Business School

Risk and Actuarial Studies

ACTL4303 AND ACTL5303

ASSET LIABILITY MANAGEMENT

Week 2
Risk, Return and Portfolio Theory
Greg Vaughan

Course Coordinators
Greg Vaughan
Sessional Lecturer for the course
Email: [email protected]
Consultation hours: by appointment (Mondays 5-6pm)

Professor Michael Sherris

Telephone: 9385 2333
Email: [email protected]
Consultation hours: by appointment (Mondays 4-5pm and Wednesdays
5-6pm)

This weeks coverage

Bodie et al
Chapter 5 Risk, Return, and the Historical Record
Chapter 6 Capital Allocation to Risky Assets
Chapter 7 Optimal Risky Portfolios
Chapter 8 Index Models

Historical returns, risk premiums, individual portfolio

optimisation

Figure 5.1 Determination of the Equilibrium Real

Rate of Interest

Is this how bond yields work in the real world?

4

Australian bond supply has expanded significantly

in recent years with no impact on yields

Supply/Demand only impacts bond yields in extreme situations (eg US

Quantitative Easing). Inflation and global context are more important.
5

Australian Bond Yields 1921-2013

Real vs. Nominal Rates

Fisher effect: Approximation
nominal rate = real rate + inflation premium
R = r + i or r = R - i
Example r = 3%, i = 6%
R = 9% = 3% + 6% or 3% = 9% - 6%
Fisher effect: Exact
r = (R - i) / (1 + i)
2.83% = (9%-6%) / (1.06)
Empirical Relationship:
Inflation and interest rates move closely together

Taxes and the Real Rate of Interest

Tax liabili(es are based on nominal income
Given a tax rate (t) and nominal interest rate (rn),
the real a8er-tax rate is:

rn (1 t ) i = ( rr + i )(1 t ) i = rr (1 t ) it
The a8er-tax real rate of return falls as the
ina(on rate rises

8

INVESTMENTS | BODIE, KANE, MARCUS

The Taylor Rule (1993)

A description of monetary policy in response to
changes in inflation and output
*

it = t + r + a
i

Where

(y y )
t

is the target short-term nominal interest rate

t + ay

is the assumed equilibrium real interest rate (eg 2-3%)

is the current rate if inflation
is the desired rate of inflation (eg 2-3%)

y is the current rate of GDP growth

t

y is the potential rate of GDP growth ( approx 3%)

a , a are policy parameters (originally proposed at 0.5)

Monetary Policy - The Taylor Rule

Implies the central bank will adjust nominal interest rates
more than one for one in response to any change in
inflation ( a > 0), known as the Taylor principle
If inflation is well managed, output fluctuation should be
well behaved so there is an argument that a y is zero
Hence monetary policy can emphasise inflation control, so
long as inflation is well behaved (but beware stagflation)
In the Australian context the exchange rate is a key
consideration in monetary policy setting, as it effects growth
and inflation.
Other versions substitute unemployment for GDP growth,
as appears to be current US policy.
Why are current Australian settings easier than this rule?
10

Recent monetary policy in the US

In the US there have been three rounds of significant
Quantitative Easing (the US Fed buys bonds with printed
money)
This has significantly increased money supply,
But the inflationary consequences have been mute
because velocity of money has been slow
Keynes first described the liquidity trap where increases in
money supply no longer have effect on real activity
However asset prices (equity and property) have risen
generating a wealth effect and indirect impact on activity
The US has begin tapering its Quantitative Easing as
unemployment has fallen

11

Monetary policy from an Australian perspective

Monetary policy setting in Australia takes into account the
strength of the currency
A strong $A makes life difficult for exporters and importcompeting companies
If the dollar is strong, monetary policy will be less restrictive
than otherwise
A rise in interest rates may mean tighter monetary policy, or
less loose monetary policy
The neutral real cash rate varies by circumstances but
2-3% is a guide.
Monetary policy transmission is more efficient in Australia
than in the US because home mortgages are typically
floating rate rather than fixed rate.
12

Australian monetary policy since inflation targeting

13

The high Australian dollar has impacted monetary policy

14

Yield curve slope and economic growth (US)

15

16

17

18

Rates of Return: Single Period

P
1 P 0 + D1
HPR =
P0
HPR = Holding Period Return
P0 = Beginning price
P1 = Ending price
D1 = Dividend during period one

This formula is applicable for individual securities where

distributed income needs to be added back. It is not the
correct formula for portfolio returns.

19

Market history or scenario analysis?

History may not represent a stationary return distribution because many
things change over time:
structure of the economy (resource relevance, manufacturing versus
service orientation)
economic policy context (eg floating currency, industry protection,
fiscal discipline, terms of trade)
inflation environment (supply side shocks versus targeting era)
credit conditions (pre or post GFC)
Market participants (institutions versus individuals)
Historical analysis to inform expectations requires accurate return data
and contextual knowledge.
How is the current environment structurally different from the past?

