CAPM issues in international investment

The Cost of International Capital P.
Sercu, International Finance: Theory into Practice Overview
Chapter 18
Estimating the Cost of International Capital
Overview
The Cost of International Capital P. Sercu, International Finance: Theory into Practice Overview
Translate rst, then discountor vv? Two procedures When to do what? The Single-Country CAPM From Asset Returns to Portfolio Return The tangency solution How the weights affect mean and variance How to make a portfolio efcient The Market Portfolio as the Benchmark The International CAPM Why do we need an InCAPM? Why Xrisk pops up in the InCAPM Do assets have a nancial nationality? Aggregating the Efciency Conditions The InCAPM Wrapping up
Overview
Overview
Overview
CAPM issues in FDI

The Cost of International Capital P. Sercu, International Finance: Theory into Practice Translate rst, then discountor vv? The Single-Country CAPM The International CAPM Wrapping up
Suppose an Australian rm considers an investment in India. Issues: Can we assess NPV in INR, as locals would do?
would look simpler no? ... but works only if AU and IN are both (part of) an integrated market
How does a multi-country market work, where investors think in different currencies depending on where they live?
risks and returns depend on the currency they are measured in, ... which violates the homogenous expectations assumption of standard CAPM
CAPM issues in FDI

CAPM issues in FDI

Outline
The Cost of International Capital P. Sercu, International Finance: Theory into Practice Translate rst, then discountor vv?
Two procedures When to do what?
The Single-Country CAPM The International CAPM Wrapping up
Two facts need reconciliation:

CFs are probably rst computed in FC (INR):

cost data in
INR ;
local sales priced to market

INR ,
all these prices are probably sticky in terms of
not
AUD
Yet commonly used CoCa is in HC (AUD) So either we translate the cashows into AUD or we shift to an INR CoCa


cost data in
INR ;

INR ,
not
AUD


cost data in
INR ;

INR ,
not
AUD
Two procedures
When in Rome, act like the Romans?

set CoCA on basis of INR risk-free rate, and add risk premium computed from INR stock returns (risk, price of risk) compute
FCPV PV
if desired, convert
into
HC
... or ... My shareholders consume/think in AUD ?

translate expected CFs into
AUD ,
including the covariance
set CoCA on basis of AUD risk-free rate, and add risk premium computed from AUD stock returns (risk, price of risk) compute
PV
in
HC
Two procedures
When in Rome, act like the Romans?

set CoCA on basis of INR risk-free rate, and add risk premium computed from INR stock returns (risk, price of risk) compute
FCPV PV
if desired, convert
into
HC
... or ... My shareholders consume/think in AUD ?

translate expected CFs into
AUD ,
including the covariance
set CoCA on basis of AUD risk-free rate, and add risk premium computed from AUD stock returns (risk, price of risk) compute
PV
in
HC
When to do what?
Does the translation/discounting procedure matter?

In practice: the local version looks easier (but only if we cut corners, as well see) In principle: no if mkts are integrated, yes if mkts arent Australians care about the value to them, not about how Indians would value the project. The two are not the same in segmented markets. We can see the Australians required return in their own ` lAustralienne capital market. So we value a
Can there be a CAPM-type no-PPP equilibrium?

presence of (real) exchange risk means expectations cannot be homogenous: (real) expected returns and risks for asset j differ across investors But this special type of investors heterogeneity is easily incorporated (InCAPM instead of CAPM)
When to do what?


When to do what?


When to do what? Overview

CoCa model 1. Foreign investments:

currency of calculations
home and host nancially integrated home and host nancially segmented home country part of larger nancial market home country totally isolated
inCAPM inCAPM
CAPM
FC HC HC
& HC only only
2. domestic investments

home country part of larger nancial market home country totally isolated
inCAPM
CAPM
HC HC
only only
Outline
The Cost of International Capital P. Sercu, International Finance: Theory into Practice Translate rst, then discountor vv? The Single-Country CAPM
From Asset Returns to Portfolio Return The tangency solution How the weights affect mean and variance How to make a portfolio efcient The Market Portfolio as the Benchmark
The International CAPM Wrapping up
From Asset Returns to Portfolio Return

