Lecture 3

Risk and Return

Anna Pavlova
London Business School

london.edu
 Outline

• Returns and Variability: A Paradox?

• Diversification
• Portfolio Theory:
• A Two-Asset Portfolio.
• Many Risky Assets

• Appendix 1: Adding a Riskless Asset

• Appendix 2: Portfolio Frontier with More than Two Assets
• Appendix 3: Expectation, Variance and Standard Deviation

london.edu Lecture 3: Risk and Return 2

 Risk = Standard Deviation in Returns

A representative year
Index • Standard deviation is 5.39%
01-Jan 7.32% Average 1.32%
01-Feb -2.47% • Also known as monthly volatility
01-Mar 2.26% Standard
(normally quoted as a percentage)
01-Apr 5.18% Deviation 5.39% • Means (roughly) that value of
01-May 9.69% portfolio can readily move up or
01-Jun -0.69% down 5.39% in any month
01-Jul -0.32%
• For a portfolio valued $20m, 5.39%
01-Aug -9.03%
is $1.078m.
01-Sep -4.89%
01-Oct -0.41% • So “typical” monthly return is 1.32%
01-Nov 6.44% ± 5.39%
01-Dec 2.72%

london.edu Lecture 3: Risk and Return 3

 Risk and Return: The value of £1 invested, 1900-2021
100,000.0

39,398
Equities 9.1 % per year
10,000.0
Bonds 5.6 % per year
Index value (start 1900 = 1.0; log scale) Bills 4.6 % per year
Inflation 3.6 % per year
1,000.0
719

234
100.0
69

10.0

1.0
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

0.1
Source: E Dimson, P Marsh and M Staunton Triumph of the Optimists (Princeton University Press, 2002) and subsequent updates

london.edu Lecture 3: Risk and Return 4

 Risk and Return: Rates of Return in UK in 1900-2021
150%

125%
Equities Bonds

100%

75%

50%

25%

0%

-25%

-50%
1900 1905 1910 1915 1920 1925 1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020
Source: E Dimson, P Marsh and M Staunton Triumph of the Optimists (Princeton University Press, 2002) and subsequent updates

london.edu Lecture 3: Risk and Return 5

 Average Returns, US 1900-2017

Mean Standard
Series Return Risk Premium Deviation
Stocks 11.5 7.7 19.7
Bonds 5.3 1.5 8.3
Bills 3.8 0.0 2.8
Inflation 3.0
Source: Brealey-Myers-Allen

1 2017
• The mean return is calculated as: RA = ∑
118 i =1900
Ri

• The risk premium in the table is computed as the difference between the
mean of asset returns and the Bills.
london.edu Lecture 3: Risk and Return 6
 Returns and Variability
Variability
45

40 Small Co’s
35

30

25

20 S&P 500
15

10
Govt LT
Corporate
5
TBills Govt IT
0
0 5 10 15 20 Return

• For well diversified portfolios, the relation between risk and return is very close
to linear.
• Does this empirical observation hold also for individual stocks?
london.edu Lecture 3: Risk and Return 7
 Individual Shares versus the Stock Market: A Paradox?

Investment Risk Premium Variability (σ)

Stock Market 7% 20%
Typical Individual Share 7% 30-40%

• The risk premium for a typical individual share is not closely related to its
volatility.
• Is this a paradox? No. Much of the risk of an individual stock can be diversified
away.

london.edu Lecture 3: Risk and Return 8

 Example: Why Buy Gold?

