Risk and return – Stocks vs.

Bonds
&PSoo (3STMEREAnzeation)
Investing stock index Ex S

versionidity/Est/morison/
- .

securities in SIZE How BIG IS us

,
a basket Size of US =
soT$
EARM = 90T$
how big the
SIZE Of
market capitalizzation is

Edustries
=

company (shell
esity

(2) Asset allocation

60% -Stocks-Real Estate
40% -bands - Alternatives

(3) Security selection

which
stocks/bonds
-

(9) Execution/Nike
(s) Performance dualysis

1
Investment management process

2
Returns
Holding-period return Real vs. nominal returns
𝑃𝑃𝑡𝑡+1 + 𝐷𝐷𝑡𝑡+1 1 + 𝑅𝑅𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑟𝑟𝑟𝑟
𝑅𝑅𝑡𝑡+1 = −1 1 + 𝑅𝑅𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 =
𝑃𝑃𝑡𝑡 1 + 𝐼𝐼𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛
𝐷𝐷𝑡𝑡+1 𝑃𝑃𝑡𝑡+1 − 𝑃𝑃𝑡𝑡
= +
𝑃𝑃𝑡𝑡 𝑃𝑃𝑡𝑡 𝑅𝑅𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 ≈ 𝑅𝑅𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑟𝑟𝑟𝑟 − 𝐼𝐼𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛
= income yield + capital gain/loss

where
𝑃𝑃𝑡𝑡 : price today
𝑃𝑃𝑡𝑡+1 : price tomorrow
𝐷𝐷𝑡𝑡+1 : dividend tomorrow

Stocks and Bonds 3

Risk
Summarize risk through standard deviation, 𝜎𝜎, as a measure
of dispersion

If 𝑅𝑅1 , 𝑅𝑅2 , … , 𝑅𝑅𝑇𝑇 are returns

𝑇𝑇
1 2
𝜎𝜎 = � 𝑅𝑅𝑡𝑡 − 𝑅𝑅�
𝑇𝑇
𝑡𝑡=1

Stocks and Bonds 4

S&P 500 over 1926 to 2023
Mean = 11.9%, StDev = 19.3%
Standard deviation gives an idea of the range of possibilities
There is roughly 2/3rd chance that the return will be in the range
[Mean−StDev Mean+StDev]=[−7.4% 31.3%]

5
U.S. historical record
Over 1926-2023
Corporate Government
Stocks Bonds Bonds T-Bills Inflation
Mean 11.9% 6.2% 5.8% 3.4% 3.0%
StDev 19.3% 8.8% 9.8% 3.1% 3.9%

Stocks and Bonds 6

Risk and return

Stocks and Bonds 7

Stocks or bonds – Nominal growth

Note the log scale

Stocks and Bonds 8

Stocks or bonds – Real growth

Note the log scale

Stocks and Bonds 9

Risk (1) – Value of $1 in one year
If history repeats itself

Stocks and Bonds 10

Risk (2) – Chances of losing money
Volatility gives you an idea of what is the chance that you will
lose next year

 If (stock) return has mean of 10% and volatility of 20%, there is a

31% chance that your return next year will be negative
 If (bond) return has mean of 6% and volatility of 10%, there is a
27% chance that you will be in the red next year
 If (T-bill) return has mean of 3% and volatility of 3%, there is a 16%
chance that you will be in the red next year

Stocks and Bonds 11

Risk (2) – Long vs. recent history
1926-2023 2000-2023

Stocks and Bonds 12

Risk (4) – What is it?
Aggregate stock market

Stocks and Bonds 13

U.S. versus World
U.S. World

Source: Credit Suisse Global Investment Returns Sourcebook 2024

Stocks and Bonds 14

U.S. versus Switzerland
Future
26 sep 2024
Stocks Vs Bonds"
Rs) vs (R)
expectern
-same as
history 09%
/
U.S. Switzerland bard
centmerlin e
-based Current

Source: Credit Suisse Global Investment Returns Sourcebook 2024

Stocks and Bonds 15

Estimating expected returns
1. Historical approach

2. Survey approach

3. Implied approach

Stocks and Bonds 16

Historical approach
Limitations
 Assume that the risk aversion of investors / riskiness of
stocks has not changed in a systematic way across time

Stocks and Bonds 17

Survey approach
Limitations
 No constraints on reasonability (the survey could produce
negative risk premiums or risk premiums of 50%)
 Forecasts are extremely volatile
 Forecasts tend to be short term; even the longest surveys
do not go beyond one year

Stocks and Bonds 18

Implied approach
Gordon growth formula
Field intere
Dividered

P = D / (R – G)
E(R) = D/P + E(G)
D/P ≈ 2%
E(G) ??
expected growth

Stocks and Bonds 19

Stock (over)valuation

Stocks and Bonds 20

P/E ratios and stock returns

Stocks and Bonds 21

Where does one go from here?
Economic analysis of the country
• Identify business cycles

Stocks and Bonds 22

Choice between risky and risk-free
Split investment funds between safe and risky assets
• Risk free asset: T-bills
• Risky asset: stock index

Risky and Riskfree 23

Portfolio statistics
Individual assets
• Portfolio weights: 𝑤𝑤 [in risky] and 1 − 𝑤𝑤 [in risk-free]
• Expected (mean) returns: 𝜇𝜇 and 𝑟𝑟𝑓𝑓
• Variances of returns: 𝜎𝜎 2 and 0

Portfolio mean return = 𝑤𝑤 × 𝜇𝜇 + 1 − 𝑤𝑤 × 𝑟𝑟𝑓𝑓

Portfolio variance = 𝑤𝑤 2 × 𝜎𝜎 2
• Standard deviation = 𝑤𝑤 × 𝜎𝜎

Risky and Riskfree 24

Example
Risk-free return of 7%
Stock with mean 15% and volatility 22%

Borrow at the risk-free rate and invest in stock using 50%

leverage
• Portfolio return = −0.5×7% + 1.5×15% = 19%
• Portfolio standard deviation = 1.5×22% = 33%

Risky and Riskfree 25

Example …

Risky and Riskfree 26

Example …

Risky and Riskfree 27

Example …
𝑈𝑈 = 𝜇𝜇 − 0.5𝐴𝐴𝜎𝜎 2

28
Optimal allocation
Optimal allocation to stock
𝑤𝑤𝑆𝑆 = 𝜇𝜇𝑆𝑆 − 𝑟𝑟𝑓𝑓 � 𝐴𝐴𝜎𝜎𝑆𝑆2
= 8%/(4 × 22%2 )
= 41.3%

Optimal portfolio statistics

𝜇𝜇𝑃𝑃 = 10.31%
𝜎𝜎𝑃𝑃 = 9.09%

Optimal utility = 0.08653

Risky and Riskfree 29

Two risky assets
Portfolio weights
• 𝑤𝑤1 and 𝑤𝑤2 [Note that 𝑤𝑤1 + 𝑤𝑤2 = 1.0]

Expected (mean) returns

• 𝜇𝜇1 and 𝜇𝜇2

Variances of returns
• 𝜎𝜎12 and 𝜎𝜎22

Covariance of returns
• 𝜎𝜎12 = 𝜌𝜌12 𝜎𝜎1 𝜎𝜎2
Two risky 30
Portfolio statistics
Portfolio mean
𝑤𝑤1 𝜇𝜇1 + 𝑤𝑤2 𝜇𝜇2

• Weighted sum of individual asset returns

Portfolio variance
𝑤𝑤12 𝜎𝜎12 + 𝑤𝑤22 𝜎𝜎22 + 2𝑤𝑤1 𝑤𝑤2 𝜎𝜎12

• Not weighted sum of individual asset variances

• The third term involving covariance is very important

Two risky 31
Example
Two assets – bond and stock
• Means of 8% and 13%
• Standard deviations of 12% and 20%
• Correlation of 0.3

Equal-weighted portfolio
• Portfolio return = 0.5×8% + 0.5×13% =10.5%
• Portfolio variance = 0.52×(12%)2 + 0.52×(20%)2
+2×0.5×0.5×(0.3×12%×20%) = 0.0172
• Portfolio standard deviation = 0.0172 = 13.11%

Two risky 32
Example …

Allocation Statistics
Standard
Bond Stock Mean deviation
1 0% 100% 13.0% 20.00%
2 10% 90% 12.5% 18.40%
3 20% 80% 12.0% 16.88%
4 30% 70% 11.5% 15.47%
5 40% 60% 11.0% 14.20%
6 50% 50% 10.5% 13.11%
7 60% 40% 10.0% 12.26%
8 70% 30% 9.5% 11.70%
9 80% 20% 9.0% 11.45%
10 90% 10% 8.5% 11.56%
11 100% 0% 8.0% 12.00%

Two risky 33
Example: Portfolio mean

Two risky 34
Example: Portfolio volatility

Two risky 35
Example: Risk-return tradeoff

Two risky 36
Example: MVP
Minimum variance portfolio: the portfolio composed of risky
assets with smallest standard deviation

𝑛𝑛𝑚𝑚𝑚𝑚 𝜎𝜎22 − 𝜎𝜎12

𝑤𝑤1 = 2
𝜎𝜎1 + 𝜎𝜎22 − 2𝜎𝜎12

𝑤𝑤𝐵𝐵𝑛𝑛𝑛𝑛𝐵𝐵 = 82%, 𝑤𝑤𝑆𝑆𝑡𝑡𝑛𝑛𝑆𝑆𝑆𝑆 = 18%

𝜇𝜇𝑛𝑛𝑚𝑚𝑚𝑚 = 8.9%
𝜎𝜎𝑛𝑛𝑚𝑚𝑚𝑚 = 11.45%

Two risky 37
Example: MVP …
With correlation = +1
𝑛𝑛𝑚𝑚𝑚𝑚 𝜎𝜎2 𝑛𝑛𝑚𝑚𝑚𝑚 −𝜎𝜎1
𝑤𝑤1 = , 𝑤𝑤2 =
𝜎𝜎2 − 𝜎𝜎1 𝜎𝜎2 − 𝜎𝜎1

With correlation = −1
𝑛𝑛𝑚𝑚𝑚𝑚 𝜎𝜎2 𝑛𝑛𝑚𝑚𝑚𝑚 𝜎𝜎1
𝑤𝑤1 = , 𝑤𝑤2 =
𝜎𝜎2 + 𝜎𝜎1 𝜎𝜎2 + 𝜎𝜎1

With correlation = 0
𝑛𝑛𝑚𝑚𝑚𝑚 𝜎𝜎22 𝑛𝑛𝑚𝑚𝑚𝑚 𝜎𝜎12
𝑤𝑤1 = 2 2, 𝑤𝑤2 = 2
𝜎𝜎2 + 𝜎𝜎1 𝜎𝜎2 + 𝜎𝜎12

38
Example: Which portfolio to choose?
𝜇𝜇𝐴𝐴 = 8.9%
𝜎𝜎𝐴𝐴 = 11.45%
(82% bond, 18% stock)

𝜇𝜇𝐵𝐵 = 9.5%
𝜎𝜎𝐵𝐵 = 11.70%
(70% bond, 30% stock)

Two risky 39
Example: Optimal portfolio
Using the utility function 𝑈𝑈 = 𝜇𝜇 − 12𝐴𝐴𝜎𝜎 2

𝑛𝑛𝑚𝑚𝑡𝑡𝑛𝑛𝑛𝑛𝑟𝑟𝑟𝑟 𝜎𝜎22 − 𝜎𝜎12 + 𝜇𝜇1 − 𝜇𝜇2 ⁄𝐴𝐴

𝑤𝑤1 =
𝜎𝜎12 + 𝜎𝜎22 − 2𝜎𝜎12

𝑤𝑤𝐵𝐵𝑛𝑛𝑛𝑛𝐵𝐵 = 51%, 𝑤𝑤𝑆𝑆𝑡𝑡𝑛𝑛𝑆𝑆𝑆𝑆 = 49%

𝜇𝜇𝑛𝑛𝑚𝑚𝑡𝑡𝑛𝑛𝑛𝑛𝑟𝑟𝑟𝑟 = 10.5%
𝜎𝜎𝑛𝑛𝑚𝑚𝑡𝑡𝑛𝑛𝑛𝑛𝑟𝑟𝑟𝑟 = 13.04%
for 𝐴𝐴 = 4

40
Introduce risk-free asset
Maximize the slope of the CAL for any possible portfolio, P

The objective function is the slope (Sharpe ratio):

𝜇𝜇𝑃𝑃 − 𝑟𝑟𝑓𝑓
𝑆𝑆𝑃𝑃 =
𝜎𝜎𝑃𝑃

Two risky 41
Example …
𝜇𝜇𝐴𝐴 = 8.9%
𝜎𝜎𝐴𝐴 = 11.45%
8.9% − 5%
𝑆𝑆𝐴𝐴 = = 0.34
11.45%

𝜇𝜇𝐵𝐵 = 9.5%
𝜎𝜎𝐵𝐵 = 11.70%
9.5% − 5%
𝑆𝑆𝐵𝐵 = = 0.38
11.70%

Two risky 42
Example …

Two risky 43
Example …
𝜇𝜇1𝑟𝑟 𝜎𝜎22 − 𝜇𝜇2𝑟𝑟 𝜎𝜎12
𝑤𝑤1𝑃𝑃 =
𝜇𝜇1𝑟𝑟 𝜎𝜎22 + 𝜇𝜇2𝑟𝑟 𝜎𝜎12 − 𝜇𝜇1𝑟𝑟 + 𝜇𝜇2𝑟𝑟 𝜎𝜎12

8 − 5 × 400 − 13 − 5 × 72
𝑤𝑤1𝑃𝑃 = = 0.40
3 × 400 + 8 × 144 − 3 + 8 × 72
𝑤𝑤2𝑃𝑃 = 0.60

𝜇𝜇𝑃𝑃 = 0.4 × 8% + 0.6 × 13% = 11%

𝜎𝜎𝑃𝑃2 = 0.42 × 144 + 0.62 × 400 + 2 × 0.4 × 0.6 × 72 = 201.6
𝜎𝜎𝑃𝑃 = 201.6 = 14.2%
11% − 5%
𝑆𝑆𝑃𝑃 = = 0.42
14.2%
Two risky 44
Example …
Optimal allocation to P
𝐴𝐴 = 4, 𝑤𝑤𝑃𝑃 = 𝜇𝜇𝑃𝑃 − 𝑟𝑟𝑓𝑓 � 𝐴𝐴𝜎𝜎𝑃𝑃2 = 0.7439

Two risky 45
Example …
𝜇𝜇𝑛𝑛𝑚𝑚𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = 𝑤𝑤𝑃𝑃 𝜇𝜇𝑃𝑃 + 1 − 𝑤𝑤𝑃𝑃 𝑟𝑟𝑓𝑓 = 9.46%

𝜎𝜎𝑛𝑛𝑚𝑚𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = 𝑤𝑤𝑃𝑃 𝜎𝜎𝑃𝑃 = 10.56%

9.46% − 5%
𝑆𝑆𝑛𝑛𝑚𝑚𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = = 0.42 = 𝑆𝑆𝑃𝑃
10.56%

CASH

Two risky 46
Example …
𝜇𝜇𝑛𝑛𝑚𝑚𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = 𝑤𝑤𝑃𝑃 𝜇𝜇𝑃𝑃 + 1 − 𝑤𝑤𝑃𝑃 𝑟𝑟𝑓𝑓 = 9.46%

𝜎𝜎𝑛𝑛𝑚𝑚𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = 𝑤𝑤𝑃𝑃 𝜎𝜎𝑃𝑃 = 10.56%

9.46% − 5%
𝑆𝑆𝑛𝑛𝑚𝑚𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = = 0.42 = 𝑆𝑆𝑃𝑃
10.56%

Two risky 46
Mean-variance optimization

1
Portfolio statistics – many assets
Portfolio mean is the weighted average, 𝑤𝑤 ′ 𝜇𝜇

Portfolio variance is the sum of (weighted) variances AND

covariances, 𝑤𝑤 ′ Σw

Example for three assets

𝑤𝑤𝑥𝑥 𝜇𝜇𝑥𝑥 + 𝑤𝑤𝑦𝑦 𝜇𝜇𝑦𝑦 + 𝑤𝑤𝑧𝑧 𝜇𝜇𝑧𝑧
𝑤𝑤𝑥𝑥2 𝜎𝜎𝑥𝑥2 + 𝑤𝑤𝑦𝑦2 𝜎𝜎𝑦𝑦2 + 𝑤𝑤𝑧𝑧2 𝜎𝜎𝑧𝑧2 + 2𝑤𝑤𝑥𝑥 𝑤𝑤𝑦𝑦 𝜎𝜎𝑥𝑥𝑦𝑦 + 2𝑤𝑤𝑦𝑦 𝑤𝑤𝑧𝑧 𝜎𝜎𝑦𝑦𝑧𝑧 + 2𝑤𝑤𝑧𝑧 𝑤𝑤𝑥𝑥 𝜎𝜎𝑧𝑧𝑥𝑥

Geometry 2
Many risky securities
Many more combinations of portfolios are possible

If you plot all the portfolio combinations, they will no longer

plot in a line. Rather, they will plot as an area

1.8%

1.6%

1.4%

1.2%

Mean 1.0%

0.8%

0.6%

0.4%

0.2%

0.0%
0% 5% 10% 15% 20% 25% 30% 35%
StdDev
Geometry 3
Minimum variance frontier

Geometry 4
Extending to include risk-free asset

Geometry 5
Extending to include risk-free asset …
The optimal combination becomes linear

A single combination of risky and risk-free assets will

dominate

Note confusing terminology

• With only risky assets, the efficient frontier is a parabola
• With risky and riskfree asset, the efficient frontier is a straight line

Geometry 6
Efficient frontier reprise

Geometry 7
Global minimum variance portfolio
Has the lowest possible variance among all combinations of
stocks
E(r)

G
Global minimum
variance portfolio
σ(r)

Geometry 8
Tangency portfolio
Maximizes the ratio of expected return to standard deviation

Tangency
portfolio
E(r)
T

σ(r)

Geometry 9
Two-fund separation
Combinations of the risk-free asset and the tangent portfolio
provide the best risk and return tradeoff available to an
investor

This means that the tangent portfolio is efficient and that all
efficient portfolios are combinations of the risk-free
investment and the tangent portfolio. Every investor should
invest in the tangent portfolio independent of his or her taste
for risk

Geometry 10
Two-fund separation …
An investor’s preferences will determine only how much to
invest in the tangent portfolio versus the risk-free investment
• Conservative investors will invest a small amount in the tangent
portfolio
• Aggressive investors will invest more in the tangent portfolio
• Both types of investors will choose to hold the same portfolio of
risky assets, the tangent portfolio, which is the efficient portfolio

Geometry 11
Diversification

Geometry 12
Diversification …
Equally-weighted portfolio
• A portfolio in which the same amount is invested in each stock

Variance of an equally-weighted portfolio of 𝑁𝑁 stocks

var 𝑅𝑅𝑝𝑝
1
= Average variance of individual stocks
𝑁𝑁
1
+ 1− Average covariance between stocks
𝑁𝑁

Geometry 13
Diversification …

Geometry 14
Limits of diversification
SOLVEn (DATA)

~
N

Volatility
(I
=

for
any portfolio P

Mp = w M, +
wip+... Warn Wh
,
=

Sp = w. Ew
Unique (Idiosyncratic) maxf(x)
Risk 6

I
%WI
min Wst g(x) =S
S =
f(x) + 2(f(x) 3)
max -

som (rights) = 200

Illagrangian) wity = =...

n+ 2 = variance

Market (Systematic) Risk

Number of
securities

Geometry 15
Portfolio math

2equationsse
n+

Math 16
Portfolio math … 6p2BM
/A +

Mp

Chyperbol
6p2
RISK FREE ASSETS

6p (Mp-Rg)Asome numbers
=

= Ep =
(Mp-Rf) a some numbers

Mp

68

Math 17
Portfolio math …

Math 18
Global MVP portfolio

Math 19
Two-fund separation

Math 20
With a riskless asset

Math 21
Tangency portfolio

Math 22
Expected returns

Math 23
Covariance
Relevant measure of risk is the covariance with the tangency
portfolio
• Why is risk covariance?
• Because it is the marginal variance or risk

Intuition
• In economics, it is the marginal cost of goods that determines their
prices, not their total or average cost
• Likewise, the marginal variance or covariance determines the
additional risk of an investment, and therefore its price (here,
expressed as returns not dollars)

Math 24
Partner’s Healthcare (Exercise to solve

Healthcare network of hospitals

• Differing needs in terms of size of their endowment assets,
operating budgets etc.
• Different allocations to various pool of assets

Two pools
• STP: Safe pool. Can be thought of as risk-free asset
▪ Average yield in Spring 2005 was 3.2%
• LTP: Risky asset pool
▪ Managed by external money managers

Excel 25
Asset allocation problem
Baseline asset mix of LTP
• Domestic Equity 55%
• International Equity 30%
• Long-Term Bonds 15%
Possible addition of real assets
• REITs (Real Estate)
• Commodities

What should be the composition of LTP?

How should the various member hospitals decide their

allocation decisions?
Excel 26
Data

&
Correlations
Exp Return StdDev US Equity Foreign EquiBonds REITs Commodities
US Equity 12.94% 15.21% 1.00 0.62 0.25 0.56 -0.02 Choose it between Real Estate
Foreign Equity 12.42% 14.44% 0.62 1.00 0.06 0.40 0.01 or commodities
Bonds 5.40% 11.10% 0.25 0.06 1.00 0.16 -0.07 for the correlations
REITs 9.44% 13.54% 0.56 0.40 0.16 1.00 -0.01 know which one
Idort
Commodities 10.05% 18.43% -0.02 0.01 -0.07 -0.01 1.00

Covariances
Exp Return StdDev US Equity Foreign EquiBonds REITs Commodities
US Equity 12.94% 15.21% 0.023134 0.013617 0.004221 0.011533 -0.000561
Foreign Equity 12.42% 14.44% 0.013617 0.020851 0.000962 0.007821 0.000196
Bonds 5.40% 11.10% 0.004221 0.000962 0.012321 0.002405 -0.001432
REITs 9.44% 13.54% 0.011533 0.007821 0.002405 0.018333 -0.000250
Commodities 10.05% 18.43% -0.000561 0.000196 -0.001432 -0.000250 0.033966

STP 3.20%

Excel 27
Risk-return tradeoff
14%
US Equity
12% Foreign Equity

Torre
LTP
Close
10%
Commodities
REITs
8%
Mean

6%
Bonds
4%
STP

2%

0%
0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20%
StdDev
Excel 28
Real assets
REITs have low risk and reasonable expected returns

Commodities have much higher risk but only slightly higher

expected returns

Ultimately, it depends on the correlation of these asset

classes with LTP

Excel 29
Optimal portfolios for Partners
MAX SP
WEIGHTS = 100% (TO SOLVE IT With EXCEL)
Y min Variance p
*
wweigert)
· St 8%
meaup
=

weights ?O

?
Any portfolio
Np = W N , + We
i
Nz + ....
+
WgUs

& op wit=
.....
+ +2wwz6 ...
solve
COMMAND + EXCEL
=

min B10
SQRT(B10)

[
#add
B9 = 8%
OPTMAL
PORTEOLIO
mene
#add m of

lights 07
weights =

.
100%

6% state

-
7% us EQUITY

to solve

#add
9 56%
7 6j% weights
= o
.
.

#tomaximate 7 66 %
.

max
SidDer =
10%

11 . 34 % (MEAN)
(Nrp) min
Var,
e
Mean 8 33 %
= .

Excel st sum
(weights) = 100 %
StaDer 7 61 %= .

30
 Tangency portfolios for Partners

Tangency Portfolios
US Equity 41% 32% 24%
Foreign Equity 47% 43% 31%
---- Sharpe
-
Ratio Bonds 12% 11% 11%
·
Mean ---------

REITs 0% 14% 10%

i!
- SRp
Meau--Ry) o
Map Commodities 0% 0% 24%

m
=

Mean 11.77% 11.40% 10.92%

StdDev 11.98% 11.38% 9.45%
Sharpe Ratio 0.715 0.721 0.816
SDelt
SR
Mean--Rf
=
=

CONTINUE T
- -

SDeVi
#dod

sharpe ratio mas

Excel 31
Efficient frontier again

Excel 32
Optimal portfolios again
Different hospitals have different target return/risk levels of
LTP. How can one accommodate this? For e.g., what should a
hospital wishing to target 12% risk of LTP do?

A portfolios exists on the tangent line that is above and to

the right of the tangency portfolio
• Has return of 13% and has a 127% position in LTP financed with
borrowing / shorting in the STP
• Can one use leverage to fund portfolio investment and enhance
returns?

Excel 33
Summary of optimization steps
Suppose we have N assets. To do the mean variance
analysis, we need
• N expected returns (one for each asset)
• N variances (one for each asset)
• N(N−1)/2 covariances (for pairs of assets)

To find the portfolio frontier we need to:

• Find the portfolio with the highest expected return for each
(feasible) standard deviation
• Find the portfolio with the lowest standard deviation for each
expected return

Excel 34
Implementation issues
Sometimes recommendations of MV analysis seem
unreasonable
• Large short (or long) positions

Generate high turnover

• Significant changes in portfolio composition from one period to the
other
• Generate high transaction costs

Investments in illiquid securities

• Securities may possess desirable risk-return characteristics but may
not trade in sufficient volume
Excel 35
3) Two Risk (15 1B) +
,
1 Risk free 2)Two Risk (1 stak + 1 Bond) S)NRisky Assets + 1 Risk free (Real Estate problem)

. ·
efficient

~
indifference

da
-
ave Frontier

-
-

pin
... Si
- -
ot
-

.
... Sz

agp : Tendendo portato

,

Rf
-

B(-50 %
·
RB , 150 % 5)

(s
%R % 6%)
95
·+ ,

A
(50% it so % T)
,
Two "Fund"
=

(50 %Pr, 20% 5 20% B , (3 : 2 S/B) Separate

Rf
VRMy =

Np-1A6 ?
4) N
Risky assets

from
indifference
curve

EFFICIENT

entire black live

·
S4 mean-variance frontier

6
CAPM, CCAPM, APT

1
Capital Asset Pricing Model (CAPM)
It is the equilibrium model that underlies all modern financial
theory

Derived using principles of diversification with simplified

assumptions

Markowitz, Sharpe, Lintner, and Mossin are researchers

credited with its development

2
Assumptions

3
CAPM derivation
These assumptions guarantee that the mean-variance
efficient frontier is linear and is the same for every investor

By the two-fund separation theorem, all investors invest in

the same tangency portfolio 𝑇𝑇 and the risk-free asset (with
different weights).

In equilibrium, this tangency portfolio 𝑇𝑇 coincides with the

“market portfolio”

4
Efficient frontier – CAL and CML

5
Expected returns
CML shows that for all “efficient” portfolios 𝑝𝑝

𝐸𝐸 𝑅𝑅𝑀𝑀 − 𝑅𝑅𝑓𝑓
𝐸𝐸 𝑅𝑅𝑝𝑝 = 𝑅𝑅𝑓𝑓 + 𝜎𝜎𝑝𝑝
𝜎𝜎𝑀𝑀

But what about an arbitrary portfolio (or an asset) 𝑞𝑞 that is

not on CML?

6
Diversifiable vs. systematic risk
A (efficient) and B (not efficient) have the same expected
return but different volatility
• B contains more diversifiable risk than A, but has the same
systematic risk as A

Portfolio
expected
return A B
Diversifiable risk

Portfolio volatility

7
Systematic risk and return tradeoff
Systematic risk earns risk premium
• Holding more systematic risk improves expected returns

Diversifiable risk does not earn a risk premium

8
Expected returns …
Recall that we derived an equation for expected returns for
any asset using tangency portfolio

𝐸𝐸 𝑅𝑅𝑝𝑝 = 𝑅𝑅𝑓𝑓 + 𝛽𝛽𝑝𝑝,𝑇𝑇 𝐸𝐸 𝑅𝑅𝑇𝑇 − 𝑅𝑅𝑓𝑓

where 𝛽𝛽𝑝𝑝,𝑇𝑇 is the beta of portfolio 𝑝𝑝 with respect to the

tangency portfolio 𝑇𝑇

We have now identified the tangency portfolio 𝑇𝑇 as the

market portfolio 𝑀𝑀

9
CAPM formula

𝐸𝐸 𝑅𝑅 = 𝑅𝑅𝑓𝑓 + 𝛽𝛽 × 𝐸𝐸 𝑅𝑅𝑀𝑀 − 𝑅𝑅𝑓𝑓

(1) (3) (2)
CALCUCATONS Of
debt , Equity *

10
Beta again
Volatility refers to total risk, sum of
• Diversifiable/Idiosyncratic risk
• Non-diversifiable/Systematic risk

Systematic risk is related to the contribution to the optimal

diversified portfolio aka the market portfolio

Market beta is related to this systematic risk

Only beta matters

Nothing else matters

11
 Diversification again

E(rt) Rf + P[E(Rm)
= -

Ry)

N CML SML

·
Mo

i 6
repo
p 12
 Security market line

Q ⑧
①
A

·
M
E(m) <
RI CML

but

6 A) O

M
E(R)) Rf =

BA 20 ·

Gaso
Saro
Ba 0
=

Sar 0
=

Rg Ro
·
A

· A

B
ToALCUA Any *
(EX NESTLE)
.

13
 CAPM inputs
Same for all projects
 𝑅𝑅𝑓𝑓 : Risk-free return
 𝐸𝐸 𝑅𝑅𝑀𝑀 − 𝑅𝑅𝑓𝑓 : Market risk premium (aka equity premium)

Project-specific
 𝛽𝛽 : Beta with respect to market

CORRELATIONS
OiL COMPANIES HIGH B(BETA)
14
CAPM implications
Expected return depends only on beta
• Not on whether the project is in hi-tech sector or utility sector
• Not on whether project size is big or small

Expected return does not depend on volatility

There is a risk-return tradeoff for individual stocks as long as

you measure risk properly through market betas

15
CAPM implications …
Expected return on a risky project that has zero-beta is equal
to the risk-free return
• Because it is assumed that investors can and do diversify this
project’s risk away completely
• What if you, as an individual do not?

16
Market betas
What do the betas mean?

17
Market beta calculation
Can be calculated using regression

Collect historical data on returns

• Stock, 𝑅𝑅𝑖𝑖
• Risk-free, 𝑅𝑅𝑓𝑓
• Market, 𝑅𝑅𝑀𝑀

Dependent variable, 𝑌𝑌𝑡𝑡 = 𝑅𝑅𝑖𝑖𝑡𝑡 − 𝑅𝑅𝑓𝑓𝑡𝑡 *

It =
a
Independent variable, 𝑋𝑋𝑡𝑡 = 𝑅𝑅𝑀𝑀𝑡𝑡 − 𝑅𝑅𝑓𝑓𝑡𝑡 =
Beta It NEITHER OF

-S
STATISMALLY

corto
m =

Regress 𝑌𝑌𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽 𝑋𝑋𝑡𝑡 + 𝑒𝑒𝑡𝑡

Xt 18
Wal-Mart beta
Monthly returns from 2010 to 2014
Date S&P500 Rfree WMT Y X
Jan-10 -3.58% 0.01% -0.04% -0.04% -3.58%
Feb-10 3.04% 0.01% 1.20% 1.19% 3.03%
Mar-10 6.10% 0.01% 3.39% 3.38% 6.09%
Apr-10 1.60% 0.01% -3.53% -3.54% 1.58%
May-10 -8.01% 0.01% -5.18% -5.19% -8.02%
Jun-10 -5.35% 0.01% -4.92% -4.93% -5.36%
Jul-10 7.05% 0.01% 6.49% 6.48% 7.03%
Aug-10 -4.54% 0.01% -1.46% -1.47% -4.56%
Sep-10 9.04% 0.01% 6.74% 6.73% 9.03%
Oct-10 3.87% 0.01% 1.21% 1.20% 3.86%
Nov-10 -0.01% 0.01% -0.15% -0.16% -0.02%
Dec-10 6.71% 0.01% 0.26% 0.25% 6.69%
Jan-11 2.33% 0.01% 3.97% 3.96% 2.32%
…

19
Wal-Mart beta …
SUMMARY OUTPUT

egression Statistics
Multiple R 0.3936
R Square 0.1549
Adjusted 0.1404
Standard 0.0409
Observat 60

ANOVA
df SS MS F gnificance F
Regressio 1 0.0178 0.0178 10.633 0.0019
Residual 58 0.0969 0.0017
Total 59 0.1146

Coefficients
andard Err t Stat P-value ower 95%
Upper 95%
ower 95.0%
pper 95.0%
Intercept 0.005 0.0056 0.904 0.3697 -0.006 0.0162 -0.006 0.0162
X Variabl 0.4612 0.1414 3.2608 0.0019 0.1781 0.7443 0.1781 0.7443

Beta is 0.46
• What does the intercept mean?
20
Wal-Mart beta …

21
Practical issues in using CAPM in

treusars
deeplyitisdiferent of

*
aggregate

Bills)
Which risk-free rate?

How to get beta?