Key relationstated in excess return terms, herebetween portfolio return rp and asset returns rj & weights xj : rp r =
N X j=1
xj ( rj r )
Example
time0 data and decisions Vj,0 nj nj Vj,0 xj 100 4 400 0.40 50 4 200 0.20 25 12 300 0.30 =900 =0.90 +100 =1000 +0.10 =1.00 time1 result Vj,1 nj Vj,1 120 480 70 280 20 240 (excess) rates of return rj rj r xj (rj r) 0.20 0.15 0.060 0.40 0.35 0.070 0.20 0.25 0.075 =0.055 +0.050 =0.105
risky :
j 1 2 3
subtotal riskfree total 0
105 =1105
From Asset Returns to Portfolio Return

Key relationstated in excess return terms, herebetween portfolio return rp and asset returns rj & weights xj : rp r =
N X j=1
xj ( rj r )
Example
time0 data and decisions Vj,0 nj nj Vj,0 xj 100 4 400 0.40 50 4 200 0.20 25 12 300 0.30 =900 =0.90 +100 =1000 +0.10 =1.00 time1 result Vj,1 nj Vj,1 120 480 70 280 20 240 (excess) rates of return rj rj r xj (rj r) 0.20 0.15 0.060 0.40 0.35 0.070 0.20 0.25 0.075 =0.055 +0.050 =0.105
risky :
j 1 2 3
subtotal riskfree total 0
105 =1105
Two assetsone risk-free, one risky

rp = x rs + (1 x)r0 = r0 + x ( rs r0 ) E( rp ) = r0 + x E( rs r0 ), sd( rp ) = |x| sd( rs )
x 6
E( rp ) 6
s 1 E( rs ) ........................................ q . . . . . . . . . . . . . . . . . . . 0 r0 . . . . . . . . . . . . . sd( rs )
sd( rp )
Two assetsone risk-free, one risky

rp = x rs + (1 x)r0 = r0 + x ( rs r0 ) E( rp ) = r0 + x E( rs r0 ), sd( rp ) = |x| sd( rs )
x 6
E( rp ) 6
s 1 E( rs ) ........................................ q . . . . . . . . . . . . . . . . . . . 0 r0 . . . . . . . . . . . . . sd( rs )
sd( rp )
Two imperfectly correlated risky assets

rp = x1 r1 + (1 x1 ) r2 ; ( E( rp ) = E( r2 ) + x [E( r1 ) E( r2 )], p 2 sd( rp ) = x1 var( r1 ) + 2x1 (1 x1 )cov( r1 , r2 ) + (1 x1 )2 var( r2 )

E( rp ) 6
sd( rp )
Many risky assets and one risk-free

many risky assets:

bound is similar, but most portfolios are now inside the bound only portfolios on the upper half o/t bound are efcient
add a risk-free asset

best risky portfolio is the tangency one:
x 6
E( rp ) 6

t 1 E( rt ) ................... .q . . . . . . . . . . . . . . . . . . 0 r0 . . . . . . . . . . . . . sd( rt )
sd( rp )
Many risky assets and one risk-free

many risky assets:

bound is similar, but most portfolios are now inside the bound only portfolios on the upper half o/t bound are efcient
add a risk-free asset

best risky portfolio is the tangency one:
x 6
E( rp ) 6

t 1 E( rt ) ................... .q . . . . . . . . . . . . . . . . . . 0 r0 . . . . . . . . . . . . . sd( rt )
sd( rp )
How weights affect mean and variance (1)

From rp = r +
N rj j=1 xj (
r):
r+
N X j=1 N X j=1
E( rp )
xj E( rj r ) ,
var( rp )
xj
N X k=1
xk cov( rj , rk ).
Understanding the variance formula: portfolio variance is a weighted average of each assets covariance with the entire portfolio: N N X X cov( rp , rp ) = cov( xk rk , rp ) = xk cov( rk , rp ) .
k=1 k=1
each of these assets portfolio covariances is, in turn, a weighted average of the assets covariance with all components of the portfolio: N N X X cov( rj , rp ) = cov( rj , xk rk ) = xk cov( rj , rk ).
k=1 k=1

