International Capital Markets

and Investment Practice

Part 5
Forecasting returns and conditional asset pricing

Dr. Peter Oertmann

[email protected]

This lecture is largely congruent in content with the video

recording of the 2020 lecture available on lecturio.de
Part 5: Forecasting returns and conditional asset pricing

Economic conditions and asset returns

Specification of a forecasting model
Exploring the predictable variation of stock and bond market returns
Introduction into conditional asset pricing models
Predicting the time-variation of risk premiums
Global market integration

© 2018-2022 Peter Oertmann 2

Part 5: Positioning in the lecture

Major fields in asset management Knowledge base

Diversification Asset allocation Stylized facts on Sources of risk

global markets and return
Empirical research Models and empirical
research

Security Investment Asset pricing Portfolio

selection strategy Theories, models and construction
empirical tests
implementation Models and empirical
tests

Risk Trading and Financial Time series

management execution economics analysis
Models and empirical Models and empirical
research procedures

© 2018-2022 Peter Oertmann 3

Part 5: Forecasting returns and conditional asset pricing

Economic conditions and asset returns

Specification of a forecasting model
Exploring the predictable variation of stock and bond market returns
Introduction into conditional asset pricing models
Predicting the time-variation of risk premiums
Global market integration

© 2018-2022 Peter Oertmann 4

Equity performance and the business cycle – returns

Source: MSCI data through 2000, Dahlquist/Harvey (2001)

© 2018-2022 Peter Oertmann 5

Equity performance and the business cycle – volatility

Source: MSCI data through 2000, Dahlquist/Harvey (2001)

© 2018-2022 Peter Oertmann 6

Equity performance and the business cycle – corellations with world

Source: MSCI data through 2000, Dahlquist/Harvey (2001)

© 2018-2022 Peter Oertmann 7

Mechanics of the yield curve

© 2018-2022 Peter Oertmann 8

The TERM spread reflects economic growth expectations

Source: Dahlquist/Harvey (2001)

© 2018-2022 Peter Oertmann 9

Average monthly stock returns and the US yield curve

Source: MSCI data through 2000, Dahlquist/Harvey (2001)

© 2018-2022 Peter Oertmann 10

Findings of Dahlquist and Harvey (2001)

§ In the US, average equity returns during recessions are about one third of the
level during expansions
§ It is impossible to hide from a U.S. recession
§ In almost every country there is a huge difference between expansion and
recession average returns, i.e. other countries seem to be more sensitive to the
US business cycle than the US equity return
§ Volatility is greater during US recessions in almost every country
§ Correlations are higher in US recessions

© 2018-2022 Peter Oertmann 11

Survey article by Wheelock and Wohar (2009)
Can the term spread predict output growth and recessions?

Research questions
§ Does the yield spread forecast output growth?
§ Does it forecast recessions?
Conclusion (quotes)
§ The answer to both questions is a qualified “yes”
§ USA: Every U.S. recession since 1953 was preceded by a large decline in the
yield on 10-year Treasury securities relative to the yield on 3-months Treasury
securities, and several recessions were preceded by an inversion of the yield
curve.
§ Germany: Germany experienced recessions beginning in 1966, 1974, 1980,
1991, 2000, and 2008. All but the 1966 recession were preceded by a sharp
decline in long-term Treasury security yields relative to short-term yields that
resulted in a flat or inverted yield curve. The only inversion that was not followed
by a recession occurred in 1970.
§ UK: (similar findings)
Source: Wheelook/Wohar (2009)

© 2018-2022 Peter Oertmann 12

 Survey article by Wheelock and Wohar (2009)
A look at the data
ock and Wohar Wheelock an

US term spread and recessions Figure 2 German term spread and recessions
ure 1
1953-2008
Term Spread and Recessions
1960-2008
German Term Spread and Recessions

Percent Percent
5 6

4
4

3
2

2
0
1

–2
0

–1 –4

–2 –6
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
20 3
19 3

19 3

20 9
19 3

19 9

19 1

19 3

19 9

01
73

19 9

09
19 9

81

05
65

95
61

85

91
19 5

19 7
19 7
55

20 7
19 7
19 7

19 7

0
6

9
5

8
7
6

8
7

0
6
5

20
19

19
19

19
19

19

20
19
19

19

Spread Between 10-Year Government Bond Yield and 3-Month Treasury Bill Yield
Spread Between 10-Year and 3-Month Treasury Security Yields

NOTE: The term spread is calculated as the difference between the yields on 10-year and 3-month Treasury securities. The sha
The term spread is calculated as the difference between the yields on 10-year and 3-month Treasury securities. The shaded
areas denote recessions as determined by the Economic Cycle Research Institute.
denote recessions as determined by the National Bureau of Economic Research.