20

Scenario Returns: Example

State
Prob. of State
r in State
Excellent
.25
0.3100
Good
.45
0.1400
Poor
.25
-0.0675
Crash
.05
-0.5200

E(r) = (.25)(.31) + (.45)(.14) + (.25)(.0675) + (0.05)( 0.52)
E(r) = .0976 or 9.76%
21

INVESTMENTS | BODIE, KANE, MARCUS

Scenario analysis can be very subjective

Do the scenarios span the distribution of possible
outcomes or are they too narrow?
How objective are the scenario probabilities?
What is the basis for asset performance under different
scenarios?
Do expected returns, variances and covariances look
sensible compared to history?
Scenario analysis is more relevant to the short term when
skewed outcomes are perceptible. As horizon lengthens,
history may be a more reliable guide.

22

Geometric Average Real Equity Returns

1960-2014

Source: Credit Suisse Global Investment Returns Sourcebook 2015

23

Geometric Average Equity Risk Premium over

Bonds 1960-2014

Source: Credit Suisse Global Investment Returns Sourcebook 2015

24

Equity Return Volatility 1900-2014

(excluding very high inflation markets)

Source: Credit Suisse Global Investment Returns Sourcebook 2015

25

Arithmetic and geometric returns

E(Geometric Average) = E(Arithmetic average)

1 2
2

IF we were observing a stationary return distribution:

The sample arithmetic average (eg over 55 years) is an unbiased
estimate of the true mean
However compounding at this rate results in an upwardly biased
forecast for a given horizon (eg over ten years)
The unbiased estimator for a projection of H years, based on a data
sample of T years is
(H/T) x Geometric + (1-H/T) x Arithmetic
For a short projection horizon, the Arithmetic mean is appropriate, but
where the horizon is longer, or the data span is shorter, the geometric
mean becomes increasingly relevant

26

Arithmetic and geometric returns

E(Geometric Average) = E(Arithmetic average)

1 2
2

Based on the global history some reasonable numbers would be:

Geometric average real equity return of 5.5%
Geometric average equity premium over bonds of 2.5%
Equity volatility relative to bonds/inflation of 20% (relative to
inflation or bonds it is slightly lower than nominal volatility)
So arithmetic average real equity return would be 7.5%
Arithmetic average premium over bonds would be 4.5%
These numbers are consistent with a historic real bond return of
3% globally
Currently real bond yields are typically below 1%
27

The Normal Distribution

Investment management is easier when returns are
normal
Standard devia(on is a good measure of risk when
returns are symmetric
If security returns are symmetric, porVolio returns will
be as well
Future scenarios can be es(mated using only the
mean and the standard devia(on
The dependence of returns across securi(es can be
summarized using only the pairwise correla(on
coecients
28

INVESTMENTS | BODIE, KANE, MARCUS

Normality and Risk Measures

What if excess returns are not normally
distributed?
Standard devia(on is no longer a complete
measure of risk
Sharpe ra(o is not a complete measure of
porVolio performance
Need to consider skewness and kurtosis

29

INVESTMENTS | BODIE, KANE, MARCUS

Normality and Risk Measures

Value at Risk (VaR)
Loss corresponding to a very low percen(le of the
en(re return distribu(on, such as the 8h or rst
percen(le return

Expected ShorVall (ES)

Also called condi(onal tail expecta(on (CTE),
focuses on the expected loss in the worst-case
scenario (le8 tail of the distribu(on)
More conserva(ve measure of downside risk than
VaR
30

INVESTMENTS | BODIE, KANE, MARCUS

Australian Equity Log Returns 1980-2012

Frequency

31

Skewness

Excess
Kurtosis

Probability
of Normality

Monthly

-3.1

30.3

0%

Quarterly

-1.8

8.8

0%

Annual

-0.5

0.8

56%

Chapter Six Overview

Risk aversion and its es(ma(on
Two-step process of porVolio construc(on
Composi(on of risky porVolio
Capital alloca(on between risky and risk-free
assets

Passive strategies and the capital market line

(CML)

32

INVESTMENTS | BODIE, KANE, MARCUS

Risk and Risk Aversion

Specula)on

Gamble

Taking considerable risk Bet on an uncertain

for a commensurate
outcome for enjoyment
gain
Par(es assign the
Par(es have
heterogeneous
expecta(ons

33

same probabili(es to
the possible outcomes

INVESTMENTS | BODIE, KANE, MARCUS

Risk and Risk Aversion

U(lity Values
Investors are willing to consider:
Risk-free assets
Specula(ve posi(ons with posi(ve risk
premiums
PorVolio afrac(veness increases with expected
return and decreases with risk

34

INVESTMENTS | BODIE, KANE, MARCUS

Table 6.1 Available Risky Portfolios

Each porVolio receives a u(lity score to assess

the investors risk/return trade o

35

INVESTMENTS | BODIE, KANE, MARCUS

Risk Aversion and Utility Values

U(lity Func(on

U = U(lity
E(r) = Expected return on the asset or porVolio
A = Coecient of risk aversion
2 = Variance of returns
= A scaling factor
2
1
U = E (r )
A
2

36

INVESTMENTS | BODIE, KANE, MARCUS

Whats missing from this Utility function ?

