Carnegie-Rochester Conference Series on Public Policy 28 (1988) 3 19-324 North-Holland
COMMENTS ON THE BOMHOFF-SCHOTMANPAPER: "THE TERM STRUCTURE IN THE UNITED STATES, JAPAN, AND WEST GERMANY" LEONARDO LEIDERMAN Tel Aviv University*
1.
INTRODUCTION
The main goal of the Eomhoff-Schotman paper is to explore empirically the link between holding-period returns based on portfolios that exploit term-structure investment opportunities and a set of explanatory variables that intend to capture elements of "news" and uncertainty. goal, the paper uses a two-stage procedure. are constructed
using
term-structure
series is then regressed, variables.
Some
of
the
in the main
To achieve this
First, latent variable indices
data
for each country.
The
index
second stage, on a set of explanatory
findings
are
that
determinants
of
term-
structure yields differ across the three countries considered, the United States term-structure index affects those of Germany and Japan in a timevarying fashion, and there is time variation in risk premia, variation that cannot be solely attributed to movements in the measured variance of bond returns. The results econometric being
work
in the paper are useful and informative. takes
time-varying
seriously
parameters
into account
in the
estimated
the
possibility
relations
multivariate Kalman filter techniques and ARCH models. focus my comments on what
Moreover, the and
of thus
there uses
In what follows,
I
I consider to be some of the limitations of the
present analysis and point out how these limitations could be overcome in future work.
*Department University
0 167
-
of
of
Economics.
currently
on
leave
at
the
Graduate
Chicago.
2231/88/$03.50
01988
Elsevier
Science
Publishers
B.V.
(North-Holland)
School
of
Business,
2. ON THE TWO-STAGE APPROACH The way
in which the two-stage approach is implemented in the paper
gives rise to questions on the extent to which the analysis in each one of the stages is consistent with that in the other one.
In the first stage,
the authors construct portfolio weights on bonds of different maturities and thus generate
a series of excess returns on mean-variance
portfolios for each country.
efficient
Though it is indicated that a limited set of
returns on foreign bonds was also considered, the main series generated in this first stage treats each country separately. risk
premium
developed
in
the
international diversification
second
stage
However, the model of the
is based
of a portfolio.
on
the
notion
of
In fact, it is not clear
that investors who have access to foreign securities will want to hold the type of portfolio generated
in the first stage of the present analysis.
Given the nature of the model, it would have been interesting to construct portfolios
in
the
first
stage
upon
allowing
investors
access
to
both
domestic and foreign term structures. An alternative and
approach, where
research
short-
and
work.
and
For
Ingersoll,
excess example
and
Ross to
returns
of
how
functions
observable
are
intertemporal model
simultaneously
could
then
be
theoretical work
(1985),
assumptions,
be not to adopt a two-stage
equilibrium
determined
as
The implications of such a model for optimal
returns
an
of
a general
long-term
functions of fundamentals. portfolios
strategy would
instead develop
obtain
who
actually
closed-form
economic
the
subject
along these showed,
solutions
of
empirical
lines, see Cox, restrictive
under
for
bond
prices
For empirical work
variables.
as
in the
spirit of this approach based on term-structure Euromarket data, see Hakkio and Leiderman (1986) and Campbell and Clarida (1987).
3. The
authors
ON THE MODEL FOR THE RISK PREMIUM
develop
in Section
III
of
their
paper
a
simple mean-
variance model with the main purpose of deriving an expression for the risk premium on bilateral
bonds
in an
setting, given
empirical analysis,
international that data
setting.
While
for three countries
they
do
are used
so
in a
in the
it would have been useful to see a derivation of the
premium in a three-currency model. The authors make a restrictive assumption about investors' access to 320
foreign
exchange
uncovered for
positions
foreign
Therefore, access the
that
markets
once
implies
a
foreign
exchange.
one
specific
pattern
for
Thus,
namely,
the
foreign
exchange
Another risk
restrictive
premium
especially one
is
of
the
that
Given
purchasing-power eighties,
it
would
depends
nominal
yields For
be
relevant
been
taken to
deviations
of
the
investors
useful
from
international
the
have
rates
of
that
are in
foreign
see
in
not
the
yields
that
the in term
which
bonds
Ingersoll,
parity.
purchasing-power
the
(or
risk,
risk of
purchasing-
the
and
context
Ross
uncertainty
For
and
inflationary
in
however,
countries
from
structures
of
how inflation
task,
is
on as
seventies
where
different
different
later
deviations
a model
inflation an easy
the
term-structure
and of
nominal
Cox,
for This
appears
structures of
model
estimated
and
of
or is
possibility
under
domestic
investigated
and
during
term
pricing
purchasing-power
portfolios
to
analysis
the
located
and
arbitrage
uncertainty.
