Exchange Risk Sensitivity and Its

Determinants: A Firm and Industry

Analysis of U.S. Multinationals
Jongmoo Jay Choi and Anita IVIehra Prasad
Jongmoo Jay Choi is Professor of
Finance at Temple University, Philadelphia,
PA. Anita Mehra Prasad is Senior
International Treasury Analyst at Lam
Research Corporation, Fremont, CA.
We develop a model of firm valuation to examine the exchange risk sensitivity of 409
U.S. multinational firms during the 1978-89 period. In contrast to previous studies,
we find that exchange rate fluctuations do affect firm value. More specifically, we
find that approximately sixty percent of firms with significant exchange risk exposure
gain from a depreciation of the dollar. We also find that cross-sectional differences
in exchange risk sensitivity are linked to key firm-specific operational variables (i.e.,
foreign operating profits, sales, and assets). Although we find limited support for
exchange risk sensitivity when we aggregate the data into 20 SIC-based industry
groups, we do observe some cross-sectional and inter-temporal variation in the
exchange risk coefficients. Subperiod analysis reveals higher number of firms with
significant exchange risk sensitivity during the weak-dollar period as compared to the
strong-dollar period.
Exchange rate variability is a major source of
macroeconomic uncertainty affecting firms in an open
economy. Exchange rate fluctuations affect operating cash
flows and firm value through the translation, transaction, and
economic effects of exchange risk exposure. In addition,
extemal shocks may create an interdependence between
exchange rates and stock retums. Therefore, it is reasonable
to expect a connection between exchange rate changes and
firm value. The importance of exchange rate variability is
also evidenced by the growing emphasis corporations place
on exchange risk measurement and management strategies.
However, compared to other macroeconomic factors,
such as inflation and interest rate risk, the research on
exchange risk and its impact on firm value is scant. Recent
studies based on portfolio data (Bodnar and Gentry, 1993,
Jorion, 1990, and Prasad and Rajan, 1995) and market-index
data (Ma and Kao, 1990) have found minimal or no evidence
of exchange rate fluctuations affecting stock retums. One
explanation for these counterintuitive results is the research
design used in these studies. We posit that, like any other
The authors would like to thank Yakov Amihud, Mark Griffiths, Richard
Levich, Anthony Saunders, Ren6 Stulz, and Arthur Warga for their helpful
comments. We also gratefully acknowledge suggestions received from the
Editors and three anonymous referees.
macroeconomic factor, the exchange risk factor will not have
the same affect on all firms. Rather, the exchange risk
sensitivity of firms will depend on their operating profiles,
financial strategies, and other firm-specific variables. Thus,
an aggregate-level analysis may not reveal the true exchange
risk sensitivity of firm value. A firm-level study is necessary
to understand whether, and why, individual fimis display
varying sensitivity to exchange risk.
Therefore, we focus our attention on individual firm value
and estimate a model of firm valuation under exchange risk
exposure using individual stock retum data for 409 U.S.
multinational firms for the 1978-89 period. We find that firm
value is significantly affected by both real and nominal
exchange rates. Further, these effects vary in terms of the
degree and direction across firms. The degree of effect refers
to the percentage change in firm value in response to a one
percent change in exchange rates; direction of effect refers
to whether a firm gains or loses from a given change in
exchange rates. We find a higher percentage of firms with
significant exchange risk exposure gain with a depreciation
of the dollar. In order to explain this cross-sectional
variation, we develop a framework linking exchange risk
exposure to firm-specific foreign operational variables; i.e.,
profits, sales, and assets. Consistent with our expectations.
Financial iVIanagement, Voi. 24, No. 3, Autumn 1995, pages 77-88.
78
FINANCIAL MANAGEMENT / AUTUMN 1995
we find a positive relationship between these foreign
operational variables and the exchange risk sensitivity of
firm value.
The intertemporal stability of exchange risk exposure
is examined by estimating these coefficients during
equally-spaced subperiods and by dividing the sample period
into strong and weak-dollar subperiods. For comparability
with prior studies, we also examine exchange risk exposure
by dividing the 409 firms into 20 SIC-based industry
portfolios. Consistent with prior studies, we observe few
industry portfolios with significant exchange risk exposure.
These results also confirm our suspicion that
aggregation of firm data into portfolios, and the resultant loss
of information, is an important reason why earlier studies fail
to document strong support for exchange risk sensitivity of
firm value.
I. Model
This study focuses on the measurement of exchange risk
exposure, as opposed to exchange risk pricing.' We employ
a simple two-factor model to estimate the exchange-risk
sensitivity coefficient of individual firms and industry
portfolios. Inclusion of the exchange risk factor is
recommended by Adler and Dumas (1984), and a similar
equation is employed by Bodnar and Gentry (1993) and by
Jorion (1990).