Asset Return Variability

Gold 8.8% 20.8
S&P 500 12.8% 18.3

Coefficient of correlation between S&P 500 and Gold = -0.4

• Gold has a lower return and a higher variability than the S&P 500.
• Is Gold a bad investment?

london.edu Lecture 3: Risk and Return 9

 Portfolio Theory: A Two-Asset Portfolio

• Investment strategy: hold a portfolio of gold and stocks: a fraction of xGold% in

gold, (100- xGold)% in the S&P500.

xGold 100% 80% 60% 45.4% 40% 20% 0%

Return 8.8% 9.6% 10.4% 10.6% 11.2% 12.0% 12.8%
Variability 20.8% 15.5% 11.7% 10.7% 10.8% 13.5% 18.3%
Hold only Minimum Hold
gold Variance only
Portfolio S&P500

london.edu Lecture 3: Risk and Return 10

 Portfolio Theory: A Two-Asset Portfolio

• Conclusion: even if gold is very risky when you hold it alone, as part of a
portfolio it reduces significantly the total portfolio risk since it works as an
insurance against some forms of risk (inflation risk, ... etc.). Technically
speaking, the low correlation coefficient with the S&P 500 makes gold a
reasonably good hedging instrument.
• Let us take a step back and derive the formulas to obtain the data in the
previous table.

london.edu Lecture 3: Risk and Return 11

 Portfolio Theory: A Two-Asset Portfolio

• Two assets: Suppose you invest a proportion xGold of your wealth in gold and
the rest (xSP500 = 1- xGold) in the S&P500. The return on your portfolio:

𝑅𝑅𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = 𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 𝑅𝑅𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 + (1 − 𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 )𝑅𝑅𝑆𝑆𝑆𝑆𝑆𝑆𝑆

• The expected return on your portfolio is:

𝐸𝐸(𝑅𝑅𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 ) = 𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 𝐸𝐸(𝑅𝑅𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 ) + (1 − 𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 )𝐸𝐸(𝑅𝑅𝑆𝑆𝑆𝑆𝑆𝑆𝑆 )

• It is a weighted average of the two expected returns

london.edu Lecture 3: Risk and Return 12

 Portfolio Theory: A Two-Asset Portfolio

• The variance of the portfolio return is:

2
𝑉𝑉𝑉𝑉𝑉𝑉 𝑅𝑅𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = 𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 ) + (1 − 𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 )2 𝑉𝑉𝑉𝑉𝑉𝑉 𝑅𝑅𝑆𝑆𝑆𝑆𝑆𝑆𝑆

+2𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 (1 − 𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 )𝐶𝐶𝐶𝐶𝐶𝐶(𝑅𝑅𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 , 𝑅𝑅𝑆𝑆𝑆𝑆𝑆𝑆𝑆 )

• or in short-hand notation:
2
𝜎𝜎𝑅𝑅2𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = 𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 𝜎𝜎𝑅𝑅2𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 + (1 − 𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 )2 𝜎𝜎𝑅𝑅2𝑆𝑆𝑆𝑆𝑆𝑆𝑆 + 2𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 (1 − 𝑥𝑥𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 )𝜎𝜎𝑅𝑅𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 ,𝑅𝑅𝑆𝑆𝑆𝑆𝑆𝑆𝑆

• with 𝜎𝜎𝑅𝑅𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 ,𝑅𝑅𝑆𝑆𝑆𝑆𝑆𝑆𝑆 = 𝜌𝜌𝑅𝑅𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 ,𝑅𝑅𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝜎𝜎𝑅𝑅𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 𝜎𝜎𝑅𝑅𝑆𝑆𝑆𝑆𝑆𝑆𝑆

london.edu Lecture 3: Risk and Return 13

 Portfolio Theory: A Two-Asset Portfolio

The relation between expected return and risk and x, the proportion of the
portfolio invested in gold (data from the previous table).
Return is Linear. Risk is Not!
25%

20%
Risk and Return

15%

10%

5%

0%
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
Portfolio Composition (Proportion invested in Gold)

Return Risk

london.edu Lecture 3: Risk and Return 14

 Portfolio Theory: A Two-Asset Portfolio

• The expected return of a portfolio varies linearly with the portfolio weights.
However the variance of the portfolio return does not vary linearly with the
portfolio weights. This is why diversification reduces risk. The crucial
parameter is the correlation coefficient ρ.
• We usually plot the previous picture in the mean-standard deviation space.

