to, and help companies issue securities, and collect
deposits or lend to households, or manage households and companies
funds. In this chapter, we shall discuss a relatively broad range of financial
assets (also known as financial instruments), and their key defining
characteristics.
The cash flow promised to bond holders comes from the cash flow
generated by real assets. Since the cash flow of real assets is often risky, it
may be that there is not enough to pay the promised amount at all times.
If this happens, the bond may default. In the example above, for instance,
the coupon promises a cash flow of 100 to be paid to the bond holders, but
if the corporate cash flow available in year 10, after coupon repayments
are made, is only 70 the bond holders stand little chance of receiving their
promised repayment of 100. The bond defaults, therefore, and the bond
holders can expect to receive at most only 70. Some bonds are, however,
practically default-free for instance, bonds issued by the government
(government bonds they are often called treasury bonds in the USA
and gilts in the UK). Bond instruments are traded in the money market
or the bond market. The distinction between these markets is essentially
that of the original maturity of the instrument. If a bond was issued with
very short maturity normally less than six months it will be traded in
the money market. If a bond has longer maturity it is traded in the bond
market. Another distinction is the denomination of the claim.
Normally, money market instruments are traded in large denominations
so as to be out of reach of normal households. They are used by banks and
corporations to lend and borrow in the short term. Bonds, on the other
hand, can be held by households.
23 Investment management
Price = 78:35 =
100
1.055
100 + 5
1.05364/365
100 + 5
1.051/365
At the time of the next coupon payment, the bonds trade at exactly par
value since the coupon rate equals the discount rate. But because of the
difference in the timing of the coupon the actual prices are different. The
accrued interest for the two bonds is given by the formula:
days since last coupon payment
Accrued interest = coupon payment
We find, therefore, the following quoted prices for the two bonds:
1
= 100.00
365
364
= 100.00
365
The adjustment for accrued interest makes the prices comparable. Bonds
are fixed securities but they often feature call provisions. Gilts often have
call provisions determining the redemption date so that the UK
government may retain flexibility to redeem the bond within given time
intervals. It is common in these circumstances to treat the redemption date
as the first date in the redemption interval if the coupon rate is greater
than the current market rate (so that the loan is relatively expensive
compared to the current rate for the UK government) and conversely as
the last date in the redemption interval if the coupon rate is less than the
current market rate.
Equity markets
Equity is, as opposed to a fixed claim like a bond, a residual claim. This
means that it has a cash flow that is in the form of the residual cash flow
of the real asset after all fixed claims with promised payments are paid
off. For instance, if a business is financed by a 10-year bond in addition
to its equity, the equity holders have a claim on the business net of the
cash flow that is promised to the bond holders. The equity claim is the
means by which ownership and control for corporations are separated.
When we refer to the owners of a corporation we mean the owners of the
corporate equity and not the owners of the corporate debt, although both
have claims on the cash flow of the firm. The owners of the equity are,
however, normally not directly involved in the running of the corporation
this is the job of the executive manager who is hired to do precisely
this job. Therefore, the ownership is separated from the control function
in corporations. The manager is hired on a long-term basis (although he
may be fired at short notice) whereas the owners of the equity can decide
for themselves whether they wish to invest long-term or short-term in
the corporation. The separation of ownership and control is, therefore, a
simple way to achieve a long-term stable management structure even if
the owners of equity are all short-term investors. In partnerships (such
as many accounting and legal practices) it would create a great deal of
operational upheaval to have ongoing ownership changes taking place.
11
23 Investment management
We can say, however, that the equity holders have more influence on the
running of the company than the debt holders. The direct influence of an
individual equity holder is nonetheless limited. An equity holder normally
gets the right to vote in general meetings. This means in practice that
he gets the chance to influence a few very important decisions such as
large investment projects or decisions related to corporate mergers and
takeover through his vote, and also to influence the choice of who sits
on the non-executive board of directors (NED). The NED has a direct
oversight on the executive management team of the corporation, and it
is essentially through representation on the NED that shareholders have
their main influence in the running of the firm. A lot of measures aimed at
strengthening corporate governance are aimed at making the NED more
effective in overseeing the executive management team.
Initially, companies issue equity that is owned privately (i.e. it is not
listed on a stock exchange) by an entrepreneur, a family, or by venture
capitalists. The process of making private equity public normally involves
the corporation seeking listing of its equity on a stock exchange. The
equity can thereafter be traded freely by all investors. The first time
a company seeks a listing is called an Initial Public Offering (IPO).
Subsequent equity issues are called seasoned issues, and these are much
less involved than the IPO since the equity has been traded for a while
before the issue. If the company sells existing equity (for instance, if the
government sells equity that is already issued but fully state owned) in the
IPO or during a seasoned issue, we call it a secondary issue. If new equity
is issued, we call it a primary issue. Sometimes the company needs to raise
additional capital when it goes public, and in this case it is necessary to
make some of the issued equity a primary issue. Otherwise, the issue is
primarily a process of transferring equity from the initial owners to the
new investors.
Equity instruments
Equity instruments consist of stocks common stocks or preferred stocks
in publicly traded companies. The two most distinctive features are that
they are residual claims and that an owner can exercise the right to limited
liability (i.e. the owner can decide to relinquish his claim on the real
underlying assets and instead hand these over to the other claim holders).
A residual claim is a claim that is unspecified, it will be determined as the
residual of the total corporate cash flow net of all fixed claims. Therefore,
if the corporate cash flow is 100m, on which the debt holders have a
fixed claim of 75m, the residual cash flow due to the equity holders is
the residual 100m 75m = 25m. The implication of the fact that equity
is a residual claim is that its value can never exceed the value of the total
real assets of the firm. The implication of the fact that the equity holders
can exercise the right to limited liability is that the value of the equity
can never become negative. Common stock and preferred stock differ in
two respects. First, common stock holders normally have voting power in
general meetings whereas preferred stock holders have not. Second, the
claim of preferred stock holders has seniority over that of common stock
holders. Thus, if the company wishes to pay dividends to its common stock
holders it must first pay a dividend to its preferred stock holders.
Common stock is often split into two classes (dual-class shares), usually
called A and B shares. These classes differ in their voting power, where
one class (normally A shares) have superior voting power relative to
the other class. The reason dual-class share structures are introduced is
12
that a controlling family may wish to retain the majority of the voting
power whilst at the same time may diversify by selling B shares to outside
investors. Dual-class share structures are relatively rare in the USA and the
UK but can frequently be found in Europe and Japan.
Derivatives markets
Bonds and equity claims are claims that perform a dual role. For the issuer
(businesses, banks or governments), these claims are a means of raising
capital used for investment or expenditure. For the investors, these claims
are means of smoothing real cash flows across time and states. Derivatives
are instruments that do not really play a direct role as a means of raising
capital that is, these instruments are in zero net supply. If no buyer exists
for a particular derivative instrument, then also no seller exists. Derivatives
are, therefore, almost exclusively used for risk management purposes.
Derivatives are also sometimes called contingent claims. The cash flow of
derivatives is almost always linked to the price of a primary asset such as a
bond or an equity claim the underlying asset. In this sense, therefore, the
cash flow is a function of, or contingent on, what happens to the price of
the underlying asset. However, recently we also observed derivatives that
had a cash flow contingent on other events, such as the event that a bond
defaults (credit derivatives), or the event that the weather is bad (weather
derivatives).
There are three broad types of derivative claims: futures (forwards),
options and swaps.
If you enter into a futures or forward agreement, you effectively
undertake the obligation to buy or sell an asset at a specified price in
the future.
An option is like a futures agreement, except that you have the right
to buy or sell rather than an obligation. This implies that you have
the right to opt out of the transaction if you own an option, but must
always carry out the transaction if you own a futures contract.
A swap is an undertaking to swap one cash flow for another cash flow.
Managed funds
Managed funds represent, in essence, a delegation of the investment
decision from the individual investor to a professional fund manager. We
distinguish between active and passive funds, fixed income and equity
funds, and open-end and closed-end funds.
An active fund is one where the fund manager typically makes investment
decisions that are in the form of bets the manager might think that
certain sectors or certain stocks are better bets than others and influences
the investments of funds to these sectors or stocks. A passive fund is
one where the fund manager typically attempts to mimic a broad stock
market index (like the FTSE 100 in London and the Standard & Poor 500
in New York). This normally amounts to physically holding the index or
a large number of stocks in the index. Open-end funds are funds where
the investors clear their holding directly with the fund. Therefore, if a
new investor comes in to buy units of the fund the fund simply issues
new units. The price the investor pays is the Net Asset Value (NAV) less
charges. The NAV is calculated as the total net value of the fund divided by
the number of units issued to investors. Closed-end funds are funds which
have a fixed number of units issued. If an investor wishes to buy units in
13
23 Investment management
14
23 Investment management
(and we need to work out the return on our initial equity position) and a
short trade (which is a loan, and we are interested in the implied loan rate
on our net loan, taking into account the deposit on our margin account).
First consider a long position. You buy 1000 shares of a stock at an initial
price of 100p per share, and one year later the price is 60p per share.
During the year, you receive dividends worth 10p per share. You hold the
position for another year, where you receive dividends of 8p per share, and
then finally at the end of year two you sell the position at a price of 110p
per share. The initial margin is 60%, and the maintenance margin is 40%.
There are three steps to the calculations here. First, you need to work out
the gross cash flows. In year 0, the investment cost is 1000 (1000 units
times the price of 100p). Then in year one, the cash flow is the dividend
payment of 100 (1000 shares times 10p) which we assume arrive at
the end of the year (this may not be the case of course but it is the most
conservative estimate). Finally, in year two, the cash flow is 1100 from
the proceeds of the sale, plus 80 from the dividend payment, a total of
1180. So the gross cash flow is (1000; +100; +1180). Next, we need
to work out the net cash flow. The initial margin is 60%, which means we
can borrow 40% on a margin loan. This gives us a cash inflow in year 0 of
400. Following the initial position there is a maintenance margin of 40%,
which means we have to keep 60% at the minimum as equity. We do not
need to check the end of year two as the margin loan is unwound in any
case, so lets look at year one. Here, the value of the position is 600, and
the margin loan is 400, i.e. an equity position of 200, which is 33.3%.
We need to maintain a 40% equity, so the maximum margin loan is 360.
Therefore, we need to pay off 40 of our margin loan. At the end of year
two the loan is repaid, and if we assume zero interest the cash flow is
360. The margin cash flow is, therefore, (+400; 40; 360). Therefore,
the net cash flow is ( 1000; +100; +1180) + (+400; 40; 360) =
( 600; +60, 820). The final step is to work out the rate of return on our
net investment. Here we use the familiar internal rate of return (IRR)
formula from course 92 Corporate finance:
600 +
60
820
+
=0
1+ IRR (1+ IRR)2
and as above it suffces to check year one only as you are unwinding the
position in year two. In year one the liability is 1600, and 40% of this
is 640. Since you have only 600 deposited, you need to put another
40 in the margin account. Ignoring interest rates on the margin account,
therefore, the margin cash flows are ( 600, 40, +640). The net cash
flow is (+1000, 100, 680) + ( 600, 40, +640) = (+400, 140;
40). The internal rate of return formula gives us here:
400
140
40
=0
1+ IRR
(1+ IRR)2
Summary
This chapter has outlined some basic facts on financial claims and
markets.
There was an overview of bond and money markets in which we trade
fixed claims, and an overview of equity markets in which we trade
equity claims which are residual claims (the exact opposite of fixed
claims).
There was an overview of derivatives markets, managed funds and
exchange traded funds.
The chapter also dealt with margin trading and how margin accounts
work.
Finally, there was a short discussion of regulation of financial markets.
Activities
1. Discuss why we need regulation of markets. Try to look for arguments
to support your discussion by looking up, for instance, issues related
to regulation on the websites of the London and New York stock
exchanges: www.londonstockexchange.com or www.nyse.com.
17
23 Investment management
2. If you buy an asset on a 50% margin, how much would you have to pay
initially if the price is 126p per share and you buy 1000 shares? How
much more do you need to pay if the price went down to 115p per
share?
3. Consider a short sales transaction on a 70% initial margin requirement
and 60% maintenance margin. You keep the transaction over five
months, and you trade 1000 shares of a stock. The initial price is 100p
per share. The price at the end of the first, second, third and fourth
month is 95p per share, 120p per share, 140p per share and 110p
per share, respectively. The price at the end of the fifth month is 98p
per share. Calculate the gross monthly profit, and the net monthly
profit taking into account the margin deposit. You can assume the
margin deposit account is interest free. What is the net monthly profit
if the deposit account pays 0.1% monthly interest rate? What is the
net monthly profit if the commission on the sale and the repurchase
transaction is 0.5% of the transaction amount?
18
Essential reading
Bodie, Z., A. Kane and A.J. Marcus Investments. (Boston, Mass.; London:
McGraw-Hill Irwin, 2008) Chapters 4 and 5.
Further reading
Allen, F. and D. Gale Financial Innovation and Risk Sharing. (Cambridge, Mass.;
London: MIT Press, 1994) Chapter 2.
Introduction
In this chapter we look at the historical and empirical evidence
surrounding financial markets and assets. The first part surveys the
innovations regarding financial instruments and the trading process in
financial markets. The second part surveys the history of investment
returns on financial assets. Three questions are addressed in particular:
What return can investors expect to earn when investing in various
types of financial assets?
What are the risk characteristics of these returns?
Are returns and risk characteristics linked?
An interesting issue is the historic relationship between the risk and return
on various instruments. If investors are risk averse we expect to find they
demand compensation for holding risky portfolios. This will be discussed
in relation to the so-called equity premium puzzle.
23 Investment management
were broken. Similarly, preferred stock has been in use for a long time.
Exchange trading of financial securities also has a surprisingly long history.
Equity was traded in Antwerp and Amsterdam in the 1600s. Moreover,
options and futures (called time bargains at the time) were traded on the
Amsterdam Bourse after it was opened in 1611.
Many of the European stock markets experienced major stock market
bubbles in the eighteenth century. A famous example is the South Sea
Bubble (1720) where the price of the South Sea Company rose from
131% of par in February to 950% by June 23, then fell back to 200%
by December. This bubble led to the so called Bubble Act which made
it illegal to form a company without a charter or to pursue any line of
business other than the one specified in the charter.
We also know of early examples of speculative bubbles.
23 Investment management
same way as currency swaps these are also netted out so that the swap
agreement effectively consists of a series of single payments.
rated instruments) have priority claim to this cash flow. The mezzanine
tranches (with intermediate rating) have seniority after the senior tranches
are serviced, and finally the equity tranche carries the residual claim.
It is commonplace that the financial institution selling off loans or debt
portfolios in this way retains the equity tranche.
It should be mentioned here that collateralised loan or debt obligations
have been cited as one of the factors causing the so-called credit crunch
which started in late 2007 and has continued to the time of writing. A
problem with collateralised loan or debt obligations is that if the financial
institution knows it will be able to sell the loan in the secondary market to
outside parties, there is little incentive to make sure its lending decisions
are sound. The liquidity in the market for collateralised loan or debt
obligations did dry up considerably in late 2007.
Geometric average
return
Arithmetic average
return
Small-company stocks
11.64%
17.74%
Large-company stocks
10.01%
12.04%
5.38%
5.68%
US T-bills
3.78%
3.82%
Inflation
3.05%
3.14%
Table 2.1
Arithmetic average
return
Standard average
return
Small-company stocks
17.74%
39.30%
Large-company stocks
12.04%
20.55%
5.68%
8.24%
US T-bills
3.82%
3.18%
Inflation
3.14%
4.37%
Table 2.2
23 Investment management
u (x)
u (x)
If the investor has CARA (constant absolute risk aversion) utility the utility
function takes the form u(x) = exp( x). The risk aversion coefficient is in
this case (as the CARA name suggests), the constant .