A. Exchange Exposure
A two-factor model, where Rjj, the retum on company i's
stock at time t, is a linear function of the retum on a market
factor, R^t' ^nd the exchange risk factor, et, is described
below:
Rjt = a-+ P R + Ye + V . (1)
The coefficients P; and Yi provide a measure of market-risk
and exchange risk sensitivity of firm i; vj, is the idiosyncratic
error term. Equation (1) is not a model of asset pricing but a
factor model that allows measurement of factor sensitivities.
A potential problem in estimating such a two-factor model
arises from the possibility that the market and exchange risk
factors may be correlated or jointly influenced by some
extemal shocks. To avoid any bias due to factor correlations,
we orthogonal ize the exchange risk factor. We follow the
standard approach (e.g., Elton and Gruber, 1991) and run a
side regression of exchange rates on the market factor.2
The issue of exchange risk pricing in the U.S. markets is examined in Choi
and Prasad (1992) and in Jorion (1991).
^During our sample period of 1978-89, the correlation between the exchange
risk factor and the market factor is statistically insignificant at the 0.05 level.
We use Equation (1) to examine the null hypothesis
that the exchange rate fluctuatians have no effect on
stock retums, i.e., HQ: YJ = 0. The; altemate hypothesis is
H ]: Yi 5^ 0. The sign of the exchange rate coefficient can be
either positive or negative depending on the net exposed
asset and liability positions of the firm. For example, firms
that use their foreign subsidiaries principally to import
finished goods and sell them in the U.S. will benefit from an
appreciation of the dollar. This benefit arises due to a
reduction in the dollar value of fcireign costs. In contrast,
firms that incur most of their ccst of production in the
U.S. and sell in foreign markets have exposed foreign
sales revenue. They find that their products become less
competitive in overseas markets, and their foreign sales
revenues decline with any appreciation of the dollar.
Similarly, multinational firms with net exposed assets
abroad will lose with a strengthening dollar, while firms
with net exposed liabilities gain.
B. Firm-Specific Determinants
In this section, we develop a framework for incorporating
the role of firm-specific variables in explaining the
cross-sectional variations in excliange risk exposure of
individual firms. We begin by defining the value of a
U.S.-based multinational firm (Vj) as the sum of its domestic
value (Vj) and its foreign value (V j) components:
V. = d.+V<'.
(2)
where each component is expressed in dollars (the home
currency of the multinational firm). These component
values, in turn, are the present valuer of their respective net
operating dollar cash flows (TI'' and K ^ ) :
(3)
fj = J 7ifi,exp(-ki,)d,.
By definition, Rjt, the rate of retum on a firm's stock, is the
percentage change in the firm value, V,t,
\ = (V. - V. ,)/(V. , ) . (4)
Similarly, the exchange-risk-sensitivity coefficient, Yi. in
Equation (1) can be expressed as:
Estimations were done with and without orthogonalization, and the results
were similar. However, following the suggestions of an anonymous referee,
we use orthogonalized variables in all the esti nations reported here.
CHOI & PRASAD / EXCHANGE RISK SENSITIVITY AND ITS DETERMINANTS
79
= cov(Rj,, e,)/var(ej). (5)
exchange risk exposure in terms of the firm's identifiable
assets:
Applying Equations (2) and (3) to Equation (4) and
substituting the resulting value of Rij in Equation (5) enables
us to express the exchange risk exposure of stock retums in
terms of the firm's net operational cash flows from domestic
and foreign sources:
To establish the connection between the exchange
exposure coefficient and firm-specific variables, we
decompose the net operating cash flows from domestic and
foreign sources as:
(7)
where S;, is firm i's sales revenue at time t, VQ, is the
variable cost, FCjt is the fixed cost, and T is the effective tax
rate (the superscripts d and f denote domestic and foreign
variables, respectively). Substitution of Equation (7) in
Equation (6) gives us the exchange exposure coefficient in
terms of revenue and cost variables:
Yj = [Sj,cov(SVSj,,ej)-VCj,cov(VCfj/VCj,,e,)]/(8)
where Sit = S'^it + S^jt and Vcn = VC^jt + i^
Alternatively, following Hodder (1982), the firm value
can also be expressed as a function of the firm's assets and
liabilities. For this purpose, we may write operating cash
flows as:
d A' ' Tif =
^i t ' ^ it' " it
(9)
where r^i, and r'^jt are the rate of returns on domestic (A''it)
and foreign assets (A^n) respectively. Substitution of
Equation (9) in Equation (6) allows us to explain the
' TO simplify covariance calculations, the domestic and foreign revenue and
cost variables are assumed to be exogenous and independent of each other.
In a more general setting, a more complicated expression is obtained. For
example, fixed costs become variable in a continuous time framework. A
violation of the assumption that domestic and foreign costs and revenues are
independent would yield additional covariance terms. These additional
covariance terms, however, do not materially change the nature of
firm-specific variables used in the empirical work here. Finally, it is also
possible to have a covariance involving financial leverage, as in Hamada
(1972), by including domestic and foreign debt expenses in Equation (7).
