London School of Economics Department of Finance

Michaelmas Term (Section A) Dr. Georgy Chabakauri

FM300 Corporate Finance, Investments, and Financial Markets

Investments and Elements of International Finance

Course Description

This section of the course examines the empirical evidence on the behavior of stock
prices from an investments perspective. After reviewing the essentials of portfolio
theory and asset pricing, we will examine empirical tests of asset pricing models and
their practical applications in determining expected returns.
The course then focuses on market efficiency and return predictability. We
will learn how to conduct an event study, how to create a portfolio strategy based on
investment anomalies, and how to interpret the evidence in the light of rational and
behavioral theories.
We will then examine the main models and methods of performance
evaluation and attribution, focusing on the main issues concerning the management
of bond and equity portfolios.
The last part of the course introduces topics in international finance.
Particular emphasis is given to issues related to international portfolio management
and hedging strategies.

Readings

Lecture notes is the main source for the preparation for the exam.

The required textbook for this section of the course is:

Zvi Bodie, Alex Kane, and Alan Marcus, Investments, McGraw-Hill, 11th ed.
(note: 10th and 9th editions are fine.)

The lecture notes also include additional materials covered in the optional
readings but not covered in the textbook.

Homework assignments

There is no continuous assessment in this course. However, in order to test

your mastery of the material as we progress through the course, we will
collect HW #3 and #4. These formative assignments are graded for feedback
only, and your final mark for the course will be determined by your
performance in the final exam. Answers to the homework problems will be
posted on Moodle in the week after the class. Week 11 is the reading week

1
during which you should revise the whole course and problem sets, and read
the optional materials to deepen your knowledge.

Contact information

Office (CON 2.08)

Email: g.chabakauri@lse.ac.uk
First line of contact: class teachers
Check online for office hours

Course Outline

1. The Capital Asset Pricing Model: overview

• Basic rules of statistics
• Brief review of portfolio theory and the CAPM
• Risk and return with a risk-free asset
• Applications of the CAPM
• Predictions of CAPM

Readings:
BKM Chapters 6, 7, 9

2. The Capital Asset Pricing Model: empirical tests

• Regression analysis
• Estimating betas and expected returns
• Empirical tests of the CAPM

Readings:
BKM Chapters 8.1, 8.2, 13.1

3. Empirical evidence of the CAPM and multifactor models

• Empirical tests of the CAPM
• Brief overview of multifactor models and empirical tests
• The Fama-French model

Readings:
BKM Chapters 10, 13.2, 13.3

2
Optional readings:
Fama, Eugene, and Kenneth French, 1992, “The cross-section of expected
stock returns”, Journal of Finance 47, pp. 427-265.
Fama, Eugene, and Kenneth French, 1993, “Common risk factors in the
returns on stocks and bonds”, Journal of Financial Economics 33, pp. 3-56.
Fama, Eugene, and Kenneth French, 2015, “A five-factor asset prising
model”, Journal of Financial Economics 116, 1-22.

4. Market efficiency: Empirical tests and evidence

• The three forms of market efficiency
• Are markets efficient? Tests of market efficiency and empirical
evidence
• Technical analysis
• Event studies

Readings:
BKM Chapters 11.1 – 11.4, 12

Optional readings:
Fama, E., 1991, “Efficient capital markets: II”, Journal of Finance 46, 1575-
1617.

5. Investment anomalies and return predictability

• Investment anomalies: size, value, momentum, post-earnings
announcement drift
• Issues on data-mining

Readings:
BKM Chapter 11.4

Optional readings:
Schwert, W., 2003, “Anomalies and market efficiency”, Handbook of the
Economics and Finance, Ch. 15, 937-972.
Jegadeesh, N. and Titman S., 2001, “Profitability of momentum strategies:
An evaluation of alternative explanations”, Journal of Finance 56, 699-720.
La Porta, R., J. Lakonishok, A. Shleifer, and R. Vishny, 1997, “Good news
for value stocks: further evidence on market efficiency”, Journal of Finance
52, 859-874.

6. Behavioral Finance and limits to arbitrage

• Behavioral biases

3
 • Investment anomalies: underreaction and overreaction
• The disposition effect
• Costly arbitrage and short-sale constraints

Readings:
BKM Chapter 12

Optional readings:
Barberis, N., A. Shleifer, and R. Vishny, 1998, “A Model of Investor
Sentiment,” Journal of Financial Economics 49, 307-343.
Daniel, K., A. Hirshleifer, and A. Subrahmanyam, 1998, “Investor
Psychology and Security Market Under- and Overreactions,” Journal of
Finance 53, 1839-1885.
Hong, H., and J. C. Stein, 1999, “A Unified Theory of Underreaction,
Momentum Trading, and Overreaction in Asset Markets,” Jounral of
Finance 54, 2143-2184.
Shleifer, A., and R. Vishny, 1997, “The limits of arbitrage”, Journal of
Finance 52, 35-55.
Shleifer, A., 2000, “Inefficient markets: An introduction to behavioral finance”,
Clarendon Lectures in Economics, Oxford University Press, New York.

7. Fixed-income portfolio management

• Interest rate risk
• Duration and convexity
• Interest rate swaps
• Passive bond management, immunization, active bond
management

Readings:
BKM Chapters 14, 15, 16

8. Active portfolio management and performance evaluation

• Measures of performance
• Market timing
• Stock picking
• The Treynor-Black model
• Performance attribution procedures

Readings:
BKM Chapters 24, 27.1, 27.2

Optional readings:
Sharpe, W., 1992, “Asset allocation: Management style and performance
measurement”, Journal of Portfolio Management, Winter, 7-19.

4
Carhart, M., 1997, “On persistence in mutual fund performance”, Journal of
Finance 52, 57-82

9. Elements of International Finance

• International financial markets: Spot and forward exchange
markets, Interest rate parity
• Exchange rate determination: Purchasing Power Parity, Forecasting
exchange rates, International risk management

Readings: lecture notes.

Optional readings:
Raman Uppal and Piet Sercu, International Financial Markets and the Firm,
South-Western College Publishing, 1995.
Adler, M., and B. Dumas, 1984, “Exposure to currency risk: Definition and
measurement”, Financial Management.

10. International asset allocation

• Exchange rate risk
• The world equity portfolio
• Benefits from international diversification
• The home equity bias

Readings:
BKM Chapter 25

Optional readings:
Michaud, R., G. Bergstrom, R. Frashure, and B. Wolahan, 1996, “Twenty
years of international equity investing”, Journal of Portfolio Management 23
Eun, C., and B. Resnick, 1988, “Exchange rate uncertainty, forward
contracts, and international portfolio selection”, Journal of Finance 43, 197-
215.
French, K., and J. Poterba, 1991, “Investor diversification and international
equity markets”, American Economic Review 81, 222-226.

5
FM300 Corporate Finance,
Investments,
and Financial Markets
Introduction
Dr. Georgy Chabakauri
London School of Economics

FM 300 Introduction 1

FM300 Section 1
Investments and Elements
of International Finance

FM 300 Introduction 2
 Administrative details
Course materials
– Self-contained lecture notes;
– Textbook: Investments and portfolio management by
Bodie, Kane, and Marcus, McGraw-Hill, 10th ed.
– Optional readings (e.g., journal articles)
Pre-requisites
– FM 212/213 or equivalent course (talk to me if unsure)
Lectures
– Monday 12-2pm, a short (10min) break;
– Week 11 is the reading week.
– Lectures are recorded.
FM 300 Introduction 3

Administrative details
Classes and Homework
– First class in MT week 3, last class in LT week 1;
– Two formative assignments: Homework 3 and Homework 4
– These assignments are marked for feedback only and
should be submitted in class to your class teachers;
– Homework 10 is a sample exam;
Final Exam
– In the summer term, 100% of the course mark;
– Solutions to two past exams will be on Moodle.

FM 300 Introduction 4
 Administrative details

Office Hours (CON 2.08)

– First line of contact, especially about problem sets: class
teachers;
– Check my office hours online;
– Course-related questions are difficult to resolve by e-mail.
So, come to the office hours instead of sending e-mails.
– There will be no office hours during breaks and holidays.

FM 300 Introduction 5

Main Topics

Benchmark asset pricing models (L1-L3)

Market efficiency and behavioral finance (L4-L6)
Portfolio management and performance evaluation (L7-L8)
International finance (L9-L10)

FM 300 Introduction 6
 Topic 1: CAPM

Basic rules of statistics;

Brief review of portfolio theory and the CAPM.

FM 300 Introduction 7

Topic 2: Empirical Tests of CAPM

Regression analysis;
Estimating betas and expected returns;
Applications of the CAPM;
Empirical tests of the CAPM.

FM 300 Introduction 8
 Topic 3: Multifactor Models

Multifactor models and empirical tests;

Arbitrage Pricing Theory;
The Fama-French model.

FM 300 Introduction 9

Topic 4: Market Efficiency

Tests of market efficiency;

Weak form efficiency: Technical analysis;
Semi-strong form efficiency: Event studies;
Strong form efficiency.

FM 300 Introduction 10
 Topic 5: Cross-sectional and
time-series predictability of returns

Portfolio strategies:
– Size;
– Short-term reversal, medium-term momentum;
– Value and growth;
– Post-earnings announcement drift.

FM 300 Introduction 11

Topic 6: Behavioral Finance and

Limits to Arbitrage

Behavioral biases
– Underreaction, overreaction
– Inattention
Limits to arbitrage
– Noise trader risk
– Riding the sentiment wave
– Short-sale constraints

FM 300 Introduction 12
 Topic 7: Fixed-income Portfolio
Management

Interest rate risk;

Duration and convexity;
Passive bond management: immunization;
Active bond management: Yield curve strategies.

FM 300 Introduction 13

Topic 8: Performance Evaluation

Measures of portfolio performance

Active asset allocation: the Treynor-Black model
Performance attribution
– Stock selection and market timing
– Style analysis

FM 300 Introduction 14
 Topic 9: International Finance

Spot and forward exchange rates

Covered Interest Parity
Uncovered Interest Parity
Purchasing Power Parity
Real exchange rates and interest rates

FM 300 Introduction 15

Topic 10: International Asset Allocation

Returns on Foreign Investment;

Exchange rate risk;
Hedging FX risk;
The International CAPM;
The home bias.

FM 300 Introduction 16
 FM 300 Section A

The Capital Asset Pricing Model:

Overview
Lecture Note 1
Dr. Georgy Chabakauri
London School of Economics

FM 300 Lecture Note 1 1

Questions

How much should we expect to receive as a return from a

risky asset?
How do we know if a trading strategy is profitable?
How do we estimate the impact of announcing an acquisition
plan?
What benchmark should we use to evaluate the performance
of a mutual fund manager?
How do we discount the cash-flows of a project?

FM 300 Lecture Note 1 2

 Outline

Basic statistics rules for portfolio returns

Brief overview of portfolio theory and the CAPM
– How are expected returns determined?
– How are they related to risk?
Risk and return with a risk-free asset

FM 300 Lecture Note 1 3

Basic statistics for asset returns

Suppose you had bought a share of GM at the end of
December 1989 for $42.25 and sold it a year later, at the end of
December 1990, for $34.375. During this year, GM paid a per-
share dividend of $3.1 Your realized return from holding GM
throughout 1990 would have been:

34.375 + 3.1 - 42.25

= -11.54%
42.25
E(ri) – rf : The expected excess return of asset i is the expected
asset’s return minus the return on the riskless asset
E(rM) – rf : The market risk premium is the expected excess
return of the market portfolio.

FM 300 Lecture Note 1 4

 Basic statistics for asset returns
GM example
Date Price Dividend Return
29-Dec-89 42.2500 -
31-Dec-90 34.3750 3.00 -11.54%
31-Dec-91 28.8750 1.60 -11.35%
31-Dec-92 32.2500 1.40 16.54%
31-Dec-93 54.8750 0.80 72.64%
30-Dec-94 42.1250 0.80 -21.78%
29-Dec-95 52.8750 1.10 28.13%
31-Dec-96 55.7500 1.60 8.46%
31-Dec-97 60.7500 5.59 19.00%
31-Dec-98 71.5625 2.00 21.09%
31-Dec-99 72.6875 14.15 21.34%
Average return 14.25%
Variance of return 0.0638
Standard deviation of return 25.25%

FM 300 Lecture Note 1 5

Basic statistics for asset returns

Expected Return (Mean Return)

N
1
N
åR
t =1
t = RA ® E ( RA )

Law of large numbers

FM 300 Lecture Note 1 6

 Expected return on a portfolio

The expected rate of return on a portfolio of stocks is:

j =n
( )
E (rP ) = å j =1 x j E rj where å j =1 x j = 1
j =n

The expected rate of return on a portfolio is a weighted

average of the expected rates of return on the individual
stocks.
In the two-asset case:
E (rP ) = x1 E (r1 ) + (1 - x1 ) E (r2 )

FM 300 Lecture Note 1 7

Portfolio returns
We measure the (realized) return rP on a portfolio in period t as:

rPt = å j =1 x j rjt
j=N

where xj = fraction of the portfolio’s total value invested in stock

j, j=1,…,N.
– xj > 0 is a long position; xj < 0 is a short position;
Stock market indices:
– Equally weighted: x1=x2=…=xN=1/N
– Value weighted: xj= Proportion of market capitalization
We measure the average return of a portfolio over time as:
1 t =T
rP = å rPt
T t =1
FM 300 Lecture Note 1 8
 A digression: short sales
A short sale consists of selling a stock that we do not own.
Steps:
– Date 0: Borrow the stock from a broker.
– Date 0: Sell the stock in the market.
– Between dates 0 and 1: Compensate the broker for any
dividends the stock pays.
– Date 1: Buy the stock in the market.
– Date 1: Return the stock to the broker.
A short sale is profitable if the stock price goes down.
Portfolio return is still weighted average of returns on the
individual stocks, but weights of shorted stocks are negative.

FM 300 Lecture Note 1 9

Variance

Variance: Average value of squared deviations from the

mean. A measure of volatility.

Var( X ) = E{[ X - E ( X )]2 }

Standard deviation: A measure of volatility that is easier to

interpret because expressed in the same units as the portfolio
return.
Std ( X ) = Var( X )

FM 300 Lecture Note 1 10

 Covariance and correlation

Covariance between two random variables X and Y:

Cov ( X , Y ) = E[{X - E ( X )}{Y - E (Y )}]

The covariance measures how returns on different stocks

move in relation to each other.
Correlation between two random variables X and Y:

Cov( X , Y )
Corr ( X , Y ) =
Std ( X ) Std (Y )

FM 300 Lecture Note 1 11

Rules of means and variances

Let a, b be two constants. Then expected values are:

E (ar1 ) = aE (r1 )
E (r1 + r2 ) = E (r1 ) + E (r2 )
E (ar1 + br2 ) = aE (r1 ) + bE (r2 )

where the third rule is implied by the first two. In our applications
the constants a and b will typically be portfolio weights.
Similarly, for variances:
Var (ar1 ) = a 2Var (r1 )
Cov(ar1 , r2 ) = aCov(r1 , r2 )
Var (r1 + r2 ) = Var (r1 ) + Var (r2 ) + 2Cov(r1 , r2 )
Var (ar1 + br2 ) = a 2Var (r1 ) + b 2Var (r2 ) + 2abCov(r1 , r2 )
FM 300 Lecture Note 1 12
 Measuring Portfolio Risk
The risk of a portfolio is measured by its standard
deviation or variance.
The variance for the two stock case is:
var( rp ) = s p2 = x12s 12 + x22s 22 + 2 x1 x2s 12
s i2 = Variance of asset i
s ij = Covariance of returns of assets i and j
or, equivalently,

var( rp ) = s p2 = x12s 12 + x22s 22 + 2 x1 x 2 r12s 1s 2

rij = Coefficient of correlation of the returns of i and j

FM 300 Lecture Note 1 13

Covariance and correlation

GM and MSFT
Date GM return MSFT return
31-Dec-90 -11.54% 72.99%
31-Dec-91 -11.35% 121.76%
31-Dec-92 16.54% 15.11%
31-Dec-93 72.64% -5.56%
30-Dec-94 -21.78% 51.63%
29-Dec-95 28.13% 43.56%
31-Dec-96 8.46% 88.32%
31-Dec-97 19.00% 56.43%
31-Dec-98 21.09% 114.60%
31-Dec-99 21.34% 68.36%
Average return 14.25% 62.72%
Variance of return 6.38% 14.43%
Standard deviation of return 25.25% 37.99%
Covariance of returns -0.0552
Correlation -0.5755

FM 300 Lecture Note 1 14

 Review of Portfolio Theory
What happens to risk if we combine more and more stocks
into a portfolio?
Combining stocks into portfolios can reduce standard
deviation.
Correlation coefficient is crucial to diversification
– Perfectly correlated;
– Imperfectly correlated;
– Perfectly and negatively correlated.
Efficient portfolios: have the smallest standard deviation given
expected return; with the highest expected return given risk.

FM 300 Lecture Note 1 15

Portfolio Risk
Portfolio standard deviation

Unique
risk

Market risk
0
5 10 15
Number of Securities

FM 300 Lecture Note 1 16

 2-stock Portfolios

We can plot frontiers for different correlations.

Diversification depends on correlations.
0.15
0.14
IBM
Expected Return (R)

0.13

0.12

0.11
0.1

0.09
ExxonMobil
0.08

0.07
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Standard Deviation (σ)

ρ=-1 ρ=-0.5 ρ=0 ρ=0.35 ρ=0.5 ρ=1

FM 300 Lecture Note 1 17

Efficient Frontier with N stocks

Formally, the problem of finding the efficient frontier of N-

stock portfolios is as follows:
N N
σ p = min
w1 ,w 2 ,...,w N
∑i=1 ∑ j=1w i w jσ iσ j ρij ,
subject to: w1 +... + w N = 1,
w1r1 +... + w N rN = rP .

Investor achieves minimal risk for given expected return r p

Hence, the optimal portfolio is on the efficient frontier.
This optimization problem can be solved explicitly. However,
this is beyond our scope. You can use Excel solver to solve
for optimal weights and σp given r p .
FM 300 Lecture Note 1 18
 Example

Consider 3 stocks with the following characteristics:

Stock Expected Return Standard Deviation

IBM 15% 29.7%
ExxonMobil 7.9% 19.2%
Starbucks 12.3% 29.9%

and correlations:

IBM ExxonMobil Starbucks

IBM 1 0.35 0.2
ExxonMobil 1 -0.1
Starbucks 1

FM 300 Lecture Note 1 19

Efficient Frontier with 3 stocks

The picture below shows 2-stock and 3-stock frontiers.
Efficient frontier is the upward sloping (dark blue) part.

16%
Efficient Frontier with 3 stocks
15%
14%
Expected Return (R)

13% IBM

12%
11%
10% Starbucks
9%
8%
ExxonMobil
7%
14% 16% 18% 20% 22% 24% 26% 28% 30%
Standard Deviation (σ)

FM 300 Lecture Note 1 20

 Lending and Borrowing allowed

How do the combinations of risky portfolios and riskless assets

look like? Consider a new portfolio Q with fraction (1-w) in riskless
asset (long or short position in bonds) and w in risky portfolio P:

rQ = (1- w )rf + wrP = rf + w (rP - rf ).

Weight w > 1 means we borrow (or sell short Treasury Bills).

Expected returns and standard deviations of returns rQ:

rQ = rf + w (rP - rf ),
sQ = (1- w ) 2 var(rf ) + w 2sP2 + 2(1- w )w cov(rf , rP )
=0 =0
= w s = ws P .
2 2
P

FM 300 Lecture Note 1 21

Lending and Borrowing allowed

Substituting out w=σQ/σP we obtain expected returns of Q as

functions of standard deviations:
æ r P - rf ö
rQ = rf + sQ ç ÷.
è sP ø
This is an equation of a straight line with the slope given by
the Sharpe ratio: (rP - rf ) / sP .

FM 300 Lecture Note 1 22

 Lending and Borrowing allowed
Portfolios on R-P line are not the best portfolios that one can
get. Portfolios on a steeper line that passes through R dominate
those on R-P line.
The efficient frontier with the riskless asset is the steepest
possible line, R-T, where T is tangency portfolio.
ng
owi
r r
Bo
Expected Return (%)

T
g
ndin
Le Q2
P
rf Q1 W>1
R
W<1

Standard Deviation (σ)

FM 300 Lecture Note 1 23

Lending and Borrowing allowed

Lending/borrowing at a riskless rate rf (or buying/selling

bonds) helps achieve higher returns for a given risk.

Efficient frontier becomes a tangent line.

tangency
portfolio ng
owi
r r
Bo
Expected Return (%)

ing
end
L
rf

Standard Deviation

FM 300 Lecture Note 1 24

 Efficient Portfolios with Multiple Assets

Return

Low Risk High Risk

High Return High Return

Low Risk High Risk

Low Return Low Return

Risk

FM 300 Lecture Note 1 25

CAPM

Capital market line (CML) passes through the market portfolio and
the riskless asset. Because market and tangency portfolios coincide
CML coincides with the efficient frontier.
CML
Expected Return (%)

rM
IBM Coca-Cola

rf ExxonMobil

Standard Deviation

FM 300 Lecture Note 1 26

 The Capital Market Line
The CML gives the trade-off between risk and return for
portfolios consisting of the risk-free asset and the tangency
portfolio M.
The equation of the CML is:
E ( rM ) - r f
E ( rp ) = r f + s p
sM

The expected rate of return on a risky asset can be thought of

as composed of two terms: risk-free rate and risk premium:
– E(r) = rf + Risk x [Market Price of Risk]

FM 300 Lecture Note 1 27

Basic assumptions of the CAPM

The goal is to determine expected returns by combining

optimal portfolio choices of all investors in the economy.
Assumptions:
– Investors have a one-period horizon
– Each investor holds a mean-variance efficient portfolio,
i.e., a portfolio with the highest expected return given its
volatility
– Investors have the same beliefs.
– Lending and borrowing at a single riskless rate.

FM 300 Lecture Note 1 28

 The intuition on the CAPM

Everybody holds the same risky portfolio, the market.

Market risk cannot be diversified.
Investors demand a premium for bearing non-diversifiable risk
only, the market risk.
Beta measures the market risk, hence it is the correct
measure for non-diversifiable risk.
Conclusion:
In a market where investors can diversify by holding many
assets in their portfolio, they demand a risk premium
proportional to beta.

FM 300 Lecture Note 1 29

CAPM
Capital Asset Pricing Model (CAPM):
cov "# , "+
! "# − "% = - (! "+ − "% )
,+
stock excess beta, βi market excess
return return

This formula can be rewritten in terms of beta:

! "# − "% = 0# (! "+ − "% )

Expected returns depend on market risk measured by beta and

not by individual risk which can be diversified away.

Market portfolio M is well-diversified, and hence, contains only the

market risk.