If asset returns are, moreover, normally distributed, we can write the
expected utility function as:
Var(x)
Expected utility = E(x)
2
Suppose a CARA investor is indifferent between holding large-company
stocks and long-term US Treasury bills over a long period of time. Then
the following expression must hold:
0.03182
0.20552 = 0.0382
0.1204
2
2
which is solved for a risk aversion coefficient of 4. This is a fairly
reasonable number, but asset returns are not normal so we cannot use
this simple model to estimate the implied risk aversion coeffcient. This
is the motivation for Mehra and Prescotts study. They fit a rigorous
theoretical model to data on the return on stock market investments and
government bonds. The model generates the risk aversion coefficient of a
representative investor (see Appendix 1 for a review of risk aversion and
the risk aversion coeffcient). They found that a reasonable estimate for the
risk aversion coefficient is between 30 and 40. This is way too high, as a
risk aversion coefficient of 30 implies, for instance, that if the investor is
facing a gamble where he has a 50% chance of doubling his wealth and a
50% chance of halving his wealth, he would be willing to pay up to 49%
of his current wealth to avoid the gamble, i.e. if his current wealth is 100,
24
Summary
This chapter surveyed the historical perspective on the financial system
and, in particular, financial innovations of various types. There was a
survey of examples of major financial innovations such as swaps and
collateralised debt obligations.
The second part of the chapter dealt with the long term return on
various classes of assets, where a strong relationship between risk and
return was uncovered.
This part also covered some controversial issues related to the
difference between equity returns and government bond returns
issues related to the so-called equity premium puzzle.
Activities
1. Explain the role of poison pill securities and discuss whether this is a
helpful innovation of financial securities.
2. The historical evidence points to the fact that riskier securities have a
greater average return. Explain why. We also know that the expected
prize in lotteries is smaller than the cost of participating (an example is
UKs Lotto: A $1 lottery ticket has an expected prize payment of around
$0.45). Can you think of a reason why people are reluctant to accept
risk in financial markets but happy to pay for risk in lotteries?
25
23 Investment management
26
Essential reading
Bodie, Z., A. Kane and A.J. Marcus Investments. (Boston, Mass.; London:
McGraw-Hill Irwin, 2008) Chapters 4 and 13.
Further reading
Lo, Andrew W. Hedge Funds. (Princeton, NJ:Princeton University Press, 2008).
Pole, Andrew Statistical Arbitrage. (Hobaken: Wiley Finance, 2007) Chapter 2.
MacKenzie, Donald An Address in Mayfair. (London Review of Books)
www.lrb.co.uk/v30/n23/mack01.html
Introduction
Investors do not always invest directly in financial assets. Sometimes they
use portfolio or fund managers to invest on their behalf. Professional
investors or fund managers now control the bulk of private investment
in financial assets, and its sheer size has made this sector a subject of a
lot of research. This chapter will provide an overview of the empirical
evidence of fund management and investment strategies. In particular,
we will discuss whether there is an empirical foundation for the notion
that fund managers provide value for money, and whether various types
of investment strategies (e.g. technical trading strategies, or the so-called
contrarian or momentum strategies) yield abnormal returns.
Among all the evidence surrounding investment strategies to gain long
term sustainable trading profits, we find a long list of anecdotal stories
of individuals making huge trading profits in inventive, ad hoc, ways.
They are of course interesting in their own right, but it should be noted
that they deal normally with trading opportunities that can be exploited
27
23 Investment management
only once. The Economist 2004 Christmas Special surveys some cases of
exceptionally profitable single trades. An interesting observation is that
the cases fall into one of two categories a normal trader spotting an
unexpected arbitrage opportunity, or a large trader using a window of
opportunity to exploit his market power. An example of the former is the
case of the Italian national Ludovico Filotti who worked for Barings Bank
in London. While on holiday in Italy in 1993 Mr Filotti discovered a new
savings scheme, guaranteed by the Italian government and issued by Italys
Post Offce offering a very high return relative to the Italian government
bonds. Although aimed at ordinary investors, the Post Offce had not
barred institutions from investing in such bonds. Having borrowed by
selling Italian government bonds, Mr Filotti flew personally to Italy with a
bankers draft of $50m to invest in the Post Offce bonds. The trade made
a huge profit and is a classic example of an arbitrage transaction where
a trader buys an asset cheaply in one market and sell it expensively in
another. An example of a large trader exploiting market power is the case
of George Soros who in 1992 betted against the UK Pound Sterling staying
in the European Exchange Rate Mechanism (ERM). He borrowed heavily
in pounds to invest in other currencies, forcing the Bank of England to use
its reserves to buy pounds to prop up the exchange rate. Eventually, the
pressure on the Bank of England reserves became so huge the government
decided to withdraw the pound from the ERM, netting Soros a profit of
around $1bn after unwinding his position.
These cases are atypical and one view is that they are also likely to become
less frequent as the financial system becomes more integrated and global,
and as the capital markets become more effcient. This view is supported by
the evidence of active portfolio management.
Although the average fund might not provide much value for money
for the investor, it may be that the best funds can. Several studies have
examined whether funds which perform better than the average over a
two-year period are also likely to perform better than the average in the
subsequent two-year period.
Study
Initial period
Top half
successive period
Bottom half
successive period
Goetzmann/
Ibbotson
Top half
Bottom half
62.0%
36.6%
38.0%
63.4%
Malkiel 1970s
Top half
Bottom half
65.1%
35.5%
34.9%
64.5%
Malkiel 1980s
Top half
Bottom half
51.7%
47.5%
48.3%
52.5%
Table 4.1
These results demonstrate that whereas winners and losers among fund
managers have a tendency to remain within their group over time, the
effect seems to be vanishing over time. The study based on the 1980s data
set shows that past winners are almost equally likely to become future
winners as future losers. Similarly, past losers are almost equally likely
to be future winners as future losers. Fund management performance
appears, therefore, to have become more and more associated with luck
than with skill.
pt+1 + dt+1
It
1+ r
where E denotes the expectations operator, pt+1 + dt+1 is the sum of next
periods price and dividends, respectively, and r is the discount rate. The
set It contains the current information. An implication of the efficient
market hypothesis is that future price innovations are unpredictable, i.e.
future prices are the forecasted price (todays price inflated by the discount
rate) plus an unpredictable pricing error.
29
23 Investment management
23 Investment management
N i
N
Since the weights are proportional to the deviation of the assets
performance relative to the equally weighted market index, they capture
the idea that the more extreme deviations lead also to more extreme
reversal and momentum effects.
i
Hedge funds
Hedge funds have become increasingly popular investment vehicles,
despite the high profile collapse of the Long Term Capital Management
hedge fund (LTCM) in 1998. Hedge funds have no specific definition,
but their activity is characterised by very flexible investment strategies
involving both long and short positions, often in complex securities.
Moreover, investors are often asked to commit their capital for a fixed term
such that the hedge fund managers can pursue their investment strategy
without the need to worry about investor redemptions. The essential
difference between hedge funds and other financial institutions is that
they are not heavily regulated.
One of the first hedge funds was set up in 1949: A.W. Jones & Co.
developed an investment strategy based on long/short positions in
equities. The idea is to buy stocks you think will do well and sell (or short)
stocks you think would do badly. If the market moves in the meantime,
the long and short positions will move together to maintain your net
portfolio value, and you make money if your stock picking is correct (in
bull as well as bear markets). As the sophistication of hedge funds grew,
so did they turn to other markets. One of the investment strategies in fixed
income (bond) markets is the on-the-run/off-the-run strategy employed
by LTCM. The on-therun/off-the-run strategy employed by LTCM is based
on the institutional feature of the US government bond market which
issues new bonds every six months. Every new auction brings, say, a new
30-year government bond to the market which investors compete to buy
(the bond goes on-the-run). When the bond is six months old, it becomes
a 29.5-year bond and a new 30-year bond is issued. The old bond goes
off-the-run. LTCM observed that the difference between a 30-year bond
and a 29.5-year bond is almost imperceptible, so they should have the
same yield. In practice, however, there was a spread that was caused by
the fact that when the new bond went on-the-run its price was bid up.
Therefore, LTCM sold short the new 30-year bond and bought the old 29.5year bond to unwind its position six months later when the spread was
32
33
23 Investment management
Hedge Fund
1.4%
3.6%
3.6%
5.8%
8.9%
18.3%
14.0%
27.0%
1.39
2.15
34
110
110
x =
80 = 110
80 B 80
The net cost of this strategy is zero, and suppose we hold the position until
the two stocks have the same price. If they converge when the price of
both A and B is 120, the profit is:
120xA 120xB = 120(110) 120(80) = 120(110 80) = 120(30) = 3600
If they instead converge when the price of both A and B is 50, the profit is:
50xA 50xB = 50(110) 50(80) = 50(110 80) = 50(30) = 1500
Regardless, if the prices converge we make a trading profit.
We shall now go through a trading strategy which takes advantage of
deviations of spreads (or differences) between two rates or assets. The
idea is that if todays spread is lower than the average spread you should
bet that it will widen tomorrow, and if the spread is higher you should
bet that it will narrow tomorrow. Under certain assumptions, this strategy
yields a profit 75% of the time, i.e.
Pr(St > St 1 and St1 < E(St)) + Pr(St < St 1 and St 1 > E(St)) = 75%
To see this, consider the following diagram:
St
Todays spread St
above yesterdays
spread St 1
E(St)
Yesterdays spread St 1
above the mean E(St 1)
Yesterdays spread
St 1 below
the mean E(St 1)
St 1
E(St 1)
From the figure we can see where profits are not made. We do not make
profit if yesterdays spread is above the mean, and todays spread is also
above yesterdays spread, and we do not make profits if yesterdays spread
is below the mean, and todays spread is also below the mean. This area is
highlighted in the next diagram:
35
23 Investment management
St
E (St)
St 1
E (St 1)
Let us assume that the spread each day is an independent draw from the
same distributions. It follows that each of the four areas of the diagram to
the left or to the right of E(St 1) and above or below E(St) (the four
rectangular shapes separated by the mean spread) has probability 25%, and
the loss-making areas exactly cuts two of these areas in half. Therefore, the
probability of making a loss is only 25% and the probability of making a
gain is the complementary probability 75%.
You should be careful with an illustration such as this, however, as the
assumption that the spread is drawn each day from the same distribution
is in fact not innocent. If the spread is unusually large one day, it is likely
to be large for a reason, and it is likely that this reason also causes the
spread the next day to be high (in expectation). Therefore, the conditional
expected spread E(StSt 1) is likely not to be equal to the long term average
E(St) (which, to those of you who are familiar with the laws of statistics,
is given by the double or iterated expectation E(St) = E(E(St St 1))). The
observation of a high spread should not, therefore, necessarily mean that
we expect the spread to narrow the following day. In fact, what happened
around the collapse in LTCM (as was also the case with Volkswagen shares)
was that spreads that were seen as unjustifiably large did not narrow over
time but kept widening further. There just is no simple way of making
money in financial markets.
Summary
This chapter took an investor perspective on the history of finance, and looked
at the historical evidence of the performance of the managed fund industry.
The main finding was that managed funds do not, on average, outperform
broad stock market indices, which is indicative that markets tend to be
informationally effcient.
The chapter went on to discuss critics of the efficient market hypothesis who
use arguments based on behavioral finance. Some trading strategies based on
behavioural finance (momentum and reversal effects) were outlined.
36
Activities
1. Describe the efficient market hypothesis. If the efficient market
hypothesis is really true, but traders nonetheless keep searching for
trading strategies that can beat the market, do you think they would
find useful trading strategies? Suppose we can construct trading
strategies that yield symmetrical risk profiles where abnormal gains
and losses are similar in magnitude and frequency, and strategies that
yield asymmetrical risk profiles where abnormal gains are small but
frequent and losses are large but infrequent. Which strategy do you
think we would be more likely to find in an efficient market if we were
out to beat the market?
2. Consider the asset prices given in the table below. One is generated
under the efficient market hypothesis and the other is not. Your task is
to identify which is which.
Asset A
Asset B
100.00
100.00
98.58
100.25
97.78
101.57
99.78
101.62
97.83
105.35
99.64
112.38
101.07
114.89
102.70
122.21
99.61
134.35
102.48
142.81
106.77
154.47
109.24
168.91
107.79
184.15
111.04
191.92
116.00
196.27
118.65
200.51
122.66
199.74
128.39
197.06
128.30
201.85
122.00
215.77
37
23 Investment management
38
Essential reading
Bodie, Z., A. Kane and A.J. Marcus Investments. (Boston, Mass.; London:
McGraw-Hill Irwin, 2008) Chapter 3.
Elton, E.J., M.J. Gruber, S.J. Brown and W.N. Goetzmann Modern Portfolio
Theory and Investment Analysis. (New York; Chichester: John Wiley & Sons,
2010) Chapter 3.
In this chapter I have listed Chapter 3 of Bodie, Kane and Marcus as the
Essential reading, but clearly for much of the analytical material in this
chapter you will probably find Hasbroucks book listed below of much
greater relevance. The reason Hasbroucks book is not Essential reading is
that market microstructure is still a relatively new area in finance, and has
yet to find its place in asset pricing and investment management.
Further reading
Hasbrouck, Joel Empirical Market Microstructure. (Oxford: Oxford University Press,
2007) Chapters 1, 2, 3, 5 and 7.
Introduction
This chapter looks at the microstructure of markets, which essentially
studies the trading mechanisms for financial securities. The microstructure
of markets deal with a number of issues:
The reasons for trade: Trade involves exchange of cash for an
asset with (generally) unknown value. The private value of assets may
be different from the cash equivalent that can be exchanged for the
asset, which motivates trading behaviour. This is normally captured by
market microstructure models where informed traders operate.
The protocol for trading: Market microstructure models are
normally very specific about the rules of the game for trading financial
securities.
39
23 Investment management
40
Buy quantity
Limit buy
100
100
Limit sell
Market buy
Market sell
Order
Sell quantity
50
110
20
80
The open price is generated on the basis of price priority. To illustrate the
orders received above, we can draw the following graph.
Price
110
100
Quantity
20
80
120
The buy orders are ordered from the highest price to the lowest (where
the highest price will stand the greatest chance of being executed first),
and the sell orders are ordered from the lowest price to the highest (where
the lowest price will stand the greatest chance of being executed first). The
lines cross at a quantity of 80 and a price of 100, so the open price will be
100. Therefore, all of the market sell order will be executed at the price of
100, and all of the market buy plus 60 units of the limit buy will also be
executed at the price of 100. This all happens at the open, and the limit
orders not executed will remain in the system to meet new incoming
orders. The limit order book will, therefore, consist of a limit buy of 40
units at a limit price of 110, and a limit sell of 50 units at a price of 100.
These orders form the bid-ask spread of the market, which is the ask price
of 110 and the bid price of 100. Following the open, new orders are
collected and executed according to two priority rules. First priority is as
before on price: the buy orders with the highest price and the sell orders
with the lowest price have always the greatest chance of execution. Next,
orders are ordered on submission time, where the oldest orders (given the
same price) have the greatest chance of execution. Fully electronic limit
order markets have now become a very popular way of organising trading
activity on stock exchanges around the world.