From rp = r +
N rj j=1 xj (
r):
r+
N X j=1 N X j=1
E( rp )
xj E( rj r ) ,
var( rp )
xj
N X k=1
xk cov( rj , rk ).
k=1 k=1
each of these assets portfolio covariances is, in turn, a weighted average of the assets covariance with all components of the portfolio: N N X X cov( rj , rp ) = cov( rj , xk rk ) = xk cov( rj , rk ) .
k=1 k=1

From rp = r +
N rj j=1 xj (
r):
r+
N X j=1 N X j=1
E( rp )
xj E( rj r ) ,
var( rp )
xj
N X k=1
xk cov( rj , rk ).
k=1 k=1
each of these assets portfolio covariances is, in turn, a weighted average of the assets covariance with all components of the portfolio: N N X X cov( rj , rp ) = cov( rj , xk rk ) = xk cov( rj , rk ) .
k=1 k=1

Example:
E( rj r) 0.200 0.122 cov( rj , r1 ) 0.16 0.05 cov( rj , r2 ) 0.05 0.09 xj 0.50 0.40
1 2
E( rp r) cov( r1 , rp ) cov( r2 , rp ) cov( rp , rp )
= = = =
0.50 0.200 + 0.40 0.122 = 0.1488 0.50 0.160 + 0.40 0.050 = 0.1000 0.50 0.050 + 0.40 0.090 = 0.0610 0.50 0.100 + 0.40 0.061 = 0.0744
Marginal effect of xj on mean and variance

Three-asset example: Asset 1s marginal contribution to portfolio expected excess return is its own expected excess return: E( rp r ) E( rp r ) x1 = = x1 E( r1 r) + x2 E( r2 r); E( r1 r).
Asset 1s marginal contribution to portfolio variance is (twice) its covariance with the portfolio return: var( rp ) var( rp ) x1
Proof:
= =
2 2 x1 var( r1 ) + 2x1 x2 cov( r1 , r2 ) + x2 var( r2 );
2cov( r1 , rp ).
= = = = 2x1 var( r1 ) + 2x2 cov( r1 , r2 ), 2[x1 cov( r1 , r1 ) + x2 cov( r1 , r2 )], 2[cov( r1 , x1 r1 ) + cov( r1 , x2 r2 )], 2cov( r1 , x1 r 1 + x2 r2 ),
var( rp ) x1

Proof:
= =
2cov( r1 , rp ).
var( rp ) x1

Proof:
= =
2cov( r1 , rp ).
var( rp ) x1
Effect of w on var( rp ): example

Example
1 2 E( rj r) 0.200 0.122 cov( rj , r1 ) 0.04 0.05 cov( rj , r2 ) 0.05 0.09
portfolio 1: x1 = .40, x2 = .40, x0 = .20 cov( r1 , rp ) = 0.. . . . . 0.. . . . . + 0.. . . . . 0.. . . . . = cov( r2 , rp ) cov( rp , rp ) = 0.. . . . . 0.. . . . . + 0.. . . . . 0.. . . . . = = 0.. . . . . 0.. . . . . + 0.. . . . . 0.. . . . . =
0.036 000, 0.056 000, 0.036 800,
0.036 400, 0.056 500, 0.037 524,
.037 524 .036 800 .000724 var = = = 0.0724 2 cov( r1 , rp ) . x1 0.01 0.01

Example
0.036 000, 0.056 000, 0.036 800,
0.036 400, 0.056 500, 0.037 524,
var .037 524 .036 800 .000724 = = = 0.0724 2 cov( r1 , rp ) . x1 0.01 0.01

Example
0.036 000, 0.056 000, 0.036 800,
0.036 400, 0.056 500, 0.037 524,
var .037 524 .036 800 .000724 = = = 0.0724 2 cov( r1 , rp ) . x1 0.01 0.01

Example
0.036 000, 0.056 000, 0.036 800,
0.036 400, 0.056 500, 0.037 524,
var .037 524 .036 800 .000724 = = = 0.0724 2 cov( r1 , rp ) . x1 0.01 0.01
How to make a portfolio efcient