Source: Wheelook/Wohar (2009)

m spread is useful for predicting recessions. tively.

Figure 1 shows the Germany
difference experienced
between the recessions beginning quarterly average of monthly observations
ermore, the relationship appears robust to in 1966,
yields on 10-year and 3-month 1974, 1980, 1991,
U.S. Treasury 2000, and 2008. All
secu- the year-over-year percentage change in rea
© 2018-2022 Peter Oertmann 13
clusion of other variables and nonlinearities rities for 1953-2008. The shaded regions indicatepreceded by a sharp
but the 1966 recession were domestic product (GDP) for the United Sta
forecasting model. recession periods asdecline
definedinbylong-term Treasury security yields rela-
the National Germany, and the United Kingdom. The ta
tive to short-term
2 yields that resulted in a flat or presents the contemporaneous correlation b
 Wheelock and Wohar

Survey article by Wheelock and Wohar (2009)

Table 1
Correlation of GDP growth and lagged and futures term spreads
Correlation of GDP Growth and Lagged and Future Term Spreads by Country
Lagged term spread
Term
t (t – 1) (t – 2) (t – 3) (t – 4) (t – 5) (t – 6)

United States –0.0449 0.0999 0.2557 0.3605 0.4141 0.3957 0.3196

(0.5047) (0.1379) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001)
Germany –0.0003 0.1641 0.2991 0.3689 0.3845 0.3649 0.3421
(0.9970) (0.0455) (0.0002) (0.0001) (0.0001) (0.0001) (0.0001)
United Kingdom 0.0723 0.1816 0.2486 0.3025 0.3379 0.3166 0.2607
(0.3319) (0.0144) (0.0008) (0.0001) (0.0001) (0.0001) (0.0005)

Future term spread

Term
t (t + 1) (t + 2) (t + 3) (t + 4) (t + 5) (t + 6)

United States –0.0449 –0.1428 –0.2374 –0.2994 –0.3372 –0.3538 –0.3421

(0.5047) (0.0335) (0.0004) (0.0001) (0.0001) (0.0001) (0.0001)
Germany –0.0003 –0.1722 –0.3414 –0.4424 –0.4548 –0.4545 –0.4110
(0.9970) (0.0357) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001)
United Kingdom 0.0723 –0.0364 –0.1366 –0.2116 –0.2306 –0.2204 –0.2261
(0.3319) (0.6244) (0.0652) (0.0040) (0.0017) (0.0001) (0.0021)

NOTE: U.S. data are for 1953:Q1–2008:Q4; German data are for 1973:Q1–2008:Q2 (West Germany, 1973-1991); U.K. data are for
1958:Q1–2008:Q2. Numbers in parentheses represent p-values.
Source: Wheelook/Wohar (2009)

ties, which
there is no universally agreed-upon theory as to © 2018-2022 causes their yields to fall relative to
Peter Oertmann 14
why a relationship between the term spread and current yields on short-term securities.
economic activity should exist. To a large extent, Many studies attribute the apparent ability
Early study by Fama and French (1989)

Source: Fama/French (1989)

© 2018-2022 Peter Oertmann 15

Early study by Fama and French (1989)

Explanatory variables, X(t)

§ Dividend yield
– D(t)/P(t) on the value-weighted NYSE portfolio
– Foundation: Dow (1920), Ball (1978), Rozeff (1984), Shiller (1984), …
§ Term spread
– Difference between the time t yield on the Aaa bond portfolio and the one-
month bill rate, TERM(t)
– Earlier work: Fama (1976, 1984), Keim/Stambaugh (1986), …
§ Default spread
– Difference between the time t yield on a portfolio of 100 corporate bonds and
the Aaa yield, DEF(t)
– Earlier work: Fama (1986), Keim/Stambaugh (1986)

Source: Fama/French (1989)

© 2018-2022 Peter Oertmann 16

Early study by Fama and French (1989)

Source: Fama/French (1989)

© 2018-2022 Peter Oertmann 17

Early study by Fama and French (1989)

Source: Fama/French (1989)

© 2018-2022 Peter Oertmann 18

Early study by Fama and French (1989)

Conclusions
§ The three variables forecast stock and bond returns
§ The variation in expected returns is largely common across securities, and is
negatively related to long- and short-term variation in business conditions
§ When business conditions are poor, income is low and expected returns on
bonds and stocks must be high to encourage substitution from consumption to
investment – when times are good and income is high, the market clears at lower
levels of expected returns

The general message is that expected returns

are lower when economic conditions are strong
and higher when conditions are weak.