This extension, however, is not pursued here because of the difficulty in
obtaining the necessary data on the breakdown between domestic and
foreign debt and interest expenses.
Yj = Aj,cov(AyAj^,e,)/var(ej) (10)
where Aj, = A'';; + Af;,. Equations (6), (8), and (10) are now
summarized in functional form as:
(11)
For empirical tests, we express the exchange risk exposure
in Equation (11) as a function of domestic and foreign
operating profits, sales, costs, and assets.
Note that foreign and domestic variables are all stated in
dollars. Thus, the stated covariances would reflect not only
the economic effects on foreign-currency-denominated
operational cash flows but also the translation impact of
restating cash flows, assets, and liabilities in U.S. dollars.
Equation (11) is a general specification that subsumes
different operational characteristics and market conditions in
the firm's output and input markets.'* The existence of a
positive relationship between exchange risk exposure and
firm-specific variables is tested based on the null hypothesis,
HQ: ai > 0, against the alternate, H]: ai< 0.
Characterization of exchange risk exposure as a function
of these firm-specific variables presumes that the exchange
risk faced by the firm is not fully eliminated by operational
or hedging strategies.^ Given imperfect hedging, we can
expect the exchange exposure of firms to increase with an
increase in the firm's overseas assets, sales revenues, or
profits. For example, if a firm generates a higher proportion
of its revenues from foreign markets, it may face a higher
level of exchange rate risk because a larger percentage
of its revenues is denominated in foreign currencies.
Consequently, ceteris paribus, the higher the foreign sales,
the greater will be the effect of exchange rate fluctuations on
firm value. The sensitivity of the firm's cost structure to
exchange rate changes also affects the exchange risk
exposure of a firm. Here, we capture this possibility by the
inclusion of foreign operating profits in the estimations.
Similarly, a firm's ownership of exposed assets abroad
affects its value in dollars through the translation effect.^
""See Choi (1986) and Errunza and Senbet (1981) for further discussion of
exchange rates and firm valuation.
^The assumption of imperfect exchange risk hedging has been observed by
Grammatikos, Saunders, and Swary (1986) in the case of commercial
banking. It is also consistent with the results obtained by Eun and Resnick
(1988) for major industrial countries based on aggregate market indices.
*Eaker (1980) examines the choice of the currency of denomination for
multinational transactions and its effect on exchange risk exposure.
80
FINANCIAL MANAGEMENT / AUTUMN 1995
II. Data
Monthly time-series of stock returns (inclusive of
dividends) were obtained from the University of Chicago
Center for Research in Security Prices (CRSP) tapes for the
period of January 1978toDecember 1989. Four hundred and
nine multinational firms that had complete price and
dividend information during the entire sample period are
included in tbe study. A multinational firm is defined as a
firm that has production facilities located in more than
two countries (Dunning, 1973). Consistent with this
definition, we determine the multinationality of a firm using
firm-specific information in the COMPUSTAT database. A
firm is considered a multinational if its foreign sales, net
operating profits, and identifiable physical assets are all 25%
or more of their respective corporate totals and exceed U.S.
$1 million in 1989.'' This method of defining multinational
firms according to all three foreign operational variables
simultaneously is similar to that used in the international
business literature. It is also more stringent than using an
arbitrary cutoff point for one of these variables on the basis
of segment data reported under SFAS No. 14.
The nominal exchange rate variable is the U.S. dollar
value of one unit of foreign currency, where foreign currency
is the multilateral trade-weighted basket of ten major
currencies as published in the Federal Reserve Bulletin.^ An
increase in the exchange rate implies an appreciation of the
foreign currency and a depreciation of the dollar. We
calculate the real exchange rate by adjusting the nominal
exchange rate for the U.S. and foreign monthly consumer
inflation rates obtained from the appropriate Federal Reserve
Bulletin. The inflation rates for the foreign country are
calculated based on the same trade-weights as used in the
multilateral-exchange rate series. The exchange risk factor
used in Equation (1) is calculated as the percentage change
in the nominal and real exchange rates orthogonal ized to the
market factor. We also used unexpected exchange rate
changes, defined as the difference between actual and
expected exchange rates. Expected exchange rates were
proxied by the forward rate or based on lagged spot rates.
The results in the latter case are consistent with those
We recognize that given our arbitrary classification date of 1989. there is
no control for firms that may have changed classification during the sample
period. Lack of data availability and incomplete data during the earlier
periods are the principal reasons we base our classifications on the last year
in our sample period.
The weights of each currency are: 0.064 Belgian franc, 0.091 Canadian
dollar, 0.131 French franc, 0.208 German mark, 0.090 Italian lira, 0.136
Japanese yen, 0.083 Dutch guilder, 0.042 Swedish krona, 0.036 Swiss franc,
and 0.119 British pound. These weights are based on the average trade shares
of the 10 countries for a five-year period: 1972-1976. {Federal Reserve
Bulletin, August 1978, p. 700).
reported here. The market factor is proxied by the percentage
change in the value-weighted, dividend-adjusted CRSP
market index. Firm-specific variables used in the study are
obtained from the COMPUSTAT database, which contains
information from firm 10-K reports.