Micro economics: budget allocation problem has as efciency condition that marginal utilityj = same across all goods j marginal costj Mean-Variance: the good side is not utility but expected return; and the bad side is variance. So the efciency condition is js contribution to E( rp ) = same across all assets j js contribution to var( rp )
E( rj r ) = , for all risky assets j=1, ... N cov( rj , rp )


(In)Efciency and Risk Aversion: examples

Asset 1 Asset 2
E( rj r ) 0.092 0.148
(co)variances cov( r1 , r1 ) = 0.04 cov( r1 , r2 ) = 0.05 cov( r2 , r1 ) = 0.05 cov( r2 , r2 ) = 0.09
portfolio with weights x1 = .40, x2 = .40, x0 = .20 is not efcient: cov( r1 , rp ) = 0.. . . . . 0.. . . . . + .. . . . . 0.. . . . . = 0.036 cov( r2 , rp ) = 0.. . . . . 0.. . . . . + .. . . . . 0.. . . . . = 0.056,
0.. . . . . 0.. . . . .
= 2.555 =
0.. . . . . 0.. . . . .
= 2.643
portfolio with weights x1 = .40, x2 = .60, x0 = .00 is efcient: cov( r1 , rp ) = 0.. . . . . 0.. . . . . + .. . . . . 0.. . . . . = cov( r2 , rp ) = 0.. . . . . 0.. . . . . + .. . . . . 0.. . . . . =
0.. . . . . 0.. . . . .
0.046 0.074,
=2=
0.. . . . . 0. . . . . .
0.. . . . . 0.. . . . .
0.023 0.037,
=4=
0.. . . . . 0. . . . . .

Asset 1 Asset 2
E( rj r ) 0.092 0.148
0.. . . . . 0.. . . . .
= 2.555 =
0.. . . . . 0.. . . . .
= 2.643
0.. . . . . 0.. . . . .
0.046 0.074,
=2=
0.. . . . . 0. . . . . .
0.. . . . . 0.. . . . .
0.023 0.037,
=4=
0.. . . . . 0. . . . . .

Asset 1 Asset 2
E( rj r ) 0.092 0.148
0.. . . . . 0.. . . . .
= 2.555 =
0.. . . . . 0.. . . . .
= 2.643
0.. . . . . 0.. . . . .
0.046 0.074,
=2=
0.. . . . . 0. . . . . .
0.. . . . . 0.. . . . .
0.023 0.037,
=4=
0.. . . . . 0. . . . . .
The Market Portfolio as the Benchmark

Homogeneous expectations and opportunities:

Peters tangency portfolio is the same as Pauls and Marys So everybody holds a combination of the tangency portfolio and riskfree assets So the aggregate portfolio of all investors is a combination of the tangency portfolio and riskfree assets, with a weight that reects the average risk-aversion In a closed economy, this aggregate portfolio must be the portfolio of all existing shares (the market portfolio).
Bottom line: the market portfolio must be efcient too; so j : E( rj r ) = m , cov( rj , rm ) E( rj r) = m cov( rj , rm ).
(1)


(1)


(1)


(1)


(1)
Identifying m the CAPM

Piece of cake. If j : xj ):
m =
E( rj r ) , cov( rj , rm )
then for any portfolio p (with weights r) E( rp r) cov( rp , rm )
Pn m = Pn
rj j=1 xj E(
rj , rm ) j=1 xj cov(
Now pick the market portfolio as our p. Then m = Substitute: E( rj r) = = = m cov( rj , rm ), E( rm r ) cov( rj , rm ), var( rm ) j E( rm r), E( rm r ) E( rm r) = . cov( rm , rm ) var( rm )
with j as in rj = j + j rm + j , the market model. j rm is the undiversiable risk or market risk, diversiable risk.
j
the rm-specic or idiosyncratic or

m =
Pn m = Pn
rj j=1 xj E(
with j as in rj = j + j rm + j , the market model. j rm is the undiversiable risk or market risk, diversiable risk.
j

m =
Pn m = Pn
rj j=1 xj E(
rm is the with j as in rj = j + j rm + j , the market model. j undiversiable risk or market risk, diversiable risk.
j

m =
Pn m = Pn
rj j=1 xj E(
rm is the with j as in rj = j + j rm + j , the market model. j undiversiable risk or market risk, diversiable risk.
j
The CAPM for dummies: just using words

The beta is a measure of the assets relative riskthat is, the assets market covariance risk cov( rj , rm ), rescaled by the portfolios total risk, var( rm ). Beta can be estimated from the market-model regression. A risky asset with beta equal to zero should have an expected return that is equal to the risk-free rate, even if the assets return is uncertain. The reason is that the marginal contribution to the total market risk is zero. If an assets beta or relative risk is non-zero, the assets expected return should contain a risk premium. The additional return that can be expected per unit of beta is the markets expected excess return above the risk-free rate.


