A.D. Martin International Review of Economics and Finance 9 2000 267–286 269
over the 1975–1987 time period. Wetmore and Brick 1994 find some U.S. bank portfolios are significantly exposed to exchange rate risk over the 1986–1991 time
period. Chamberlain, Howe, and Popper 1997 report that approximately 30 of U.S. banks and 10 of Japanese banks are significantly exposed over the 1986–1993
time period. Choi and Elyasiani 1997 find 80 of the largest U.S. banks are signifi- cantly exposed over the 1975–1992 time period.
The present research contributes to the literature in the following ways. First, exchange rate exposure is assessed for the key financial institutions that comprise the
interbank FX market. Differences in exchange rate exposure across the institutions in this study may be attributed to differing degrees of risk aversion and levels of
proficiency in managing the exposure. The market should recognize significant expo- sure for those institutions that are less risk averse andor less proficient in managing
their foreign exchange exposure.
1
Second, differences in exposure across countries are analyzed. Eleven different countries are represented by the institutions in the sample. Differences in exposure
across countries may be attributed to differing regulatory and supervisory requirements e.g., Chamberlain, Howe, and Popper, 1997. Even though the Basle Accord of
1988 initiated uniform minimum capital standards for internationally active banks, it provides only guidelines. In reality, it is unlikely that consistent practices are followed
Barth, Nolle, Rice, 1997.
Lastly, this study assesses whether exchange rate exposure exists at a global level. A portfolio comprised of the key financial institutions involved in the FX market may
be viewed as a system in which all FX trading is conducted. Gains by one institution would be offset by losses of another. A simplified example may help clarify this point.
Assume there are two institutions Trader A and Trader B whose only business is trading foreign exchange with each other. The variance of a portfolio that contains
these two companies would be: s
2 p
5 s
2 A
W
2 A
1 s
2 B
W
2 B
1 2s
A
s
B
r
AB
. Since a trade would consist of one trader winning and the other trader losing, r
AB
5 2 1. Furthermore,
there exists a portfolio with proportions W
A
and W
B
that minimizes the variance, where the portfolio variance is zero. Therefore, it can be argued that there is no
foreign exchange exposure from a global portfolio perspective.
2. Data and estimation
Exchange rate exposure is assessed for the key FX institutions and for various portfolios. More specifically, a multi-factor model similar to Madura and Zarruk 1995
and Choi and Elyasiani 1997 is used. The model is estimated using the seemingly unrelated regression SUR methodology of Zellner 1962:
R
it
5 b
0i
1 b
1i
R
mt
1 b
2i
I
jt
1 b
3i
X
jt
1 m
it
i 5 1 . . . I; j 5 1 . . . J; t 5 1 . . . T 1
where
270 A.D. Martin International Review of Economics and Finance 9 2000 267–286
R
it
5 return on individual institution or portfolio i at week t;
R
mt
5 return on the Dow Jones World Stock Index at week t;
I
jt
5 return on the interest rate index for country j at week t;
X
jt
5 return on the exchange rate for country j at week t;
b
0i
5 intercept for individual institution or portfolio i;
b
1i
5 coefficient measuring the exposure of individual institution or portfolio i
to world market risk; b
2i
5 coefficient measuring the exposure of individual institution or portfolio i
to interest rate risk; b
3i
5 coefficient measuring the exposure of individual institution or portfolio i
to exchange rate risk; and m
it
5 residual for the ith equation at week t.
Eq. 1 is estimated using weekly data over the 1994-1996 time period i.e. T 5 156.
2
The competitive nature of the FX market makes it difficult to generalize about the extent of involvement of these particular financial institutions in the FX market
prior to 1994. The Federal Reserve Bank of New York 1996, p. 120 substantiates this claim:
The market structure statistics suggest that the foreign exchange market is highly competitive. Among the top ten dealers, only four dealers’ ranking remained
unchanged between 1992 and 1995. Among the dealers who were in the top ten in either 1992 or 1995, four dealers saw their ranking fall by five or more places,
while four dealers saw their ranking rise by five places or more.
Although most previous studies use monthly data to estimate exchange rate expo- sure, Wetmore and Brick 1994 employ weekly data. Because this study examines a
three-year period, weekly observations are considered more appropriate than monthly observations.
Stock prices are gathered from the Wall Street Journal WSJ. Deutsche Morgan Grenfell, Goldman Sachs, Indosuez, and Bank of Scotland are excluded either because
they are not listed on a major exchange or their stock prices are not provided in the WSJ. Thus, 26 of the top 30 FX participants are analyzed [i.e., I 5 26 in Eq. 1]. All
interest rates and exchange rates are taken from the Economist. Past studies involving global financial institutions also have used these data sources e.g., Madura Zarruk,
1995. With the exchange rate measured in foreign currency units per domestic cur- rency units, positive exchange rate exposure coefficients indicate unhedged short
long foreign currency domestic currency positions.
The interest rate returns are calculated as the rate of change in the nominal long- term interest rates of the countries in the sample. There are 11 countries represented
by the financial institutions in the sample [J 5 11 in Eq. 1]: Australia, Canada, France, Germany, Holland, Hong Kong, Japan, Sweden, Switzerland, the United
Kingdom, and the United States.
3
The use of long-term interest rates is consistent with studies that suggest banks are more sensitive to long-term rates than short-term
rates e.g., Kane Unal, 1988; Madura Zarruk, 1995. Actual interest rates are
A.D. Martin International Review of Economics and Finance 9 2000 267–286 271
used since studies such as Choi, Elyasiani, and Kopecky 1992 and Madura and Zarruk 1995 find the exposure to actual and unanticipated interest rate changes to
be quite similar. Because of evidence that suggests actual exchange rate changes may obscure expo-
sure e.g., Choi, Elyasiani, and Kopecky, 1992, unanticipated exchange rate changes are also used in this study. More specifically, the mean unanticipated return in the
value of the domestic currency, measured in terms of the foreign currencies represented by the institutions in the sample, is used. The unanticipated return on the currency
of country j, X
jt
, is defined in Eq. 2 as: X
jt
5 A
X
jt
2 EX
jt
2 where
A X
jt
5 actual rate of change in the value of the domestic currency of country j relative to the foreign currency at week t; and
E X
jt
5 expected rate of change in the value of the domestic currency of coun- try j relative to the foreign currency at week t, where the expected rate
of change is based on the international Fisher effect IFE. The actual and expected rates of change are calculated relative to each of the 10
foreign currencies separately. The mean of these unanticipated exchange rate changes as well as the mean of actual exchange rate changes are ultimately used to represent
the exchange rate factor in Eq. 1.
4
The expected return is estimated according to IFE. Thus, EX
jt
is projected for each country j based on nominal interest rate differentials
5
between country j and each of the remaining 10 foreign countries:
E X
jt
5 1 1 I
jt
1 1 I
ft
3 where
I
jt
5 interest rate for domestic country j at week t; and
I
ft
5 interest rate for foreign country f at week t.
3. Results