Source: Fama/French (1989)

© 2018-2022 Peter Oertmann 19

Part 5: Forecasting returns and conditional asset pricing

Economic conditions and asset returns

Specification of a forecasting model
Exploring the predictable variation of stock and bond market returns
Introduction into conditional asset pricing models
Predicting the time-variation of risk premiums
Global market integration

© 2018-2022 Peter Oertmann 20

Modelling asset returns

Traditional approach
§ Random walk with constant drift
§ The price at time t-1 plus a drift is the best forecast for the price at time t

Pt = drift + Pt -1 + e t
§ The return is equal to a constant drift plus “white noise” – not forecastable!

R t = drift + e t

More general approach

§ Random walk with time-varying drift
§ Lagged information can be used to forecast a part of the drift

R t = d 0 + d1 (Z t -1 ) + e t

© 2018-2022 Peter Oertmann 21

Conditioning information

Full information set

f t -1 = (j1, t -1, j2, t -1,..., j k, t -1) Î Â k

§ all information on the state of the global economy

§ publicly available at time t-1

Information set applied for return conditioning

Z t -1 = (Z1, t -1, Z2, t -1,..., Zh, t -1) Ì f t -1

§ instrumental variables representing the global information set

§ observable at time t-1

© 2018-2022 Peter Oertmann 22

Specification of an instrumental forecasting model

Linear regression of observed returns on lagged instruments

Rit = w i0 + w i1 × Z1,t -1 + w i 2 × Z 2,t -1 +...+ w ik × Zk,t -1 + eit

where Z j,t -1 level of the jth instrument at time t - 1

w ij impact of the jth instrument on the return of the ith asset
w i0 constant component of the ith asset return

How to explore predictability

§ Analyze estimates and t-values for w-coefficients
§ Calculate the R-square

© 2018-2022 Peter Oertmann 23

Specifying instrumental variables

Starting point
§ Be devotedly aware of the data mining problem
§ Instruments must make economic sense
§ Fundamental model of asset prices is a good “anchor point”

Dividend Discount Model (DDM)

Model to determine the “intrinsic value” of a stock

E(D t +1 )
Pr ice t =
d - E(g)
where
D t +1 next dividend
d risk adjusted discount factor
g dividend growth rate

© 2018-2022 Peter Oertmann 24

Searching for instrumental variables

… related to the time-variation of

§ Dividends
– Expected corporate earnings and cash flows
– Expected dividend growth

§ Discount factor
– Risk free interest rate
– Risk exposure of the firm
– Expected risk premium for risk exposure

© 2018-2022 Peter Oertmann 25

Economic categories for instrumental variables

§ Inflation
§ Business expectations
§ Default risk
§ Fundamental valuation of firms
§ Microstructure of market
§ World market integration
§ Political risk
§ Momentum
§ Investors’ sentiment

© 2018-2022 Peter Oertmann 26

Inflation

Goal: Measure expected inflation as one of the most pervasive forces affecting
business decisions

E(D t +1 )
Effect on dividends possible? YES
Pr ice t =
d - E(g)
Effect on discount factor possible? YES

Instrument candidates
§ Interest rates
§ Term structure of interest rates
§ Pricing of inflation-linked bonds
§ …

© 2018-2022 Peter Oertmann 27

Business expectations

Goal: Measure expected economic growth

E(D t +1 )
Effect on dividends possible? YES
Pr ice t =
d - E(g)
Effect on discount factor possible? (YES)

Instrument candidates
§ Expected GDP growth
§ Capacity utilization
§ Money supply aggregates
§ Housing starts
§ Term structure of interest rates
§ …

© 2018-2022 Peter Oertmann 28

Default risk

Goal: Measure confidence in corporate solvency

E(D t +1 )
Effect on dividends possible? (YES)
Pr ice t =
d - E(g)
Effect on discount factor possible? YES