E(Rp) =
Rg +
[E(Rm) Rf] Ap +
WE CAN
TO ESTIMATE WE CAN
Look To
wok to
BLOMBERG
BLOOMBERG
What is the equity premium?
estimation of debt

historical average

22
(1) Risk-free rate
CAPM has no concept of multiple periods, and therefore of
different risk-free rates based on horizon

In practice, choose a risk-free rate with similar duration to

your project
• For valuing stocks, a 10- or 20-year rate is often used
REASOMABLE

23
(2) Beta
Beta should be the beta of your project
• If the firm engages in a project different from its core competency,
do not use firm beta to evaluate the project

In M&A, use the beta of the target firm, not that of the
acquirer, to determine how much to pay

In international projects, calculate beta with respect to local

(foreign) market index
• Especially for emerging, segmented markets

24
Beta estimation tips
Get estimates for your company from websites such as
Bloomberg, Yahoo, or Google

Get estimates for comparables (adjusting for differences, e.g.

leverage) from websites

Estimate your own

Take averages
If estimate too far from 1.0, shrink it closer to 1.0

25
 Average betas
Betas add-up

If a firm has two divisions, A and B, with betas, 𝛽𝛽𝐴𝐴 and 𝛽𝛽𝐵𝐵 ,
and relative weights (values) of 𝑤𝑤𝐴𝐴 and 𝑤𝑤𝐵𝐵

𝛽𝛽Firm = 𝑤𝑤𝐴𝐴 𝛽𝛽𝐴𝐴 + 𝑤𝑤𝐵𝐵 𝛽𝛽𝐵𝐵

CAPM = there is one

Big systematic "Risk Factor" that happens

of to the market
OUTSIDE
POTENTIAL RISK FACTORS

S
pandemics e
unemploymente climate chance
*

Rate
Swan of Companyanflation/ interest
* BLACK

BLONLY Mis One Chance POLITICL FACTORS

26
 Consumption CAPM (CCAPM)
Higher Beta => Higher
Financial markets are inherently dynamic, not static E
(RE
• Assets are traded every day, not only once
• Investors’ decisions are made sequentially, not only once
• Investors take the future into account: they know that their
investment decisions today will affect tomorrow’s situation, etc.

Financial assets’ prices/returns are linked to the real side of

the economy
• In particular, there should be feedback loops between the assets’
prices and the investors’ consumption behavior

risker unhappy
=
higher B
=
higher marginal utility =
higher
Casset efected
is
risk
marginal utility) 27
CCAPM …

28
CCAPM …

29
CCAPM …

30
CCAPM …

31
CCAPM …

32
Arbitrage pricing theory (APT)
APT of Ross (1976) is a philosophically different theory in
that it does not rely on equilibrium but only on the absence
of arbitrage
• More powerful since it does not rely on assumptions about investor
preferences
• Less powerful since it sacrifices economic identification

33
Factor model
Returns are assumed to be generated from a factor model
with 𝐾𝐾 factors

𝑟𝑟𝑖𝑖𝑡𝑡 − 𝑟𝑟𝑓𝑓𝑡𝑡 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝐹𝐹𝑖𝑡𝑡 + 𝛽𝛽𝑖𝑖𝑖 𝐹𝐹𝑖𝑡𝑡 + ⋯ + 𝛽𝛽𝑖𝑖𝑖𝑖 𝐹𝐹𝑖𝑖𝑡𝑡 + 𝑒𝑒𝑖𝑖𝑡𝑡

Returns of assets consist of two components

• Systematic Risk: Small number of common factors that proxy for
economic events (changes in interest rates, inflation, and
productivity)
• Non-systematic Risk: Unique to each asset (new product
innovations, changes in management, lawsuits, labor strikes, etc.)

34
Factor model …
Factor model is just a statistical description of data

Describes the time-series variation in returns of a single

security 𝑖𝑖

The intercept (𝛼𝛼) in the model (time-series regression) is left

as a free parameter

35
Asset pricing model
Need theory to go from a factor model to an asset pricing
model

APT is one such theory. It says that if the time-series of

returns is described by

𝑟𝑟𝑖𝑖𝑡𝑡 − 𝑟𝑟𝑓𝑓𝑡𝑡 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝐹𝐹𝑖𝑡𝑡 + 𝛽𝛽𝑖𝑖𝑖 𝐹𝐹𝑖𝑡𝑡 + ⋯ + 𝛽𝛽𝑖𝑖𝑖𝑖 𝐹𝐹𝑖𝑖𝑡𝑡 + 𝑒𝑒𝑖𝑖𝑡𝑡

then the cross-section of average returns is described by

𝐸𝐸 𝑟𝑟𝑖𝑖 − 𝑟𝑟𝑓𝑓 = 𝛽𝛽𝑖𝑖𝑖 𝜆𝜆𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝜆𝜆𝑖 + ⋯ + 𝛽𝛽𝑖𝑖𝑖𝑖 𝜆𝜆𝑖𝑖

36
APT
Essentially, the absence of arbitrage is enough to ensure that
the intercept (alpha) in the factor model equation is zero
• Strictly speaking, Ross’ APT does not require the intercept to be
zero but only to be “small”
▪ Additional equilibrium conditions need to be imposed to get strict equality. In
this sense, to get a return equation, one still needs the absence of arbitrage
and equilibrium

37
APT …
Note that the key is the absence of alpha. Unlike the free
intercept in the time-series regression, the theory imposes
the constraint that the 𝛼𝛼 should be zero for all firms

𝜆𝜆 is called as the factor premium

• Need not be equal to 𝐸𝐸 𝐹𝐹 . For example, if 𝐹𝐹𝑡𝑡 = Inflation𝑡𝑡 , then
𝜆𝜆Inflation ≠ 𝐸𝐸(Inflation)
• However, if 𝐹𝐹 represents excess returns on a traded factor, then 𝜆𝜆 =
𝐸𝐸 𝐹𝐹 . For example, if 𝐹𝐹𝑡𝑡 = 𝑅𝑅𝑀𝑀,𝑡𝑡 − 𝑅𝑅𝑓𝑓,𝑡𝑡 , then 𝜆𝜆𝐹𝐹 = 𝐸𝐸(𝑅𝑅𝑀𝑀 − 𝑅𝑅𝑓𝑓 )

38
CAPM and market model
CAPM Market model
Says that Says that
E[r − rf ] = β (E[rM ] − rf ) rt − rft = α + β (rMt − rft ) + et
• This is a statement about • This is a statement about
expected returns returns every month

Implies that Implies that

rt − rft = β (rMt − rft ) + et E[r − rf ] = α + β (E[rM ] − rf )
• This is a statement about • This is a statement about
returns every month expected returns

Difference is in alpha 39
How to choose factors?
1. Factor analysis (purely statistical)
• Factor analysis constructs a limited set of abstract factors that best
replicate the estimated variances and covariances
• Throws no light on underlying economic determinants of the
covariances
2. Use of macroeconomic variables
• Business cycle risk
• Confidence risk
• Term premium risk
3. Use of firm specific attributes: size, B/M ratio

40
Example of an APT model
Chen, Roll, and Ross (1985) examine the following 5 macro
variables

Industrial Production (IP), risk premium on corporate bonds (CG), unanticipated

inflation (UI) appear to be factors that have significant explanatory power

41
Risk decomposition
Factor models can be used for risk decomposition

𝑟𝑟𝑖𝑖𝑡𝑡 − 𝑟𝑟𝑓𝑓𝑡𝑡 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽𝑖𝑖 𝑓𝑓𝑡𝑡 + 𝑒𝑒𝑖𝑖𝑡𝑡

𝜎𝜎𝑖𝑖𝑖 = 𝛽𝛽𝑖𝑖𝑖 𝜎𝜎𝑓𝑓𝑖 + 𝜎𝜎𝑒𝑒𝑖𝑖
𝑖

where
𝜎𝜎𝑖𝑖𝑖 = total variance
𝛽𝛽𝑖𝑖𝑖 𝜎𝜎𝑓𝑓𝑖 = systematic (factor related) variance
𝑖
𝜎𝜎𝑒𝑒𝑖𝑖 = unsystematic (firm-specific) variance

42
Example
𝑅𝑅𝑀𝑀𝑀𝑀𝐹𝐹𝑇𝑇,𝑡𝑡 − 𝑅𝑅𝑓𝑓,𝑡𝑡 = 0.01 + 0.93 𝑅𝑅𝑀𝑀,𝑡𝑡 − 𝑅𝑅𝑓𝑓,𝑡𝑡 + 𝑒𝑒𝑀𝑀𝑀𝑀𝐹𝐹𝑇𝑇,𝑡𝑡
𝜎𝜎𝑀𝑀 = 3.92%, 𝜎𝜎𝑒𝑒 = 5.53%
𝑅𝑅𝑖 (why?)
Systematic risk = 𝛽𝛽𝑖 𝜎𝜎𝑚𝑚
𝑖

= 0.932(3.92)2 = 13.20 (30%)

Unsystematic risk = 𝜎𝜎𝑒𝑒𝑖

= (5.53)2 = 30.62 (70%)

Total risk = 43.82

Standard deviation of MSFT = 6.62%
43
Risk decomposition with 𝐾𝐾 factors
If there are 𝐾𝐾 factors, represent the 𝐾𝐾 loadings by vector 𝐵𝐵𝑖𝑖
and the factors by vector 𝐹𝐹𝑡𝑡

𝑟𝑟𝑖𝑖𝑡𝑡 − 𝑟𝑟𝑓𝑓𝑡𝑡 = 𝛼𝛼𝑖𝑖 + 𝐵𝐵𝑖𝑖′ 𝐹𝐹𝑡𝑡 + 𝑒𝑒𝑖𝑖𝑡𝑡

𝜎𝜎𝑖𝑖𝑖 = 𝐵𝐵𝑖𝑖′ Σ𝑓𝑓 𝐵𝐵𝑖𝑖 + 𝜎𝜎𝑒𝑒𝑖𝑖

𝑖

where Σ𝑓𝑓 is the 𝐾𝐾 × 𝐾𝐾 covariance matrix of the factors

44
Factor models & mv-analysis
Assume 1,000 listed stocks

MV analysis requires 500K numbers

• 1,000 means
• 1,000 variances
• 499,000 covariances

With 1 factor, you need only 3,001 numbers

• 1,000 means
• 1,000 variances
• 1,000 betas
• 1 factor variance
45
Factor models and covariances
Consider two stocks
RA = α A + β A × F + e A
RB = α B + β B × F + eB

Then the covariance between the two stocks is

cov( A, B ) = β A × β B × var( F )

The variance of factors and the covariance between factors

are more robust, statistically speaking, than the variance of
individual assets and the covariance between assets

46
 MULTIFACTOR
=
(Rz) =
Ry + Bp E(F)
,
CAPM

BpE(E2) 2
-
+ INFLATION
FACTOR MODEL

·
-

+
BpzE(F3) APT
CARBITRAGE PRICING
THEORM)

tr =
Equity risk Premium =
E(R) -ne
E(rp) Rp = + B m.

E(RA) Rp + Bp
=
,
2
,
+
Bpx + Bpzt
Yes & NO
B =

(x) EXTENSION

GR
Bp =
of CAPM

Pin =
BPz =
con (BETERFACTORS MODEL)
 M SML

tuy
E(RA) =
theory says

E
Risk-adjusted Rf +
B(E(Rn) Re] -

⑧
shouldere
it everysimb meaning
------

Risk profiles e se

"
⑧

Ba B
in
theory < =
0

for all assets

edite
long o
or

desirable
buy
B
so

undesirable
shortore

STOCK 17/10/24
Market efficiency

1
What is an efficient market?
Efficient market is one where the market price is an unbiased
estimate of the true value of the investment

Market efficiency does not require that the market price be

equal to true value at every point in time. All it requires is
that errors in the market price be unbiased, i.e., that prices
can be greater than or less than true value, as long as these
deviations are random
• Randomness implies that there is an equal chance that stocks are
under or over valued at any point in time

2
Information and market efficiency
Prices are correct and fully reflect all available information
𝐷𝐷𝑡𝑡+1 𝐷𝐷𝑡𝑡+2
𝑃𝑃𝑡𝑡 = 𝐸𝐸𝑡𝑡 + 2
+⋯
1 + 𝑅𝑅 1 + 𝑅𝑅
Investors use all available information in forming
expectations about future cash flows

Prices react to new information quickly and to the right

extent

The discount rate is right for the riskiness of the cash flows

3
EMH and information release

Pr ice

Information release

4
Undereaction and Overreaction

Pr ice

Underreaction and
eventual price adjustment

Price
correction

Overreaction

Pr ice

Gradual overreaction

5
No abnormal returns
The only way you can get higher returns is by taking on
more risk

There is no information out there that can be used to

construct strategies that earn returns higher than required
for their risk

Market efficiency does not say that you cannot predict

returns. It says that you can’t predict abnormal returns.

6
Information

Type of Form of
information efficiency
Past prices Weak
Public Semi-strong
Private Strong

7
Weak and semi-strong form
Weak form Semi-strong form
 Current prices fully  Current prices fully
reflect all information in reflect all past prices
past prices and all publicly
available information
 Using past prices,  Fundamental analysis
returns, volumes will (using economic and
produce no predictable accounting information)
patters that can be will not produce profits
exploited to yield better
returns in the future

8
Filter rules

9
Technical analysis

10
Semi-strong form: Stock research

11
Strong (Private information)
 Current prices fully reflect all information, public and private
 Insider trading will not produce profits
• Knowing a merger is going to take place before it is announced
publicly will not produce profits
 Although illegal, evidence that prices move before public
announcements, suggesting insider information
 Insider trading appears profitable, indicating markets are
not strong form efficient
• These profits are short-lived, suggesting the market may be close
to efficient

12
Implications of EMH
No group of investors should be able to consistently beat
the market using a common investment strategy

In an efficient market, equity research and valuation would

be a costly task that provided no benefits. The odds of
finding an undervalued stock should be random (50/50). At
best, the benefits from information collection and equity
research would cover the costs of doing the research

13
Implications of EMH (contd.)
In an efficient market, a strategy of randomly diversifying
across stocks or indexing to the market, carrying little or no
information cost and minimal execution costs, would be
superior to any other strategy, that created larger
information and execution costs. There would be no value
added by portfolio managers and investment strategists

14
Implications of EMH (contd.)
In an efficient market, a strategy of minimizing trading, i.e.,
creating a portfolio and not trading unless cash was needed,
would be superior to a strategy that required frequent
trading.

15
What are NOT the implications
No investor will beat the market in any time period
• To the contrary, approximately half of all investors, prior to
transactions costs, should beat the market in any period

16
What are NOT the implications (contd..)
No group of investors will beat the market in the long term
• Given the number of investors in financial markets, the laws of
probability would suggest that a fairly large number are going to
beat the market consistently over long periods, not because of their
investment strategies but because they are lucky

17
How do markets become efficient
Markets do not become efficient automatically

Actions of investors, sensing bargains and putting into effect

schemes to beat the market, that make markets efficient

New information about securities comes to markets in a

random fashion

18
Necessary conditions for EMH
The market inefficiency should provide the basis for a
scheme to beat the market and earn excess returns
• The asset (or assets) which is the source of the inefficiency has to
be traded
• The transactions costs of executing the scheme have to be smaller
than the expected profits from the scheme

19
Necessary conditions for EMH (contd..)
There should be profit maximizing investors who
• recognize the potential for excess return
• can replicate the beat the market scheme that earns the excess
return
• have the resources to trade on the stock until the inefficiency
disappears

20
Contradiction
There is an internal contradiction in claiming that there is no
possibility of beating the market in an efficient market and
then requiring profit-maximizing investors to constantly seek
out ways of beating the market and thus making it efficient.
If markets were efficient, investors would stop looking for
inefficiencies, which would lead to markets becoming
inefficient again. It makes sense to think about an efficient
market as a self-correcting mechanism, where inefficiencies
appear at regular intervals

21
Tests of efficiency
Broad principle is to look for stock-picking strategies based
on some past information which have earned high returns
with little risk

Unfortunately, we can never be sure of in-efficiency

• It is always possible that we are not measuring risk properly
▪ We don’t know what the right discount rate is
• This is the “Joint Hypothesis Problem”

22
Evidence for
Stock prices appear to move randomly

New information appears to be quickly incorporated into

prices
• Event studies

−t 0 +t
Days relative to announcement date

Professional money managers do not beat the market on

average
23
Evidence against
Year 2008 (and 2009)

Inexplicable market crashes (October 1987)

Volume of trading is too high

Volatility is too high

Existence of many mutual funds

Market anomalies
24
Financial crisis of 2007–2009
The Economist: The efficient-markets hypothesis has underpinned
many of the financial industry’s models for years. After the crash,
what remains of it?

Robert Lucas: If an economist had a formula that could reliably

forecast crises a week in advances, say, then that formula would
become part of generally available information and prices would
fall a week earlier. (The term “efficient” as used here means that
individuals use information in their own private interest. It has
nothing to do with socially desirable pricing; people often confuse
the two.)

25
Financial crisis of 2007–2009 …
Richard Thaler: concedes that in some ways the events of the
past couple of years have strengthened the EMH. The hypothesis
has two parts, he says: the “no-free-lunch part and the price-is-
right part, and if anything the first part has been strengthened as
we have learned that some investment strategies are riskier than
they look and it really is difficult to beat the market.” The idea
that the market price is the right price, however, has been badly
dented.

Markus Brunnermier: Since funding frictions limit arbitrage

activity, the fact that you can’t make money does not imply that
the “price is right.”

26
Covid-19

Who knew?
27
But …

28
So …

29
PASSIVE ACTIVE
17 10 2024
-

e
-

① VALUE STRATEGY

·
MIGHEST 10 %

BACK TEST TO SHOW EFFICACY THE STRATEGYS

GETA ME DAT
A

1950

Mines
Portfolio Value Stocks
10% Rungso
Ettom
PIE

........unti
ers
- 20

Highest 10% Ringso

793 -

Stocks R
Sigsi
817
EXCEL

SPREADSHETT
RETURN RETURN
on
AWE onWTH If I'm
1950
right
1951 Russ

i Rv Rj Pr
=

If THE
=

MARKET IS
EFFICIENT

2023
Ru Ro
CAPM – cross-sectional
CAPM says that the only
source of risk is the market
factor

Average returns should line

up perfectly against the betas
(in a cross-sectional sense)

1
CAPM – time-series
Alpha from a regression 𝑅𝑅𝑡𝑡 − 𝑅𝑅𝑓𝑓𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽 𝑅𝑅𝑚𝑚𝑡𝑡 − 𝑅𝑅𝑓𝑓𝑡𝑡 + 𝑒𝑒𝑡𝑡
should be zero (in a time-
series sense)
Excess
returns
β
.
on stock
This is the same as saying
. . .
. . . . .
that the “average” returns . . . . ...
should be completely .. ..α . . . . . . .
.. . .. . Excess
explained by beta
.... . ......
returns on
market

. . . .. .

2
CAPM – time-series …
Zero alpha should be true for all stocks / portfolios

This provides the link between the time-series and the cross-
section

3
Time-series regressions
When running regressions of portfolio returns

𝑅𝑅𝑡𝑡 − 𝑅𝑅𝑓𝑓𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽𝐹𝐹𝑡𝑡 + 𝑒𝑒𝑡𝑡

Dependent variable should be 𝑅𝑅 − 𝑅𝑅𝑓𝑓 , not 𝑅𝑅; else intercept is

not an “alpha”

NO assumption of normality, homoscedasticity, etc. in

running OLS

4
Time-series regressions …
The claim is not really about what is left unexplained
• The residuals from the regression are what we have called as
idiosyncratic return
• The point of CAPM is not that there is no idiosyncratic risk

The horse race is not really to reduce the idiosyncratic risk

(not about 𝑅𝑅� 2 )

The horse race is all about explaining the average returns (all
about 𝛼𝛼)

5
Time-series tests of CAPM
Run the following regression for every stock/portfolio

𝑅𝑅𝑖𝑖𝑡𝑡 − 𝑅𝑅𝑓𝑓𝑡𝑡 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽𝑖𝑖 𝐹𝐹𝑡𝑡 + 𝑒𝑒𝑖𝑖𝑡𝑡

and check that all alphas are statistically indistinguishable

from zero

Infeasible in practice; tests are usually done on a subset of

stocks and/or carefully chosen portfolios

6
Cross-sectional tests of CAPM
1st stage: Run the following time-series regression for every
stock

𝑅𝑅𝑖𝑖𝑡𝑡 − 𝑅𝑅𝑓𝑓𝑡𝑡 = 𝛼𝛼�𝑖𝑖 + 𝛽𝛽̂𝑖𝑖 𝐹𝐹𝑡𝑡 + 𝑒𝑒𝑖𝑖𝑡𝑡

2nd stage: Run the following cross-sectional regression

𝑅𝑅�𝑖𝑖 − 𝑅𝑅�𝑓𝑓 = 𝜆𝜆̂ 0 + 𝜆𝜆̂1 𝛽𝛽̂𝑖𝑖 + 𝑢𝑢𝑖𝑖

• The estimated coefficients are 𝜆𝜆’s

• Null hypothesis is that 𝜆𝜆0 = 0 and 𝜆𝜆1 = 𝐹𝐹�
7
Cross-sectional tests of CAPM …
2nd stage problems

1. Errors-in-variable due to pre-estimated betas

• One solution to reduce the magnitude of the problem is to use
portfolios instead of individual stocks
▪ Betas for individual stocks are estimated more imprecisely than those for
portfolios

2. Cross-correlation in residuals
• Solution is to run Fama-MacBeth regressions

8
Fama-MacBeth regressions
Run cross-sectional regressions period by period

𝑅𝑅𝑖𝑖𝑡𝑡 − 𝑅𝑅𝑓𝑓𝑡𝑡 = 𝜆𝜆̂ 0𝑡𝑡 + 𝜆𝜆̂1𝑡𝑡 𝛽𝛽̂𝑖𝑖 + 𝑢𝑢𝑖𝑖𝑡𝑡

Inference is performed on 𝜆𝜆̂̅0 and 𝜆𝜆̂1̅

Can add additional variables on the right-hand-side

𝑅𝑅𝑖𝑖𝑡𝑡 − 𝑅𝑅𝑓𝑓𝑡𝑡 = 𝜆𝜆̂ 0𝑡𝑡 + 𝜆𝜆̂1𝑡𝑡 𝛽𝛽̂𝑖𝑖 + +𝜆𝜆̂ 2𝑡𝑡 𝑍𝑍𝑖𝑖(𝑡𝑡) + 𝑢𝑢𝑖𝑖𝑡𝑡

Null hypothesis is 𝜆𝜆̂̅0 = 0, 𝜆𝜆̂1̅ = 𝐹𝐹,

� 𝜆𝜆̂̅2 = 0 9
Roll critique
The only testable hypothesis of CAPM is whether the market
portfolio is mean-variance efficient
• Sample betas can be made to conform to the SML relationship
exactly because all samples contain an infinite number of ex post
mean-variance efficient portfolios
• CAPM is not testable unless we know and use the exact composition
of the true market portfolio
• Benchmark error due to proxy for market

10
Anomalies/strategies
Investment strategies which seem to earn high returns
without being very risky

Form a portfolio based on observable characteristics, and

measure its returns over time

If the strategy gives high returns on average

• The strategy may be risky, and the high average returns are just
fair compensation for that risk
▪ How do we measure risk?
• If risk does not explain the high returns, is it evidence of market
inefficiency?
1
Portfolio strategies
In some investment strategies, firms with specific
characteristics are viewed as more likely to be undervalued,
and therefore have excess returns, than firms without these
characteristics

In these cases, the strategies can be tested by creating

portfolios of firms possessing these characteristics at the
beginning of a time period, and examining returns over the
time period

2
Steps involved
1. The variable on which firms will be classified is defined,
using the investment strategy as a guide
• This variable has to be observable, though it does not have to be
numerical
2. The data on the variable is collected for every firm in the
defined universe at the start of the testing period, and
firms are classified into portfolios based upon the
magnitude of the variable
• The returns are collected for each firm in each portfolio for the
testing period, and the returns for each portfolio are computed

3
Steps involved (contd..)
3. Portfolios can be equal- or value-weighted (what are the
relative advantages?)
4. Excess returns for these portfolios are computed based on
a factor model
• This involves regressing the portfolio’s returns against market
returns (factor returns) over the sample period
• Check for statistical significance (t-statistic)
5. As a final test, the extreme portfolios can be matched
against each other to see whether there are statistically
significant differences across these portfolios

4
1. Value investing
The conventional definition: A value investor is one who
invests in low price-book value or low price-earnings ratios
stocks

The generic definition: A value investor is one who pays a

price which is less than the value of the assets in place of a
firm

5
Value screens
 Price to book ratios: Buy stocks where equity trades at less
than or at least a low multiple of the book value of equity
(or high book-to-market)
 Price earnings ratios: Buy stocks where equity trades at a
low multiple of equity earnings (or high earnings to price
ratio)
 Price to cash-flow ratio: Buy stocks where equity trades at a
low multiple of cash flows
 Dividend yields: Buy stocks with high dividend yields

6
Price/Earnings ratio
Investors have long argued that stocks with low price
earnings ratios are more likely to be undervalued and earn
excess returns

Higher the P/E, the more the investors are paying for
earnings and the larger the implied expectations for future
earnings growth

This is one of Ben Graham’s primary screens

• And that of his disciple Warren Buffett’s

7
 Value versus growth 17-10-2024

(1)v Fo NOT ENOUGH TO INVEST (THE PLOT) NEED

1964-2020 2000-2020 MORE
Mean SDev Mean SDev
(2) Entro INFORMATION
Growth 12.1 22.7 10.5 25.0
2 10.5 16.4 8.7 17.2
3 12.2 17.0 10.6 17.9
4 11.8 16.4 10.2 17.3
6v > O and L
,
20
5 11.5 16.8 8.6 16.7
6 13.0 15.8 9.3 12.7
7 14.7 17.7 11.5 18.1
8 14.4 20.1 10.4 20.9
9 15.0 19.6 11.1 16.7
Value 16.0 20.9 12.6 19.6
V−G 3.9 20.6 2.1 18.2

8
 Value versus growth consistency Couly MARKET

US) (Che200 stocks)

AVERAGE = 3 9.

LONG PORTFOLO

4 % otherwise It Will BE 0

* TO INVEST
the masicumsed
P/E Q not espectation
can see
VALE : Low optimistic we
minimum need to

NORMAL
and we
choose wich ones are in the
lunderpriced) middle

france goys for

the return)
PIE = avred
Coad for
optimistic 20
GROWTH High PIE pice
: over
corving Despectation
Coverpriced) 9
 Why value is better than growth?
Rational (Value stocks are ‘fallen angels’)
• Represents a distress factor in the economy. Value stocks are more
prone to this source of risk than growth stocks
▪ Fama and French factor model

Physicology

Irrational (Over-reaction)
• Growth stocks are ‘glamorous.’ People tend to want to buy these
and stampede towards them, pushing up the price, and depressing
future returns. Value stocks have been neglected, causing their
price to fall, and expected returns to rise
·
FED PRESIDENT
"irrational esuberes" (famous phrase)

* INVIDIA /
TESLA (price is too high)

10
Top value/growth stocks
https://finviz.com/screener.ashx?v=121&f=cap_largeover,geo_usa&o=forwardpe

11
2. Size investing

1964-2020 2000-2020
Mean SDev Mean SDev
Small 15.9 30.4 13.3 28.6
2 14.6 26.0 11.3 23.2
3 14.9 22.8 12.2 21.2
4 14.3 22.2 10.4 19.3
5 14.8 21.5 11.1 20.9
6 14.0 19.6 11.8 19.0
7 14.2 20.2 11.8 20.9
8 13.6 18.0 11.9 19.6
9 12.7 17.4 11.2 19.6
Big 11.3 16.8 7.7 17.9
S−B 4.6 25.1 5.6 21.1

12
Small versus big consistency

13
January effect
Much of the small firm effect occurs in January

14
January effect analysis
Often explained by tax-loss selling
• However, effect is widespread in international markets also, even
when there’s no capital gains tax
• And, there still seems to be a size effect after controlling for this

May not accord with efficient markets

• Why don’t people buy in December in anticipation?

Window dressing by institutional investors

15
Small cap premium
Small stocks are really “small”
• Market cap of stocks in the smallest decile (end of 2018) is around
$120M (those in big cap is around $110B)

The transactions costs of investing in small stocks is

significantly higher than the transactions cots of investing in
larger stocks

Differential transactions costs are unlikely to explain the

magnitude of the premium across time

16
Size and Value
Size and value screens can be combined

Growth 2 3 4 Value
Small 10.6 16.2 15.6 18.0 19.3
2 12.1 15.1 16.1 16.2 17.1
3 12.4 15.2 14.4 16.3 17.6
4 13.7 13.0 13.7 15.5 15.6
Big 12.2 11.4 12.0 11.1 12.9

Long-short return is 7.0% over 1964-2020 and 4.4% over

2000-2020

17
3. Profitability investing
Invest in profitable companies
• Gross profitability: Sales minus the cost of goods sold to total
assets
• Return on assets: Earnings to total assets

Shouldn’t this already be reflected in current price?

Currently profitable firms often generate significantly higher

equity returns than unprofitable firms, despite often having
higher valuation ratios

18
Profitability investing results (1)

1964-2020 2000-2020
Mean SDev Mean SDev
Unprofitable 10.8 26.5 6.3 30.4
2 10.7 20.6 5.8 25.0
3 10.7 18.2 7.8 20.4
4 11.6 17.2 9.3 19.1
5 12.5 16.0 10.3 17.9
6 11.4 17.5 6.8 20.0
7 11.2 15.9 8.9 17.0
8 13.1 16.2 10.9 14.7
9 14.0 17.9 9.4 18.1
Profitable 13.2 18.8 11.3 17.2
P−U 2.4 17.1 5.0 17.3

19
Profitability investing results (2)

20
Profitability as a hedge
In periods of severe volatility, the profitability factor has positive
returns
• Might reflect investors’ preference for firms that generate revenue and earnings
more efficiently
▪ In these market environments, current profits may take on paramount importance, with projected,
longer-term revenues taking a backseat

21
Why profitability?
The profitability premium is higher among firms with smaller
capitalization, higher volatility, less analyst coverage, fewer
institutional holdings, lower dollar trading volume, higher
bid-offer spread, lower credit ratings, higher illiquidity

22
4. Investment investing
Invest in companies with low investment
• Asset growth
• Net stock issuance

23
Investment investing results (1)

1964-2020 2000-2020
Mean SDev Mean SDev
Conservative 14.9 22.0 11.1 22.3
2 14.9 18.2 12.1 20.9
3 13.4 16.0 10.8 16.2
4 12.1 14.8 8.9 15.6
5 12.2 15.3 11.0 16.2
6 12.0 16.8 8.2 17.0
7 12.8 17.4 8.8 16.2
8 12.1 18.1 10.8 18.9
9 13.4 21.9 9.6 23.2
Aggressive 10.0 24.1 7.0 25.7
C−A 4.8 15.9 4.1 17.4

24
Investment investing results (2)

25
Why investment investing?
Risk-based arguments suggest that firms with growth options
(low asset growth firms) are riskier than firms that have
converted their growth options into assets (high asset growth
firms)

Mispricing explanation is that investors overreact to transient

asset-growth rates only to be disappointed when growth and
stock returns revert to a more normal level

26
5. Momentum investing
At the beginning of each month (say Nov 1st, 2020) compute
returns of each stock over the past 12 months

Rank all stocks by their past 12-month returns

Divide the stocks into 10 portfolios. Top 10% of the stocks

with the highest past returns are referred to as winners.
Bottom 10% of the stocks with the lowest past returns are
referred to as losers

27
Momentum investing results (1)

1964-2020 2000-2020
Mean SDev Mean SDev
Losers 4.6 32.8 8.6 45.1
2 9.3 23.6 8.4 29.1
3 11.4 21.3 9.7 25.2
4 11.6 18.3 11.6 20.8
5 10.8 16.5 11.5 18.1
6 11.8 17.2 10.1 18.4
7 11.7 15.7 9.6 15.0
8 13.8 17.3 9.5 16.0
9 14.5 18.1 9.5 16.9
Winners 19.7 23.9 11.4 21.9
W−L 15.1 27.3 2.9 35.3

28
Momentum investing results (2)

29
Momentum profits
Violation of weak-form efficiency

Let no academic (aka me!) scoff at technical analysis while

singing virtues of momentum

30
Momentum funds

31
Momentum crashes
Strategy had returns of −42% in Jan 2001, and −40%,
−46%, and −19% in Mar-May 2009, respectively

Returns for year 2009 were −120%

Excluding 2009, momentum returns are still 9%/year in this

century

Volatility-scaled momentum strategies avoid these crashes

32
6. Reversal investing
Actually, in momentum investing a better signal is stock
return over past 12 months excluding the most recent month

What if we look at most recent one-month winners and

losers?