FM 300 Lecture Note 1 30

 CAPM
Excess return is a linear function of β. This function is called
Security Market Line. All possible portfolios will be somewhere on
this line.
– Note that market portfolio has β = 1.

r = rf + b (rM - rf ),
Expected Return (%)

rM

rf

1.0 Beta (Risk)

FM 300 Lecture Note 1 31

Beta
The appropriate measure of risk for an individual stock is
beta.
Beta measures the stock’s sensitivity to market risk factors.
– The higher the beta, the more sensitive the stock is to
market movements.
The relevant risk for pricing asset i is its contribution to the
variance of the market portfolio:
s iM Asset i's contribution to Market variance
bi = =
s M2 Total Market variance

FM 300 Lecture Note 1

 Summary
Optimal investments depend on trading off risk and return
– Investors with higher risk tolerance invest in riskier assets;
– Only risk that cannot be diversified matters.
If investors can borrow and lend, then everybody holds a
combination of two portfolios: the market portfolio of all risky
assets and the the riskless asset.
In equilibrium, all stocks must lie on the security market line.
– Beta measures the amount of non-diversifiable risk;
– Expected returns reflect only market risk;
– These expected returns can be used as required returns in
project evaluation, performance measurement, etc.

FM 300 Lecture Note 1 33

Appendix: CAPM Derivation (Not Required)

Derived from the fact that market portfolio coincides with tangency
portfolio => is on the steepest line with highest Sharpe ratio.

Denote by wf proportion of wealth in riskless asset, and wi

proportion of wealth in asset i. Then:
w f + w1 + ... + w N = 1.
The return on a portfolio with expected value rP (after substituting
out wf) and volatility σP is given by:
n n
rP = w f rf + å w i ri = rf + å w i (ri - rf ),
i =1 i =1 (1)
N N
s = åå w i w j s i s j rij .
2
P
i =1 j =1

FM 300 Lecture Note 1 34

 … CAPM: Derivation (Not Required)
Consider now a portfolio P with stock weights equal to the weights
of market portfolio, except for stock k:
w1 = w1M ,..., w k -1 = w kM-1, w k ¹ w kM , w k +1 = w kM+1,..., w N = w NM .
Suppose, we fix all portfolio weights except for weight wk and
consider the Sharpe ratio (rP - rf ) / sP as a function of wk.
We know, that the market portfolio has highest possible Sharpe
ratio. Hence, Sharpe ratio of portfolio P will be maximized when
w k = w kM .
Therefore, the following first order condition should hold:
d [(rP - rf ) / sP ] (2)
= 0.
dw k w k =w kM

FM 300 Lecture Note 1 35

… CAPM: Derivation (Not Required)

Using chain and product rules we obtain:
d [(rP - rf ) / sP ] 1 d (rP - rf ) rP - rf dsP
= - 2 .
dw k sP dw k sP dw k
From this equation and (2), evaluating the expressions at
w k = w kM we obtain:
d (rP - rf ) rM - rf dsP
= , (3)
dw k sM dw k

where all derivatives are evaluated at w k = w kM .

FM 300 Lecture Note 1 36

 … CAPM: Derivation (Not Required)
Taking into account (1), computing the derivatives with respect to
wk and evaluating them at w k = w k we obtain:
M

d (rP - rf ) dsP 1 dsP2 cov(rk , rM ) (4)

= rk - rf , = = .
dw k dw k 2sP dw k sM

The last equality holds because:

N
s =w s +2
2
P
2
k
2
k åw kw iM cov(rk , ri ) + terms without w k .
i =1, i ¹k
Hence, the derivative evaluated at w k = w kM is given by:
dsP2
= 2(w 1M cov(rk , r1 ) + ... + w kM cov(rk , rk ) + ... + w NM cov(rk , rN ) )
dw k
= 2 cov(rk , w1M r1 + ... + w NM rN ) = 2 cov(rk , rM ).
FM 300 Lecture Note 1 37
 FM 300 Section A

The Capital Asset Pricing Model:

Empirical Tests
Lecture Note 2
Dr. Georgy Chabakauri
London School of Economics

FM 300 Lecture Note 2 1

Outline

Regression analysis (estimate beta)

Use of CAPM
Empirical tests of the CAPM
Roll’s Critique

FM 300 Lecture Note 2 2

 Regression analysis
Linear regression is a technique for fitting a line to data.
Consider the relation between two variables X and Y:
Y = α + βX + ε
– α, β are constants;
– ε is a random variable, cov(X,ε) = 0 and E(ε) = 0.
Variation in Y is decomposed into
– Var(βX): variation that can be “explained” by X.
– Var(ε): “unexplained” variation.
cov(Y,X)= β×cov(X,X) => β is given by:
cov(', ))
!=
var(')
FM 300 Lecture Note 2 3

Regression analysis
In general, we do not know α and β. We can estimate them
using regression analysis.
Need to fit the best line to scatterplot of Y vs. X.
Method of ordinary least squares:
6

min 2 )3 − 8 − 9'3 :
0,1
345

Let a and b be estimates of α and β. The line of the best fit is:
Y = a + bX.
This method also works for multiple explanatory variables.

FM 300 Lecture Note 2 4

 … Regression analysis

Estimates a and b are random variables, and their precision

depends on the number of data points N.

The estimates are unbiased: E[a|X]=α and E[b|X]=β.

Denote their standard deviations by: >=? and >=A .

Suppose, we want to test a hypothesis H0: ; = ;.

<

Consider the following t-statistic:

8 − ;<
C=
>=?
Reject the hypothesis if |t| is too large. But how large exactly?

FM 300 Lecture Note 2 5

… Regression analysis

Reject H0 if |C| > C,̅ where the bound is such that:

Prob C ≥ C̅ = N
Under some distributional assumptions and large N, C̅ = 1.96
for the significance level J=0.05.
Practical rule: reject the hypothesis when |t|>2.
Typically, H0: α = 0 or β = 0. If H0 is rejected => the coefficient
is statistically significant.
R-squared measures how well we fit the data:
explained var var(!')
O: = =
explained var + unexplained var var !' + var(V)

FM 300 Lecture Note 2 6

 Regression and asset returns
Regression equation for stock i:
( rit - rf ) = ai + b i ( rMt - rf ) + e it
εit = firm-specific component of return, not due to market
movements.
rMt – rf = component due to movement in the market portfolio.
Variation in return of asset i is decomposed into:
– Systematic risk, perfectly correlated with the market
portfolio: Var ( b i (rMt - rf ))
– Idiosyncratic risk, uncorrelated with the market portfolio:
Var(e it )

FM 300 Lecture Note 2 7

Regression and asset returns

Example: Suppose we estimate the regression of MSFT

monthly excess returns on the S&P500 excess returns, and
we obtain the following fitted line:
Y = -0.0424 +1.4693 X
Interpretation:
– on average, a 1% increase in the S&P monthly return
causes a 1.4693% increase in MSFT monthly return.
– in months when the S&P 500 doesn’t “move,” MSFT’s
return tends to be -4.24%

FM 300 Lecture Note 2 8

 Uses of the CAPM

Estimate expected returns for:

– Project evaluation;
– Portfolio selection;
– Benchmark to evaluate investment strategies (risk
adjustment). It tells us what the required return on a stock
should be, after accounting for risk;
– Performance evaluation and attribution.

FM 300 Lecture Note 2 9

Estimating expected returns

The SML gives us a way to estimate the expected (or

required) rate of return on equity.
( ) [
E rj = r f + b j E(rM ) - r f ]
We need estimates of three things:
– Risk-free interest rate, rf.
– Market risk premium, [E(rM)-rf].
– Stock’s beta, βj.

FM 300 Lecture Note 2 10

 …Estimating expected returns

The risk-free rate can be estimated by the current yield on

Treasury bills (one-year, one-month, depending on frequency
of data).
– Assume one-year yield on Treasury bills is 5.0% .
The market risk premium can be estimated by looking at the
historical difference between the return on stocks and the
return on Treasury bills.
– This difference has averaged about 8% since 1926.
The betas are estimated by regression analysis.

FM 300 Lecture Note 2 11

Estimating Betas

General Motors
GM return (%)

Price data: May 91- Nov 97

R2 = .07
β = 0.72

Market return (%)

FM 300 Lecture Note 2 12

 Estimating Betas

General Motors

GM return (%)
Price data: Dec 97 - Apr 04

R2 = .29
β = 1.21

Market return (%)

FM 300 Lecture Note 2 13

Betas of selected common stocks

Stock Beta Stock Beta

AT&T 0.96 Ford Motor 1.03
Boston Ed. 0.49 Home Depot 1.34
BM Squibb 0.92 McDonalds 1.06
Delta Airlines 1.31 Microsoft 1.20
Digital Equip. 1.23 Nynex 0.77
Dow Chem. 1.05 Polaroid 0.96
Exxon 0.46 Tandem 1.73
Merck 1.11 UAL 1.84
Betas based on 5 years of monthly returns.

FM 300 Lecture Note 2 14

 …Estimating expected returns

E(r) = 5.0% + b (8.6%)

Stock E(r) Stock E(r)

AT&T 13.3% Ford Motor 13.9%
Boston Ed. 9.2% Home Depot 16.5%
BM Squibb 12.9% McDonalds 14.1%
Delta Airlines 16.3% Microsoft 15.3%
Digital Equip. 15.6% Nynex 11.6%
Dow Chem. 14.0% Polaroid 13.3%
Exxon 9.0% Tandem 19.9%
Merck 14.5% UAL 20.8%

FM 300 Lecture Note 2 15

…Estimating expected returns

What is the beta and expected rate of return of an equally-

weighted portfolio consisting of Exxon and Polaroid?
Portfolio Beta

b p = (1 / 2 )(.46) + (1 / 2 )(.96)
b p = 0.71

Expected Rate of Return

E ( r p ) = 5.0% + ( 8.6% )(0.71) = 11.1%

FM 300 Lecture Note 2 16

 Expected returns: CAPM regression

The regression equation implies:

E ( ri ) - rf = ai + b i ( E ( rM ) - rf )

αi represents the abnormal return of stock i.

The intercept in the time-series regression is also the
deviation of the asset from the SML.
If the CAPM holds, then αi=0 for all assets.

FM 300 Lecture Note 2 17

…Expected returns: CAPM regression

FM 300 Lecture Note 2 18

 The SML and mispriced stocks

Suppose for a particular stock:

cov X3 , X[
W X3 < XZ + : (W X[ − XZ )
>[

Remember the definition of expected returns:

W(\3,5 + ]3,5 )
1 + W(X3 ) =
\3,^

Then P0 falls, so that E(ri) increases until disequilibrium

vanishes and the equation holds!

FM 300 Lecture Note 2 19

Disequilibrium example

Assume:

E ( rM ) = 0.11
rf = 0.03
bi = 1.25
E ( ri ) = .03 + 1.25(.08) = 0.13

Suppose that stock i is offering an expected return of 0.15.

According to the CAPM, this stock is underpriced – its rate of
return is too high for its level or risk.

FM 300 Lecture Note 2 20

 Does the CAPM hold?

Implications of the CAPM:

W X3 − XZ = !3 (W X[ − XZ )
1. Expected excess returns are linear in beta.
2. Slope of the line is market risk premium.
3. Expected returns depend only on beta.

FM 300 Lecture Note 2 21

Testing CAPM

Two main types of tests, both based on the same equation

E[rit – rft] = bi E[rMt – rft],
or the “average” pricing relationship.

FM 300 Lecture Note 2 22

 Time Series Approach

The first test is based on

E[rit – rft] = bi E[rMt – rft]
Consider the regression of a single firm’s excess returns on the
market’s excess returns:
rit – rft = ai + bi (rMt – rft) + eit
This implies
E[rit – rft] = ai + bi E[rMt – rft]
The CAPM therefore imposes the restriction that ai = 0 for all i.

FM 300 Lecture Note 2 23

…Time Series Approach

FM 300 Lecture Note 2 24

 Two-Pass Approach

The second test also uses the equation of the SML:

E[rit – rft] = bi E[rMt – rft]
1. Estimate betas for each stock (the “first pass regression”)
and for all i compute sample averages:
rit – rft
2. Run the “second pass” regression

rit – rft = g 0 + g 1bˆi + ei

FM 300 Lecture Note 2 25

…Two-Pass Approach

If the CAPM holds: N^ = 0 and N5 = X[` − XZ` .

To run this test, we consider 2904 stocks traded on the NYSE,
AMEX, and NASDAQ between 1989 and 1998.
Average excess market return was .0103.
Regress each stock on a value-weighted index to get betas (the
“first pass regression”), and use 1-month T-bills as the riskless
asset.

FM 300 Lecture Note 2 26

 …Two-Pass Approach

FM 300 Lecture Note 2 27

… Two-Pass Approach
Second pass regression results:
g0 g1
Estimate: 0.00122 0.00676
Std. Err.: (0.00037) (0.00036)
T-statistic for g0 = 0:
(.00122 – 0) /.00037 = 3.30
T-statistic for g1 = (average excess market return):
(.00676 – .0103) / .00036 = –9.82
CAPM is rejected on both counts!

FM 300 Lecture Note 2 28

 … Two-Pass Approach

FM 300 Lecture Note 2 29

…Testing CAPM

Summary of testing procedures

(1) Time series regression test
– Advantage: better statistical properties
(2) Two-pass regression test
– Advantages: many securities may be considered, easy to
do, easy to extend
– Disadvantage: weaker statistical properties

FM 300 Lecture Note 2 30

 Empirical evidence

The first tests of the CAPM asked:

– Is the risk/return relation linear?
– Does it have a positive slope?
– Does it have the correct intercept and slope?
– Does expected return depend on anything other than
beta?
Fist tests found that the relation between beta and average
excess returns was too flat.

FM 300 Lecture Note 2 31

Testing CAPM: Fama-MacBeth

Fama-MacBeth (1973) estimate monthly cross-sectional regression
of returns vs. betas for portfolios sorted by beta values estimated
from a prior period, then average the estimates of the risk premium
(slope) and the intercept.
– To estimate beta for year t they use 5 years of preceding data.
– Then, sort stocks by beta and form N portfolios.
– Using portfolios instead of individual assets leads to more
precise estimation because error terms are diversified away.

FM 300 Lecture Note 2 32

 Testing CAPM: Fama-MacBeth
Summary of the procedure:
– Estimate portfolio betas using rolling 5-year regressions;
– Next, run cross-sectional regression for each t to find a` :
X3` − XZ = bcdeC` + !3 a` + ;3` , f = 1, 2, … i

– Calculate the average lambda:

a5 + ⋯ + a l ;35 + ⋯ + ;3l
aj = ;= 3 =
m m
– Calculate standard deviations (see Fama-MacBeth, 1973).

FM 300 Lecture Note 2 33

…Testing CAPM: Fama-MacBeth

This procedure is easy to extended to multifactor models.
They also allow for possible non-linearity as well as the impact
of idiosyncratic risk on expected returns, which we will see later.
They find that:
– Average returns and betas are (reasonably) linearly related.
– The expected return on a portfolio with a beta of zero is
higher than the return on one-month T-bills.
– Therefore, the risk premium (slope) is smaller than the
CAPM model would predict.

FM 300 Lecture Note 2 34

 …Testing CAPM

The relationship for CAPM is weaker since the mid-

1960s, especially for high-beta portfolios.
1931-1965 1966-2005
30 30
Avg Risk Premium, %

SML
Portfolios 1,2,3, ….
20 20

10 10
Market portfolio

0 0

1 β 1 β
Based on: F. Black, Beta and Return, Journal of Portfolio Management, 1993

FM 300 Lecture Note 2 35

Summary

Strong evidence that expected returns increase with risk.

Expected excess returns are approximately linear in beta.
Risk-return relation is flatter than what CAPM predicts:
– high beta stocks have lower returns relative to the CAPM;
– low beta stocks have higher returns relative to the CAPM.

FM 300 Lecture Note 2 36

 Roll’s critique (1977)
Main idea: a test of the CAPM that does not use the true
market portfolio is not a test of the CAPM.
For any efficient portfolio of stocks P we may write:
(1) W Xn = W Xo + !n (W Xp − W Xo )
where rZ is the return on a portfolio such that: cov(rz, rP)=0
and !n =cov(Xn ,Xp )/var(Xp ) is beta with respect to portfolio P.
Roll pointed out that previous tests might be testing equation
(1) rather than CAPM.
This is just due to the property of the efficient frontier.
Suppose, the true market portfolio is M*, but instead we use
an approximate efficient portfolio M.

FM 300 Lecture Note 2 37

Roll’s critique (1977)

Then, we have the following two equations for return rj:
(2) W Xn = W Xo + !n (W X[ − W Xo )

(3) W Xn = XZ + !n∗ (W X[∗ − XZ )

Equation (3) is exact CAPM. Finally, CAPM for portfolio z:

(4) W Xr = XZ + !o∗ (W X[∗ − XZ )
Substituting (4) into (2), after some algebra, we find:
W Xn = XZ + !o∗ W X[∗ − XZ + !n (W X[ − XZ − !o∗ (W X[∗ − XZ ))

N^ N5
Hence, regressing returns on betas we get biased coefficients.
In particular N^ >0.
FM 300 Lecture Note 2 38
 Roll’s critique (1977)
Moreover, equation (1) shows that the average return on an
asset or portfolio is a linear function of beta, if betas are
computed using any ex-post efficient portfolio.
Note that (1) is not an equilibrium equation.
Hence, if one chooses an ex-post efficient portfolio one may
wrongly accept the CAPM, even if the true market portfolio is
inefficient.
If the portfolio used is not efficient then the return is not a
linear function of beta.
Hence, if one chooses an ex-post inefficient portfolio as an
approximation to the market portfolio one may wrongly reject
the CAPM.
FM 300 Lecture Note 2 39

Roll’s critique (1977)

A test of the CAPM that does not use the true market portfolio
is not a test of the CAPM. It is only a test of whether the
chosen index is ex-post efficient.
The conclusion from Roll’s critique is that the CAPM is not
testable!
We need to include ALL assets in the sample. Using a proxy
like the S&P500 is not good enough.
The market should include bonds, real estate, foreign assets,
human capital, …

FM 300 Lecture Note 2 40

 Summary of empirical evidence

Why do tests reject CAPM?

– Theory says that true betas are related to average returns.
– We regress average returns on estimated betas.
– Estimation error in betas may explain part of this problem
– We use a proxy for the market return (the “Roll critique”)
– Market should include bonds, real estate, foreign assets,
human capital, …

FM 300 Lecture Note 2 41

 FM 300 Section A

Empirical Evidence of the CAPM and

Multifactor Models
Lecture Note 3
Dr. Georgy Chabakauri
London School of Economics

FM 300 Lecture Note 3 1

Outline
More tests of CAPM
Multifactor models
– Introduction
– Arbitrage Pricing Theory (APT)
– Macroeconomic factors
– The three-factor model

FM 300 Lecture Note 3 2

 More tests of the CAPM

Consider a second pass regression similar to the one in

Lecture 2 where we additionally include squared beta and a
measure of idiosyncratic risk:

rit – rft = g 0 + g 1bˆi + g 2 bˆi + g 3sˆ i + ei

2

CAPM implies that:

g 1 = rMt – rft g0 =g2 =g3 = 0

Fama and MacBeth find that g2 and g3 are insignificant, but
that the empirical SML is still too flat.

FM 300 Lecture Note 3 3

Betas, Size, and Book-to-Market

In fact, any number of variables could be added to the

regression:

rit – rft = g 0 + g 1bˆi + g 2 X i + ei

CAPM implies g2 = 0.
This test is immune to Roll’s critique.

FM 300 Lecture Note 3 4

 Betas, Size, and Book-to-Market

Banz (1981) adds firm size to the regression and finds a

negative g2.
Fama and French (1992) add firm size and the ratio of book
value to market value (B/M). They find that both are
significant but that beta is not!
In explaining the cross-sectional variations of average
returns,
– beta does not seem to be enough.
– Size and book-to-market dominate beta.

FM 300 Lecture Note 3 5

Fama and French (1992)

Since size and beta are highly correlated empirically, Fama
and French (1992) do a double sort, first on size, then on
market β:
– This helps detect variation in β unrelated to size;
– Split stocks up into ten size decile portfolios;
– Break each size decile portfolio up into ten sub-portfolios
sorted on pre-formation (preliminary) β’s;
– To determine β of portfolio, use time-series of portfolio
returns, post-formation (post-β).
Next table shows that keeping size decile fixed, variation in
beta produces little variation in return. Moreover, higher-β
stocks do not have higher returns, contrary to CAPM.

FM 300 Lecture Note 3 6

 … Fama and French (1992)

Average returns for portfolios formed on size and beta: July

1963-Dec 1990

FM 300 Lecture Note 3 7

… Fama and French (1992)

Similar results hold for portfolios sorted on BM and beta.
More formally, they use Fama-MacBeth procedure to
study how returns depend on beta, size, and BM.
Beta is not significant in 1963-1990 sample even when
size and BM are not included in the regression.
Size and BM are highly significant.
Table below reports the coefficients from regression of
average stock returns on variables such as β, log size
(ME), and log book-to-market (BE/ME).

FM 300 Lecture Note 3 8

 … Fama and French (1992)

FM 300 Lecture Note 3 9

Betas, Size, and Book-to-Market

Size effect (E[rsmall] – E[rlarge]) is found in

– Japan: 14.4% per year
– France: 10.8% per year
– Finland: 9.1% per year
– Taiwan, Spain, Belgium, Switzerland, New Zealand,
Germany, Iceland, Canada, and the UK: between 4.8%
and 4.8% per year
In all studies, beta alone cannot explain differences in returns.
[from Hawawini and Keim (1992)]

FM 300 Lecture Note 3 10

 Betas, Size, and Book-to-Market

Firms with a high ratio of book value to market value (B/M) tend
to earn a higher rate of return than predicted by the CAPM.

Book-to-market effect has been found in France, Germany,

Japan, Switzerland, and the U.K.
High B/M firms are “value” stocks.
– lots of tangible assets in place; cash flows are stable;
possible history of bad returns
Low B/M firms are “growth” stocks
– few physical assets; cash flows volatile and expected to
grow; also called “glamour” stocks

FM 300 Lecture Note 3 11

Risk Proxies

The existence of other variables besides beta that influence

expected returns challenges the CAPM.
Do size, B/M, and other variables represent proxies for risk
factors?
– If yes, then some other model relating risk and return may
be appropriate.

FM 300 Lecture Note 3 12

 Multifactor models

Recall the SML equation:

E (ri ) = rf + b i (E (rM ) - rf ) = rf + b i RPM

What if stock i is exposed to more risk factors than just the

market?
Then the contribution of stock i to the risk of a portfolio would
be measured not only by the beta with respect to the market,
but also by the betas relative to the other factors’ risk premia.

FM 300 Lecture Note 3 13

… Multifactor models

Examples of other (potential) factors:

– Gross domestic product
– Interest rates
– Expected inflation
The pricing equation becomes:

E (ri ) = rf + b i ,M RPM + b i ,GDP RPGDP + bi , IR RPIR + ...

K
= rf + å b i ,k RPk
k =1

FM 300 Lecture Note 3 14

 Arbitrage Pricing Theory
Developed by Ross (1976). Main assumptions:
– Returns can be described by factors;
– Idiosyncratic risk is diversified;
– There are no arbitrage opportunities.
An arbitrage opportunity arises when an investor can make
sure profits without making a net investment – i.e., an investor
could construct a zero investment portfolio with a sure profit.
Since no investment is required, an investor can create large
positions to secure large levels of profit.
Arbitrage Pricing Theory (APT), like CAPM, also stipulates a
relationship between expected return and risk, but it uses
different assumptions and techniques.
FM 300 Lecture Note 3 15

… Arbitrage Pricing Theory

Portfolios vs. zero-investment portfolios.
− A portfolio has weights that sum to 1.
− A zero investment portfolio, or arbitrage portfolio, has
weights that sum to 0.
Zero investment portfolios
− require no initial net investment
− have initial “size” defined by the sum of all long positions
− factor exposures: their beta with respect to factor k is the
weighted average of the sensitivities of the securities in the
portfolio to factor k, Si wi bki.