Some background readings on limit order markets can be found in Elton,
Gruber, Brown and Goetzmann Chapter 3, which will probably be useful
when reading the more technical material in Hasbrouck Chapter 2.
23 Investment management
Glosten-Milgrom
, where Pr(V = _V )
There is a security with pay-off V which is either _V or V
= . The population of traders consists of uninformed noise traders and
informed speculators, with the proportion of informed speculators equal
to . The market maker (or dealer) quotes bid and ask quotes, B and A,
respectively. The traders are drawn randomly from the population. If
and sells if
an informed speculator is drawn, the trader buys if V = V
V = _V . If the trader is an uninformed noise trader, the trader buys or sells
with probability one half each. The market maker cannot tell whether
the trader drawn is informed or not. We find, therefore, the following
probability structure for this model:
42
1
1
Figure 5.1: The tree shows the probability structure of the Glosten-Milgrom
model
A buy transaction always takes place at the ask price A, and a sell
transaction always takes place at the bid price B. The probability of a
low asset price _V and a sell transaction at the bid price B by an informed
speculators is , and the probability of a low asset price _V and a sell
transaction at the bid price B by an uninformed noise trader is (1) .
The market makers problem is to determine bid and ask prices B and A,
such that the market maker makes zero profit on the transaction. This does
not mean the market maker is going to set the same price as the bid ask
price, however. The market maker thinks it is more likely that somebody
is willing to buy the asset at the ask when the value is high, since an
informed trader will never sell when the value is high. The event that
somebody wants to buy at the ask is, therefore, good news and the event
that somebody wants to sell at the bid is bad news, caused by so-called
adverse selection. The prior beliefs of the market maker is that the value
is low with probability and high with probability 1 , and a bid or an
ask transaction leads to revision of the market makers beliefs. The revision
process is governed by Bayes law:
Pr(V = _V )Bid)Pr(Bid) = Pr(BidV = _V )Pr(V = _V )
which yields:
Pr(V = V |Bid) =
Pr(Bid|V = V ) Pr(V = V )
Pr(Bid)
Here, the probability of a bid transaction given a low asset value is (that
an informed trader wants to sell at the bid) plus (1 ) (that an
uninformed noise trader wants to sell at the bid). The unconditional
probability of a bid transaction is ( + (1 ) ) (that there is a bid when
the asset value is low) plus (1 )(1 ) (that there is a bid when the
asset value is high). We notice that the unconditional probability of a bid
transaction is:
) Pr(V = V
)
Pr(Bid) = Pr(BidV = _V )Pr(V = _V ) + Pr(BidV = V
The unconditional probability is, therefore, simply the sum of the
conditional probabilities weighted by the unconditional probabilities of the
event on which we are conditioning. Putting it all together, therefore, we
find:
43
23 Investment management
Pr(V = V |Bid) =
( + (1 ) )
( + (1 ) ) + (1 ) (1 ))
Pr(V = V |Bid) =
1 (1 2)
(1 )
1 + (1 2)
The formula for the bid and ask prices are, therefore:
B=
A=
(1 + )
V +
1 (1 2)
(1 )
1 + (1 2)
(
(
V +
1
(1 + )
1 (1 2)
(1 )
1 + (1 2)
)
)
The bid ask spread is defined as the difference between the higher ask
price and the lower bid price:
Bid-Ask Spread = AB
(
(
(
[(1 )]
1 + (1 2)
[(1 + )]
1 (1 2)
[(1 + )]
) (
) (
V +
V +
1
1 (1 2)
(1 )
1 + (1 2)
(1 + )
1 (1 2)
(1 )
1 + (1 2)
) )
) )
V
(V V )
Let 0 denote the initial beliefs of the market maker and set 0 = . If the
first transaction is at the bid, the market maker must set bid and ask prices
also for the second round. The market maker has at this stage revised his
beliefs to Pr(V = _V )Bid) = 1. The bid-ask prices in the second round can
then be worked out by repeating the process above setting 1 = as the
market makers prior beliefs. In general, therefore, we find the
relationship:
k(Ask) =
k(Bid) =
44
k1(1 )
1 + (1 2k1)
k1(1 + )
1 (1 2k1)
Kyle
The Kyle model is a dynamic model of security prices where the market
maker is setting informationally efficient prices (as in the GlostenMilgrom) model, but where additionally an informed trader chooses
optimal trading strategies. This extends the Glosten-Milgrom model in the
sense that the insiders can choose how much to trade in the Kyle model
in the Glosten-Milgrom model they simply trade a single unit of the stock
as they are drawn to trade.
We consider a security with value which is distributed v~N(p0;20) (i.e.
normally distributed with mean p0 and variance 20). A single informed
trader knows the realisation of v, and submits a market order x(v). There
are also other traders in the market who do not know v and who trade
for reasons independent of v. The net market orders from these traders
the liquidity traders is u ~ N(0, u2) (i.e. normally distributed with mean
zero and variance u2). The market maker observes the total net order
flow y = x + u and sets a price p(y) (notice the price is not a function of
x and u separately the market maker can only see the net order flow).
The market maker is risk neutral and is facing perfect competition, so in
clearing the market, the market price p(y) must be equal to the expected
asset value conditional on y:
p(y) = E(v\y)
The informed trader assumes the market maker uses a linear price setting
strategy: p(y) = + (y E(y)). The profits of the informed trader are
(v) = (v p)x. Using the fact that the price depends on y and not on x, the
trading profits are stochastic because the market price is stochastic:
~
~
~
(v) = (v ( + y))x = (v ( + (x + u ))) x . Taking expectations, the
expected profits equal:
~
= (v ( + (x + E(u ))))x
= (v ( + x))x
= vx x x2
Maximising the expected profits, we find (the first order conditions for
maximum is that the derivative (in x) is set to zero):
~
d E((v))
= v 2x = 0
dx
(v )
2
We notice that this leads to a linear strategy also for the informed trader:
1
x = + v, where = 2 and = 2 . What we need now is to tie the linear
trading strategy to the linear price formula, using the condition that prices
~
are informationally effcient: p(y) = E( vy). First, however, we notice that
~
~
~
~
x ~ N( + p0, 220 ), and also that y = u + x ~ N( + p0, 220 + u2).
45
23 Investment management
~
~
Cov(v, y) (y E(y))
~
Var(y)
E(v| y) = E(v) +
E( v) = p0
~
E( y) = + p0
~
Cov( v, y ) = Cov( v, u + x )
~
= Cov( v, u + + v)
~
= Cov( v, v) = Var( v)
2
= 0
~
Var( y) = Var(u + x )
~
= Var(u + + v)
~
= Var(u ) + 2Var( v)
= 2u + 220
so we find:
20
~
E(v|
y) = p0 +
= p0
(y p0)
2u + 220
2 20
2u + 220
20
2u + 220
20
2u + 220
We also know that + y = E( vy) so our guess that pricing rule is linear is
satisfied. Now we just need to put all the information together. Our
information is that:
= 2
1
=
2
(
(
) (
= p0 1
= p0 1
1
42
2u +
20
1
2
42 0
2 20
2u + 220
1
42
2u +
= p0 1
1
42
2u +
20
1
2
42 0
20
1
2
42 0
)
)
)
20
2u + 220
1
2
22 0
(which, rearranging, leads to...)
2 2 + 1 2
2
u
4 0
2 20
2u + 220
1
22
2u +
1
46
20
1
2
42 0
2u +
1 2
1 2
4 0 = 2 0
2u =
1 2
4 0
2 =
2
1 0
4 2u
1 0 (an then everything unravels from top and the model is solved ...)
2 u
=
=
p0 u
0
u
0
1 0
y
2 u
and the equilibrium trading strategy by the informed investors is given by:
u
v
x(v) = p0 u +
0
0
We notice that as the variance of noise trading increases relative to the
variance of the asset pay-off, the more aggressive trading will be observed
by the informed trader as there is an increase in the coefficient on v. At the
same time, however, the market prices become less sensitive to the order
flow as the coefficient on y decreases. A market with a lot of noise trade
has, therefore, one of the characteristics of a liquid market: traders can
trade large quantities without moving prices a great deal.
We can also work out the equilibrium using a regression interpretation
on the projection (actually, a linear regression model is just a linear
projection). If we imagine the market maker can regress asset pay-offs v on
order flow y, the expected price conditional on order flow is given by:
~
E( vy) = a + by + e
where a and b are regression coeffcients and e is an error term with zero
expectation. The regression coefficient is given by:
b=
~ ~
Cov(v,
y)
=
Var(y~ )
20
+ 220
2
u
20
(a + p0 )
+ 220
2
u
which corresponds to the findings above and the rest of the model follows.
There is more detail on the Kyle model in Hasbrouck Chapter 7.
23 Investment management
The variance of the asset pay-off is, by construction, equal to one. Consider
a single noise trader who trades the asset in the quantity:
q = + with probability
with probability
The constant represents the standard deviation of noise trade. To see
this, work out the variance:
1
1
Var(q) = 2 + 2 = 2
2
2
and the standard deviation is just the square root of 2 which is .
There exists an insider who has perfect knowledge of the realisation of x,
and who wants to trade to benefit from his privileged information. Let y
denote the order of the insider. The noise trader and the insider submit
their order to a market marker, who is risk neutral and will clear the
market at a price p which is informationally efficient given the order q
and y:
p = E(x|q, y)
The problem for the insider is to pick y such as to maximise expected
trading profits, and the problem for the market maker is to make rational
inferences given the observation of the order flow.
It is easiest to consider the market makers problem first. The market
maker observes q and y, and since the absolute value of q is known:
|q| =
the market maker can identify the insiders order, provided:
|y|
In this case, therefore, the market price will always be fully revealing, i.e.
p = E(x|q, y) = x, so the insider cannot ever make trading profits. Therefore,
it is optimal for the insider to set |y| = also.
Now we consider the market makers problem. Since |q| = |y| = , the
market maker observes one of three aggregate orders: (i) the insider and
the market maker both submit orders of so the aggregate order is 2.
In this case the market maker knows the insider is buying, so the market
price p = E(x|, ) = 1 is fully revealing; (ii) the insider and the market
maker submit orders that offset each other, so that the aggregate order is
= 0. In this case the market maker cannot tell whether the insider is
selling or buying (both equally likely) and does not infer anything from
the order flow. Therefore, the market price p = E(x|, ) = E(x| , ) = 0
reveals no information; and finally (iii) the insider and the market maker
submit orders so the aggregate order is 2. In this case the market
maker knows the insider is selling, so the market price p = E(x| , ) = 1
is fully revealing.
when we are planning to buy, we will incur liquidity costs that reduce
the return on our investments. The second implication is that when
deviations from the fundamental value occur, we may be fooled into
making judgements about assets that are incorrect. An example is the
market squeeze in Volkswagen stock that occurred in 2008 (described
briefly in the Hedge funds section in Chapter 3), that made Volkswagen
briefly the largest company in the world. In such situations we are right
in being sceptical about asset prices. A good strategy for protecting
yourself against liquidity effects and the risk of price deviations from
fundamentals is never to trade in such a way that you may need to carry
out a fire-sale of your portfolio, and never to trade too large a quantity at
any one time. A cautious trading strategy where you keep a liquid reserve
of funds available for liquidity shocks, and placing small amounts in the
market at regular intervals is likely to protect you against these market
microstructure effects.
Second, if you are a relatively unsophisticated trader with poor
information, you are likely to incur specific costs of trading against more
sophisticated traders, the so-called adverse selection cost of trading.
There is no obvious way of detecting and protecting yourself from the
activity of well-informed investors as they tend to be clever at hiding
their information as they trade (as predicted by the Kyle model). This is
of course not necessarily an argument against trading per se, but it is an
argument against trading very often (e.g. frequent buy-sell transactions)
as, in this case, you increase the likelihood of ending up at the opposite
side of a clever, well-informed investor.
Summary
This chapter dealt with market microstructure which, broadly speaking,
is the process by which investors intention to trade is ultimately
transformed into actual transaction volume and price.
The chapter had a brief outline of the workings of a limit order
market, which is now a common market structure for exchange-traded
instruments such as bonds and equities.
The Roll, the Glosten-Milgrom and the Kyle models were briefly
outlined, and these models contain most of the relevant concepts that
appear in relation to market microstructure.
There was also a relatively simple discrete outline of the Kyle model to
complement the original set-up, which is somewhat technical.
Finally, the chapter discussed why market microstructure is relevant for
investors.
Activity
1. Imagine the following game show format: you are invited to choose
one box out of three, where one of the boxes contains a prize.
Before you open your box, the game show host opens one of the two
remaining boxes that he knows does not contain a prize, and invites
you to swap your box with the remaining unopened box. When this
game show is run the majority of people keep their original box and
refuses the swap. However, explain why you maximise your chances of
winning the prize by making the swap. Explain also how this relates to
adverse selection and market microstructure, where if you buy assets
you are more likely to trade an overvalued asset than if you sell.
49
23 Investment management
50
Chapter 6: Diversification
Chapter 6: Diversification
Learning outcomes
By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
accurately compute the expected return, risk premium, variance and
the standard deviation of a risky portfolio
accurately compute the expected return, risk premium, variance and
the standard deviation of a combination of a risky portfolio and a risk
free asset
clearly define mean-variance preferences and coherently explain how
investors with mean-variance preferences choose portfolios
derive the optimal portfolio with the knowledge of the investors' risk
aversion coefficient with confidence
define the portfolio frontier in detail
formally explain how the existence of a portfolio frontier on which
investors choose their optimal portfolios implies that the CAPM pricing
formula holds
concisely define the concepts of systematic and unsystematic risk, and
cogently explain how these concepts are used to simplify the problem
of estimating the covariance between asset returns
discuss the implications of the single index model to effective
diversification in detail
review the Treynor-Black model
thoroughly define factor models, and illustrate well-established pricing
formulas
knowledgeably explain why myopic portfolio choice may sometimes be
optimal even though the investors have a long investment horizon.
Essential reading
Bodie, Z., A. Kane and A.J. Marcus Investments. (Boston, Mass.; London:
McGrawHill Irwin, 2008) Chapters 6, 7, 8, 9, 10 and 27.
Elton, E.J., M.J. Gruber, S.J. Brown and W.N. Goetzmann Modern Portfolio
Theory and Investment Analysis. (New York; Chichester: John Wiley & Sons,
2010) Chapters 4, 5, 6, 7, 8, 9, 12 and 13.
Further reading
Campbell, J.Y. and L.M. Viceira Strategic Asset Allocation. (New York: Oxford
University Press) Chapter 2.
Introduction
Now we turn to the problem of using the underlying pricing theory
(which is not reviewed in this guide but is extensively outlined in the
Corporate finance subject guide) as a tool for investment analysis.
The topic for this chapter is the principle of diversification. Investors
can eliminate free of cost a great deal of portfolio risk by spreading
their investment to a broad portfolio. Some movements in stock prices
51
23 Investment management
where E(rP ) is the expected return of the portfolio, wi is the weight of asset
i in the portfolio, and E(ri) is the expected return on asset i. In our example
above, we find that the expected return is given by:
E(rP) = 30%E(rBT)+50%E(rLloyds Bank)+20%E(rMarks & Spencer)
The next important formula is the one that expresses the variance of the
return on a portfolio. Here we find:
Var(rP) =
i
ji
where we sum over all assets i = 1, 2, ..., n and j = 1, 2, ..., n. The righthand
side can be written in two ways. The most familiar way to express the
variance of a portfolio of assets is as the sum of the variances of the assets
multiplied by the portfolio weights squared, plus two times the sum of all
covariances multiplied by the two corresponding portfolio weights. This
corresponds to the rightmost expression in the formula above. However,
the variance of the return on an asset is simply the covariance of the
return with itself so we can express all variance terms as covariance terms.