33
Reversal investing results (1)

1964-2020 2000-2020
Mean SDev Mean SDev
Losers 12.8 25.9 8.6 29.6
2 14.5 20.9 10.5 23.5
3 15.0 19.3 10.7 19.9
4 13.3 18.1 11.4 19.0
5 12.9 16.4 10.8 16.6
6 11.8 16.7 8.7 16.7
7 11.8 17.2 9.8 18.2
8 11.8 19.0 8.4 20.0
9 10.2 18.7 6.9 21.9
Winners 9.6 22.6 8.3 28.1
L−W 3.2 19.5 0.3 18.7

34
Reversal investing results (2)

35
Reversal investing
This is an extremely high turnover strategy

Turnover for long-short strategy is 150%

Strategy works better for small, illiquid stocks

36
7. Long-term reversal investing
Define winners and losers based on returns over the past 5
years (excluding the most recent year to not confound the
effects of momentum)

37
Long Reversal investing results (1)

1964-2020 2000-2020
Mean SDev Mean SDev
Losers 16.2 28.4 11.9 31.3
2 13.4 20.1 7.2 18.6
3 13.5 17.0 9.5 16.8
4 13.3 18.8 9.0 17.9
5 12.6 16.0 8.8 15.2
6 13.0 16.1 11.1 16.9
7 12.7 15.8 10.1 14.7
8 13.0 16.9 11.4 16.0
9 11.3 18.9 9.9 17.8
Winners 12.5 24.1 11.6 26.2
L−W 3.7 25.5 0.3 26.3

38
Long Reversal investing results (2)

39
8. Accruals investing
Accounting rules prescribe what is called “accrual accounting”
rather than “cash flow” accounting

40
Accruals (1)
Account for revenues generated during a particular period (fiscal
quarter or fiscal year) based on “completion of sale” and not based on
sales for which the firm has collected payments
• A credit sale during a period is recorded as revenue for that period, although it
does not generate cash inflows in that period

Costs are recorded only to the extent that they are incurred to
generate the completed sale
• If a firm purchases goods in a particular period but does not sell them in that
period, the purchase in not recorded as a cost. This purchase goes into inventory

41
Accruals (2)
The difference between accounting earnings and cash flows is
called accruals
• Earnings = Cash flow + accruals
• Accruals = ∆ Working capital – D&A

While accrual accounting is good in principle, it allows

managers some discretion in accounting for revenues and
costs in the income statement

42
Accruals (3)
Two important components of accruals

1. Change in inventory
• Business reason: Firm anticipates a large sales increase
• Possible manipulation
▪ Inventory hard to sell but the management has not fully accounted for its
decline in value (e.g. out of fashion clothes)
▪ Obsolete items not written off (e.g. outdated computers or cell phones)

43
Accruals (4)
2. Change in accounts receivables
• Business reasons: Firm anticipates a large sales increase
• Possible manipulation
▪ Bogus sales booked to meet earnings or revenue targets
▪ Uncollectable receivables not written off

44
Accruals (5)
Earnings accompanied by high current accruals (and
therefore low current cash flows) tend to have lower future
earnings and cash flows than net income accompanied by low
(or negative) current accruals

Therefore, net income accompanied by high accruals is of

lower earnings quality while net income accompanied by low
accruals is of higher earnings quality

45
Accruals (6)
If investors are not able to fully distinguish between low
quality and high quality earnings, they would overvalue
stocks with high accruals and undervalue stocks with low
accruals (Sloan, 1996)

46
Accruals investing results (1)

1964-2020 2000-2020
Mean SDev Mean SDev
Good EPS 14.3 21.0 11.3 23.2
2 14.4 20.0 10.6 22.6
3 12.9 17.4 9.0 20.3
4 12.3 16.1 10.4 17.5
5 12.3 15.6 9.5 14.7
6 12.3 18.0 8.0 20.0
7 12.5 17.7 9.8 18.8
8 12.4 18.1 9.9 17.3
9 11.5 20.7 10.5 19.9
Bad EPS 10.3 23.8 9.6 25.7
G−B 3.9 9.8 1.7 11.3

47
Accruals investing results (2)

48
9. Earnings drift
Every quarter (in the U.S) and less frequently elsewhere,
firms report their earnings for the most recent period

When firms make earnings announcements, they convey

information to financial markets about their current and
future prospects. The magnitude of the information, and the
size of the market reaction, should depend upon how much
the earnings report exceeds or falls short of investor
expectations

49
Earnings announcements
In an efficient market, there should be a “quick” reaction to
the earnings report, if it contains surprising information, and
prices should increase following positive surprises and down
following negative surprises

50
Earnings drift
Slow price response to earnings announcements

Referred to as post-earnings announcement drift (PEAD)

Violation of semi-strong form efficiency

51
Earnings drift …

52
Earnings profits

Chan, Jegadeesh, and Lakonishok (Financial Analysts Journal, 1999)

53
10. Stock issuance
Managers may repurchase stock when they believe that the
equity is underpriced and issue stock when they believe that
equity is overpriced

54
Equity issuance investing results (1)

1964-2020 2000-2020
Mean SDev Mean SDev
Repurchase 14.2 16.3 10.0 15.8
1 12.7 18.6 11.7 19.4
2 11.2 16.9 7.8 17.5
3 11.2 17.0 6.9 17.7
4 13.0 17.3 12.0 19.1
5 13.9 18.6 13.1 18.9
6 14.2 20.8 11.5 24.4
7 12.9 22.8 10.2 27.3
8 12.0 23.3 11.4 29.7
9 9.4 21.1 7.1 25.5
High Issue 7.7 23.5 4.5 26.2
Repu−Issu 6.5 13.8 5.4 15.6
55
Equity issuance investing results (2)

56
11. Volatility investing
Finance theory says that
Higher risk  Higher return

Total risk
E(RX) > E(RY)

Total risk

X Y

57
Volatility investing results (1)

1964-2020 2000-2020
Mean SDev Mean SDev
Safe 12.2 14.6 10.4 13.5
2 12.7 16.7 8.8 16.4
3 12.3 16.8 9.5 16.9
4 13.2 17.4 10.1 17.9
5 13.5 19.1 9.6 20.2
6 14.8 20.9 9.4 21.1
7 14.2 23.8 11.0 28.0
8 16.5 27.3 15.8 32.9
9 13.8 31.0 9.9 34.4
Risky 7.3 38.1 8.4 44.6
S−R 4.9 33.8 2.1 37.3

58
Volatility investing results (2)

59
Risk anomaly
This is taking finance back to old ages
Even more, it is overturning the basic rules of finance
Higher total risk  Lower return

Expected return Theory

Data

60
Why risk anomaly?
Why do high volatility stocks underperform?

Excess demand for high volatility stocks

• Maybe (excess) preference
▪ Lottery like payoffs

Limits to arbitrage
• Long-only investors cannot help in correcting the mispricing of
high-beta stocks
• If subject to tracking error, they might even be unwilling to invest
in low-beta stocks

61
12. Low beta investing
Finance theory during enlightenment
• Higher systematic risk  Higher return

62
Betting against beta
Data show that the slope of CML is too flat

Theory

Expected return
Data
Market

63
Implementing low beta strategies

Theory

Expected return
Data
-α
Market

+α

β
β=0.5 β=1.5

Which securities do you want?

• Beta = 0.5 or beta = 1.5?

What is the alpha being generated?

64
Implementing low beta strategies …

Source: Frazzini and Pedersen (2011)

65
Cumulative growth

Quality as a combination of (Profitability + Growth + Safety + Payout) [AQR]

66
Seasonal effects
Day of the week effect
Returns are on average negative on Monday

Returns are on average positive Wednesday−Friday

Turn of the month effect

Average return for the four days around the turn of the
month, starting with the last day of the prior month is
significantly more positive than the average returns on any
other four-day period

67
Seasonal effects …
Holiday effect
Returns are much larger in the days preceding market
closures for holidays

Daylight saving
Kamstra, Kramer, and Levi (2000) show that Friday-Monday
return is significantly lower on daylight saving weekends than
other weekends

68
Seasonal effects …
Sunshine effect
Hirshleifer and Shumway (2001) find that sunshine is
positively related to stock returns

Lunar effect
Yuan, Zheng, and Zhu (2005) find that stock returns are
lower on days around a full moon than on days around a new
moon

69
Seasonal effects …
Weekend effect
Volume and volatility are higher on Friday than Monday,
despite the fact that information is released over the
weekend that cannot be traded on until the market opens

70
Seasonal effects …
What does all this mean?

“October. This is one of the peculiarly dangerous months to

speculate in stocks. The others are July, January,
September, April, November, May, March, June, December,
August, and February.“
Mark Twain

71
Why anomalies?
If anomalies really are anomalies, i.e., these strategies earn
returns higher than is commensurate with their risk
• Why do they exist?
• Aren’t there smart people wanting to get rich?
• Wasn’t there something called ‘arbitrage’?

72
Limited arbitrage
Arbitrageurs (e.g. hedge funds) do not have infinite
investment horizons, they are risk averse, and they have
limited capital.

Therefore, strategies that are volatile and take a long time to

converge to fundamental value are quite risky for
arbitrageurs (think of the internet bubble). Their principals
(investors in their funds) may not wait sufficiently long to see
if prices indeed converge in the long run

73
Limited arbitrage …
After they invest in these strategies, it is possible that prices
diverge further from fundamental values in the short run and
the arbitrageurs may run out of capital prior to convergence.
Because of this possibility, there is a limit on the size of the
positions money managers are willing to take.

Also, most of these inefficiencies are bigger for small and

mid-cap stocks than lager cap stocks. The transaction costs
are higher for small and mid-cap stocks.

74
Interpreting anomalies
 Poor risk measurement

 Frictions to trading and exploiting the anomaly (bid-ask

spread, transactions costs, liquidity, taxes, etc.)

 Data mining
• If you try many strategies, some of them will do great in historical
data
▪ Doesn’t tell you anything about future performance
• Many anomalies appear in international markets

 Behavioral explanations
75
Key take-aways
Sorting on screens can deliver extra profits
• May not be explained by risk factors (or, at least, the risk factors
that we know of)

How to validate?
• Back-testing
• Plausible ‘stories’ for why this happened and why this is expected to
continue

Limitations
• Not a pure arbitrage
• Can/should be supplanted with fundamentals and common sense
76
Performance measurement

1
Risk-adjusted measures
Sharpe measure rp − rf
Sp =
σp

Jensen’s alpha α p = rp −  rf + β p ( rm − rf )

Appraisal ratio αp
IR p =
σe

2
More on alpha
CAPM alpha
𝑅𝑅𝑖𝑖𝑖𝑖 − 𝑅𝑅𝑓𝑓𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑅𝑅𝑓𝑓𝑖𝑖 + 𝑒𝑒𝑖𝑖𝑖𝑖

Fama and French (1993) 3-factor alpha

𝑅𝑅𝑖𝑖𝑖𝑖 − 𝑅𝑅𝑓𝑓𝑖𝑖 = 𝛼𝛼𝑖𝑖3 + 𝛽𝛽𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑅𝑅𝑓𝑓𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝑆𝑆𝑀𝑀𝐵𝐵𝑖𝑖 + 𝛽𝛽𝑖𝑖3 𝐻𝐻𝑀𝑀𝐿𝐿𝑖𝑖 + 𝑒𝑒𝑖𝑖𝑖𝑖

• SMB (small minus big): Size factor

• HML (high minus low): Value factor

3
More on alpha …
Fama and French (2015) 5-factor alpha
𝑅𝑅𝑖𝑖𝑖𝑖 − 𝑅𝑅𝑓𝑓𝑖𝑖
= 𝛼𝛼𝑖𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑅𝑅𝑓𝑓𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝑆𝑆𝑀𝑀𝐵𝐵𝑖𝑖 + 𝛽𝛽𝑖𝑖3 𝐻𝐻𝑀𝑀𝐿𝐿𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝑅𝑅𝑀𝑀𝑊𝑊𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝐶𝐶𝑀𝑀𝐴𝐴𝑖𝑖 + 𝑒𝑒𝑖𝑖𝑖𝑖

• SMB (small minus big): Size factor

• HML (high minus low): Value factor
• RMW (robust minus weak): Profitability factor
• CMA (conservative minus aggressive): Investment factor

4
More on alpha …
Can also add a momentum factor. Example, a 6-factor alpha
𝑅𝑅𝑖𝑖𝑖𝑖 − 𝑅𝑅𝑓𝑓𝑖𝑖
= 𝛼𝛼𝑖𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑅𝑅𝑓𝑓𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝑆𝑆𝑀𝑀𝐵𝐵𝑖𝑖 + 𝛽𝛽𝑖𝑖3 𝐻𝐻𝑀𝑀𝐿𝐿𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝑅𝑅𝑀𝑀𝑊𝑊𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝐶𝐶𝑀𝑀𝐴𝐴𝑖𝑖
+ 𝛽𝛽𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖 + 𝑒𝑒𝑖𝑖𝑖𝑖

Or add a liquidity factor. Example, another 6-factor alpha

𝑅𝑅𝑖𝑖𝑖𝑖 − 𝑅𝑅𝑓𝑓𝑖𝑖
= 𝛼𝛼𝑖𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑅𝑅𝑓𝑓𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝑆𝑆𝑀𝑀𝐵𝐵𝑖𝑖 + 𝛽𝛽𝑖𝑖3 𝐻𝐻𝑀𝑀𝐿𝐿𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝑅𝑅𝑀𝑀𝑊𝑊𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝐶𝐶𝑀𝑀𝐴𝐴𝑖𝑖
+ 𝛽𝛽𝑖𝑖𝑖 𝐿𝐿𝐿𝐿𝑄𝑄𝑖𝑖 + 𝑒𝑒𝑖𝑖𝑖𝑖

5
More on Information ratio
Two ways to calculate information ratio

1. Use a factor model

Calculate alpha and standard deviation of residuals

𝑟𝑟𝑝𝑝𝑖𝑖 − 𝑟𝑟𝑓𝑓𝑖𝑖 = 𝛼𝛼𝑝𝑝 + 𝛽𝛽𝑝𝑝 𝑟𝑟𝑚𝑚𝑖𝑖 − 𝑟𝑟𝑓𝑓𝑖𝑖 + 𝑒𝑒𝑝𝑝𝑖𝑖

Calculate information ratio

𝛼𝛼𝑝𝑝
𝐿𝐿𝑅𝑅 =
𝜎𝜎𝑒𝑒𝑝𝑝

6
Information ratio …
2. Use benchmark-adjusted returns
Calculate “alpha” and standard deviation of residuals

𝑟𝑟𝑝𝑝𝑖𝑖 − 𝑟𝑟𝑏𝑏𝑖𝑖 = 𝛼𝛼𝑝𝑝 + 𝑢𝑢𝑝𝑝𝑖𝑖

Fund excess returns

Calculate information ratio

𝛼𝛼𝑝𝑝
𝐿𝐿𝑅𝑅 =
𝜎𝜎𝑢𝑢𝑝𝑝

When would this approach make sense?

7
Investment styles and benchmarks
Most popular procedure to measure performance is to
compare fund returns with returns on a benchmark index
selected based on fund’s investment style

 Equity Funds
• Passive Funds
• Active Funds – Style classification
 Bond Funds
• High-yield bond funds
• Money market funds
• Municipal bond funds

8
Equity styles

9
Index characteristics

Russell S&P 500

10
Fixed income indices
 Barclays Aggregate
 Barclays Govt/Credit
 Barclays Long Govt/Credit
 Merrill Lynch Govt/Corp 1-3 Years
 Merrill Lynch High Yield Master
 Citigroup Broad Inv Grade

11
Fixed income style box
Short Duration Long

Short-Term Interm-Term Long-Term

High
High High High

Quality
Short-Term Interm-Term Long-Term
Medium Medium Medium

Short-Term Interm-Term Long-Term

Low

Low Low Low

Risk Low Moderate High

12
Style analysis
Some funds may follow combinations of different styles. How
would you choose the appropriate benchmarks?

The particular style followed by a fund manager may well

differ from the style he/she claims to follow

13
Style analysis …
Institutional funds hire fund consultants to evaluate the style
of the manager to ensure that the fund performance is
consistent with the style that the manager follows

Style drift can lead to termination

• Eg. First Quadrant’s contract was terminated by MI Municipal EES in
Dec, 2001 for style drift

14
Sharpe’s style analysis
Style revealed by portfolios’ returns
𝑟𝑟𝑖𝑖𝑖𝑖 = 𝛼𝛼⏟𝑖𝑖 + [𝛽𝛽𝑖𝑖𝑖 𝐹𝐹𝑖𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖 𝐹𝐹𝑖𝑖𝑖 + ⋯ + 𝛽𝛽𝑖𝑖𝑖𝑖 𝐹𝐹𝑖𝑖𝑖𝑖 ]
Active return Style return
+ 𝑒𝑒�
𝑖𝑖𝑖𝑖
Tracking error

Return can be decomposed into

• Style : term in the square brackets
• Selection: Residual component

It is useful to think of these regressions as not “factor

models” but rather as style models
15
Sharpe’s style analysis …
The factors (styles) should be exhaustive with respect to the
manager’s universe and mutually exclusive
• The term in the brackets is the “normal benchmark”

Betas are the weights on the styles / factors

• Factor sensitivities (betas) sum to 1.0

16
Example
ACP fund is a large-cap growth fund

Style analysis over the last 3 years reveals an R2 of 91.9%

with weights in 4 styles as :

Do you agree that ACP is accurately described as a large-cap

growth fund? 17
Example …
ACP is indeed actively managed
• Only 91.9% of the return is attributable to style

Fund is essentially large-cap

• Only 4% in small-cap growth stocks

However, ACP invests so as to obtain results similar to those

from a portfolio with 39% in large-cap value stocks, 57% in
large-cap growth
• Better described as a large-cap product with a growth bias

18
Example …
Further rolling historical analysis reveals the following with
an active return of −0.38% and a tracking error of 6.58%
(both annualized)

What would you do about ACP?

19
Example …
Historical weights generally support the previous conclusion
of style (large-cap with growth bias)

Information ratio is −0.38%∕6.58% = −0.06

• For each % of tracking risk, ACP earned −0.06% active return
• Portfolio’s active risk is unrewarded

20
Example …
An even better choice of indices might have been
• Russell Top 200 Growth
• Russell Top 200 Value
• Russell Mid-cap Growth
• Russell Mid-cap Value
• Russell Small-cap Growth
• Russell Small-cap Value

21
Mutual funds classification

22
Empirical evidence – Mutual funds
Do mutual funds outperform their benchmarks?
• As a group (average performance)
• Individually (cross-sectional variation)

Do individual funds exhibit persistence in performance (“hot

hands”)?
• Is past performance an indicator of future performance?

23
Average performance
Chen, Jegadeesh, and Wermers (2000) find
• Raw returns using stocks purchased by funds outperform stocks sold by them
• Provides prima facie evidence of stock picking skills (before expenses and trading
costs)

Wermers (2000) examines the difference between returns based on

stock holdings and net returns
• Difference is around 2.3%
• 0.7% due to lower returns on non-stock holdings
• 0.7% due to expenses and trading costs

Gross returns based on stocks’ holdings cover expenses and trading

costs; other fund assets reduce the overall return

24
Average performance …
U.S. data 1984-2006

Source: Fama and French (2010)

25
Cross-sectional variation
Kosowski et al. (2005) find that cross-sectional standard
deviation of alphas is high
• Indicates the possibility that some funds are performing well (and
some rather badly!)

What about tails of distribution?

• What if I told you that a particular fund has an average abnormal
return of 10% pa?
• What if this return was achieved by best performing fund out of
1000 funds?
• Have to control for ‘luck’

26
Cross-sectional variation …

Source: Fama and French (2010) 27

Performance persistence

28
Persistence – Carhart (1997)

29
Performance and flows
People act as if there were much more persistency in
performance

30
Institutional investors performance

Source: Amit Goyal and Sunil Wahal, “The Selection and Termination of Investment
Management Firms by Plan Sponsors,” Journal of Finance 63 (2008): 1805–1847

31
Performance bottom-line

32
By the way, costs of investing in funds
Front-end load (charged once)
• Commission for purchasing shares ( ≤ 8.5%)
▪ Typically lower than 6%
• Low-load (3%), no-load
• Used primarily to pay brokers who sell the funds
• Reduce the amount of money invested
▪ If front-end load is 5%, then need 5.3% to break even!

Back-end load (charged once)

• Exit fees (5%-6% going down by 1% per year)

Front-end and back-end loads are explicitly charged to the

investor 33
Costs of investing …
Operating expenses (charged annually)
• Range from 0.2%-2%
• Include admin expenses and fees paid to the manager

Distribution costs (charged annually)

• Distribution costs ( ≤ 1%)
• Include advertising etc., and commissions paid to brokers

Investors are charged explicitly; these expenses are

deducted periodically from the assets of the fund

34
Impacts of costs on investment
performance

35
Costs of investing …
More expenses: “Soft dollars”

Fund manager directs the fund’s trades to a broker and earns

soft dollars
• Broker pays for databases, computer hardware etc.
• Fund gains by not reporting these expenses and can advertise low
expense ratio
• Fund pays high commission to broker; trading costs not reflected in
expenses; however net performance suffers

36
Costs of investing …

37
Bottom-line
Without strong evidence to the contrary, a useful prior is to
between firm and mild believer in market efficiency
• Trivial arbitrage does not exist
• Trivial great bets do not exist either

“If you’ve been playing poker for half an hour and you still
don’t know who the patsy is, you’re the patsy.” (Warren
Buffett)

38
Behavioral finance
People in classical finance are rational

People in behavioral finance are normal

1
Behavioral finance – What is it?
Financial practitioners are not fully rational
• Rely on rules of thumb
• Use psychology-based models for decision making

Form (description as well as frame of a decision problem) as

well as substance influences practitioners

Prices of financial assets may deviate from fundamentals

• Investment decisions and risk allocation may be suboptimal

2
Behavioral finance – Why study it?
Markets may not be (are not?) efficient

Practitioners need to know because they make particular

types of (costly) mistakes

Misallocation of capital can lead to economic crises

3
Why study it? …
Study of behavioral finance can help practitioners in
• Recognizing their own mistakes and those of others
• Avoid mistakes (maybe!)
• Understanding that one investor’s mistakes can become another’s
profits or another’s risk

“If you don’t know who you are, the stock market is an
expensive place to find out” Adam Smith, The Money Game

4
Traditional finance
Agents are risk-averse, self-interested, utility maximizers

1. When they receive new information, they update their

beliefs correctly (in a Bayesian fashion)

2. Given their beliefs, they make choices that are consistent

with utility theory

5
E.g., Modern portfolio theory
1. Investors have correct beliefs about future expected
return and risk
2. Given these beliefs, they choose a mean-variance efficient
portfolio

6
Implication – EMH
 Price ≡ Fundamental value
 No free lunch
 Markets are efficient

E.g., If fundamental value of Ford is $15 but irrational

pessimistic traders push the price to $10, rational
arbitrageurs will buy Ford at its bargain price and, perhaps,
at the same time, hedge their bet by shorting a “substitute”
security, such as GM

7
EMH …
Do not need everyone to be rational

All that is needed is a sufficient number of smart investors

But …

The arbitrage may be unattractive/risky making the

mispricing persist

8
What might prevent arbitrage? (1)
Fundamental risk: Bad news about Ford’s fundamental value
causes the stock to fall further, leading to losses
• Substitute securities are rarely perfect

9
What might prevent arbitrage? (2)
Noise trader risk: Even if GM is a perfect substitute security
for Ford, there is still the risk that the pessimistic investors,
causing Ford to be undervalued in the first place, become
even more pessimistic, lowering its price even further
• Arbitrageurs might have to liquidate their positions early at steep
losses
• Delegated money management makes this worse
• Creditors can exacerbate the problem

10
What might prevent arbitrage? (3)
Implementation costs: Commissions, bid–ask spreads, price
impact, short-sale constraints
• Maybe costly to even learn about mispricing
• Resources required to exploit it are expensive

11
Real life: Limits to arbitrage
Price is right ⇒ No free lunch

No free lunch ⇒ Price is right

Even if professional money managers do not beat the market

(no free lunch), this does not mean that the price is right

12
Evidence on LTA
Anomalies are not a good evidence because they could be
compensation for some yet unknown risk
• Joint hypothesis problem

Need examples where mispricing can be established without

any doubt

13
LTA evidence (1)
Twin shares: Shares of Royal Dutch, which are primarily
traded in the USA and in the Netherlands, are a claim to 60%
of the total cash flow of the two companies, while Shell,
which trades primarily in the UK, is a claim to the remaining
40%

⇒ MV(Royal Dutch) = 1.5×MV(Shell)

14
LTA evidence (1) …

15
LTA evidence (2)
Internet carve-outs: In March 2000, 3Com sold 5% of its
wholly owned subsidiary Palm Inc. in an IPO. After the IPO, a
shareholder of 3Com indirectly owned 1.5 shares of Palm

Palm=$95 ⇒ 3Com≥$142

Actual, 3Com=$81 ⇒ 3Com(other than Palm)=−$60

16
Information (1)
Huberman and Regev (2001): On Sunday, 3rd May 1998, the
NYTimes carried an article on a potential new drug being
researched by EntreMed. EntreMed’s stock rose to $85 from
close on previous Friday of just over $12. Stock subsequently
fell during trading on Monday, May 4th to close at $52. Three
weeks later, the stock was still trading at $30

The potential breakthrough had been reported no less than 5

months earlier in Nature and other newspapers, including the
NYTimes!

17
Information (2)
Rashes (2001) MCI-MCIC “Massively Confused Investors
Making Conspicuously Ignorant Choices”
 MCI was a telecom company: traded on Nasdaq under the
ticker MCIC
 Massmutual Corporate Investors is a closed-end mutual
fund: traded on NYSE under the ticker MCI
 There was an unusual comovement between their stock
prices over 1996-97 when MCI was involved in merger talks

18
Information (3)
What’s in a name? Cooper, Dimitrov, and Rau (2001), “A
Rose.com by Any Other Name”
 Investors revalued companies upward by about 53% in a
11-day window around the announcement of the name
change when they simply changed their name to include
“dot-com” in 1998-1999
• Even for firms whose business is unrelated to internet, 23%
abnormal return!

19
Why are investors not rational? (1)
Bounded rationality

 People have informational, intellectual, and computational

limitations

 Use heuristics in decision-making

 Satisfice (Satisfy + Suffice) to create a stop rule to the

decision process

20
Why are investors not rational? (2)
Psychological biases

Classified as
• Biases in beliefs
• Biases in preferences

Or as
• Cognitive errors
• Emotional biases

21
Behavioral errors and biases

1
Types of errors
Cognitive errors are basic statistical, information-
processing, or memory errors

Emotional biases arise as a result of attitudes and feelings

that cause the decision to deviate from rational utility
maximization

2
Cognitive errors
Blind spots in human mind

Not due to predisposition towards some judgements

Investors can adapt their behavior if the source of bias is

logically identifiable

Better information, education, and advice is useful for

moderating the biases

3
Emotional biases
Stem from impulse, sometimes unreasoned judgements

Individuals may be unable to control these emotions even if

they want to

Only possible to adapt to these biases

4
Cognitive errors
1. Belief perseverance biases
i. Conservatism
ii. Confirmation
iii. Representativeness
▪ Base-rate neglect
▪ Sample-size neglect
iv. Illusion of control
v. Hindsight
2. Information processing biases
i. Anchoring
ii. Mental accounting
iii. Framing
iv. Availability
5
Emotional biases
1. Loss aversion
• Disposition effect
2. Overconfidence
3. Self-control
4. Status quo
5. Endowment
6. Regret-aversion

6
Cognitive errors (1)
Belief perseverance biases
i. Conservatism
ii. Confirmation
iii. Representativeness
▪ Base-rate neglect
▪ Sample-size neglect
iv. Illusion of control
v. Hindsight

7
Conservatism
Maintaining your prior views (or forecasts) by failing to
properly incorporate new information as it becomes available

Me continuing to believe in EMH even after compelling evidence to the contrary 

8
Conservatism …
Diagnosis
 Say that after you have made an investment based on your
own research, an advisor gives you information that
contradicts your opinion. What do you do?
 How frequently do you take in new information about the
market?

Mitigation
 React decisively and fully to any new information and seek
unbiased counsel

9
Confirmation
Seeking out evidence that confirms your beliefs and ignoring
evidence that contradicts them
• Researchers frame their data in ways that tend to confirm their
hypotheses
• During an election season, people tend to seek positive information
that paints their favored candidates in a good light. They will also
look for information that casts the opposing candidate in a negative
light

10
Confirmation …

11
Confirmation …
Classic example: Capital punishment studies
• In one experiment (Lord, Lepper, and Ross, 1979), 24 pro-death
penalty students and 24 anti-death penalty students critically
evaluated “studies” on capital punishment
• These students found that studies which supported their pre-
existing view were superior to those which contradicted it, in a
number of detailed and specific ways
• In fact, the studies all described the same experimental procedure
but with only the purported result changed

12
Confirmation …
Consequences for investments
 Only reaffirming evidence is considered
 Information is ignored that refutes the validity of the screen
 Under-diversified portfolios if you fall in love with a
particular stock or sector
 Holding too much of your own-company’s stock because
you are convinced of its growth prospects

13
Confirmation …
Mitigation
 Actively look for information that challenges beliefs
 Obtain corroborating support for investment decision
• If a stock breaks through 52-week high, obtain supporting
information to assure good value
• Confirming an investment idea through purchase is not a good
strategy

14
Representativeness
Classifying new information based on past experiences or the
way things have happened in the past

Rely on “best-fit” approximation

15
Representativeness …
Linda is 31, single, outspoken, and very bright. She majored
in philosophy. As a student she was deeply concerned with
issues surrounding equality and discrimination. Is it more
likely that Linda is:

 A bank clerk, or

 A bank clerk and active in feminist movement?

16
Representativeness …
Steve, a 30-year old American, has been described by a
former neighbor as follows: “Steve is a very shy and
withdrawn, invariably helpful, but with very little interest in
people or the social world. A meek and tidy soul, he has a
need for order and structure and a passion for detail.” Which
occupation is Steve currently more likely to have:

 Salesman, or

 Librarian?

17
Representativeness …
What is the probability that Company A (ABC, a 75-year old
steel manufacturer that is having some business difficulties)
belongs to group X (value stocks that will likely recover)
rather than group Y (companies that will go out of business)?

What is the probability that a AAA-rated municipal bond

(issued by an “inner city” and a racially-divided county)
belongs to group X (risky municipal bonds) rather than to
group Y (safe municipal bonds)?

18
Representativeness …
May explain long-term reversal: Stocks that have been
extreme losers in the preceding three years do much better
than extreme past winners over the subsequent three years
• Investors become unduly pessimistic about the prospects of the
past losers, driving down their prices. Prices revert back giving
exceptional returns

19
Representativeness: Sample size neglect
Hot hands: Imagine that you’re the coach of a basketball
team. There’s 10 sec left, and your team is down by a
basket. Your star player (5-year career average of 55% shots
made) is only 2 for 10 today. Another veteran player (5-year
career average of 45% shots made) is 10 for 10 today.
Whom do you give the ball to?

20
Representativeness: Hot hands

21
Representativeness: Sample size neglect …
Law of small numbers!

Which of the following two sequence of coin tosses from an

unbiased coin is more likely?
HTHTHTHT
HHHHTTTT

Gambler’s fallacy: If a fair coin generates four heads in a

row, people will say that “tails are due.” Since they believe
that even a short sample should be representative of the fair
coin, there have to be more tails to balance out the large
number of heads
22
Representativeness: Regression to mean
Flight training example (Kahneman)

Instructors of pilots found that bad landings improved after

verbal punishment, while good landings did not after praise

Concluded that verbal reward was detrimental to learning,

while punishment was beneficial

23
Representativeness: Regression to mean
It’s been better to have been a novice than a professional the
past few years, because people with the most experience
have been the most cautious. But markets do regress back to
the mean, and I agree that we are late in the ball game. This
is the longest period we’ve ever had with such high returns
from equities, and I can’t believe it’s a new era that will just
keep going forever. I don’t know if returns going forward will
be 7% or 8%, but I’m pretty sure that they will be below
average.
Excerpt from an interview that appeared in the August 18, 1997 issue of Fortune magazine, with global
strategist Barton Biggs of Morgan Stanley and senior investment advisor Robert Farrell of Merrill Lynch

24
Representativeness: Regression to mean
Regression to mean implies that future returns will be closer
to their historical average. But is does not mean that they
will be below their historical average

25
Representativeness …
 Lack of performance persistence but performance-fund flow
relationship in mutual funds
 Difficulty in market timing

26
Illusion of control
When people believe that they can control or influence
outcomes
• A mid-level manager may believe that he can personally influence
his employer's stock price
• People permitted to select their own numbers in lottery are willing
to pay a higher price than those using randomly assigned numbers

27
Illusion of control …
Consequences
 Excessive trading
 Inadequate diversification (often due to over-investing in
own-company stock)

Mitigation
 Keep detailed records of your predictions and refer back to
these in order to get an honest assessment of your
predictive powers

28
Hindsight bias
Seeing past events as having been inevitable and predictable
• People tend to remember their own predictions of the future as
more accurate than they actually were

29
Hindsight bias …
Consequences
 Overestimating how well they predicted various events
 Unfairly assessing money manager or security performance

Mitigation
 Keep detailed records and refer back to them
 Seek out contrary and independent views

30
Cognitive errors (2)
Information processing biases
i. Anchoring
ii. Mental accounting
iii. Framing
iv. Availability

31
Anchoring
When required to estimate a value with unknown magnitude,
people generally begin by envisioning some initial default
number—an “anchor”—which they then adjust up or down to
reflect subsequent information and analysis
• Regardless of how the initial anchor was chosen, people tend to
adjust their anchors insufficiently and produce end approximations
that are, consequently, biased
• Related to conservatism bias

32
Anchoring …
Real estate appraisal: Two groups of professional real-estate
agents were shown the same house. One group was given a
list price of $65,900, the other $83,900. Their average
appraisals were $67,811 and $75,190, respectively.
• When asked to explain their decisions, less than 25% mentioned
listing price as one of the factors

Do you anchor on market price while doing a DCF?

33
Anchoring …
Consequences
 Holding onto a stock to attain a price that you are anchored
to, such as the purchase price or a high-water mark
(instead of rational analysis)
 Making a market or security forecast anchored to last year’s
market levels or ending securities prices

34
Mental accounting
A process to code, categorize, and evaluate economic
outcomes by grouping their assets into any number of non-
fungible (non-interchangeable) mental accounts
• Segregating investments by source of funds
▪ Bonuses, salaries, etc. being invested in different accounts or managed
separately

35
Mental accounting …
Imagine that you bought a ticket to a hit Broadway play. At
the theater you realize that you have lost your ticket which
cost $250. Do you spend another $250 to see the
performance?