london.edu Lecture 3: Risk and Return 15

 Portfolio Theory: Mean-Variance Analysis

Portfolio
Minimum
Frontier
variance
Expected portfolio
Return
0% Gold
12.8%

10.62%
8.8% 100% Gold

10.65% 18.3% 20.8% Standard

deviation
london.edu Lecture 3: Risk and Return 16
 Portfolio Theory: Mean-Variance Analysis

• The portfolio frontier (or efficient frontier) is the set of portfolios that
maximise expected return for a given level of standard deviation.
• Would you ever own Gold on its own? No. Gold only is strictly dominated by
the S&P500. Similarly all portfolios between 100% Gold and the minimum
variance portfolio are strictly dominated.

• Where is the magic? The crucial parameter that is driving this diversification
effect is the correlation ρ between the two stocks. Let us look at what happens
when ρ = +1 and ρ = -1.

london.edu Lecture 3: Risk and Return 17

 Portfolio Theory: Mean-Variance Analysis
Minimum
variance ρ = +1: No benefits from
portfolio diversification.
Expected
Return
0% Gold
12.8%

8.8% 100% Gold

18.3% 20.8% Standard

deviation
london.edu Lecture 3: Risk and Return 18
 Portfolio Theory: Mean-Variance Analysis

Minimum Efficient ρ = -1: Maximum benefits

variance Frontier from diversification
Expected portfolio
Return
0% Gold
12.8%

10.54%
8.8% 100% Gold

0% 18.3% 20.8% Standard

deviation
london.edu Lecture 3: Risk and Return 19
 Portfolio Theory: Mean-Variance Analysis

ρ = -1
Expected
Return
0% Gold
12.8% ρ = +1
ρ = -0.4

8.8% 100% Gold

18.3% 20.8% Standard

deviation
london.edu Lecture 3: Risk and Return 20
 More than Two Risky Assets

• When there are more than two risky assets, there is more than one possible
portfolio for each level of expected return.
• We need to consider an optimisation problem that selects the lowest standard
deviation portfolio among all portfolios with the same level of expected return.
• This problem is described in Appendix 2.

london.edu Lecture 3: Risk and Return 21

 A Slide from Appendix 2: Deriving the Portfolio Frontier

• The optimization problem: Among all portfolios with a given expected return
(E), which is the portfolio with the minimum variance (or standard deviation)?
• Choose portfolio weights xn, n=1,…N, to minimize
𝑁𝑁

𝜎𝜎𝑅𝑅2𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = � 𝑥𝑥𝑛𝑛2 𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑛𝑛 ) + 2 � 𝑥𝑥𝑛𝑛 𝑥𝑥𝑚𝑚 𝐶𝐶𝐶𝐶𝐶𝐶(𝑅𝑅𝑛𝑛 , 𝑅𝑅𝑚𝑚 )

𝑛𝑛=1 𝑛𝑛<𝑚𝑚

𝑁𝑁

Subject to 𝐸𝐸(𝑅𝑅𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 ) = � 𝑥𝑥𝑛𝑛 𝐸𝐸(𝑅𝑅𝑛𝑛 ) = 𝐸𝐸

𝑁𝑁 𝑛𝑛=1

and � 𝑥𝑥𝑛𝑛 = 1
𝑛𝑛=1

london.edu Lecture 3: Risk and Return 22

 More than Two Risky Assets (cont.)

• The mean-variance optimization problem can be solved easily in Excel.

• The next slide plots a portfolio frontier with many risky assets. Each dot on the
scatter plot corresponds to an individual asset.

london.edu Lecture 3: Risk and Return 23

 More than Two Risky Assets

london.edu Lecture 3: Risk and Return 24

 Which Portfolio to Choose?

• We should always choose a portfolio on the Portfolio Frontier (assuming that

we care only about mean and variance).
• Depending on institutional rules and taste for risk any investor will choose one
particular Optimal portfolio on the Portfolio Frontier.

london.edu Lecture 3: Risk and Return 25

 Conclusions

• Measuring risk of an individual stock:

• When the stock is examined in isolation: standard deviation of stock returns
• When the stock is examined as part of a portfolio: standard deviation is
NOT a good measure of risk
• Diversification reduces risk:
• Specific Risk can be diversified
• These insights inspired a lot of research and lead to the CAPM.

london.edu Lecture 3: Risk and Return 26

 Appendix 1: What if We Add a Riskless Asset?