The CAPM for yuppies: APT interpretation

You can always avoid a stocks idiosyncratic risk:

form js shadow portfolio j: invest a weight j in m, and (1 j ) riskfree now compare the returns on j and on j: rj r j = = j + j rm + j ,
(1 j ) r + j rm .
You cannot seriously expect to be rewarded for a risk that you can easily avoid. So j and j have the same expected return:
E( rj )
eq.
= =
E( r j ), (1 j ) r + j E( rm ), r + j E( rm r ) .
CAPM: diversiable risk gets no reward.


(1 j ) r + j rm .
E( rj )
eq.
= =
E( r j ), (1 j ) r + j E( rm ) , r + j E( rm r).


(1 j ) r + j rm .
E( rj )
eq.
= =


(1 j ) r + j rm .
E( rj )
eq.
= =
How and
when to use the single-country CAPM
Finding beta?
single-asset betas are noisy; many prefer industry betas.
http://pages.stern.nyu.edu/ adamodar/New Home Page/datale/Betas.htm See e.g.
intervaling effect: monthly-return beta estimates are higher than daily-return ones correct for thin trading:
Scholes-Williams-Fowler-Rorke; Dimson;
correct for leverage: CoCa assumed full-equity nancing
Finding the market risk premium?

Take long-term estimates: 20 yrs is not enough Correct past data for ination? Issue of survivorship bias Just assume 4 or 5%??
When to use the model?

When local mkt is segmented from RoW Or assuming that RoW is not important?
How and
Finding beta?


How and
Finding beta?


Outline
The Cost of International Capital P. Sercu, International Finance: Theory into Practice Translate rst, then discountor vv? The Single-Country CAPM The International CAPM
Why do we need an InCAPM? Why does it contain Xrisk? Do assets have a nancial nationality? Aggregating the Efciency Conditions The InCAPM
Wrapping up
Why do we need an InCAPM?

Standard country-by-country CAPM?

Summing until the national levele.g. Canadathe countrys aggregate portfolio should be efcient in CAD terms, ... but it can no longer be equated to the stocks issued issued by local companies:
Canadian investors hold many foreign stocks Many Canadian stocks are held abroad
Aggregating to the world level solves this problem, but

we might run into problems with homogeneous opportunities we do run into problems with homogeneous expectations
Canadian T-bill is risky to Americans, not to Canadians US T-bill: vv and these are just the most obvious examples
Wrapping up
We need a CAPM that takes into account the fact that investors think in different currencies.



Wrapping up



Wrapping up



Wrapping up
Why Xrisk pops up in the InCAPM

Obvious down-to-earth reasons:

We need to translate The
FC FC
returns into the investors
HC
T-bill even has s (=S/S) as its sole source of risk
An analytical problem:
Demand equations (or efciency conditions) are in different currenciessay CAD and USD ... so how can we aggregate and nd the link between aggregate demand and the world market portfolio w? we need to translate the US investors efciency condition (which is in USD) into CADor vice versabefore we can aggregate
Wrapping up


FC FC
HC
Wrapping up


FC FC
HC
Wrapping up
How to translate returns and their moments?