FM 300 Lecture Note 3 16

 … Arbitrage Pricing Theory

The APT makes a distinction between two types of risk:

– Factor risk is systematic and is impossible to diversify
away.
– Idiosyncratic risk is completely irrelevant for well-
diversified investors.

FM 300 Lecture Note 3 17

… Arbitrage Pricing Theory

The APT assumes:

1. !"# -!$ = &" + ()" *)# +…+ (+" *+# + ,"#

2. Residuals ," are uncorrelated across assets. This is not

an assumption of the CAPM.
3. Returns are possibly non-normal (which CAPM usually
assumes).
4. Investors might care about more than just mean and
variance (In CAPM investors care only about mean
and variance)

FM 300 Lecture Note 3 18

 … Arbitrage Pricing Theory

The CAPM is preference-based:

– If investors care about their portfolio’s mean and variance,
they dislike assets that increase the variance. Hence they
demand a higher expected return for investing in them.
The APT is arbitrage-based:
– If expected return was not related to factor betas, then we
would be able to find arbitrage opportunities.
– APT does not tell us where the factors come from

FM 300 Lecture Note 3 19

… Arbitrage Pricing Theory

Multiple types of systematic risk:
– Macro factors: GNP growth, inflation;
– Security factors: Market risk, small stock risk, value versus
growth risk.
Betas can be estimated as regression coefficients.

FM 300 Lecture Note 3 20

 … Arbitrage Pricing Theory

Consider an example of a one-factor model.

Suppose, asset returns are given by:
!" − !$ = &" + (" !0# − !$ + ,"#

APT result: for most assets & ≈ 0

Result is similar to CAPM but is derived under different

assumptions and using different techniques.

FM 300 Lecture Note 3 21

… Arbitrage Pricing Theory

The intuition is as follows. Assume for a moment that ,"# =0.
Then, if &" >0, we can make arbitrage:
– Consider the following portfolio that consists of a risk-free
asset and the market portfolio with weight (" :

!1 = 1 − (" !$ + (" !0 = !$ +(" (!0 − !$ )

– Asset j has return
!" = !$ + &" + (" !0# − !$

– Consider an arbitrage portfolio that shorts p and goes

long j. This portfolio will have return &" >0.

No-arbitrage implies &" =0.

FM 300 Lecture Note 3 22
 … Arbitrage Pricing Theory

Suppose, ,"# ≠ 0. Then, exact arbitrage not possible.

However, idiosyncratic risk is irrelevant for well-diversified

investors and is not compensated.
Consider a well-diversified portfolio A of assets with & > 0
and a well-diversified portfolio B of assets with & ≤ 0.
Idiosyncratic risks are (almost) diversified away.
Using similar strategy as on the previous slide we will get
arbitrage. Hence, by no-arbitrage, & ≈ 0.

FM 300 Lecture Note 3 23

… Arbitrage Pricing Theory

Here is the arbitrage strategy for well-diversified portfolios
that have different alphas:
– Find two well-diversified portfolios with different
expected returns but the same factor betas.
− Buy the portfolio with the high expected return and
short the same amount of the other.
− Offsetting betas eliminates systematic risk.
– We are left with a zero-cost, zero-variance position
with positive mean.
Idiosyncratic risk is irrelevant for well-diversified investors
and is not compensated.

FM 300 Lecture Note 3 24

 … Arbitrage Pricing Theory
Example:
rf = 4%; Factor F = 6%; Portfolio C with βC = 0.5 and E[RC] = 6%.
C does not belong to the set of portfolios described by the factor
return! It lies below the line 4%+0.5(6%).
We can find an arbitrage opportunity. Take a portfolio whose
returns are described by factor F. For example, portfolio A with βA =
1, EA=4% + βA(6%) = 10%.
Create portfolio D, a combination of portfolios A and RF, so that it
has the same beta as C:
βD = wA βA+(1-wA) βRf = βC = 0.5. So wA = 0.5.
E[RD]= 0.5(10%) + 0.5(4%) = 7%.
Now buy D and short C for the same amount. E.g. $1m.
Profit = $10,000, Arbitrage return = 1%.
FM 300 Lecture Note 3 25

… Arbitrage Pricing Theory

E(r)% Short portfolio C

Use funds to construct D
(comprised of A and
10 risk-free asset).
A
D Arbitrage profit = 1%
7
6 C

risk-free 4

.5 1.0 Beta for F

FM 300 Lecture Note 3 26

 … Multifactor Pricing Models

First, specify the factors and estimate the sensitivities:

– Construct a series of factors Fkt through time;
– Examples of factors: macroeconomic and financial
variables (e.g., Chen, Roll, and Ross 1986); return
spreads between index portfolios formed according to
cross-sectional sorts (e.g., Fama and French, 1993)
Estimate the sensitivities bik for each asset i by running a
time-series regression of Rit on the Fkt.
We will discuss two examples.

FM 300 Lecture Note 3 27

… Macroeconomic factors
Important factors:
1. monthly growth rate of industrial production;
2. change in expected inflation;
3. unexpected inflation;
4. change in risk premium - measured as the difference in
returns between junk bonds (rated below Baa) and long-
term Treasury bonds;
5. change in the term structure of interest rates - measured
as the difference in returns between long-term Treasury
bonds and bills;
6. the return on the equally weighted portfolio of NYSE
stocks is also included as a sixth factor.
FM 300 Lecture Note 3 28
 The Three-Factor Model

The Fama and French (1993) three-factor model is the most

popular multifactor model. The three factors are:

1. rM–rf : the market return minus the riskless rate;

2. SMB= “Small Minus Big”: return on small-cap stocks

(bottom 50%) minus return on large-cap stocks (top 50%).

3. HML= “High Minus Low”: the return on high book-to-market

stocks (top 30%) minus the return on low book-to-market
stocks (bottom 30%). All three are zero-investment
portfolios.

FM 300 Lecture Note 3 29

… The Three-Factor Model

SMB (small minus big) and HML (high minus low) portfolios:
Sort all firms independently into two size groups (S and B) and
three book-to-market groups (L, M, and H);
Form six value-weighted portfolios, LS, MS, HS, LB, MB, and
HB, as the intersections of the size and book-to-market groups;
HML is constructed to be neutral with respect to size:
– HML = (HS + HB)/2 − (LS + LB)/2
SMB is constructed to be neutral with respect to book-to-
market:
– SMB = (HS + MS + LS)/3 − (HB + MB + LB)/3

FM 300 Lecture Note 3 30

 The Three-Factor Model
1. For each stock, run the regression
rit–rft = ai + bi (rMt–rft) + si RSMBt + hi RHMLt + eit
2. The pricing relation:
E[ri–rf]= bi E[(rM–rf)] + si E[RSMB]+ hi E[RHML]
3. An equivalent representation of the pricing relation is
ai = 0
where ai is the intercept in the time-series regression.

FM 300 Lecture Note 3 31

The Three-Factor Model

Slope estimates for: rit–rft = ai + bi (rMt–rft) + si RSMBt + hi RHMLt + eit
B/M quintile

32
FM 300 Lecture Note 3
 The Three-Factor Model
Alpha estimates for: rit–rft = ai + bi (rMt–rft) + si RSMBt + hi RHMLt + eit

B/M quintile

(t-statistics)

FM 300 Lecture Note 3 33

The Three-Factor Model

Alpha estimates for: rit–rft = ai + bi (rMt–rft) + eit

B/M quintile

(t-statistics)

FM 300 Lecture Note 3 34

 CAPM vs Fama-French

The figure plots average actual returns vs returns predicted by CAPM and
the FF model for 25 size and book-to-market double-sorted portfolios. FF
has better fit because portfolios on panel B are along 45-degree line.
Source: Goyal (2012), see also BKM
FM 300 Lecture Note 3 35

The Three-Factor Model

Fama and French (1993) find that SMB and HML factors are
priced, and the loadings vary a lot across portfolios sorted on
size and B/M.
– Small stocks load positively on SMB
– Value stocks (high B/M) load positively on HML
The intercepts are generally close to zero for most assets
(sorted portfolios)
– Except for smallest and largest growth firms
Their results are replicated in cross-country studies, and in
data before 1963

FM 300 Lecture Note 3 36

 The Three-Factor Model

The three Fama-French factors are empirically motivated

– they were chosen on the basis of variables that, based on
past evidence, seem to predict high average returns;
– thus may be capturing risk-premia;
Are they proxies for yet unknown risk factors?
– Maybe.
Data-snooping problem
– Unlikely.

FM 300 Lecture Note 3 37

The Five-Factor Model

Fama and French (Journal of Financial Economics, 2015)

propose a five-factor model:
– market, HML, SMB + RMW, CMA;

old factors profitability and investment

– RMW is difference between returns on diversified portfolios

of stocks with robust and weak profitability;
– CMA is the difference between returns on diversified
portfolios of stocks of conservative and aggressive
investment firms.

FM 300 Lecture Note 3 38

 … The Five-Factor Model
Average excess monthly returns, July 1963-Dec 2013

OP-operating profitability; Inv - % change in total assets

FM 300 Lecture Note 3 39
 FM 300 Section A

Market efficiency:
Empirical tests and evidence
Lecture Note 4
Dr. Georgy Chabakauri
London School of Economics

FM 300 Lecture Note 4 1

Outline

The Efficient Market Hypothesis

Weak form efficiency
– Serial correlation in stock returns
– Technical analysis
Semi-strong form efficiency
– Event studies
Strong form efficiency

FM 300 Lecture Note 4 2

 The Efficient Market Hypothesis

Efficient Market Hypothesis (EMH):

– Fama (1991): “I take the market efficiency hypothesis to be
the simple statement that security prices fully reflect all
available information.”
– A market is efficient with respect to a given information set
It if no investor can make economic profits by trading on
the basis of It. (Malkiel 1992)
– All stocks are fairly valued at all times based on all
currently available information.

FM 300 Lecture Note 4 3

The Efficient Market Hypothesis

Prices reflect all available information

The market responds only to new information
It is not possible to forecast price changes
– Prices react to information
– Information is unpredictable (news)
– Therefore, price changes are unpredictable

FM 300 Lecture Note 4 4

 Why should markets be efficient?
There are many sophisticated investors (arbitrageurs) who
trade on information and eliminate inefficiency—they eliminate
any abnormal returns (more in Lecture 6).
These investors have unlimited capital and unlimited risk
capacity, so they compete fiercely for any profitable opportunity.
Competition assures that prices reflect all available information.

FM 300 Lecture Note 4 5

Efficient Markets Hypothesis

In efficient markets predictable trends in stock prices will be
quickly eliminated by investors.

$90
Microsoft Stock Price

Predicted
price of $90

Actual price as soon as

70
upswing is recognized

50

Last This Next

Month Month Month

FM 300 Lecture Note 4 6

 Two conflicting views

A Random Walk Down Wall Street, by Burton Malkiel:

– Stock returns are unpredictable, at least in the short run;
– Portfolio managers provide no benefit;
– Best strategy is to reduce costs by buying index.
A Nonrandom Walk Down Wall Street, by Andrew Lo and
Craig MacKinlay:
– There are predictable components in stock returns;
– Profit opportunities may exist.

FM 300 Lecture Note 4 7

Three forms of Market Efficiency

Economists distinguish three forms of market efficiency:
– Weak form: cannot achieve abnormal returns given observed
history of past stock prices, trading volume and other trading
data.
– Semi-strong form: cannot achieve abnormal returns given all
publicly available information (quality of management, balance
sheet composition, accounting practices etc.).
– Strong form: cannot achieve abnormal returns given publicly
and privately available information. Note however, that insider
trading by corporate officers, directors and their relatives is
illegal.

FM 300 Lecture Note 4 8

 Information sets
The different forms of market efficiency are a result of
changing the definition of “available information” It :
– Weak form: It = history of prices, returns, trading volumes;
– Semistrong form: It = public information;
– Strong form: It = public and private information.
If the market is strong-form efficient, it is also semi-strong and
weak form efficient:
– Strong form => Semi-strong form => Weak form.

FM 300 Lecture Note 4 9

Weak Form Efficiency

A market is weak form efficient if current prices reflect all

information contained in past prices and price movements.
Implications
– Past prices cannot predict price movements in the future.
– Trading rules based on technical analysis cannot yield
superior returns.
Tests
– Tests of technical trading rules.
– Tests of correlation of returns.

FM 300 Lecture Note 4 10

 Technical analysis
Is based on the belief that historical price series exhibit
regularities (patterns) that can be profitably exploited to
forecast future price movements. For this reason technical
analysts are often called chartists.
Examples:
– Support level: A barrier that the stock price has not fallen
below and is unlikely to fall below in the future.
– Resistance level: A barrier that the stock price has not
risen above and is unlikely to eclipse in the future.

FM 300 Lecture Note 4 11

Technical analysis

Brock, Lakonishok, and LeBaron (1992) test two popular

decision rules and find support for some technical rule:
– moving average (buy when the short-period MA rises above
the long-period MA and vice versa);
– trading range break (buy when the price rises above its last
peak and vice versa);
– However, these strategies generate high trading costs.

FM 300 Lecture Note 4 12

 Filter rules
Example of an x% filter rule: if the daily closing price of a particular
security moves up by at least x%, buy and hold the security until its
price moves down at least x% from a subsequent high, at which time
simultaneously sell and go short. The short position is maintained
until the price rises at least x% above subsequent low, at which time
cover the short position and buy.
Fama and Blume (1966) present a more detailed analysis of filter
rules (correcting for dividends and trading costs), and conclude that
such rules do not perform as well as the buy-and-hold strategy.
Very small filters (0.5%-1.5%) yield superior returns, but generate
frequent trading. They show that even low transaction costs are
enough to eliminate the profits from such filter rules.

FM 300 Lecture Note 4 13

Filter rules

FM 300 Lecture Note 4 14

 Review of serial correlation
Consider the random variable rt, the time series of returns.
Serial correlation (autocorrelation) is defined as:

Cov ( rt , rt -k )
Corr( rt , rt -k ) = , where
Var( r )

Cov ( rt , rt -k ) = E [{rt - E ( r )}{rt -k - E ( r )}]

The autocorrelation summarizes the manner in which current

rt and past rt-k values of the same variable tend to move in
relation to one another.

FM 300 Lecture Note 4 15

Review of serial correlation

These studies are based on a simple regression of the form:
"# = ' + !"#$% + )#
Statistically significant ! implies predictability of returns.
Positive ! means positive correlation between "# and "#$% :
– If "#$% is low (high) then "# tends to be low (high);
Negative ! means negative correlation between "# and "#$% :
– If "#$% is low (high) then "# tends to be high (low);

FM 300 Lecture Note 4 16

 Weak Form
Scatter of stock returns for S&P500 and FTSE 100 stock indices.
Each dot shows returns on week t and week t+1. No predictability.
S&P Composite FTSE 100
(correlation = -.07) (correlation = -.08)

Return in week t + 1, (%)

Return in week t + 1, (%)

Return in week t, (%) Return in week t, (%)

Source: BMA textbook

FM 300 Lecture Note 4 17

Serial correlation in returns

Medium horizons (1-12 months)

Returns exhibit positive serial correlation:
– winners stay winners, losers stay losers;
– momentum strategies are profitable.
Long horizons (over 1-5 years)
Returns exhibit negative serial correlation called reversals:
– winners become losers, and vice versa.
We will talk more about momentum and reversal in
Lectures 5 and 6.

FM 300 Lecture Note 4 18

 Weak Form: Summary
Implications:
Technical rules are useless
E.g., if the price of a stock has just increased or decreased, it
does not follow that it will increase or decrease in the future
If technical rules worked, everyone would use them. As a
result they would not work anymore.
Evidence:
Small serial correlation in stock returns at short horizons
Larger serial correlation at longer horizons (we will examine
this evidence in Lecture 5)

FM 300 Lecture Note 4 19

Semi-Strong Form Efficiency

A market is semi-strong form efficient if all publicly
available information is reflected in market prices
Implications:
Prices react immediately to economic events
No systematic under/overshooting after announcement
Examples of events:
– Macroeconomic news
– Firm-specific news: earnings announcements, mergers
and acquisitions, issues of new debt or equity, dividend
announcements, accounting changes…
Tests:
Fundamental analysis, event studies of reactions to news.
FM 300 Lecture Note 4 20
 Fundamental analysis

Fundamental analysis is the practice of using firm earnings,

dividends, and other sources of accounting and financial data
to forecast future returns of the stock.
For example, we may attempt to identify and buy firms whose
earnings are likely to be higher than the industry forecast.
The success of fundamental analysis would be a violation of
the semi-strong form of market efficiency.

FM 300 Lecture Note 4 21

Event studies: an example

Suppose that you wish to check for the effect of an analyst
downgrade on stock returns
Under efficient markets:
– What should the contemporaneous effect of the
downgrade be?
– What should the future effect of the downgrade be?
After the close of trading on September 28, 2000, Cisco
was downgraded to “market perform” from “outperform” by
analyst Paul Sagawa of the investment firm Sanford C.
Bernstein.

FM 300 Lecture Note 4 22

 Event studies: an example

Is this an evidence that the downgrade affected returns?

– Difficult to say…
– What if the whole market did poorly on that day?
– Need a methodology to identify the impact of news.
FM 300 Lecture Note 4 23

Event studies methodology

Are there abnormal returns following a news
announcement?
– ARit = Rit − E[Rit|It], where Rit = realized excess return.
How do we calculate expected returns?
For example, using the CAPM regression:
",# − "3# = +, + ., "5# − "3# + ),#
Suppose, we have estimates +* , and .-, before the event.
Abnormal returns are obtained from:
̂ = ",# − "3# − (+* , + .-, ("5# − "3# ))
/0,# = ),#

FM 300 Lecture Note 4 24

 Event studies methodology

Define the event window

– Consider the event of interest and identify the period over
which security prices of the firms involved will be
examined.

FM 300 Lecture Note 4 25

Event studies methodology

Estimate the expected return model
– Estimate the parameters of the “normal” (expected)
return model using data from the estimation window
Measure Abnormal Returns in the event window
– The abnormal return is the realized return of a security
during the event window, after subtracting the normal
return
Aggregate abnormal returns in the event window to
compute cumulative abnormal return for company i:
:<

7/0, = 8 ),#
̂
#9:;
FM 300 Lecture Note 4 26
 Event studies: an example
Consider the Cisco example. The negative return on the day of
the downgrade was –7.4% but this might not be surprising:
– NASDAQ composite fell about 3% on the same day;
– Cisco has a beta of 1.31 with respect to NASDAQ.
The estimated single factor model is
RCSCO,t = .001 + 1.31 RNASDAQ,t + εt, where σ(εt) = .035

Given the market return of –3%, the abnormal return at time t is:
ARt = RCSCO,t – [.001 + 1.31 RNASDAQ,t]
= –.074 – [.001 + 1.31 (–.03)] = –.036.
The standard error of the firm-specific component is .035. Can’t
say whether the downgrade had any significant effect.

FM 300 Lecture Note 4 27

Event studies: an example

We can increase the power of our test by considering a set of
companies that are affected by the same type of event.
– Cisco downgrade represents just one out of thousands of
similar events.
To do this, we will have to average results over many different
firms with downgrades at different times.
– Cisco downgrade was on 28 Sep 2000;
– Yahoo! was downgraded on 11 Oct 2000.
Averaging these results diversifies away idiosyncratic noises
unrelated to the event.

FM 300 Lecture Note 4 28

 Event studies: an example
Time is measured relative to the downgrade date:
– t=0 denotes the day of the downgrade for stock i.
– We will look at cumulative returns starting, for example,
100 days prior to the announcement date and 50 days
following the announcement.
For stock i, the CAR from 100 days before to 50 days after the
downgrade is:
>=

7/0, = 8 ),#
̂
#9$%==

FM 300 Lecture Note 4 29

Event studies methodology

Aggregate abnormal returns across individual firms/events

N
1
CAR =
N
å CAR
i =1
i

Test if the event has price impact, by looking at the magnitude

of abnormal returns. The null hypothesis is that CAR=0.
Test if prices reflect new information immediately:
– For each day in the event and post-event window, plot the
average CAR across firms.

FM 300 Lecture Note 4 30

 Event studies methodology

CAR plots as functions of time reveal the timing and impact

of information releases on prices.
A jump in CAR on the event date is consistent with an
information release.
A trend in CAR after the event date is a sign of market
inefficiency.
Any change in CAR prior to the event date is a sign of
information leakage.
If prices underreact or overreact, then price changes can be
predicted using publicly available information!

FM 300 Lecture Note 4 31

Event studies: jump

FM 300 Lecture Note 4 32

Event studies: underreaction

FM 300 Lecture Note 4 33

Event studies: overreaction

FM 300 Lecture Note 4 34

 Event studies: information leakage

FM 300 Lecture Note 4 35

Semi-Strong Form

Example: In most takeovers the acquiring firms pay premium

over current market prices to acquire a firm.
The announcement of takeover increases the stock prices.
– Keown and Pinkerton (1981) study cumulative abnormal
returns before and after takeover attempts for 194 firms.
– They show that the new information is very quickly
incorporated by the end of the trading day.
– Moreover, there is an evidence of information leakages
generated by insiders.

FM 300 Lecture Note 4 36

 Semi-Strong Form
Price response of 194 takeover targets.
– Prices increased before date 0 due to information leakages.

Announcement Date
39
Cumulative Abnormal Return
34
29
24
19
(%)

14
9
4
-1
-6
-11
-16
Days Relative to annoncement date

FM 300 Lecture Note 4 37

Semi-Strong Form

Carhart (1997) shows that mutual funds fail to provide higher

returns than stock market indices.
Average Annual Return on 1493 Mutual Funds and the Market Index

40
30
20
Return (%)

10
0
-10
-20 Funds
Market
-30
-40
62

77

92
19

19

19

Source: Carhart (1997) and BMA textbook

FM 300 Lecture Note 4 38

 Semi-Strong Form

Sometimes mutual funds beat the market.

– However, their superior performance is not very persistent:
past winners may become losers next year.
– Very few mutual funds over-perform consistently.
As a result, some funds turned into index funds that invest in
stock market index.
– They “buy the index” in such a way as to minimize the cost of
managing the portfolio.
Inability to beat an index is an evidence in favour of EMH.

FM 300 Lecture Note 4 39

Semi-Strong Form
Mutual funds are unable to earn abnormal returns because
managers compete fiercely against each other.
Hence, it should not be surprising that smart managers (e.g.,
mutual funds, hedge funds) cannot earn large abnormal returns,
because if they do, investors will give them more money to
manage, hence the more fierce the competition for abnormal
returns.

FM 300 Lecture Note 4 40

 Semi-Strong Form: Summary
Implications
Fundamental analysis is valueless
– Unless it incorporates private information.
Prices respond to news quickly
Evidence
Rapid adjustment to new information.
Pre-announcement information leakage.
In general, evidence favors market efficiency (e.g. takeover
and dividend announcements). There are exceptions.

FM 300 Lecture Note 4 41

Strong Form Efficiency

A market is strong form efficient if all relevant information

(public or private) is reflected in market prices.

Implications:
No profits from insider trading.
Tests:
Profitability of trading on inside information.
E.g., corporate insiders.