III. Exchange Risk Sensitivity of
Individual Firms
We estimate Equation (1) using the ordinary least
squares (OLS) method to obtain exchange risk sensitivity
coefficients for the 409 multinational firms.^ From an
econometric standpoint, the generalized least square (GLS)
approach, which accounts for the cnjss-sectional correlation
in residuals, is superior to OLS. However, GLS limits the
number of firms (cross-sectional series), which must be
smaller than the number of observations (time-series).
Therefore, rather than grouping data and losing valuable
information, we use OLS on individual firms, achieving
greater economic information at the potential expense of
econometric inefficiency. We apply ihe GLS approach, in the
form of seemingly unrelated regression, to industry data later
in this paper.
Table 1 presents the results foi" the nominal and real
exchange risk exposures of individual firms. The results are
summarized in terms of the sign iind significance of the
exposure coefficient. Panel A shows that 61 firms have
significant exchange risk sensitivities at the 0.10 level
(two-tailed test). Of the firms witti significant exchange
exposure, 64% benefit from a depreciation of the dollar.
The others have a negative exchange exposure
coefficient. A positive (negative) coefficient indicates that
firms experience an increase (decrease) in stock retums when
the dollar depreciates against the forsign currency. The null
hypothesis^that exchange rate exposures are zero for all
firms (Yi = 0)is rejected at the 0.01 level (F - 372.57), thus
establishing the existence of exchange risk sensitivity at the
firm level.
The nominal and real exchange risk sensitivities of our
sample firms are very similar; we observe only a marginal
difference in the exact value of the nominal and real
exchange risk sensitivity coefficient. The firms that are
exposed significantly to nominal exchange risk are also the
ones that exhibit significant real exchange risk coefficients.
Of the 63 firms with a significant real exchange risk
coefficient, 59% have a positive coefficient. Given these
Q
We examined the original stock return data and the residuals from the
regression procedure to detect outliers. Forty-two of the 409 firms exhibited
one to three outliers (3 deviations from the mean) from a total of 140
observations per firm. In reestimating the regressions for these 42 firms, we
excluded the outliers.
CHOI & PRASAD / EXCHANGE RISK SENSITIVITY AND ITS DETERMINANTS 81
Table 1. Exchange Risk Exposure of Individual Firms: 1978-89
The exchange risk exposure coefficient, Yp is estimated individually for 409 multinational firms using monthly time-series data for the
1978-89 period. Estimations use the OLS approach. Rj, is the rate of retum on stock i, R^,; is the percentage change in the CRSP
value-weighted market index, and e, is the percentage change in dollar value of one unit of a trade-weighted basket of currencies. The
exchange rate factor is orthogonal to the market factor. The real exchange rate is calculated by adjusting the nominal data for the U.S.
and foreign inflation rates. Panel A provides information about the nominal and real exchange risk exposure coefficients. Panel B tests
for the overall exchange risk exposure of all 409 fitms.
Total Firms
PanelA. Exchange Risk Exposures
Significant Exposure
f/o Total Firms)
(i) The Nominal Exchange Exposure Estimations
409 61 (15%)
(ii) The Real Exchange Exposure Estimations
409 63 (15%)
Positive Exposure
f/o Significant)^
39 (64%)
37 (59%)
Negative Exposure
f/o Significant)^
22 (36%)
26(41%)
Panel B. Test of Overall Fxcliange Risk Exposures
Ho: No exchange exposure exists. (Yi = 0) F value = 372.57***
***Significant at the 0.01 level.
' ^he exchange exposure coefficients in Panel A are evaluated at the 0.10 level using a two-tailed test.
Table 2. Exchange Risk Sensitivity Coefficients Based on Nominal Data for the 1978-1989 Period
This table reports summary statistics based on nominal exchange risk exposure coefficients reported in Panel A of Table 1. Only the 61
firms with significant nominal exchange risk exposures are included in this analysis. Quartiles are formed by ranking the exposure
coefficients in descending order.
1. All 61 Firms with Significant Exposure
2. First Quartile
3. Second Quartile
4. Third Quartile
5. Fourth Quartile
Mean
0.1567
0.9043
0.5565
-0.0057
-0.8780
Std. Deviation
0.7350
0.2653
0.0522
0.4756
0.1608
iVIaximum
1.6677
1.6677
0.6711
0.4905
-0.6758
Minimum
-1.1965
0.6764
0.4981
-0.5998
-1.1965
results, and the fact that it is the nominal exposure that is some of the sample firms gain, while others lose, when the
reported in financial statements, the following estimations dollar depreciates. These variations, and their relationship to
use only nominal data.
firm-specific variables, is examined in detail in the following
We observe cross-sectional variations in the exchange section. Table 2 provides summary statistics on the exchange
risk sensitivity of individual firms. As reported in Table 1, exposure coefficients of the 61 firms with significant
82
FINANCIAL MAN/VGEMENT / AUTUMN 1995
nominal exchange exposures. In addition to the benefits or
loss (direction) effect, the information in Table 2 reflects
differences in the degree of exchange risk sensitivity of
firms. The nominal exchange risk coefficients vary from
-1.1965 to 1.6677, with a mean of 0.1567 for the group
of 61 firms. Thus, on average, a 1% depreciation of the
dollar is accompanied by a 0.15% Increase in the stock
return. Examination by quartiles provides a better
understanding of the distribution of these coefficient values.