Exact formula:
1 + rj 1 + rj = =
j j,t+1 V V ,t+1 St+1 = = (1 + rj )(1 + s) , Vj,t Vj,t St 1+ rj . 1 + s
Quadratic approximation
rj = rj s rj s + s2
Example
input data: r s 0.1 0.000 0.1 0.025 0.1 0.050 0.1 0.075 0.1 0.100 0.1 0.125 0.1 0.150 0.1 0.175 0.1 0.200 true r (r s)/(1 + s) 0.1000 0.0732 0.0476 0.0233 0.0000 -0.0222 -0.0435 -0.0638 -0.0833 linear approx rs 0.1000 0.0750 0.0500 0.0250 0.0000 -0.0250 -0.0500 -0.0750 -0.1000 quadr approx (r s)(1 s) 0.1000 0.0731 0.0475 0.0231 0.0000 -0.0219 -0.0425 -0.0619 -0.0800
Wrapping up

Exact formula:
1 + rj 1 + rj = =
j j,t+1 V V ,t+1 St+1 = = (1 + rj )(1 + s) , Vj,t Vj,t St 1+ rj . 1 + s
rj = rj s rj s + s2
Example
input data: r s 0.1 0.000 0.1 0.025 0.1 0.050 0.1 0.075 0.1 0.100 0.1 0.125 0.1 0.150 0.1 0.175 0.1 0.200 true r (r s)/(1 + s) 0.1000 0.0732 0.0476 0.0233 0.0000 -0.0222 -0.0435 -0.0638 -0.0833 linear approx rs 0.1000 0.0750 0.0500 0.0250 0.0000 -0.0250 -0.0500 -0.0750 -0.1000 quadr approx (r s)(1 s) 0.1000 0.0731 0.0475 0.0231 0.0000 -0.0219 -0.0425 -0.0619 -0.0800
Wrapping up

rj = rj s rj s + s2
Approximation to mean & variance:

E(rp )
= =
d
E( rp ) E( s) E( rp s) + E( s2 ) E( rp ) E( s) [E( rp )E( s) + cov( rp , s)] + [E( s)2 + var( s)], E( rp ) E( s)cov( rp , s) + var( s). var( rp s) = var( rp )2cov( rp , s) + var( s).
USD
var( rp )
Example: do Americans-in-Paris like

Example 1: Positive covariance Wus (in CAD) S (CAD/USD) Wus (in USD)
E(Wus ) stdev(Wus )
exposure?
Comment same distribution for Wus ... and same distribution for S; but the large-cov case has ...
a lower mean Wus ... and a lower stdev Wus
Example 2: Negative covariance 12,000 1.50 8,000 16,000 1.00 16,000 4,000
Wrapping up
12,000 1.00 12,000 667
16,000 1.50 10,667
11,333
12,000

rj = rj s rj s + s2
Approximation to mean & variance:

E(rp )
= =
d
E( rp ) E( s) E( rp s) + E( s2 ) E( rp ) E( s) [E( rp )E( s) + cov( rp , s)] + [E( s)2 + var( s)], E( rp ) E( s)cov( rp , s) + var( s). var( rp s) = var( rp )2cov( rp , s) + var( s).
USD
var( rp )
Example: do Americans-in-Paris like

Example 1: Positive covariance Wus (in CAD) S (CAD/USD) Wus (in USD)
E(Wus ) stdev(Wus )
exposure?
Comment same distribution for Wus ... and same distribution for S; but the large-cov case has ...
a lower mean Wus ... and a lower stdev Wus
Example 2: Negative covariance 12,000 1.50 8,000 16,000 1.00 16,000 4,000
Wrapping up
12,000 1.00 12,000 667
16,000 1.50 10,667
11,333
12,000
Do assets have a nancial nationality?

Consider rj = j + sCAD/USD + j , with = relative exposure (or elasticity of V wrt S).
Relative exposures of various assets

asset type CAD TBill USD TBill Canadian importer Canadian exporter US exporter US importer gamma 0 1 <0 >0 <1 >1 effect of rising USD no effect one-to-one effect CAD price should fall CAD price should rise USD price should fall USD price should rise
Wrapping up
Q: is gold, or oil, a dollar investment because it is quoted in USD?