FM 300 Lecture Note 4 42

 Strong Form

Strong form efficiency: cannot achieve abnormal returns given

publicly and privately available information.

Theoretically, there are no reasons to believe in strong form

efficiency because corporate insiders have superior information
which is not available to outside investors, although trading on
insider information is illegal.

For example, they can profit on various announcements.

FM 300 Lecture Note 4 43

Joint Hypothesis

Fama (1991): “… market efficiency per se is not testable.

It must be tested jointly with some model of equilibrium, an
asset-pricing model.”
Every test of market efficiency is a joint test of
– The efficiency of the market
– The asset pricing model used to calculate expected
returns
If the model of computing E[Rit|It] is incorrect we may find
large abnormal returns ARit = Rit − E[Rit|It].

FM 300 Lecture Note 4 44

 Conclusion on market efficiency

In general, empirical evidence supports the semi-strong form

of market efficiency:
– Most public information is taken into account.
However, violations of the weak form are sometimes
observed.

FM 300 Lecture Note 4 45

 FM 300 Section A

Investment anomalies and

Return predictability
Lecture Note 5
Dr. Georgy Chabakauri
London School of Economics

FM 300 Lecture Note 5 1

Outline

Definition of anomalies
Some important anomalies:
– Reversal and Momentum effect;
– Post-earnings announcement drift;
– Market predictability;
– Size and Value;
An introduction to behavioural finance

FM 300 Lecture Note 5 2

 What is an anomaly?

Definition: Anomaly is an empirical result that seems to be

inconsistent with maintained theories of asset-pricing.
Investment anomalies suggest either profit opportunities or
the need for better models (or better tests) – recall that any
test of market efficiency is a joint hypothesis test.
To be sure we found an anomaly we need to study data from
independent samples (e.g., other periods or other countries).
If results disappear, it could reflect data-mining in the
discovery of the anomaly, or it could mean that the market
learned and competed it away.
If results persist, it could reflect risk or market frictions.

FM 300 Lecture Note 5 3

Puzzles and Anomalies

In 1978, Michael Jensen wrote:
I believe there is no other proposition in economics which has more
solid empirical evidence than the Efficient Markets Hypothesis.
However, in the following decades, there is mounting evidence
that challenges the efficient market hypothesis.
We will focus on the following:
– Reversals and Stock Price Momentum;
– Post Earnings Announcement Drift;
– Bubbles, Twin-Stock Puzzle, etc.

FM 300 Lecture Note 5 4

 Reversals
There is abundant evidence that returns are negatively
autocorrelated at long horizons (3-5 years).
DeBondt and Thaler (1985) find that “trends” in stock returns
reverse over long periods.
– Stocks are allocated to portfolios based on their market-
adjusted cumulative abnormal returns during the previous
3-5 years;
– Assign the top 35 stocks (or the top 50 or top decile) to the
winner portfolio and the bottom 35 (or the bottom 50 or
bottom decile) to the loser portfolio;
– Measure the performance of the extreme portfolios over
the following 1-5 year periods.

FM 300 Lecture Note 5 5

… Reversals
Over the last 50 years, loser portfolios of 35 stocks
outperform winner portfolios of 35 stocks by 25% in the 3-year
period following portfolio formation.
Most of the trading gains come from the loser portfolios
(losers outperform the market by 19.6%, and winners
underperform by 5%).
So, losers rebound and winners fade back.
Might be because of market overreaction to news. After the
overreaction is recognized, investment performance reverses.
Strategy: invest in recent losers, avoid recent winners.

FM 300 Lecture Note 5 6

… Reversals

FM 300 Lecture Note 5 7

… Reversals

FM 300 Lecture Note 5 8

 Momentum
Jegadeesh and Titman (1993) found that recent past winners
outperform recent past losers.
Each month t rank stocks on the basis of returns over the
ranking period of past J months, where J = 3, 6, 12 months.
Form equally weighted portfolios of winners (top 10%) and
losers (bottom 10%).
Zero-investment momentum strategy: each month t buy
winners and shorts losers.
Hold this position for a holding period of K months, where K = 3,
6, 12 months.
Earn abnormal returns: winners tend to have positive alpha,
losers tend to have negative alpha.
FM 300 Lecture Note 5 9

Momentum

Original sample: NYSE-AMEX stocks, 1965-1989.

Average returns are reported for various combinations of the
ranking periods J and holding periods K.
Table below reports the returns of buy (winners), sell (loser),
momentum (buy-minus-sell) portfolios.
The J = 6 and 9 month formation periods produce returns of
about 1% per month regardless of the holding period.

FM 300 Lecture Note 5 10

 … Momentum

FM 300 Lecture Note 5 11

… Momentum

Jegadeesh and Titman 2001 update their initial sample with

more recent data: NYSE, AMEX, NASDAQ stocks, 1965-1998.
Consider 6-month ranking period and 6-month holding period.
Results confirm earlier findings: momentum profits are large
and significant, 1.23% per month in the overall sample (1.42%
for small stocks, 0.86% for large stocks).
They also look at post-holding period (more on this next week).

FM 300 Lecture Note 5 12

 … Momentum

FM 300 Lecture Note 5 13

… Momentum

Momentum portfolio returns have large t-statistics.

Table I shows monotone relationship between the returns of
portfolios and their performance rank.
Momentum holds for 1990-1998 data, hence:
– momentum was not eliminated by traders after its
discovery in the 1965-1989 sample;
– momentum is less likely to be a result of data mining.
Momentum holds for large cap stocks, which are more liquid
and can be traded with smaller transaction costs.

FM 300 Lecture Note 5 14

 … Momentum

FM 300 Lecture Note 5 15

… Momentum

Previous table shows that CAPM and 3 factor Fama-French

alphas are large and significant => these models cannot
explain momentum.
Momentum profits are still present after taking into account
transaction costs.
We will study some behavioural explanations for the
profitability of momentum strategies.

FM 300 Lecture Note 5 16

 Earnings Announcement Puzzle

Fundamental principle of EMH is that new information should be

quickly reflected in stock prices.

– Therefore, news should not generate predictable patterns.

Rendleman, Jones, and Latane (1982) calculate cumulative

abnormal returns for a large sample of firms around the
announcement days.

FM 300 Lecture Note 5 17

… Earnings Announcement Puzzle

They sort all firms into 10 groups according to the magnitude of
earnings surprises (i.e., earnings this year minus earnings last
year).
– As expected, on the day of announcement the abnormal
returns jump according to the size of surprise.
– However, prices continue to drift in the same direction for
many months (post-earnings announcement drift).
– One potential explanations is that investors underreact to the
information in earnings surprises (limited attention).

FM 300 Lecture Note 5 18

 … Earnings Announcement Puzzle
Post-announcement drifts:

Cumulative abnormal returns to announcements

Days from announcement

Source: Rendleman, Jones, and Latane (1982)

FM 300 Lecture Note 5 19

… Post-earnings announcement drift

Significant post-earnings announcement price drift in the

direction of the earnings surprise, also called earnings
momentum
The absolute magnitude of the drift is larger for small firms
and the drift continues beyond the 60-trading-day interval.
Risk cannot explain the complete picture.
The dominant view is that an incomplete response to earnings
is the true cause of post-earnings-announcement drift.
This seems to contradict market efficiency, which says that
investors use all available information to price securities.

FM 300 Lecture Note 5 20

 Twin Stocks Puzzle
“Siamese twins” stocks: Royal Dutch and Shell
– RD incorporated in Netherlands, Shell in England;
– RD trades mostly in the Netherlands/US, Shell in UK.
Due to 1907 merger agreement, Royal Dutch got 60% and Shell
40% of dividends and earnings of the joint company.
– Efficient markets hypothesis implies that by no-arbitrage the
prices should be related as follows:
3
PRoyal Dutch = PShell .
2

– In reality, this is NOT what we see.

FM 300 Lecture Note 5 21

… Twin Stocks Puzzle

Picture below shows ln(PRoyal Dutch /(1.5 ´ PShell )).

Source: Froot and Dabora (1999)

FM 300 Lecture Note 5 22

 Bubbles and market efficiency

Tulip mania in 17th century Holland: tulip prices peaked at

several times the annual income of a skilled worker.
In 2007 NASDAQ index fluctuated at a level of around 2600
while its value around 2001 was around 5000.
– Hence, it seems that some companies, primarily internet
companies, were grossly overvalued in 2001.
– I.e. their market value exceeded the PV of dividends.
– Thus, there might have been a bubble in the market, i.e.
increases in prices not justified by predicted increases in
dividends.
Hard to believe that security prices represented unbiased
rational values of assets.

FM 300 Lecture Note 5 23

Dot Com Bubble

NASDAQ (monthly data):

5000

4500

4000

3500

3000

2500

2000

1500

1000

500

0
31/12/1997 15/05/1999 26/09/2000 08/02/2002 23/06/2003

Source: www.yahoofinance.com

FM 300 Lecture Note 5 24

 Size effect
Banz (JFE, 1981) and Reinganum (JFE, 1981) showed that
small capitalization firms on the NYSE earned higher average
returns than is predicted by the CAPM from 1936-75.
Dimensional Fund Advisors started a small cap index fund in
1982 to mimic the Banz strategy.
Quote: “Dimensional Investing is about implementing great
ideas in finance for our clients”.
DFA US 9-10 Small Company Portfolio had an alpha (risk
adjusted excess return) of 20 basis points from 1982-2002 -
About half as large as the estimate from 1936-75, and not
reliably greater than zero (t-stat = 0.67)
Thus, this anomaly might have disappeared after its discovery.

FM 300 Lecture Note 5 25

Book-to-market ratios

As we discussed, firms with the highest B/M ratio (value

stocks) have higher returns than firms with low B/M (growth/
glamour stocks).
10% of stocks with highest B/M have avg return 16.87%
10% of stocks with lowest B/M have avg return of 10.92%
This difference cannot be explained by beta.
Two explanations: either 1) high B/M firms are underpriced or
2) B/M is a risk factor affecting equilibrium stock returns.
Risk-based explanation (Fama and French, 1992, 1993):
– Value stocks offer higher returns because of higher risk,
e.g., distress risk;
FM 300 Lecture Note 5 26
 … Book-to-market ratios
Behavioral explanation (e.g., Lakonishok, Shleifer and Vishny,
LSV 1994; La Porta, Lakonishok, Shleifer and Vishny, LLSV
1997):
– “…Some investors tend to get overly excited about stocks
that have done very well in the past and buy them up, so
that these "glamour" stocks become overpriced.”
– “Similarly, they overreact to stocks that have done very
badly, oversell them, and these out-of-favor "value” stocks
become underpriced.”
– “Value stocks provide superior returns because the market
slowly realizes that earnings growth rates for value stocks
are higher than it initially expected and conversely for
glamour stocks ”
FM 300 Lecture Note 5 27

… Book-to-market ratios

If it is a risk story then value stocks should underperform in

some states of the economy.
– LSV 1994 show that this is not the case: value stocks do
not perform worse than growth stocks in bad states like
recessions.
– The difference in betas also cannot explain the difference
in returns of value and growth stocks.

FM 300 Lecture Note 5 28

 … Book-to-market ratios
LLSV 1997 test the expectations errors hypothesis.
– They allocate securities to portfolios according to proxies
for “value” (E/P, C/P, B/M), and track the performance of
these portfolios over 5 years.
– Value returns are significantly larger than glamour returns
for five years after portfolio formation.
– The difference is large at earnings announcements.
– Event returns for glamour stocks are significantly lower
than glamour returns on an average day.
– Hence, glamour stocks get negative surprise at
announcement days, consistent with behavioural story that
investors had wrong expectations.
FM 300 Lecture Note 5 29

Time-series predictability of returns

Consider the predictive regression
RM,t+1 = a + b Xt + εt+1
where RM,t+1 is the market excess return and Xt is a vector of
predictive variables known at time t.
Fama and Schwert (1977): negative relation between market
returns and short-term interest rates.
Campbell (1987): using spreads between long- and short-term
interest rates (term premium).
Fama and French (1988): using dividend yields on aggregate
stock portfolios.

FM 300 Lecture Note 5 30

 Calendar anomalies
January effect (turn of the year)
– Most of the premium for holding small cap or high B/M
portfolios comes in January.
Monday effect (turn of the week)
– Several studies have shown that returns on Monday are
lower than returns on other days of the week.
Turn of the month effect
– Stocks consistently show higher returns on the last day
and first four days of the month.

FM 300 Lecture Note 5 31

Introduction to behavioral finance

Behavioral finance addresses a number of issues:
– Helps identify scenarios in which markets are not perfectly
efficient (in the weak and semi-strong form sense).
– Provides explanations for market inefficiencies.
There are two main building blocks:
– Some investors in the market are affected by various
psychological/social biases and constraints;
– Limits to arbitrage due to market frictions, limited arbitrage
capital, limited arbitrage risk capacity.
Psychological biases, heuristics, or other forms of irrational
behavior can cause various investment anomalies such as.

FM 300 Lecture Note 5 32

 Example

Linda is 31 years old, single, outspoken, and very bright. She

majored in philosophy. As a student, she was deeply concerned
with issues of discrimination and social justice, and also
participated in anti-nuclear demonstrations.

Which of the following two statements is more likely to be true?

A: Linda is a bank teller and also active in the feminist movement.
B: Linda is a bank teller.

FM 300 Lecture Note 5 33

Psychological biases

Conservatism:
Individuals are slow to change their beliefs in the face of new
evidence. Conservatism implies underweighting of new
evidence.
Representativeness:
Tendency to identify a single uncertain event by the degree to
which it (i) is similar in essential characteristics to its parent
population, and (ii) reflects the salient features of the process
by which it is generated.
Example: when a company has a consistent history of
earnings growth over several years, investors might conclude
that it is representative of an underlying growth potential.
FM 300 Lecture Note 5 34
 Psychological biases
Overconfidence:
The most documented bias in experimental settings.
Investors overestimate their ability, or the precision of
their information.
Since people fail more often than they expect to, rational
learning over time would tend to eliminate overconfidence.

Biased self-attribution:
Individuals attribute good outcomes to their ability, and
bad outcomes to external circumstances (e.g., bad luck).
This mechanism hinders a rational learning process and
induces individuals to learn to be overconfident.

FM 300 Lecture Note 5 35

Psychological biases

People have limited attention/computational ability.

They often focus on a subset of all available information.
Theoretically, investors could potentially gather all possible
information about a firm and test all kind of statistical and
forecasting models to assess the value of the firm.
However, their time is limited and therefore they end up
making an investment decision based on limited information.

FM 300 Lecture Note 5 36

 FM 300 Section A

Behavioral finance and

limits to arbitrage
Lecture Note 6
Dr. Georgy Chabakauri
London School of Economics

FM 300 Lecture Note 6 1

Outline

Psychological Biases (and price implications)

– Underreaction and overreaction
Limits to Arbitrage
– Implementation costs
– Noise trader risk
– Riding the sentiment wave

FM 300 Lecture Note 6 2

 Under/over-reaction and stock returns

Underreaction: If investors underreact to news, prices adjust

slowly to the release of new information. Underreaction
implies return continuation, until prices fully incorporate all
available information.
Overreaction: If investors overreact to news, prices are
pushed beyond the levels warranted by fundamentals.
– Overreaction could imply return continuation at shorter
horizons;
– But it implies negative autocorrelations at longer horizons
(because prices eventually will reflect the fundamental
value of a stock and prices would go down).

FM 300 Lecture Note 6 3

Under/over-reaction and stock returns

Theoretical models that simultaneously account for investors’

under- and over-reaction to news:
– Daniel, Hirshleifer and Subrahmanyam (DHS 1998);
– Barberis, Shleifer, and Vishny (BSV 1998);
– Hong and Stein (HS 2000);
They argue that momentum and reversal are due to
behavioural biases in the process of interpreting information.

FM 300 Lecture Note 6 4

 Underreaction

Underreaction might be due to conservatism and limited

attention.
Conservatism: Investors underweight new information, and
rely too heavily on their prior beliefs.
Limited attention: Investors do not have time/resources to
take into account every piece of information.
– HS 2000: investors disregard information in prices, and
information diffuses slowly among investors. Gradual
information diffusion drives underreaction of prices.

FM 300 Lecture Note 6 5

Overreaction

Overreaction may arise because of representativeness bias.

Representativeness heuristic:
– BSV 1998: investors over-extrapolate from a sequence of
growing earnings, overreacting to a long trend.
– HS 2000: momentum traders disregard fundamental values
and extrapolate from past prices, pushing the price of
winners above their intrinsic value.

FM 300 Lecture Note 6 6

 DHS 1998
DHS 1998 propose a theory based on overconfidence about
the precision of private information and biased self-attribution.
Suppose, true asset value is !.
– At date 1 informed investors receive a noisy private signal
! + #, where # is noise with true variance $%& ;
– They later receive a public signal ! + (, where ( is noise;
Private signals can be interpreted as results of their analysis
of market conditions, security prices etc.
Public signal is some public information such as earnings
announcements etc.

FM 300 Lecture Note 6 7

DHS 1998
Investors have overconfidence and self-attribution biases.
Overconfidence: they exaggerate the precision of their signal.
– They believe that their noise # has variance $)%& , which is
lower than the true variance $%& ;
Self-attribution bias: if the public signal has the same sign as
their private signal, they see it as evidence of their skill.
– This reinforces their belief that private signal is correct.
– They become more overconfident and believe that the
variance of # is $)%& −+. That is, they think that the precision
of their signal improves.
– However, they underweight public signal when the public
signal contradicts their private signal.
FM 300 Lecture Note 6 8
 … DHS 1998
Consistent with psychological evidence, they interpret good
news as further evidence of their skills and attribute bad news
to external errors or sabotage.
– When investors receive confirming news, their confidence in
the private signal rises too much;
– When they receive disconfirming news, confidence declines
too little.
Suppose, they get good private signal and buy stocks at date 1.
– If later they receive good news they buy even more;
– If they receive bad news, they do not sell much;
– Hence, on average, the price increases.

FM 300 Lecture Note 6 9

… DHS 1998

The reaction to a favorable initial shock (the private information

signal) is hump-shaped.
– Price on average rises further as public information arrives,
because confidence about the private signal on average
grows => short-lag positive autocorrelation;
– More accumulated evidence forces investors back to a
more reasonable expectation => reversal at long horizons.
Similarly for an unfavorable shock.

FM 300 Lecture Note 6 10

 … DHS 1998

FM 300 Lecture Note 6 11

BSV 1998
Actual earnings D for a risky asset follow a random walk:
70 = 7012 + #0
Earnings follow a random sequence with no predictability, but
investors see patterns (consistent with representativeness
bias). Investors mistakenly believe that the earnings process
randomly fluctuates between two regimes (or models):
– M1 is a mean-reversion in earnings: positive shock to
earnings is likely to be followed by negative shock, and
vice versa, that is, cov #012 , #0 < 0;
– M2 is a regime with expected trend: cov #012 , #0 > 0;
– Investors do not know whether they are in regime 1 or 2;
– They observe D and update probability of being in 1 or 2.
FM 300 Lecture Note 6 12
 … BSV 1998

Assume investors put more probability on being in regime 1.

If recent earnings changes #0 reverse, investors erroneously
believe the firm is in a mean-reverting state, and underreact to
recent news, consistent with conservatism.
If investors see a sequence of growing earnings, they tend to
conclude (wrongly) that the firm is in a growth regime, and
over-extrapolate trends, consistent with representativeness.

FM 300 Lecture Note 6 13

… BSV 1998

After observing a long sequence of positive shocks investors

overreact and mistakenly conclude they are in regime 2.
– Stock prices increase;
– If next shock is positive, stock return is positive but small,
because this positive shock was expected by investors,
and hence, incorporated in prices in the previous period;
– If shock is negative, the investors are surprised, and
hence, there is a large negative return;
Thus, there is a reversal of long-term returns and delayed
correction of previous biases.

FM 300 Lecture Note 6 14

 … BSV 1998

Suppose, investors believe they are in regime 1 and #0 > 0.

– Investors expect that #092 < 0 is more likely whereas in
reality negative or positive shocks are equally likely;
– If #092 < 0, as expected => realized return is not large;
– If #092 > 0, this is a surprise => return is large and positive;
Hence, good news at t predicts large return at t+1 on average,
consistent with underreaction stories and momentum.
Price underreacts to information about #0 because investors
think that positive shock will be followed by a negative shock.

FM 300 Lecture Note 6 15

HS 2000

There are two types of traders in the model: newswatchers

and momentum traders. Each group of traders is risk averse,
and is able to process only a subset of available information.
A firm pays dividend D at time T, and then closes down.
Information about the dividend slowly rotates among different
groups of newswatchers.
– Information is gradually incorporated in prices (in contrast
to the efficient markets hypothesis);
– Underreaction arises because investors learn the
information gradually over time.
– Hence, there might be trends and predictability in prices.

FM 300 Lecture Note 6 16

 … HS 2000

Newswatchers condition on their own private signals (news)

but ignore information in prices, causing underreaction.
Asset prices, in general, contain useful information:
– For example, if prices are going up it might be because
many newswatchers got positive news;
– This is an important source of information which is ignored
by newswatchers;
– They only learn the information from each other and ignore
the information in prices.

FM 300 Lecture Note 6 17

… HS 2000

Momentum traders, in contrast, condition on the cumulative

price change over the last k periods: :0 − :01; .
Momentum traders exploit the underreaction of newswatchers
by buying in response to price increases.
This accelerates the reaction to news, but also causes
overshooting: prices increase even more, making momentum
stronger.
Eventually, the overshooting is corrected when newswatchers
learn all the information and act accordingly.
As a result, we get momentum with eventual reversal.

FM 300 Lecture Note 6 18

 Empirical evidence: JT 2001
Jegadeesh and Titman (JT, 2001) confirm their earlier
evidence on momentum and study what happens with the
winners and losers in the post-holding periods (13-60 months).
Looking at the returns in the post-holding periods allows them
to test whether momentum is due to one of the following:
1. Differences in expected returns;
2. Underreaction;
3. Overreaction;

FM 300 Lecture Note 6 19

… Empirical evidence: JT 2001

1. Differences in expected returns:
Momentum profits are the result of differences in unconditional
expected returns across stocks (e.g., winners are stocks with
higher returns at all times). Hence, momentum profits should be
the same in any post-ranking period.
2. Underreaction:
If investors underreact to news, prices adjust slowly to
information. Once information has been fully incorporated in
prices, there is no further predictability in stock returns. The post-
holding period momentum profits should be zero.
3. Overreaction:
If investors overreact to news, prices are pushed beyond the
levels warranted by fundamentals. Eventually, there will be a
correction. Momentum profits should eventually reverse as
prices revert to fundamental values.
FM 300 Lecture Note 6 20
 … Empirical evidence: JT 2001

Predictions of various theories:

Cumulative Returns
differences in returns

underreaction

overreaction

t
holding period post-holding period

FM 300 Lecture Note 6 21

… Empirical evidence: JT 2001

At the end of each month rank stocks based on previous 6
month returns (months -5 to 0);
Group stocks into 10 portfolios based on these rankings and
hold them for 6 months or 12 months;
– They confirm momentum in this holding period;
Then, look at portfolio returns in the post-holding period
(month 13 to month 60);
Formation Period Holding Period Post-Holding Period
(month -5 to month 0) (month 1 to month 6 or 12) (month 13 to month 60)

They find that the return on momentum portfolio is negative

in the post-holding period, consistent with overreaction story.