Now imagine the same scenario, but you are planning to buy
the $250 ticket when you arrive. At the box office, you
realize that you have lost $250 somewhere in the parking lot.
Still, you have more than enough to buy the ticket. Do you?

36
Mental accounting …
Imagine that you go a store to buy a lamp which sells for
$100. At the store, you discover that the same lamp is on
sale for $75 at a branch of the store five blocks away. Do you
go to the other branch to get the lower price?

Now imagine that you go the same store to buy a dining

room set which sells for $1,775. At the store you discover
that you can buy the same table and chairs for $1,750 at a
branch of the store five blocks away. Do you go to the other
branch to get the lower price?

37
Mental accounting …
Mentally account for money as too sacred or special to
become too conservative with it
• Retirement money: 401(k), 403(b), 457 plans.
• $10K saved will grow to only $43K in bonds @5% but $174K in
stocks @10% over a period of 30 years
• Mentally accounting for retirement money as too special could
mean that you haven’t saved enough for your retirement

Retirement portfolio treated as MC Hammer portfolio!

38
Mental accounting …
Mental accounting can cause investors to irrationally
distinguish between returns derived from income and those
derived from capital appreciation
• Can cause some investors to chase income streams and unwittingly
erode principal in the process

39
Framing
Making skewed decisions based on how a question is framed

40
Framing …
Imagine that you are the commander of an army, threatened with
a superior enemy force. Your staff say that your soldiers will be
caught in an ambush in which 600 soldiers will die unless you lead
them to safety by one of two available routes. If you take route A,
200 soldiers will be saved. If you take route B, there is a one-
third chance that 600 soldiers will be saved and a two-thirds
chance that none will be saved? Which route do you take?

Imagine that once again you are the commander of an army,

threatened with a superior enemy force. Once again, your staff
tells you that if you take route A, 400 soldiers will die. If you take
route B, there is a one-third chance that no soldier will die and a
two-thirds chance that 600 soldiers will perish. Which route do
you choose? 41
Framing …
Choose a portfolio

42
Framing …
Consequences
 Misidentifying risk tolerance
 Choosing sub-optimal investments
 Focus on short-term price fluctuations

43
Availability
Estimating the probability of an outcome based on how easily
it comes to mind

Consequences
 Making choices based on advertising or reputation
 Limited range of investments are considered, which leads to
an insufficiently diversified portfolio and inappropriate asset
allocation

44
Availability …
“All accidents”
and “all disease”
the same; in
reality 1:16.

Most overstated:
 Rare but
spectacular

Most
understated: “Rare” “Common”

 Common,
mostly non-fatal Slovic, Fischhoff, Lichtenstein (1982)
45
Availability …
A. Estimate the percentage of words in the English language
that begin with the letter “a”
B. Estimate the percentage of words in the English language
that have the letter “a” as their third letter

Most people believe A (true 6%) is more likely than B (true

9%)
• Easier to retrieve first letter words from memory

46
Availability …

47
Availability …
Buying decision: Almost 5,000 stocks to choose from (U.S.).
Seek among those which easily come to mind. E.g., large
price changes

30

25
Percent Order Imbalance

20
Large Discount Brokerage

15 Small Discount Brokerage

10 Large Retail Brokerage

0
1a 1b 2 3 4 5 6 7 8 9 10a 10b
-5

Partitions of Stocks Sorted on Previous Day's Return

Barber and Odean (2001) 48

Emotional biases
1. Loss aversion
• Disposition effect
2. Overconfidence
3. Self-control
4. Status quo
5. Endowment
6. Regret-aversion

49
Loss aversion
Making increasingly risky bets in order to avoid suffering or
recognizing losses

When comparing absolute values, utility derived from a gain

is much lower than the utility given up with an equivalent
loss

50
Loss aversion …
Imagine that you have just been given $1,000 and have
been asked to choose between two options. Option A:
Guaranteed additional $500. Option B: heads you get
additional $1000, tails gain nothing?

Now imagine that you have been given $2000 and have to
choose between two options. Option A: Guaranteed losing
$500. Option B: heads you lose $1000, tails lose nothing?

51
Loss aversion …
Reference point: Investor A owns a block of share which she
originally bought at $100 per share. Investor B owns a block
of share of the same stock for which she paid $200 per
share. The value of stock was $160 yesterday and today it
dropped to $150 per share. Who do you think is more upset?

52
Loss aversion …

Disposition effect: Holding (not selling) of losers too long and

selling (not holding) of winners too quickly

53
Loss aversion …
People are pre-disposed to “get-even-itis”
• Have difficulty in making peace with their losses
• Disposition effect is the propensity to lock in sure gains than to lock
in sure loss

54
Loss aversion: Disposition effect

55
Loss aversion …
Individuals are more likely to sell stocks that have risen in
price rather than those that have fallen in price (Odean,
2001)
• Stock that is up in value is 70% more likely to be sold than a stock
that is down
• A losing stock is held for a median of 124 days while a winning
stock is held for a median of 102 days
• Stocks that investors sold outperform the stocks that they held by
3.4% over the next 12 months

56
Loss aversion …
Coval and Shumway (2005) investigate morning and
afternoon trades of 426 traders of CBOT
• Assume significantly more risk in the afternoon trading following
morning losses than gains

Locke and Mann (1999) find that CME traders also display
disposition effect/loss aversion
• Best traders are the ones who are least loss-averse (sold their
losers and rode their winners)

57
Overconfidence
Believing that you have superior knowledge, abilities, and
access to information (illusion of knowledge)

Intensified when combined with self-attribution bias

• Taking credit for successes and assigning responsibility for failures

58
Overconfidence …
Prediction overconfidence
 Give high and low estimates for the weight in pounds of an
empty Boeing 747 aircraft. Choose numbers far apart to be
90% certain that the true answer lies in between.

Certainty overconfidence
 How good a driver are you? Compared to the drivers you
encounter on the road, are you above-average, average, or
below-average?

59
Overconfidence …
May have biological evolutionary roots

May be part of human condition

• Five to six times as many optimistic adjectives as pessimistic
adjectives in virtually all languages

60
Overconfidence …
Individuals who trade the most frequently post exceptionally
poor investment results

Source: Brad Barber and Terrance Odean, 2000, “Trading Is Hazardous to Your Wealth: The Common Stock
Investment Performance of Individual Investors,” Journal of Finance 55, 773–806.
61
Overconfidence …
Barber and Odean (2001)
 Performance of stocks picked by men and women was
about the same

 Men traded 45% more than women

• And chose stocks in smaller companies, higher price-to-book, and
higher betas
• Earned 1.4% less on risk adjusted basis

 Single men traded 67% more and earned 3.5% less on a

risk-adjusted basis

62
Overconfidence …
Day traders abandoned regular jobs to trade full time from
their personal computers

Most online traders were between 25 and 43 years old

75% of online traders were men; most active cohort was in

30-34 age group

63
Overconfidence …
Mitigation

64
Self-control
Inability to delay gratification

Failing to act in pursuit of your long-term financial objectives

due to a lack of self-discipline

Failing to save for retirement

65
Self-control …
Hyperbolic discounting: Tendency to prefer small payoffs now
than large payoffs in the future, where the tendency
increases closer to the present both payoffs are

66
Self-control …
Consequences
 Failing to save for retirement
 Taking on too much risk in an attempt to catch up
 Undiversified portfolio (e.g., too many equities, etc.)

67
Status quo
Doing nothing rather than making optimal investment
decisions

Consequences
 Failing to explore new opportunities, specifically when a
change might the best decision for the portfolio
 Maintaining a portfolio that has drifted away from its
optimal allocation

68
Endowment
Valuing assets more when you own them than when you
don’t

69
Endowment …
Imagine that you recently found a ticket to the Final Four.
You very much want to go. Now a stranger offers to buy your
precious ticket. What is the smallest amount for which you
would sell?

Now imagine that you don’t have a ticket to the event, but
you really want one. How much would you be willing to pay?

70
Endowment …
Investors hold onto securities that they have inherited,
regardless of whether retaining these securities makes sense

Investors hold onto securities that they have purchased

because they are familiar with these, even though this may
not make rational sense

71
Endowment …
Fail to contribute to your pension plan
• Typically employers match 50¢ to each dollar of your contribution
up to 6% of your salary
• 12 million people choose not to accept this free money
▪ Costs employees $6 billion a year in missed employer matches
• Overvalue what they have today (today’s salary) and fail to value
what they could have

72
Regret aversion
When past investment decisions - notably the regret
associated with them - have irrational influence over future
investment decisions

Being afraid to make a mistake because of regret associated

with past decisions

Markowitz had a 50-50 stock-bond portfolio “I visualized my

grief if the stock market went way up and I wasn’t in it—or it
went way down and I was completely in it. My intention was
to minimize my future regret.”

73
Regret aversion …
Consequences
 Overly-conservative investing
 Investing in the familiar
 Herding behavior
 Hanging on to a losing investment in order to avoid
realizing a loss

74
Skepticism
Don’t people make fewer mistakes when given proper
incentives?

Don’t people learn quickly and stop making mistakes?

Don’t people with expertise in some areas make fewer

mistakes?

75
Skepticism …
Behavioral biases cost you money but reflect psychological
tendencies that bring benefits in other areas
• Tendency to weigh losses more than gains is doubtless useful from
evolutionary standpoint
• Tendency to set up mental accounts can serve you well, for
instance in saving for future

Many behavioral biases appear to conflict with each other

• People routinely overestimate their own abilities
• People blindly follow the actions of others

76
Practical advice
Once a behavioral bias has been identified, it may be
possible to either moderate the (cognitive) bias or adapt to
the (emotional) bias so that the resulting financial decisions
more closely match the rational financial decisions assumed
by traditional finance

77
Interest rates

1
Present value formula
Present value is the discounted value of future cash flows

Assume that the project is risk-free

Then discount all the cash flows at the risk-free rate:

𝑇𝑇
𝐶𝐶𝐹𝐹 (𝑡𝑡) 𝐶𝐶𝐹𝐹 (1) 𝐶𝐶𝐹𝐹 (2) 𝐶𝐶𝐹𝐹 (3)
𝑃𝑃𝑃𝑃 = � 𝑡𝑡
= 1
+ 2
+ 3
+⋯
1 + 𝑟𝑟 1 + 𝑟𝑟 1 + 𝑟𝑟 1 + 𝑟𝑟
𝑡𝑡=1

Spot rates 2
Different risk-free rates
Relax the assumption of constant risk-free rate. Assume that
(1)
the risk-free rate for the first year (𝑟𝑟0 ) is not the same as
(2)
that for the second year (𝑟𝑟0 ) and so on:

𝑇𝑇
𝐶𝐶𝐹𝐹 (𝑡𝑡) 𝐶𝐶𝐹𝐹 (1) 𝐶𝐶𝐹𝐹 (2) 𝐶𝐶𝐹𝐹 (3)
𝑃𝑃𝑃𝑃0 = � = + + +⋯
(𝑡𝑡) 𝑡𝑡 (1) 1 (2) 2 (3) 3
𝑡𝑡=1 1+ 𝑟𝑟0 1+ 𝑟𝑟0 1+ 𝑟𝑟0 1+ 𝑟𝑟0

(𝑡𝑡)
Notation 𝑟𝑟0 :subscript 0 means that these are rates today,
superscript (𝑡𝑡) means the ‘maturity’ for the rate

Spot rates 3
Spot rates
Risk-free rates for different years are also called as spot
rates

(1)
Interest rate for the first year (𝑟𝑟0 ) is the spot rate for the
first year or 1-year spot rate

(3)
Interest rate for the third year (𝑟𝑟0 ) is the spot rate for the
third year or 3-year spot rate

Spot rates 4
Spot rates …

0 1 2 3 4 5

(1)
𝑟𝑟0
(2)
𝑟𝑟0
(4)
𝑟𝑟0

Spot rate for 𝑇𝑇-years is valid only for the cash-flow occurring
at 𝑇𝑇 years
(3)
Do not use to discount everything from now to the end of
𝑟𝑟0
the third year
(3)
Perhaps, 𝑟𝑟0 should be called as third-year spot rate instead
of three-year spot rate
Spot rates 5
Spot rates …
Spot rates are known today
(3)
• Thus, 𝑟𝑟0 is not the rate that you would know only at the end of the
third year
• Rather it is the rate today that you will use to discount cash flow at
the end of the third year
▪ The subscript 0 is to remind us that these are rates known today

Annualized for comparison purposes

Spot rates are different for different maturities

Spot rates 6
Discount bonds
Bonds that do not have a coupon

Also called as discount bonds (Why?)

Completely specified by
• Maturity
• Face value (principal)

We will denote the prices of discount bonds by symbol 𝐵𝐵

(3)
• 𝐵𝐵0 is the price today of a 3-year zero-coupon bond

Bond pricing 7
Discount bond pricing
Consider a discount bond with face amount $100 that will be
paid in 5 years. Suppose that the five-year spot rate is 3.3%.
What should be the price of the discount bond?

(5) $100
𝐵𝐵0 = 5
= $85.0 < $100
1 + 3.3%

Bond pricing 8
Discount bond pricing …
In general, if the face value of 𝑇𝑇-years to maturity zero-
(𝑇𝑇)
coupon bond is 𝐹𝐹 and the 𝑇𝑇-year spot rate is 𝑟𝑟0 , then

(𝑇𝑇) 𝐹𝐹
𝐵𝐵0 =
(𝑇𝑇) 𝑇𝑇
1+ 𝑟𝑟0

Bond pricing 9
Coupon bonds
Coupon bonds return a face or principal amount after 𝑇𝑇
periods along with coupons at regular periods in between

Bond pricing 10
Coupon bonds pricing
Suppose that you are given the option to purchase a 5-year
coupon bond with annual coupon rate of 7% and a face
amount of $100. Also you know the following information for
spot rates
Year 1 2 3 4 5
Spot rate 2.04% 2.60% 2.82% 2.96% 3.30%

What is the price of the coupon bond?

Bond pricing 11
Coupon bonds pricing …

Price
7 7 7 7
= 1
+ 2
+ 3
+ 4
1 + 2.04% 1 + 2.60% 1 + 2.82% 1 + 2.96%
107
+ 5
= $117.1 > $100
1 + 3.30%

Bond pricing 12
Coupon bonds pricing …
You need information about spot rates to calculate the prices
of coupon-bonds

Coupon rate is used only for determining the cash flows

(numerator)

Do not use the coupon rate for discounting (denominator)

Bond pricing 13
Coupon bonds pricing …
In general, the price of a coupon bond is:

𝑇𝑇
𝐶𝐶 𝐹𝐹
𝑃𝑃0 = � +
(𝑡𝑡) 𝑡𝑡 (𝑇𝑇) 𝑇𝑇
𝑡𝑡=1 1+ 𝑟𝑟0 1+ 𝑟𝑟0

where 𝐶𝐶’s are the coupons, 𝐹𝐹 is the face value, and 𝑟𝑟’s are
the spot rates

(𝑡𝑡)
Price of a zero-coupon bond with face value of $100 is 𝐵𝐵0 =
(𝑡𝑡) 𝑡𝑡
100� 1 + 𝑟𝑟0
Bond pricing 14
Coupon bonds pricing …
Combining the above two equations, we get:
𝑇𝑇
1 1
𝑃𝑃0 = � 𝐶𝐶 × + 𝐹𝐹 ×
(𝑡𝑡) 𝑡𝑡 (𝑇𝑇) 𝑇𝑇
𝑡𝑡=1 1+ 𝑟𝑟0 1+ 𝑟𝑟0

(𝑡𝑡) (𝑇𝑇)
𝐵𝐵0 1 𝐵𝐵0 1
= (𝑡𝑡) 𝑡𝑡
and = (𝑇𝑇) 𝑇𝑇
100 1+𝑟𝑟0 100
1+𝑟𝑟0

𝑇𝑇 (𝑡𝑡) (𝑇𝑇)
𝐵𝐵0 𝐵𝐵0
𝑃𝑃0 = � 𝐶𝐶 × + 𝐹𝐹 ×
100 100
𝑡𝑡=1
Bond pricing 15
Coupon bonds pricing …
(𝑡𝑡)
Think of 𝐵𝐵0 /100 as the present value factor
𝑇𝑇 (𝑡𝑡) (𝑇𝑇)
𝐵𝐵0 𝐵𝐵0
𝑃𝑃0 = � 𝐶𝐶 × + 𝐹𝐹 ×
100 100
𝑡𝑡=1

is the same as:

𝑇𝑇

Price = � 𝐶𝐶 × Present value factor for year 𝑡𝑡

𝑡𝑡=1
+𝐹𝐹 × Present value factor for year 𝑇𝑇

Bond pricing 16
Coupon bonds – Bootstrapping
Suppose that there are no discount bonds trading with
exactly one and two years to maturity, but there are coupon
bonds with these maturities trading. Can we infer the spot
(1) (2)
rates 𝑟𝑟0 and 𝑟𝑟0 ?
Years to Maturity 1 2
Face value 100 100
Coupon Rate 5% 8%
Price 99.75 104.80

Bond pricing 17
Coupon bonds – Bootstrapping …
Price of 1-year coupon bond
5 100
99.75 = (1)
+ (1)
(1 + 𝑟𝑟0 ) (1 + 𝑟𝑟0 )

105
99.75 = (1)
(1 + 𝑟𝑟0 )

(1) 105
𝑟𝑟0 = − 1 = 5.26%
99.75

Bond pricing 18
Coupon bonds – Bootstrapping …
Price of 2-year bond:
8 8 100
104.80 = 1 + +
1 (2) 2 (2) 2
1 + 𝑟𝑟0 1+ 𝑟𝑟0 1+ 𝑟𝑟0
8 108 8 108
= 1+ 2 = +
1 (2) (1 + 5.26%) (2) 2
1 + 𝑟𝑟0 1 + 𝑟𝑟0 1+ 𝑟𝑟0

(2) 108
𝑟𝑟0 = − 1 = 5.41%
104.8 − 7.6

Bond pricing 19
YTM
Yield to maturity (YTM) is the IRR of the bond

YTM 20
YTM on a discount bond
By definition

(𝑇𝑇) 𝐹𝐹 𝐹𝐹 𝐹𝐹
𝐵𝐵0 = = =
(𝑇𝑇) 𝑇𝑇 1 + 𝐼𝐼𝐼𝐼𝐼𝐼 𝑇𝑇 (𝑇𝑇) 𝑇𝑇
1+ 𝑟𝑟0 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0

YTM on a zero-coupon bond is the same as the spot rate

(𝑇𝑇)
• Call the YTM on a 𝑇𝑇-year zero-coupon bond as 𝑌𝑌𝑇𝑇𝑀𝑀0
• Superscript (𝑇𝑇) is to remind us that this is the YTM for a zero-
coupon bond with maturity of 𝑇𝑇 years

YTM 21
YTM on coupon bonds
Consider the previous 5-year coupon bond with annual
coupon rate of 7% and a face amount of $100 and price of
$117.1. What is the YTM on this bond?

It turns out that

117.1
7 7 7 7
= 1
+ 2
+ 3
+ 4
1 + 3.24% 1 + 3.24% 1 + 3.24% 1 + 3.24%
107
+
1 + 3.24% 5

YTM 22
YTM on coupon bonds …
YTM on the coupon bond is 3.24%
• Lower than the coupon rate

In general:
𝑇𝑇
1 1
𝑃𝑃0 = � 𝐶𝐶 × + 𝐹𝐹 ×
(𝑡𝑡) 𝑡𝑡 (𝑇𝑇) 𝑇𝑇
𝑡𝑡=1 1+ 𝑟𝑟0 1+ 𝑟𝑟0
𝑇𝑇
1 1
= � 𝐶𝐶 × + 𝐹𝐹 ×
(𝑇𝑇) 𝑡𝑡 (𝑇𝑇) 𝑇𝑇
𝑡𝑡=1 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0

YTM 23
YTM on coupon bonds …
YTM on a coupon bond provides no information about the
spot rates
• Is a weighted average of the spot rates

(𝑇𝑇)
𝑌𝑌𝑇𝑇𝑀𝑀0 on a 𝑇𝑇-year coupon bond
≠
spot rate(s)

YTM 24
Terminology is confusing!!
We talk about YTM for both the discount bond and coupon
bond
(𝑇𝑇)
𝑌𝑌𝑇𝑇𝑀𝑀0on a 𝑇𝑇-year coupon bond ≠
spot rate(s)
(𝑇𝑇)
𝑌𝑌𝑇𝑇𝑀𝑀0 on a 𝑇𝑇-year zero-coupon
(𝑇𝑇)
bond ≡ spot rate (𝑟𝑟0 ) for 𝑇𝑇-years

Should be more precise and say that the “YTM for a coupon
bond” or “YTM for a discount bond” instead of just “YTM”

YTM 25
YTM on coupon bonds …
To calculate the YTM of a coupon bond, we do not need the spot
rates, need only the price

Cannot go back from the YTM (or price) to calculate the individual
spot rates

YTM on two coupon bonds with the same maturity may not even
be the same if the bonds have different face values and coupon
rates

E.g., using the previous spot rates, YTM of a bond with coupon
rate of 1% (10%) is 3.29% (3.21%)
• Why does the bond with lower coupon have higher YTM?

YTM 26
Bond price calculation vs. YTM
For discount bonds

(𝑇𝑇) 𝐹𝐹 𝐹𝐹
𝐵𝐵0 = =
(𝑇𝑇) 𝑇𝑇 (𝑇𝑇) 𝑇𝑇
1+ 𝑟𝑟0 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0

YTM 27
Bond price calculation vs. YTM …
For coupon bonds

𝑇𝑇
1 1
𝑃𝑃0 = � 𝐶𝐶 × + 𝐹𝐹 ×
(𝑡𝑡) 𝑡𝑡 (𝑇𝑇) 𝑇𝑇
𝑡𝑡=1 1+ 𝑟𝑟0 1+ 𝑟𝑟0
𝑇𝑇
1 1
= � 𝐶𝐶 × + 𝐹𝐹 ×
(𝑇𝑇) 𝑡𝑡 (𝑇𝑇) 𝑇𝑇
𝑡𝑡=1 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0

YTM 28
YTM and coupon rate
Suppose the coupon rate is 𝑐𝑐, so the coupon amount is 𝐶𝐶 =
𝑐𝑐 × 𝐹𝐹

𝑇𝑇
1 𝐹𝐹
𝑃𝑃0 = � 𝐶𝐶 × +
(𝑇𝑇) 𝑡𝑡 (𝑇𝑇) 𝑇𝑇
𝑡𝑡=1 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0

𝑐𝑐𝐹𝐹 1 𝐹𝐹
= 1− +
(𝑇𝑇)
𝑌𝑌𝑇𝑇𝑀𝑀0 (𝑇𝑇) 𝑇𝑇 (𝑇𝑇) 𝑇𝑇
1+ 𝑌𝑌𝑇𝑇𝑀𝑀0 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0

YTM 29
YTM and coupon rate: Par value
If the bond is trading at par, this means that price is equal to
face value, 𝑃𝑃0 = 𝐹𝐹

𝑐𝑐𝐹𝐹 1 𝐹𝐹
𝑃𝑃0 = 1− +
(𝑇𝑇)
𝑌𝑌𝑇𝑇𝑀𝑀0 (𝑇𝑇) 𝑇𝑇 (𝑇𝑇) 𝑇𝑇
1+ 𝑌𝑌𝑇𝑇𝑀𝑀0 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0

𝑐𝑐 1 1
1= 1− +
(𝑇𝑇)
𝑌𝑌𝑇𝑇𝑀𝑀0 (𝑇𝑇) 𝑇𝑇 (𝑇𝑇) 𝑇𝑇
1+ 𝑌𝑌𝑇𝑇𝑀𝑀0 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0

(𝑇𝑇)
𝑐𝑐 = 𝑌𝑌𝑇𝑇𝑀𝑀0
YTM 30
YTM and coupon rate …
If coupon bond is trading at par, its coupon rate is equal to its
YTM
(𝑇𝑇)
𝑃𝑃0 = 𝐹𝐹 ⇔ 𝑐𝑐 = 𝑌𝑌𝑇𝑇𝑀𝑀0

It is easy to see (just go through the same calculations but with

inequality sign) that:

For a coupon bond that trades at discount

(𝑇𝑇)
𝑃𝑃0 < 𝐹𝐹 ⇔ 𝑐𝑐 < 𝑌𝑌𝑇𝑇𝑀𝑀0

For a coupon bond trades at premium

(𝑇𝑇)
𝑃𝑃0 > 𝐹𝐹 ⇔ 𝑐𝑐 > 𝑌𝑌𝑇𝑇𝑀𝑀0

YTM 31
YTM and perpetual bonds
If 𝑇𝑇 → ∞ and we have a perpetual bond, then

𝑐𝑐𝐹𝐹 1 𝐹𝐹 𝑐𝑐𝐹𝐹
𝑃𝑃0 = 1− + →
(∞)
𝑌𝑌𝑇𝑇𝑀𝑀0 (∞) 𝑇𝑇 (∞) 𝑇𝑇 (∞)
𝑌𝑌𝑇𝑇𝑀𝑀0
1+ 𝑌𝑌𝑇𝑇𝑀𝑀0 1+ 𝑌𝑌𝑇𝑇𝑀𝑀0

(∞)
The 𝑌𝑌𝑇𝑇𝑀𝑀0 of the perpetual bond equals 𝑐𝑐𝐹𝐹/𝑃𝑃0

YTM 32
Current yield
For a coupon bond the ratio 𝑐𝑐𝐹𝐹/𝑃𝑃0 is called current yield

For a coupon bond trading at par the current yield is equal to

YTM

For a very long bond the current yield is not too bad an
approximation of the YTM too

YTM 33
Holding period returns (HPR)
Do not confuse YTM with HPR
(1) (2)
If 𝑟𝑟0 = 2.04%, 𝑟𝑟0 = 2.60%, then price of a 2-year 5% coupon
(2) 5 105
bond today is 𝑃𝑃0 = + = 104.65
1.0204 1.02602
Holding period return depends on next year’s 1-year spot
rate

Next year Next year HPR ℎ𝑝𝑝𝑟𝑟

�1 =
(1) (1)
rate 𝑟𝑟1̃ price 𝑃𝑃�1
1
𝑃𝑃�1
2 −1
𝑃𝑃0

2.04% 102.90 3.11%

2.60% 102.34 2.57%
3.16% 101.78 2.04%
3.26% 101.69 1.95%
YTM 34
YTM and HPR
YTM HPR
 Is the average return if the  Is the rate of return over a
bond is held to maturity particular investment period

 Depends on coupon rate,  Depends on the bond’s price

maturity, and par value at the end of the holding
period, an unknown future
value

 Is readily observable  Can only be forecasted

YTM 35
Yield curve
YTMs on zero-coupon bonds with different maturities is called
the term structure of (spot) interest rates or yield curve

This is the yield curve for today and only today

(1)
• Rate for 1-year (𝑟𝑟0 ) is the rate applicable today for an investment
of 1-year
• Will change tomorrow: Should not interpret to mean that in one
(1) (1)
year, the rate will be 𝑟𝑟0 (it will be 𝑟𝑟1̃ )

Yield curve tomorrow is unknown today

Yield curve 36
Yield curve examples

Yield curve 37
Current yield curve – U.S.

Yield curve 38
Current yield curve – Switzerland

Yield curve 39
Yield curve under certainty
Suppose investors believed that 1-year interest rates in
future years will be as follows

Year Expected interest rate

(1)
0 (today) 𝑟𝑟0 2.04%

1 year from now 𝑟𝑟1̃

(1) 3.16%

2 years from now 𝑟𝑟2̃

(1) 3.26%

3 years from now 𝑟𝑟3̃

(1) 3.37%

4 years from now 𝑟𝑟4̃

(1) 4.71%

Yield curve 40
Yield curve under certainty …
(1) 100 (1)
𝐵𝐵0 = = 98.0; 𝑌𝑌𝑇𝑇𝑀𝑀0 = 2.04%
1.0204
Yield curve
(2) 100 (2) 3.5%
𝐵𝐵0 = = 95.0; 𝑌𝑌𝑇𝑇𝑀𝑀0 = 2.60%
1.0204 � 1.0316 3.0%

2.5%
(3) 100 (3)
𝐵𝐵0 = = 92.0; 𝑌𝑌𝑇𝑇𝑀𝑀0 = 2.82%
1.0204 � 1.0316 � 1.0326 2.0%

1.5%
(4) 100
𝐵𝐵0 = = 89.0; 1.0%
1.0204 � 1.0316 � 1.0326 � 1.0337
(4)
𝑌𝑌𝑇𝑇𝑀𝑀0 = 2.96% 0.5%

0.0%
(5) 100 1 2 3 4 5
𝐵𝐵0 = = 85.0; Years to Maturity
1.0204 � 1.0316 � 1.0326 � 1.0337 � 1.0471
(5)
𝑌𝑌𝑇𝑇𝑀𝑀0 = 3.30%

Yield curve 41
Yield curve shape (1)
In the previous example, the yield curve was upward sloping
because we knew that the 1-year rate in future was going to
be higher

In the real world with uncertainty, is it true that an upward

sloping yield curve signals that, in the future, interest rates
will be higher?

Yield curve 42
Yield curve shape (2)
Why might the market think that interest rates will be higher
in the future?

 Central bank’s actions may prompt the market to revise

upwards the interest rates
• Economy may be heating up
• Need to curb excessive speculation (prevent bubbles)

 Inflation will be higher in the future

Yield curve 43
Yield curve shape (3)
Alas, in the real world with uncertainty, it is not true that
today’s upward sloping yield curve is indeed followed by
higher interest rates next year

In fact, even inflation-adjusted (real) interest rates slope up

• In other words, upward sloping yield curve may not signal higher
inflation in the future

Yield curve 44
Detour: Real interest rates
Exact

(1+Real) = (1+Nominal)/(1+Inflation)
(1+Nominal) = (1+Real)×(1+Inflation)

Approximate

Real ≈ Nominal − Inflation

Nominal ≈ Real + Inflation

Yield curve 45
Detour: TIPS
Treasury Inflation-Protected Securities (TIPS) started in 1997
• Provide protection against inflation

Principal increases with inflation (measured by the Consumer

Price Index: CPI-Urban, Non-Seasonally-Adjusted with a 3-
month lag)
• When a TIPS matures, the investor is paid the inflation-adjusted
principal or original principal, whichever is greater

Yield curve 46
Detour: TIPS …
Nominal bond TIPS
 1-year T-bill paying coupon of 4% on  1-year TIPS paying coupon of 4% on
$100 $100
 Inflation at the end of year turns out to  Inflation at the end of year turns out to
be 2% be 2%

 You get paid  You get paid

• Principal: $100 • Principal: $100×(1+2%) = $102
• Coupon: 4%×$100=$4 • Coupon=4%×$102=$4.08

 Rate of return
 Rate of return • Nominal = $106.08/$100 − 1 ≈ 6%
• Nominal = $104/$100 − 1 = 4% • Real ≈ 6% − 2% = 4%
• Real ≈ 4% − 2% = 2%

Yield curve 47
Yield curve shape (4)
So, why does the yield curve slope up?

 There is a risk that the prices of government bonds, even

those not subject to default risk, will change (because of
changes in interest rates)

 This risk is higher for long-maturity bonds than that for

short-maturity bonds

 Yield curve usually slopes up to compensate investors for

this additional risk
Yield curve 48
Yield curve shape (5)
Define yield spread as the difference between a long-term
rate (say 5-year rate) and a short-term rate (say a 3-month
rate)

This yield spread has been historically successful in

predicting U.S. recessions

Yield curve 49
Yield curve and recessions …

Yield curve 50
Forward rates
Suppose that you observe the following pattern of discount
bond prices
Year (𝑡𝑡) 1 2 3 4 5
𝑌𝑌𝑇𝑇𝑀𝑀0
(𝑡𝑡) 2.04% 2.60% 2.82% 2.96% 3.30%
(𝑡𝑡)
𝐵𝐵0 98.0 95.0 92.0 89.0 85.0

Forward rates 51
Forward rates …
Forward rate between time 𝑇𝑇 and 𝑇𝑇 + 1 is given by:
(𝑇𝑇+1) 𝑇𝑇+1
1+ 𝑟𝑟0 (𝑇𝑇)
(𝑇𝑇→𝑇𝑇+1) 𝐵𝐵0
𝑓𝑓0 = −1= −1
(𝑇𝑇) 𝑇𝑇 𝐵𝐵0
(𝑇𝑇+1)
1+ 𝑟𝑟0

Forward rate between time 𝑇𝑇 and 𝑇𝑇 + 𝑁𝑁 is given by:

1
𝑇𝑇+𝑁𝑁 𝑁𝑁 1
𝑇𝑇+𝑁𝑁 𝑇𝑇 𝑁𝑁
(𝑇𝑇→𝑇𝑇+𝑁𝑁)
1 + 𝑟𝑟0 𝐵𝐵0
𝑓𝑓0 = 𝑇𝑇 −1= 𝑇𝑇+𝑁𝑁
−1
1 + 𝑟𝑟0
𝑇𝑇 𝐵𝐵0

Forward rates 52
Forward rates calculation

Year (𝑡𝑡) 1 2 3 4 5
𝑌𝑌𝑇𝑇𝑀𝑀0
(𝑡𝑡) 2.04% 2.60% 2.82% 2.96% 3.30%
(𝑡𝑡)
𝐵𝐵0 98.0 95.0 92.0 89.0 85.0
(𝑡𝑡−1→𝑡𝑡)
𝑓𝑓0 3.16% 3.26% 3.37% 4.71%

Do they remind you of something that we have seen before?