Instrument candidates
§ Credit spreads (BBB-AAA)
§ Credit insurance premiums
§ …

© 2018-2022 Peter Oertmann 29

Microstructure of the market

Goal: Measure the liquidity of the market or a market segment

E(D t +1 )
Effect on dividends possible?
Pr ice t =
d - E(g)
Effect on discount factor possible? YES

Instrument candidates
§ Trading volume
§ Market capitalization concentration ratios
§ Volatility
§ …

© 2018-2022 Peter Oertmann 30

Investors’ sentiment

Goal: Measure the investors’ attitude towards capital market risk

E(D t +1 )
Effect on dividends possible? (YES)
Pr ice t =
d - E(g)
Effect on discount factor possible? YES

Instrument candidates
§ Sentiment indices
§ Consumer confidence measures
§ Put/Call ratio
§ …

© 2018-2022 Peter Oertmann 31

Part 5: Forecasting returns and conditional asset pricing

Economic conditions and asset returns

Specification of a forecasting model
Exploring the predictable variation of stock and bond market returns
Introduction into conditional asset pricing models
Predicting the time-variation of risk premiums
Global market integration

© 2018-2022 Peter Oertmann 32

Application of a predictive regression

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 33

Application of a predictive regression – global instruments

§ Global dividend yield (iDYG7): GDP-weighted average of aggregate DY in the

G-7 stock markets, the dividend-to-price (D/P) ratio at time t-1 is the average
value of dividends paid from month t-12 to t-1, divided by the respective market
index value at t-1.
§ Global inflation rate (iING7): GDP-weighted average of inflation rates in the G-7
countries, expressed in a log form.
§ Global term spread (iTSG7): GDP-weighted average difference between the
yield on long-term government bonds and the 1-month Eurocurrency short-term
interest rate for the G-7 countries.
§ U.S. default spread (iDSUS): Difference between the yield on U.S. low-grade
corporate bonds and the U.S. (default-free) government bond yield with the same
maturity.
§ TED spread (iTEDS): Difference between the 3-month Eurodollar rate and the
90-day yield on the U.S. Treasury bill.

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 34

Regression results – international stock markets

© 2018-2022 Peter Oertmann 35

Regression results – international stock markets (cont.)

© 2018-2022 Peter Oertmann 36

Regression results – international bond markets

© 2018-2022 Peter Oertmann 37

Results for stock markets

§ Excess returns on international stock markets are to some extent predictable by

the global instruments specified.
§ R-square:
– Average: 0.053
– 0.028 (Austria) to 0.095 (Japan)
– Relatively high R-square for Belgium (0.086), France (0.071), Hong Kong
(0.078), and the Netherlands (0.070)
§ Coefficients’ signs
– Global dividend yield (iDYG7): positive
– Global inflation rate (iING7): negative
– Global term spread (iTSG7): negative
– U.S. default spread (iDSUS): mixed signs
– TED spread (iTEDS): mostly negative

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 38

Results for bond markets

§ The predictable variation of international bond market returns seems to be related

only to some few of the global instruments prespecified here.
§ R-square:
– Average: 0.071
– 0.043 (U.K.) to 0.117 (USA)
– For Canada, Germany, the Netherlands, Switzerland, and the United States,
the fraction of return variance captured is larger than with the stock market
§ Coefficients’ signs
– Global dividend yield (iDYG7): positive
– Global inflation rate (iING7): not significant
– Global term spread (iTSG7): mixed signs
– U.S. default spread (iDSUS): negative
– TED spread (iTEDS): not significant

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 39

Part 5: Forecasting returns and conditional asset pricing

Economic conditions and asset returns

Specification of a forecasting model
Exploring the predictable variation of stock and bond market returns
Introduction into conditional asset pricing models
Predicting the time-variation of risk premiums
Global market integration

© 2018-2022 Peter Oertmann 40

Conditional asset pricing

Goal
§ Exploitation of the relationship between business conditions and expected asset
returns across the universe of assets
§ Application of observable information variables (instruments) to model time-
variation in global risk premiums in pricing relationships

Some milestone studies

§ Keim & Stambaugh (1986)
§ Campbell (1987)
§ Gibbons & Ferson (1989)
§ Harvey (1991), Ferson/Harvey (1991, 1993)
§ Ilmanen (1995)