FM 300 Lecture Note 6 22

 …Empirical evidence: JT 2001

FM 300 Lecture Note 6 23

Limits to arbitrage
Arbitrageurs are sophisticated investors who make profits
by exploiting arbitrage opportunities. In so doing, they
eliminate mispricing, and ensure that stock prices are
aligned with fundamental values.
In reality, arbitrageurs face several constraints and risks:
– Implementation costs;
– Noise trader risk;
An alternative view: arbitrageurs may exacerbate/cause
mispricing! If mispricing is predictable, they may find it
optimal to ride a sentiment wave rather than correcting it.

FM 300 Lecture Note 6 24

 Implementation costs: short sales
We will focus on the short-sell constraint.
Arbitrage trades usually require taking short positions.
– Recall the definition of arbitrage: buy the cheaper asset
and sell short the more expensive asset, with the same
future payoffs.
– Costly short selling can deter arbitrage.
- If gains from arbitrage < costs of shorting/borrowing,
arbitrageurs do not have the incentives to trade.
– Sometimes arbitrageurs are unable to find shares to
borrow (i.e., infinite shorting costs).
This may explain the violation of the law of one price during
3Com spinoff discussed in Lecture Note 5;

FM 300 Lecture Note 6 25

… Implementation costs: short sales

How short selling works:
– A borrows a share from B (usually her broker) and sells the
share in the market (the term of a stock loan is usually one
day, but is typically renewed many times)
– A must post collaterals with B, with the collateral value >
stock value (the difference is the “margin requirement”)
– A pays B the dividends of the stock; B pays A an interest on
the collateral.
– The interest B pays A (called the “rebate rate”) is determined
by the supply and demand for shares to short.
– In most cases, rebate rate < risk free rate (the difference is
called the “haircut”).

FM 300 Lecture Note 6 26

 …Implementation costs: short sales
The cost of short selling
– The foregone investment return on the collateral
– The haircut
There is another risk involved
– the lender may recall the stock loan at any time, called
“buy-in risk”;
– e.g., if the lender wants to sell shares after a price
increase, the short-seller has to close position at a loss if
he can’t find other shares to borrow
Overall, only 10% of the shares outstanding are borrowed and
sold short.

FM 300 Lecture Note 6 27

Noise trader risk

Another important limit to arbitrage is the so-called noise

trader risk
– Noise traders buy and sell stocks without accounting for
fundamental information:
- people may sell their stocks to finance their
expenses (e.g., buy a house);
- inexperienced investors make investment mistakes.
– Consequently, stock prices reflect not only information
about fundamental values, but also unpredictable
demand/supply shocks caused by noise traders.

FM 300 Lecture Note 6 28

 Twin Stocks Puzzle
Picture below shows ln(PRoyal Dutch /(1.5 ´ PShell )).

Source: Froot and Dabora (1999)

FM 300 Lecture Note 6 29

… Noise trader risk

Noise traders can introduce risk to the market. Let’s
consider the Royal Dutch/Shell example
– Suppose, we think the price of Royal Dutch is too low
relative to Shell in Jan 1980, given that
A
PRoyal Dutch
» 0.85
1.5 ´ P A
Shell

Shell is overpriced, RD is underpriced:

– Sell short 1.5 units of Shell and buy 1 unit of RD;
– When the prices converge, close our positions, by selling
a unit of Dutch and buying back 1.5 units of Shell.

FM 300 Lecture Note 6 30

 … Noise trader risk

Arbitrageurs’ hope is that prices will converge and the parity

will be restored soon.
– Indeed, if everyone starts taking long position on RD and
short positions on Shell, then PRoyal Dutch ↑ and PShell ↓.
– Market forces should work to restore the parity.
But this is NOT always the case
– The mispricing can get even worse before being corrected.
– e.g., at the end of 1980.
J.M. Keynes: ”Markets can remain irrational longer than you
can remain solvent.”

FM 300 Lecture Note 6 31

… Noise trader risk

Most arbitrageurs worry about mispricing getting worse;
They may need to close their positions prematurely:
– Their investors may withdraw capital;
– They may face margin calls from their brokers.
The implication is that:
– Arbitrageurs may have to close their arbitrage positions
precisely when the mispricing is the most severe;
– In addition, when arbitrageurs close their arbitrage
positions, they are effectively magnifying the mispricing.
Hence, noise-trader risk deters arbitrage activities.

FM 300 Lecture Note 6 32

 Performance based Arbitrage

Shelifer and Vishny (SV 1997) show how noise trader risk
creates limits to arbitrage for active portfolio managers.
Arbitrageurs (e.g., hedge funds) manage capital provided by
outside investors.
– Their arbitrage ability is limited if capital is limited.
Outside investors do not have full understanding of the
manager’s strategies. Hence, it may be rational for investors
to allocate funds based on past performance of the manager
and to withdraw some capital after poor performance
(“performance based arbitrage”).
Responsiveness of funds under management to past
performance is called performance based arbitrage.
FM 300 Lecture Note 6 33

… SV 1997

This means that the arbitrageurs may face capital constraints

precisely when their investments have the best prospects, i.e.
when the mispricing they bet against initially deepens.
Anticipating these potential capital constraints, rational
arbitrageurs will be less aggressive in their initial bets against
mispricing.
As a result, arbitrage may not fully enforce market efficiency.

FM 300 Lecture Note 6 34

 … SV 1997
Suppose, the arbitrageur knows that:
– the value of the asset at date T is equal to V;
– the asset is underpriced, and current price P0 is too low.
Arbitrage: buy at price P0 and sell at price V at date T.
Noise traders generate random shocks to prices.
Signal extraction problem for the investor is difficult, and poor
performance in the intermediate period may be due to:
– Random shock from noise traders;
– Deepening of the mispricing;
– Lack of skill.

FM 300 Lecture Note 6 35

… SV 1997
Consider two potential paths of prices between 0 and T:

V
1

P0 2

0 T

Price tends to increase along the first path (solid line).

Price significantly decreases along the second path (dashed
line) before going up again.

FM 300 Lecture Note 6 36

 … SV 1997

Investors may start withdrawing money when they see prices

going down along path 2 on the picture because they worry
that the manager lacks skills.
Note, that withdrawals happen when the price is even lower
than P0, and hence, arbitrage is more profitable than before.
Funds liquidate assets at price lower than P0 => lose money.
The managers worry about prices going down before going up.
– They reduce their positions as the price moves further away
from the fundamental value V.

FM 300 Lecture Note 6 37

Bubbles and Overconfidence

The presence of irrational investors can help explain certain
puzzles like Dot Com bubble.
– It seems that NASDAQ in late 90s was a bubble;
– Alan Greenspan called it “irrational exuberance”;
– One explanation can be that investors were overly optimistic
about the prospect of internet companies.
– Seeing increasing stock prices investors irrationally
extrapolated trends into future.
Evidence in favour of behavioural story: firms that simply added
“.com” to their names managed to increase their stock prices.

FM 300 Lecture Note 6 38

 Riding the sentiment wave
Conventional view: rational speculators would find it
optimal to attack price bubbles, thus exerting a correcting
force on prices.
Hedge funds were riding the technology bubble rather
than fighting it because:
– Other investors were overconfident, and the trend was
expected to continue for some time;
– Attacking the bubble too early would have made them
lose potential profit and clients, as in SV 1997 model;
Riding the bubble makes bubble worse in the short-run:
– Speculators push prices in the direction of bubble’s
growth, which inflates it further.
FM 300 Lecture Note 6 39

Riding the sentiment wave

Brunnermeier and Nagel (2004) use Price to Sales ratio

(P/S) to identify candidates for mispricing.
– High P/S means asset is overpriced.
Hedge funds were riding the tech bubble, but understood
that prices would eventually deflate:
– Hedge fund portfolios were heavily tilted towards tech
stocks during 1998-2000;
– Looking at individual stocks, hedge funds reduced their
holdings before stock prices collapsed.

FM 300 Lecture Note 6 40

Hedge funds and the tech bubble

FM 300 Lecture Note 6 41

Hedge funds and the tech bubble

FM 300 Lecture Note 6 42

 Hedge funds and the tech bubble

Many funds (e.g., Soros) rode the bubble for some time.

FM 300 Lecture Note 6 43

 FM 300 Section A

Fixed-income portfolio management

Lecture Note 7
Dr. Georgy Chabakauri
London School of Economics

FM 300 Lecture Note 7 1

Outline

Effects of interest rate changes on bond prices

Duration and Convexity
Immunization
Bond portfolio management
– Passive strategies: Immunization
– Active strategies: Bond swaps

FM 300 Lecture Note 7 2

 Interest rate risk

Assume that you are a bond portfolio manager: how can you
attempt to shield the value of your portfolio from risk?
What risk does a bond portfolio manager face?
– The key source of risk to the entire portfolio is interest rate
risk: the effect that interest rate movements can have on
the prices of bonds and bond portfolios.

FM 300 Lecture Note 7 3

Bond price and yield

Recall that the price of a bond with T years to maturity, face

value F and coupon rate c is:

cF cF cF (c + 1)F
P= + + + ... +
1 + r1 (1 + r2 ) 2 (1 + r3 )3 (1 + rT ) T

The yield to maturity is the “hypothetical constant” discount

rate that makes the PV of the bond’s cash flows equal to the
bond’s price. It is the bond’s IRR:
T -1
cF (1 + c)F
P=å +
i =1 (1 + y) (1 + y)T
i

FM 300 Lecture Note 7 4

 Relationship between price and yield

Price

Yield

FM 300 Lecture Note 7 5

Inference from the term structure

Yields of zero coupon bonds are equal to !" .

A plot of zero-coupon yields !" against time to maturity t is
called a yield curve (or term structure).
Yield curve varies over time.
If we look at a plot of the yield curve and notice that it is
upward/downward sloping, what can we conclude?
We need some assumptions on investor behavior

FM 300 Lecture Note 7 6

 Inference from the term structure

Various shapes of term structure:

FM 300 Lecture Note 7 7

Explaining Term Structure: Forward

Rates
Forward rates can be found from equation:

(1+ r s ) s (1+ fs,t ) t−s = (1+ r t ) t

Otherwise, arbitrage is possible:
– E.g., suppose (1+ r s ) s (1+ fs,t ) t−s > (1+ r t ) t
– To make arbitrage borrow £1 at rate rt for t years, lend £1 at
rate rs for s years, and then £(1+rs)s at rate fs,t for t − s years.

FM 300 Lecture Note 7 8

 Explaining Term Structure

Expectations Hypothesis: Forward rates are unbiased

predictors of future spot rates: fs,t=E[rs,t].
This hypothesis helps explain various shapes of term structure.
Term structure reflects expectations about future interest rates:
(1+r2)2=(1+r1)(1+f1,2), f1,2=E[r1,2].

E.g., upward sloping term structure means that r2>r1. Therefore,

f1,2 > r1 => investors expect future interest rates to increase.

FM 300 Lecture Note 7 9

Explaining Term Structure

Liquidity Premium Hypothesis:

– Lenders prefer liquidity (short-term), borrowers prefer long-
term => market provides a “term premium”: long-term
bonds offer risk premium over short-term bonds => higher
long-term rates.
Segmentation Hypothesis
Every maturity has its own clientele (lender & borrower), and
there is no explicit relation between short-term and long-term
rates.

FM 300 Lecture Note 7 10

 Inference from the term structure

If the yield curve is downward sloping, we could infer:

– The forward rate for the coming period is smaller than the
yield at the same maturity
– Investors anticipate short rates to fall in the future.
– A downward sloping yield curve is not likely to be due to
term premia because they are positive under the liquidity
preference story.
Conclusion: the theories of term structure determination
provide a framework within which we can use the information
contained in yield curves.

FM 300 Lecture Note 7 11

Inference from the term structure

FM 300 Lecture Note 7 12

 Interest rate changes and bond prices

To see the influence of interest rate changes on bond prices,

assume that term structure is flat, that is, !" =r.
Assume a menu of £1 zeros with different periods to maturity.
The price of a T period zero is Pk = 1/(1+r)T.
Similar for coupon bearing bonds
Notice that r = y in this case
Price Pk more sensitive to r when T is large

FM 300 Lecture Note 7 13

Duration

Definition: The duration of a bond is the elasticity of its price

with respect to changes in yields.
Duration is the % change in price associated with a 1%
change in yield.

FM 300 Lecture Note 7 14

 Duration
Bond
Value

y-Δy y y+Δy
Interest rate (%)

The sensitivity to interest rate changes is determined by the

steepness (slope) of the bond price curve.

FM 300 Lecture Note 7 15

Calculating durations

From the definition, we can write the duration of a bond with

price P and yield y as
dP
P 1 + y dP
D=− =−
d(1 + y) P dy
1+y

Change in the price of a bond, given a change in yield:

D
dP = − ×P×dy,
1+y
D
ΔP ≈ − ×P×Δy,
1+y

FM 300 Lecture Note 7 16

 Calculating Duration

Modified duration: D∗ = D/(1 + y)

Calculate duration of a zero-coupon bond with price P = (1+y)-T:

Duration equals maturity T:

1+y T
D=T=
P (1 + y)456

FM 300 Lecture Note 7 17

…Calculating durations
Consider an arbitrary portfolio of coupon and zero-coupon
bonds which has cash flows Cj for j=1,…,N. Given a yield of y,
this implies a portfolio value of
N
Cj
P=å
j=1 (1 + y) j

Using the definition of duration we obtain

(1 + y)dP (1 + y) N Cj N C j /(1 + y) j
D=-
Pd(1 + y)
= åj =åj
P j=1 (1 + y) j+1 j=1 P

N C j /(1 + y) j
D=åj
å
N
j=1
j=1
C j /(1 + y) j

FM 300 Lecture Note 7 18

 …Calculating durations

The duration (called Macaulay duration) for bonds when term

structure is flat is given by:

PV(C6 ) PV(C: ) PV(C4 )

D = 1× + 2× + ⋯ + T×
P P P

Duration is a weighted average of payment dates.

– Weights PV(Ct)/P sum up to 1.
– Cash flows with larger present values are given larger
weights PV(Ct)/P in this average.

FM 300 Lecture Note 7 19

Facts about duration

The duration of a coupon bond is lower than maturity because

a lot of the cash flows accrue before maturity.
Duration generally increases with maturity
Duration falls as coupon rate rises: the weight on the early
payments is higher.
Duration falls as yield rises: a higher yield reduces the value
of later cash flows by a greater proportional amount.
– Weights of earlier cash flows in the total become greater.

FM 300 Lecture Note 7 20

 Immunization
Immunization techniques are strategies used by financial
institutions to shield their financial status from exposure to
interest rate fluctuations.
Banks and other financial institutions have a natural mismatch
between asset and liability maturity structures.
By carefully choosing the structure of assets it is possible to
reduce the interest rate sensitivity of the net worth (= asset
value – liability value).
This is achieved by matching the durations of assets and
liabilities. Hence, net worth is not sensitive to interest rate
fluctuations and firms do not lose money because of interest
rate fluctuations.
FM 300 Lecture Note 7 21

… Immunization

Suppose, you are a manager of an insurance company. Your

assets consist of £30mln in cash and the liabilities consist of
four £10mln payments in years 5, 6, 7, and 8 from now.
Current interest rate (which is also your discount rate) is r =
5%. However, you are concerned that the interest rates may
go down. What is the immunization strategy in year t = 0?

Balance Sheet
Assets Liabilities
£30 mln year: 5 6 7 8

£10 £10 £10 £10

PV=L=29.173 mln

FM 300 Lecture Note 7 22

 … Immunization

Net worth W is defined as the difference between the values

of assets and liabilities:
10 10 10 10
W = 30 - - - -
(1+ 0.05) (1+ 0.05) (1+ 0.05) (1+ 0.05)8
5 6 7

= £0.82 mln

If r falls to 4% W becomes:
10 10 10 10
W = 30 - - - -
(1+ 0.04)5 (1+ 0.04)6 (1+ 0.04)7 (1+ 0.04)8
= -£1.02 mln

The company loses money if the interest rate drops!

FM 300 Lecture Note 7 23

… Immunization

Suppose, to immunize the net worth the manager invests £30

mln and buys x3 units of 3-year zeros at price P3 and x10 units of
10-year zeros at price P10.
3-year and 10-year bonds with face value £100. Hence, their
prices are:
– P3 =100/(1+0.05)3=86.384; P10 =100/(1+0.05)10=61.391

Need to find x3 and x10 such that the net worth is not sensitive
to interest rates.

FM 300 Lecture Note 7 24

 … Immunization

Modified duration of the stream of liabilities is: DL* = 6.13.

Hence, changes in liabilities are given by:
DL » -DL*LDr
The value of assets is given by:
< = => ?> + =6@ ?6@
Approximate change in asset value is given by:
∆< ≈ −B>∗ => ?> ∆! − B6@
∗
=6@ ?6@ ∆!
change in value change in value
of 3-year zeros of 10-year zeros

FM 300 Lecture Note 7 25

… Immunization

Objective: choose x3 and x10 such that:

DW = DA - DL » 0
Hence:
B>∗ => ?> + B6@
∗
=6@ ?6@ = BI∗ J

Moreover, we should have enough money to buy zeros:

=> ?> + =6@ ?6@ = £30 mln

total money
investment available

FM 300 Lecture Note 7 26

 … Immunization
Putting numbers to these equations we get:
246.811x3 + 584.676 x10 = 178.899mln,
86.384 x3 + 61.391x10 = 30mln.
Hence, x3=185476 , x10=227684.
Net worth (=assets – liabilities) the same as before: W =
£0.82mln. How does it change when r = 4%?

18.547 22.768
W= +
( 1+ 0.04 ) ( 1+ 0.04 )10
3

10 10 10 10
- - - -
( 1+ 0.04 )5 ( 1+ 0.04 )6 ( 1+ 0.04 )7 ( 1+ 0.04 )8

= £0.842 mln we no longer lose money when r ↓!

FM 300 Lecture Note 7 27

… Immunization

Note that immunization is a dynamic strategy and has to be

revised at future dates.
This is because as time passes 3-year and 10-year bonds
become 2-year and 9-year bonds and hence their durations
change.
Duration of the liabilities also changes when 1 year passes. In
particular, the stream of liabilities becomes:

year: 4 5 6 7

£10 £10 £10 £10

PV=L=32.269

FM 300 Lecture Note 7 28

 Duration and convexity

How does immunization perform for large yield changes?

Now approximate the change in price of a bond associated
with a given change in yields Δy using a second order Taylor
approximation:
∆P 1 dP 1 d: P
≈ ∆y + 0.5 :
(∆y):
P P dy P dy

Define convexity as:

1 d: P
C= × :
P dy

FM 300 Lecture Note 7 29

…Duration and convexity

Using the formula for modified duration and convexity, we have

∆P
≈ −D∗ ∆y + 0.5C(∆y):
P

Using durations only, we get a linear approximation to bond

price changes as yields change. This is why duration-based
immunization strategies only work for small changes in yield.
If we introduce convexity, we can approximate bond (portfolio)
price changes far more accurately.
We can construct immunization strategy with respect to both
duration and convexity.

FM 300 Lecture Note 7 30

 …Duration and convexity

The expression for the convexity is as follows:

4
1 t(t + 1)CN
C= M
P (1 + y)N5:
NO6

FM 300 Lecture Note 7 31

…Duration and convexity

Price

convexity

Duration

Yield

FM 300 Lecture Note 7 32

 Swaps
Swap is an important tool for managing interest rate risks.
Swap is a contract that allows two parties to exchange their
cash flows in the future.
– A company that pays fixed interest rate on its debt may
prefer to pay a floating rate;
– A company that receives cash flows in pounds may prefer
to receive cash flows in dollars;
– The market for swaps is large. In 2003 the notional amount
of swaps outstanding was $120 trillion.

FM 300 Lecture Note 7 33

Interest Rate Swaps

In interest rate swaps two parties agree to exchange a stream of
interest payments of one type (fixed or floating) for a stream of
payments of the other type.
Consider the following example.
– Firm A agrees to pay firm B a floating rate note (FRN):
t=0 t=1 t=2 t=3 t=4 t=T-1 t=T

£N r0,1 £N r1,2 £N r2,3 £N r3,4 .... £N rT-2,T-1 £N rT-1,T+£N

– In exchange, firm B pays firm A a constant coupon bond:

t=0 t=1 t=2 t=3 t=4 t=T-1 t=T

£N k £N k £N k £N k .... £N k £N k+£N

FM 300 Lecture Note 7 34

 … Interest Rate Swaps
Swaps have the following characteristics:
– Payments are based on a notional amount £N;
– On payment dates net difference (rt -1,t - k ) ´ N is
exchanged;
– At the inception, fixed interest rate k is set such that the
swap has zero value;
– There is a secondary market for swaps;

FM 300 Lecture Note 7 35

… Interest Rate Swaps

The floating rate in many swap contracts is based on London

Interbank Offer Rate (LIBOR) plus/minus several basis points.
LIBOR is the rate at which banks are willing to lend to each
other.
In practice, swaps are arranged via financial intermediaries
such as banks, that charge certain commission.

LIBOR LIBOR
A Bank B
k k

FM 300 Lecture Note 7 36

 … Interest Rate Swaps
Fixed interest rate k is determined from the condition that the
value of floating rate note equals the value of coupon bond
with coupon yield k:
VFRN =Vbond(k).
Fixed rate k is similar to forward interest rate because it is
determined at t = 0.
The value of fixed coupon bond can be found given the term
structure at time 0 (i.e. interest rate r0,t between date 0 and
date t):

£N ´ k £N ´ k £N ´ k £N ´ k + £N
Vbond = + + + ... + .
1 + r0 ,1 (1 + r0 ,2 ) (1 + r0 ,3 )
2 3
(1 + r0,T )T

FM 300 Lecture Note 7 37

… Interest Rate Swaps

The price of FRN is £N because £N is enough to replicate the
stream of cash flows:
– Invest £N at t = 0, get £N+£N r0,1 at t =1, pay coupon
£Nr0,1;
– Invest £N at t =1, get £N+£N r1,2 at t =2, pay coupon £Nr1,2
etc.

t=0 t=1 t=2 t=3 t=4 t=T

£N r0,1 £N r1,2 £N r2,3 £N r3,4 .... £N rT-1,T+ £N

£N £N r0,1+ £N £N r1,2+ £N £N r2,3+ £N £N r3,4+ £N £N rT-1,T+ £N

....
invest £N invest £N invest £N invest £N
at r0,1 at r1,2 at r2,3 at r3,4

FM 300 Lecture Note 7 38

 … Interest Rate Swaps
Fixed interest rate k is then determined form the condition that
the value of floating rate note equals the value of coupon
bond with coupon yield k:
£N ´ k £N ´ k £N ´ k £N ´ k + £N
£N = + + + ... + .
1 + r0 ,1 (1 + r0 ,2 )2 (1 + r0 ,3 )3 (1 + r0,T )T
The equation for k can be rewritten as follows:
k k k k +1
1= + + + ... + .
1 + r0 ,1 (1 + r0 ,2 ) (1 + r0 ,3 )
2 3
(1 + r0,T )T

FM 300 Lecture Note 7 39

Quick Question
Consider a 2-year swap with annual payments, notional
amount $1. Floating coupon rate is the 1-year rate
observed one year earlier. Current data: r0 ,1 = 7%, r0 ,2 = 10%.
– What is the appropriate fixed rate k to set on this
swap agreement?
t =0 t =1 t =2

r0 ,1 = 7%

r0 ,2 = 10%

Cash Flows:
receive fixed k k
pay floating - 0.07 - r1,2

FM473 Finance I 40
Forwards, Futures and Swaps
 Answer
Fixed rate of return solves the following equation:
k k +1
+ = 1.
1 + r0 ,1 (1 + r0 ,2 )2

– Substituting the interest rates we obtain:

k k +1
+ = 1 Þ k = 9.855%.
1 + 0.07 (1 + 0.1)2

FM473 Finance I 41
Forwards, Futures and Swaps
 FM 300 Section A

Active portfolio management and

performance evaluation
Lecture Note 8
Dr. Georgy Chabakauri
London School of Economics

FM 300 Lecture Note 8 1

Outline

Active portfolio management

Measures of performance
Treynor/Black Model
Tests for stock selection/market timing
Style benchmarks (with multi-factors or styles)
Empirical evidence

FM 300 Lecture Note 8 2

 Why need asset management

Diversification with lower transactions costs;

Tailoring portfolio to meet specific objectives;
Selecting securities to achieve a superior risk/return
tradeoff;
…

FM 300 Lecture Note 8 3

Investment objectives
Mutual funds (passive or active) are required by law to state their
investment objectives. Some common ones:
– Growth funds: stock funds looking for high expected returns;
– Income funds: stock and/or bond funds looking for high
dividends and coupons with relatively low risk;
– Balanced funds: conservative stock and/or bond funds that
attempt to minimize losses;
– Money market funds: invest in short-term low-risk debt
instruments;
– Municipal bond funds: invest only in municipal bonds to obtain
entirely tax-free income;
– Corporate bond funds: goal is to outperform Treasuries while
minimizing bankruptcy risk.
FM 300 Lecture Note 8 4
 Active management
Index Management: Construct a portfolio P that has fewer and
more liquid assets and mimics as closely as possible the market
portfolio M. Need a passive portfolio strategy, but not the market
portfolio because of transaction costs.
Portfolio weights minimize tracking error variance: var(%& − %( )
Optimal portfolio weights solve the following problem:
7

min var 3 84 %4 − %&

-. ,-0 ,…,-2
456

The solution to this problem is a set of portfolio weights, w1, . . . ,

wn which define the tracking portfolio.