Forward rates 53
A short digression (towards confusion!)
Parameters of interest rates

 When are they determined?

 When do they start?

 For what period are they valid?

• When do they end?

Forward rates 54
Timing issues – Spot rates today

0 1 2 3 4 5

(1)
𝑟𝑟0
(2)
𝑟𝑟0
(4)
𝑟𝑟0
These rates are known today and start today

Determined from YTMs of zero-coupon bonds

(2) (2)
𝑌𝑌𝑇𝑇𝑀𝑀0 = 𝑟𝑟0

Annualized for comparison purposes

Forward rates 55
Timing issues – Forward rates today

0 1 2 3 4 5

(3→4)
(1→2) 𝑓𝑓0
𝑓𝑓0
(2→4)
𝑓𝑓0

These rates are known today and start in the future:

(4) 4
(2→4)
1+ 𝑟𝑟0
𝑓𝑓0 = 2 −1
(2)
1 + 𝑟𝑟0

Forward rates 56
Timing issues – Spot rates in future

0 1 2 3 4 5

(1)
𝑟𝑟1̃
(2)
𝑟𝑟1̃
Tilde (~) to indicate that they are unknown

These rates are unknown today (no formula!) and start in

future. These will become known in future. For example, in
(1) (1)
one year’s time we will call 𝑟𝑟�1 as 𝑟𝑟0

Forward rates 57
Interest rates varieties
Can you distinguish these kinds of interest rates?

 Spot rates
 YTM
 Coupon rates
 Forward rates
 Future spot rates

Forward rates 58
Interest rates varieties …
0 1 2 3 4 5

(1) (3→4)
𝑟𝑟0 𝑓𝑓0
(2)
𝑟𝑟0
(4)
𝑟𝑟0
(2→4)
𝑓𝑓0
(1)
𝑟𝑟1̃
(2)
𝑟𝑟1̃

2 (2→4) 2 4
2 4
1 + 𝑟𝑟0 1+ 𝑓𝑓0 = 1 + 𝑟𝑟0

Forward rates 59
What do the forward rates signal?
(1)
Annualized spot rate today for one year is 𝑟𝑟0
(2)
and that for two years is 𝑟𝑟0

(1→2) (1)
Is 𝑓𝑓0 = 𝐸𝐸0 𝑟𝑟�1 ?

Forward rates 60
Theories of term structure
1. Expectations Hypothesis
Forward rates are unbiased predictors of future spot rates
(1→2) (1)
𝑓𝑓0 = 𝐸𝐸0 𝑟𝑟�1

(2) (1)
Upward sloping yield curve means > 𝑟𝑟0 implying that
𝑟𝑟0
(1→2) (1) (1) (1)
𝑓𝑓0 > 𝑟𝑟0 implying, in turn, that 𝐸𝐸0 𝑟𝑟�1 > 𝑟𝑟0 (investors
expect future interest rates to rise)

Term structure 61
 Theories of term structure …
2. Risk premium hypothesis
Long-term bonds are riskier than short-term bonds making
term structure upward sloping
(1→2) (1)
𝑓𝑓0 = 𝐸𝐸0 𝑟𝑟�1 + Bond risk premium

3. Random walk hypothesis (RW)

Upward sloping yield curve signals only risk premium and has
NO expectations of rising interest rates
2026

(1) (1)
𝐸𝐸0 𝑟𝑟�1 = 𝑟𝑟0
2025

Ecoza(Co #C2025
Se Elisabet(no Arbitrage)]
EU =
3 15 %
En(Return) = 12 Fro02-12026)
Ross)lachos
(2)
=

04 %
.

3 16 %

/
.

-
.

Ro
RW
E2024(RCo)) Rose SERT
36 %
2 04 %
= Erw(Return) y
.

(1) a
By
1
year
from O
Term structure 62
 Yield curve examples

Panel A:
Constant Expected Short Rate
CURRY
Constant Liquidity Premium
STRATEGY

i E
MONY
loss

-
2024
next
year

I ↓

Panel B:
Declining Expected Short Rate
Increasing Liquidity Premiums
-

- -
GROUP 1 Group 2

63
Yield curve examples …

Panel C:
Declining Expected Short Rate
Constant Liquidity Premiums

Panel D:
Increasing Expected Short Rates
Increasing Liquidity Premiums

64
 Upward sloping yield curve
The yield curve reflects expectations of future interest rates,
but the forecasts are clouded by risk/liquidity premiums

An upward sloping curve could indicate:

• Rates are expected to rise
and/or
• Investors require large risk premiums to hold long term bonds

EXAMPLE Ro = 2
.
04 % = Ro = 2 60 %
.

2 /bond today (2024)

Buy 2-year
Sell this bond next year in 202s

what webmofe
I

E,
E( Return) =

. 16%
3 2 .
04%
= 96 93 .
= 98

293
04%983
_
1 = 2
.
.

95

Term structure 65
Inverted yield curve
Preferred habitat
• Beginning of expansion
▪ Many investment opportunities → Demand for loanable funds is high → Upward pressure on long-
term yields
▪ Central bank trying to stimulate the economy → chooses low short-term yields
▪ Yield curve is steeply sloped at the beginning of expansion
• End of expansion
▪ Few investment opportunities → Demand for loanable funds is weak → Downward pressure on
long-term yields
▪ Central bank has anti-inflation contractionary policies → chooses high short-term yields
▪ Yield curve is flat or inverted at the end of expansion
Expectations
• End of expansion
▪ Central bank has anti-inflation contractionary policies → chooses high short-term yields
• Even if the central bank does not actually increase the short-term rates, if it has historically done so, the market may flatten
the yield curve if it expects economic weakness ahead

Term structure 66
Interest rate risk
Interest rates fluctuate

Fixed-income instruments are risky even if coupon and

principal are guaranteed for Treasuries

Interest rate risk 67

Our old example
Suppose that you are given the option to purchase a 5-year
coupon bond with annual coupon rate of 7% and a face
amount of $100 and spot rates

Year 1 2 3 4 5
Spot rate 2.04% 2.60% 2.82% 2.96% 3.30%
Price
7 7 7 7
= 1
+ 2
+ 3
+ 4
1 + 2.04% 1 + 2.60% 1 + 2.82% 1 + 2.96%
107
+ 5
= $117.1
1 + 3.30%

Interest rate risk 68

Yield curve shifts
Increase all spot rates by 5bps
Price
7 7 7 7 107
= 1
+ 2
+ 3
+ 4
+ 5
1 + 2.09% 1 + 2.65% 1 + 2.87% 1 + 3.01% 1 + 3.35%
= $116.9

YTM also goes up by 5bps (from 3.24% to 3.30%)

116.9
7 7 7 7 107
= 1
+ 2
+ 3
+ 4
+ 5
1 + 3.30% 1 + 3.30% 1 + 3.30% 1 + 3.30% 1 + 3.30%

Interest rate risk 69

Interest rate risk – Some general rules
Prices and yields are inversely related

Prices are convex in YTM: when yields increase, prices

decline by less than the gain associated with yields decrease

Prices of long-term bonds are more sensitive than those of

short-term bonds

Prices of high-coupon bonds are less sensitive than those of

low-coupon bonds

Interest rate risk 70

Change in bond price

Interest rate risk 71

Duration as slope

Bond
value
Bond with high duration

P+
Bond with low duration

P
P−
-riskier

r−Δr r r+Δr
Interest rate
Interest rate risk 72
Duration as slope …
Bond
value

𝑑𝑑𝑃𝑃 Δ𝑃𝑃
P+ slope = ≈
𝑑𝑑𝑟𝑟 Δ𝑟𝑟

tangent line
P
P−

r−Δr r r+Δr
Interest rate

Interest rate risk 73

Duration definition
A measure of average maturity of a bond making many
payments

Duration measures the sensitivity of bond price to changes in

interest rates

Interest rate risk 74

Duration measures
Let 𝑦𝑦 be the yield to maturity on the bond under Duration
defined
slide 75 onwards

y = 1 + +m

consideration and let 𝑌𝑌 = 1 + 𝑦𝑦

y
~

Il Related
for riskiness
Measure of risk

TreasurementTrisk

𝑑𝑑𝑃𝑃/𝑃𝑃 Δ𝑃𝑃⁄𝑃𝑃 1+𝑦𝑦 Δ𝑃𝑃

(Macaulay) 𝐷𝐷 = − ≈ − ⁄ = −
𝑑𝑑𝑑𝑑/𝑑𝑑 Δ𝑑𝑑 𝑑𝑑 𝑃𝑃 Δ𝑦𝑦
𝐷𝐷 1 𝑑𝑑𝑃𝑃 1 Δ𝑃𝑃
(Modified) 𝐷𝐷∗ = = − ≈ −
1+𝑦𝑦 𝑃𝑃 𝑑𝑑𝑦𝑦 𝑃𝑃 Δ𝑦𝑦
1 𝑑𝑑𝑃𝑃 1 Δ𝑃𝑃
(Dollar) 𝐷𝐷𝑃𝑃𝐷𝐷 = − ≈ −
10,000 𝑑𝑑𝑦𝑦 10,000 Δ𝑦𝑦

Interest rate risk 75

Duration measures …
All three measures use the slope of the tangent line 𝑑𝑑𝑃𝑃/𝑑𝑑𝑦𝑦

𝐷𝐷𝑃𝑃𝐷𝐷 is the change in bond price when the YTM changes by

1bp

Percentage change in bond’s price is just the product of

modified duration and the change in bond’s YTM
Δ𝑃𝑃
≈ −𝐷𝐷∗ × Δ𝑦𝑦
𝑃𝑃

Interest rate risk 76

Duration …
For zero-coupon bonds
Duration = Maturity

For coupon bonds (or any bond portfolio) with price given by
𝑃𝑃 = ∑𝑇𝑇𝑡𝑡=1 𝐶𝐶𝐹𝐹𝑡𝑡 ⁄𝑌𝑌 𝑡𝑡 , duration is given by
𝑇𝑇
*
𝐶𝐶𝐹𝐹𝑡𝑡 ⁄𝑌𝑌 𝑡𝑡
𝐷𝐷 = � 𝑡𝑡 ×
𝑃𝑃
𝑡𝑡=1
𝑇𝑇
Present Value of 𝐶𝐶𝐹𝐹𝑡𝑡
= � Duration of 𝐶𝐶𝐹𝐹𝑡𝑡 ×
Total Value
𝑡𝑡=1
P
=.. G Interest rate risk
+...

-- )/p 77
Our old example once again
𝑌𝑌 = 1 + 𝑌𝑌𝑇𝑇𝑀𝑀 = 1.0323

𝐷𝐷 = [1 × 7⁄1.0323 + 2 × 7⁄1.03232
+3 × 7⁄1.03233 + 4 × 7⁄1.03234
+5 × 107/1.03235 ] ∕ 117.1 = 4.44

- =p .
D =
P

D =P (D
modified
duration

it derivative

APEDAY
APROXIMAMON
CONVEXITY Da Se

( 2 derivatives)
+

Interest rate risk 78

Duration of portfolio
Duration of any bond is the value-weighted average of
durations of the individual cash flows

Duration of the sum of two portfolios:

Value of A Value of B
D of portfolio = D of A × + D of B ×
Portfolio Value Portfolio Value

Interest rate risk 79

Duration rules
1. The duration of a zero-coupon bond equals its time to
maturity
2. Holding maturity constant, a bond’s duration is higher
when the coupon rate is lower
3. Holding the coupon rate constant, a bond’s duration
generally increases with its time to maturity
4. Holding other factors constant, the duration of a coupon
bond is higher when the bond’s yield to maturity is lower
5. The duration of a level perpetuity is equal to (1+y)/y

Interest rate risk 80

 Duration rules …

(SOLID BLUE)
10 years YTM 15 % corPon 15 %
BOND A
: :

duvation = 3 ,
8
CRECIVE MONEY EARLIER)

(DOTED BUE)
B to Yer 15/ componibl
BOND years :

has
higher
duration = Sio

Interest rate risk 81

Convexity
Duration is only an approximate measure of change in bond
prices
• Formula is valid only for small changes in yields

For large swings in yields, use convexity

Convexity is the second derivative of the bond price function

(slope of slope)

1 Δ2 𝑃𝑃
Convexity =
𝑃𝑃 Δ𝑦𝑦 2
Interest rate risk 82
Convexity …
Can be computed using the formula:

𝑇𝑇
1 2 + 𝑡𝑡 𝐶𝐶𝐹𝐹 ⁄𝑌𝑌 𝑡𝑡
Convexity = 2
� 𝑡𝑡 𝑡𝑡
𝑃𝑃 1 + 𝑦𝑦
𝑡𝑡=1

Interest rate risk 83

Our old example
𝑌𝑌 = 1 + 𝑌𝑌𝑇𝑇𝑀𝑀 = 1.0323

7 7
(12 + 1) × 2
+ (2 + 2) ×
1.0323 1.03232
2 7
+(3 + 3) ×
1.03233
2 7 2 107
+(4 + 4) × 4 + (5 + 5) × 5
𝐶𝐶 = 1.0323 1.0323 = 23.96
117.1 × 1.0323 2

Interest rate risk 84

Price changes
Accounting for convexity, the percentage change in price is
given by:

Δ𝑃𝑃
% change in price =
𝑃𝑃
1
≈ −Modified Duration × Δ𝑦𝑦 + Convexity × Δ𝑦𝑦 2
2

Interest rate risk 85

Old example
If the yield changes from 3.23% to 4.23%, the price declines
by 4.183%

Accounting for duration

Δ𝑃𝑃
≈ −4.30 × 1% = −4.300%
𝑃𝑃

Accounting for duration and convexity

Δ𝑃𝑃 1
≈ −4.30 × 1% + × 23.96 × 1%2 = −4.180%
𝑃𝑃 2

Interest rate risk 86

Price changes …
Price changes for a 8% coupon 30-year bond with an initial YTM of 8%

100%

80%

60%
% change in bond price

&
Exact
40%
D*
20% D* and C

Riskier 0%
3% 4% 5% 6% 7% 8% 9% 10% 11% 12% 13%
-20%

Da DB=
-40%
Ca +
C
-60% YTM

Interest rate risk 87

More on convexity
Convexity is considered desirable – Bonds with greater
curvature gain more in price when yields fall than they lose
when yields rise

If interest rates are volatile, this is an attractive asymmetry

However, comes at a cost

• Investors pay more and accept lower yields on more convex bonds

Interest rate risk 88

Beyond duration/convexity
Assumption so far has been parallel shifts in yield curve

More advanced techniques for price changes involve principal

component analysis
• Model risk factors
• Calculate sensitivity of bond to these risk factors
• Calculate change in bond price as the product of the sensitivities
and expected changes in risk factors

Interest rate risk 89

 14-11-2024

pi
return
Interest Rate variaties Rate of Forward Rate
I spot vate
today
Fo2S-20
(1)Spot Rates 2
year
today R &Osse
year spot
rate
,
(2) Coupon 1 pe2026 2024 202S
2026
org-A
=

"Future Interest Rate" 2024

(3) YTM the 1-year rale 1
year
D
from now
, ~(2026 EXPECTATOI
R 202s POSSIBLE
(4) Forward rates
↑
3 11 % Answers

iiiiii
.

=
D
(202s - 2026)
RANDOM
WALK (Ru) Fro
D
(s)
↑
posto
Ro(3)

(3-s
D

For -
TRUE

f(2025
+ 2026)
(1 + R)" =
(1 + R .)3(1 +(3) Is
224 E(20
Theories of term structure
1. Expectations Hypothesis
Forward rates are unbiased predictors of future spot rates
(1→2) (1)
𝑓𝑓0 = 𝐸𝐸0 𝑟𝑟�1

(2) (1)
Upward sloping yield curve means 𝑟𝑟0 > 𝑟𝑟0 implying that
(1→2) (1) (1) (1)
𝑓𝑓0 > 𝑟𝑟0 implying, in turn, that 𝐸𝐸0 𝑟𝑟�1 > 𝑟𝑟0 (investors
expect future interest rates to rise)

1
Theories of term structure …
2. Risk premium hypothesis
Long-term bonds are riskier than short-term bonds making
term structure upward sloping
(1→2) (1)
𝑓𝑓0 = 𝐸𝐸0 𝑟𝑟�1 + Bond risk premium

3. Random walk hypothesis (RW)

Upward sloping yield curve signals only risk premium and has
NO expectations of rising interest rates
(1) (1)
𝐸𝐸0 𝑟𝑟�1 = 𝑟𝑟0

2
Holding period returns
Buy a 2-year zero-coupon bond today and sell it one-year
later (when it becomes a 1-year bond)
(2) 1
Price today, 𝑃𝑃0 = (2) 2
1+𝑌𝑌0

(1) 1
Price next year, 𝑃𝑃1 =
� (1)
1+𝑌𝑌�1
(1) (2)
Return, 𝑅𝑅1 = 𝑃𝑃1 �𝑃𝑃0 − 1
� �
2 1
≈ 2𝑌𝑌0 − 𝑌𝑌�1
1→2 1 � (1)
≈ 𝐹𝐹0 + 𝑌𝑌0 − 𝑌𝑌1

3
Returns …
(1) (2)
On Dec 2020 𝑌𝑌2020 = 0.74% and 𝑌𝑌2020 = 1.14%

(1) (2)
This means that 𝑃𝑃2020 = $99.27 and 𝑃𝑃2020 = $97.76

Buy the 2-year bond today at $97.76

Next year 2021, this bond will become a 1-year bond. Sell it
(1)
next year for 𝑃𝑃�2021 , which is unknown today in 2020

4
Returns …

IF THEN AND
Future yield Future price Return
(1)
𝑌𝑌�2021
(1)
𝑃𝑃�2021 𝑅𝑅�2021
(RW) 0.740% 99.27 1.54%
1.140% 98.87 1.14%
(EH) 1.549% 98.47 (1)
0.73% = 𝑌𝑌2020
2.000% 98.04 0.29%

5
Return if EH
If EH is true, then interest rates will increase in the future

(1→2) (2) 1
𝐹𝐹0 = 2𝑌𝑌0 − 𝑌𝑌0 = 1.549%
� (1) (1→2)
𝐸𝐸2020 𝑌𝑌2021 = 𝐹𝐹2020 = 1.549%
(1)
𝑃𝑃�2021 = $98.47

The bond that you bought for $97.76 now trades for $98.47
giving a return of 0.73%

� 1→2 1 � (1) 1
𝑅𝑅2021 = 𝐹𝐹2020 + 𝑌𝑌2020 − 𝑌𝑌2021 = 𝑌𝑌2020
6
Return in EH …
If EH is true, you do not make any extra money by buying a
cheaper 2-year bond (vs. a 1-year bond)

Future increase in interest rates erodes the value of bond,

resulting in a rate of return that is the same as that on a 1-
year bond

7
Return if RW
If RW is true, then interest rates do not change in the future

� (1) 1
𝐸𝐸2020 𝑌𝑌2021 = 𝑌𝑌2020 = 0.74%
� (1)
𝑃𝑃2021 = $99.27

The bond that you bought for $97.76 now trades for $99.27
giving a return of 1.54%

1→2 1 (1) 1→2

𝑅𝑅�2021 = 𝐹𝐹2020 + 𝑌𝑌2020 − 𝑌𝑌�2021 = 𝐹𝐹2020

8
Return if RW – Roll-down return
We will call this the roll-down return

� 2 � 1 2 1 2 2 1
𝑅𝑅2021 = 2𝑌𝑌2020 − 𝑌𝑌2021 = 2𝑌𝑌2020 − 𝑌𝑌2020 = 𝑌𝑌2020 + 𝑌𝑌2020 − 𝑌𝑌2020

2
One earns the “yield” or the “carry”, 𝑌𝑌0 , plus the “roll-
2 1
down”, 𝑌𝑌0 − 𝑌𝑌0

9
Roll-down return
Expected return increases with maturity

This may not be an arbitrage as not only the expected return

increases but also the risk increases with maturity

Increase in risk is exactly compensated for by expected

return

10
Rolling down the yield curve
In an upward sloping yield curve and low-rate environment,
an opportunity exists to earn higher yields, while still
minimizing the risks to principal. The strategy, known as
“rolling down the yield curve”, involves the purchase of a
bond with a maturity in the higher yielding section of the
yield curve and selling the bond prior to maturity when it
reaches a lower yielding section

11
Yield curve strategy
Forward rates are good predictors of holding period returns
(not of future yield changes)

Thus, rolling-down the yield curve (RW) is quite a good

strategy compared to investing based on EH

12
Carry investing (1)
Beekhuizen, Duyvesteyn, Martens, and Zomerdijk (2018)
[Robeco] find that global curve carry factor has strong
performance that cannot be explained by other factors

13
Carry investing (2)

14
Appendix

Appendix 15
Holding period returns
Buy a 𝑁𝑁-year zero-coupon bond today (𝑡𝑡) and sell it one-year
later

(𝑁𝑁) 1
Price today, 𝑃𝑃𝑡𝑡 = (𝑁𝑁) 𝑁𝑁
1+𝑌𝑌𝑡𝑡

(𝑁𝑁−1) 1
Price next year, 𝑃𝑃𝑡𝑡+1 =
�
(𝑁𝑁−1) 𝑁𝑁−1
1+𝑌𝑌�𝑡𝑡+1

(𝑁𝑁) (𝑁𝑁−1) (𝑁𝑁)

Return, 𝑅𝑅� 𝑡𝑡+1 = 𝑃𝑃�𝑡𝑡+1 �𝑃𝑃𝑡𝑡 − 1

Appendix 16
Formulas with continuous compounding
(1) (2) (𝑛𝑛)
Yields are 𝑦𝑦𝑡𝑡 , 𝑦𝑦𝑡𝑡 , … , 𝑦𝑦𝑡𝑡

(1) (2) (𝑛𝑛)

Prices are 𝑝𝑝𝑡𝑡 , 𝑝𝑝𝑡𝑡 , … , 𝑝𝑝𝑡𝑡
(𝑛𝑛) (𝑛𝑛) (𝑛𝑛) (𝑛𝑛)
𝑝𝑝𝑡𝑡 = exp −𝑛𝑛𝑦𝑦𝑡𝑡 or 𝑦𝑦𝑡𝑡 = − ln 𝑝𝑝𝑡𝑡 ⁄𝑛𝑛

(𝑛𝑛−1→𝑛𝑛)
Forward rates are 𝑓𝑓𝑡𝑡
(𝑛𝑛−1→𝑛𝑛) (𝑛𝑛) (𝑛𝑛−1)
𝑓𝑓𝑡𝑡 = 𝑛𝑛𝑦𝑦𝑡𝑡 − (𝑛𝑛 − 1)𝑦𝑦𝑡𝑡
(𝑛𝑛) 𝑛𝑛 𝑛𝑛−1
= 𝑦𝑦𝑡𝑡 + 𝑛𝑛 − 1 𝑦𝑦𝑡𝑡 − 𝑦𝑦𝑡𝑡

Appendix 17
Formulas …
Holding period returns are

� (𝑛𝑛) (𝑛𝑛−1) (𝑛𝑛)

ℎ𝑝𝑝𝑟𝑟𝑡𝑡+1 ≡ ln 𝑝𝑝�𝑡𝑡+1 �𝑝𝑝𝑡𝑡
(𝑛𝑛) 𝑛𝑛−1
= 𝑛𝑛𝑦𝑦𝑡𝑡 − 𝑛𝑛 − 1 𝑦𝑦�𝑡𝑡+1
(𝑛𝑛−1→𝑛𝑛) 𝑛𝑛−1 (𝑛𝑛−1)
= 𝑓𝑓𝑡𝑡 − 𝑛𝑛 − 1 𝑦𝑦�𝑡𝑡+1 − 𝑦𝑦𝑡𝑡

Expected holding period returns are

� (𝑛𝑛) (𝑛𝑛−1→𝑛𝑛) 𝑛𝑛−1 (𝑛𝑛−1)

𝐸𝐸𝑡𝑡 ℎ𝑝𝑝𝑟𝑟𝑡𝑡+1 = 𝑓𝑓𝑡𝑡 − 𝑛𝑛 − 1 𝐸𝐸𝑡𝑡 𝑦𝑦�𝑡𝑡+1 − 𝑦𝑦𝑡𝑡

Appendix 18
Roll-down return
If the yield curve does not change, then

(𝑛𝑛−1) (𝑛𝑛−1)
𝐸𝐸𝑡𝑡 𝑦𝑦�𝑡𝑡+1 = 𝑦𝑦𝑡𝑡
� (𝑛𝑛) (𝑛𝑛−1→𝑛𝑛)
𝐸𝐸𝑡𝑡 ℎ𝑝𝑝𝑟𝑟𝑡𝑡+1 = 𝑓𝑓𝑡𝑡

(𝑛𝑛−1→𝑛𝑛) (𝑛𝑛)
We will call this the roll-down return. Since 𝑓𝑓𝑡𝑡 = 𝑦𝑦𝑡𝑡 +
𝑛𝑛 𝑛𝑛−1
𝑛𝑛 − 1 𝑦𝑦𝑡𝑡 − 𝑦𝑦𝑡𝑡 , we also see that one earns the “yield”,
𝑛𝑛 𝑛𝑛 𝑛𝑛−1
𝑦𝑦𝑡𝑡 , plus (a multiple of) the “roll-down”, 𝑦𝑦𝑡𝑡 − 𝑦𝑦𝑡𝑡

Appendix 19
Expected return
Expected holding period returns are
� (𝑛𝑛) (𝑛𝑛−1→𝑛𝑛) 𝑛𝑛−1 (𝑛𝑛−1)
𝐸𝐸𝑡𝑡 ℎ𝑝𝑝𝑟𝑟𝑡𝑡+1 = 𝑓𝑓𝑡𝑡 − 𝑛𝑛 − 1 𝐸𝐸𝑡𝑡 𝑦𝑦�𝑡𝑡+1 − 𝑦𝑦𝑡𝑡
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑑𝑑𝑅𝑅𝑑𝑑𝑛𝑛 𝑟𝑟𝑟𝑟𝑡𝑡𝑟𝑟𝑟𝑟𝑛𝑛 𝐶𝐶𝐶𝐶𝐶𝑛𝑛𝐶𝐶𝑟𝑟 𝑖𝑖𝑛𝑛 𝑦𝑦𝑖𝑖𝑟𝑟𝑅𝑅𝑑𝑑

Appendix 20
Expected return …
We can approximate the part of return due to yield changes
from the formula
Δ𝑃𝑃
% change in price =
𝑃𝑃
1
≈ −Modified Duration × Δ𝑦𝑦 + Convexity × Δ𝑦𝑦 2
2

Appendix 21
Expected return …
Combining, we get

(𝑛𝑛) (𝑛𝑛−1→𝑛𝑛) 1 ∗
�
𝐸𝐸𝑡𝑡 ℎ𝑝𝑝𝑟𝑟𝑡𝑡+1 ≈ 𝑓𝑓𝑡𝑡 + −𝐷𝐷∗ 𝐸𝐸𝑡𝑡 Δ𝑦𝑦 + 𝐶𝐶 𝑉𝑉𝑡𝑡 Δ𝑦𝑦
2
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑑𝑑𝑅𝑅𝑑𝑑𝑛𝑛 𝑟𝑟𝑟𝑟𝑡𝑡𝑟𝑟𝑟𝑟𝑛𝑛
𝑅𝑅𝑟𝑟𝑡𝑡𝑟𝑟𝑟𝑟𝑛𝑛 𝑑𝑑𝑟𝑟𝑟𝑟 𝑡𝑡𝑅𝑅 𝑑𝑑𝑟𝑟𝑟𝑟𝐶𝐶𝑡𝑡𝑖𝑖𝑅𝑅𝑛𝑛
𝐶𝐶𝑛𝑛𝑑𝑑 𝑐𝑐𝑅𝑅𝑛𝑛𝑐𝑐𝑟𝑟𝑐𝑐𝑖𝑖𝑡𝑡𝑦𝑦

Appendix 22
Carry
It is typical to think of funding the purchase of a long-term
bond with a short-term bond. In this case, it is appropriate to
think of excess returns (bond risk premium) and their
relation to “carry” costs

� (𝑛𝑛) (1)
𝐵𝐵𝑅𝑅𝑃𝑃 ≡ 𝐸𝐸𝑡𝑡 ℎ𝑝𝑝𝑟𝑟𝑡𝑡+1 − 𝑦𝑦𝑡𝑡 =
(𝑛𝑛−1→𝑛𝑛) (1) ∗
1 ∗
≈ 𝑓𝑓𝑡𝑡 − 𝑦𝑦𝑡𝑡 + −𝐷𝐷 𝐸𝐸𝑡𝑡 Δ𝑦𝑦 + 𝐶𝐶 𝑉𝑉𝑡𝑡 Δ𝑦𝑦
2
𝐶𝐶𝐶𝐶𝑟𝑟𝑟𝑟𝑦𝑦
𝑅𝑅𝑟𝑟𝑡𝑡𝑟𝑟𝑟𝑟𝑛𝑛 𝑑𝑑𝑟𝑟𝑟𝑟 𝑡𝑡𝑅𝑅 𝑑𝑑𝑟𝑟𝑟𝑟𝐶𝐶𝑡𝑡𝑖𝑖𝑅𝑅𝑛𝑛
𝐶𝐶𝑛𝑛𝑑𝑑 𝑐𝑐𝑅𝑅𝑛𝑛𝑐𝑐𝑟𝑟𝑐𝑐𝑖𝑖𝑡𝑡𝑦𝑦

Appendix 23
What if the EH was true?
In this case, all rates of return will equal the one-year return
� (𝑛𝑛) (1)
𝐸𝐸𝑡𝑡 ℎ𝑝𝑝𝑟𝑟𝑡𝑡+1 = 𝑦𝑦𝑡𝑡

This implies that

𝑛𝑛−1 (𝑛𝑛−1) 1 (𝑛𝑛−1→𝑛𝑛) (1)

𝐸𝐸𝑡𝑡 𝑦𝑦�𝑡𝑡+1 = 𝑦𝑦𝑡𝑡 + 𝑓𝑓𝑡𝑡 − 𝑦𝑦𝑡𝑡 or
𝑛𝑛−1

(𝑛𝑛−1→𝑛𝑛) (1) 𝑛𝑛−1 (𝑛𝑛−1)

𝑓𝑓𝑡𝑡 = 𝑦𝑦𝑡𝑡 + (𝑛𝑛 − 1)𝐸𝐸𝑡𝑡 𝑦𝑦�𝑡𝑡+1 − 𝑦𝑦𝑡𝑡

Appendix 24
EH and RW
EH says that carry costs are exactly equal to return due to
duration/convexity (there is no risk premium)
� (𝑛𝑛) 1
𝐸𝐸𝑡𝑡 ℎ𝑝𝑝𝑟𝑟𝑡𝑡+1 − 𝑦𝑦𝑡𝑡 = 0

RW says that expectations of yield changes are zero

� (𝑛𝑛) 1 𝑛𝑛−1→𝑛𝑛 1
𝐸𝐸𝑡𝑡 ℎ𝑝𝑝𝑟𝑟𝑡𝑡+1 − 𝑦𝑦𝑡𝑡 = 𝑓𝑓𝑡𝑡 − 𝑦𝑦𝑡𝑡

Appendix 25
Rolling down the yield curve
In an upward sloping yield curve and low rate environment,
an opportunity exists to earn higher yields, while still
minimizing the risks to principal. The strategy, known as
“rolling down the yield curve”, involves the purchase of a
bond with a maturity in the higher yielding section of the
yield curve and selling the bond prior to maturity when it
reaches a lower yielding section

Appendix 26
Example

𝑛𝑛 𝑦𝑦 𝑝𝑝 𝑓𝑓
1 1.00% 99.00
2 1.15% 97.73 1.30%
3 1.25% 96.32 1.45%
4 1.75% 93.24 3.25%
5 2.20% 89.58 4.00%
6 2.55% 85.81 4.30%
7 3.05% 80.78 6.05%
8 3.70% 74.38 8.25%
9 4.10% 69.14 7.30%
10 4.30% 65.05 6.10%
11 4.60% 60.29 7.60%
12 5.00% 54.88 9.40%
13 5.20% 50.86 7.60%

Appendix 27
Example …

Appendix 28
Roll-down return example
Last column gives the expected holding period return if the
yield curve does not change
• Forward rates are the 𝐸𝐸[ℎ𝑝𝑝𝑟𝑟]
▪ 𝐸𝐸[ℎ𝑝𝑝𝑟𝑟] from buying a 5-year bond today and selling it one-year later is
� (5) = 5𝑦𝑦 (5) − 4𝐸𝐸0 𝑦𝑦� 4 = 5𝑦𝑦 5 − 4𝑦𝑦 4 = 𝑓𝑓 (4,5) = 4.0%
𝐸𝐸0 ℎ𝑝𝑝𝑟𝑟
1 0 1 0 0 0

𝐸𝐸[ℎ𝑝𝑝𝑟𝑟] increases with maturity

• This may not be an arbitrage as not only the 𝐸𝐸[ℎ𝑝𝑝𝑟𝑟] increases but
also the risk increases with maturity
▪ Increase in risk is exactly compensated for by 𝐸𝐸[ℎ𝑝𝑝𝑟𝑟] [RW hypothesis]

Appendix 29
What if the interest rates increase?
4 (4)
If 𝑦𝑦�1 > 𝑦𝑦0 = 1.75%, then ℎ𝑝𝑝𝑟𝑟 from buying a 5-year bond today and selling it one-year
later is

� (5) = 5𝑦𝑦 (5) − 4𝑦𝑦� 4 = 5 ⋅ 2.2% − 4 ⋅ 2.0% = 3.0% if 𝑦𝑦� 4 = 2.0%

ℎ𝑝𝑝𝑟𝑟
1 0 1 1
� (5) = 5𝑦𝑦 (5) − 4𝑦𝑦� 4 = 5 ⋅ 2.2% − 4 ⋅ 2.25% = 2.0% if 𝑦𝑦� 4 = 2.25%
ℎ𝑝𝑝𝑟𝑟
1 0 1 1
� (5) = 5𝑦𝑦 (5) − 4𝑦𝑦� 4 = 5 ⋅ 2.2% − 4 ⋅ 2.5% = 1.0% if 𝑦𝑦� 4 = 2.5%
ℎ𝑝𝑝𝑟𝑟
1 0 1 1

Fixed income investors concerned with the possibility of rising interest rates and the
related principal loss, but also seeking more generous yields than those provided by
short term money market instruments, should consider the strategy of “rolling down
the yield curve.” The higher yields provide an element of protection, as it can help
offset losses from price declines that result from an increase in interest rates

Appendix 30
Example with EH
E(hpr) from buying a 5-year bond today and selling it one-
(1) (4→5)
year later is only 𝑦𝑦0 = 1.0% and not 𝑓𝑓0 = 4.0% calculated
earlier

• One-year later, the 4-year yield will be

4 (4) 1 (4→5) (1) 1
𝐸𝐸0 𝑦𝑦�1 = 𝑦𝑦0 + 𝑓𝑓0 − 𝑦𝑦0 = 1.75% + 4.0% − 1.0% = 2.5%
4 4
• The hpr will be
� (5) = 5𝑦𝑦 (5) − 4𝐸𝐸0 𝑦𝑦� 4 = 5 ⋅ 2.2% − 4 ⋅ 2.5% = 1.0%
𝐸𝐸0 ℎ𝑝𝑝𝑟𝑟
1 0 1

Yields rise in the future cutting the advantage of long-term

bonds
Appendix 31
EH vs. RW
Regress
(𝑛𝑛) 1 𝑛𝑛−1→𝑛𝑛 1
ℎ𝑝𝑝𝑟𝑟𝑡𝑡+1 − 𝑦𝑦𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽 𝑓𝑓𝑡𝑡 − 𝑦𝑦𝑡𝑡 + 𝜖𝜖𝑡𝑡+1

Null under EH: 𝛽𝛽 = 0 and RW: 𝛽𝛽 =1

Appendix 32
Corporate bonds

1
Credit ratings

CONFLICT Of

Te

-
-

SPECULATIVE

2
Corporate bonds’ returns
Quick look at history (1927 to 2019): Returns in excess of T-
bills

Mean StDev Sharpe

S&P500 8.0% 18.8% 0.42
Small-cap 12.0% 30.9% 0.39
Government 2.4% 8.5% 0.28
Corporate 2.8% 7.6% 0.35

3
More recent history

HighYield looks pretty good

4
Recent history …

From 1983 to 2019

Mean StDev Sharpe
S&P500 8.2% 14.6% 0.56
Small-cap 7.0% 20.4% 0.34
Government 3.0% 4.4% 0.68
Corporate 4.2% 5.5% 0.76
High Yield 5.3% 8.2% 0.65
HighYield is high yield for a reason!