Wrapping up


Wrapping up

The Cost of International Capital P. Sercu, International Finance: Theory into Practice
USD
The picture:
Translate rst, then discountor vv? The Single-Country CAPM The International CAPM
US exporters 0 Canadian importers Canadian exporters CAD T-bill 1
T-bill US importers gamma
Summary
exposures are not either 0 (CDN ) or 1 (US), but spread around these values lots of overlap in the middle: internationally competing rms have little nationality
Wrapping up

The Cost of International Capital P. Sercu, International Finance: Theory into Practice
USD
The picture:
Translate rst, then discountor vv? The Single-Country CAPM The International CAPM
US exporters 0 Canadian importers Canadian exporters CAD T-bill 1
T-bill US importers gamma
Summary
exposures are not either 0 (CDN ) or 1 (US), but spread around these values lots of overlap in the middle: internationally competing rms have little nationality
Wrapping up
Aggregating the Efciency Conditions

The original conditionsa Babylonian chaos:

CDN s US
p:
E( rj r ) E( rj r )
= cov( rj , rp ),
). = cov( rj , rp
s p :
Peace and Harmony Restored:

CDN s US
p : E( rj r )
= =
cov( rj , rp ) , cov( rj , rp ) + (1 ) cov( rj , s).
s p : E( rj r )
Aggregate (after weighting by invested wealths):

world: E( rj r) = cov( rj , rw ) + cov( rj , s),
Proof: Wca E( rj r) = Wus E( rj r) = (Wca + Wus )E( rj r) = E( rj r) = cov( rj , Wca rp ) cov( rj , Wus rp ) cov( rj , Wca rp + Wus rp ) cov( rj , rw ) +Wus (1 ) cov( rj , s) +Wus (1 ) cov( rj , s) + Wus Wca + Wus | {z (1 ) cov( rj , s). }
Wrapping up


CDN s US
p:
E( rj r ) E( rj r )
= cov( rj , rp ),
). = cov( rj , rp
s p :

CDN s US
p : E( rj r )
= =
s p : E( rj r )

Wrapping up


CDN s US
p:
E( rj r ) E( rj r )
= cov( rj , rp ),
). = cov( rj , rp
s p :

CDN s US
p : E( rj r )
= =
s p : E( rj r )

Wrapping up


CDN s US
p:
E( rj r ) E( rj r )
= cov( rj , rp ),
). = cov( rj , rp
s p :

CDN s US
p : E( rj r )
= =
s p : E( rj r )

Wrapping up
A 2-country InCAPM
Need to identify two prices of risk: ,

So we need two benchmarks We wake the world market, and the
USD
T-bill
Lets cheat a bit and assume that cov( rw , s) = 0

general: world:
USD
E( rj r) = cov( rj , rw ) + cov( rj , s) E( rw r) = cov( rw , rw ) + 0 = E( sj + r r) = 0 + cov( s, s) = E( rj r ) =

E( rw r) var( rw ) E( rw r) var( rw ) E( s+r r) var( rs )
TBill:
general:
cov( rj , rw ) +
E( s+r r) var( rs )
cov( rj , s)
= j E( rw r) + j E( s + r r)
Wrapping up
The no-cheating model: same, except that the regression slopes are from a multiple regression,
rj = j,w,s + j;s rw + j;w s+
j;w,s .
A 2-country InCAPM

USD
T-bill

general: world:
USD

TBill:
general:
cov( rj , rw ) +
E( s+r r) var( rs )
cov( rj , s)
= j E( rw r) + j E( s + r r)
Wrapping up
j;w,s .
A 2-country InCAPM

USD
T-bill

general: world:
USD

TBill:
general:
cov( rj , rw ) +
E( s+r r) var( rs )
cov( rj , s)
= j E( rw r) + j E( s + r r)
Wrapping up
j;w,s .
The Solnik-Sercu-Adler-Dumas model

N countries
E(rj r) rj = =
j;.. E(rw r) + j,1;.. E(s 1 + r1 r ) + j,2;.. E(s 2 + r2 r ) +...j,n;.. E(s n + rn r ),
j,w;s + j,w;.. rw + j,s1 ;.. s n+ 2 + j,s2 ;.. s 2 + ...j,sn ;.. s
j;w,s .
Resticted versions (1): trim the list of countries

because s are hard to estimate and Tbill premia small
small countries exchange rates have a small risk premium l = P Wl
k
Wk
( 1 ) < 1
unconnected countries (no X, M links; no competitors): cov must be close to zero anyway
Wrapping up
Restricted Version (2): cut out all Xrisk terms, just keep the beta multivariate
Assumes that Tbill premia are zero in the long run So produces almost a world CAPM, except for the beta

N countries
E(rj r) rj = =
j;w,s .

k
Wk
( 1 ) < 1
Wrapping up

N countries
E(rj r) rj = =
j;w,s .

k
Wk
( 1 ) < 1
Wrapping up
Outline
Practical implications
When do we use what model?