• Consider a portfolio of a risky asset (or a risky portfolio) and a risk-free asset.
The weight on the risky asset is x.

• The return of this portfolio is

RPortfolio = x Rrisky + (1 − x ) R f

• The expected return is

E ( RPortfolio ) = x E ( Rrisky ) + (1 − x ) R f

london.edu Lecture 3: Risk and Return 27

 What if We Add a Riskless Asset?

• The standard deviation of the portfolio is (from Slide 13)

σ 2
RPortfolio = x Risky σ
2 2
RRisky + (1 − xRisky ) σ 2 2
Rf + 2 xRisky (1 − xRisky )σ RRisky , R f

• In this case, the standard deviation is a weighted sum of the standard

deviations of the portfolio components. The portfolio frontier is then a straight
line.
σ 2
RPortfolio =x σ 2 2
RRisky + 0 + 0 or σ RPortfolio = x σ RRisky

london.edu Lecture 3: Risk and Return 28

 What if We Add a Riskless Asset?

Portfolio Frontier w ith a Riskless Asset

Portfolio
Frontier

Expected Return

Tangent Portfolio Line 1>Line 2

Portfolio 1

Line 2
Portfolio 2

Riskless Asset
Standard Deviation

The Portfolio frontier is a straight line straight line connecting the Risk-free rate
and the Tangent Portfolio.

london.edu Lecture 3: Risk and Return 29

 Which Portfolio to Choose?

• We should always choose a portfolio on the Portfolio Frontier (assuming that

we care only about mean and variance).
• The Portfolio Frontier is a straight line connecting the risk-free rate and the
Tangent portfolio.
• The straight line is called the Capital Market Line.
• Depending on institutional rules and taste for risk any investor will choose one
particular Optimal portfolio on the Capital Market Line.

london.edu Lecture 3: Risk and Return 30

 Appendix 2: Portfolio Frontier with More than Two Assets

• Let’s consider indices from the Percentage of World

largest seven stock markets. These
Country Market Capitalization
indices are constructed by Morgan
Stanley Capital International US 29.9
(MSCI). Japan 8.2
UK 6.8
China 5.4
France 4.4
Hong Kong 4.3
Canada 3.7
Total 62.7
Source: MSCI, August 2008

london.edu Lecture 3: Risk and Return 31

 Statistical Properties of Returns: 1/1970-1/2008

Canada China France Hong Kong Japan UK US

Mean Return(%) 13.2 6.5 14.4 23.7 13.2 14.0 11.4
Std. Dev.(%) 19.06 38.1 22.0 36.2 21.8 22.2 15.1

Correlations Canada China France HK Japan UK US

Canada 1
China 0.49 1
France 0.48 0.31 1
HK 0.38 0.65 0.31 1
Japan 0.33 0.17 0.40 0.31 1
UK 0.52 0.38 0.58 0.39 0.37 1
US 0.71 0.43 0.50 0.36 0.31 0.54 1

london.edu Lecture 3: Risk and Return 32

 Deriving the Portfolio Frontier

• The optimization problem: Among all portfolios with a given expected return
(E), which is the portfolio with the minimum variance (or standard deviation)?
• Choose portfolio weights xn, n=1,…N, to minimize
N
σ 2
RPortfolio = ∑ x n Var ( Rn ) + 2 ∑ xn xmCov ( Rn , Rm )
2

n =1 n< m
N
Subject to E ( RPortfolio ) = ∑ xn E ( Rn ) = E
n =1
N
and ∑x
n =1
n =1

london.edu Lecture 3: Risk and Return 33

london.edu Lecture 3: Risk and Return 34
 Computation in Excel

• Program the equations in Excel.