© 2018-2022 Peter Oertmann 41

Ferson and Harvey (1991)
The variation of economic risk premiums

Motivation
§ Evidence that stock and bond returns are to some extent predictable
§ Debate on the reason for that predictability
– Market inefficiencies?
– Changes in required returns?
§ Attempt to calibrate the relative importance of these two explanations

Approach
§ Assuming a rational asset pricing model: Expected returns of assets are related
to their sensitivity to changes in the state of the economy
§ Sensitivity is measured by the assets’ beta coefficients
§ For each of the relevant state variables there is a market-wide risk premium
(increment to the expected return per unit of beta)
§ Focus on the time-series behavior of the risk premiums
Source: Ferson/Harvey (1991)

© 2018-2022 Peter Oertmann 42

Ferson and Harvey (1991)

Source: Ferson/Harvey (1991)

© 2018-2022 Peter Oertmann 43

Ferson and Harvey (1991)

Conclusions
§ Much of the predicted variation of monthly excess returns of common stock
portfolios is associated with their sensitivity to economic variables
§ The risk premium associated with exposure to a stock market index captures the
largest component of the predictable variation in stock returns
§ Risk premiums associated with term structure shifts and default spreads are the
most important for fixed-income securities
§ Time-variation in the premium for beta risk is more important than changes in the
betas

Our findings strengthen the evidence that the

predictability of returns is attributable to time-
varying, rationally expected returns.

Source: Ferson/Harvey (1991)

© 2018-2022 Peter Oertmann 44

Pricing restriction of a conditional beta pricing model

E(rit f t -1 ) = bi1 × l1t (f t -1 ) + bi 2 × l 2 t (f t -1 ) + ... + bik × l kt (f t -1 )

where E(rit f t -1 ) ith asset' s excess return expected for period t

based on informatio n set at time t - 1
l1t (f t -1 ) jth factor risk premium expected for period t
based on informatio n set at time t - 1
bij exposure of the ith asset' s return to the jth factor
f t -1 informatio n set at time t - 1

© 2018-2022 Peter Oertmann 45

Modeling time-varying risk premiums

l jt (Z t -1) = w j0 + w j1 × Z1, t -1 + w j2 × Z2, t -1 +... + w jh × Zh, t -1

where Z v , t -1 level of the vth instrument at time t - 1

w jv impact of the vth instrument on the jth risk premium
w j0 constant component of the jth risk premium

Assumption of …
§ a linear relationship between instrument levels and risk premium levels
§ a constant long-rum risk premium mean

© 2018-2022 Peter Oertmann 46

Part 5: Forecasting returns and conditional asset pricing

Economic conditions and asset returns

Specification of a forecasting model
Exploring the predictable variation of stock and bond market returns
Introduction into conditional asset pricing models
Predicting the time-variation of risk premiums
Global market integration

© 2018-2022 Peter Oertmann 47

Recall the result of our unconditional pricing test

L L

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 48

Specification of a global conditional beta pricing model

System of pricing equations

æ r1t ö æ b11 & b13 ö æ w10 & w15 ö æ Z0 , t -1 ö æ b11 & b13 ö æ d1t ö æ e1t ö
ç ÷ ç ÷ ç ÷ ç ÷ ç ÷ ç ÷ ç ÷
ç% ÷ = ç% ÷ × ç% ÷ × ç% ÷ + ç% ÷ × ç% ÷ + ç% ÷
çr ÷ ç b &b ÷ ç w & w ÷ ç Z ÷ ç ÷ ç ÷ çe ÷
è nt ø è$n!#1 n3 ø è 30 35 ø è 5, t -1 ø è bn1 & bn3 ø è d 3 t ø è$#nt ø
!" $!!!!#!!!! " $!#!" $#" "
factor time - varying risk premia factor factor residual
exposures exposures returns returns
$!!!!!!
!#!!!!!!!
" $!!!!!#!!!!!
"
conditiona lly expected returns unexpected returns

§ 3 global sources of risk

– WDSTR, ILG7C, CHG7C
§ 5 instruments to model time-variation in the 3 risk premiums
§ Joint estimation of factor exposures and w-coefficients
Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 49

Set of instruments

§ Global dividend yield (iDYG7)

§ Global inflation rate (iING7)
§ Global term spread (iTSG7)
§ U.S. default spread (iDSUS)
§ TED spread (iTEDS)