Is host a segmented market? immaterial to the foreign investor! Is this a domestic or a foreign investment? immaterial! Is home a segmented market?
if yes: CAPM if no: InCAPM, with world dened as markets total portfolio
When do we use what currency?

home and host integrated? use either home and host not integrated: use
HC HC
or
FC
(home)
Estimated risks and returns

Market: use long-term, lowish estimates beta: use your priors too, or even exclusively
capital goods v consumer goods; operating leverage; growth opportunities
gamma: use your common sense, incl. setting many =0 TBill premia: tiny and hard to estimate


HC HC
or
FC
(home)



HC HC
or
FC
(home)



HC HC
or
FC
(home)



HC HC
or
FC
(home)


CAPM issues in international investment

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CAPM issues in international investment

Uploaded by

Copyright:

Available Formats

The Cost of International Capital P.

Sercu, International Finance: Theory into Practice Overview

Estimating the Cost of International Capital

CAPM issues in FDI

CAPM issues in FDI

CAPM issues in FDI

The Single-Country CAPM The International CAPM Wrapping up

Two facts need reconciliation:

CFs are probably rst computed in FC (INR):

local sales priced to market

all these prices are probably sticky in terms of

The Single-Country CAPM The International CAPM Wrapping up

Two facts need reconciliation:

CFs are probably rst computed in FC (INR):

local sales priced to market

all these prices are probably sticky in terms of

The Single-Country CAPM The International CAPM Wrapping up

Two facts need reconciliation:

CFs are probably rst computed in FC (INR):

local sales priced to market

all these prices are probably sticky in terms of

The Single-Country CAPM The International CAPM Wrapping up

When in Rome, act like the Romans?

The Single-Country CAPM The International CAPM Wrapping up

... or ... My shareholders consume/think in AUD ?

including the covariance

When in Rome, act like the Romans?

The Single-Country CAPM The International CAPM Wrapping up

... or ... My shareholders consume/think in AUD ?

including the covariance

Does the translation/discounting procedure matter?

The Single-Country CAPM The International CAPM Wrapping up

Can there be a CAPM-type no-PPP equilibrium?

Does the translation/discounting procedure matter?

The Single-Country CAPM The International CAPM Wrapping up

Can there be a CAPM-type no-PPP equilibrium?

Does the translation/discounting procedure matter?

The Single-Country CAPM The International CAPM Wrapping up

Can there be a CAPM-type no-PPP equilibrium?

When to do what? Overview

CoCa model 1. Foreign investments:

The Single-Country CAPM The International CAPM Wrapping up

& HC only only

The International CAPM Wrapping up

From Asset Returns to Portfolio Return

The International CAPM Wrapping up

subtotal riskfree total 0

From Asset Returns to Portfolio Return

The International CAPM Wrapping up

subtotal riskfree total 0

Two assetsone risk-free, one risky

rp = x rs + (1 x)r0 = r0 + x ( rs r0 ) E( rp ) = r0 + x E( rs r0 ), sd( rp ) = |x| sd( rs )

The International CAPM Wrapping up

Two assetsone risk-free, one risky

rp = x rs + (1 x)r0 = r0 + x ( rs r0 ) E( rp ) = r0 + x E( rs r0 ), sd( rp ) = |x| sd( rs )

The International CAPM Wrapping up

Two imperfectly correlated risky assets

rp = x1 r1 + (1 x1 ) r2 ; ( E( rp ) = E( r2 ) + x [E( r1 ) E( r2 )], p 2 sd( rp ) = x1 var( r1 ) + 2x1 (1 x1 )cov( r1 , r2 ) + (1 x1 )2 var( r2 )

The International CAPM Wrapping up

Many risky assets and one risk-free

many risky assets:

add a risk-free asset