Professional Documents
Culture Documents
Min and McDonald (2006) - The Portfolio Balance Model of Exchange Rates in Korean
Min and McDonald (2006) - The Portfolio Balance Model of Exchange Rates in Korean
To cite this article: Professor Hong-Ghi Min & Judy McDonald (1993) The Portfolio-
Balance Model of Exchange Rates: Short-Run Behavior and Forecasting (The
Korean Won/U.S. Dollar Case), International Economic Journal, 7:4, 75-87, DOI:
10.1080/10168739300000015
Taylor & Francis makes every effort to ensure the accuracy of all the information
(the “Content”) contained in the publications on our platform. However, Taylor
& Francis, our agents, and our licensors make no representations or warranties
whatsoever as to the accuracy, completeness, or suitability for any purpose
of the Content. Any opinions and views expressed in this publication are the
opinions and views of the authors, and are not the views of or endorsed by
Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and
Francis shall not be liable for any losses, actions, claims, proceedings, demands,
costs, expenses, damages, and other liabilities whatsoever or howsoever caused
arising directly or indirectly in connection with, in relation to or arising out of the
use of the Content.
This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan, sub-
licensing, systematic supply, or distribution in any form to anyone is expressly
forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Downloaded by [University of Auckland Library] at 16:05 15 October 2014
INTERNATIONAL ECONOMIC JOURNAL
Volume 7, Number 4. Winter 1993
HONG-GHI MIN*
Kyungnam University
Downloaded by [University of Auckland Library] at 16:05 15 October 2014
JUDY MCDONALD
Lehigh University
1. INTRODUCTION
*The authors would like to thank the editor and two anonymous referees for their helpful
comments. All remaining errors are our own.
76 HONG-GHI MIN AND JUDY McDONALD
In this study, we empirically test the asset-market model of the exchange rate to
explain and forecast the monthly Korean won-U. S. dollar exchange rate for the
1980's. As a methodology for testing the asset-market approach, we concentrate on
the forecasting performance of the asset-market structural model. Since moments of
any order (mean and variance etc.) for the estimated coefficients do not exist in the
dynamic simultaneous equations model (see Basmann, 1974) we can not judge
whether estimated coefficients in the system are significantly different from zero. For
the same reason, it is also impossible to judge whether a certain variable has its
Downloaded by [University of Auckland Library] at 16:05 15 October 2014
expected sign. What is more important for the dynamic simultaneous equations
model is the performance of the system as a whole. In this regard, we use a
likelihood-ratio test to evaluate the structural model and to investigate the forecasting
accuracy of the asset-market model for the 1980's Korean won-U. S. dollar rate.
Meese and Rogoff (1983) discuss the fact that a major focus of attention in exchange-
rate economics has been the forecasting performance of different structural models.
The organization of the rest of this paper is as follows. In section 2, we derive an
asset-market model of exchange-rate determination from the financial market's
equilibrium conditions. In section 3, we derive an alternative asset-market model
using the observationally-equivalent transformation of the initial structural model,
since the initial structural model is poorly specified according to the likelihood-ratio
test criteria. Since the alternative model passes the very stringent likelihood-ratio test
and is also found to be dynamically stable, we compare the forecasting performance
of our asset-market structural model with the random walk model. Section 4 presents
the conclusions.
The signs of the partial derivatives are listed in each equilibrium condition. The
total differentiation of equations (1) and (2), with (4) written in for W , yields:
A B
since h, > m, and (1 - m) > h (due to adding-up constraint and dominance of own rate
effect).
An increase in the domestic money stock, M, either through a budget deficit or a
swap transaction with foreign assets, directly raises the exchange rate. This increase
in the exchange rate is required for short-run financial market equilibrium.