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 50

We are interested in the Ω matrix

Ω coefficients describe the

conditional time-variation
of global risk premiums

æ r1t ö æ b11 & b13 ö æ w10 & w15 ö æ Z0 , t -1 ö æ b11 & b13 ö æ d1t ö æ e1t ö
ç ÷ ç ÷ ç ÷ ç ÷ ç ÷ ç ÷ ç ÷
ç% ÷ = ç% ÷ × ç% ÷ × ç% ÷ + ç% ÷ × ç% ÷ + ç% ÷
çr ÷ ç b &b ÷ ç w & w ÷ ç Z ÷ ç ÷ ç ÷ çe ÷
è nt ø è$n!#1 n3 ø è 30 35 ø è 5, t -1 ø è bn1 & bn3 ø è d 3 t ø è$#nt ø
!" $!!!!#!!!! " $!#!" $#" "
factor time - varying risk premia factor factor residual
exposures exposures returns returns
$!!!!!!
!#!!!!!!!
" $!!!!!#!!!!!
"
conditiona lly expected returns unexpected returns

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 51

Ω matrix for stock markets

Determinants of time-varying factor risk premia

l! ( Z )
jt t -1
w-coefficients
Global instruments WDSTR ILG7C CHG7C
Constant -0.013 0.039 ** -0.012 *
-0.785 1.800 -1.329
iDYG7 4.672 *** 1.412 -0.348
4.583 1.047 -0.616
iING7 -2.423 *** -0.806 0.438 *
-4.581 -1.151 1.494
iTSG7 -1.001 *** -1.461 *** -0.049
-2.153 -2.374 -0.192
iDSUS 0.064 -2.273 *** -0.269
0.090 -2.415 -0.683
iTEDS -2.817 *** -3.250 *** 0.946 **
-3.150 -2.745 1.906
GMM chi-square 89.538
p-value 0.319 Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 52

Ω matrix for bond markets

Determinants of time-varying factor risk premia

l! ( Z )
jt t -1
w-coefficients
Global instruments WDSTR ILG7C CHG7C
Constant -0.032 0.072 *** 0.001
-0.888 2.420 0.135
iDYG7 -0.535 -2.623 ** 0.130
-0.271 -1.625 0.231
iING7 0.223 0.845 0.253
0.231 1.068 0.916
iTSG7 1.815 ** -1.328 ** -0.522 **
1.893 -1.692 -1.903
iDSUS 2.206 * -2.670 *** -1.603 ***
1.389 -2.024 -3.476
iTEDS 0.968 0.721 0.533 *
0.657 0.597 1.263
GMM chi-square 23.146
p-value 0.809 Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 53

Time-variation of the world stock market risk premium

15.00%

10.00%

5.00%

0.00%

-5.00%

-10.00%

-15.00%
Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan
82 83 84 85 86 87 88 89 90 91 92 93 94 95

Stock markets Bond markets

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 54

Time-variation of the global interest rate risk premium

15.00%

10.00%

5.00%

0.00%

-5.00%

-10.00%

-15.00%
Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan
82 83 84 85 86 87 88 89 90 91 92 93 94 95

Stock markets Bond markets

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 55

Time-variation of the global FX risk premium

15.00%

10.00%

5.00%

0.00%

-5.00%

-10.00%

-15.00%
Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan Jan
82 83 84 85 86 87 88 89 90 91 92 93 94 95

Stock markets Bond markets

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 56

Part 5: Forecasting returns and conditional asset pricing

Economic conditions and asset returns

Specification of a forecasting model
Exploring the predictable variation of stock and bond market returns
Introduction into conditional asset pricing models
Predicting the time-variation of risk premiums
Global market integration

© 2018-2022 Peter Oertmann 57

Concept of market integration

„Markets are integrated if assets with the same risk in terms of an exposure
to common systematic global risk factors have the same expected returns
irrespective of the market in which they are traded“

Interpretation
§ Sources of global risk are the same across international markets
§ Rewards for global risks are the same in each market
§ If the factor risk premiums were different across markets, an investor could
increase the expected return of his portfolio without altering her risk exposure by
simply investing in those countries that provide higher rewards for the same risk
(case of market segmentation)

© 2018-2022 Peter Oertmann 58

Integration can be measured

... across
§ Countries
§ Geographic regions
§ Industrial sectors
§ Emerging (converging) and developed markets
§ Stock and bond markets

© 2018-2022 Peter Oertmann 59

How can integration be measured?