The 61 coefficients are ranked in descending order and
grouped into quartiles. The first quartile, which includes the
firms with the highest positive-exposure coefficients, has an
average coefficient of 0.9043, while the fourth quartile, with
lowest negative-coefficient values, has a mean
coefficient of -0.8780.
The fact that we do not find a large percentage of firms
with significant exchange risk sensitivity is not inconsistent
with the theory posited in the paper. A fundamental
motivation for our study is grounded in the insight that
exchange rate fluctuations, like any other macroeconomic
factor, should have varying effects on firm value. We argue
that variations in exchange risk sensitivity of firm value are
a reflection of differences in firm-specific economic,
operational, and policy variables.' ^ For example, the impact
of exchange rate fluctuations on firm value should depend
on whether the firm has net exposure on the foreign cost side
(cash outflow) or on the foreign revenue side (cash inflow).
Firms with net cash outflow exposure (import-oriented
firms) should benefit, while firms with net cash inflow
exposure should lose from appreciation of the dollar.
Similarly, some firms may have foreign cash infiows almost
offsetting foreign cash outflows. The value of these firms
will be unaffected by changes in exchange rates even if they
have extensive foreign operations. For example, Hewlett
Packard, a representative firm, derives a significant portion
of its revenues from foreign markets (approximately 50% in
1994) and has production facilities in several countries
(more than ten in 1994) but displays an insignificant
exchange risk sensitivity coefficient.''
Another firm-specific operational variable affecting
exchange risk sensitivity of a firm is its policy regarding the
degree of risk aversion and its hedging strategies. Such
hedging decisions can effectively disguise the potential
exchange risk exposure of a firm by sheltering cash flows
from exchange rate fluctuations. For example, Eastman
Kodak, known to be innovative and sometimes
aggressive in its exchange risk management strategies,
displays a significant exposure of -0.4386, while Polaroid, a
firm in the same SIC classification, is more conservative and
has an insignificant exchange risk exposure.
Bartov and Bodnar (1994) provide an additional
justification for finding insigrificant exchange risk
exposure. They suggest that firms that can respond to
exchange rate changes and overill international market
conditions at low cost will tend to have insignificant
exchange risk exposure. Consisteni. with our arguments, we
find evidence suggesting that foreign cash inflows
(revenues) and outfiows (costs) are exposed to exchange
rate fluctuations. However, offsetting cash fiows can
minimize the exposure to exchange risk. Supporting
empirical evidence is presented in the following section,
where we examine the effect cif several firm-specific
international-business operational variables on the exchange
risk sensitivity of a firm. However, due to the lack of detailed
information, the effect of management profiles and exchange
risk hedging strategies is left for fui;ure investigation.
Before moving on to firm-specific estimation, a caveat is
in order. Given the possibility that estimations based on the
two-factor model may be biased due to the existence of
omitted variables, the robustness of the results presented
here is evaluated using an altemative model specification.
We specify a three-factor model where interest rate risk
augments the market and exchange risk factors.'^ Use of the
interest rate variable is well-sufported in the existing
literature (for example, Choi, Elyasiani, and Kopecky, 1992,
and Sweeny and Warga, 1986). Th(i interest-rate risk factor
is calculated as the change in the three-month U.S. Treasury
bill rate. Results obtained reveal 61 firms with significant
exchange risk exposures. These are the same firms that were
significantly exposed to exchange risk according to the
two-factor model. The sign of the exchange risk coefficient
is also consistent with the two-factor model; 59% of the
significant exposures are positive and 41% negative.
IV. Firm-Specific Detorminants of
Exchange Risk Sensitivity
Based on Equation (11), we estimate the following
linearized equation to test the null hypothesis that exchange
risk coefficients are positively correlated with variables that
indicate the extent of a firm's international operations:
(12)
and Bodnar (1994) present similar arguments in their recent study
on the relationship between exchange rate changes and firm performance.
'Quantitative information about specific firms was obtained from annual
reports and documents submitted to the SEC.
X., =
We recognize that the use of a three-factor model does not preclude the
possibility of the existence of other fundamental economic factors.