If the country is running a surplus on its current account, so that net foreign assets,
F, are increasing, this tends to decrease the exchange rate. A country experiencing a
deficit in its current account, with decreasing F, would experience an increase in its
exchange rate. When we move to a bilateral model for a particular rate, for example,
the Korean won/U. S. dollar rate, the natural extension of the. theory is a model that
78 HONG-GHI MIN AND JUDY McDONALD
includes movements in the Korean and U. S. money stocks as well as net foreign
assets.' We can write down a bilateral2 model for the Korean won/U. S. dollar rate,
where the exchange rate (e) is a function of measures of the U. S. and Korean stocks
of money and net foreign assets, like equation (9):
where the subscripts k and a stand for Korea and U. S. respectively. Assuming that
this function is linear, we can write down the estimating equation as equation (10):
Downloaded by [University of Auckland Library] at 16:05 15 October 2014
e(W I $ ) = a, + a,M, + a,M, +a,& + a,F, + u, where u is the random error term. (10)
B. Policy Reaction Functions
In this section we derive a money equation and a policy reaction function for the
policy maker's foreign-exchange market intervention. The policy variables in this
model are assumed to be the authorities' holdings of foreign assets (FG) and the
money stock (M). First, the monetary authority wants to influence domestic financial
conditions, that is, the money supply or the interest rate. Second, the exchange rate
may be a target of policy: targeting the exchange rate is important for a small open
economy that wants to decrease uncertainty in import and export prices. Finally, the
level of official holdings of foreign reserves (FG) may be a target.
The policy maker's objective function takes the form:
Minimize L = w , ( M - M * ) ~ + w , ( F G - F G * ) ~ + w ~ ( ~ - ~ * ) ~
(M,FG)
where w , - w, are the weights for each target, a superscript * indicates a target value,
FP is the private holdings of foreign reserves, q is the money multiplier and HG is the
'Omitting the domestic bond stocks for the econometrically spurious reason that the sign of
their coefficients cannot be determined a priori (see Branson, et al., 1977) will, by deleting the
relevant information in the regression, bias the regression coefficients. In spite of this, since
data is not available for the private holdings of the partner country's bond stock, H and H*
could not be included in the estimation.
'A bilateral model, in spite of its generality, does not permit one to assign a priori the sign of
the effect of a change in the asset stocks on the exchange rate (see Bisignano and Hoover,
1982).
31n the short-run analysis of exchange-rate determination, the dynamics of current-account
adjustments are not included in our model. Instead, we concentrate on a purely short-run
horizon in which income and prices are exogenous and the exchange rate is determined simply
by the interaction of asset markets (see Branson, et al., 1977).
PORTFOLIO-BALANCE MODEL OF EXCHANGE RATES
A. Data
All stock variables are seasonally adjusted by X-I1 ARIMA, and period averages
are used. Monthly current account data is generated from the quarterly current
account figures based on monthly trade-balance interpolation (Chow and Lin, 1971).
Variable definitions are based on Branson, et al. (1977). See Appendix B for the
definitions of all the variables used. Since the Korean won-U. S. dollar rate has
shown greater variation since 1981,4 we use the period 1981: 10 - 1988: 12 for the
estimation and 1989: 1 - 1989: 12 for the comparison of forecasting accuracy. Since
the target value for the exchange rate and for official holdings of foreign reserves are
not available, we use one period lagged values of the exchange rate and official
reserves as proxies for these variables' target values. For the target level of the money
stock, we convert the target growth rate of the total money stock into the target level
of the monthly money stock (MQ. (The data used in this study will be sent to readers
on request).
The money, intervention and exchange-rate equation (11-1 through 11-3) are
estimated using two-stage least squares (2SLS). We include M,(-1) in equation (11-1)
and e(-1) in equation (11-3) to correct serial correlation. Empirical results are
reported in Table 1.
4The Bank Korea imposed new regulations on the exchange position of foreign exchange
banks (called the "overall position management system") in April, 1981 (Kwag, 1989).
'The difference between the likelihood-ratio (LR) test in this paper and the usual specifica-
tion error test is that the LR test in this paper, when combined with the observationally-
equivalent transforrnat~ons,shows us the way to build an alternative model. Also regularity
conditions for the LR test are satisfied, given a normality assumption.
HONG-GHI MIN AND JUDY McDONALD
Mk Constant MT Ae(W/$)
,6994 ,01664 -.43567 .99416 Adj R2 = ,9993
(1.2082) i.03178) (.38671) (.03390) D. W = 2.0823
Intervention Equation (equation 11-2) I
1
Downloaded by [University of Auckland Library] at 16:05 15 October 2014
Notes: All equations are estimated by 2SLS. Figures in the parentheses are standard errors. Ae
is the first difference of the exchange-rate time series and AM is the difference between targeted
money and the total money stock.
We evaluate the structural model using the likelihood-ratio test5 (see Appendix B
for derivation). The significance probability of this test statistic is less than 1 percent.