observed
embody
model
be
domestic
movements
uncertainty,
models,
portfolio
for
structures
inflation
price
were
inflation
affects
Hakkio
forward
can
price-level
in
of
purchasing-power
international must
that
intertemporal
Incorporating
no
unexpected
a theoretical
uncertainty
equilibrium
is that
implicitly
expectations.
of
of
made when developing
variables
have
on
term and
structures
magnitude
parity
premium
power)
the
three
by
interest
structure
markets
assumption constructed
yields
covered
may be
the
those
stressed
for
volatility
rates.
there given
explanatory
regressions.
are
assumption
surprising
as
of
covered
(1984)
of
under
from
only
markets
take
considerations
differ
term
by financial term
these
fact,
to
a period
term-structure
the
there
this
forward
In
take
see Stulz
in
be optimal
assumption
determined
context:
would
may well
determines
the
simultaneously
and of
sample,
assumption.
securities,
risk;
that
to
forward
agents
exchange-rate
portfolios
allowed
however,
enable
seems plausible paper’s
forward
(1986),
foreign
these
the
alternative
Leiderman and
to
in
are
In reality,
as
against It
as
investors
markets. used,
hedge
rates,
relevant.
under
these
analysis.
exchange
no
Specifically,
are
and hence
a theoretical
of
in
exchange
positions
in
markets.
of
(1985).
into in
price and that
that
an
account
deflators
are
there
may be
analysis
of
optimal
see Stulz
(1984).
4. ON THE ESTIMATED TERM-STRUCTURE MODELS Tables
8,
9,
and
10
in
the
paper
321
report
the
results
of
regressing
term-structure indices on a constant, unexpected change in the short-term interest rate, unexpected inflation, unexpected money growth, the expected level of short interest rate, expected long-term inflation, expected longterm real money growth, and the variance in short-term interest rates.
The
equations for Germany and Japan include also explanatory variables aimed at capturing effects of United States yields on the term structures in these countries. The before what
specifications used
give
interpreting the results,
the
"correct"
signs
for
rise to
a
number
of
issues.
First,
the authors provide their priors as to
the
explanatory
variables
should
be.
A
feature of models in which expectations about future variables may affect current variables the
(like the current slope of the term structure) is that
sign of a given explanatory
variable may depend
content of that variable for future variables. sign on interest-rate innovations.
on the
information
Consider for example the
That sign is likely to depend on how
the arrival of interest-rate news affects expectations of future interest rates,
and
this
in
turn,
under
rational
expectations,
stochastic process followed by interest rates.
depends
on
the
Thus, unless one considers
the stochastic processes of the "news" variables in conjunction with the term-structure model,
it may be difficult to determine a priori what the
"correct" signs should be. growth me.
in real money
Second, the rationale for including expected
balances
as an explanatory
variable
is unclear
In most models, real money balances are endogenously determined
typically depend on income (or wealth) and interest rates. already
included
as
explanatory
variables
in
the
to and
The latter are
equations,
so
one
possibility would be to include some measure of income as an explanatory variable.
With monthly data, industrial production could be a proxy for
such a variable, and in fact Chen, Roll, and Ross (1986) have shown that movements in industrial production contain explanatory power for movements in some financial involving
real
returns--
money
though
balances
whether
is an
open
this issue.
is due
to
Third,
a mechanism it would
be
interesting to investigate the role of forward interest and exchange rates as explanatory variables, since these variables typically reflect current expectations about future developments. changes
in fiscal
policies
term-structure movements. estimated equations, for
attempting
to
are
Similarly, variables proxying for
likely to contain
explanatory
power
Last, while these are detailed comments on the
I would like to emphasize that my own preference
provide
for
a
tighter
equations. 322
link between
the
theory
and
is
these
REFERENCES Campbell, J.Y. and Clarida, R.H. (1987)
The Term Structure of Euromarket Interest Rates: An Empirical Investigation.
Journal of Monetary Economics,
19: 25-44.
Chen, N.F., Roll, R., and Ross, S.A. (1986)
Economic Forces and the Stock Market.
Journal
of Business,
59:
283-404. Cox, J.C., Ingersoll, Jr., (1985)
A
Theory
and Ross, S.A.
J.E.,
the
of
Term
Structure
of
Interest
Rates.
53: 385-408.
Econometrica, Hakkio, C. and Leiderman, L. (1986)
Intertemporal
Asset
Pricing
and
the
Term
Structures
of
Exchange Rates and Interest Rates: The Eurocurrency Market. European
Economic
30: 325-44.
Review,
Stulz, R.M. (1984)
Currency
Preferences,
Determination
of
Purchasing
Exchange
Journal of Money, Credit,
Rates
and Banking,
323
Power in
an
Risks,
and
Optimizing
16: 302-16.
the
Model.