Also, for each covariance term between the return on asset A and asset B,
52
Chapter 6: Diversification
Numerical example
Consider the three stocks BT, Lloyds Bank and Marks & Spencer given
above. Suppose the expected return of the stocks and the variance are:
Stock
Expected return
Variance
BT
10%
9%
Lloyds Bank
13%
16%
8%
8%
Lloyds Bank
BT
Lloyds Bank
5%
2%
3%
53
23 Investment management
5%
= 0.42
9% 16%
which implies that if you mix BT and Lloyds in a portfolio some of the risk
in the individual stocks will be offset by the fact that the correlation
coefficient is not perfect (i.e. it is not 1).
A portfolio that consists of a fraction w invested in A and the remaining
fraction 1 w invested in B, has expected return E(r) = wE(rA) + (1 w)E(rB)
and variance of return Var(r) = w2Var(rA) + 2w(1 w)Cov(rA, rB) +
(1 w)2Var(rB). If E(rA) = E(rB), an investor with meanvariance preferences
chooses to hold A and B such that the variance is minimised. We know that
the minimum variance portfolio has first derivative with respect to w equal
to zero, so we find the first order condition which implies:
w=
B (B ABA )
(B A )2 + 2AB (1 AB)
where we use the fact that 2i = Var(ri) and ABAB = Cov(rA, rB).
If we consider two assets A and B with expected return 10% and variance
16% and 9%, respectively, with correlation coefficient 0.5, the minimum
variance portfolio consists of:
w=
30%(30% 0.5(40%))
(30% 40%)2 + 2(40%)(30%)(1 0.5)
= 0.23
Chapter 6: Diversification
Er
Er rF
rF
Figure 6.1
Meanvariance preferences
This section assumes some knowledge of utility theory. If you are
unfamiliar with utility theory you should first consult Appendix 1
which has a brief outline of this theory. If investors have meanvariance
preferences they have a utility function over portfolios that take the
form u(, 2) where is the expected return of the portfolio and 2 is the
variance of the return of the portfolio. This function is increasing in and
decreasing in 2. A rational investor picks a portfolio that maximises his
utility, that is, a portfolio that has an optimal riskreturn tradeoff. This
tradeoff can be found in the following way (using standard optimisation
techniques). Consider a general twoasset example where = w1 + (1 w)
2 and 2 = w221 + 2w(1 w)12 + (1 w)222 (here i is the expected return on
asset i, 2i is the variance of asset i, and 12 is the covariance between 1 and 2).
The first order condition for utility maximisation (see Appendix 1) is:
2
+ 2
=0
w
w
We find by working out the derivatives in the second brackets of the two
terms above and rearranging, that:
= (1 2)
w
and
2
= 2w (21 21212 + 22) 22(1 121)
w
55
23 Investment management
ErM rF
2w2M
u(, 2) = 2
2
56
Chapter 6: Diversification
and
u
=
2 2
such that the marginal rate of substitution between risk and return is:
u/2
=
u/
2
and if we put this into the left-hand side in the relationship above, we find
the optimal weight on the market portfolio as:
ErM rF
1
w* = =
2M
The more risk averse the individual is, therefore, the smaller the weight he
puts on the market portfolio, as the risk aversion coefficient enters into the
denominator on the right hand side.
You can read more about this approach to selecting among optimal
portfolios in Elton, Gruber, Brown and Goetzmann Chapter 12.
Figure 6.2
23 Investment management
figure we have illustrated three such combinations which all cross the
risk free asset and the risky portfolio. Here, any rational investor who
has meanvariance preferences will choose linear combinations of the
risk free asset and the tangency portfolio. The reason is that any other
portfolio he might think of either has the same or less expected return but
greater variance, or it has the same or greater variance but lower expected
return. The tangency portfolio is, therefore, the optimal risky portfolio for
investors to hold. This portfolio is normally a highly diversified portfolio,
and it achieves the minimum variance of portfolio return given its
expected return.
ErM
rF
Figure 6.3
The graph shows the investment opportunity set in the plane with
a risk free asset. We notice that the portfolio frontier is now linear, and
consists of a linear combination of the risk free asset and the market
portfolio. Also, we notice that the portfolio frontier is formed as a V lying
to the right, as there is not only a limit to the maximum return for any
level of risk, there is also a limit to the minimum return. From Figure 6.3
we can see that investors optimally choose to hold two funds only the
risk free asset and the market portfolio (or the tangency portfolio). This
property is called twofund separation. What is the market portfolio?
The law of supply and demand tells us that all stock issued must be held
by investors investing optimally in the market. Since all investors invest
optimally when they hold the tangency portfolio, this portfolio must
simply be the (valueweighted) index of all risky assets issued. Since this
portfolio is observable, we can estimate its expected rate of return. This
has further implications as we shall see below.
We can also find the optimal risky portfolio algebraically, although here we
will just sketch a method (a full derivation can be found in the appendix).
Suppose we know two risky portfolios on the portfolio frontier in Figure
6.2 say A and B. Then, an important property (that we do not prove
here) is that any portfolio that is formed by taking positions in A and B is
also a portfolio on the same frontier, with expected return
E(r) = wE(rA) + (1 w)E(rB)
where w is the weight on A. The standard deviation of the portfolio equals:
= w22A + 2w(1 w)ABAB + (1 w)22B
58
To find the frontier portfolio we need to maximise the risk premium per
unit of risk, i.e. we need to solve the problem:
max E(r) rF
w
Chapter 6: Diversification
Substituting for the expressions above, we can solve this programme and
find an exact expression for the optimal weights. This necessitates,
however, that we already have identified two frontier portfolios A and B.
In this case that may be unrealistic but we make use of this result below
when we look at optimal diversification in a framework where some assets
earn abnormal returns.
F
When we wish to maximise the slope
the first order condition
Cov(ri , rM)
Var(rM)
(ErM rF ) = i (ErM rF )
which is the CAPM pricing formula. This result tells us, therefore, what the
required expected rate of return on assets should be, given their risk
characteristics. The risk characteristics can be estimated through
estimating the beta factor for assets, and through estimating the aggregate
risk premium of the market portfolio. The market portfolio is simply the
index of all risky assets.
Estimation issues
This section contains an overview of the index model, which yields
considerable benefits in estimating the variancecovariance structure
of asset returns. If there are n assets, we need to work out n variances
and n(n 1)/2 covariances. If n is large this becomes computationally
demanding, and the index model gives us a simple method for reducing
the number of estimations. The idea behind index models is to decompose
the risk in asset returns into two types: the systematic (or market) risk and
the unsystematic (or idiosyncratic) risk. The systematic risk is the part that
is correlated with the risk of the market index, and the unsystematic risk is
uncorrelated with the risk of the market index. This model is very closely
related to the CAPM, and is indeed suggested by the CAPM expected
returns relationship above. The return on any risky asset can according to
the CAPM be written as:
ri = rF + i(rM rF) + i = (1 i)rF + i rM + i
where i is an error term with zero mean and zero covariance with rM. The
reason why the error term is uncorrelated with rM comes from the fact
59
23 Investment management
that rF + i(ErM rF) is the predictable part of the expected return on the
stock. Any error relative to the predictable part must be uncorrelated to
any random variable correlated with the predictable part. The single index
model is slightly more general as it states that asset returns are written as:
ri = ai + birM + ei
with ai a constant, bi is the regression coefficient with estimate equal to the
CAPM beta: Cov(ri , rM), and ei is an error term, again with zero mean and
Var(rM)
zero covariance with rM.
Taking the variance on both of the asset returns yields the following
relationship:
var(ri) = Var(i rM) + Var(i) = Var(birM) + Var(ei)
where the two last terms are identical as long as the estimated beta
bi is the same as the true CAPM beta i . This relationship allows the
decomposition of total risk into market risk and idiosyncratic risk:
(Total risk)Var(ri) = (Market risk)2Var(rM) + (Idiosyncratic risk)Var(i)
Market risk is sometimes called systematic risk, and idiosyncratic risk is
called unsystematic risk. We can now derive the fractions of market risk
and idiosyncratic risk:
Proportion market risk = 2Var(rM)
Var(ri)
and
Proportion idiosyncratic risk = Var(i)
Var(ri)
which gives us the decomposition into market risk and idiosyncratic risk in
percentage terms.
If we now assume that the idiosyncratic risk term i is uncorrelated with
the idiosyncratic risk term j for two assets i and j, we have the single
index model written in factor form (we will talk about factor models
later). The covariance between return on two assets can be written as:
Cov(ai +birM +ei, aj +bjrM +ej) = Cov(birM, bjrM) = bibjVar(rM)
since the error terms ei and ej are uncorrelated with all other variables.
The index model therefore yields a very effective method of estimating the
variancecovariance structure of asset returns as it relies mainly on beta
estimates. For n assets, the number of beta estimates is n, and in addition
we need to work out the variance of the market index. The variance
covariance structure can be constructed from n + 1 estimates, therefore,
instead of n + n(n 1)/2.
Chapter 6: Diversification
E(r) = 1 E(ri)
n i
This portfolio has a beta factor equal to the average beta of the assets in
the portfolio. If we expand the right hand side and remove the
expectations operator, we find:
( )
r = 1 ai + 1 bi rM + 1 ei
n i
n i
n i
Letting A = 1n iai and B = 1n ibi and E = 1n iei we get r = A + BrM + E which
allows a risk decomposition Var(r) = B2Var(rM) + Var(E). If Var(ei) = i2 = 2
(i.e. the idiosyncratic risk term has the same variance across all assets), we
find:
2
() () ()
Var (E) = 1
n
= 1
n
i
2
i
= 1
n
i
2
n2 =
2
n
Therefore, the idiosyncratic risk of the portfolio will go towards zero for n
large, and we achieve full diversification. Provided the index model is true,
therefore, we can create a fully diversified portfolio by simply adding a
suffcient number of assets to our equally weighted portfolio.
61
23 Investment management
expect are positive for positive expected abnormal return and negative for
negative expected abnormal return.
We know that if we hold a passive portfolio exactly replicating the index,
the unsystematic risk of all assets cancels out. Once we start to load up
on assets with positive alphafactor, however, we incur diversification
costs that are linked to the unsystematic risk component of that asset.
Therefore, our active portfolio weights must reflect this.
The starting portfolio is the market index, which we call P denoting the
passive portfolio with expected return ErP and standard deviation P .
Suppose a new active portfolio has been identified (for the time being we
do not go into details about how the only constraint is that this portfolio
is on the portfolio frontier) and lets call this portfolio A denoting an
active portfolio. The active portfolio has expected return ErA = A + (1 A)
rF + AErP and standard deviation A = A2 2P + Var(eA), where eA is the
idiosyncratic risk term of the active porfolio. The covariance between the
active portfolio and the passive portfolio is given by PA =AP2 .
The optimal risky portfolio is in this case neither the passive portfolio
P nor the active portfolio A, but some linear combination of the two.
Suppose we form a portfolio with weight w in the active portfolio and
1 w in the passive portfolio. Suppose further we wish to find the portfolio
that achieves the greatest expected risk premium given the standard
deviation is incurred. This portfolio is found as the solution to:
wErA + (1 w) ErP rF
2 2
A
Mixture portfolio
Active
portfolio
Passive portfolio
rF
Figure 6.4
Chapter 6: Diversification
Factor models
The index model can be interpreted as a onefactor model of asset returns.
It assumes that asset returns are influenced by a systematic risk factor that
is common to all assets, and unsystematic risk specific to each asset. This is
fairly restrictive since in reality there probably is more than one common
risk factor. The generalised factor model assumes, therefore, that asset
returns can be written as:
ri = ai + bi1 f1 + bi2 f2 + . . . + bik fk + ei
where bin are factor loadings for factor n = 1, 2, . . . k; fn are the factors; and
ei is the idiosyncratic risk term with zero expected value. The essential
assumption is that the covariance between any two idiosyncratic risks ei
and ej is always zero. It is also common to assume that the factors have
zero expected value. If we were to make the CAPM into a onefactor
model, therefore, we would have to create a factor f with zero mean. A
good candidate is the factor f = rM E(rM). It is easy to see that Ef = E(rM)
E(rM) = 0. Therefore, we find:
ri = (1 i)rF + i rM + i = (1 i)rF + i(f + E(rM)) + i = (rF + i(E(rM) rF))
+ i f + i
This illustrates that for factor models, the expected return is always
captured by the regression constant ai.
The arbitrage pricing theory tells us that if assets have a factor structure,
expected returns on stocks can be written as a linear combination of risk
premia, one for each factor, as in the kfactor case:
E(ri) = ai = rF + bi1 1 + bi2 2 + . . . + bik k
where n are the risk premia. The betas here are factor loadings.
Factor models can apply the principle of diversification to create portfolios
that contain a large number of assets (such as to reduce the unsystematic
risk term) and also to be immune to all but a single factor. To see how we
make use of such portfolios, consider a twofactor model and assume we
have already identified two fully diversified portfolios, A and B, that have
returns:
rA = aA + bA f1 and rB = aB + bB f2
These portfolios contain so many assets that the unsystematic risk
component simply vanishes. If we create a portfolio of these portfolios
with weight w in A and (1 w) in B, the return can be written as:
r = waA + (1 w)aB + wbA f1 + (1 w)bB f2
This portfolio is sensitive only to factor risk, since the original portfolios
are already fully diversified. The portfolios A and B can be used, therefore,
to diversify optimally and choose optimal factor loadings to suit the
individual investor.
63
23 Investment management
64
Chapter 6: Diversification
Summary
In this chapter we have seen how the principle of diversification arises
from the underlying fundamental pricing theory in finance.
Initially, we looked at how portfolios that include more than one risky
asset can reduce risk, and we expanded this simple example into a fully
fledged theory of optimal investment in risky and risk free assets.
We also looked at how we should accommodate abnormal asset returns
into this framework, and how to deal with factor models of asset
returns.
The final parts of this chapter looked at how to incorporate private
information in the diversification setup (in the context of the Treynor
Black model) and when it is permissible to incorporate a myopic
(static) perspective on asset allocation problems and when it is not.
Activities
1. Suppose you toss a coin, and let tails represent a gain of 1 and heads
a loss of 1. What is the expected pay-off of a single coin toss? What is
the variance?
2. Now suppose you toss the coin 100 times, and assume you receive
1
of the pay-off each time. What is the expected pay-off in this case?
100
What is the variance? Use these lotteries to explain diversification in
stock markets.
3. You can simulate risk using a computer package such as Microsoft
Excel. Try to simulate the effects of dependence between random
outcomes. For instance, if you use the formula RAND() in Excel it will
return a randomly chosen number between 0 and 1. If you generate
two random independent outcomes x and y using the RAND() function,
then you can generate a new random variable z = k(ax + (1 a)y)
(k 1)Ex for some numbers a and k. The expected value of z is k(aEx +
(1 a)Ey) (k 1)Ex = k(Ey + a(Ex Ey)) (k 1)Ex = kEx (k 1)Ex =
Ex = Ey. The variance of z is k2a2Varx + k2(1 a)2Vary (since x and y are
uncorrelated), which is k2(1 2a(1 a))Var(x) = k2(1 2a(1 a))Var(y). If
we choose k2(1 2a(1 a)) = 1, or k2 = (1 2a(1 a))1 the new random
variable z has the same expectation and variance as x and y. However,
in this case z is correlated with both x and y. Pick a value of a (and
work out k from the formula above), and make 100 draws of x and y
which enables you to generate 100 draws of z. Estimate the covariance
between x and y, and then the covariance between x and z. Are the
results what you expected?