CHOI & PRASAD / EXCHANGE RISK SENSITIVITY AND ITS DETERMINANTS 83
where S^^ is the foreign sales revenue, A^jj is the foreign
identifiable assets, and n^^i is the foreign operating profit. All
firm-specific variables are measured in billions of U.S.
dollars. Data on firm-specific variables are plagued with
problems, such as missing data items, and observations are
missing in the series that are available. This limitation forces
us to drop the cost variable in Equation (11) from the
empirical tests. The effects of the cost variable on the
exchange risk sensitivity, however, should be reflected in the
operating profit variable. We find usable information for the
three variables listed in Equation (12) for the five-year period
1985 to 1989. Cross-sectional estimations are carried out for
each of these five years and for the average data during
1985-89.
Prior studies have defined firm-specific variables either
as ratios or as levels in evaluating the multinationality of a
firm. However, high foreign sales ratios generally do not
indicate greater exchange exposure across firms of different
sizes. For example, a firm with $200 million in total sales
and a 20% foreign sales ratio will have a larger cash flow
exposed to exchange rate risk than a firm with $100 million
in total sales and a 20% foreign sales ratio. Therefore, we use
the level of foreign variables rather than ratios of foreign to
domestic variables. To avoid multicollinearity, each of these
variables is estimated separately.
Following Fama and French (1992), estimations are
carried out in two steps, using non-overlapping periods.
Exchange risk betas are estimated using time-series data for
1978-84, while cross-sectional estimates use the 1985-89
time period. These estimations are based on the 61 firms with
significant exchange exposure coefficients. Firms with
insignificant exposures are excluded from the present
analysis. If the exposure coefficient is insignificant, we
cannot use it to derive any reliable conclusion about its
relationship with firm-specific variables. Initial estimations
based on the absolute value of the exchange risk coefficient,
Yj, show that foreign sales and foreign assets are significant
during the overall sample period of 1985-89 (Panel A of
Table 3). Foreign sales and assets also yield significant
results for three of the five individual years1987, 1988,
and 1989. Foreign operating profits are significant for the
first three years.
Next, we separate the data based on the sign of the
exchange exposure coefficient. Estimations based on firms
with a positive coefficient (Panel B) and those with a
negative coefficient (Panel C) yield similar results. For the
sample of firms with positive coefficients, all three foreign
variables are significant for the overall sample period
and for two of the five individual years. We observe some
sensitivity to the particular firm-specific variable selected;
the foreign sales variable shows superior performance
relative to the other two variables. Foreign operating profits
are significant for two years for the sample of firms
with positive coefficients and for three years for the
sample of firms with negative coefficients. Overall, all three
firm-specific variables have positive coefficients supporting
a positive association between foreign operations and
exchange risk exposure.
V. Exchange Risk Sensitivity of
Firms During Sub-Periods
The overall sample period of 1978-89 reveals subperiods
with different secular trends. There is a steep increase in the
value of the dollar until March 1985. The following period,
characterized by the Plaza Accord in 1985 and the Louvre
Accord in 1987, indicates a declining trend until early 1987,
followed by a relatively mixed pattern. For example, a
dollar depreciation implies an increase in domestic
prices of foreign goods, while a dollar appreciation portends
a decrease in domestic prices. The falling value of the dollar
makes exporters more price competitive, while a rising dollar
benefits importers. However, given the short-run downward
price rigidity in the economy, a price decrease is not as likely
as a price increase. Therefore, a dollar depreciation and a
dollar appreciation may not bring about symmetric changes
in firm value. Thus, we test for any variations in the exchange
risk sensitivity during different dollar regimes. The overall
sample period is divided into the strong-dollar period of
January 1978 to March 1985 and the weak-dollar period of
April 1985 to December 1989.
Estimation results in Panel A of Table 4 indicate that
exchange rate effects vary during the two dollar
regimes; the number of firms with significant exchange risk
sensitivity is higher during the weak-dollar period. The
exposure coefficients range from 1.832 to -1.255 for the
strong-dollar period and from 1.702 to -2.044 for the
weak-dollar period. The direction of exchange rate effects
for both subperiods is consistent with that for the overall
sample period; the majority of firms (68%-71%) gain
from a depreciation in the value of the dollar. The falling
value of the dollar makes exporters more price competitive,
while a rising dollar benefits importers. However, given
the downward price rigidity, both changes do not have
symmetrical effects. Estimations based on equally-spaced
subperiods yield similar results. However, these results need
to be interpreted with caution, given that each subperiod is
too small for broad generalizations.