Indicative measures
§ Correlation between financial market returns
§ Volatility spillovers between financial markets
§ Factor exposures of financial markets

Theoretically correct measures

§ Global risk premiums
– Significance (= Existence)
– Ability to price (explain) the cross-section of international assets (= Pricing
error)
§ Correlation between expected returns

© 2018-2022 Peter Oertmann 60

Starting points

Increasing global market integration implies an ...

§ increasing comovement of the returns on financial markets
– Analysis of return correlations
§ increasing influence of global factors driving returns
– Analysis of markets‘ factor exposures
§ increasing comovement of expected returns on financial markets
– Analysis of the correlation of expected returns
§ improving fit of global asset pricing models
– Analysis of pricing consistency
§ increasing comovement of global risk premiums across asset classes
– Analysis of the comovement of global risk premiums

© 2018-2022 Peter Oertmann 61

Analysis of pricing consistency

Major steps
§ Identification of factors driving returns
§ Estimation of global risk premiums and tests of significance
§ Analysis of the magitude of pricing errors

Calculation of pricing errors (PE) across assets

æ PE1 ö æç R1 ö÷ é æ b11 ö æ b12 ö æ b1k ö ù

ç ÷ ê ç ÷ ç ÷ ç ÷ ú
ç% ÷ ç% ÷ ê ç% ÷ ç% ÷ ç% ÷ ú
ç PE ÷ = ç Ri ÷ - êR + ç b ÷ × l + ç b ÷ × l + ... + ç b ÷ × l ú
ç i ÷ ç ÷ ê f ç i1 ÷ 1 ç i2 ÷ 2 ç ik ÷ k ú
ç% ÷ ç% ÷ ê ç% ÷ ç% ÷ ç% ÷ ú
ç PE ÷ çç ÷÷ ê çb ÷ çb ÷ çb ÷ ú
è n ø è Rn ø ë$!! è! n1 ø
!!!! è! n2 ø
! #!!!!! è! nk ø
!!! "û
Risk premium based expected return

© 2018-2022 Peter Oertmann 62

Analysis of pricing consistency

Mean pricing error for international asset returns

Model alternatives
§ Global CAPM
– MSCI world market return
§ Global 3-factor model
– MSCI world market return
– G7 long-term interest rate changes
– FX rate changes of G7 currencies vs CHF
Setting
§ Viewpoint of a Swiss investor
§ Simultaneous estimation of factor exposures and risk premiums
§ 3 sub-periods

© 2018-2022 Peter Oertmann 63

Analysis of pricing consistency

Mean pricing error for international asset returns

82.02 - 86.12 87.01 - 89.12 90.01 - 95.02
Stock markets CAPM 3-factor CAPM 3-factor CAPM 3-factor
Australia 0.464 0.148 0.960 0.146 0.049 0.560
Austria 1.137 0.812 1.038 0.426 0.608 0.216
Belgium 1.249 0.658 0.313 0.006 0.089 0.063
Canada 0.984 0.642 0.214 0.216 0.663 0.116
Denmark 0.282 0.015 1.493 0.343 0.376 0.074
France 0.865 0.048 0.129 0.129 0.336 0.126
Germany 1.128 0.463 0.581 0.349 0.172 0.109
Hong Kong 0.274 0.631 1.727 0.143 1.378 0.532
Italy 0.963 0.169 1.010 0.118 0.787 0.613
Japan 0.726 0.174 0.593 0.282 0.997 0.549
Netherlands 0.673 0.315 0.029 0.053 0.353 0.200
Norw ay 0.198 0.324 0.042 0.005 0.406 0.490
Spain 0.714 0.488 0.079 0.172 0.508 0.088
Sw eden 0.871 0.502 0.245 0.475 0.106 0.654
Sw itzerland 0.652 0.113 0.771 0.087 0.518 0.203
United Kingdom 0.028 0.204 0.187 0.116 0.001 0.151
United States 0.685 0.146 0.450 0.186 0.130 0.282
Median 0.714 0.315 0.450 0.146 0.376 0.203