23 Investment management
derive the optimal portfolio with the knowledge of the investors' risk
aversion coefficient with confidence
define the portfolio frontier in detail
formally explain how the existence of a portfolio frontier on which
investors choose their optimal portfolios implies that the CAPM pricing
formula holds
concisely define the concepts of systematic and unsystematic risk, and
cogently explain how these concepts are used to simplify the problem
of estimating the covariance between asset returns
discuss the implications of the single index model to effective
diversification in detail
review the Treynor-Black model
thoroughly define factor models, and illustrate well-established pricing
formulas
knowledgeably explain why myopic portfolio choice may sometimes be
optimal even though the investors have a long investment horizon.
66
Essential reading
Bodie, Z., A. Kane and A.J. Marcus Investments. (Boston, Mass.; London:
McGraw-Hill Irwin, 2008) Chapters 10, 14, 15 and 16.
Elton, E.J., M.J. Gruber, S.J. Brown and W.N. Goetzmann Modern Portfolio
Theory and Investment Analysis. (New York; Chichester: John Wiley & Sons,
2010) Chapter 21.
Introduction
In this chapter we outline the principle of immunisation. From the
previous chapter we recall that investors who invest in diversified
portfolios do nonetheless bear risk market risk. Sometimes, they may
find it useful to eliminate some or all of this risk as well. For instance,
bond portfolios denominated in foreign currency are affected by both
interest rate risk and currency risk. Both are types of market risk that
any such portfolio is exposed to. It may be that you want to take a bet on
the future developments in the interest rate market but at the same time
avoid the interfering effects of currency risk. To achieve this you need to
immunise your portfolio to currency risk; this can be achieved through
taking offsetting positions in the currency derivatives markets. How do
investors offoad risk from their portfolios to other investors?
We can achieve this through trading financial instruments, and we need to
make sure that our trading activity achieves the objective of transferring
the right type of risk without introducing new ones. This chapter discusses
immunisation strategies that make this type of trading possible.
67
23 Investment management
Bond maths
Before we proceed we will review quickly some bond maths. A bond is
characterised by its future cash flows, which normally come in the form of
annual or semi-annual coupon payments, and a final capital payment. The
price of the bond is the sum of all future discounted cash flows receivable
to the bond holder. We can, therefore, link the prices on bonds to the future
cash flows through the yield-to-maturity of the bond. Assuming annual
coupons, the formula is given by:
c
P= c +
+ . . . + c + 100
1 + y (1 + y)2
(1 + y)T
for a T-year bond paying coupons of c per 100 nominal value. The quantity
y is the yield to maturity, and it belongs to the particular bond in question.
We may wish to define other quantities such as the spot rates, which can be
used to price all bonds. If we have all spot rates from the one-year rate to
the T-year rate, we can also write the bond price as:
P=
c
c +
+ . . . + c + 100
1 + r1 (1 + r2)2
(1 + rT)T
The spot rates r1, r2, . . . , rT will also price other bonds correctly, therefore
these rates belong to the maturity dates rather than to any particular bond.
How do we find the spot rates from the yields on bonds? The simplest case
is when we have available a large number of zero coupon bonds. A zerocoupon bond has only a single cash flow, the capital repayment in year T, so
in this case the yield-to-maturity must coincide with the T-year spot rate:
P=
100
= 100
(1 + y)T (1 + rT)T
so the T-year spot rate is given by the yield on a T-year zero coupon bond.
By looking, a zero coupon bonds across a wide range of maturities,
therefore, we can unravel the spot rates.
What if we have no zero coupon bonds available? There is no real problem
here; all we need to do is to make one up using coupon bonds. Consider the
following example: suppose we have three bonds available with cash flows
given by the following table:
Bond
Yr 1 cash flow
Yr 2 cash flow
Yr 3 cash flow
105
106
104
If we wish to receive a three-year zero coupon bond with cash flow 100, we
need to hold a portfolio consisting of xA units of A; xB units of B, and 1 xA
xB units of C such that:
105xA + 6xB + 4(1 xA xB) = 0
106xB + 4(1 xA xB) = 0
68
which yields a solution xA = 0.03880 and xB = 0.04074. The year one and
two pay-offs are zero (or close to zero since there is rounding error),
and year three pay-off is (1 + 0.03880 + 0.04074)104 = 112.27. We need to
100
hold 112.27 units of the portfolio above, therefore, to achieve a `synthetic
three-year zero coupon bond with pay-off 100 in year three. In practice,
statistical techniques are applied to bond prices to work out the spot rates
where as many bonds as possible are used. We will not go further into
this issue here apart from mentioning two key points. The first is that
many bonds are included to minimize the risk of estimation error due to
market mispricing, as the reliance of individual bond prices is reduced.
The second modification is done to tackle the problem that the cash flows
of different bonds do not fall on the same day. Therefore, the discount
rates we derive from the shorter bonds may not be directly applicable
when we seek to derive discount rates from longer bonds. The way to deal
with this problem is to fit a smooth curve of discount rates to a large set of
bonds. Individual bonds will determine certain points on this curve, and by
including a large number of bonds the scheme will smooth out the curve
to achieve the best fit. There is a variety of fitting schemes that can be
used for this purpose.
We note also that in the example above we could easily extract the term
structure from the prices of the bonds A, B and C. The price of bond A
would yield the one-year spot rate directly:
PA =
105
1 + r1
Then, using this rate and the price of bond B we can extract the two-year
spot rate:
PB =
106
6
+
1 + r1 (1 + r2)2
and finally, using r1 and r2 and the price of bond C we can extract the
three-year spot rate:
PC =
104
4
4
+
+
(1 + r3)3
1 + r1 (1 + r2)2
which give us the term structure up to year three. For instance if PA = 100
then r1 = 5.00%; if PB = 101 then also r2 = 5.47%; and if PC = 96 then also
r3 = 5.49%.
Once spot rates are extracted, we can work out the forward rates. Suppose
we work out that the one-year spot rate is 5% and the two-year spot rate is
5.2%. Suppose we can borrow 100 in one years time for one year against
paying a rate of 5.3%, i.e. repaying 105.3 in two years time. Should we
take this borrowing opportunity? The way we should look at this is that
100
to offset the loan agreement, all we need to do is to borrow 1.05
today on a
one-year deal, which generates a cash flow of 100 in year one. This cash
flow will be repaid on our loan agreement, so the net outflow in year one
is zero. Next, we invest 100 (1.053)
on a two-year deal, which generates a cash
2
1.052
flow of 100(1.053) in year two. This enables us to repay the loan, so the
net cash flow in year two is also zero. The current cash flow is
100 100 (1.053) = 0.09 which is positive. Therefore, the loan rate of 5.3%
2
1.05
1.052
offered is a good rate, and you would take the loan, offsetting it in the way
described above, and make arbitrage profits, again and again, until the
rate of 5.3% would change. What is the fair level? Let f1 be the one-year
forward rate implied by the loan agreement. Intuitively, the relationship is
1 + f1
1
that 1.05
must be equal to
to avoid arbitrage, or that:
2
1.052
69
23 Investment management
1 + f1 =
1.0522 (1 + r2)2
=
1.05
1 + r1
(1 + rt+1)t+1
(1 + rt)t
(1 + f ) =
(1 + rt+n)t+n
(1 + rt )t
1
n
We notice that all information about forward rates is fully embedded in the
spot rates and vice versa, therefore it does not matter whether we specify the
spot rates or the forward rates.
Duration
An important objective for a bank might be to immunise a bond portfolio
against movements in the term structure of interest rates. The simplest
form of immunisation is against upward or downward shifts in a flat term
structure. Suppose all spot rates are equal rt = r for all maturities t, so that all
bonds are have the same yield to maturity r. We are looking at the problem of
immunising a bond portfolio against movements in the yield.
70
There are several effects in play here. The first is that bond prices and yields
are inversely related as the yield increases the bond price falls. Therefore,
bond portfolios tend to fall in value following an increase in the yield.
Second, an increase in the portfolios yield results in a smaller value change
than a similar decrease in the yield. This phenomenon is called convexity
bond prices are convex in the yield to maturity. This can most easily be seen
by looking at a T-period zero coupon bond. The price is given by:
100
PT =
(1 + r)T
The change in the price corresponding to a change in the yield r is given by
the first derivative:
dPT
= T 100
dr
(1 + r)T +1
dr.. We note that
so that the dollar value changes by dPT = T 100
(1 + r)T + 1
dPT
=
PT.
( )
T
T (1 + r)
PT.
dr
We notice that both the change in absolute (dollar value) terms and
relative (percentage) terms depend on the current yield r, and this
relationship is convex. The convexity property can be confirmed by taking
the second derivative:
d 2PT
= T (T + 1) 100T+2
dr 2
(1 + r)
which is positive. We also note that prices of long-term bonds are more
sensitive to interest rate changes than prices of short-term bonds, since
the value in the numerator depends on T in a positive way, and the value in
the denominator depends on T in a negative way (as it enters in the form
(1 + r)T+1).
The effects of yield changes are very easy to trace when working with
zero coupon bonds. It becomes more diffcult once we allow arbitrary
bonds into our portfolio. The duration concept makes this job easier. The
Macaulays duration concept is computed such as to generate a weighted
average of the maturity dates of the payments of the bond. Consider cash
flows ct received at time t = 1, 2, . . . , T. Compute the present value for each:
ct
PV(ct) =
(l+r)t
Add all cash flows together to compute the price of the bond:
P = PV(ct)
t
Then, work out the contribution of the individual cash flows relative to the
total price of the bonds. This is the duration weight:
PV(ct)
PV(ct)
dt =
=
P
t PV(ct)
Finally, find the weighted average of the maturity dates through the
calculation:
PV(ct)
D = dt t = t
t PV(ct)
t
t
D is the duration of the bond and it is a measure of the weighted average
maturity for the bond payments. If we are working with a zero-coupon
bond of maturity T (where ct = 0 for t = 1, . . . , T 1), we find that:
71
23 Investment management
D = dt t = dTT = T
t
P=
cT
c2
c1
+ (1 + r)2 + . . .+ (1 + r)T + 1
1+r
dP
=
P
c1
cT
c2
. . . T
2
(1 + r)2
(1 + r)T + 1
(1 + r)3
P
dr
and we can now recognise the duration D of the bond inside the bracket:
dP
=
P
( )
The quantity
D
.
as D* =
1+r
D
1+r
dr = D*dr
1+r
Numerical example
Let us look at an example. Consider a five-year 5% annual coupon bond
with yield to maturity 5%. This bond is selling at exactly par value 100
(confirm for yourself), so the weights are for t = 1, . . . , 4: dt =
105/1.055
5
1.05t
100
, and
The figure shows the effects of convexity for a 20-year zero coupon bond.
We notice that for large changes in the yield there is a big discrepancy
between the predicted bond price change and the actual bond price
change. If we were to consider a change in the yield from 5% to 4%, the
predicted price change would be a 4.33% increase in the bond price (since
everything in the formula above stays the same except dr = 0.01 instead
of dr = 0.01) from 100 to 104.33. In this case, the actual price would be
P4% yield = 5 + 5 + . . . + 105 = 104.45
1.045
1.042
1.04
so we are in this case undershooting the true price change. This is due to
the effect of convexity. We can illustrate this geometrically; see Figure 7.1.
Convexity
If the current yield is r0 and the current bond price P(r0) is known, we
know from Taylor approximations that the price of the bond for nearby
yields r is:
P(r) P(r0) + P'(r0)(r r0) + 1 P''(r0)(r r0)2
2
We recap the bond price and work out the derivatives:
P(r0) =
c1
(1 + r0 )
P'(r0) =
P''(r0) =
c1
(1 + r0 )2
2 c1
(1 + r0 )3
cn
c3
c2
+. . .+
+
(1 + r0 )n
(1 + r0 )3
(1 + r0 )2
3 c3 . . . n cn
2 c2
(1 + r0 )n + 1
(1 + r0 )4
(1 + r0 )3
3 2 c2
+
(1 + r0 )4
4 3 c3 + . . . + (n + 1) n cn
(1 + r0 )n + 2
(1 + r0 )5
According to this, therefore, the absolute change in the bond price P(r0) =
P(r) P(r0) is:
P(r0) P'(r0) (r r0) + 1 P''(r0) (r r0)2
2
and the relative change, which is more interesting for our purposes:
P(r0) P'(r0) (r r ) + 1 P''(r0) (r r )2
0
0
P(r0)
P(r0)
2 P(r0)
Here, we recognise the modified duration:
P'(r0)
D* =
P'(r0)
and we define the convexity as:
C* =
1 P''(r0)
2 P(r0)
1
2
2 5 + 3 2 5 + 4 3 5 + 5 4 5 + 6 5 105
1.053 1.054
1.055
1.056
1.057
1
= 11.97
100
Therefore, the relative bond price for a yield change from 5% to 4%, is:
73
23 Investment management
Liabilities
Bond 100
Equity 100
The risk on the asset side from movements in interest rates will feed
through to the equity, so the equity has duration equal to the duration
on the asset side, i.e. 4.55. If the equity holders are unhappy with this
high exposure, they can reduce it by selling (i.e. borrowing) a bond on
the liability side. Suppose the equity holders seek to reduce the duration
by selling a five-year zero coupon bond. Since the company will receive
cash when taking the short position, it will have to decide what to do
with the proceeds. Suppose the company decides to hold the proceeds
in cash (with zero duration as cash values are insensitive to interest rate
movements). Then the balance sheet takes the following form:
Assets
Liabilities
Cash x
Bond 100
Equity 100
Duration equity
74
If we decide on our target duration for our equity exposure (say we wish
to reduce the duration of the equity to 2) then we can find the required
short position x:
100
(4.55 2) = 20(2.55) = 51
x=
5
Liabilities
Bond 100 + x
The duration on the asset side now remains at 4.55 since we hold 100%
in the original bond, and the duration on the liability side is as above.
Balancing them out we find:
4.55 =
x
5 + 100
100 + x
100 + x
Duration equity
or
(100 + x)4.55 = 5x + Duration equity 100
Solving with respect to x we find:
x = 100(4.55 Duration equity)
5 4.55
If our target is still 2, we can solve for x:
x=
100(4.55 2)
= 255
0.55
5 4.55
= 463.6
D* =
1
100
C* =
1 1
2 100
5 + 2 5 + . . . + 5 105
1.052 1.053
1.056
= 4.33
2 5 + 3 2 5 + . . . + 6 5cdot105
= 11.97
1.053 1.054
1.057
For the zero coupon bond we find the modified duration and convexity
equal to:
75
23 Investment management
D*z =
1 10 100
= 9.52
61.39 1.0511
C* =
1
1 11 10 100
= 49.89
2 61.39
1.0512
76
Assets
Liabilities
Equity
Total assets
Total liabilities
Now consider a rebalancing of the asset side where we borrow risk free
at the rate rF and invest a fraction x1 of our wealth in portfolio one and a
fraction x2 in portfolio two. The net loan is x1 + x2. The balance sheet now
looks like:
Assets
Liabilities
Risky portfolios
x1 at return r1
and x2 at return r2
Risky portfolio with
with return r1
Total assets 1 + x1 + x2
Total liabilities 1 + x1 + x2
Liabilities
Equity
Total assets
Total liabilities
23 Investment management
bi1
1 bi1 bi2
and x2 =
bi2
1 bi1 bi2
. A potential problem
h = Cov(S.F)
VarF
to derive the optimal hedge. If you have a spot portfolio of 100,000 and
wish to use the futures contract F to hedge this exposure, you simply sell h
units of the futures for each unit of spot, i.e.