84
FINANCIAL MAN>^GEMENT / AUTUMN 1995
Table 3. Firm-Specific Determinants of Exchange Exposure
l ^ i t ' -^i t - ^ it' ^ it' ^ it
The exchange risk sensitivity coefficient, Yi,, is estimated for individual firms over the 1978-84 period, using the two-factor model
(Equation 1). The effect of the firm-specific variables on the exchange risk sensitivity is examined using cross-sectional data for the
1985-89 period. Separate estimations are conducted for each of the five years, 1985 to 1989, and on average data for the combined five-year
period. Panel A reports results based on the 61 firms with significant exchange risk exposures. Given that firms exhibit both positive and
negative coefficients, Panel B evaluates the 39 firms with positive coefficients, and Panel C examines the 22 firms with negative exposure
coefficients. We expect to find positive correlation between firm-specific variables and the exchange exposure coefficient. Therefore, a
one-tail test is used to test the null hypothesis of a, > 0 against the altemate of a, < 0. All firm-specific variables are measured in billions
of U.S. dollars. S is the foreign sales revenue of firm i of dollars, A is the foreign identifiable assets of firm i, and 71*^ is the foreign
operating profits of firm i.
Panel A. Estimations Using Absolute Exchange Rate Exposure Coeffiicienti^
1985 1986 1987 1988 1989
1985-1989
-0.007
(-0.29)
0.0018
(0.067)
-0.292*
(-1.48)
-0.015
(-0.71)
-0.004
(-0.19)
-0.219*
(-1.47)
0.0192**
(1.722)
0.0185**
(1.824)
0.1671*
(1.641)
0.0114*
(1.514)
0.0117**
(1.688)
0.0745
(1.276)
O.OICO*
(1.469)
0.0114*
(1.626)
0.0724
(1.222)
0.0112*
(1.520)
0.0159*
(1.438)
0.0810
(1.193)
Panel B. Estimation Using Positive Exchange Rate Exposure Coeffiicients"
1985 1986 1987 1988 1989
1985-1989
Sf 0.0156*
(1.615)
Af 0.0266
(0.915)
Jlf -0.215
(-1.06)
0.0059
(0.243)
0.0183
(0.702)
-0.150
(-0.91)
0.0133**
(1.650)
0.0129**
(1.782)
0.0881*
(1.277)
Panel C. Estimations Using Negative Exchange
1985
Sf 0.0079*
(1.352)
Af 0.0067
(1.195)
Jlf 0.8710*
(1.489)
**Significant at the 0.05 level.
*Significant at the 0.10 level.
^Numbers in parentheses are t values.
1986
0.0061*
(1.520)
0.0065*
(1.334)
0.4561*
(1.391)
1987
0.0538*
(1.569)
0.0047*
(1.285)
0.4275*
(1.408)
0.0129*
(1.638)
0.0130**
(1.787)
0.0847*
(1.362)
0.0119**
(1.664)
0.0133**
(1.860)
0.0814
(1.280:i
Rate Exposure Coefficients'^
1988
0.0037**
(1.663)
0.0244
(0.920)
0.2180
(1.060)
1989
0.002''**
(1.660;
O.OOli!
(0.868]
0.1732
(0.827)
0.0131**
(1.739)
0.0209**
(1.810)
0.0954*
(1.357)
1985-1989
0.0053*
(1.631)
0.0041
(1.200)
0.4175*
(1.382)
CHOI & PRASAD / EXCHANGE RISK SENSITIVITY AND ITS DETERMINANTS 85
Table 4. Exchange Risk Exposure of Individual Firms During the Subperiods
This table reports the variability in the nominal exchange risk exposure coefficients during various subperiods. The overall sample period
is segmented based on the trends in the dollar and then into equally-spaced subperiods. Estimations are based on the two-factor model
for the 409 multinational firms. R^, is the rate of retum on stock i, R^i is the percentage change in the CRSP value-weighted market index,
and e( is the percentage change in exchange rates. The exchange rate factor is orthogonal to the market factor.
Firms with
Significant Exposure
(i) Strong-Dollar Period (1/78-3/85)
34 (8%)
(ii) Weak-Dollar Period (4/85-12/89)
55(13%)
Firms with
Significant Exposure
(i) First Subperiod (1/78-12/83)
31 (8%)
(ii) Second Subperiod (1/84-12/89)
44(11%)
Panel A. Subperiods Based on Dollar Value
Average Vaiue of TJ
0.4042
0.3950
Positive Exposure
(% Significant^
23 (68%)
39(71%)
Panel B. Equally-Spaced Subperiods
Average Vaiue of y^
0.3826
0.6401
Positive Exposure
(% Significant)^
20 (65%)
38 (86%)
Negative Exposure
(% Significant)^
11(32%)
16(29%)
Negative Exposure
(% Significant)"
11 (35%)
6(14%)
"The significance of exchange exposure is evaluated at the 0.10 level using a two-tailed test.
VI. Exchange Risk Sensitivity of
Industry Groups
In this section, we shift our focus from individual
firms to industry groups. We explore the possibility that
the exchange risk exposure patterns are industry-specific.
We recognize that exchange exposure patterns may be
examined by using other classification schemes, such as
capitalization-ranked portfolios or the export- or
import-orientation of firms. However, given our focus on
examining industry-specific exchange risk sensitivity, and to
facilitate comparison with Jorion's (1990) results, we use
two-digit SIC codes to group the 409 firms into 20 industry
portfolios.