FM 300 Lecture Note 8 5

Active management
Consider a portfolio P with returns given by:
%( = :( + <( %& − %> + ?(

Then, the tracking error is: %( − %& = :( + (<( −1)%& − <( %> + ?(

The standard deviation of the tracking error is then given by:

B
@A = <( − 1 B C& + CDB

This expression gives some intuition for the tracking error. If

portfolio P has beta equal to 1, then the tracking error is equal to
the residual risk.

FM 300 Lecture Note 8 6

 Active management
Performance Enhancement:
– There is evidence that returns are affected by factors
such as size, book-to-market, and momentum.
– Investors can try to use this fact to beat the market.
To maximize the probability of beating an index, managers
maximize portfolio return relative to the benchmark:
7

max E 3 84 %4 − %&
-. ,-0 ,…,-2
456
7

s. t. var 3 84 %4 − %& ≤ CK B
456

FM 300 Lecture Note 8 7

Active management

Note that this problem is the same as minimizing the variance

of (rP−rM), subject to achieving a given expected return.
The solution to this problem is again a set of portfolio
weights, w1, …, wn.
The only difference from standard mean-variance
optimization is that portfolio return is replaced with (rP − rM).

FM 300 Lecture Note 8 8

 Active management

Market Timing. Market timing is based on the premise that

investors can forecast the market, using variables such as
dividend yield, macroeconomic indicators etc.
You shift funds between a market index portfolio and a safe
asset based on your forecasts:
rP = wt · rM + (1 − wt) · rf = rf + wt (rM - rf)
=β

In good times, wt increases => beta increases.

Market Timing ability is to take high beta positions before the
market goes up and low beta positions before it goes down.

FM 300 Lecture Note 8 9

Active management

Factor Tilting is more general than market timing because it

presumes a multi-factor structure.
Tilting is forming a portfolio to take advantage of a forecast of
a factor. Assuming we have this ability, we can earn superior
profits by varying the factor betas/loadings of our portfolio.
We want to increase the loading when we think that the factor
is likely to have a positive realization. We want to decrease
the loading when we think that the factor is likely to have
negative returns.
rP,t = :( + bP,1,t f1,t + · · · + bP,n,t fn,t + ep,t

FM 300 Lecture Note 8 10

 Active management
Stock selection is a portfolio management technique that
presumes superior knowledge about expected stock returns.
Portfolio managers spend a lot of resources on information,
and buy stocks they think are underpriced, i.e. stocks with
positive alphas.
They attempt to construct a well diversified portfolio of
positive alpha stocks. (Similarly for overpriced stocks.)
Individual stocks, even if mispriced, still have firm specific
risk. Therefore, a risk-averse portfolio manager will want to
hold a portfolio of well-diversified mispriced securities.

FM 300 Lecture Note 8 11

The Treynor-Black model

The intuition and logic of the Treynor-Black model:

1. Select a small set of securities you think are mispriced (i.e.,
have positive or negative alphas). This is active portfolio A.
2. Active portfolio is likely to be under-diversified.
3. Combine this active portfolio with the passive benchmark
portfolio M (e.g., the market) to diversify.
4. Calculate a new capital allocation line (CAL).
5. Use your utility function to determine you optimal portfolio.

FM 300 Lecture Note 8 12

 The Treynor-Black model
Consider a portfolio: rp = w · rA + (1 − w) · rM
Find optimal weight that maximizes the Sharpe ratio:

A %N − %> 8A %L + 1 − 8 A %& − %>

max =
- CN 8 B CLB + (1 − 8)B C&
B
+ 28(1 − 8)CL&

Suppose, active portfolio has positive alpha and return given

by %LM − %> = :L + <L %&M − %> + ?M .

Positive α means that the security would plot above the CML.
An α either > 0 or < 0 means that a combination of the
benchmark and portfolio A has a higher Sharpe Ratio than the
benchmark.

FM 300 Lecture Note 8 13

The Treynor-Black model

The optimal portfolio weight on portfolio A is given by
8Q :L /CDB
8∗ = , where 8Q = B
1 + (1 − <L )8Q A %& − %> /C&

weight of (1 − w*) is invested in benchmark portfolio M,

and w* in active portfolio A.
The Sharpe Ratio of the resulting portfolio:
VW(B = VW&
B
+ XWLB
:L
XWL =
CD
ARA is the Appraisal Ratio of portfolio A.

FM 300 Lecture Note 8 14

 Frontier for the Treynor-Black

A
T

E[rp] CAL

rf

FM 300 Lecture Note 8 15

Measuring performance
Though more and more money is in passive funds, there is
still over $1 trillion in active funds.
Active funds have higher fees than passive funds.
For example, the Vanguard S&P 500 Index fund has
expenses of 0.20% per year, while the Fidelity Magellan Fund
has an initial load of 3%, expenses of 0.95% per year.
The average expense ratio of active funds was 130 bp
(Carhart 1997); this suggests U.S. investors spend over $10
billion/year on active management.
In deciding where to invest money, and how much to pay the
fund, it is crucial to see how much value fund managers add.

FM 300 Lecture Note 8 16

 Measuring performance

FUND NAME Category YTD RET

• Franklin Biotechnology Discovery Health 62.24

• Lord Abbett Micro Cap Growth Small Growth 60.42
• Legg Mason Opportunity Trust Mid-Cap Value 51.16
• RidgeWorth Aggressive Growth Large Growth 50.57
• Rydex Series Trust Internet Technology 38.72
• Oppenheimer Global Opportunities World Stock 32.80
• S&P 500 Index 22.85

• YTD RET means year-to-date return.

Data through 18 Sep 2013

FM 300 Lecture Note 8 17

Measuring performance

If fund managers are generating high returns by taking on

more risk, they should not be compensated for it.
If we could use other well-known rules to get the same
returns, we should not pay the manager.
– Buying small stocks, high book-to-market stocks;
– Buying high momentum stocks.
We should only be willing to pay managers the extra returns
that they can generate in excess of what we could have
generated. The manager can do this in two ways:
– Market timing or factor tilting;
– Stock selectivity.

FM 300 Lecture Note 8 18

 Sharpe Ratio
Sharpe Ratio is the ratio of excess returns to volatility:

A[%( ] − %>
VW =
C(

Sharpe ratio captures trade-off between risk and return.

Idea: use the Capital Market Line. If markets are efficient
and CAPM holds, return/risk ratio should be the same for
all (diversified) portfolios.
Compare the Sharpe ratios of portfolio P and the market
portfolio M. If SRp > SRM then P has outperformed the
market; otherwise, P has underperformed the market.

FM 300 Lecture Note 8 19

M-squared measure
Sharpe ratio ranks the portfolios, but the numerical value is
difficult to interpret. E.g., is the difference of 0.04
economically big or small?
Mix the managed portfolio P with T-bills so that the
resulting portfolio P* has the same volatility as market M.
Let w be the weight of P. It can be found from the condition
8C( = C& (see Lecture 1). The expected return of P* is:
A %(∗ = 8A %( + (1 − 8)%>
Then, the M2 measure compares the returns:
[ B = A[%(∗ ] − A[%& ]
Such comparison is meaningful because P* and M have
the same standard deviation.
FM 300 Lecture Note 8 20
 Other Measures
The Sharpe measure is not appropriate for funds that are
part of a larger portfolio, or when deciding how much to
compensate managers.
Three alternative measures are useful in this case:
– Jensen Measure, Treynor Measure, Appraisal Ratio.
These measures are based on the SML (Security Market
Line), as opposed to the CML (Capital Market Line).
Each of these measures is asking the question of how well
the fund would have done relative to a portfolio of the
market and risk-free asset with the same systematic risk.

FM 300 Lecture Note 8 21

Jensen’s Alpha
From the CAPM:
a p = E[rpt - rft ] - b p E[rMt - rft ]
Alpha shows the attractiveness of a portfolio, and is the
intercept from the following regression:
rpt - rft = a p + b p (rMt - rft ) + e pt
Jensen’s α is also the maximum you should be willing to pay a
portfolio manager (remember that α is in units of return)
If, for example, a fund has a pre-expense α of 0.0015 when
calculated with a monthly regression, this means that we
should be willing to pay up to 0.15%/month (or approximately
1.8%/year) in expenses.

FM 300 Lecture Note 8 22

 Treynor’s Ratio

The Treynor’s ratio is used when the portfolio P you are

evaluating is one actively managed portfolio out of many in
your overall portfolio.
Suppose, the return of portfolio P is given by:
%(M = %> + :( + <( %&M − %> + ?M

The standard deviation of this portfolio is C( = <(B C&

B
+ CDB

However, if P is part of a bigger portfolio idiosyncratic risk is

diversified away, i.e. CDB is irrelevant.
We need to evaluate expected excess return only relative to
systematic risk measured by <( .
FM 300 Lecture Note 8 23

Treynor’s Ratio

This reasoning gives rise to the Treynor’s ratio:

A %( − %>
@( =
<N

Idea: use the security market line.

– If markets are efficient and CAPM holds, all assets should
lie on SML and @( = A[%&M ] − %> .

– Hence, if @( > A %&M − %> = @& portfolio is above the SML

and its Tretnor’s ratio exceeds that of the market portfolio.
It is the slope of the SML for the actively managed portfolio.

FM 300 Lecture Note 8 24

 Treynor Measure

E[rp]

AB
C M
XYZ
rf
Ex post Treynor index
lower than the market

bp

FM 300 Lecture Note 8 25

Treynor vs. Jensen

Treynor's measure can be viewed as evaluating performance

using the ratio of α to systematic risk:
A %( − %> :(
@( = = + (A %& − %> )
<N <(
= @&
Note that @& = A %& − %> , because <& = 1.
Hence, Treynor’s measure and alpha always agree in their
under/outperformance judgments relative to the market M.
However, they may not rank portfolios in the same fashion.
Ratio :/< is abnormal return per unit of systematic risk.

FM 300 Lecture Note 8 26

 Sharpe vs. Treynor

Both measure the risk vs. excess return tradeoff, but

definitions of risk are different
– Sharpe ratio uses total risk, and is appropriate for
characterizing the entire portfolio.
– Treynor measure uses beta, or covariance with the
market, and is useful for describing investments that may
form a small part of a portfolio.
Both are useful for ranking investment opportunities.

FM 300 Lecture Note 8 27

Appraisal ratio

The problem with the Jensen and Treynor measures is that

they do not adjust for the idiosyncratic risk in the portfolio.
The Treynor/Black model of active portfolio management
suggests that the following appraisal ratio is a guide to how
much of an asset/fund you want to add to your portfolio:

:L
XWL =
CD

FM 300 Lecture Note 8 28

 When is each measure appropriate?
Compensation:
Jensen’s α: is the maximum amount you should be willing to pay
a manager
If you are using past α, this assumes that the manager’s future
performance will be the same as their past performance!
Optimal Portfolio Choice:
Use the Sharpe Ratio when the portfolio represents the entire
investment fund
Use the Treynor Measure when the portfolio represents one of
many actively managed portfolios that you are adding to a
passive portfolio
Use the Appraisal Ratio when the portfolio represents a single
actively managed portfolio to be optimally mixed with the
passive portfolio
FM 300 Lecture Note 8 29

Stock picking ability

Specific Stock Selection Ability

Excess
Return
of Fund

}a
Excess Return
of Market

Stock-picking ability: Can the manager identify which stocks will

outperform the market or the industry?
FM 300 Lecture Note 8 30
 Market timing ability

Market Timing Ability

Excess
Return
of Fund

Linear
Regression

}a
Excess Return
of Market

Market-timing ability: Can the manager identify turning

points between bull and bear markets?
FM 300 Lecture Note 8 31

Market timing

Managers have market timing ability if they can forecast when

to be in the equity market and when to be out of the market
(hold cash or risk-free asset):
– Their portfolio has a high β when they expect the return on
the market to exceed the risk-free rate.
– Their portfolio has a low β when they expect the return on
the market to be dominated by the risk-free rate.
This implies that timing ability manifests itself as a non-linear
relationship between excess portfolio returns and excess
market returns: the excess portfolio return is more sensitive to
the market in good times.

FM 300 Lecture Note 8 32

 Market timing
Two popular methods for estimating market timing ability.
The first, developed by Henriksson and Merton, uses the
following regression:
rpt - rft = a p + b p (rMt - rft ) + g p I t (rMt - rft ) + e pt
where the dummy variable It=1 if rM,t > rf,t and It=0 otherwise.
If ]N > 0 and significant then there is evidence of timing ability.
Beta is time-varying and given by <N + ]N _M .
Henriksson finds that 62% of funds have negative timing ability.

FM 300 Lecture Note 8 33

Market timing
An alternative measure, developed by Treynor and Mazuy,
uses the following quadratic regression:

rpt - rft = a p + b p (rMt - rft ) + g p (rMt - rft ) + e pt

2

If the estimate of ]N is positive then there is timing ability

The value of stock selection is the α from this regression.
If the manager has timing ability, α will be biased unless the
squared market return is added to the regression.
Treynor and Mazuy find no evidence of market timing ability.
Factor timing can be tested similarly by including the square
of each factor-portfolios return as a right-hand side variable.

FM 300 Lecture Note 8 34

 Multifactor-based measures
All of the measures discussed so far are based on the
CAPM as a benchmark. For a multifactor model, however,
we can use the same basic measures.
Sharpe measure: since the definition of total risk has not
changed, the Sharpe measure is unchanged.
Jensen measure: same intuition as with the CAPM, but
alpha is calculated based on all the factors.
Appraisal ratio: use a multi-factor regression to get αp and
the standard deviation of the residual.
Treynor measure: with the APT we do not get one unique
measure because there are multiple sources of risk. We
get one Treynor measure for each factor or characteristic.
FM 300 Lecture Note 8 35

Style analysis

Loadings on a small number of factors (or characteristics)

can influence the return on a mutual fund.
Some of the fund return can be attributed to the fund’s
time-varying exposures to these factors.
The remaining portion of a fund’s return is attributable to
the security selection.

FM 300 Lecture Note 8 36

 Two approaches to analyzing style

1. Compare an actively managed fund with a passive

fund that holds similar types of stocks:
- Compare international stock fund with Morgan-
Stanley World Index;
- Compare a small-cap fund with the DFA small cap
index fund;
- Compare growth or value funds with the Vanguard
growth or value index funds; etc.

FM 300 Lecture Note 8 37

…Two approaches to analyzing style

2. Use regression analysis:

– Regress fund returns on broad indices of particular
styles of securities (e.g. use indices of small/large
caps, value/growth, momentum, bonds, international
stocks, etc.);
– Sharpe finds that 97.5% of returns can be attributed to
style, the remaining 2.5% to security selection.

FM 300 Lecture Note 8 38

 Regression analysis of style

Sharpe suggested the following regression:

%(M = : + <6 %6M + <B %BM + ⋯ + <7 %7M + ?M
where coefficients are estimated subject to constraints:
<6 + <B + ⋯ + <7 = 1,
<6 ≥ 0, <B ≥ 0, … , <7 ≥ 0.

In this regression, rP is the return on the portfolio, and r1, r2, …

are returns on various asset classes such as small cap, large
cap, low P/E stocks etc.
Betas can be interpreted as portfolio weights.

FM 300 Lecture Note 8 39

Empirical evidence
The average active manager adds very little/no value (Jensen
1968, Malkiel 1995).
Marcus (1990), in “The Magellan Fund and Market Efficiency,”
shows that Peter Lynch exhibited statistically significant
abnormal performance.
One way to identify good active managers is by past
performance. The evidence is mixed:
– Chevalier and Ellison (1999) find that there is manager
persistence, yet not fund persistence;
– Baks, Metrick, and Wachter (1999) find sufficient persistence
in active manager performance that it is worthwhile adding
some actively managed portfolios with strong past
performance and low costs and expenses to your portfolio.
FM 300 Lecture Note 8 40
 FM 300 Section A

Elements of international finance

Lecture Note 9
Dr. Georgy Chabakauri
London School of Economics

FM 300 Lecture Note 9 1

Outline
FX markets
Spot exchange rates
Forward exchange rates
Covered Interest Parity
Expectations theory of the exchange rate
Uncovered Interest Parity
Purchasing Power Parity
Real exchange rates
Real interest rates (The Fisher equation)

FM 300 Lecture Note 9 2

 Types of FX markets

Spot market
– delivery in one or two days
Forward market
– contract now to buy in future
– 1,2,3,6 and 12 months most common

FM 300 Lecture Note 9 3

FX turnover

Global foreign exchange market turnover was $4.0 trillion in

April 2010, reaching a 20% increase from April 2007.

– This increase is largely due to the increased trading

activity of financial institutions such as banks, hedge
funds, pension funds, mutual funds, insurance companies
and central banks, among others.

Foreign exchange spot turnover rose to $1.5 trillion in April

2010 from $1.0 trillion, an increase of 48% at current
exchange rates.

FM 300 Lecture Note 9 4

 Managing FX risk
Assets and liabilities exposed to FX rate movements will face
large risk over short to relatively long horizons.

The FX risk is extremely variable: managers should use a

variety of forecasts and tools to estimate future uncertainty.

Managers are exposed to FX risk that they cannot forecast.

How should they hedge?

Managers can translate future Foreign Currency cash flows

into future Domestic Currency cash flows using forward rates.

FM 300 Lecture Note 9 5

Forward Exchange Rate

Denote by S the spot exchange rate of home (UK) currency to
foreign currency (US), measured in units of home (UK)
currency. E.g., S=£0.8 for $1.
Denote by F the forward exchange rate of home (UK)
currency to foreign currency (US) in units of home currency.
The forward exchange rate is the rate fixed now for an
exchange of currencies at a specified future date.
Forward Ft,t+k: Enter into contract today (time t) to exchange
currencies k periods from now (time t+k) at this rate.
If Ft,t+k = £0.9, then after k periods you can exchange pounds
into dollars at this rate paying £0.9 for $1.

FM 300 Lecture Note 9 6

 Forward Premium/Discount
Forward Premium: the annualized percentage difference
between the forward rate and the spot rate

!#,% − (# 365
!"#,% =
(# ,−-

If FP > 0, then it takes more £ to buy 1$ in the future than now

=> £ is at a “discount” and $ is at a “premium”.
If FP < 0, then it takes fewer £ to buy 1$ in the future than now
=> £ is at a “premium” and $ is at a “discount”.

FM 300 Lecture Note 9 7

Pricing Forward Contracts

Forward rates are priced by “interest arbitrage”.

Suppose, an investor can borrow or invest (risk free) at home
or in the foreign country. If the investor eliminates exchange
risk by a forward contract, then there should be no difference
in the payoff of the two transactions.
Denote by: r the interest rate in the home country (UK)
and by: r* the interest rate in the foreign country (US).
If there is no arbitrage, borrowing in one country and lending
in the other country should not be profitable.

FM 300 Lecture Note 9 8

 Covered Interest Parity

Let F be one-year forward price, r, and r* one-year rates.

Then, no-arbitrage pricing implies that:

1+1
! = (×
1 + 1∗
Suppose, the equation is violated, and we have the following:
1+1
! < (×
1 + 1∗

Then, we can construct an arbitrage strategy.

Violation in the other direction is addressed similarly.

FM 300 Lecture Note 9 9

…Covered Interest Parity

Then, arbitrage can be constructed as follows:
1. At t = 0 borrow $1 at interest rate r*;
2. At t = 0 convert $1 into £S, lend £S in UK (or buy 1-year
zero with payoff in £) at interest rate r ;
3. At t = 0 take long position in currency forward (i.e. commit
to buy $ in one year at price £F);
4. At t =1 convert your profit in £ into $ at the futures price
and repay your loan.

FM 300 Lecture Note 9 10

 …Covered Interest Parity
In year 1 the investor will receive £S(1+ r).
– Then, investor will buy £S(1+ r)/F dollars;
– In year 1 need to repay $(1+r*);
– Net arbitrage profit: £S(1+ r)/F – (1+r*) > 0;

Consequently, in the absence of arbitrage the covered interest

parity should hold.

FM 300 Lecture Note 9 11

…Covered Interest Parity

More generally, to avoid arbitrage at maturity T, we have:

,−-
1 + 1#,%
!#,% = (# × 365
∗ , −-
1 + 1#,%
365

In this formula t and T are measured in days, and interest rates

are annualized.
Annualized forward premium is given by (for small rates):

!#,% − (# 365 ∗
≈ 1#,% − 1#,%
(# ,−-

FM 300 Lecture Note 9 12

 Does CIP hold?

CIP is a theory: what is required for it to work in reality?

– free movement of capital across borders.
Examples of capital controls:
– US, Germany before 1974
– Britain before 1978
– Japan before 1980s
– France after 1981

FM 300 Lecture Note 9 13

Unbiased Expectations Hypothesis

Forward rate is also potentially linked to unknown spot rate at

t+1 through investors’ expectations.