78
This method does not rely on a factor structure or that factor risk can be
traded directly.
The fact that spot and futures rates have moved closely together in the
past does not imply that they will continue to do so in the future. The
famous case of Metallgesellschaft illustrates this point. This German
company went into a crisis in late 1993 because they had lost a lot of
money on oil futures that were designed to hedge their underlying
business that was highly exposed to the oil price through selling refined
oil products to retail customers on very long term fixed contracts. The
main vehicle for hedging this exposure was, however, short oil futures
contracts which were much more liquid than the long term contracts. The
correlation with the long contracts were so high that this was not seen as
a problem. It did become a problem when the short futures prices started
to detach themselves from the historical pattern, and what was set up as a
hedge actually became a speculative position, and the company lost over
$1 billion on its derivatives trading, leading to the firing of the CEO and a
massive debt and asset restructuring to salvage the company.
Summary
This chapter looked at risk immunisation, which deals with specific
trading strategies that can eliminate or reduce all or part of the risk of
portfolios.
The first part dealt with immunisation of interest rate risk of bond
portfolios, using duration (and to some extent convexity).
The second part dealt with immunisation of equity portfolios, including
the use of hedge ratios and futures trading.
Activity
1. Try to find data on bond yields over various maturity dates, for instance
from Bloomberg: www.bloomberg.com/markets/rates/index.html.
What shape best describes the term structure?
23 Investment management
80
x
(1 x)2
1x
Essential reading
Bodie, Z., A. Kane and A.J. Marcus Investments (Boston, Mass.; London:
McGraw-Hill Irwin, 2008) Chapters 24 and 27.
Elton, E.J., M.J. Gruber, S.J. Brown and W.N. Goetzmann Modern Portfolio
Theory and Investment Analysis(New York; Chichester: John Wiley & Sons,
2010) Chapters 25 and 27.
Further reading
Grinblatt, M. and S. Titman Financial Markets and Corporate Strategy. (Boston,
Mass.; London: McGraw-Hill Irwin, 2008) Chapter 22.
Embrechts, P., C. Klppelberg, and T. Mikosch Modelling Extremal Events.
(Berlin: Heidelberg; New York: Springer Verlag, 1997) Note: this book is
very advanced and is not really drawn on in this chapter except for some
initial observations made in the very beginning.
Introduction
This chapter discusses the problem of measuring the risk. For most of
the time we will talk about the risk of a portfolio, which is particularly
relevant in situations where investors delegate their portfolio decisions to
a professional fund manager. Investors have a fundamental choice between
direct investment (the DIY alternative) and delegated investment (the
fund management alternative). Any decision to hand over your money to
a fund manager should be judged on the value added compared to direct
investment. It should be said, however, that the problem of measuring risk
is more general than that, and we will talk briefly about other types of risk
as well.
Even if we restrict ourselves to talking about portfolio risk, the problem
is not as easy as it sounds. In financial markets the expected return on
81
23 Investment management
Types of risk
A fund manager can be exposed to a whole range of different types of
risk, and obviously we have to make a choice which types of risk we aim
to discuss here. Often we separate risk into five categories, although these
are not entirely mutually exclusive:
market risk: risk of unexpected changes to asset prices or rates such
as exchange rates and interest rates
credit risk: risk of changes in value that comes from unexpected
changes in credit quality of trading counterparties
liquidity risk: risk that the cost of adjusting a portfolio will
increase unexpectedly, or that access to credit becomes unexpectedly
significantly more costly
operational risk: the risk of fraud or failures in operations due to
systems breaking down or human errors being made
systemic risk: the risk of meltdown in the financial system, for
instance caused by a chain reaction of events causing liquidity crises or
defaults.
82
In this chapter we focus primarily on market risk that is the risk that the
asset values of the fund managers portfolio will change unexpectedly. For
financial firms it has traditionally been the case that risk management has
essentially been identical to the management of market risk. For industrial
firms, in contrast, the traditional role of risk management has been to
manage operational risk and also, to some extent, liquidity and credit
risk (particularly ensuring that creditors are able to pay on time, and that
credit lines or working capital is available when necessary). In recent
years we have witnessed a convergence of risk management practices
where financial firms are increasingly aware of credit risk, liquidity risk
and operational risk, and where industrial firms are increasingly aware of
market risk. Here, we restrict our discussion mainly to the management of
market risk. Market risk is measured in terms of investment returns (see
Appendix 1), so all discussions about risk management will deal with the
management of the variations of investment returns. We discuss various
methods of measuring this variability in the following pages.
Risk decomposition
When interested in measuring the exposure of investment returns to
risk we are keen to work with measures that are relevant and have clear
meaning. Unfortunately, there are many measures available each are
relevant in their own way and have their own meaning but there exists
no all-encompassing risk measure that generalises all the other. One of
the most intuitive and flexible ways of measuring risk is by decomposing
the risk of an investment portfolio into risk that is correlated with
outsider factors (such as the market index) and risk that is idiosyncratic.
A convenient way to do such decomposition is by regression methods
(see Appendix 1 for a review of such methods). Suppose the return on
a portfolio depends on a number of known factors, for example the $/
exchange rate and the return on the FTSE 100 stock market index. The
regression model where we regress the return of a portfolio r on the return
on the market index rM of the type:
r = a + brM + e
provides estimates of the coeffcients that can be useful measures of the
risk of the portfolio. In particular, this regression model will decompose
the risk of the portfolio into market risk with variance Var(brM) = b2Var(rM)
and idiosyncratic risk Var(e). If we use factor portfolios instead of the
market index, we get a regression model of the type:
r = a + b1 f1 + b2 f1 + . . . + bk fk + e
which allows a similar decomposition, only in this case we get:
Market risk = Var(b1f1 + . . . + bk fk)
and
Idiosyncratic risk = Var(e)
The residual risk e will be uncorrelated with all of the factors f1, . . . , fk
but the factors may be correlated so it is not necessarily the case that
we can decompose the market risk into its individual components as the
covariance terms may not vanish completely.
Value-at-Risk
The Value-at-Risk (VaR) method for measuring risk exposure has become
one of the most popular, particularly as a measure of risk in corporations
and financial institutions. The VaR measure is defined as the worst loss
83
23 Investment management
1.28
1.65
1.96
2.33
x
Figure 8.1
measures that adjusts the performance of a given portfolio for the risk
it takes. The first is the Sharpe measure which is the ratio of the excess
return to the standard deviation of the portfolio:
Sharpe ratio = Er rF
where Er is the expected return of the portfolio, rF is the risk free return,
and is the standard deviation of the portfolio. The Sharpe ratio for a
portfolio needs to be measured against the Sharpe ratio of the market
index, which forms a benchmark:
Sharpe ratio for the market = ErM rF
M
This comparison tells us whether the portfolio lies above, on, or below, the
capital market line. The capital market line is illustrated in the following
figure:
rF
Figure 8.2
Any arbitrary capital market line portfolio has, therefore, the same Sharpe
ratio.
Treynors ratio
The second is Treynors measure which measures the expected excess
return on the portfolio relative to the beta risk of the portfolio.
E r
Treynor ratio = r F
85
23 Investment management
and this measure should also be measured against the market index
(which has unit beta):
Treynor ratio for the market =
ErM rF
1
= ErM rF
It is easy to see that Treynors ratio is really a test of whether the asset is
on the security market line. The CAPM model predicts that assets should
be priced according to the formula:
Er = rF + (ErM rF)
which gives us the security market line. By rearranging, we find:
Treynor for asset =
Er rF
rF
Figure 8.3
The graph shows the security market line in the plane. The security
market line has a slope equal to ErM rF.
If the asset has a Treynor ratio that is greater than the market, then it is
located above the security market line, and vice versa. Assets with Treynor
greater than the market are considered good buys, but they may carry
idiosyncratic risk for which the investor is not compensated.
x= M
The mixed portfolio has now the same risk as the market portfolio, and we
can compare the return of the mixed portfolio with the market portfolio.
The difference is the M2 measure:
M2 = 1
) r +
Er ErM
Information ratio = P
Sharpe vs Treynor
What is the connection between the Sharpe ratio and the Treynor ratio?
To see this, we note that in order to compare two portfolios with different
risk characteristics it is important that the risk premium is linear in the
risk measure. We know that this is true for the Sharpe ratio if we use
standard deviation of total risk as the risk measure. We shall now see that
this is true for the Treynor ratio if we use beta risk (i.e. the market risk, or
systematic risk, component of total risk) as the relevant risk measure, and
if the investor already holds a large diversified portfolio with the same risk
characteristics of the market portfolio.
Consider an investor who holds a large portfolio P with return rP and is
considering making a small investment m in a new portfolio Q with return
rQ, by borrowing at the risk free rate rF . His new position is:
r = rP + m(rQ rF)
The marginal expected return from this operation is:
lim ErP + m(ErQ rF) ErP = dErP + m(ErQ rF) = ErQ rF
m
dm
m0
23 Investment management
2
M
1 (Er r )
M
F
2M2
Changing risk
Most funds change their portfolios significantly, perhaps as often as once a
year, which means that the process of measuring the long-run performance
is complicated further by the discontinuities in the asset allocation
decisions of the fund. In effect, a significant rebalancing of the asset
allocation decisions will change the corresponding return distributions
also, and consequently the performance measures discussed above may
give misleading results.
Consider the following example. Suppose the Sharpe ratio of the market
index is 0.4. A fund manager follows a low risk strategy for his fund
over the first year, where he takes an annual expected excess return of
1% against a standard deviation of 2%. This yields a Sharpe ratio of
0.5, which beats the market. Over the following year, the fund manager
switches to a high risk strategy where he takes an annual expected excess
return of 9% against a standard deviation of 18%. The Sharpe ratio is still
88
0.5, so he still beats the market. Suppose we break the returns down into
quarterly returns, where the fund earns (in annualised returns) 1%, 3%,
1%, and 3% in the first year. This is consistent with an average return
of 1% and a standard deviation of 2%. The following four quarters the
numbers are 9%, 27%, 9% and 27% (again consistent with a mean of
9% and a standard deviation of 18%). If we now take the average return
and standard deviation of the full two-year period, ignoring the structural
break caused by the switch in investment strategy, we find that the fund
earns an average return of 5% against a standard deviation of 13.42%,
which yields a Sharpe ratio of 0.37. This looks inferior to the market
index. The bias is caused by the fact that we are calculating our Sharpe
ratio on the basis of aggregating both low-risk and high-risk periods. Since
the aggregate variance is going to be biased towards the high risk period
(just like outliers get a disproportionate weight when working out the
sample variance), so the aggregate variance is too high compared to the
aggregate return.
Market timing
Market timing is the practice of switching between safe and risky portfolios
at different points in time. For instance, a fund manager might decide
to switch a large portion of his capital into safe government bonds if he
thinks the stock market on the whole is overvalued and the risk of a stock
market correction is imminent. Similarly, he might switch back again when
he thinks the market is undervalued. If market timing is significant, we
should expect that the market risk of the fund is greater in periods where
the excess return of the market is high, than in periods where the excess
return of the market is low. This suggests a convex relationship between
the excess return on the fund and the excess return on the market. For
a fixed portfolio, the CAPM predicts that this relationship is linear (and
given by the beta of the portfolio). To identify market timing, therefore,
we can use a regression based methodology where we regress the excess
return of the fund on the excess return of the market plus a second
variable which tends to be high when the excess return of the market is
also high. There are two obvious candidates of this regression.
The first looks at measuring market timing for a portfolio P:
(rP rF) = aP + bP (rM rF) + cP (rM rF)2 + eP
The model regresses the excess return of the portfolio (the left hand side)
on the excess return of the market (the right hand side). Thus far, this is
the standard CAPM model. The regression model then includes a second
variable which is the squared of the excess return of the market. This
variable is never negative, and tends to be high when the excess return of
the market is also high and positive, and low when the excess return of the
market is low and positive. If the coefficient cP in this regression turns out
significant and positive, then that is indicative that the fund manager is
successfully implementing a market timing strategy as he takes advantage
of positive market movements and does not suffer negative returns for
negative market movements.
The second regression model measures market timing for a portfolio Q:
(rQ rF) = aQ + bQ(rM rF) + cQ(rM rF)D + eQ
This model is structured in exactly the same way as the first, only that the
second variable on the right hand side is equal to the excess return of the
market multiplied by a dummy variable which is zero when the excess
return of the market is negative, and one otherwise. This specification
89
23 Investment management
takes care of the cases where the excess return of the market is negative by
loading weight onto the constant term. A significantly positive regression
coefficient cQ is also here indicative of the implementation of a successful
market timing strategy.
Summary
This chapter dealt primarily with recognising and classifying risk
in financial investments, and how to take risk into account when
measuring the performance of the investment.
There was a brief outline of the VaR risk measure.
The bulk of the chapter looked at the various versions of the Sharpe
ratio and the Treynor ratio. The relationship between these two
measures was discussed, as well as problems associated with using
these measures in situations with changing risk.
Finally, the chapter discussed various aspects of market timing.
Activity
1. Try to measure the Sharpe ratio of a portfolio which contains
asymmetric risk. A way to do this is a portfolio where you top up your
investment by selling deep out-of-the money options. For instance, you
start with a capital of 100 which you invest in the index. You sell put
options and invests the proceeds risk free. At the end of the period you
will most likely have both your original investments in the index plus
the proceeds from the sale of the puts - i.e. you have boosted your risky
investment which is likely to boost the Sharpe ratio of your portfolio.
Here is how you can simulate the results:
start with simulating the index return this can be done by a
binomial approach for instance, assume the index increases by
20% or decreases by 10% per four months, which implies that
over the year there is a 18 probability it increases by 72.8%; a 38
probability it increases by 29.6%; a 38 probability it decreases by
2.8%; and a 18 probability it decreases by 27.1%. These numbers can
be simulated by the RAND() function in Microsoft Excel a draw
between 0.0000 and 0.1250 represents a decrease of 27.1%; a draw
between 0.1251 and 0.5000 represents a decrease of 2.8%; a draw
between 0.5001 and 0.8750 represents an increase of 29.6%; and
finally a draw between 0.8751 and 1.0000 represents an increase of
72.8%.
next simulate the option prices. Suppose you pick an initial index
value of 100, and work out the value of a one-year put option with
exercise price 80 (the subject guide for course 92 Corporate
finance tells you the details of how this is done). If the risk free
return per four months is 1%, the option price is:
3
1 19
(80 72.9) = 1.75
p(80) =
1.01 30
i.e. for each 100 you invest in the index you can invest an additional
1.75 risk free.
simulate your portfolio returns for each draw of the index you
can work out the pay-off of your index investment, the pay-off of
your put position (remember you sell the put), and the pay-off of
the risk free part of your investment. Each period you recalibrate
your investment to 100 and work out the return, and eventually
90
you will have a return time series of which you can carry out a
statistical analysis. Finally, work out the Sharpe ratio of the index,
and the Sharpe ratio of your portfolio. Which is greater? Did you
have outcomes where your put-investment came up with a negative
pay-off (there is only a one-in-eight chance this happens in any one
year)?
91
23 Investment management
Notes
92
Essential reading
Bodie, Z., A. Kane and A.J. Marcus Investments (Boston, Mass.; London:
McGraw-Hill Irwin, 2008) Chapter 27.