• Use Solver to perform the minimisation.

london.edu Lecture 3: Risk and Return 35

london.edu Lecture 3: Risk and Return 36
london.edu Lecture 3: Risk and Return 37
 Appendix 3: Expectation, Variance, and Standard Deviation

• We can define expectation, variance and st. dev. in a statistics sense, for a
data series X1, X2, … XN.
• A measure of the average value (mean) of the series is the sample average

𝑿𝑿𝟏𝟏 + ⋯ + 𝑿𝑿𝑵𝑵
�=
𝑿𝑿
𝑵𝑵

london.edu Lecture 3: Risk and Return 38

 Sample Variance and Standard Deviation

• Two measures of the dispersion of the series around the average value are
the sample variance

(𝑿𝑿𝟏𝟏 − � )𝟐𝟐 + ⋯ + (𝑿𝑿𝑵𝑵 − 𝑿𝑿

𝑿𝑿 � )𝟐𝟐
𝑺𝑺(𝑿𝑿)𝟐𝟐 =
𝑵𝑵 − 𝟏𝟏
• and the sample standard deviation

𝑺𝑺(𝑿𝑿) = 𝑺𝑺(𝑿𝑿)𝟐𝟐

• The sample standard deviation is in the same units as the data series. It
measures the “typical" distance of the series elements from the average value.

london.edu Lecture 3: Risk and Return 39

 Probability and Random Variables

• We can also define expectation, variance, and st. dev. In a probability sense,
for a random variable Z.

• Suppose that Z takes the values Z1, .., ZK, with probabilities p1, .., pK.

• The expectation of Z is

𝐸𝐸 𝑍𝑍 = 𝑝𝑝1 𝑍𝑍1 + ⋯ + 𝑝𝑝𝐾𝐾 𝑍𝑍𝐾𝐾

london.edu Lecture 3: Risk and Return 40

 Variance and Standard Deviation

• The variance of Z is

𝑉𝑉 𝑍𝑍 = 𝑝𝑝1 (𝑍𝑍1 − 𝐸𝐸 𝑍𝑍 )2 + 𝑝𝑝2 (𝑍𝑍2 − 𝐸𝐸 𝑍𝑍 )2 + ⋯ + 𝑝𝑝𝐾𝐾 (𝑍𝑍𝐾𝐾 − 𝐸𝐸 𝑍𝑍 )2

and the standard deviation of Z is

𝜎𝜎 𝑍𝑍 = 𝑉𝑉(𝑍𝑍)

london.edu Lecture 3: Risk and Return 41

 Example

• Consider the random variable Z describing the outcomes of rolling a dice. Z

can take the values 1, 2, 3, 4, 5, 6, each with probability 1/6.
• The expectation of Z is
1 1 1 1 1 1
𝐸𝐸 𝑍𝑍 = × 1 + × 2 + × 3 + × 4 + × 5 + × 6 = 3.5
6 6 6 6 6 6

• The variance of Z is
1 2 1 2 1 2
𝑉𝑉 𝑍𝑍 = × (1 − 3.5) + × (2 − 3.5) + ⋯ + × 6 − 3.5 = 2.9167
6 6 6

• The standard deviation of Z is

𝜎𝜎 𝑍𝑍 = 2.9167 = 1.7078

london.edu Lecture 3: Risk and Return 42

 Statistics vs Probability Definitions

• The statistics and probability definitions of expectation, variance, and standard

deviation, are related.
• Suppose that we roll a dice a number of times, and record the data series
formed by the outcomes.
• For that data series, we can compute a sample average, sample variance and
sample standard deviation.
• These will generally be different than the expectation (3.5), variance (2.9167)
and standard deviation (1.7078) of Z.
• However, if we roll the dice very many times, they will get very close. If we roll
the dice an “infinite” number of times, they will become equal.
• General result (Law of Large Numbers): Statistics concepts converge to
their probability counterparts as the sample grows.

london.edu Lecture 3: Risk and Return 43