© 2018-2022 Peter Oertmann 64

Analysis of pricing consistency

Mean pricing error for international asset returns

82.02 - 86.12 87.01 - 89.12 90.01 - 95.02

Band markets CAPM 3-factor CAPM 3-factor CAPM 3-factor
Canada 0.129 0.083 0.479 0.404 0.550 0.394
France 0.448 0.245 0.097 0.180 0.010 0.109
Germany 0.538 0.069 0.165 0.141 0.105 0.032
Japan 0.469 0.095 0.104 0.110 0.182 0.205
Netherlands 0.439 0.095 0.208 0.157 0.050 0.076
Sw itzerland 0.097 0.082 0.002 0.024 0.076 0.119
United Kingdom 0.226 0.278 0.080 0.009 0.174 0.002
United States 0.159 0.055 0.731 0.666 0.374 0.205
Average 0.333 0.089 0.135 0.149 0.140 0.114

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 65

Analysis of the comovement of global risk premiums

Example
Correlations between global risk premiums in stock and bond returns

Risk premiums
Market Interest rate FX rate

1982-1995 0.077 0.468 0.751

Sub-periods
1982-1986 -0.218 0.465 0.696
1987-1989 (crash period) -0.030 0.509 0.796
1990-1995 0.511 0.360 0.918

Source: Oertmann (1997)

© 2018-2022 Peter Oertmann 66

Integration and asset pricing

§ Segmented markets
– Tests using assets from one country
– CAPM: Sharpe (1964), Lintner (1965), Black (1965)
§ Perfectly integrated markets
– Tests on the cross-section of international assets
– World CAPM with FX risk: Dumas (1994), Dumas/Solnik (1995)
– World multibeta models: Solnik (1983); Ferson/Harvey (1993, ...)
§ Mild market segmentation
– Assuming a certain degree of segmentation
§ Time-varying market integration
– Allowing the degree of integration to change through time
– Bekaert/Harvey (1995)
– Bekaert/Harvey/Lundblad/Siegel (2011)

© 2018-2022 Peter Oertmann 67

Multifactor models – an overview

Model Input and Output Applications

Multifactor model IN Time series of returns and • Analyze the risk profile of assets
k factor changes • Explain return variance
Rit = ai + å bij × d jt + eit OUT Factor sensitivities (β‘s) • Play scenarios
j =1

Multifactor asset pricing model IN Risk-free rate, assets’ β‘s • Generate (a vector of) expected
k and risk premiums returns as an input for Strategic
E(Ri ) = R f + å bij × l j OUT Cross-sectional consistent Asset Allocation (SAA)
j =1 expected returns

Instrumental forecasting model IN Time series of returns and • Forecast asset returns
k lagged instruments
Rit = w i1 + å w ij × Z j,t -1 + eit OUT Instrument sensitivities (w’s)
j =1

Conditional asset pricing model IN Risk-free rate, assets’ β‘s • Generate (a vector of)
k and risk premiums at t-1 conditionally expected return as
E t (Ri ) = R ft + å bij × l jt (Z t -1) OUT Cross-sectional consistent an input for Tactical Asset
j =1 expected returns at time t-1 Allocation (TAA)

© 2018-2022 Peter Oertmann 68

Part 5: Selected references

Bekaert, G. and C. R. Harvey (1995), Time-varying world market integration, The Journal of Finance
Bekaert, G. et al. (2011), What segments equity markets, working paper, Duke University
Dahlquist, M. and C. Harvey (2001), Global Tactical Asset Allocation, The Journal of Global Capital Markets
Fama, E. F. and K. R. French (1989), Business conditions and expected returns on stock and bonds, Journal
of Financial Economics
Ferson, W. and C. Harvey (1991), The variation of economic risk premiums, Journal of Political Economy
Ferson, W. and C. Harvey (1993), The risk and predictability of international equity returns, The Review of
Financial Studies
Heston, S. L., Rouvenhorst, K. G. and R. E. Wessels (1995), The structure of international stock returns and
the integration of capital markets, Journal of Empirical Finance, 2, 173-197.
Oertmann, P. (1997), Global Risk Premia on International Investments, Gabler.
Oertmann, P. and H. Zimmermann (1997), Wieviel Noise erträgt ein Prognosemodell für die taktische Asset
Allokation?, Finanzmarkt und Portfolio Management, Vol. 11.
Wheelook, D.C. and M.E. Wohar, Can the Term spread predict output growth and recessions? A survey of
the literature, Federal Reserve Bank of St. Louis Review

© 2018-2022 Peter Oertmann 69