5
Factors in bonds
We know about style investing in stocks
• Size, Value, Profitability, Investment, Momentum, …

Do similar styles/factors exist in bonds?

If so, how to construct these factors?

• From bonds, stocks, both, …

6
Bond factors from bonds
Houweling and Zundert (2017) [Robeco] construct factors
using bond characteristics
• Size
• Low-Risk (maturity, rating)
• Value (maturity, rating, 3-month change in spread)
• Momentum (6-month return on bond)

7
Bond factors from bonds …

8
Bond factors from bonds …
Long-short alphas of 1-2% on investment grade bonds and
5-8% on junk bonds

Long-only Sharpe ratios of 0.2-0.3 on investment grade and

0.4-0.6 on junk

Results robust to transaction costs and liquid bonds

9
Bond factors from bonds …
Can be combined with equity factor investing

10
Bond factors from stocks
Bektić, Wenzler, Wegener, Schiereck, and Spielmann (2019)
[Deka] construct factors using only stock characteristics
• Size, value, profitability, investment

11
Bond factors from stocks …

High yield

12
Bond factors from stocks …
Investment grade – U.S.

Investment grade - Europe

13
Bond factors from stocks (2)
Chordia, Goyal, Nozawa, Subrahmanyam, and Tong (2017)
consider a laundry list of factors from stocks and bonds
• Many stock factors do not work but some stock and bond factors do

14
Bond factors from stocks (2) …
Long only and all bonds
From 1983 to 2014
Mean StDev Sharpe
Corporate 4.3% 5.7% 0.75
High Yield 5.3% 8.5% 0.62
Stock Factors
‒ Momentum 6.9% 4.8% 1.45
‒ Reversal 9.4% 5.2% 1.82
‒ Profitability 7.6% 5.0% 1.54
‒ Investment 6.6% 4.6% 1.42
Bond Factors
‒ Momentum 8.7% 7.5% 1.16
‒ Reversal 13.3% 6.5% 2.05 15
Bond factors from stocks (2) …
Long only and HY bonds
From 1983 to 2014
Mean StDev Sharpe
Corporate 4.3% 5.7% 0.75
High Yield 5.3% 8.5% 0.62
Stock Factors
‒ Momentum 9.9% 5.6% 1.77
‒ Reversal 13.1% 6.1% 2.15
‒ Profitability 11.0% 10.2% 1.08
‒ Investment 8.5% 5.9% 1.43
Bond Factors
‒ Momentum 15.8% 19.1% 0.83
‒ Reversal 16.9% 7.3% 2.33 16
Bond factors from bonds/stocks
Israel, Palhares, and Richardson (2018) [AQR] construct
factors using bond and stock characteristics
• Carry (spread from bonds)
• Value (duration, rating, volatility from bonds)
• Momentum (6-month return on bond and stocks)
• Defensive (duration from bonds, leverage and profitability from
stocks)

17
Bond factors from bonds and stocks …

18
Bond factors from bonds and stocks …
-

-
factors
credit
:

↓
for lang
maturity
-

tima

Sharpe ratio of 2.2 for

combined strategy

19
Momentum in corporate bonds
Jostova, Nikolova, Philipov, and Stahel (2013) explore
whether past corporate bond returns forecast future
corporate bond returns

20
Momentum …

21
Information from options
Maybe option traders are more sophisticated in impounding
material information in options and bond traders react with a
delay

Cao, Goyal, Xiao, and Zhan (2021) find that changes in

implied volatility of options predict bond returns
• Monthly return of 0.6% (but monthly rebalancing needed)
• Similar predictability for trading CDS

22
Machine learning portfolios
Bali, Goyal, Huang, Jiang, and Wen (2021) use ML
techniques to construct portfolios

Long-short return in % per month

Investment Speculative
grade grade
OLS 0.18 (0.89) 0.20 (0.88)
PCA 0.35 (1.68) 0.41 (3.04)
PLS 0.42 (1.72) 0.53 (2.78)
LASSO 0.35 (2.01) 0.56 (3.13)
Ridge 0.45 (2.56) 0.44 (2.91)
ENet 0.34 (1.86) 0.53 (3.01)
RF 0.61 (1.95) 0.69 (1.89)
FFN 0.87 (3.34) 0.87 (3.04)
LSTM 0.90 (3.63) 0.56 (2.66)
Combination 0.96 (3.52) 0.75 (2.97)
23
23
Which characteristics matter?

24
24
Bottomline
Traditional view
• Stocks give equity premium
• Government bonds give term premium
• Corporate bonds give default premium

New and improved view

• Factors give factor premia

Traditional bonds portfolios can benefit from focusing on

factors (instead of just duration and rating)

25
TRADING STRATEGY FOR CORPORATE BONDS

GOVERNENT BOND - FUTURE INTEREST PATE TO PREDICT

TAKE A VARIABLE IC =
for stocks

↓ valute
for
X is
corporate
IC = risk of default/credit Rate
Xs from Maturity
YTM leverage there is a
bands

Cap
Market tu correlation between
bond and stocks

t's from PIE

stocks
volatility
monnentura
of stock
Currencies

1
Exchange rate quotations
A currency exchange rate is the rate used to exchange two
currencies. An exchange rate states the price of one currency
in terms of units of another currency

2
Quote convention
All quotes are presented as
a:b = S
• a is the quoted currency
• b is the currency in which the price is expressed
• S is the price of the quoted currency a in units of currency b

For example, $:¥ = 130 means the U.S. dollar is quoted at

130 Japanese yen (¥) per dollar. Or the U.S. dollar is priced
at 130 yen

3
Direct quotes
A direct exchange rate is the domestic price of foreign
currency
a:b = FC:DC
• For example, an American investor seeing a direct quote €:$=1.25
knows she will pay $1.25 for one euro. To a European investor, the
direct quote is $:€=0.80 which says that 1 dollar (foreign currency)
is worth 0.80 euro
▪ Direct quote tells us how much it will cost to purchase a certain amount of the
foreign currency (1 unit): if a European needs to buy something in the U.S.
for $1000, she would use a direct quote to know how much it will cost her in
euros
An appreciation of the foreign currency causes an increase in
the direct quote
4
Indirect quotes
An indirect exchange rate is the amount of foreign currency
that one unit of domestic currency will purchase
a:b = DC:FC
• For an American investor, the indirect quote $:€=0.80 says that 1
dollar will purchase 0.80 euro
▪ Indirect quote tells us how much of the foreign currency we can get for a
certain amount of domestic currency

An appreciation of the foreign currency causes a decrease in

the indirect quote

5
Currency movements

6
Forward rates
Spot rates are quoted for immediate currency transactions
(although in practice delivery takes place 48 hours later)

Forward exchange rates are contracted today but with

delivery and settlement in the future

In a forward, or futures, contract a commitment is

irrevocably made on the transaction date, but delivery takes
place later, on a date set in the contract

7
Forward rate determination
Current spot rate S=$:€=0.8 (direct quote for a European,
indirect quote for an American)

Interest rates are I$=10% and I€=14%

What is the forward rate?

F = S×(1+I€)/(1+I$) = 0.8291

8
Covered interest rate parity

9
CIP
With the convention that the exchange rates are quoted as
a:b
1 + 𝐼𝐼𝑏𝑏
𝐹𝐹 = 𝑆𝑆 ×
1 + 𝐼𝐼𝑎𝑎

If everything is expressed in logs, we can write

𝑓𝑓 − 𝑠𝑠 = 𝑖𝑖𝑏𝑏 − 𝑖𝑖𝑎𝑎

10
Forward premium/discount
With the convention that the exchange rates are quoted as
a:b

if 𝐹𝐹 > 𝑆𝑆 (same as 𝐼𝐼𝑏𝑏 > 𝐼𝐼𝑎𝑎 )

• Currency a trades at a premium relative to currency b (“a is
expected to appreciate”)
▪ Currency a has lower interest rates than currency b

if 𝐹𝐹 < 𝑆𝑆 (same as 𝐼𝐼𝑏𝑏 < 𝐼𝐼𝑎𝑎 )

• Currency a trades at a discount relative to currency b (“a is
expected to depreciate”)
▪ Currency a has higher interest rates than currency b

11
Forward premium/discount …
If a:b is FC:DC (direct quote) then forward premium is
associated with higher domestic interest rates (“FC is
expected to appreciate”)
• What is gained on higher domestic rates is lost on its discount
(recall that domestic currency depreciation is associated with
increase in direct quote)

If a:b is DC:FC (indirect quote) then forward premium is

associated with higher foreign interest rates (“DC is expected
to appreciate”)

12
Example
For an American, direct quote is €:$=1.25
1 + 𝐼𝐼$ 1.10
𝐹𝐹 = 𝑆𝑆 × = 1.25 × = 1.2061 < 1.25
1 + 𝐼𝐼€ 1.14
• Euro trades at a discount relative to Dollar
• “Euro is expected to depreciate”

For a European, direct quote is $:€=0.80

1 + 𝐼𝐼€ 1.14
𝐹𝐹 = 𝑆𝑆 × = 0.80 × = 0.8291 > 0.80
1 + 𝐼𝐼$ 1.10
• Dollar trades at a premium relative to Euro
• “Dollar is expected to appreciate”

13
Parity relations
Purchasing power parity (PPP) relation, linking spot exchange
rates and inflation

Uncovered interest rate parity (UIP) relation, linking spot

exchange rates, expected exchange rates and interest rates

14
PPP relation
PPP states that the spot exchange rate adjusts perfectly to
inflation differentials between two countries

There are two versions of PPP:

• Absolute PPP
▪ This claims that the exchange rate should be equal to the ratio of the average
price levels in the two economies
• Relative PPP
▪ Focuses on the general across the board inflation rates

15
Absolute PPP
The theory states that the spot exchange rate between two
currencies should be equal to the ratio of the two countries’
price levels

PPP exchange rate is found as a ratio of prices of identical

products in the two countries
𝑃𝑃𝑃𝑃𝑃𝑃
𝑃𝑃𝑏𝑏
𝑆𝑆 =
𝑃𝑃𝑎𝑎
With a:b as FC:DC (direct quote)
𝑃𝑃𝑃𝑃𝑃𝑃
𝑃𝑃𝐷𝐷𝐷𝐷
𝑆𝑆 =
𝑃𝑃𝐹𝐹𝐷𝐷

16
Real exchange rate
Ratio of actual to PPP-implied exchange rate
𝑆𝑆 𝑃𝑃𝑎𝑎
𝑅𝑅𝑅𝑅𝑅𝑅 = 𝑃𝑃𝑃𝑃𝑃𝑃 = 𝑆𝑆 ×
𝑆𝑆 𝑃𝑃𝑏𝑏
• RER is technically a unitless measures (i.e. price of goods relative
to the price of goods); generally expressed in index form with
respect to some base period (time of assumed PPP)
• A country’s RER provides a measure of a country’s export
competitiveness: a rise in the index implies a fall in
competitiveness, and vice versa

If absolute PPP holds, then RER = 1

17
Big Mac index
The “Big Mac Index” by The Economist is a prime example of
this law of one price:
• Assuming that the Big Mac is identical in all countries, it serves as a
comparison point as to whether the currencies are trading at
market prices

https://www.economist.com/news/2020/07/15/the-big-mac-index 18
Big Mac index …
Example: Big Mac costs CHF 6.50 in Switzerland and $5.28 in
the US (indirect quotes for American)
• PPP-implied exchange rate is $:CHF = 6.50/5.28 = 1.23106 (it
should cost CHF 1.23 to buy one dollar)
• Actual exchange rate $:CHF = 0.96085 (it costs only CHF 0.96 to
buy one dollar)
• RER = 0.96085/1.23106= 0.78 < 1
▪ Swiss Franc is overvalued

𝑆𝑆
If 𝑅𝑅𝑅𝑅𝑅𝑅 = > 1, then FC is overvalued (DC is undervalued)
𝑆𝑆 𝑃𝑃𝑃𝑃𝑃𝑃
𝑆𝑆
If 𝑅𝑅𝑅𝑅𝑅𝑅 = < 1, then FC is undervalued (DC is overvalued)
𝑆𝑆 𝑃𝑃𝑃𝑃𝑃𝑃
19
Big Mac index …

20
Big Mac index changes …

21
Relative PPP
Percentage movement of the exchange rate should be equal
to the inflation differential between the two economies

1 + Π𝑏𝑏
𝑆𝑆𝑡𝑡+1 = 𝑆𝑆𝑡𝑡 ×
1 + Π𝑎𝑎

𝑠𝑠𝑡𝑡+1 − 𝑠𝑠𝑡𝑡 = 𝜋𝜋𝑏𝑏 − 𝜋𝜋𝑎𝑎

22
Relative PPP …
If a:b is FC:DC (direct quote)

𝑠𝑠𝑡𝑡+1 − 𝑠𝑠𝑡𝑡 = 𝜋𝜋𝐷𝐷𝐷𝐷 − 𝜋𝜋𝐹𝐹𝐷𝐷

PPP says that, if domestic inflation is higher than foreign

inflation, then prices in DC will increase. Therefore, DC must
depreciate (exchange rate increases) in order for DC to
remain competitive with FC

23
Fisher interest rate relation
Nominal interest rates in each country are equal to the
required real rate of return with compensation for expected
inflation
1 + 𝐼𝐼 = (1 + 𝑅𝑅) × (1 + Π) or 𝑖𝑖 = 𝑟𝑟 + 𝜋𝜋

On an international level, the Fisher model assumes that the

real rate requirement is similar across major industrial
countries. Thus any observed market interest rate differences
between counties according to this model is accounted for on
the basis of differences in inflation expectations

24
UIP relation
This is a theory combining relative PPP and the international
Fisher relation. It claims the expected change in the indirect
exchange rate approximately equals the foreign minus the
domestic interest rate

1 + 𝐼𝐼𝑡𝑡,𝑏𝑏
E𝑡𝑡 𝑆𝑆𝑡𝑡+1 = 𝑆𝑆𝑡𝑡 ×
1 + 𝐼𝐼𝑡𝑡,𝑎𝑎

E𝑡𝑡 𝑠𝑠𝑡𝑡+1 − 𝑠𝑠𝑡𝑡 = 𝑖𝑖𝑡𝑡,𝑏𝑏 − 𝑖𝑖𝑡𝑡,𝑎𝑎

Similar to EH for bonds

25
UIP relation …
If a:b is FC:DC (direct quote)

E𝑡𝑡 𝑠𝑠𝑡𝑡+1 − 𝑠𝑠𝑡𝑡 = 𝑖𝑖𝑡𝑡,𝐷𝐷𝐷𝐷 − 𝑖𝑖𝑡𝑡,𝐹𝐹𝐷𝐷

UIP says that foreign currency movements are equal to

interest rate differentials. If domestic interest rates are
higher than foreign rates, then DC must depreciate
(exchange rate increases)

26
Forward unbiasedness hypothesis
Combine CIP (no arbitrage relation)
𝑓𝑓𝑡𝑡 − 𝑠𝑠𝑡𝑡 = 𝑖𝑖𝑡𝑡,𝑏𝑏 − 𝑖𝑖𝑡𝑡,𝑎𝑎
and UIP (a theory)
E𝑡𝑡 𝑠𝑠𝑡𝑡+1 − 𝑠𝑠𝑡𝑡 = 𝑖𝑖𝑡𝑡,𝑏𝑏 − 𝑖𝑖𝑡𝑡,𝑎𝑎
to get
𝑓𝑓𝑡𝑡 = E𝑡𝑡 𝑠𝑠𝑡𝑡+1

Forward rates are unbiased predictors of future spot rates

(FUH)

27
Prices, interest rates, & exchange rates
(A) Purchasing power parity
• forecasts the change in the spot rate based on differences in expected rates of
inflation
(B) Fisher effect
• nominal interest rates are the required real rate of return (r) plus expected
inflation (π)
(C) International Fisher effect
• spot exchange rate should change in an amount equal to but in the opposite
direction of the difference in interest rates between countries
(D) Interest rate parity
• difference in the national interest rates should be equal to, but opposite in sign
to, the forward rate discount or premium for the foreign currency
(E) Forward rate as an unbiased predictor
• forward rate is an efficient predictor of the future spot rate

28
Recap of CIP and UIP
Direct quote FC:DC For a European $:€

(CIP) (CIP)
𝑓𝑓𝑡𝑡 = 𝑠𝑠𝑡𝑡 + 𝑖𝑖𝑡𝑡,𝐷𝐷𝐷𝐷 − 𝑖𝑖𝑡𝑡,𝐹𝐹𝐷𝐷 𝑓𝑓𝑡𝑡 = 𝑠𝑠𝑡𝑡 + 𝑖𝑖𝑡𝑡,€ − 𝑖𝑖𝑡𝑡,$
If 𝑖𝑖𝐷𝐷𝐷𝐷 > 𝑖𝑖𝐹𝐹𝐷𝐷 , FC trades at a If 𝑖𝑖€ > 𝑖𝑖$ , $ trades at a premium
premium to DC to €

(UIP) (UIP)
E𝑡𝑡 𝑠𝑠𝑡𝑡+1 = 𝑠𝑠𝑡𝑡 + 𝑖𝑖𝑡𝑡,𝐷𝐷𝐷𝐷 − 𝑖𝑖𝑡𝑡,𝐹𝐹𝐷𝐷 E𝑡𝑡 𝑠𝑠𝑡𝑡+1 = 𝑠𝑠𝑡𝑡 + 𝑖𝑖𝑡𝑡,€ − 𝑖𝑖𝑡𝑡,$
If 𝑖𝑖𝐷𝐷𝐷𝐷 > 𝑖𝑖𝐹𝐹𝐷𝐷 , FC is expected to If 𝑖𝑖€ > 𝑖𝑖$ , $ is expected to
appreciate appreciate

29
UIP and RW (direct quote)
With the convention a:b is FC:DC, Sell FC in the forward
market at 𝑡𝑡 and buy FC in the spot market at 𝑡𝑡 + 1
𝑟𝑟�𝑟𝑟𝑡𝑡+1 = −𝑓𝑓𝑡𝑡 + 𝑠𝑠̃𝑡𝑡+1 = − 𝑓𝑓𝑡𝑡 − 𝑠𝑠𝑡𝑡 + Δ𝑠𝑠̃𝑡𝑡+1
= 𝑖𝑖𝑡𝑡,𝐹𝐹𝐷𝐷 − 𝑖𝑖𝑡𝑡,𝐷𝐷𝐷𝐷 + Δ𝑠𝑠̃𝑡𝑡+1

Under UIP (similar to EH for bonds)

E𝑡𝑡 Δ𝑠𝑠̃𝑡𝑡+1 = 𝑖𝑖𝑡𝑡,𝐷𝐷𝐷𝐷 − 𝑖𝑖𝑡𝑡,𝐹𝐹𝐷𝐷 and E𝑡𝑡 𝑟𝑟�𝑟𝑟𝑡𝑡+1 = 0
• You earn higher interest rate on FC, but it depreciates in the future

Under RW
E𝑡𝑡 Δ𝑠𝑠̃𝑡𝑡+1 = 0 and E𝑡𝑡 𝑟𝑟�𝑟𝑟𝑡𝑡+1 = 𝑖𝑖𝑡𝑡,𝐹𝐹𝐷𝐷 − 𝑖𝑖𝑡𝑡,𝐷𝐷𝐷𝐷
30
UIP and RW (indirect quote)
With the convention a:b is DC:FC, Buy FC in the forward
market at 𝑡𝑡 and sell FC in the spot market at 𝑡𝑡 + 1
𝑟𝑟�𝑟𝑟𝑡𝑡+1 = 𝑓𝑓𝑡𝑡 − 𝑠𝑠̃𝑡𝑡+1 = 𝑓𝑓𝑡𝑡 − 𝑠𝑠𝑡𝑡 − Δ𝑠𝑠̃𝑡𝑡+1
= 𝑖𝑖𝑡𝑡,𝐹𝐹𝐷𝐷 − 𝑖𝑖𝑡𝑡,𝐷𝐷𝐷𝐷 − Δ𝑠𝑠̃𝑡𝑡+1

Under UIP
E𝑡𝑡 Δ𝑠𝑠̃𝑡𝑡+1 = 𝑖𝑖𝑡𝑡,𝐹𝐹𝐷𝐷 − 𝑖𝑖𝑡𝑡,𝐷𝐷𝐷𝐷 and E𝑡𝑡 𝑟𝑟�𝑟𝑟𝑡𝑡+1 = 0
• You earn higher interest rate on FC but it depreciates in the future
Under RW
E𝑡𝑡 Δ𝑠𝑠̃𝑡𝑡+1 = 0 and E𝑡𝑡 𝑟𝑟�𝑟𝑟𝑡𝑡+1 = 𝑖𝑖𝑡𝑡,𝐹𝐹𝐷𝐷 − 𝑖𝑖𝑡𝑡,𝐷𝐷𝐷𝐷

31
Regressions
Regress

Δ𝑠𝑠̃𝑡𝑡+1 = 𝛾𝛾0 + 𝛾𝛾1 𝑓𝑓𝑡𝑡 − 𝑠𝑠𝑡𝑡 + 𝑒𝑒𝑡𝑡+1

Null under UIP 𝛾𝛾1 = 1, under RW 𝛾𝛾1 = 0

Actual empirical evidence is 𝛾𝛾1 ≈ −1

• Fama (1984), Cumby and Obstfeld (1984)

𝑅𝑅𝑡𝑡 𝑟𝑟�𝑟𝑟𝑡𝑡+1 ≈ 2 𝑓𝑓𝑡𝑡 − 𝑠𝑠𝑡𝑡 = 2 𝑖𝑖𝑡𝑡,𝐹𝐹𝐷𝐷 − 𝑖𝑖𝑡𝑡,𝐷𝐷𝐷𝐷

32
Carry trade
Invest in high interest rate currencies
• Under UIP, what you gain on high interest rate, you should lose on
currency depreciation
• Actually, not only do you earn high interest rate, but also the
currency actually appreciates

Short (borrow) in low interest rate currencies

• Under UIP, what you lose on low interest rate, you should gain on
currency appreciation
• Actually, not only do you pay low interest rate, but also the
currency actually depreciates

33
Empirical evidence
Doskov and Swinkels (JIMF 2015)

34
Currency carry and equities

35
Carry over the years

36
Carry trades in recent years
Menkhoff, Sarno, Schmeling, and Schrimpf (JF 2012)

Carry returns are much higher after Bretton Woods

37
Carry trades during recent crisis
Melvin and Taylor (JIMF 2009)

38
Aggregate carry
(𝑘𝑘) 𝑘𝑘
Usual carry is to go long in currencies 𝑘𝑘 where 𝑓𝑓𝑡𝑡 − 𝑠𝑠𝑡𝑡 <0
(€) 𝑘𝑘 (𝑘𝑘)
(equivalently 𝑖𝑖𝑡𝑡 < 𝑖𝑖𝑡𝑡 ) and short currencies 𝑘𝑘 where 𝑓𝑓𝑡𝑡 −
𝑘𝑘 (€) 𝑘𝑘
𝑠𝑠𝑡𝑡 > 0 (equivalently 𝑖𝑖𝑡𝑡 > 𝑖𝑖𝑡𝑡 )

Construct aggregate carry: Go long in all currencies when

(𝑘𝑘) 𝑘𝑘
where ∑𝑘𝑘 𝑓𝑓𝑡𝑡 − 𝑠𝑠𝑡𝑡 ⁄𝐾𝐾 < 0 and short otherwise
• Buy currencies by shorting (borrowing) Euro when the average
forward premium is negative (€ is expected to appreciate against a
basket of currencies)

39
Aggregate carry …
Lustig, Roussanov, and Verdelhan (JFE 2014)

USD: Aggregate
carry

FX: Country-level
carry

HML: long-short
carry

40
Aggregate carry …
Risk premia is high in bad times (counter-cyclical)

41
Currency momentum
Menkhoff, Sarno, Schmeling, and Schrimpf (JFE 2012)

Similar to equity momentum

42
Currency momentum …
Even at 1-month horizons (no short-term reversal)

43
Carry and momentum
Are lagged returns the same as lagged high interest rate
differentials?

Low correlation between carry and momentum

But, high T-costs as profitability comes from “minor”

currencies
44
Currency value
Asness, Moskowitz, and Pedersen (JF 2013)

Invest in “good value” currencies based on RER (buy weak

currencies with low RER that have become cheap in real
terms compared to $; 5-year change in RER)

45
Currency value enhanced
Menkhoff, Sarno,
Schmeling, and
Schrimpf (RFS 2017)

Adjust RER for

fundamentals

46
Currency value enhanced …
Adjusted strategy has higher Sharpe ratio and lower
drawdowns

47
Value and momentum combo
Asness, Moskowitz, and Pedersen (JF 2013)

48
Technical trading rules
Hsu, Taylor, and Wang (JFM 2016)

Examine 21,000 technical trading rules on 30 currencies over 1971-

2015

Out-of-sample performance to the right on vertical line

49
Summary of trading strategies
Carry
 Go long in high interest rate currencies, short low interest
rate currencies
Momentum
 Go long in past winner (increase in spot rate) currencies,
short past loser currencies
• Past horizon can be between 1 and 12 months
Value
 Go long in cheap currencies, short expensive currencies
• Cheap/Expensive defined as the percentage change in RER over the
last 5 years

50
Commodities

1
Size of commodity futures markets

Open
Interest:
Total
number of
contracts
either long
or short.
Each
contract
has both
sides, but
for (OI)
only one is
counted.

Note: Measures of total open interest are generally much smaller than the value of production
and inventories, but turnover in futures markets can be significantly larger than measures of
physical market size. Although a good deal of this turnover is likely to be speculative, the
comparatively small level of open interest suggests much of this trading is very short term in
nature.
Source: RBA (2011), http://www.rba.gov.au/publications/bulletin/2011/jun/pdf/bu-0611-7.pdf
2
Commodity markets
Hedgers
• Natural exposure to hedge
• BP swap dealers

Processors Consumers
Producers

Commodity Market
• Physical
• Financial (OTC, CME, ICE,...)

Banks/Dealers Investors
Commodity
Traders
Intermediaries Arbitrageurs Speculators
• look for ‘quasi-riskless’ trades • No natural exposure to commodity
• Predict future price

3
Physical and Financial Trading

Physical Trading Financial Trading/Derivatives Contracts

Spot Trading Forward Futures

• Commercial contract • Unilateral agreement, OTC • Standardized, Exchange traded

• Flexible covenants • Flexible covenants • Centrally cleared, margins

• Actual/physical • No margin/lower capital • Trade repository
cost
• Juridical commitment of the buyer • replace spot transaction on • Necessity of a physical delivery
and seller until execution of the many occasions (e.g. non- or termination of the position
contract storable commodity such as before maturity
electricity

• Long transaction • form of contracting • CCP generates ‘market prices’,

appropriate for commodities marked to market
• Illiquid and discontinuous market • Flexible regarding optimal • Liquidity, no credit risk unless
(also swaps) transfer of goods clearing house fails
• Transfer of goods in conditions • Low transaction costs but
suiting demand higher capital charges

4
Example of Brent Crude Futures
If a seller of one Aug 13 futures contract where F(t=July 5, 2013, T=Aug
2013)=107.48 takes/chooses physical delivery, this means selling
1,000 barrels for $107,480

1 US oil barrel
=42 US gallons
=159 litres

1 barrel of WTI
weighs 138.8
kg

Capacity of a
ULCC is 4m
barrel or at
$107.46 per
barrel, $430m
per tanker

5
Forward contract
An agreement to buy or sell an asset at a certain future time
for a certain price specified today (forward price)
• The party that has agreed to buy has what is termed a long position
• The party that has agreed to sell has what is termed a short
position

6
Forward price
The forward price for a contract is the delivery price that
would be applicable to the contract if were negotiated today
(i.e., it is the delivery price that would make the contract
worth exactly zero today)
• The forward price may be different for contracts of different
maturities
• We will denote the forward price with symbol 𝐹𝐹0,𝑇𝑇

7
Payoff with forwards

Long
𝐹𝐹0,𝑇𝑇
position

Profit

0 𝑃𝑃�𝑇𝑇
Short
Loss position
𝐹𝐹0,𝑇𝑇
−𝐹𝐹0,𝑇𝑇

8
Hedging using forwards
A long futures hedge is appropriate when you know you will
purchase an asset in the future and want to lock in the price
Payoff = 𝑃𝑃�𝑇𝑇 − 𝐹𝐹0,𝑇𝑇

A short futures hedge is appropriate when you know you will

sell an asset in the future and want to lock in the price
Payoff = 𝐹𝐹0,𝑇𝑇 − 𝑃𝑃�𝑇𝑇

9
Short hedge example
 Today is May 15. An oil producer has negotiated a contract
to sell 1m barrels of oil on Aug 15. If the oil price is lower
by 1¢, then profits are lower by $10K
 Can hedge by selling futures. Price today of Aug futures is
$59/barrel
 If the spot price on Aug 15 is $55/barrel
• Payoff w/o future: $55m
• Payoff w/ future: $55m + ($59−$55)×1m = $59m
 If the spot price on Aug 15 is $65/barrel
• Payoff w/o future: $65m
• Payoff w/ future: $65m + ($59−$65)×1m = $59m

10
Long hedge example
 Today is Jan 15. A copper fabricator knows it will require
100K pounds of copper on May 15. The spot price of copper
is 340¢.
 Can hedge by buying futures. Price today of May futures is
320¢/pound
 If the spot price on May 15 is 325¢/pound
• Cost w/o future: $325K
• Cost w/ future: $325K + (320¢−325¢)×1K = $320K
 If the spot price on May 15 is 310¢/pound
• Cost w/o future: $310K
• Cost w/ future: $310K + (320¢−310¢)×1K = $320K

11
Why hedge?
In the previous examples, there was a distinct possibility that
we would lose money by hedging. So why hedge?
• Companies should focus on the main business they are in and take
steps to minimize risks arising from interest rates, exchange rates,
and other market variables

Some possible reasons:

• Shareholders less informed and higher costs
• Industry practice (e.g. gold industry)
• Bankruptcy costs

12
Long or short hedge

Condition today Risk Appropriate

hedge
Plan to sell Asset price may Short
asset fall
Hold asset Asset price may Short
fall
Plan to buy Asset price may Long
asset rise
Sold short Asset price may Long
asset rise

13
Hedged payoff with forwards

Final payoff
𝐹𝐹0,𝑇𝑇 𝐹𝐹0,𝑇𝑇
Underlying
Long
Profit Profit
hedge
0 𝑃𝑃�𝑇𝑇 0 𝑃𝑃�𝑇𝑇
𝐹𝐹0,𝑇𝑇 Short 𝐹𝐹0,𝑇𝑇
hedge
Loss Loss
−𝐹𝐹0,𝑇𝑇 −𝐹𝐹0,𝑇𝑇
Final cost
Underlying

𝑃𝑃�𝑇𝑇 − 𝑃𝑃�𝑇𝑇 − 𝐹𝐹0,𝑇𝑇 = +𝐹𝐹0,𝑇𝑇 −𝑃𝑃�𝑇𝑇 + 𝑃𝑃�𝑇𝑇 − 𝐹𝐹0,𝑇𝑇 = −𝐹𝐹0,𝑇𝑇

14
Convergence of forward price to spot
If the asset is liquid then at maturity, 𝐹𝐹�𝑇𝑇,𝑇𝑇 = 𝑃𝑃�𝑇𝑇 but not before.
What happens to hedge initiated at time 𝑡𝑡1 and closed out at
time 𝑡𝑡2 ?