An individual investor (assuming risk neutrality) compares

Et [St+1] with Ft,t+1, and makes profit if Et [St+1]≠ Ft,t+1.
If Ft,t+1 > Et [St+1], then (e.g., Ft,t+1 = £0.9, Et [St+1]=£0.8 for $1):
– Buy pounds forward at time t (buy £0.9 forward for $1) ;
– Expect to sell at time t+1 (sell £0.9 for $1.125=0.9/0.8) .
If Ft,t+1 < Et [St+1], then:
– Sell forward at time t;
– Expect to buy at time t+1.
FM 300 Lecture Note 9 14
 Unbiased Expectations Hypothesis

If investors have identical expectations and are risk neutral,

that is, do not require risk premium, Ft,t+1 will rise or fall to
Et [St+1].
This leads to the Unbiased Expectations Hypothesis:
!#,% = 5# (%

FM 300 Lecture Note 9 15

Uncovered interest parity

From CIP and UEH we obtain the Uncovered Interest Parity:

,−-
1 + 1#,%
5# (% = (# × 365
∗ ,−-
1 + 1#,%
365
The expected (annualized) capital gain on foreign currency is
approximately equal to the difference in interest rates.
(% − (# 365 ∗
5# ≈ 1#,% − 1#,%
(# , − -

FM 300 Lecture Note 9 16

 Testing the Expectations Theory
This is theory. Does it hold in the data?
We need Et [St+1], but we can’t observe this variable. We can
observe the actual realized spot rate St+1. So we can test:

(#78 − (# !#,#78 − (#
= + 6#78
(# (#
where 5# 6#78 =0.
This equation is equivalent to the UEH !#,878 = 5# (#78 , which
can be verified by taking expectation in the equation above
and cancelling (# .

FM 300 Lecture Note 9 17

Testing the Expectations Theory

If prices are correct on average => error term has zero
average (i.e., an intercept in the regression zero).
Consider a regression:
(#78 − (# !#,#78 − (#
=9+: +6#78
(# (#

If UEH is true => a=0.

FM 300 Lecture Note 9 18

 Testing the Expectations Theory
Result: 9 ≠ 0. Why does the theory fail?
One explanation is risk: the actual spot rate next period may
be very different from the forward rate from the period before.
In well functioning markets risk will be rewarded to the extent
that it cannot be diversified away (remember the CAPM).
In well-functioning financial markets there will be a
relationship between risk and expected return:
E[return]= risk free rate + risk premium
Coefficient ’a’ is the estimate of the risk premium.

FM 300 Lecture Note 9 19

Exchange Rate Movements

and Interest Rate Differentials 1976-2001

FM 300 Lecture Note 9 20

 Purchasing Power Parity, Real Exchange
Rates and Real Interest Rates
So far we have focused on asset (FX) markets.

Now we look at international goods markets.

Corporations sell goods and provide services: need to study

how they are priced internationally.

What is the relationship between international goods and

asset markets?

FM 300 Lecture Note 9 21

Commodity Price Parity (CPP)

If nothing prevents arbitrage in international goods markets,

then identical goods should sell for the same price (when
evaluated in a common currency).
a.k.a “Law of One Price”
If goods do not sell for the same price, then there is an
arbitrage opportunity:
– Buy low;
– Sell high.

FM 300 Lecture Note 9 22

 Commodity Price Parity (CPP)

E.g. Toyota Camry in the US costs USD 23,000; if

S[JPY/USD] = 125, then Toyota Camry in Japan should cost
USD 23000 JPY 125 /USD = JPY 2.875 m.
What if it costs JPY 3.00m?
Buy Camrys in the US for USD 23,000, sell in Japanese car
market for JPY 3.00m.
Profit = S[USD/JPY] JPY 3m – USD 23,000 – cost of
shipping.
CPP works for standardized commodities traded in liquid
markets (e.g. oil, gold) but fails for most other commodities.

FM 300 Lecture Note 9 23

Absolute PPP

Suppose CPP holds on average.

Measure price of bundle of goods using Consumer Price Index.
"# = price of basket of goods in the home country at time t.
"#∗ = price of basket of goods in the foreign country at time t.
Absolute purchasing power parity (PPP):

"# = "#∗ (# ⇒ (# = "# /"#∗

This is a theory of exchange rate determination: exchange
rates are determined in goods markets, not in asset markets.

FM 300 Lecture Note 9 24

 Absolute PPP
Absolute PPP may fail:
– CPP may not hold, even on average;
– Systematic differences between prices.
Represent deviations from PPP by Rt:
(# = ?# "# /"#∗
This relative PPP is a weaker condition than absolute PPP: it
may work when absolute PPP fails.
What if systematic deviations Rt are constant over time?
Then, we can look at relative movements in exchange rates
and price levels over time ignoring changes in Rt.

FM 300 Lecture Note 9 25

Relative movements
Use changes to represent relative movements:
(% − (#
A#,% =
(#
"% − "#
@#,% =
"#

∗
"%∗ − "#∗
@#,% =
"#∗
@#,% = domestic inflation from t to T;
∗
@#,% = foreign inflation from t to T;
Change in R is 0.

FM 300 Lecture Note 9 26

 Relative PPP
Rewriting the equation in terms of changes for constant R:

1 + @#,%
1 + A#,% = ∗
1 + @#,%

Over time, exchange rate moves to offset differences in

inflation rates between the two countries:
∗
A#,% ≈ @#,% − @#,%

Relative PPP fails in short run (1 day to 5 years).

– Exchange rates depend not only on goods markets.
– Exchange rates behave like stock prices over short run:
they have high volatility and are difficult to predict.

FM 300 Lecture Note 9 27

Relative PPP

Some tendency of nominal exchange rates to move with

inflation differentials towards PPP levels in the 5-10 year range.
Absolute PPP holds better in long horizon (10-20 years).
Depends on supply shocks, technology shocks, non-traded
goods and services.

FM 300 Lecture Note 9 28

 Real Exchange Rates

Need a measure that would tell us something about the

relative purchasing power of money in the two countries.
This measure is called real exchange rate, and is given by ?# :
(# ×"#∗
?# =
"#

Intuition: you could buy a number R of UK baskets for every

foreign basket purchased.
If investors can trade freely across countries and the absolute
PPP is satisfied then R=1.

FM 300 Lecture Note 9 29

Real interest rates

Real interest rate is the nominal interest rate r adjusted
for inflation @:
1 + 1#,%
1 + B#,% =
1 + @#,%
For small rates we can have the following approximation:
B#,% ≈ 1#,% − @#,%
This is the real rate of interest earned on a deposit in the
home country over the period from t to T.
Nominal rate is given by (assuming we know inflation):

1#,% ≈ B#,% + @#,%

FM 300 Lecture Note 9 30

 The Fisher equation

Assume investors estimate expected inflation, and care about

expected real rates.
Investors want nominal rates to keep up with inflation.
We get the following equation for the nominal interest rate:
1#,% ≈ 5# [B#,% ] + 5# [@#,% ]
Investors price bonds or loans to maintain their expected real
rate of return.

FM 300 Lecture Note 9 31

The Fisher Open Relation

Investors compare expected real returns in the home and

foreign economies. Then, absent any frictions, we have:
∗
5# B#,% ≈ 5# [B#,% ]
Hence, investors demand a higher nominal interest rate from
countries with higher expected inflation rates.
The above equation and the equation for nominal rates imply:
∗ ∗
1#,% − 1#,% ≈ 5# [@#,% − @#,% ]

FM 300 Lecture Note 9 32

 Appendix: Approximations

The approximations are obtained using the following idea.

Consider, for example, the formula for real interest rates:
1 + 1#,%
1 + B#,% =
1 + @#,%
This formula implies that:
(1 + B#,% )(1 + @#,% ) = 1 + 1#,%
B#,% = 1#,% − @#,% − @#,% ×B#,%
Assuming that @#,% and B#,% are small, we obtain:

B#,% ≈ 1#,% − @#,%

FM 300 Lecture Note 9 33

 FM 300 Section A

International asset allocation

Lecture Note 10
Dr. Georgy Chabakauri
London School of Economics

FM 300 Lecture Note 10 1

Outline

Return on Foreign Investment

Foreign exchange risk
International CAPM
Benefits of international diversification
International investment choices
The home bias

FM 300 Lecture Note 10 2

 Why a global perspective?

A larger range of risk-return choices

– Higher returns on equities can be explained by higher
growth rates in some countries.
Diversification with foreign securities reduces portfolio risk
– Since foreign investments can be influenced by different
factors than domestic investments, risks can be reduced.
Barriers to global investing, both for companies and for
individual investors, are getting smaller.

FM 300 Lecture Note 10 3

Issues

Emphasis for our analysis:

– Risk assessment;
– Diversification benefit.
We will address the following questions:
– What are the risks involved in investment in foreign
securities?
– Are there benefits to diversification in foreign securities?

FM 300 Lecture Note 10 4

 Return on foreign assets

Investing in foreign assets exposes investors to the

variations in exchange rates.

Total return on foreign assets is determined by both the

investment return and movement in exchange rates.

It is often not possible to completely hedge a foreign

investment against the exchange rate risk.

FM 300 Lecture Note 10 5

… Return on foreign assets

Return of a foreign security is a function of two factors:

1. Return on foreign market;
2. Return on the foreign currency.
Example: U.S. investor invests $20,000 in the UK.
– Initial Investment: $20,000;
– Initial Exchange: $2.00/ £1;
– Initial Investment in Pound Sterling: £10,000;
– Risk-free Rate in U.K. 10%;
– Future Value in Pound Sterling: 11,000.

FM 300 Lecture Note 10 6

 … Return on foreign assets

Consider 3 scenarios:
– Pound Depreciates to $1.80, 11,000 1.8 = $19,800;
Return in $ (-200 / 20,000) = -1%;
– Pound Remains at $2.00, 11,000 2.0 = $22,000;
Return in $ (2,000 / 20,000) = 10%;
– Pound Appreciates to $2.20; 11,000 2.20 = $24,200
Return in $ ( 4,200 / 20,000) = 21%.
Movements in foreign exchange can have a major influence.
Both factors must be considered in international investing.

FM 300 Lecture Note 10 7

… Return on foreign assets

Let Pt* be the price of a foreign asset in the foreign currency.
The price in the domestic currency is: Pt = Pt*St, where St is
the exchange rate. Returns on foreign assets in units of
domestic and foreign currency are: r and r*.
Let st be the % change in St.
The return on foreign asset in units of domestic currency r is:

&" &"∗ ,"

!" = −1= ∗ × −1
&"'( &"'( ,"'(
= 1 + !"∗ 1 + $" − 1 = !"∗ + $" + !"∗ $"

Approximation when !"∗ and $" are small: !" ≈ !"∗ + $" .
FM 300 Lecture Note 10 8
 Risk and return in international
investment
The expected dollar return on a foreign asset is:
/[!" ] = /[!"∗ ] + E[$" ] + /[!"∗ $" ]
The dollar variance is:
var !" = var[!"∗ + $" + !"∗ $" ]
= var !"∗ + var $" + interaction terms

What fraction of the return variance comes from:

– asset return variability;
– exchange rate variability?

FM 300 Lecture Note 10 9

…Risk and return in international

investment
Variance of dollar return on foreign stocks

The dominant risk in foreign stock markets comes from the asset return (rf);
Exchange rate variability is less important;
The interaction terms are close to zero.

FM 300 Lecture Note 10 10

 …Risk and return in international
investment
Variance of dollar return on foreign bonds

The dominant risk in foreign bond markets is exchange rate variability;

Bond return variability is less important;
The interaction terms are close to zero.
FM 300 Lecture Note 10 11

Country-specific risk
Country-specific risk can include for example
– political risk, financial risk and economic risk;
– can affect asset returns or exchange rates, or both.
Political Risk Variables:
– Government stability, corruption, etc.
Financial Risk Variables:
– Foreign debt (%GDP), Exchange rate stability, etc.
Economic Risk Variables:
– GDP growth, annual inflation, etc.

FM 300 Lecture Note 10 12

 International Information Risk
Difficult to obtain information on international investments due
to, for example:
– different political, legal systems;
– different financial reporting techniques.

Easier for an ‘insider’ to understand information than an

‘outsider’ because of:
– cultural differences;
– language barriers.

FM 300 Lecture Note 10 13

Hedging FX risk
What happens when we hedge the FX risk?

Consider decomposition: Pt* = E[Pt*] + (Pt* – E[Pt*])

Consider a simple FX hedging strategy at date t-1: home

investor sells forward the expected foreign currency payoff from
the foreign asset E[Pt*], at forward price F (e.g., £F for $1).

Pt* – E[Pt*] is the unexpected foreign currency payoff, and this

must be sold at the future spot rate, St. The hedged proceeds in
home currency are (using superscript “h” for “hedged”):
&"> = / &"∗ ×? + &"∗ − / &"∗ ,"

FM 300 Lecture Note 10 14

 …Hedging FX risk
The hedged return (in home currency) becomes:
&"> / &"∗ ×?" + &"∗ − / &"∗ ,"
!"> = > −1= ∗ −1
&"'( &"'( ,"'(
/ &"∗ ?" &"∗ − / &"∗ ,"
= ∗ × + ∗ × −1
&"'( ,"'( &"'( ,"'(

= 1 + / !"∗ 1 + ?&" + !"∗ − / !"∗ 1 + $" − 1

= !"∗ + ?&" + / !"∗ ?&" − $"

where FP is forward premium.

Then, we get the following approximation:

!"> ≈ !"∗ + ?&"

FM 300 Lecture Note 10 15

…Hedging FX risk
Expected return:
– Unhedged: E[rt] = E[rt*] + E[st];
– Hedged: E[rth] = E[rt*] + FPt;
Variance: since FP is known at date t-1, then:
var(rth) < unhedged var(rt)
This is why FX risk hedging can enhance the risk-return
efficiency of international stock market investment.
Contracts that hedge the whole currency (FX) risk are called
“quanto contracts”. It is like buying forwards with variable
notional amount.

FM 300 Lecture Note 10 16

 International Asset Pricing

Asset prices depend on the level of market segmentation.

Integrated financial markets:
– There are no barriers to financial flows;
– Asset prices in different countries are determined jointly.
Segmented financial markets:
– Prices are set independently in each national market.
Reality is somewhere between these two extremes.

FM 300 Lecture Note 10 17

…International Asset Pricing

Implications for International Asset Pricing are as follows.

Completely integrated markets:
– Systematic risk is measured by the world market.
Completely segmented markets:
– Systematic risk should be measured against the local
(e.g., national) market portfolio.

FM 300 Lecture Note 10 18

 International CAPM
International CAPM explicitly recognizes that investors care
about returns in their own currency:
– US investors care about returns in US $;
– German investors in euro;
– Japanese investors in yen;
– Because the consumption goods are sold in the local
market with prices denominated in the local currency.
This implies that the U.S. T-bond is riskless for U.S. investors,
but not for German or Japanese investors.

FM 300 Lecture Note 10 19

…International CAPM
How do we proceed, then?
– Take some numeraire, say the U.S. dollar (this can be any
currency).
– Convert returns on all assets including foreign ones into $:
1 + ri$ = (1 + ri€)(1 + s)
– For a German investor, convert everything to Euros:
1 + ri€ = (1 + ri$)/(1 + s)
The world aggregate investor, comprising of the U.S., German
and other nationals, care about both the asset returns in the
chosen numeraire (e.g., $) and exchange rate variations.

FM 300 Lecture Note 10 20

 … International CAPM

The market portfolio M in the International CAPM includes all

risky assets in the world weighted according to their market
values (under the same numeraire);
Investors also hold a portfolio of domestic and foreign bonds;
– domestic bonds offer risk-free investment;
– foreign bonds are risky because of FX risk.
In the absence of cross-border frictions, CAPM is given by:
/ !B − !C = DB (/ !@ − !C )

!@ is global market portfolio, !@ , !B , !C are in the same currency.

FM 300 Lecture Note 10 21

…International CAPM

In the case of frictions and PPP violations, expected return on

asset is the sum of the risk-free rate plus the market risk
premium plus various currency risk premia:

/ !B − !C = DB / !@ − !C + G( H&( + ⋯ + GJ H&J ,

RPK are currency risk premia (see Adler and Dumas, 1983).

FM 300 Lecture Note 10 22

 Tests of the ICAPM
Empirical researchers have explored several questions:
– Is the global market risk priced?
– Is domestic market risk priced (segmentation)?
– Is currency risk priced?
A summary of current research tends to support the conclusion
that assets are priced in an integrated global financial market.
The evidence is sufficiently strong to justify using the ICAPM
as an anchor in structuring global portfolios.
However, the evidence can be somewhat different for
emerging smaller markets, in which constraints are still serious.

FM 300 Lecture Note 10 23

Historical stock market performance

1970-2002

FM 300 Lecture Note 10 24

 International diversification

Can we reduce risk through international diversification?

– The lower the correlation coefficient between investments,
the greater the benefit of diversification.
– Foreign stock markets are imperfectly correlated, implying
possible gains from international diversification.
– It is possible to expand the efficient frontier above the
domestic only frontier.
– It is possible to reduce the systematic risk level below the
domestic only level.

FM 300 Lecture Note 10 25

…International diversification

Correlations vary greatly between pairs of countries.

– The degree of independence of a stock market is directly
linked to the independence of a nation’s economy and
governmental policies.
– In general, long-term bond or equity returns are not very
highly correlated across countries
Correlations are increasing over time, with globalization and
as various regulatory barriers have fallen.

FM 300 Lecture Note 10 26

 Diversification benefits
Mean Return (monthly)
3

2.5

2 Efficient set
UK U.S. investing
1.5 JP
FR
US
1 GM
R CN
0.5
f
0
0 2 4 6 8

Standard Deviation (monthly)

FM 300 Lecture Note 10 27

…Diversification benefits

1 stock (30-40% risk)

Portfolio Risk (%)

U.S. stocks (15-20%)

U.S. systematic risk

International stocks (10-15%)

1 10 20 30 40 50
Number of Stocks in portfolio

FM 300 Lecture Note 10 28

 …Diversification benefits

To sum up: considerable benefit of international

diversification.
– Portfolios that are diversified internationally tend to
have substantially lower standard deviations.
Correlations vary greatly between pairs of countries.
– correlation between U.S. and Canadian returns > that
between U.S. and European returns.
Correlations are increasing over time.

FM 300 Lecture Note 10 29

…Diversification benefits

Caveat: Correlations increase dramatically in periods of

crises. So the benefits of international risk diversification
disappear when they are most needed. A phenomenon
referred to as “correlation breakdown”.

FM 300 Lecture Note 10 30

 …International investment choices
Fixed-Income Investments
– Eurobond: an international bond that pays cash flows in a
currency not native to the country of issue.
– Yankee bond: a bond denominated in U.S. dollars, sold in
the U.S., but issued by a foreign corporation/government.
Mutual Funds (issued in the US):
– Global funds: invest in both U.S. and foreign stocks/bonds;
– International funds: invest mostly outside the U.S.;
– Funds can specialize on:
- Diversification across many countries;
- Concentrate in a segment of the world (Global Energy);
- Concentrate in a specific country (Brazil);
- Concentrate in types of markets (Emerging Markets);
FM 300 Lecture Note 10 31

International investment choices

Direct stock/bond purchases.
– The most difficult approach: purchasing stock in the foreign
country (in the foreign currency) and transferring back to the
investor’s home country.
American depository receipts (ADRs)
– Certificates issued by a U.S. bank (represent indirect
ownership of shares of a foreign firm on deposit in the bank).
– Traded in the US market and quoted in U.S. dollars.
– Very popular way of investing.

FM 300 Lecture Note 10 32

 Home bias
Home bias: despite the potential benefits of international
portfolio diversification, most investors tilt their portfolios
towards domestic securities.
National equity as % of total equity holdings:
– Japanese investors: 95%
– U.S. investors: 90%
– Canadians: 89%
– Germans: 82%
– British: 78%
(Tesar & Werner, 1998)

FM 300 Lecture Note 10 33

… Home bias
Portfolio holdings have been changing rapidly:
– U.S investors held only 4% foreign stocks in 1987;
held 10% as of 1996.
– Reflected in balance of payments statistics:
- Large gross purchases of foreign stocks & bonds by
U.S. investors, firms and banks in 1990’s;
- Even larger gross purchases of U.S. stocks & bonds by
foreigners in 1990’s (net capital inflow).
However, investors still mostly hold their own country’s assets.
There are several factors that tilt investor holding toward
domestic assets, discussed below.
FM 300 Lecture Note 10 34
 … Home bias

1. Domestic assets used as a hedge against domestic risk.

– Domestic stocks can hedge domestic inflation risk;
– Banks and insurance companies with domestic liabilities
have an incentive to hedge with domestic assets.
2. Market frictions.
– Government controls: limitations on foreign ownership of
domestic assets intended to keep capital at home.
– Taxes on cross-border transactions (capital gains, etc.)
– Transaction costs influence informational and allocational
efficiency.

FM 300 Lecture Note 10 35

… Home bias
3. Unequal access to information, due to, for example, time
difference, language barrier, etc:
– Difficult to get and interpret information from distant
markets;
– Once invested, it is difficult to monitor the actions of
distant managers.
4. Investor irrationality
– Individuals prefer investments that are culturally similar
and geographically nearby;
– Seeking psychological comfort.

FM 300 Lecture Note 10 36

 … Home bias
Coval and Moskowitz find that U.S. fund managers prefer
local firms, especially small firms that produce non-traded
goods.
Kang and Stulz report that Japanese multinational
corporations (traded in Japan) with greater international
presence (e.g., large export volume) have larger foreign
ownership.

FM 300 Lecture Note 10 37

… Home bias
Grinblatt and Keloharju find that investors in Finland are
more likely to own firms that are:
– located nearby, communicate in their native tongue
(Swedish vs Finnish), or have CEOs of the same
cultural background.
Cohen reports that employees tend to overweight own
company stocks in their personal portfolios.

FM 300 Lecture Note 10 38

 FM 300: Corporate Finance, Investments, and Financial Markets

Section A - Michaelmas Term

Classwork 1

1.
You estimate the index regression for excess returns of stock A, and you get the
following results:
RA = .01 + .80 × RM + εA
σM = 0.20, σεA = .10 (σεA is the residual standard deviation).
What is the volatility of the return for stock A?

2.
Consider two stocks, A and B, with betas of 0.5 and 1.5, respectively. You believe
that the annualized expected return of the market is E[rM] = 10%, the annual standard
deviation of the market is σM = 0.20, and the risk-free rate is 5%.
Finally, you believe that returns of A, B, and the market over the next year are
represented by the regression:
(ri − rf) = αi + βi(rM − rf) + ei, for i = A, B,
in which rA-rf is 0.04, rB-rf is 0.06, the R2 is 0.95 for both A and B, and where eA and
eB are uncorrelated.
Find the abnormal returns of these stocks (in excess of the returns predicted by the
CAPM).
Find the variance of portfolio P, composed of A and B with weights wA=70% and
wB=30%.