Higher moments of the return distribution (although not
always significant)
Liquidity
Income preference over capital appreciation (eg postretirement investment)
Inflation hedging characteristics of assets

37

Risk, Uncertainty and Risk Aversion

What is risk?
Not solely about assets
Depends on the liability profile
Probability of loss relative to liabilities
Knightian Uncertainty
Probabilities not known (eg shape of distribution tails)
Risk aversion (Von Neumann and Morgenstern)
Utility increases with wealth
Per-dollar increment to utility decreases with wealth
Losses are more painful than gains are enjoyable

38

Figure 6.4 The Investment Opportunity Set

39

Figure 6.5 The Opportunity Set with

Differential Borrowing and Lending Rates

40

Portfolios of One Risky and a Risk-Free Asset

r

= y r p + (1 y ) r f

( (r ) r )

E (r c ) = r f + y E

1
2
U = E (r c ) A c
2

U = r f + y #$ E

(r )
p

2
1
2
r f %& A y p
2

Utility is maximized where:

y

41

(r ) r
p

2
p

Figure 6.6 Utility as a Function of

Allocation to the Risky Asset, y

42

INVESTMENTS | BODIE, KANE, MARCUS

Figure 6.8 Finding the

Optimal Complete Portfolio

43

INVESTMENTS | BODIE, KANE, MARCUS

Chapter 7 Overview
The investment decision:
Capital alloca(on (risky vs. risk-free)
Asset alloca(on (construc(on of the risky
porVolio)
Security selec(on

Op(mal risky porVolio

The Markowitz porVolio op(miza(on model
Long- vs. short-term inves(ng
44

INVESTMENTS | BODIE, KANE, MARCUS

Portfolios of Two Risky Assets:

Risk
PorVolio variance:
p2 = wD2 D2 + wE2 E2 + 2wD wE Cov ( rD , rE )
2

D
= Bond variance


E2 = Equity variance

Cov
( r D , r E ) = Covariance of returns for bond
and equity
45

INVESTMENTS | BODIE, KANE, MARCUS

Figure 7.8 Determination of the Optimal

Overall Portfolio

46

INVESTMENTS | BODIE, KANE, MARCUS

Figure 7.10 The Minimum-Variance

Frontier of Risky Assets

47

INVESTMENTS | BODIE, KANE, MARCUS

Markowitz Portfolio Optimization

Model
Search for the CAL with the highest reward-to-
variability ra(o
Everyone invests in P, regardless of their
degree of risk aversion
More risk averse investors put more in the risk-
free asset
Less risk averse investors put more in P

48

INVESTMENTS | BODIE, KANE, MARCUS

Markowitz Portfolio Optimization

Model
Capital Alloca(on and the Separa(on Property
PorVolio choice problem may be separated into
two independent tasks
Determina(on of the op(mal risky porVolio is
purely technical
Alloca(on of the complete porVolio to risk-free
versus the risky porVolio depends on personal
preference

49

INVESTMENTS | BODIE, KANE, MARCUS

Australian Super Investment Choices

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%

Cash
Fixed Interest
Private Equity
Infrastructure
Direct Property
International Shares
Australian Shares

50

Australian Super Investment Choices

Expected
frequency
of negative
years in 20

Short Term
Risk

Long Term
Risk of
lagging
inflation

2 to 3

Medium

Low

Conservative 3 to 4
Balanced

Medium to
High

Low

Balanced

High

Low

High Growth

4 to 5

High

Low

Stable

51

Markowitz Portfolio Optimization

Model
Op(mal PorVolios and Nonnormal Returns
Fat-tailed distribu(ons can result in extreme
values of VaR and ES and encourage smaller
alloca(ons to the risky porVolio
If other porVolios provide suciently befer VaR
and ES values than the mean-variance ecient
porVolio, we may prefer these when faced with
fat-tailed distribu(ons

52

INVESTMENTS | BODIE, KANE, MARCUS

Chapter 8 Overview
Advantages of a single-factor model
Risk decomposi(on
Systema(c vs. rm-specic