Elton, E.J., M.J. Gruber, S.J. Brown and W.N. Goetzmann Modern Portfolio
Theory and Investment Analysis(New York; Chichester: John Wiley & Sons,
2010) Chapter 27.
Further reading
Grinblatt, M. and S. Titman Financial Markets and Corporate Strategy. (Boston,
Mass.; London: McGraw-Hill Irwin, 2008) Chapters 21 and 22.
Duffe, D. and K.J. Singleton Credit Risk: Pricing, Measurement and Management.
(Princeton, NJ: Princeton University Press: 2003) Chapter 1.
Stulz, R. Risk Management & Derivatives. (Mason, Ohio: Thomson SouthWestern, 2003) Chapters 2, 3 and 4.
Introduction
We buy accident insurance because losses associated with accidents may
be detrimental to our financial situation. The same argument goes for
portfolio insurance. Portfolio managers care much more about losses to
the value of their portfolios than they care about corresponding gains.
Sometimes, therefore, they choose trading strategies that offer an implicit
insurance effect against losses whilst keeping the potential for making
portfolio gains (as opposed to portfolio immunisation which protects
against both upside and downside potential). This chapter looks at
strategies that offer such insurance effects.
What is the economic rationale for insuring against losses? Utility theory
(see Appendix 1) argues that when individuals are risk averse they are
willing to pay to avoid the risk. If insurance is available from individuals
who are less risk averse, it is possible to transfer the risk away from the
individuals who are more risk averse to those who are less risk averse,
93
23 Investment management
23 Investment management
1m in one years time. You can buy a one-year put option that offers
perfect protection to your portfolio losses below 1m if you buy 1m/X
units of the put option above with T = 1 year and X equal to the current
index level. If the current index level is 4800, we set X = 4800, and the
portfolio pay-off equals either 1m (if the put option is in the money and
X > S) or the value of the unprotected portfolio which is (S=X)m (if
the put option is out of the money and S > X), whichever is greater. The
portfolio earns, therefore, either the percentage increase in the index, if
positive, or guarantees a pay-off of the initial value of 1m. The initial
investment cost of the protected portfolio is, however, more than 1m
since the put option is costly to acquire at the outset.
d ln V = 1 2 dt + dZ
2
The increments of the log-value ln VT ln Vt are then normally distributed:
ln VT ln Vt N
((
1 2 (T t), 2(T t)
2
From the above it follows that a 1%, 30-day VaR is given by the solution y* to:
N x
(y
1 2 30
365
2
30
365
= 0.01
20
252
The motivation for this is that there is much less volatility during nontrading days than during trading days. Also, we still use the correct days in
the numerator as the difference in expected return between trading days
and non-trading days is negligible.
Consider the following data: = 0.10 (expected return is 10%
continuously compounded roughly 10.5% annually compounded); =
0.3 (volatility or standard deviation is 30%), which yields (using 30=365
in the numerator and 22=252 in the denominator):
N x
96
y 0.00452
0.0845
= 0.01
y* 0.0192
23 Investment management
how we can replicate this using call options. Suppose we hold a portfolio
which is currently worth $100,000 and we are interested in buying a put
option which guarantees that the value of our portfolio is at least equal
to $100,000 also in one years time. The cost of this is a put option with
exercise price $100,000, which gives us a net pay-off of:
pay-off in one years time = X + (100,000 X) = 100,000 if X < 100,000
X
if X 100,000
Non-linear pay-offs
Some portfolios contain assets with non-linear risk, normally derivative
securities, which creates particular problems for risk management. To
illustrate the problem with non-linear risk, we shall consider an example
of the risk of a call option on a stock. Suppose the stock has constant risk
where it either increases by 40% or decreases by 10% each year. If the
current stock price is 100, next years price is either 140 or 90, and year
two price is either 196, 126 or 81. Assume the up-movements and the
down-movements are equally likely. The expected return on the stock is
always 12 (40% 10%) = 15%, and the variance of the stock is
1
(0.4)2 + 12 (0.1)2 0.152 = 6.25%. This is true regardless of the price
2
of the stock. Suppose the risk free rate is 5% and the market index has
expected return 15%, implying the stock has a beta of 1.
Consider a two-year call option on the stock with exercise price 100.
This option can be priced by the Cox-Ross-Rubinstein binomial pricing
model, where we can make use of the risk-neutral probabilities. These
probabilities take the value
q = 1.05 0.9 = 3
1.4 0.9
10
for up-movements and
1q= 7
10
for down-movements. We can verify that these probabilities are correct
by pricing the stock. The current price of 100 should be year ones
expected price discounted by the risk free rate if we use the risk neutral
probabilities:
100 =
1
1.05
7
3
90
140 +
10
10
1
1.052
(( )
3
10
(196 100) + 2
( )( )
3
10
())
7
7
(126 100) +
10
10
= 17.74
CU1 =
1
1.05
3
7
(126 100)
(196 100) +
10
10
= 44.76
1
1.05
3
7
0 = 7.43
(126 100) +
10
10
If we want to work out the expected return on the call, we find that the
expected return currently is:
1 44.76 + 1 7.43 1 = 47.09%
2 17.74 2 17.74
the expected return after one up-movement is:
1 96 + 1 26 1 = 36.28%
2 44.76 2 44.76
and finally, the expected return after one down-movement is:
1 26 + 1 0 1 = 75.00%
2 7.43 2
We notice that these numbers change all the time. The expected return on
the call tends to increase if the stock price decreases, and decrease if the
stock price increases. We can use the expected return to calculate the beta
of the call option. Currently, the beta of the call is given by:
C0 = 47.09 5 = 4.2
15 5
If the stock price increases in the following period, the beta of the call
becomes:
36.28 5 = 3.1
CU
1 =
10
and if the stock price decreases, the beta becomes:
75 5 = 7.0
CD
1 =
15 5
What is common in situations like these is to work out the current hedge
ratio and use this to hedge the exposure over a short time interval i.e. to
estimate the local beta of the asset. In the example above, we should start
out hedging the call as a 4.2 beta asset, then be prepared to rebalance
our hedge to a 3.1 beta hedge or a 7.0 beta hedge depending on the
movements of the underlying stock. Only if there is little variation in
hedge ratios over time can you increase the duration of the hedge.
Extreme risk
There are certain situation in which we really only care about extreme
events and we are happy to ignore the more regular variability. An
example is flood-defences. The regular variability in water-levels caused
by tidal flows is not really of interest to us, what we want to know is the
probability of extreme floods caused by high waves or storm surges. In this
case we often need to use a special statistical method to make inferences
as by the nature of the problem almost all observations are within the
regular but uninteresting range. This applies also to the world of finance.
The risk we are exposed to in normal times tend to be different from the
risks we are exposed to in extreme situations, and often it can be hard
to estimate the behaviour of the probability distribution in these extreme
circumstances.
99
23 Investment management
pT
(1 + r)T t
It
i.e. the price of a claim with an expected cash flow (at time t) of its
future cash flow (at time T), Et pT , discounted by some discount rate r. In
general, the discount rate r is risk adjusted if we take expectations with
respect to actual probabilities, but is equal to the risk free rate if we take
expectations with respect to the so-called risk-neutral probabilities. A
payment obligation cT has actual cash flow ~c T at time T, implying that we
should write the current value ct = Et (~c T (1 + r)Tt It), where the expected
cash flow typically is less than the actual payment obligation, i.e. Et ~c T cT.
This is, however, not standard. Instead we write the pricing formula:
cT
ct =
(1 + y)T t
where we do not work out the expected cash flow Et ~c T at all but use the
contractual payment obligation. This necessitates, however, that we make
adjustments to the discount rate, which now typically is not equal to risk
adjusted (buy-and-hold) discount rates nor equal to the risk free rate, but
is equal to the yield-to-maturity which is the risk free rate plus a premium:
y = rF + credit spread
100
where the premium represents the credit spread. If there is no chance the
debtor will default on the payment obligation, the credit spread is zero,
as in this particular case Et ~c T= cT and the risk adjustment to the discount
rate is zero. This is, however, a very special case that only applies to
government issued payment obligations (government debt).
Investors are typically interested in hedging one of two things related
to credit risk. First, they want to hedge against a shortfall in cash flow
linked to the failure of receiving the full amount of the contractual
payment obligation. This can be achieved by buying so-called credit
default swap instruments triggered by the occurrence of credit events.
Second, they want to hedge against sudden changes (typically increases)
in the credit spreads of debt obligations that are not yet due. This can be
achieved by buying credit-spread based derivatives, which give the
holder the right or the obligations to make certain transactions triggered
by movements in the credit-spreads of the underlying asset or a reference
asset. The trading of such instruments makes it feasible to hedge a whole
new class of risks in addition to the usual risks linked to the movements in
asset prices.
Hedging volatility
Often, risk management strategies involve trading derivative securities
such as options, as we have seen above. We know from the pricing
literature on options that their prices are sensitive not only to the price
of the underlying asset (the asset whose price ultimately determines the
pay-off of the option), but also to the volatility of the underlying asset.
Consult, for instance, the subject guide for 92 Corporate finance for
an explanation of this. Active option trading strategies can, therefore,
be constructed to be equivalent to buying and selling volatility of the
underlying asset.
We illustrate this with an example using the Black-Scholes option pricing
formulas. Recall that the Black-Scholes call option price can be written as:
c = SN (d1) Xe r (T t) N (d2)
where S is the current stock price; X is the exercise price; T t is time to
maturity, and r is the risk free rate (continuously compounded). The two
parameters d1 and d2 are given by:
d1 =
ln( XS
) + (r + 12 2 T t)
T t
d2 = d1 T t
By put-call parity, the put price is:
p = c + Xe r (T t) S
= SN(d1) Xe r (T t) N(d2) + Xe r (T t) S
= Xe r (T t) (1 N(d2)) S(1N(d1))
= Xe r (T t) N(d2) SN(d1)
We wish to purchase x calls and y puts, so that our portfolio consists of:
Option portfolio = xc + yp
= S(xN (d1) yN (d1))
Xe r (T t) (xN (d2) yN (d2))
101
23 Investment management
dOption portfolio
= xN (d1) yN (d1)
dS
1
60
i=1
+ SRt
where VaR(1%; 10 days) is the 1%, 10-day value at risk for day t, the
number St is a multiplier, and the charge SRt is the additional capital charge
for idiosyncratic risk. We notice that the average VaR over the 60-day
period is used, and the multiplier St is used to compensate for the accuracy
of the banks VaR model. This multiplier should be worked out on the basis
of past data. If the bank consistently gets it wrong it will normally show up
by some backtesting routine, and adjustments can be made to the banks
capital charge by making adjustments to this multiplier.
Summary
This chapter studied risk management, from the perspective of
investors, corporations and banks or financial firms.
Some risk management strategies were outlined and contrasted, such
as, for instance, put option protection and value at risk.
Next, some problem areas were highlighted, such as the problem of
portfolios with non-linear risk, extreme risk and credit risk.
The chapter also looked at a simple strategy for hedging changes in
volatility (as opposed to changes in market values).
Finally, there was a brief outline of the capital allocation mechanism
common for the bank regulatory framework outlined in the Basel
accord.
102
Activity
1. Suppose you have a 10% after tax discount rate, and you have a profit stream of 1, 0,
3, and 2 over the next three years. The tax rate is 30% and you can accumulate losses
to offset next years profits against taxes. Work out the value of a risk management
programme where you can transfer profits over time at the discount rate (assume, for
instance, that you can create a maximum loss (profit) of 1 today by creating an offsetting
profit (loss) next year of 1.1 these transfers are priced by the 10% discount rate so will
be zero NPV projects on a before tax basis however, you should be able to save on the
total tax bill by carrying out such transfers cleverly).
2. Why might a straddle trading strategy ahead of announcements (such as scheduled
earnings announcements) not yield long-run abnormal average profits? Explain.
3. Discuss whether VaR based risk management might be preferable to standard portfolio
insurance.
Return
1%
-15%
4%
-5%
5%
0%
20%
5%
40%
10%
20%
15 %
5%
20%
4%
25%
1%
30%
103
23 Investment management
What is the expected return on this portfolio? Suppose you can hold risk
capital at a risk free return of 4%. How much capital should you hold if
your target is to lose at most 5% of your portfolio value with probability
99%?
104
Investment returns
To work out investment returns, we review an important formula from
corporate finance the internal rate of return (IRR) formula for a
T-period investment that costs I0 and has cash flows c1, c2, ..., cT:
cT
c2
c1
...
0 = I0 +
1 + IRR + (1 + IRR)2 + + (1 + IRR)T
For one-period investments the application of this formula is particularly
simple:
IRR =
c1
1
I0
23 Investment management
Year
Annual return
log return
10%
9.53%
-4.55%
-4.65%
-10%
-10.54%
5.56%
5.41%
The initial portfolio is split 50:50 between the two assets, so the annual
return on the part invested in A is exactly offset by the part invested in B,
so the total return over year 1 is (50%)10% + 50%(10%) = 0. At the
end of the first year, the amount invested in A is 55,000, so the split now
is 55:45. The return for year two is, therefore (55%)(4.55%) + (45%)
(5.56%) = 0. We have here made use of the annual returns.
When it comes to the assets return over the entire two-year period we find
that the log returns are more accurate. The log return of asset A over the
entire two-year period is given by:
Log return A = 9.53% 4.65% = 4.88%
so the 50,000 invested in A has grown to 50,000e0.0488 = 52,500.
Similarly, the log return of asset B over the entire period is:
Log return B = 10.54% + 5.41% = 5.12%
so the 50,000 invested in B has grown to 50,000e 0.0512 = 47,500.
The holding in A and B aggregate to 100,000.
(1+rt )
t=1
1
N
N t=1 t
106
These averages do not give the same result. To see this, consider a stock
that has a starting price of 100, then goes up to 110, and then down
again to 99. The return in the first period is 10%, and the return in the
second period is 10%. The arithmetic average is, therefore, equal to 0. The
geometric average is, in contrast, less than zero since:
1
((1.10)(0.90))2 1 = 0.005%
The arithmetic average is a useful measure for investments that have the
same starting balance in every period, whereas the geometric average is a
useful measure for an investment that carries its starting balance in every
period over from the previous period.
Taylor approximation
In many instances we might be able to work out the price of a bond for
a given yield accurately; however, realising that the yield may change
we need to have a general formula for the bond price in `nearby
yields as well. To do this we can use a mathematical tool called Taylor
approximation. For a function f (x), if the value f (x0) is known we can
approximate the value at x close to x0 with infinite accuracy by:
f (x) f (x0) + f ' (x0) (x x0) + 1 f ''(x0) (x x0)2 + 1 f '''(x0) (x x0)3
2
6
+ ... + 1 f (n)(x0) (x x0)n + ...
n!
where f ', f '', f ''' and f (n) are the first, the second, the third and the nth
1
1 is one over n factorial, or
derivative of f; and where n!
123 n
Since the distance x x0 is small, the term (x x0)n will quickly vanish as n
increases, therefore, we can ignore the higher order terms. It is common
to end at n = 1 or n = 2. For bonds, for instance, we work out bond prices
using duration only (where we stop at n = 1) and sometimes for increased
accuracy using duration and convexity (where we stop at n = 2).
Optimisation
In economics we frequently need to work out problems where we seek to
maximise or minimise something, and sometimes we need to maximise
or minimise something subject to not violating some constraint (e.g.
maximise the expected utility from our investment strategy, subject to not
investing more money than we have available). Optimisation is an area
that is highly mathematical, where we can make use of the full power of
calculus. Here we will review the basics of optimisation (without going
into too much detail).