3.
Suppose that the index model for stocks A and B is estimated from excess returns
with the following results:

RA = 3% + .7RM + εA;
RB = -2% + 1.2RM + εB;
σM = 20%; Rsq(A)=.20; Rsq(B)=.12;

a. What is the standard deviation of each stock?

b. Break down the variance of each stock into the systematic and firm-specific
components.
c. What are the covariance and correlation coefficient between the two stocks?
d. Are the intercepts of the two regressions consistent with the CAPM? Interpret
their values.
4.
The file Data1.xls (on the course website) contains annual returns for 10 companies,
from 1990 to 2007. Using the functions in excel, perform the following calculations:
a) Find each stock’s average return (you can use the function “AVERAGE”)
b) Find each stock’s standard deviation (you can use the function “STDEV”)
c) Estimate each stock’s alpha, beta, R-squared (you can use the functions
“INTERCEPT”, “SLOPE”, “RSQ”; note that Y is the dependent variable, and
X the independent variable).
d) Find the average return of an equally-weighted portfolio of the 10 stocks
e) Find the standard deviation of the portfolio
f) Check whether this portfolio earns any risk-adjusted excess returns
g) Check whether this portfolio gives any diversification benefit (hint: look at its
R-squared).
Note: to get regression estimates, you can also use the regression tool. Go to “Tools –
Data analysis – regression” (it may be in the Add-ins if you can’t find it). This
estimation tool also gives you standard errors for alpha and beta, as well as the
residual standard error.

Note: in this classwork, the notation for returns and excess returns is the same as in
your textbook:
ri = return of stock i
Ri = ri – rf = excess return of stock i.
 FM 300: Corporate Finance, Investments, and Financial Markets

Section A - Michaelmas Term

Classwork 2

1.
Assume that all stock market returns have the market index as a common factor, and
that all stocks in the economy have a beta of 1 on the market index. The firm-specific
components of stock returns are uncorrelated and have a standard deviation of 30%.
Suppose that an analyst studies 50 stocks, and finds that one-half have an alpha of
2.6%, and the other half have an alpha of -4%. Suppose the analyst buys £1 million of
an equally weighted portfolio of the positive alpha stocks, and sells £1 million of an
equally weighted portfolio of the negative alpha stocks.
a) What are the expected profit (in pounds) and the standard deviation of the analyst’s
profit (in pounds)?
b) What would your answer be if the number of stocks was 100 instead of 50? Explain
your results.

2.
Describe the empirical evidence on the two-pass cross-sectional tests of the CAPM.
Within the framework of these tests, comment on the results of Fama and French
(1992), who find that size and Book-to-Market are significant in explaining the cross-
section of stock returns in the US.

3.
The excel file Data2.xls (on the course website) contains monthly returns for Disney
(RET), from 1995 to 2005. The file also contains the monthly series of T-bill rates
(RF), excess return on the market (MKTRF), and Fama-French factors SMB and
HML. Estimate alphas and factor loadings for the following asset pricing models, and
comment on your results:
a. The CAPM model, using all the data in the series.
b. The three-factor model, using all the data in the series.
c. The three-factor model, using data until December 2000.
 FM 300: Corporate Finance, Investments, and Financial Markets

Section A - Michaelmas Term

Classwork 3

1.
a) Explain why any test of market efficiency is always a joint test.

b) Very risky stocks on average yield higher returns than relatively safe stocks. Is this
a violation of market efficiency?

c) Stocks that have appreciated unusually in the recent past continue to do so. Is this a
violation of market efficiency?

d) A successful firm like Microsoft has consistently generated large profits for years.
Is this a violation of market efficiency?

e) You can earn abnormal returns by buying stocks in December and selling them in
January. Which form of market efficiency does this strategy violate?

2.
Massey Energy (MEE) is a mining company that processes and ships coal, operating
30 underground mines and 13 surface mines in West Virginia, Kentucky and Virginia.
The Company's steam coal is primarily purchased by utilities and industrial clients as
fuel for power plants. On October 24, 2002, after the close of trading, the company
announced its third quarter earnings.

The file data3.xls (on the course web page) contains daily excess returns (i.e. daily
returns in excess of the risk-free return) for MEE, as well as daily excess returns for
the market index (i.e. market returns in excess of the risk-free rate of return).

Plot the daily abnormal returns and the daily cumulative abnormal returns for MEE
over a 41 trading day window starting 10 days before the announcement and ending
30 days after the announcement (from day –10 to day +30). Use the returns during
210 trading days before the event window (days -220 to -11) to estimate the expected
returns for MEE.

Is there any unusual activity at announcement?

Are your results consistent with the efficient market hypothesis?
 FM 300: Corporate Finance, Investments, and Financial Markets

Section A - Michaelmas Term

Classwork 4

1.
The monthly rate of return on T-bills is 1%. The market went up this month by 1.5%.
AmbChaser, Inc. (AC), which has an equity beta of 2, surprisingly just won a lawsuit
that awards it $1 million immediately.
If the original value of AC equity were $100 million, what would you guess was the
rate of return of its stock this month?

2.
Consider a small set of stocks to examine the size anomaly and the momentum
anomaly. The file Data4.xls contains monthly returns on a set of U.S. stocks from July
1998 to December 2000. It also contains the stocks’ market capitalization (cap, in
millions of USD) and cumulative 6-month past return measured at the end of June
1998 (pastret). Furthermore, the file contains the series of monthly market returns in
excess of the risk-free rate (MKT_RF), the Fama-French factors (SMB and HML),
and the risk-free rate (RF).
You are forming portfolios at the beginning of July 1998 (i.e. you are selecting stocks
based on their characteristics as of the end of June 1998).

a) Based on information available at the end of June 1998, select the 3 largest stocks
and the 3 smallest stocks. Form an equally weighted portfolio of large stocks and an
equally weighted portfolio of small stocks. Calculate the average returns of these two
portfolios starting in July 1998 until the end of the sample period. Using the Fama-
French model, estimate the risk-adjusted returns for the portfolio of small stocks, the
portfolio of large stocks, and the strategy that buys the small stocks and sells the large
stocks.
Is there any evidence of a size anomaly?

b) Now consider the past 6-month cumulative return measured at the end of June 1998
to select the top 3 winner stocks and the bottom 3 loser stocks. Form a portfolio of
winners and a portfolio of losers at the beginning of July 1998 and compute the
average future returns to these portfolios. Using the Fama-French model, estimate the
risk-adjusted returns for the winner and the loser portfolios.

Form a momentum portfolio starting in July 1998 and compute the average returns to
this portfolio in the following 3, 6, 12 months. What do you notice?

Estimate alphas and factor loadings from the Fama-French model for the momentum
portfolio. Is there evidence of momentum profits?
 FM 300: Corporate Finance, Investments, and Financial Markets

Section A - Michaelmas Term

Classwork 5

1. You know that firm XYZ is very poorly run. On a scale of 1 (worst) to 10 (best),
you would give it a score of 3. The market consensus evaluation is that the
management score is only 2. Should you buy or sell the stock?

2. What would happen to market efficiency if all investors attempted to follow a

passive strategy?

3. Stocks A, B, and C have a dividend announcement on June 4 2001, September 15

2002, and December 6 2003, respectively. These announcement dates have been
aligned at 0, following standard event study methodology. The table below contains
abnormal returns for the three stocks around the announcement day, as well as the
average abnormal return across the stocks.

A B C Average
-2 -0.0047 -0.0586 0.0077 -0.0185
-1 -0.0238 0.1874 0.1292 0.0976
0 0.1446 0.3536 0.2604 0.2528
1 -0.0516 -0.0353 0.0143 -0.0242
2 -0.1277 -0.0396 0.0577 -0.0366

Plot the average Cumulative Abnormal Return and analyze this event study. Explain
the market efficiency implications of this event study and discuss possible
interpretations in behavioral finance.

4. Assume that the true and unknown return R* of a stock (the return that should
contain all information about a firm’s fundamentals) is independently and identically
distributed over time, with standard deviation σ. Assume further that the observed
return Rt is a combination of last period’s true return and this period’s true return:

Rt = a Rt-1* + (1-a) Rt*.

Think of 0<a<1 as underreaction and a<0 as overreaction by investors to information.

(a) Derive formulas for the variance and the autocorrelation of observed returns.
(b) Evaluate the autocorrelation obtained above for a = 1/3 and a = –1/3 and compare.
What conclusions can we draw?

Hint: Derive the autocorrelation of observed returns starting from the autocovariance,
defined as Cov(Rt , Rt-1), and substitute the formula above. Remember that, for i.i.d.
returns, the covariance is always zero (independence) and the variance is always σ2
(identical distribution).
 FM 300: Corporate Finance, Investments, and Financial Markets

Section A - Michaelmas Term

Classwork 6

1. Assume a constant annual interest rate of 5%. A bank has an asset base that
delivers cash flows for the next 5 years as shown in the table below. It wishes to
structure its liabilities so as to match the value and duration of its assets. Two bonds
(A and B) are available to the bank as liabilities and their respective cash flows are
shown below.

Time 1 2 3 4 5
Assets 400 450 600 200 300
Bond A 8 8 8 8 108
Bond B 0 0 0 100 0

a. Derive the quantities of the two bonds that the bank should use to meet its
objectives.
b. Check that the value and duration of the bond portfolio match those of the asset
cash flow stream.

2. You are managing a bond portfolio of $1 million. Your target duration is 10 years,
and you can choose from two bonds: a zero-coupon bond with maturity 5 years, and a
perpetuity currently yielding 5%.
a) How much of each bond will you hold in your portfolio?
b) How will these fractions change next year if target duration becomes 9 years?

3. A 30-year maturity bond making annual coupon payments with a coupon rate of
12% has duration of 11.54 years and convexity of 192.4. The bond currently sells at a
yield to maturity of 8%. Use a financial calculator or spreadsheet to find the price of
the bond if its yield to maturity falls to 7% or rises to 9%.
What prices for the bond at these new yields would be predicted by the duration rule
and the duration-with-convexity rule? What is the percentage error of each rule? What
do you conclude about the accuracy of the two rules?
 FM 300: Corporate Finance, Investments, and Financial Markets

Section A - Michaelmas Term

Classwork 7

1. Consider the following information regarding the performance of a money manager

in a recent month. The table shows the actual return of each sector of the manager’s
portfolio, the fraction of the portfolio allocated to each sector, the benchmark or
neutral sector allocations, and the return of sector indices.

Actual return Actual w Benchmark w Index return

Equity 2% 0.70 0.60 2.5%
Bonds 1% 0.20 0.30 1.2%
Cash 0.5% 0.10 0.10 0.5%

a) What was the manager’s return in the month? What was her overperformance or
underperformance?
b) What was the contribution of security selection to relative performance?
c) What was the contribution of asset allocation to relative performance? Confirm that
the sum of selection and allocation contributions equals the total excess return
(relative to the benchmark).

2.
Consider the following time-series regression estimates of the monthly returns of two
funds W and Z:
rW-rf = -0.1 + 1.1(rM-rf) + 0.7(rM-rf)2 + eW
rZ-rf = +0.1 + 0.9(rM-rf) + 0.0(rM-rf)2 + ez

What do these regressions suggest about the timing and security selection ability of
the two funds? (Assume all coefficient estimates are statistically significant at 5 per
cent level)
Do you know of alternative ways to estimate the market timing of an asset manager?
Explain.

3.
Assume the CAPM properly prices all assets.
However, your analyst has chosen three active fund managers who are supposedly
very capable at stock picking. Your analyst supplies you with the following data on
the three funds, the market, and the risk-free asset. All fees have already been
deducted from these returns. Further, assume that none of these managers have any
market or factor timing ability.
 Security Expected Beta Standard
Return Deviation
Risk-Free Asset 5% 0.0 0%
Market Portfolio 15% 1.0 10%
Fund A 36% 2.0 35%
Fund B 41% 3.0 35%
Fund C 10% 0.5 10%

a) If you could hold either the market or one of the funds, what would you choose?
Explain.
b) If you could combine one of the funds with the market portfolio, what would you
choose? Explain. Can you calculate the Sharpe-Ratio that your overall investment
would obtain?
c) Now assume that you are in charge of a pension fund, and your goal is to maximize
the Sharpe Ratio of your fund. However, you are required to use some combination of
Fund A and the Market. Assume that Fund A’s fees are currently 1% per year.
However, Fund A is going to raise its fees. How large would Fund A’s fees have to be
so that you would put nothing into Fund A?
 FM 300: Corporate Finance, Investments, and Financial Markets

Section A - Michaelmas Term

Classwork 8

1. Jason Smith is a foreign exchange trader with Citibank. He notices the following
quotes:
Spot exchange rate SFr/$ 1.6627
6-month forward exchange rate SFr/$ 1.6558
6-month $ interest rate 3.5% p.a.
6-month SFr interest rate 3.0% p.a.

(Consider the $ as the domestic currency here)

a) Ignoring transaction costs, is covered interest rate parity holding? Is there an
arbitrage opportunity?
b) If yes, what steps would be needed to make an arbitrage profit? Assuming that
Jason Smith is authorized to work with $1,000,000 for this purpose, how much would
the arbitrage profit be in dollars?

2. A French company is importing some equipment from Switzerland and will need to
pay 10 million Swiss francs 3 months from now. The current spot exchange rate is
SFr/€ 1.5543. The treasurer of the company expects the franc to appreciate in the next
few weeks, and is concerned about it. The 3-month forward rate is SFr/€ 1.5320.
(Treat the Euro as the domestic currency).
a) Given the treasurer’s expectation, what action can he take using the forward
contract?
b) Three months later, the spot exchange rate turns out to be SFr/€ 1.5101. Did the
company benefit from the treasurer’s action?

3. Use the table below to answer the following questions:

Price of copper in the Spot and Futures Markets

(in units of 25,000 lbs)
Delivery date Copper in NY Copper in London
Today (t) USD 100 GPB 40
In one year (t+1) USD 150 GBP 50

a) What is the current exchange rate of the dollar (in USD/GBP) according to the
commodity purchasing power parity theory?
b) If the actual exchange rate today is S[USD/GBP]=2.00, different from the answer
in part (a) solely because of transportation costs, what is the transportation cost for
one unit of copper in USD?
c) Using the actual exchange rate of S[USD/GBP]=2.00, what is the exchange rate
expected to be in one year according to relative PPP, assuming copper is the only
commodity in the world?

4. Assume that the one-year interest rate is 12% in the U.K. The expected annual rate
of inflation for the coming year is 10% in the U.K. and 4% in Switzerland. The
current spot exchange rate is SFr/£ 3.
Using the precise form of the international parity relations, compute the one-year
interest rate in Switzerland, the expected Swiss franc to pound exchange rate in one
year, and the one-year forward exchange rate.
 FM 300: Corporate Finance, Investments, and Financial Markets

Section A - Michaelmas Term

Classwork 9

1.
Calculate the contribution to total performance from currency, country, and stock
selection for the manager in the example below.

EAFE Return on S1/S0 Manager’s Manager’s

weight equity index weight return
Europe 0.30 20% 0.9 0.35 18%
Australia 0.10 15% 1.0 0.15 20%
Far East 0.60 25% 1.1 0.50 20%

2.
You work at a large US money management firm, where you are in charge of
investing in foreign corporate bonds and stocks and managing the foreign exchange
risk of your portfolio. Your company has purchased bonds and stocks issued by a
large Japanese company, Fuji Electronics. You need to evaluate your company’s
positions in Japanese assets.
One-year US government bonds trade at a yield of 4.75% p.a. One-year Japanese
government bonds trade at a yield of 1.00% p.a. The term structures are flat in both
countries. The current spot exchange rate is S[JPY/USD] = 100.00.
(Throughout this problem, it is easier to work with prices and returns in Japanese
Yen).

(i) You estimate the following regression:

rt = a + b rMt + c ΔS[JPY/USD] t + et
where rt is the percent return (in JPY) on Fuji’s stock at time t, rMt is the percent
return (in JPY) on the Japanese stock market at time t, and ΔS[JPY/USD] t is the
percent change in the Yen value of the US dollar at time t. The regression results are:

Coefficient Estimate t-statistic

a 0.05 1.25
b 0.85 2.35
c 1.23 2.50

Interpret the results of this regression.

(ii) You need to calculate the expected one-year holding period return on the position
in Fuji’s stock. You use the following forecast of the global economies based on
whether there will be a global recession next year. The likelihood of a global
recession is estimated to be 50%.

If global recession If NO global recession

Japanese market return 5.00% 15.00%
US market return 10.00% 30.00%
S[JPY/USD] at end of year 120.00 (or +20%) 72.84 (or -27.16%)

What should your company expect to earn on Fuji’s stock over the next year? (Use
the regression in part (i) and the table in part (ii)).

(iii) You do not know whether to trust these forecasts, and in particular you want to
understand the view of the economic forecaster on the expected exchange rate.
Explain what theory of exchange rates the economist did or did not use and whether it
is reasonable.

3.
Suppose that you are an investor based in Switzerland, and you expect the U.S. dollar
to depreciate by 2.75% over the next year. The interest rate on one-year risk-free
bonds is 5.25% in the U.S. and 2.75% in Switzerland. The current exchange rate is
SFr/USD=1.62.
Calculate the foreign currency risk premium from the Swiss investor’s viewpoint.
 FM 300: Corporate Finance, Investments, and Financial Markets
Section A - Michaelmas Term

Classwork 10: Sample exam questions for Section A

Note: Each question is worth 25 marks; you will need to answer TWO questions in each
section of the exam.

Question 1

(a) (5 marks) An investor who puts $10,000 in T-bills and $20,000 in the market portfolio
will have a beta of 2.0. Is this statement true or false? Explain.

(b) (6 marks) Investors expect the market return in the coming year to be 12%. The
Treasury bill rate is 4%. Procter & Gamble’s (PG) stock has a beta of 0.50. The
market value of its outstanding equity is $100 million.
Calculate your estimate of the expected return on PG stock. If the market return in the
coming year actually turns out to be 10%, what is your best guess of the return that
will be earned by PG stock?
Suppose now that PG wins a major lawsuit during the year. The settlement is $5
million. PG’s stock return during the year is 10%. Assume that the magnitude of the
settlement is the only unexpected firm-specific news during the year. What is your
estimate of the settlement that the market previously expected PG to receive from the
lawsuit? Continue to assume that the market return in the year is 10%.
Carefully explain your reasoning in the framework of the market efficiency tests
based on event studies.

(c) (6 marks) The following table represents the main result of the paper “The interaction
of value and momentum strategies” by C. Asness, Financial Analyst Journal, 1997. It
reports monthly percentage returns of 25 portfolios, sorted on 5 Book-to-Market
groups (from low B/M to high B/M) and 5 past returns groups (from losers to
winners). All t-statistics (not reported in this table) are greater than 2.
Is there evidence of an interaction between momentum and value strategies? Can
higher profits be obtained by taking both past returns and B/M into account? Explain.

Low 2 3 4 High Return of

B/M B/M high-low BM
1 - loser 0.03 0.49 0.80 0.83 1 0.97
2 0.61 0.59 0.90 1.25 1.35 0.74
3 0.52 0.93 0.80 1.19 1.44 0.92
4 0.99 0.97 1.17 1.45 1.68 0.69
5 - winner 1.50 1.44 1.49 1.60 1.62 0.13
Return of winner-loser 1.47 0.95 0.69 0.76 0.62
 (d) (8 marks) Assume that:
 the market portfolio is a 60-40 combination of stocks and bonds;
 the standard deviations of the returns on stocks and bonds are 20% and 10%;
 the correlation between the returns on stocks and bonds is 0.25;
 the Sharpe ratio of the market portfolio is 0.40.
Under the CAPM, what are the expected returns on stocks and bonds, in excess of the
risk-free rate?

Question 2

(a) (i) (6 marks) What is “momentum”? Describe in detail how momentum strategies are
implemented. Why can momentum portfolios be called hedge portfolios?
(ii) (6 marks) Discuss one rational explanation and one behavioral explanation for the
profitability of momentum strategies.

(b) (5 marks) Consider the following two excess return index-model regression results for
stocks A and B. The risk-free rate over the period was 6%, and the market’s average
return was 14%. Performance is measured using an index model regression on excess
returns.
Stock A Stock B
Index Model Regression Estimates 1% + 1.2(rM-rf) 2% + 0.8(rM-rf)
Residual Standard Deviation 10.3% 19.1%
Standard Deviation of Excess Return 21.6% 24.9%

For each stock, calculate Jensen’s alpha, the appraisal ratio, the Sharpe measure, and
the Treynor measure.
Discuss which stock is the best choice under the following circumstances:
 This is the only risky asset to be held by the investor.
 This stock will be mixed with the rest of the investor’s portfolio, currently
composed solely of holdings in the market index fund.
 This is one of many stocks that the investor is analyzing to form an actively
managed stock portfolio.

(c) (8 marks) A global equity manager is assigned to select stocks from a universe of
large stocks throughout the world. The manager will be evaluated by comparing his
returns to the return of the MSCI World Market Portfolio, but he is free to hold stocks
from various countries in whatever proportions he finds desirable. Results for a given
month are contained in the following table:

Country Weight in Manager’s Manager’s return Return of index

MSCI index Weight in country x for country x
U.K. 0.15 0.30 20% 12%
Japan 0.30 0.10 15% 15%
U.S. 0.45 0.40 10% 14%
Germany 0.10 0.20 5% 12%
 Calculate the total value added of all the manager’s decisions this period, and
distinguish between the value added by the manager’s country allocation decisions
and the value added by the manager’s stock selection ability within countries.
Explain why it is important to distinguish between stock picking ability and asset
allocation ability in performance attribution.

Question 3

(a) You are advising the owner of a Spanish trading company that has just signed a
contract for a sale in the U.S. The revenue of $ 50 million is due in three months. The
owner believes that the dollar is overvalued, and is worried about potential losses
from the sale. The current spot exchange rate between the dollar and the Euro is
S[USD/EUR]=0.900. The three-month Euro-dollar interest rate is 2% per annum, and
the three-month Euro-Euro interest rate is 3% per annum.

(i) (5 marks) Suppose you advise the owner of the trading company to use a forward
contract to hedge the risk he faces. What theory do you use to calculate the forward
rate? How many Euros will the owner of the company receive in three months? Can
investors trust the theory you just used for your calculations? Explain.

(ii) (5 marks) A trader working at a hedge fund has speculated by doing a “carry
trade”: he has borrowed USD at the Euro-Dollar rate for three months to invest them
in the higher yielding Euro-deposits for the same period of time. His forecast of the
exchange rate in three months is EtS[USD/EUR]t+3 = 0.850. If he takes a USD 100
million position, how much money does he expect to make? What risk is he
compensated for?
If he believed in the theory of Uncovered Interest Parity, what would be his forecast
of the spot exchange rate in three months? Can investors trust this theory?

(b) (5 marks) You are a U.S. investor considering the purchase of one of the following
securities. Assume that the currency risk of the Canadian government bond will be
hedged, and the 6-month discount on Canadian dollar forward contracts is –0.75%
versus the U.S. dollar.

Bond Maturity Coupon Price

U.S. 6 months 6.50% 100
Canadian 6 months 7.50% 100

Calculate the expected price change required in the Canadian government bond which
would result in the two bonds having equal returns in U.S. dollars over a 6-month
horizon. Assume that the yield on the U.S. bond is expected to remain unchanged.

(c) (10 marks) You are a U.S. investor who is considering investments in the French
(stocks A and B) and Swiss (stocks C and D) stock markets. The world market risk
premium is 6%. The currency risk premium on the Swiss franc is 1.25%, and the
currency risk premium on the euro is 2%. The interest rate on one-year risk-free
bonds is 3.75% in the U.S. In addition, you are provided with the following
information:

Stock A B C D
βW 1 0.90 1 1.5
γ€ 1 0.80 -0.25 -1.0
γSFr -0.25 0.75 1.0 -0.5

Calculate the expected return for each of the stocks (the U.S. dollar is the base
currency). Explain the differences in the expected returns of the four stocks in terms
of βW, γ€, and γSFr. Carefully explain your arguments.