Single-index model and its es(ma(on

Op(mal risky porVolio in the index model
Index model vs. Markowitz procedure

53

INVESTMENTS | BODIE, KANE, MARCUS

A Single-Factor Market
Advantages
Reduces the number of inputs for diversica(on
Easier for security analysts to specialize

Model

ri = E (ri ) + i m + ei

i = response of an individual securitys return to

the common factor, m; measure of systema(c risk
m = a common macroeconomic factor
ei = rm-specic surprises
54

INVESTMENTS | BODIE, KANE, MARCUS

From a single-factor model to a single-index

model
The point of the single-factor model is to vastly reduce the
number of inputs required in the Markowitz model by
simplifying the covariance structure.
A reasonable approach to making the single factor
operational is to assert that the rate of return on a broad
index of securities .. is a valid proxy for the common
macroeconomic factor

55

The security characteristic line

RHP (t ) = HP + HP RS &P500 (t ) + eHP (t )

56

Single-Index Model
Regression equa(on:

Ri (t ) = i + i RM (t ) + ei (t )
Expected return-beta rela(onship:

E (Ri ) = i + i E (RM )

57

INVESTMENTS | BODIE, KANE, MARCUS

Single-Index Model
Variance = Systema(c risk + Firm-specic risk:
2
i

2
i

2
M

= + (ei )
Covariance = Product of betas Market index
risk:

Cov ( ri , rj ) = i j
58

2
M

INVESTMENTS | BODIE, KANE, MARCUS

Index Model and Diversication

Variance of the equally-weighted porVolio of
rm-specic components:
n

1 2
1 2
(e p ) = (ei ) = (e )
n
i =1 n
2

When n gets large, 2(ep) becomes negligible

and rm specic risk is diversied away
59

INVESTMENTS | BODIE, KANE, MARCUS

Figure 8.1 The Variance of an Equally

Weighted Portfolio with Risk Coecient p

60

INVESTMENTS | BODIE, KANE, MARCUS

Portfolio Construction and

the Single-Index Model
Alpha and Security Analysis
1. Use macroeconomic analysis to es(mate the risk
premium and risk of the market index.
2. Use sta(s(cal analysis to es(mate the beta
coecients of all securi(es and their residual
variances, 2(ei).
3. Establish the expected return of each security absent
any contribu7on from security analysis.
4. Use security analysis to develop private forecasts of
the expected returns for each security.
61

INVESTMENTS | BODIE, KANE, MARCUS

Portfolio Construction and

the Single-Index Model
Single-Index Model Input List
1. Risk premium on the S&P 500 porVolio
2. Es(mate of the SD of the S&P 500 porVolio
3. n sets of es(mates of
Beta coecient
Stock residual variances
Alpha values

62

INVESTMENTS | BODIE, KANE, MARCUS

Portfolio Construction and the Single-Index Model

Managers are assessed on their information ratios: (e )

These relate to their active portfolios. A stocks weight in the active
portfolio is the difference between its portfolio weight and its weight in
the index
A

A
i

= wi wi

Key result: If not for the long only constraint (

index model implies

A
i

P
i

>0

) the single-

(e )
i

..we are concerned only with the aggregate beta of the active
portfolio, rather than the beta of each individual security.
Normally this is zero by design, so that portfolio beta is one.
Active weights are scaled based on target tracking error
63

Portfolio Construction and the Single-Index Model

For example if a manager estimates the alpha of two
stocks to be the same (5%), but the stock specific volatility
of stock A is 25% compared to 40% for stock B, which of
the following active weight combinations is approximately
correct:
a) 4% overweight in A and B
b) 4% overweight in A and 2.5% in B
c) 2.5% overweight in A and 4.0% in B
d) 5% overweight in A and 2% in B
The correct answer is d)

64

Portfolio Construction and the Single-Index Model

We have sound theory and solid empirical evidence across many
markets to support an expectation of a positive market risk premium

E( R M ) > 0

However there is no theory nor consistently reliable empirical

evidence to suggest that portfolio management activity adds value,
especially after fees

A > 0?
These expectations are of very different qualities. The framework of
Section 8.4 supports a significant amplification of active risk under
typical assumptions (eg 1% alpha, 2% tracking error). In practice
funds are more measured with active risk.
65

Next Week
Bodie, Kane and Marcus
Chapter 9 . Study 9.1 carefully. Rest of chapter is
background
Chapter 10. Study full chapter.
Chapter 11. Study 11.1 . Rest of chapter is background
Read assigned papers, focusing on conclusions and
summaries.

66

Next Week
Student guest contributions:
Group B: CAPM v APT
Group B: Reading, Understanding Momentum
Group C: Reading, Where is the Value Premium?
Group C: Reading, The Low Risk Anomaly: Decomposition
into Micro and Macro Effects

67

6-67

Thank you for your attention

Greg Vaughan
School of Risk and Actuarial Studies
University of New South Wales

68