The objective function is what we want to optimise, f(x) if we have one
variable x we wish to adjust to make f(x) as large or small as possible, and
f(x1 x2,..., xn) if we wish to simultaneously adjust n variables x1,..., xn to make
f() as large or small as possible. The fundamental method is the same:
we are at an optimum point (maximum or minimum) if we are moving
sideways only for small changes to x or x1,..., xn. This is where we use
calculus: going sideways is equivalent to putting the derivative to zero. In
the case of maximising or minimising f(x), therefore, it is necessary that:
f '(x) = 0
In the case of maximising or minimising f(x1, ..., xn), it is necessary that:
f
f
f
=
==
=0
x1
x2
xn
107
23 Investment management
Regression methods
When making observations of two random variables, we are often
interested in measuring the relationship between them. Simple ways of
doing this constitutes correlation or covariance measures, which measures
the degree of relatedness in the variability of the two random variables.
We have used these terms in several places within this guide and technical
definitions can be found towards the end of Bodie, Kane and Marcus in an
appendix. Regression methods can best be understood as a more advanced
and powerful method of measuring the relationship between two random
variables than is possible through correlation or covariance. Regression
models attempt to pick up linear relationship between two random
variables of the type:
yt = a + bxt + et
Here, denotes a series of observations of a (dependent) random variable
yt over time and a series of corresponding observations of another
(independent) random variable xt. The regression coeffcients a and b are to
be determined by the regression analysis, and the error term et measures
the deviations between the actual value of yt and the predicted value a
+ bxt. The crucial question is how the coeffcients a and b are determined.
There are two criteria the regression analysis uses to determine these
coeffcients. The first is that the error term is on average zero. The second
is that the sum of the squared errors te2t is minimised. These two criteria
determine uniquely the coeffcients a and b. The coefficient b is given by
the formula:
Cov (yt , xt)
b=
Var (xt)
We can extend the regression model to multiple independent random
variables that yield the multiple regression model:
yt = a + b1x1t + b2x2t + + bkxkt + et
where the coeffcients a, b1,..., bk are to be determined.
108
Utility theory
Utility theory is a concept used in economics to model human behaviour
using mathematical functions called utility functions. These functions can
be defined over many goods and services, but when talking about financial
investments we define these functions over money. If an individual,
through some investment choice, ends up with a final cash balance of w,
we say his utility is u(w), where u is the utility function. If w is a random
variable, we measure his utility by the expected utility, which in general
is not equal to the utility of his expected cash balance. In fact, if the
individual is risk averse the expected utility is always lower than the utility
of the expected cash balance:
E(u(w)) < u(E(w))
Where do utility functions come from? Utility functions are not an
inherent characteristic of human beings; they should rather be thought of
as a representation of preferences over outcomes. Why do we use utility
functions? The simple answer is that utility functions are a lot easier to
handle than preferences. What utility theory does, essentially, is to show
that when our preferences are suffciently structured (or, perhaps more
accurately: rational) we can represent these by a utility function. An
example of the type of structure we impose on rational preferences is the
so-called transitivity property. If an individual prefers A to B and also B to
C, it follows that he also prefers A to C. This property is fairly obvious, but
sometimes the structure we impose is more subtle (and controversial), and
there is evidence that what we assume about preferences in utility theory
may not be empirically true.
ARA =
u''(w)
u'(w)
which is minus the second derivative over the first derivative. If the utility
function is linear, the risk aversion coefficient is zero, indicating risk
neutral preferences.
The relative risk aversion coefficient is defined by:
RRA =
u''(w)
w = ARA w
u'(w)
109
23 Investment management
Whereas the absolute risk aversion coefficient tells us how much risk
we are prepared to bear in absolute terms, the relative risk aversion
coefficient tells us how much risk we are prepared to bear in relative terms
(relative to wealth). Therefore, if we experience a doubling of our wealth,
we will still bear the same amount of risk if we have a constant absolute
risk aversion coeffcient, but we will double the amount of risk if we have
constant relative risk aversion coeffcient.
v
2
E(w) = m
E(r) = r0+
x (r r )
i
i =1
If we collect the expected returns in a vector r = (r1, ..., rn) and the
portfolio weights in another vector x = (x1,..., xn), we can write the
expected return as:
E(r) = r0 + x(r r01)
where 1 = (1, 1, ,,,, 1). Suppose the variance of each asset i is ii and the
covariance between any two assets i and j is ij. Let us collect all variancecovariances in:
11 12 . . . 1n
...
2n
...
..
.
...
n1 n2
...
nn
...
110
21 22
r0
(r r01) 1(r r01)
The right hand side contains only known quantities, the (r r01) vector is
just the vector of excess return for the risky assets, and the quantity 1 is
the inverse of the variance-covariance matrix. Therefore, we are able to
work out the optimal portfolio:
r0
1(r r01)
x* =
(r r01) 1 (r r01)
The variance of x* is:
2 = x* x*
=
x* (1 (r r01))
r0
(r r01) 1 (r r01)
( r0)2
(r r01) 1 (r r01)
so, if we take the square root on both sides to find the expected return as a
function of standard deviation, we find:
= r0 (r r01) 1 (r r01)
which indicates a linear relationship between the standard deviation and
expected return on portfolios along the portfolio frontier.
The tangency portfolio can be found at the point where x0 = 0, or ni =1 xi = 1,
which implies a Lagrange multiplier such that:
1x* = T1 1(r r01) = 1
so the tangency portfolio is:
x*T = T1 (r r01) =
1
1 (r r01)
1 1 (r r01)
T2 and T
T2 = xT* x*
= T( T r0 )
111
23 Investment management
r r01 = xT*
T r0
T2
Row i of this equation states that the expected return less the risk free rate
on the left hand side, ri r0, is equal to the covariance of asset i with the
tangency portfolio divided by the variance of the tangency portfolio times
the expected return of the tangency portfolio minus the risk free rate:
ri r0 =
Cov(ri , rT )
(T r0)
T2
1
e
f (x) =
2
x2
2
N(t) =
1
e
2
x2
2
dx
but unfortunately there exists no exact function for this integral. Normally,
we use tables or some approximation algorithm to get a numerical
expression for the function N(t).
A simple versatile algorithm is:
1
N
where p(t) = t(1.5976 + 0.070566t2)
Simple(t) =
1 + e p(t)
and a more complicated one is:
N
1
e
2
1
z=
1 + pt
Complicated(t) = 1
t2
2
p = 0.2316419
b1 = 0.319381530
b2 = 0.356563782
b3 = 1.781477937
b4 = 1.821255978
b5 = 1.330274429
Whereas NSimple(t) can be used for any value of t over the entire range
< t < , the function NComplicated(t) works only for the positive range
112
E (y)
=
=0
and
Var (x) =
1
2
=1
2 Var (y) =
= Pr x
[
[
= Pr
Pr [y t] = N
113
23 Investment management
Figure 10.1
This graph shows the pay-off of a European call prior to maturity against
the value of exercise max(S X, 0). Figure 10.1 shows that it will never be
optimal to exercise an American call early if the underlying stock does not
pay dividends. The value of the European call is always higher than the
value you will get through exercise. When the stock pays dividends this
may change, however, as a dividend payment leads to a jump in the stock
price at the ex-dividend day. For American calls deep in the money, the
value of exercising the option prior to the jump may exceed the value of the
European call after the jump.
We now turn to puts, where the picture is different. For puts deep in the
money, the European put value is actually lower than the value you get
through exercise. The reason for this is that the American put option has
a maximum value X which is reached for S = 0. If the stock price goes
low enough, therefore, it is impossible to earn more from the put option.
Therefore, if S goes suffciently low, it is optimal to exercise the American
put early. The American put is, therefore, worth more than the European
put, and put-call parity will not hold for American options. This is
illustrated in the following figure.
p
Figure 10.2
This graph shows the pay-off of a European put prior to maturity against
the value of exercise max(X S, 0).
114
23 Investment management
3. a. What are the various types of risk? How does our perspective on
risk management change with whether we are individual investors
or corporations?
b. There are three bonds available in the market, the data are given in
the table below.
Bond
Price
100
105
99.6
4.75
104.75
101
5.25
5.25
105.25
Work out the yield to maturity, the spot rates, and the one-year
forward rates, using the bond data in the table.
c. Suppose you are given the following option prices on a stock that is
currently trading at a price of 100 per share.
Exercise price
90
100
110
Call price
14
10
Put price
12
What does your optimal trading strategy look like in this case?
8. a. What do we understand by convexity in the context of bond prices?
How can we make use of convexity when estimating bond price
changes following yield shifts?
b. A portfolio has expected return 12%, total variance 16% and
beta 0.8. The market portfolio has expected return 10% and total
variance 9%, and the risk free rate of return is 4%. What is the
Sharpe-ratio, the Treynor ratio, the M2 measure and the Jensens
alpha of this portfolio? If you are to advise investors about the
attractiveness of this portfolio, what would you advice be? Explain
whether you would make use of the Treynor-Black model in your
advice.
c. We have witnessed several corporate failures caused by securities
trading based on sound risk management or arbitrage arguments,
and the subject guide briefly mentions two of these cases: the case
of hedge funds taking positions in Volkswagen in early 2008 and
the crisis of Metallgesellschaft in 1993 related to hedge positions in
oil futures. Explain why, in both cases, the trading positions can be
described as sound, and also explain what when wrong.
END OF PAPER
117
23 Investment management
Notes
118
(1.10)(0.9) 1 = 0.5%
The arithmetic average is a useful measure for investments that have the
same starting balance in every period, whereas the geometric average is
a useful measure for investments that carries its starting balance in every
period over from the previous period.
Investment returns
For one-period investments the definition of investment returns is simple.
The return on an investment is simply the total proceeds from holding the
asset over the period divided by the initial investment cost, minus 1:
Total proceeds
Dividends + Sales price
Dividends + Capital gains
1=
1=
Return =
Initial price
Initial price
Initial price
where we have used the fact that Capital gains = Sale price Initial price.
When we consider a multi-period framework we need to consider interim
investments and asset sales also.
The log-returns are similarly defined as:
Initial price
Total proceeds
= ln
Log Return = ln
Returns are useful in the sense that they aggregate easily across assets. For
instance, the return on a portfolio is simply the weighted average returns
on the individual assets. It is generally not true that the log return on a
portfolio can be worked out as the weighted average of the log return
on the individual assets. Log returns are useful in the sense that they
aggregate easily over time. The log return over two periods, for instance,
is simply the sum of the log returns over the two individual periods. This
is also generally not true for returns. We illustrate this with the following
example. Consider an investment where you invest 100,000 in two
assets, A and B, with the original investment split equally between the two
119
23 Investment management
assets. A and B are both priced at 100 initially, and the price of A is 110 at
the end of the first year and 105 at the end of the second year. Similarly,
the price of B is 90 at the end of the first year and 95 at the end of the
second year. The value of the portfolio is as follows:
Initial value
100,000
100,000
100,000
The return on the portfolio is, therefore, zero percent across any individual
period, and also over the entire two-year horizon. Looking at the data for
the individual assets, we find that we can recover the portfolio returns by
taking a weighted average across assets, and that we can recover the longrun return of the assets by summing the log returns over time. The return
and log returns of the two assets are:
Return asset A
First year
Second year
10%
4.5%
9.5%
4.7%
Return asset B
10%
5.6%
10.5%
5.4%
1
1
10% = 0
10%
2
2
The average log return is, however, negative over the same period. Over
the second year, the portfolio holdings are (initially) 55,000/100,000 in
asset A and 45,000/100,000 in asset B. The second year portfolio return
is, therefore,
Portfolio return second year =
9
11
5.6% = 0 (excluding rounding error)
4.5% +
20
20
When it comes to the assets return over the entire two-year period we find
that the log returns are more accurate. The log return of asset A over the
entire two-year period is given by:
105
100
ln
and the log return of asset B over the entire two-year period is similarly
given by:
95
100
ln
(1 + r ) 1
t
120
For asset A, the geometric average return over the two-year period is:
Geometric average asset A = (1.10)(0.955) 1 = 2.5%
The geometric average takes into account the compounding effect when
holding assets over several periods. The (arithmetic) average does not
take this into account and will consequently overestimate the return over
several periods.
cov(yt, xt)
var(xt)
What do we do when the true relationship is not linear? If, for example,
the true relationship between the two observations is given by:
yt = a + bxt 2 + et
the linear regression method above is not very good at picking up the
relationship. However, if we regress yt on ln(xt) instead, we are able to
capture the non-linear true relationship by a linear model:
yt = c + d ln(xt) + ut
where the coefficient are a = c and d = 2b. Therefore, the regression
method is a very powerful method to measure the relationship between
random variables in a simple and straightforward way.
We can also extend the regression model to multiple independent random
variables that yield the multiple regression model:
yt = a + b1x1t + b2x2t + ... + bKxKt + et
where K+1 regression coefficients a, b1, ,bK are to be determined.
121
23 Investment management
Utility theory
Utility theory is a concept used in economics to model human behaviour
using mathematical functions called utility functions. These functions can
be defined over many goods and services, but when talking about financial
investments we define these functions over money. If an individual,
through some investment choice, ends up with a final cash balance of w,
we say his utility is u(w), where u is the utility function. If w is a random
variable, we measure his utility by the expected utility, which in general is
not equal to the utility of his expected cash balance:
Expected utility = E(u(w)) u(E(w))
Where do these utility functions come from? Utility functions are not an
inherent characteristic of human beings, they should rather be thought of
as a representation of preferences over outcomes. Why do we use utility
functions? The simple answer is that utility functions are a lot easier to
handle than preferences. What utility theory does, essentially, is to show
that when our preferences are sufficiently structured (or, perhaps more
accurately: rational) we can represent these by a utility function.
An example of the type of structure we impose on rational preferences is
the so called transitivity property. If an individual prefers A to B and also B
to C, it follows that he also prefers A to C. This property is fairly obvious,
but sometimes the structure we impose is more subtle (and controversial),
such as the so-called independence property. Consider the following
example. Suppose an individual can choose first between two lotteries,
lottery A paying 1m with 50% probability and 0 with 50% probability,
and lottery B paying 4m with 25% probability and 0 with 75%
probability. Then, the agent is asked to choose between a lottery C paying
1m for sure, and a lottery B paying 4m with 25% probability, 1m with
50% probability, or 0 with 25% probability. The independence property
states that if the individual chooses A before B then he also should choose
C before D. In practice, however, many people find these two choice
situations very different and may make contradicting choices.
The reason the agents should make the same choices is the following.
Consider lottery A and B. We can think of these situations by the following
table.
Outcome A
Probability A
Outcome B
Probability B
50%
50%
1m
50%
25%
4m
25%
Now look at the lotteries C and D. We can represent these by the following
table.
Outcome C
Probability C
Outcome D
Probability D
1m
50%
1m
50%
1m
50%
25%
4m
25%
We notice that the second and third lines are identical for both choice
situations, and that the first line gives the same outcome. Therefore, if our
preferences are sufficiently rational we should prefer C to D if and only if
we also prefer A to B.
122
u''(x)
u'(x)
which is minus the second derivative over the first derivative. If the utility
function is linear, the risk aversion coefficient is zero, indicating risk
neutral preferences.
v))
E(u(x)) = E( exp(x)) = exp((m
2
Therefore, this individuals expected utility maximisation problem is very
easy to state: the individual seeks to maximise the expected wealth minus
the variance of his wealth times half his risk aversion coefficient. Suppose,
for instance, that an individual with CARA utility and risk aversion
coefficient 2 seeks to maximise the expected utility from investing in
123
23 Investment management
124