Futures price

Spot price

Time
𝑡𝑡1 𝑡𝑡2 𝑇𝑇

15
Basis
Basis is the difference between the spot and futures price:

𝐵𝐵𝑡𝑡 ≡ 𝑃𝑃𝑡𝑡 − 𝐹𝐹𝑡𝑡,𝑇𝑇

where 𝐹𝐹𝑡𝑡,𝑇𝑇 is the futures price at date 𝑡𝑡 and 𝑃𝑃𝑡𝑡 as the spot
price at date 𝑡𝑡 where 𝑡𝑡 ≤ 𝑇𝑇 and 𝑇𝑇 is the maturity
• At time 0: 𝐵𝐵0 = 𝑃𝑃0 − 𝐹𝐹0,𝑇𝑇
• At maturity: 𝐵𝐵𝑇𝑇 = 𝑃𝑃𝑇𝑇 − 𝐹𝐹𝑇𝑇,𝑇𝑇 = 0

16
Basis risk
For a position opened at time 0 and closed at time 𝑡𝑡 (before maturity 𝑇𝑇)

Condition today Appropriate Hedged payoff

hedge (Cost/Revenue)
Plan to sell asset Short 𝑃𝑃𝑡𝑡 + 𝐹𝐹0,𝑇𝑇 − 𝐹𝐹𝑡𝑡,𝑇𝑇
= 𝐵𝐵𝑡𝑡 + 𝐹𝐹0,𝑇𝑇
Hold asset Short 𝑃𝑃𝑡𝑡 − 𝑃𝑃0 + 𝐹𝐹0,𝑇𝑇 − 𝐹𝐹𝑡𝑡,𝑇𝑇
= 𝐵𝐵𝑡𝑡 − 𝐵𝐵0
Plan to buy Long −𝑃𝑃𝑡𝑡 + 𝐹𝐹𝑡𝑡,𝑇𝑇 − 𝐹𝐹0,𝑇𝑇
asset = −𝐹𝐹0,𝑇𝑇 − 𝐵𝐵𝑡𝑡
Sold short asset Long 𝑃𝑃0 − 𝑃𝑃𝑡𝑡 + 𝐹𝐹𝑡𝑡,𝑇𝑇 − 𝐹𝐹0,𝑇𝑇
= 𝐵𝐵0 − 𝐵𝐵𝑡𝑡

17
Basis and hedging
Profits from hedges are simply the change in basis
• Hedge substitutes change in basis for change in spot price

Basis change is usually less variable than spot price change

• Hedged position is less risky than unhedged position
• Hedging could still be considered as a speculative activity but at a
risk level lower than that of an unhedged position

18
More on basis risk
Sometimes the basis risk arises because the forward contract
is not exactly on the underlying that one wants to hedge
• An airline wanting to hedge price of jet fuel might have futures only
on crude oil and heating oil

In this case

𝐵𝐵𝑇𝑇 = 𝑃𝑃𝑇𝑇 − 𝐹𝐹𝑇𝑇,𝑇𝑇 ≠ 0

and one has an additional source of risk

19
Hedge ratio
The objective is to find the hedging ratio that minimizes the
cash-flow volatility of the hedged position

If ℎ is the hedge ratio, Δ𝑃𝑃 is the change in the spot price and
Δ𝐹𝐹 is the change in the futures price, then the change in the
cash-flow is
|Δ𝑃𝑃 − ℎ × Δ𝐹𝐹|

and the variance of the cash-flow is

𝑉𝑉 = var Δ𝑃𝑃 + ℎ2 var Δ𝐹𝐹 − 2ℎcov(Δ𝑃𝑃, Δ𝐹𝐹)

20
Hedge ratio …
The hedge ratio that minimizes the variance of cash-flow is:
∗
cov(Δ𝑃𝑃, Δ𝐹𝐹)
ℎ =
var(Δ𝐹𝐹)

which is the same as the estimator of the slope of an OLS of

Δ𝑃𝑃 on Δ𝐹𝐹
Δ𝑃𝑃𝑡𝑡 = constant + ℎ∗ × Δ𝐹𝐹𝑡𝑡 + 𝑒𝑒𝑡𝑡

One can write the variance of the hedged cash-flow as

𝑉𝑉 = var(𝑒𝑒) = var(Δ𝑃𝑃) 1 − 𝑅𝑅2

21
Hedge ratio …
Performance of the hedge depends on the random error
associated with the statistical relation between Δ𝑃𝑃 and Δ𝐹𝐹

This approach is only useful when the variation in Δ𝐹𝐹 can

predict a significant proportion of the variation in Δ𝑃𝑃
• When the 𝑅𝑅2 of the OLS regression is high
• With a low 𝑅𝑅2 , it is important to check whether other futures
contract might not be more useful for hedging
▪ An airline with jet fuel exposure might find that crude oil contracts are better
than heating oil contracts

22
Alternative definitions of hedge ratio
So far, 𝑃𝑃 and 𝐹𝐹 have referred to as total dollar values

Defining in terms of unit prices

• Let 𝑞𝑞 denote quantities and small-case letters prices, 𝑃𝑃 = 𝑞𝑞𝑝𝑝 × 𝑝𝑝 and
𝐹𝐹 = 𝑞𝑞𝐹𝐹 × 𝑓𝑓
∗
cov(Δ𝑝𝑝, Δ𝑓𝑓) 𝑞𝑞𝑝𝑝
ℎ = ×
var(Δ𝑓𝑓) 𝑞𝑞𝑓𝑓

Defining in terms of returns

• Let 𝑅𝑅 denote returns
cov(𝑅𝑅𝑝𝑝 , 𝑅𝑅𝑓𝑓 ) 𝑞𝑞𝑝𝑝 × 𝑝𝑝
ℎ∗ = ×
var(𝑅𝑅𝑓𝑓 ) 𝑞𝑞𝑓𝑓 × 𝑓𝑓
23
Example
Delta Airlines expects to purchase two million gallons of jet
fuel in one month and decides to use heating oil futures for
hedging

How many futures contracts on heating oil should this airline

long to achieve the minimum variance hedge? Note that
heating oil futures are traded on NYMEX, and each contract is
for 42,000 gallons

24
Example …
For the last 15 days, data on daily changes in jet fuel spot
prices and heating oil futures can be summarized by
regression slope of 0.78

2,000,000
Optimal hedge ratio ℎ∗ = 0.78 × = 37.14
42,000

Thus, Delta should go long 37 futures contracts

25
Tailing the hedge
We derived everything using forwards. With futures that are
marked-to-market, one has to account for non-constant
interest rates

A heuristic solution is to multiply the hedge ratio, ℎ∗ , by the

ratio of spot to futures price

𝑝𝑝
Tailed hedge ratio ℎ∗∗ = ℎ∗ ×
𝑓𝑓
• In the previous example, if the spot and futures prices were $1.94 and
$1.99 per gallon, then the tailed hedge ratio would be
37.14×1.94/1.99=36.22

26
Futures
Forwards Futures

 Private contract between  Exchange traded

two parties
 Non-standard contract  Standard contract
 Usually one specified  Range of delivery dates
delivery date
 Settled at maturity  Settled daily
 Delivery or final cash  Contract usually closed out
settlement usually occurs prior to maturity

27
Marking to market (MTM)
Assume you hold one silver contract. One contract is for 5000 ounces
and assume futures price today for delivery in 5 days is $17.10. The
daily MTM gain/loss is …
Day Profit(Loss) Daily proceeds
Day Futures
per ounce (=5000×P/L
price
per ounce)
0 $17.10
1 17.20−17.10 5000×0.10
1 17.20 =0.10 =500
2 17.25 2 17.25−17.20 5000×0.0
3 17.18 =0.05 =250

4 17.18 3 17.18−17.25 5000×(−0.07)

=−0.07 =−350
5 17.21
… and the sum of all 4 17.18−17.18 5000×0
daily proceeds is 𝑃𝑃𝑇𝑇 − 𝐹𝐹0,𝑇𝑇 =0 =0
5 17.21−17.18 5000×0.03
=0.03 =150
Sum=$550 28
Forwards vs. futures
Forward and futures prices are usually assumed to be the
same

When interest rates are uncertain, the two prices are slightly
different
• A strong positive correlation between interest rates and the asset
price implies the futures price is slightly higher than the forward
price (Why?)
• A strong negative correlation implies the reverse

29
Futures pricing (1)
Consider an asset that doesn’t pay any interim cash-flow, has no
storage costs etc. Consider two strategies that you enter at time 0:
 A: Go long forward contract at price 𝐹𝐹0,𝑇𝑇
 B: Go long the asset and borrow 𝑃𝑃0 as a risk-free loan promising to
pay it back at time 𝑇𝑇
Strategy A Strategy B
Time 0 0 0
Time 𝑇𝑇 𝑃𝑃�𝑇𝑇 − 𝐹𝐹0,𝑇𝑇 𝑃𝑃�𝑇𝑇 − 𝑃𝑃0 1 + 𝑟𝑟 𝑇𝑇

Since both strategies cost the same (nothing) at time 0, they must be
worth the same at time 𝑇𝑇, meaning 𝑃𝑃�𝑇𝑇 − 𝐹𝐹0,𝑇𝑇 = 𝑃𝑃�𝑇𝑇 − 𝑃𝑃0 1 + 𝑟𝑟 𝑇𝑇 . Thus,

𝑇𝑇
𝐹𝐹0,𝑇𝑇 = 𝑃𝑃0 1 + 𝑟𝑟 = 𝑃𝑃0 𝑒𝑒 𝑟𝑟𝑇𝑇
30
Futures pricing (1 …)
Say that the actual futures price is 𝐹𝐹0,𝑇𝑇
∗
which is different from the
theoretical futures price 𝐹𝐹0,𝑇𝑇 . Then, there exists an arbitrage

 If 𝐹𝐹0,𝑇𝑇
∗
> 𝐹𝐹0,𝑇𝑇 , then “short” strategy A and “long” strategy B
• Short forward, long asset, and borrow 𝑃𝑃0
• Payoff at time 𝑇𝑇 is 𝐹𝐹0,𝑇𝑇
∗
− 𝑃𝑃�𝑇𝑇 + 𝑃𝑃�𝑇𝑇 − 𝑃𝑃0 1 + 𝑟𝑟 𝑇𝑇 ∗
= 𝐹𝐹0,𝑇𝑇 − 𝐹𝐹0,𝑇𝑇 > 0

 If 𝐹𝐹0,𝑇𝑇
∗
< 𝐹𝐹0,𝑇𝑇 , then “long” strategy A and “short” strategy B
• Long forward, short asset, and invest 𝑃𝑃0
• Payoff at time 𝑇𝑇 is −𝐹𝐹0,𝑇𝑇
∗
+ 𝑃𝑃�𝑇𝑇 − 𝑃𝑃�𝑇𝑇 + 𝑃𝑃0 1 + 𝑟𝑟 𝑇𝑇 ∗
= −𝐹𝐹0,𝑇𝑇 + 𝐹𝐹0,𝑇𝑇 > 0

31
Futures pricing (2)
Now consider an asset that pays an interim cash-flow 𝐷𝐷 (coupon,
dividends, etc.) at time 𝑇𝑇. Same two strategies as before

Strategy A Strategy B
Time 0 0 0
Time 𝑇𝑇 𝑃𝑃�𝑇𝑇 − 𝐹𝐹0,𝑇𝑇 𝑃𝑃�𝑇𝑇 + 𝐷𝐷 − 𝑃𝑃0 1 + 𝑟𝑟 𝑇𝑇

𝐹𝐹0,𝑇𝑇 = 𝑃𝑃0 1 + 𝑟𝑟 𝑇𝑇 − 𝐷𝐷 = 𝑃𝑃0 𝑒𝑒 (𝑟𝑟−𝑑𝑑)𝑇𝑇

where 𝑑𝑑 is the (continuously compounded) cash-flow yield

• Example, FX forward price (𝑟𝑟𝑑𝑑 is domestic rate and 𝑟𝑟𝑓𝑓 is foreign rate) is
1 + 𝑟𝑟𝑑𝑑 𝑇𝑇 𝑟𝑟𝑑𝑑 −𝑟𝑟𝑓𝑓 𝑇𝑇
𝐹𝐹0,𝑇𝑇 = 𝑋𝑋0 𝑇𝑇 = 𝑋𝑋0 𝑒𝑒
1 + 𝑟𝑟𝑓𝑓
32
Futures pricing (3)
Now consider an asset on which you incur storage costs of 𝑆𝑆 that are
paid at time 𝑇𝑇.
Strategy A Strategy B
Time 0 0 0
Time 𝑇𝑇 𝑃𝑃�𝑇𝑇 − 𝐹𝐹0,𝑇𝑇 𝑃𝑃�𝑇𝑇 + 𝐷𝐷 − 𝑆𝑆 − 𝑃𝑃0 1 + 𝑟𝑟 𝑇𝑇

𝑇𝑇
𝐹𝐹0,𝑇𝑇 = 𝑃𝑃0 1 + 𝑟𝑟 − 𝐷𝐷 + 𝑆𝑆 = 𝑃𝑃0 𝑒𝑒 (𝑟𝑟−𝑑𝑑+𝑠𝑠)𝑇𝑇

where 𝑠𝑠 is the (continuously compounded) storage cost yield. 𝜃𝜃 = 𝑟𝑟 −

𝑑𝑑 + 𝑠𝑠 is called as the carry yield. We can write
𝐹𝐹0,𝑇𝑇 = 𝑃𝑃0 + cost of carry
• Cost of carry is positive if cost of storage exceeds net interest and negative
otherwise
33
Commodities as consumption assets
The arbitrage in the previous expression required the ability to buy or
sell the asset freely

Commodities are special kind of assets in the sense that they are
consumption assets

This means that individuals may be reluctant to sell the commodity in

the spot market and buy futures
• Oil futures cannot be used to feed a refinery!

Thus, the arbitrage works only in one direction and all we can say is
that
𝐹𝐹0,𝑇𝑇 ≤ 𝑃𝑃0 1 + 𝑟𝑟 𝑇𝑇 − 𝐷𝐷 + 𝑆𝑆 = 𝑃𝑃0 𝑒𝑒 (𝑟𝑟−𝑑𝑑+𝑠𝑠)𝑇𝑇
34
Futures pricing (4)
Now consider a commodity. Holding the physical commodity is
valuable. These benefits are refereed to as convenience yield. Let 𝐶𝐶 be
the positive value of this convenience yield. Then the futures price is
given by:

𝐹𝐹0,𝑇𝑇 = 𝑃𝑃0 1 + 𝑟𝑟 𝑇𝑇
X + 𝑆𝑆 − 𝐶𝐶 = 𝑃𝑃0 𝑒𝑒 (𝑟𝑟−𝑑𝑑+𝑠𝑠−𝑐𝑐)𝑇𝑇
− 𝐷𝐷

where 𝑐𝑐 is the (continuously compounded) convenience yield

• In these cases, the futures price may be less than spot price plus the cost of
carry

35
Convenience yield
Can be viewed as the number that restores the equality
between futures price and spot price plus cost of carry

Reflects market’s expectations concerning future availability

of commodity
• Greater the possibility that shortages will occur, higher the
convenience yield
▪ If inventories are high, chance of shortages is low, convenience yield is low
▪ If inventories are low, shortages are more likely, convenience yield is high

36
Contango/Backwardation
Contango Backwardation
(Negative basis) (Positive basis)

𝐹𝐹0,𝑇𝑇 > 𝑃𝑃0 𝐹𝐹0,𝑇𝑇 < 𝑃𝑃0

 Financial securities that do not  Financial securities that pay

pay coupon or dividends high coupon or dividends
 Commodities with low  Commodities with high
convenience yield (high convenience yield (low
inventories) inventories)

37
Example
Both may exist on the same commodity (with different
futures)

38
Is 𝐹𝐹0,𝑇𝑇 = 𝐸𝐸0 𝑃𝑃�𝑇𝑇 ?
Start with a simple financial asset that pays no interim cash-
flow. We know that 𝐹𝐹0,𝑇𝑇 = 𝑃𝑃0 𝑒𝑒 𝑟𝑟𝑇𝑇 . The question, thus, can be
restated as is 𝑃𝑃0 𝑒𝑒 𝑟𝑟𝑇𝑇 = 𝐸𝐸0 𝑃𝑃�𝑇𝑇 or is 𝑃𝑃0 = 𝐸𝐸0 𝑃𝑃�𝑇𝑇 𝑒𝑒 −𝑟𝑟𝑇𝑇 ?

The answer is no as we know that present value must take

into account a risk-premium

39
Futures price and future
Futures prices are biased expectations of future spot prices
• Bias due to risk premium
• Risk premium in futures markets exists only because it is
transferred from spot markets
𝐹𝐹0,𝑇𝑇 < 𝐸𝐸0 (𝑃𝑃�𝑇𝑇 )
• Called “normal backwardation”
▪ Not to be confused with backwardation (𝐹𝐹0,𝑇𝑇 < 𝑃𝑃0 )
• Can have contango and normal backwardation at the same time
𝑃𝑃0 < 𝐹𝐹0,𝑇𝑇 < 𝐸𝐸0 (𝑃𝑃�𝑇𝑇 )

40
Convenience yield again
Recall that, in theory, backwardation does not mean that the
futures price is too low; it only means that convenience yield
is high

However, convenience yield is basically a plug-in number and

depends on inventories

Futures commodity prices may, thus, contain information

relevant for future spot prices

In the data, change in spot prices and basis are positively

related
41
Futures return and basis

Source: Gorton, Hayashi, and Rouwenhorst (2012) 42

Commodities carry
Koijen, Moskowitz, Pedersen, and Vrugt (2018) find returns
of 11% with a Sharpe ratio of 0.60 in global commodities

𝐶𝐶𝐶𝐶𝑟𝑟𝑟𝑟𝐶𝐶 = 𝑃𝑃0 − 𝐹𝐹0,𝑇𝑇 �𝐹𝐹0,𝑇𝑇

43
Commodities carry …

44
Commodities value and momentum
Asness, Moskowitz, and Pedersen (2013)

Value: Change in spot price from 5 years ago to today

Momentum: Return over past 12 months

45
Basis momentum
Boons and Prado (2019): Combine basis and momentum
(Difference between momentum of first and second-nearby
futures contract)

46
Trading

1
Market sizes regional, $millions
Domestic Market Capitalisation Europe - Middle East - Africa
(USD millions) Abu Dhabi Securities Exchange 137,617.5 124,532.6
Amman Stock Exchange 22,740.0 23,924.6
Exchange End 2018 End 2017 Athens Stock Exchange 38,370.8 48,346.2
Bahrain Bourse 21,862.7 21,723.5
Beirut Stock Exchange 9,675.2 11,491.8
Americas BME Spanish Exchanges 723,691.0 849,161.9
B3 SA Brasil Bolsa Balcao 916,824.4 815,165.2 Borsa Istanbul 149,263.6 162,886.8
Barbados Stock Exchange 3,541.6 3,350.9 Botswana Stock Exchange 4,034.1 4,583.5
Bermuda Stock Exchange 2,590.1 2,874.8 Bourse de Casablanca 61,080.8 65,754.4
Bolsa de Comercio de Buenos Aires 45,986.1 54,712.0 BRVM 8,453.3 12,485.7
Bolsa de Comercio de Santiago 250,739.6 261,259.2 Bucharest Stock Exchange 20,853.8 23,621.3
Bolsa de Valores de Colombia 103,848.4 111,293.8 Budapest Stock Exchange 28,934.6 31,553.8
Bolsa de Valores de Lima 93,385.4 95,227.2 Cyprus Stock Exchange 3,313.5 2,696.4
Bolsa de Valores de Panama 15,647.7 15,023.7 Deutsche Börse AG 1,755,172.8 2,161,242.1
Bolsa Mexicana de Valores 385,051.4 417,729.5 Dubai Financial Market 86,208.3 107,291.7
Bolsa Nacional de Valores 2,401.1 3,010.8 Euronext 3,730,398.3 4,196,901.7
Jamaica Stock Exchange 12,057.1 9,404.8 Iran Fara Bourse 26,912.2 14,954.7
Nasdaq - US 9,756,836.1 10,039,335.6 Irish Stock Exchange 110,154.4 140,012.3
NYSE 20,679,476.9 22,081,367.0 Johannesburg Stock Exchange 865,327.7 1,057,528.5
TMX Group 1,937,902.7 2,179,674.9 Kazakhstan Stock Exchange 37,005.3 39,489.8
34,206,288.6 36,089,429.5 Ljubljana Stock Exchange 7,266.5 6,318.4
LSE Group 3,637,996.0 4,455,429.5
Asia - Pacific Luxembourg Stock Exchange 49,482.6 65,564.4
Australian Securities Exchange 1,262,800.3 1,363,193.3 Malta Stock Exchange 5,050.6 4,940.4
BSE Limited 2,088,431.4 2,185,195.6 Moscow Exchange 576,116.3 517,049.1
Bursa Malaysia 398,018.7 444,445.6 Muscat Securities Market 18,782.4 21,304.4
Chittagong Stock Exchange 38.2 42.8 Namibian Stock Exchange 2,461.9 2,915.2
Colombo Stock Exchange 15,575.0 15,903.3 Nasdaq Nordic Exchanges 1,322,817.5 1,465,045.4
Dhaka Stock Exchange 39,761.9 43,935.1 Nigerian Stock Exchange 31,520.5 36,536.8
Hanoi Stock Exchange 8,308.4 9,845.0 Oslo Stock Exchange 267,382.2 271,781.5
Hochiminh Stock Exchange 124,344.6 113,041.4 Palestine Exchange 3,734.9 3,891.5
Hong Kong Exchanges and Clearing 3,819,215.4 4,341,403.9 Qatar Stock Exchange 163,047.4 130,729.4
Indonesia Stock Exchange 486,765.9 488,768.3 Saudi Stock Exchange (Tadawul) 496,353.2 451,174.2
Japan Exchange Group 5,296,811.1 6,372,337.3 SIX Swiss Exchange 1,441,160.5 1,672,955.7
Korea Exchange 1,413,716.5 1,697,377.1 Stock Exchange of Mauritius 9,847.5 9,523.8
National Stock Exchange of India 2,056,337.3 2,157,607.3 Tehran Stock Exchange 145,970.8 108,634.7
NZX Limited 86,132.6 87,174.5 Tel-Aviv Stock Exchange 187,466.4 213,550.7
Philippine Stock Exchange 258,155.7 276,209.4 The Egyptian Exchange 41,940.9 46,148.2
Shanghai Stock Exchange 3,919,420.3 4,818,288.9 Trop-X (Seychelles) Limited 283.7 NA
Shenzhen Stock Exchange 2,405,459.5 3,428,556.5 Tunis Stock Exchange 8,329.0 8,922.6
Singapore Exchange 687,257.2 771,835.9 Ukrainian Exchange 4,415.4 5,198.0
Taipei Exchange 92,477.8 108,524.4 Warsaw Stock Exchange 170,230.2 210,181.5
Taiwan Stock Exchange 959,219.7 1,041,453.9 Wiener Borse 116,802.0 150,646.7
The Stock Exchange of Thailand 500,741.0 552,263.3 Zagreb Stock Exchange 20,509.0 22,764.8
25,918,988.4 30,317,402.7 16,570,037.4 18,981,390.2
Source: World Federation of Exchanges Overview 76,695,314.4 85,388,222.4
2
Number of companies
Number of listed companies Europe - Middle East - Africa
2018 Abu Dhabi Securities Exchange 70 67 3
Exchange Domestic Foreign Amman Stock Exchange 195 195 0
Total
companies companies Athens Stock Exchange 187 183 4
Bahrain Bourse 44 43 1
Americas Beirut Stock Exchange 10 10 0
B3 SA Brasil Bolsa Balcao 339 334 5 BME Spanish Exchanges 3,007 2,980 27
Barbados Stock Exchange 20 16 4 Borsa Istanbul 378 377 1
Bermuda Stock Exchange 51 13 38 Botswana Stock Exchange 35 26 9
Bolsa de Comercio de Buenos Aires 99 93 6 Bourse de Casablanca 76 75 1
Bolsa de Comercio de Santiago 285 205 80 BRVM 45 45 0
Bolsa de Valores de Colombia 68 66 2 Bucharest Stock Exchange 87 85 2
Bolsa de Valores de Lima 223 209 14 Budapest Stock Exchange 42 42 0
Bolsa de Valores de Panama 31 30 1 Cyprus Stock Exchange 102 91 11
Bolsa Mexicana de Valores 145 140 5 Deutsche Börse AG 510 462 48
Bolsa Nacional de Valores 10 10 0 Dubai Financial Market 67 51 16
Jamaica Stock Exchange 76 76 0 Euronext 1,208 1,059 149
Nasdaq - US 3,058 2,622 436 Iran Fara Bourse 114 114 0
NYSE 2,285 1,775 510 Irish Stock Exchange 54 43 11
TMX Group 3,383 3,330 53 Johannesburg Stock Exchange 360 289 71
10,073 Kazakhstan Stock Exchange 110 97 13
Ljubljana Stock Exchange 31 31 0
Asia - Pacific LSE Group 2,479 2,061 418
Australian Securities Exchange 2,146 2,004 142 Luxembourg Stock Exchange 162 27 135
BSE Limited 5,233 5,232 1 Malta Stock Exchange 25 25 0
Bursa Malaysia 912 902 10 Moscow Exchange 225 221 4
Chittagong Stock Exchange 282 282 0 Muscat Securities Market 110 110 NA
Colombo Stock Exchange 297 297 NA Namibian Stock Exchange 44 10 34
Dhaka Stock Exchange 311 311 0 Nasdaq Nordic Exchanges 1,019 974 45
Hanoi Stock Exchange 376 376 0 Nigerian Stock Exchange 164 163 1
Hochiminh Stock Exchange 373 373 0 Oslo Stock Exchange 237 237 0
Hong Kong Exchanges and Clearing 2,315 2,161 154 Palestine Exchange 48 48 0
Indonesia Stock Exchange 619 619 0 Qatar Stock Exchange 46 46 NA
Japan Exchange Group 3,657 3,652 5 Saudi Stock Exchange (Tadawul) 200 200 NA
Korea Exchange 2,207 2,186 21 SIX Swiss Exchange 270 236 34
National Stock Exchange of India 1,923 1,922 1 Stock Exchange of Mauritius 103 99 4
NZX Limited 137 131 6 Tehran Stock Exchange 323 323 0
Philippine Stock Exchange 267 264 3 Tel-Aviv Stock Exchange 449 421 28
Shanghai Stock Exchange 1,450 1,450 NA The Egyptian Exchange 250 249 1
Shenzhen Stock Exchange 2,134 2,134 NA Tunis Stock Exchange 82 82 0
Singapore Exchange 741 482 259 Ukrainian Exchange 80 78 2
Taipei Exchange 766 732 34 Warsaw Stock Exchange 852 823 29
Taiwan Stock Exchange 945 855 90 Zagreb Stock Exchange 127 127 0
The Stock Exchange of Thailand 704 704 NA 14,027
27,795
51,895
Overview 3
Value of share trading
Value of share trading. Electronic order book
(USD millions) Europe - Middle East - Africa
2018 Amman Stock Exchange 2,365.3 2,365.3 0.0
Exchange Domestic Foreign Athens Stock Exchange 12,715.1 12,254.7 460.4
Total
companies companies Bahrain Bourse 969.1 969.1 NA
Cboe Europe 2,812,273.2 0.0 2,812,273.2
Americas Beirut Stock Exchange 377.8 377.8 0.0
B3 SA Brasil Bolsa Balcao 814,126.8 811,250.2 2,876.7 BME Spanish Exchanges 647,383.8 643,067.1 4,316.6
Barbados Stock Exchange 11.4 11.2 0.2 Borsa Istanbul 403,214.2 403,093.3 120.9
Cboe Global Markets 16,036,009.8 16,036,009.8 0.0 Botswana Stock Exchange 177.1 158.4 18.7
Bermuda Stock Exchange 49.7 49.2 0.5 Bourse de Casablanca 3,963.5 3,962.7 0.9
Bolsa de Comercio de Buenos Aires 6,778.7 5,870.9 907.7 BRVM 475.9 475.9 0.0
Bolsa de Comercio de Santiago 46,970.0 46,895.6 74.4 Bucharest Stock Exchange 2,857.0 2,844.6 12.4
Bolsa de Valores de Colombia 14,353.6 14,317.4 36.2 Budapest Stock Exchange 10,216.2 10,216.2 0.0
Bolsa de Valores de Lima 3,376.6 2,003.8 1,372.8 Cyprus Stock Exchange 57.5 57.5 0.0
Bolsa de Valores de Panama 319.1 319.1 0.0 Deutsche Börse AG 1,815,087.8 1,666,358.0 148,729.7
Bolsa Mexicana de Valores 115,263.8 96,841.5 18,422.4 Dubai Financial Market 15,642.3 13,105.9 2,536.3
Bolsa Nacional de Valores 52.3 48.5 3.8 Euronext 2,200,719.9 2,200,560.6 159.3
Jamaica Stock Exchange 566.0 557.1 8.9 Iran Fara Bourse 8,845.7 8,845.7 0.0
Nasdaq - US 16,789,810.0 14,985,300.0 1,804,510.0 Irish Stock Exchange 33,118.1 33,080.0 38.0
NYSE 19,340,880.0 18,041,900.0 1,298,980.0 Johannesburg Stock Exchange 392,535.9 319,729.5 72,806.4
TMX Group 1,447,590.5 1,440,684.8 6,905.7 Kazakhstan Stock Exchange 1,534.2 1,405.8 128.4
54,616,158.2 Ljubljana Stock Exchange 386.7 386.7 0.0
LSE Group 2,541,532.0 2,343,350.0 198,182.0
Asia - Pacific Luxembourg Stock Exchange 96.3 81.9 14.3
Australian Securities Exchange 860,757.5 818,169.0 42,588.6 Malta Stock Exchange 102.2 102.2 0.0
BSE Limited 115,717.0 115,717.0 NA Moscow Exchange 166,525.9 162,481.5 4,044.5
Bursa Malaysia 136,760.8 136,514.7 246.0 Muscat Securities Market 1,916.2 1,916.2 NA
Chittagong Stock Exchange 1,039.6 1,039.6 0.0 Namibian Stock Exchange 902.0 66.9 835.1
Colombo Stock Exchange 671.9 671.9 NA Nasdaq Nordic Exchanges 847,989.7 813,979.6 34,010.1
Dhaka Stock Exchange 15,933.3 15,030.6 902.7 Nigerian Stock Exchange 2,598.0 2,581.8 16.2
Hanoi Stock Exchange 7,722.4 7,722.4 0.0 Oslo Stock Exchange 145,885.2 126,610.2 19,274.9
Hochiminh Stock Exchange 45,894.7 45,894.7 0.0 Palestine Exchange 353.5 353.5 0.0
Hong Kong Exchanges and Clearing 2,340,132.1 2,264,866.1 75,266.0 Qatar Stock Exchange 18,946.8 18,946.8 NA
Indonesia Stock Exchange 106,002.5 106,002.5 0.0 Saudi Stock Market (Tadawul) 229,668.1 229,668.1 NA
Japan Exchange Group 6,290,598.6 6,288,593.1 2,005.5 SIX Swiss Exchange 964,623.8 945,715.5 18,908.3
Korea Exchange 2,508,491.1 2,488,255.2 20,235.9 Stock Exchange of Mauritius 462.6 460.0 2.6
National Stock Exchange of India 1,163,820.3 1,163,793.9 26.4 Tehran Stock Exchange 27,846.8 27,846.8 0.0
NZX Limited 12,672.2 12,518.7 153.4 Tel-Aviv Stock Exchange 64,952.0 64,952.0 NA
The Philippine Stock Exchange 29,127.6 29,106.6 21.0 The Egyptian Exchange 14,489.4 14,488.5 0.8
Shanghai Stock Exchange 6,037,193.3 6,037,193.3 NA Tunis Stock Exchange 938.6 938.6 0.0
Shenzhen Stock Exchange 7,498,573.2 7,498,573.2 NA Ukrainian Exchange 12.1 12.1 0.0
Singapore Exchange 221,840.3 221,840.3 NA Warsaw Stock Exchange 56,843.0 56,362.9 480.1
Taipei Exchange 269,326.3 258,747.7 10,578.5 Zagreb Stock Exchange 254.3 254.3 0.0
Taiwan Stock Exchange 967,279.8 903,912.0 63,367.8 13,451,854.7
Stock Exchange of Thailand 388,004.6 388,004.6 NA

Overview
29,017,558.9 97,085,571.8
4
Trading

Trading process 5
Trading …
Traditionally, a buy-side initiated trade is placed as an order
with a broker, who then communicates it to a trader/dealer.
In turn, the dealer would trade against own inventory or
work the order on an exchange (pathway A)

Trading process 6
DMA
Direct market access is where the broker allows clients to
access their order routing infrastructure (pathway B)

Allows buy-side to issue electronic orders directly to

exchange, effectively giving them the same level of control
as the sell-side

Sponsored access takes this to the next level, for clients

whose high-frequency trading needs ultra-low latency
(pathway E)

Trading process 7
Algorithmic trading
Algorithmic trading is a computerized system responsible for
executing orders, rather than being worked by a trader
• Initially such algorithms were used by busy sell-side traders
(pathway C)
• Now clients use their own algorithms (pathway D)

Trading process 8
ECNs
Generally based on central limit order books with continuous
auctions

Came about due to a 1997 SEC ruling that private inter-

dealer prices should also be available to the public

ECN resembles an exchange but operates like an electronic

broker
• ECN must be registered as a broker/dealer or as a self-regulated
securities exchange

Trading process 9
MTFs
Multilateral trading facility are European counterparts of
ECNs
• An MTF must be registered with the relevant national regulator
before operations can commence

Chi-X is the major MTF in Europe

Trading process 10
ATS/Dark pools
Alternative trading systems provide opaque order-driven
trading platforms that aggregate pools of liquidity

Anonymously match buy and sell orders of a pool of

participants, generally institutional investors and broker-
dealers, at pre-specified times and at prices determined in
the primary market for the security

Trading process 11
Dark pools …
Advantages
• Confidential
• Low transaction costs
▪ Important for managing large blocks

Disadvantages
• No trading immediacy
• Darkness is “pre-trade” only
▪ Trades have to be ultimately reported
• There is no mechanism for price determination
▪ Prices for all trades are driven by best bid/offer
▪ Dark pools cannot exist without visible markets

Trading process 12
Market structure

Trading process 13
Price formation
Usually done through some form of fundamental valuation
analysis (discounted cashflow, multiples, etc.)