Since the likelihood-ratio(LR) test is a model selection criteria (see Min, 1991) poor
support from this likelihood-ratio test does not imply, for the initial asset-market
structural model, the rejection of asset market hypothesis for the 1980's Korean won-
U. S. dollar rate. However this poor support does imply that the initial structural
model is misspecified even though it has microeconomic foundations. Also, since this
initial structural model has poor support from the data, we can not make reliable
statistical inferences.
Because of the poor support from the data, we derive an alternative asset-market
model from the initial structural model. The alternative model has been built using
the observationally-equivalent transformation6 of the initial structural model (see
Basmann, 1988 and Min, 1989, 1991).
models as the initial structural model since it retains all the relevant asset-market
variables. Ilowever, the alternative model is independent of the initial structural
model. This is because all the cross-equation restrictions in the observationally-
equivalent model are removed and thus this alternative model has a better support
from the likelihood-ratio test.
In this section we build an observationally-equivalent model based on the research
question that there is no feedback effect from changes in the exchange rate to the
money supply. According to McNee (1986), the inclusion of exchange rates in the
Downloaded by [University of Auckland Library] at 16:05 15 October 2014
monetary policy reaction function does not increase the explanatory power of the
model.
(-1) and e(-2) from the intervention equation; and FP, from the exchange-rate
equation. The resultant equations are reestimated using 2SLS. In Table 3, we report
the estimation results of the alternative model.
Table 2. Observationally Equivalent Model for the Initial Structural Model:
1981: 10-1988: 12
-. -.
From the exchange-rate equation, we can see that Korean private holdings of
foreign reserves (FP,) has the expected sign and the U. S. money stock (M,) has an
unexpected sign. However, as discussed earlier, the means and variances of the
PORTFOLIO-BALANCE MODEL OF EXCHANGE RATES 83
estimated coefficients do not exist, and we can not judge whether these estimated
coefficients are significantly different from zero based on the reported standard errors.
Since the alternative model has lagged dependent variables in each equation, our
natural concern is the stability of this dynamic system (Samuelson, 1983). Since all
zeros of the characteristic equation lie inside the unit circle of the complex space, the
system is dynamically stable.
When we evaluate the alternative model using the LR test, the probability that chi-
squared(9) is larger than 8.18655 is greater than 50 percent. Since the alternative
Downloaded by [University of Auckland Library] at 16:05 15 October 2014
model is well specified (compared to the initial structural model) and dynamically
stable, we will use the alternative model for forecasting.
From Table 4, we can see that the asset-market model, which has good support from
the likelihood-ratio test, forecasts better than the random walk model. From this, we
can see that the structural model derived from the asset-market approach to exchange-
rate determination is useful for forecasting.
7~MS= E JC(Yf--, where Yf is the simulated value and Yo is the actual value of
value of the exchange rate-index, and T is the number of periods in the simulation.
RMAE f 0 ~ / y awhere
= x ~ y- y , variable definitions are the same as in footnote 7 above
HONG-GHI MIN AND JUDY McDONAiD
4. CONCLUSION
This study has constructed an asset-market model for the Korean won-U. S. dollar
exchange rate to study the forecasting accuracy of the structural model. An
observationally-equivalent transformation of the initial structural model was used to
build an alternative structural model and this alternative model was used for the
dynamic simulation. One important empirical finding in this study is that the asset-
market structural model provides better forecasts of the monthly Korean won-U. S.
Downloaded by [University of Auckland Library] at 16:05 15 October 2014
dollar exchange rate than the random walk model. This implies that the asset-market
model has good support from the 1980's Korean won-U. S. dollar rate when the
structural model is properly specified. We can also conclude that the foreign-
exchange market interventions by the Korean monetary authority during the 1980's
were fairly efficient and they successfully compensated the excess demand and excess
supply conditions of Korea's foreign exchange market. In future research we plan to
develop a comprehensive vector-autoregression model to increase the explanatory
power of our results. Based on the findings in this paper, we expect that the asset-
market approach will provide superior forecasting as well as a better explanation of
exchange-rate movement under the market average exchange-rate system when
Korea's full-scale capital-market liberalization is initiated.
APPENDIX A
APPENDIX B
The constrained statistics for the structural form equation (1) estimated by 2SLS are
From (5) and (8), the likelihood - ratio test statistic can be written as follows
REFERENCES