In a frictionless world, “excess buying” does not lead to

higher prices

Price formation 14
Price formation …
Market microstructure focuses of one particular aspect of this
process in the real world (with intermediaries), namely the
determination of the bid-ask spread

Price formation 15
Price
Bid quote is the highest price that someone is publicly willing
to pay

Ask or offer quote is the lowest price at which someone is

publicly willing to sell
• Depending on the market convention the bid and ask might be
indicative or firm commitments

Price formation 16
Spreads
The difference between the ask and bid quotes is the spread
• the spread is the cost incurred by someone who bought and
immediately reversed
• Alternatively, if the bid and ask quote were posted by a single
agent, the spread could represent the realized profit

Spreads are determined based on two factors

• Transaction cost component (inventory based models)
• Adverse selection component

Price formation 17
Spreads ...
Transaction cost component
 Assuming all traders know the asset value with certainty,
prices would bounce between bid and ask prices

 Spread is equal to the cost of doing business

• Normal business costs (office, technology etc.)
• Costs of maintaining an inventory (desired position is zero)
▪ When the inventory is too low, raise bid to encourage sellers (or increase the
bid size)
▪ When the inventory is too high, lower ask to encourage buyers (or increase the
ask size)

Price formation 18
Spreads …
Adverse selection component
 Some traders are better informed than others
• Ask price is set higher than the fundamental value to reflect the
fact that the next trade is an informed buyer
• Bid price is set lower than the fundamental value to reflect the fact
that the next trade is an informed seller

 Spread allows dealers to recoup from uninformed traders

what they lose to informed traders

Price formation 19
Types of orders
Instructions to the brokers on how to complete the order

 Market
• Executed immediately at current market prices

 Limit
• Specify prices at which willing to buy/sell

Price formation 20
Market order
Instruction to trade a given quantity at the best possible
price
• No specific price conditions
• Focus is on completing the order
Market orders demand liquidity
• Buy order will execute at the offer/ask price (101)
• Sell order will execute at the bid price (100)
• Immediate cost is half the bid-ask spread (0.5)

Price formation 21
Market order …
For orders larger than the current BBO size, market orders
“walk the book”
• Eg. A buy order for 2,000 may be crossed with S1 and S2

• If the order still can’t be completed, some venues (LSE) will cancel
it, while some (Euronext) will leave the residual on the order book

Price formation 22
Market order …
Risk is the execution price
• If suddenly S2 is cancelled, then our order will cross against S1 and
S3 raising the average execution price to 102.5

Sometimes market orders can achieve better prices than

expected
• Market order executes against a hidden order

Price formation 23
Limit order
Instruction to buy/sell a given quantity at a specified price or
better
• Buy limit order must execute below this price
• Sell limit order must execute above this price

If there are no orders that match an acceptable price, limit

order is left in the book till it expires or is canceled

Limit orders 24
Limit order book

Limit orders 25
Limit order …
Buy limit order for 1,000 shares at a price of 100 or better

Order cannot currently execute

Our order is left as B4, after B1 but before B2

Limit orders 26
Limit order …
Limit order sacrifices execution to ensure price
Compare limit order to buy 2,000 shares at a price of 101
versus a market order

Limit orders 27
Limit order …
 Limit orders provide liquidity
 Since traders can choose whether they want to trade with a
limit order, standing limit orders are options to trade (but
not option contracts)
• Sell limit orders are call options
• Buy limit orders are put options
 These options are freely granted to the public
• Option value of a limit order is the value of the order to other
traders
▪ Depends on limit price, how long the order will stand, and price volatility

Limit orders 28
Limit order markets
 All traders participate equally
 Traders publicly post their orders and the transaction price
is the result of the equilibrium of supply and demand
 All buy and sell orders are entered in a central order book
and a new order is immediately matched with the book of
limit orders previously submitted
• Limit order book is publicly visible
• Execution of market orders at bid/ask is not guaranteed
 Majority of the worldwide markets today are limit order
markets

Limit orders 29
Priority within the book
 Price
• A buy order with a relatively high price is said to be (relatively)
aggressive. (“The buyer is willing to pay more”)
• A sell order with a relatively low price is aggressive. (“The seller is
willing to accept less”)
• More aggressive limit orders have priority over less aggressive
orders
 Time
• An order that arrives earlier has priority over an order that arrives
later
 Other
• Visibility
• Type of trader
Limit orders 30
Hidden orders
A trader who submits a limit order can decide whether or not
it is to be displayed (the default) or hidden

A hidden order is held in the book and is available to be

executed

We can only see it when and if it is executed

• A hidden order is a form of dark liquidity

For orders at the same price, displayed orders have priority

over hidden orders
Limit orders 31
Spreads in limit order markets
Visible information in limit order book affects price formation
(spread)
• Market orders demand liquidity but pay bid/ask
• Limit orders provide liquidity but risk non-execution

Determinants of spreads
• Information asymmetry ↑
• Volatility ↑
• Limit order management costs ↑
• Value of trader time ↑

Limit orders 32
Margin trading
Using only a portion of the proceeds for an investment;
Borrow remaining component

Margin trades 33
Stock margin trading
Maximum margin is currently 50%; you can borrow up to
50% of the stock value

Maintenance margin: minimum amount equity in trading can

be before additional funds must be put into the account

Margin call: notification from broker you must put up

additional funds

Margin trades 34
Margin trading – initial conditions
GOOG $720
50% Initial Margin
30% Maintenance Margin
100 Shares Purchased

Initial Position
Stock $72,000 Borrowed $36,000
Equity $36,000

Margin trades 35
Margin trading – maintenance margin
Stock price falls to $500 per share

New Position
Stock $50,000 Borrowed $36,000
Equity $14,000

Margin% = $14,000/$50,000 = 28%

Margin trades 36
Margin trading – margin call
How far can the stock price fall before a margin call?

Equity / Investment = Margin

(100P – $36,000) / 100P = 30%

Solving, we get
P = 360/0.7 = $514.3

Margin trades 37
Margin trading – cover margin call
Newer position (raise cash)
Stock $50,000 Borrowed $36,000
Cash $1,000 Equity $15,000
Margin% = 15,000/50,000 = 30%

OR

Newer position (sell stock)

Stock $45,000 Borrowed $31,000
Equity $14,000
Margin% = 14,000/45,000 = 31%
Margin trades 38
Short sales
Purpose
• To profit from a decline in the price of a stock or security

Mechanics
• Borrow stock through a dealer
• Sell it and deposit proceeds and margin in an account
• Closing out the position: buy the stock and return to the party
from which is was borrowed

Margin trades 39
Short sale mechanics
Four guys: Amit, X, Y, and Z. One share currently trading at
$50 but falls to $40

Original Amit borrows Amit sells to Y Amit buys Amit returns

position from X at $50 from Z at $40 to X
Amit 0 1 $50 1, $10 $10
X 1 0 0 0 1
Y 0 0 1 1 1
Z 1 1 1 0 0

Amit makes a profit of $10. X (lender) retains her position

Margin trades 40
Short sale – initial conditions
FB 1,000 Shares
50% Initial Margin
30% Maintenance Margin
$120 Initial Price

Cash $120,000 Stock owed $120,000

T-bills $60,000 Equity $60,000

Margin trades 41
Short sale – maintenance margin
Stock Price Rises to $150

Cash $120,000 Stock owed $150,000

T-bills $60,000 Equity $30,000

Margin = (30,000/150,000) = 20%

Margin trades 42
Short sale – cover margin call
Add collateral (T-bills)
Cash $120,000 Stock owed $150,000
T-bills $80,000 Equity $50,000
Margin = (50,000/150,000) = 33.3%

OR

Buy stock
Cash $70,000 Stock owed $100,000
T-bills $60,000 Equity $30,000
Margin = (30,000/100,000) = 30%
Margin trades 43
Short sale – margin call
How much can the stock price rise before a margin call?

($180,000* − 1,000P) / (1,000P) = 30%

P = $138.5

* Initial margin plus sale proceeds

Margin trades 44
Other details
You pay interest on margin loan

You receive interest on short sale proceeds but you pay the
stock lender
• Net is called short sale rebate
• In exceptional circumstances, stock can go special and have
negative rebate

On borrowed stocks
• You have to pay the dividend (manufactured dividend) to the lender
• Who owns the voting rights?

Margin trades 45
Trading algorithm
Defines the steps required to execute an order in a specific
way
1. How should we slice the order?
2. How should each slice be managed?

Example: Order to buy 6,000 shares of asset ABC might have

an algorithm such as:
• Buy 1,000 shares every 10 minutes (and then sleep for 10 minutes)
• Order will get executed in one hour

Trading algorithms 46
TWAP
Time weighted average price benchmark is the average price

Algorithms that attempt to meet this benchmark are based

on a uniform time-based schedule
• Unaffected by other factors such as market price or volume

Example: Trade 10,000 shares of ABC

• Buy 500 every 15 minutes (takes 5 hours)
• Buy 1,000 every 15 minutes (takes 2.5 hours)

Trading algorithms 47
TWAP …

Trading algorithms 48
TWAP …
Trading leads to signaling risk
• Only thing other participants do not know is the total size of the
order

Can have poor execution quality

• Prices could become unfavorable
• Liquidity drops

Trading algorithms 49
VWAP
Volume weighted average price benchmark for a given time
span is the total traded value divided by total traded quantity
• Fair reflection of the market conditions throughout the day

VWAP = ∑
vn pn
n = ∑ n f n pn
∑ n vn

Large trades have more impact than small ones

Trading algorithms 50
VWAP …
In contrast to TWAP, we will have to trade in the correct
proportions
• These proportions are not known beforehand

Use historical volume profiles to estimate fn

Example: Buy 10,000 ABC

• Trading algorithm needs to place sufficient orders in each interval
to keep up with the target execution profile
• Not affected by actual market volume or price changes

Trading algorithms 51
VWAP …

Trading algorithms 52
VWAP …
Vulnerable to sudden shits in trading volume or liquidity

Some versions allow monitoring market conditions

• Makes them hybrid between static VWAP and dynamic volume
participation approach

Trading algorithms 53
Trading costs
Explicit costs
• Commissions, stamp duties, taxes

Implicit costs
• Caused by how trades move prices
• Dependent on order size, market liquidity for the security and the
speed of execution desired by the investor

Opportunity costs
• Loss (or gain) incurred as the result of delay in completion of, or
failure to complete in full, a transaction following an initial decision
to trade
Trading algorithms 54
Iceberg of trading costs

Direct Costs Commission

3-4 ¢ (9 bp)

Impact
Execution Shortfall
5 ¢ (15 bp)

Delay
Pre-trade Costs (PC)
10 ¢ (30 bp)
Missed Trades
4 ¢ (10 bp)

Source: Plexus Group Inc., 2007

Trading algorithms 55
High frequency trading
Generally refers to computer-driven trading strategies that
process information and react to market events very quickly
• Human reaction time: 200 millisecond
• Electronic message transmission, NY to Chicago, round trip: about
10 millisecond

Trading algorithms 56
An example

Trading algorithms 57
Need for speed

58
Flash crash of May 6th, 2010
Markets are more liquid because of HFTs
• But, not required by fiat to supply liquidity
• Because risk capital is finite, liquidity can dry up just when the
demand is high
▪ Even worse, suppliers can become demanders

Trading algorithms 59
Hedge Funds

1
HFs
 Private investment vehicles for wealthy individuals or
institutional investors
 Typically organized as limited partnerships
 Investors are limited partners
• Charged a performance-based fee
 Managers are general partners
• Invest a significant portion of their own wealth
• Potential payout can be significantly higher than the fixed
management fee

Introduction 2
Fees
Management fees
• Range from 1% to 3% of AUM
• Intended to meet operating expenses

Incentive fees
• Range from 15% to 25% of the annual realized returns
• Intended to encourage managers to achieve maximum returns

Others
• Hurdle rate
• High water marks

Introduction 3
Global industry (2015)

Introduction 4
Biggest hedge funds

Introduction 5
Strategies

Introduction 6
Shifts in styles

March 2013 Historical

Introduction 7
 Performance – Average returns

Sample Period:
Jan 1994 – Feb 2024

S&P 500 = 9.11%

Introduction 8
 Performance – Standard deviation

Sample Period:
Jan 1994 – Feb 2024

S&P 500 = 15.16%

Introduction 9
Performance – Correlation

Introduction 10
Long/Short strategy
Combined purchase and sale of two securities

Focus on equities

Also referred to as “equity hedge”

One of the most popular hedge fund strategies

Long/Short 11
Shorting advantages (1)
Ability to use negative information

Take S&P500
• 478 stocks had less than 1% weight (Dec 2010)
• In Dec 2010, only 3 stocks had weights > 2%
• Median weight is 0.09%
• If the manager thinks that this stock will decline by 10%
▪ Long constraint: Can only underweight it to 0%. This gives an additional
10%×0.09% = 0.9bps advantage
▪ Short allowed: −1% short position gives an additional 10bps advantage

Long/Short 12
Shorting advantages (2)
Shorts may provide more opportunities than longs

Short selling restrictions can lead to few investors searching

for overvalued stocks

Biased analyst recommendations

• Issue more buy than sell recommendations
• Reluctant to express negative opinions for keeping their line of
communication open with the management and to protect their
corporate finance business (underwriting etc.)

Long/Short 13
Mechanics (1)
A fund has a hypothetical initial equity capital of $1000

Two potential investments: undervalued stock A and

overvalued stock B

Long/Short 14
Mechanics (2)
Manager deposits $1000 with a custodial prime broker

Purchases $900 worth of stock A: his position with the broker

is long position in stock A for $900 and a long cash position
of $100

Long/Short 15
Mechanics (3)
Sells $800 worth of stock B: increases his cash position by
$800; since it is a short sale, need to borrow the shares

Primer broker borrows $800 worth of stock B: lender asks for

collateral and lending fee of say 1% per annum ($8)

Primer broker needs a liquidity buffer to take care of price

increases for stock B

Long/Short 16
Leverage
Fund’s assets
• $900 of stock A (long)
• $800 of stock B (short)
• $800 collateral
• $100 liquidity buffer

Total of $1700 (long + short) compared with its initial $1000

of equity capital

Fund has 90% long exposure and 80% short exposure.

170% (=90+80) gross exposure and 10% (=90−80) net
long exposure
Long/Short 17
Sources of returns (1)
Spread between the long and short position. Double alpha
strategy with alphas from both the undervalued and
overvalued stocks

Interest rebate on the proceeds from the short sale minus

the lending fee

Lending fee: small for liquid stocks but can be large for
stocks with small original issuance size and for stocks where
large blocks are held by unwilling lenders

Long/Short 18
Wrong bets
How do the managers reduce the consequences of wrong
bets?

 Diversify the portfolios on both the long and short sides –

typically 100 to 200 positions

 Set concentration limits, e.g. fix the maximum size for each
position to 5% of the total portfolio

Long/Short 19
Drawbacks
Higher trading costs
 Gross exposure is usually more than the initial capital ⟹
trading costs as a percentage of initial capital are higher
 In the example, fund would face trading costs on a $1700
position and a borrowing cost for $800 worth of shares
• To reduce the costs, the gross exposure can be made similar to that
of a long-only strategy, e.g., 50% of capital in each of the long and
short positions

Long/Short 20
Drawbacks (2)
Higher turnover
• Values of long and short positions change over time, leading to
rebalancing

Delays in execution
• Rules allow short sales only on an up tick (i.e., at a price higher
than the last traded price) or zero tick (i.e., at the same price as
the last traded price if that price is higher than the previous price)

Long/Short 21
Dedicated short (1)
Take more short positions than long positions and earn
returns by maintaining net short exposure in equities

Long/Short 22
Dedicated short (2)
Detailed individual company research typically forms the core
alpha generation driver of dedicated short bias managers,
and a focus on companies with weak cash flow generation is
common
• Weak financials but a high share price
• Change auditors or regularly delay their filings
• In industries with overcapacity
• Involved in failed merger
• Too high P/E ratios

Long/Short 23
Examples

Long/Short 24
Equity market neutral
Avoid having net long or net short exposure to have zero net
exposure to the market

Quintessential hedge fund strategy that promises true

absolute returns (alpha) without having to bear the market
sensitivity (beta)

Equity Neutral 25
Market neutrality (1)
How do we make a plain-vanilla long/short equity portfolio
with $10 million of initial capital market neutral? It has $9
million long and $6 million short shares

Dollar neutrality
• Equal dollar investments in the long and short positions ⟹ increase
the short position by $3 million
• Is such a portfolio truly market neutral?

Equity Neutral 26
Market neutrality (2)
Beta neutrality
• Equal betas of the long and short positions
• In the above example, if beta of long (short) position is 1.4 (0.7),
the net beta will be 0.35 (0.5×1.4 − 0.5×0.7)
• Need to double the size of the short position to make the portfolio
beta neutral

Equity Neutral 27
Market neutrality (3)
Sector neutrality
• Fund can lose if the long positions are concentrated in a sector that
goes down and short positions are in a sector that goes up
• Can balance the long and short positions in the same sector or
industry
• Can also use capitalization and value/growth attributes to ensure
there are no biases

Equity Neutral 28
Market neutrality (4)
Factor neutrality
• Ultimate and most quantitative step of equity market neutral
strategies
• Identify the factors that influence the returns of individual stocks

Equity Neutral 29
Examples
Pairs trading

Statistical arbitrage

High-frequency trading

Equity Neutral 30
Pairs trading
Most primitive form of equity market neutral strategy

Investment rules are simple:

• Find stocks whose prices move together
• Take a long/short position when their prices diverge sufficiently
• Hold the position until the two stock prices have converged, or a
stop level loss has been hit

Equity Neutral 31
Pairs trading …
Does not explicitly require market neutrality but is often
constrained to be at least dollar neutral by hedge funds that
implement it

Variety of approaches to execute

• Fundamental valuation: Long the undervalued stock and short the
overvalued stock in a sector
• Statistical: Use time-series models to identify pairs whose two
components deviate sufficiently

Equity Neutral 32
Pairs trading …
Statistical models: Use some sort of distance function to
measure the co-movements between the two securities

Simplest measure is the tracking variance – sum of squared

differences between the two normalized price series

Equity Neutral 33
Pairs trading …
Entry and exit rules

Open a position in a pair when prices diverge by more than

two historical standard deviations estimated during the 12-
month formation period

Close the position upon convergence or end of the six-month

period, whichever is earlier

Equity Neutral 34
Pairs trading example

Equity Neutral 35
Pairs trading concerns
Data snooping

Entry and exit rules should be based on sensible

assumptions, and not based on only back-testing and
simulations

In-sample optimal trading rule may not be optimal out-of-

sample

Equity Neutral 36
Pairs trading profitability
Gatev et al. (2006) find average annualized excess returns of
11%

Profits exceed the conservative estimates of transaction costs

Equity Neutral 37
Statistical arbitrage
Extension of the pairs trading approach to relative pricing

Due to irrational, historical, or behavioral factors, temporary

discrepancies can arise

Equity Neutral 38
Very-high-frequency trading
Strategies capturing very short-term opportunities, i.e.,
momentum lasting for a few minutes or seconds

Requires four elements

1. Brainpower (to design the trading rules or the learning
algorithms)
2. High-frequency historical data (to test the trading rules)
3. Computing power (to apply the selected trading rules in real time)
4. Best execution (to minimize transaction costs and slippage)

Equity Neutral 39
James Simons’ Medallion Fund

Equity Neutral 40
HFT

Cost: $1.5b

Equity Neutral 41
Flash crash

Equity Neutral 42
Merger arbitrage
Attempt to capture the spreads in merger or acquisition
transactions involving public companies after the terms of
the transaction have been announced

The spread is the difference between the transaction bid and

the trading price

Typically, the target stock trades at a discount to the bid in

order to account for the risk of the transaction not closing
successfully

Merger Arbitrage 43
Example
 On Oct 5, 1999, WorldCom Inc (WCOM) and Sprint Inc.
(FON) formally announced their plans to merge in a stock-
for-stock exchange
 Exchange deal would have given approximately $76 in
WCOM stock for each share of FON held
 Shares of FON, never reached $76 during the entire deal
period
• FON peaked to a high of only $75.32 briefly on Nov 9, 1999
 Why did the spread persist?

Merger Arbitrage 44
Example …
Since any potential gain from the spread would not be
realized until months later when shares were to be
exchanged at closing, investors demanded a time value of
money premium as a compensation for committing capital to
the arbitrage investment for the deal period

Merger Arbitrage 45
Example …
Investors were concerned about whether the merger, which
faced serious anti-trust regulatory hurdles, would actually
close

Completion risk discouraged investors from bidding FON up

to $76
• If the true probability of completion is higher than that implied by
the spread, arbs can earn a premium

Merger Arbitrage 46
Basic principle

Merger Arbitrage 47
Merger arbitrage versus Insider trading
Arbitrageurs invest only in publicly announced transactions

Merger arbitrageurs provide liquidity to those who don’t want

to wait to see the deal going through

Some arbitrageurs may trade on unannounced transactions

based on rumors. This is not the typical merger arbitrage
strategy

Merger Arbitrage 48
Cash tender offer
Acquirer offers a fixed amount of cash in exchange for a
share of target

Premium with respect to the current share price of the

target:
• 30-50% at a date 30-90 days before the announcement
• 5-15% at the time of announcement
▪ Shrinkage due to “astute” analyst trading in anticipation of M&A as well as
insider trading

Merger Arbitrage 49
Cash tender offer …
Target’s market price moves up after the announcement but
does not reach the bid price

Arbitrage spread = % difference between the initial bid price

and the target’s closing price on the day after the
announcement

Merger Arbitrage 50
Cash tender offer …
Successful acquisition: arbitrage spread converges to zero

Strategy: Purchase the target stock at the announcement of

the takeover and hold it until the end of the offer period

If the deal fails, the arbitrageur must sell the target stock at
a loss

Merger Arbitrage 51
Example
Mar 22, 1999: Atlanta-based First Data Corp. offered $25.50
in cash for each share of Paymentech Inc.

Paymentech Inc. shares closed at $24 on Mar 22

Peak in trading between Mar 22nd and 24th due to the risk
arbitrage activity

Merger Arbitrage 52
Example …

Return of $1.50 (25.50−24) or 6.25% in 4 months

Merger Arbitrage 53
Risk
Market risk of long positions
• If equity market collapses in the middle of a deal, the long holdings
will fall in value, and the likelihood of the deal going through is
greatly reduced

To hedge the risk, arbitrageur can sell short equity index

futures. Usually done at the portfolio level rather than on a
transaction by transaction basis
• May not work as the companies involved in a merger may behave
differently from the overall equity market

Merger Arbitrage 54
Stock-for-stock offer
Bidder offers a fixed quantity of its own common stock in
exchange for a fixed quantity of target shares

Slightly more complicated case than the cash offer as the

reference price for target is not fixed and depends on the
bidder’s stock price

Merger Arbitrage 55
Stock-for-stock offer …
No longer sufficient to only buy the target stock as the
bidder’s stock may fall significantly once the merger is
completed

Necessary to consider the relative evolution of both stocks to

establish the arbitrage position

Merger Arbitrage 56
Stock-for-stock offer …
Spread between the two companies involved in merger is
expected to narrow in relative terms

Bidder’s stock price is expected to fall relative to the target’s

stock price

Strategy: Buy the target stock and sell short the bidder’s
stock ⟹ hedge the market risk in the process

Merger Arbitrage 57
Example
Sept 15, 1999: Microsoft announced it would acquire Visio
Corp using a fixed exchange ratio of 0.45 shares of Microsoft
for each share of Visio

Visio shares closed at $39.875 and Microsoft closed at

92.625

According to merger terms, Visio share was worth

0.45×92.625 = $41.681

Spread = $41.681−39.875 = $1.806

Merger Arbitrage 58
Example …
Strategy: Buy Visio shares and sell short 0.45 shares of
Microsoft for each Visio share

Merger Arbitrage 59
Assessing risk
Double price risk: If the transaction is called off, the arb
faces price risk on both long and short positions

Acquirer’s stock price (which has been shorted) could rise,

and the target’s stock price (which is held long) can fall to
pre-deal trading levels

Merger Arbitrage 60
Assessing risk …
Pre-deal prices may not be necessarily limited to prices at
which investments is made
• Prices immediately preceding the announcement may reflect a
premium as a result of rumors

A busted transaction might result in price levels far worse

than those at which arb established the positions
• Existence of leverage amplifies these losses

Merger Arbitrage 61
Assessing risk …
 Change in the terms: If the exchange ratio is altered or a
change is made in the nature of securities offered, original
arbitrage positions may have to be unwound or rebalanced
at a high transaction cost
 Dramatic increase in acquirer’s price: This will deteriorate
profit in the short position and may threaten the deal if the
acquirer tries to renegotiate better terms
 Dramatic decrease in acquirer’s price: This may cause
target’s shareholders to re-evaluate the value they are
receiving, threatening the deal

Merger Arbitrage 62
Assessing risk …
 Competing bids: This situation may result in a liquidity
problem as investors rush to cover short positions in the
event that the original acquirer is not successful
• Can be positive for the arb since they increase the chances of a
higher price on the target
 Material adverse change clause: Most definitive merger
agreements include clauses that allow the acquirer to walk
away if a material adverse change arises in the target’s
business prior to merger completion

Merger Arbitrage 63
Assessing risk …
 Tax approval: IRS approval for attempted tax-free
reorganizations is not guaranteed
• An adverse ruling could dramatically change the economics of the
situation
 Regulatory agency interference: In addition to plain
blocking the deal, this could lengthen the time to
completion, which may reduce the arb’s return

Merger Arbitrage 64
Cash plus stock
Dec 14, 2007: Ingersoll-Rand will acquire all outstanding
common stock of Trane

Holders of Trane’s will receive $36.50 in cash and 0.23

Ingersoll-Rand shares of common stock per each Trane share

Trane’s shares closed at $45.24 while those of Ingersoll-Rand

closed at $43.60

Merger Arbitrage 65
Cash plus stock …
According to merger terms, Trane share was worth
0.23×43.60 = $10.028

Spread = ($10.028+36.50) − $45.24 = $1.288

Strategy: Buy Trane shares and sell short 0.23 shares of

Ingersoll-Rand for each Trane share

Merger Arbitrage 66
Risks in merger arbitrage
Merger arbitrage is essentially a bet on whether the merger
will be successful or not

Two risks
• Transaction risk: likelihood of the transaction going through
• Calendar risk: Timing of the deal completion

Median probability of successful consummation of mergers is

89%

Merger Arbitrage 67
Transaction risk
 Acquirer’s attitude: Friendly negotiated offer is 20 times
more likely to succeed compared to hostile bid (B&Y 2003)
 Type of deal: Success rate higher for flexible stock-for-stock
exchanges (93%) compared to cash and fixed stock-for-
stock exchanges (87% and 88%)
 Takeover premium: Higher the premium, better the chances
that the deal goes through
 Ownership structure: If target company has more
arbitrageurs as shareholders, probability of success goes up

Merger Arbitrage 68
Transaction risk …
 Bidder’s toehold: Positively related to success
 Target management attitude: Friendly attitude relates
positively to success
 Number of arbitrageurs involved: Positively related to
success
 Presence of anti-trust considerations: Negatively related to
success
 Economic conditions: Deteriorating economy is negatively
related to success

Merger Arbitrage 69
Calendar risk
Uncertainty relative to the time that will elapse between the
announcement and the consummation of the merger (if the
deal goes through)

Deal with larger spreads involve longer period between

announcement and consummation

Higher spreads consistent with greater uncertainty about

final resolution

Merger Arbitrage 70
A failed deal
October 2000: GE announced intention to buy Honeywell
International in a stock-for-stock transaction valued at $45b
• 1.055 shares of GE for each share of Honeywell

Merger Arbitrage 71
Other considerations
Hold diversified portfolios by spreading bets over several
deals
• Excessive diversification can hurt performance because the
profitability of best deals is diluted

Setting position limits as well as strict stop losses for each

transaction to limit risk

Merger Arbitrage 72
Other considerations …
Arbitrageurs can take reverse positions if they believe that
deal will fail
• Long acquirer and short target in the stock deal

Merger arbitrage is not a buy-and-hold strategy

• Change the positions as the likelihood of favorable outcome
changes

Merger Arbitrage 73
Hedge fund performance
In the early years, performance was measured
• Against peers, or
• Against an absolute benchmark

Focus on recent years has moved to performance beyond

what simple, easily replicable hedge fund factors can produce

Performance and Replication 74

An example
Eight-year (Jan 1992 to Dec 1999) track record of CDP (Capital

Decimation Partners), LP

CDP S&P500
Monthly Mean (%) 3.7 1.4
Monthly Std Dev (%) 5.8 3.6
Minimum month (%) -18.3 -8.9
Maximum month (%) 27 14
Annual Sharpe ratio 1.94 0.98
Number of negative months 6/96 36/96
Correlation with S&P500 0.599 1
Total return (%) 2721.3 367.1

Performance and Replication 75

CDP
Investment strategy is shorting out-of-money puts on S&P
500 on each monthly expiration for maturities less than or
equal to three months, and with strikes approximately 7%
out-of-money

Returns are positive most of the time

Losses are infrequent

• When they occur, they are extreme

Risk not well summarized by stdev

Performance and Replication 76
Which factors?
Fung and Hsieh (2004, FAJ) Agarwal and Naik (2004, RFS)
 Equity  Equity
• S&P 500 • R1000
• Small-cap (R2000) • MSCI World ex-US and Emerging
• Emerging Markets Markets
 Bond • SMB, HML, and MOM
• 10-year Treasury  Bond
• BAA bonds • U.S. Govt. and Corp.
 Option • World Govt.
• High yield
• Bond straddles
• Currency straddles  Other
• Commodity straddles • Currency and Commodity
 Options on S&P 500
• ATM calls and puts
• OTM calls and puts
Performance and Replication 77
How does the alpha look like?

Performance and Replication 78

Hedge fund replication
 Fund-of-funds

 Multi-strategy

 Hedge fund replication

Performance and Replication 79

Fund-of-funds
Advantages
 Less onerous liquidity rules
 Lower minimum investment
 Access to niche or closed strategies
 Value creation by choosing skilled hedge fund managers

Disadvantages
 Additional layer of fees
 Unclear if there is more diversification

Performance and Replication 80

Multi-strategy funds
Advantages
 No additional fees
 Better liquidity than individual funds
 Access to underlying managers

Disadvantages
 No real diversification

Performance and Replication 81

Factor-based replication (1)
Advantages
 Give exposure to hedge fund beta through liquid
investments
 Low fees
 No issues of liquidity or minimum investment

Disadvantages
 Backward looking
• Do not account for dynamically changing strategies
 No alpha

Performance and Replication 82

Factor-based replication (2)
At time 𝑡𝑡, run the following regression using historical data
on hedge fund index return
𝑅𝑅𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽𝑖𝑖′ 𝐹𝐹𝑖𝑖 + 𝑒𝑒𝑖𝑖𝑖𝑖 , 𝑠𝑠 = 𝑡𝑡 − 𝑘𝑘, … , 𝑡𝑡

Create the portfolio 𝛽𝛽̂𝑖𝑖′ 𝐹𝐹

The predicted return for time 𝑡𝑡 + 1 is given by

𝐸𝐸 𝑅𝑅𝑖𝑖𝑖𝑖+1 = 𝛽𝛽̂𝑖𝑖′ 𝐹𝐹𝑖𝑖+1
• In back testing, one can compare actual return 𝑅𝑅𝑖𝑖𝑖𝑖+1 to the
predicted return 𝐸𝐸 𝑅𝑅𝑖𝑖𝑖𝑖+1 and see how well the model does

Performance and Replication 83