Table 5 provides the SIC-codes and the number of firms
per industry portfolio. Once the industry portfolios are
formed, we employ the seemingly unrelated regressions
(SUR) technique to estimate exchange rate sensitivity for
each of these industry portfolios. This method incorporates
the cross-sectional interdependency of residuals. Results
from the two-factor model during the 1978-89 period reveal
only two industries, mining and other retail, with positive
exchange exposure coefficients that are significant at the
0.10 level (two-tailed test). These results are consistent with
Jorion (1990). The exchange exposure coefficients range
from -0.270 for department stores to 0.361 for mining. In
terms of absolute-values, the high end of exposure
coefficients is dominated by textiles and apparel, retail, and
department stores. The sign of the coefficient is negative for
each of these groups. This suggests that as a group, firms in
these industries face a greater exposure on their cost side and,
thus, lose when the dollar depreciates. The low degree of
exchange risk exposure for utilities is explained by the fact
that firms in this industry are protected by dollar pricing of
energy products and by govemment regulations. The F
statistic of 5.128 rejects the null hypothesis of equal
exposures across industries at the 0.02 level.
The fact that a lot of variation is lost in the aggregation
process provides an econometric justification for finding
few industries with significant exchange risk sensitivity.
Another explanation is that firms within an industry group
are not necessarily homogeneous in their operational
86 FINANCIAL MANAGEMENT / AUTUMN 1995
Table 5. Exchange Risk Exposure of Industry Portfolios
Two-Factor Model: R-, =
Three-Factor Model: R,, = aj
5,g, +
A two-factor and a three-factor model are employed to estimate the exchange risk sensitivity of 20 SIC-based portfolios. Estimations use
the SUR approach to allow for cross-sectional correlations in the residuals. Rj; is the rate of retum on stock i, Rmt is the percentage change
in the CRSP value-weighted market index, e, is the percentage change in exchange rates, and 5, is the change in the three-month treasury
bill rate. The interest rate and exchange rate factors are orthogonal to the market factor."
Industry
Mining
Food & Beverages
Textile & Apparel Products
Paper Products
Chemical
Petroleum
Stone, Clay & Glass
Primary Metals
Fabricated Metals
Machinery
Electrical Equipment
Transport Equipment
Miscellaneous Manufacture
Railroads
Other Transport
Utilities
Department Stores
Other Retail
Finance, Real Estate
Other
**Significant at the 0.05 level.
*Signif icant at the 0.10 level.
"Numbers in parentheses are t values.
SIC
10-14
20
22,23
26
28
29
32
33
34
35
36
37
38,39
40
41-47
49
53
50-52,
54-59
60-69
Other
No. of
Firms
27
23
9
II
50
10
10
15
18
45
42
20
24
2
5
11
5
18
16
48
based on a two-tailed test.
1978-89
(2-Factor)
Y i
0.361**
(1.98)
-0.091
(-0.88)
-0.151
(-0.79)
0.162
(1.36)
0.083
(1.26)
0.172
(0.91)
-0.061
(-0.43)
0.089
(0.49)
0.049
(0.43)
0.043
(0.35)
-0.032
(-0.26)
-0.180
(-1.53)
-0.036
(-0.33)
0.045
(0.24)
-0.093
(-0.54)
0.082
(0.85)
-0.270
(-1.53)
-0.245*
(-1.90)
0.059
(0.77)
-0.117
(-1.26)
1978-89
(3-Factor)
Ti
0.351*
(1.87)
-0.099
(-0.92)
-0.0860
(-0.44)
0.170
(1.39)
0.088
(1.31)
0.159
(0.81)
-0.029
(-0.20)
0.128
(0.70)
0.051
(0.42)
0.072
(0.57)
-0.031
(-0.24)
-0.164
(-1.38)
-0.021
(-0.18)
0.036
(0.19)
-0.111
(-0.63)
0.073
(0.73)
-0.265
(-1.48)
-0.241*
(-1.83)
0.058
(0.75)
-0.084
(-0.90)
1/78-3/85
(2-Fa:tor)
Ti
0.633**
(2.56)
-0.180
(-1.34)
0.077
(0.31)
0.011
(0.06)
0.022
(0.22)
0.195
(0.72)
-0.021
(-0.141
0.293
(1.31)
O.I 12
(0.63)
0.087
(0.551
0.03:5
(0.19:i
-0.03:)
(-0.21)
O.O3t5
(0.02;.
0.299
(1.12;
0.174
(0.74;
-0.00(i
(-0.04;
-0.25!!
(-0.93;
-0.252
(-1.22)
0. 184*
(1.82)
-0.011
(-0.08)
4/85-12/89
(2-Factor)
Y i
0.048
(0.17)
-0.011
(-0.06)
-0.281
(-0.89)
0.357**
(2.06)
0.146*
(1.70)
0.121
(0.44)
-0.072
(-0.28)
-0.042
(-0.13)
-0.010
(-0.06)
0.132
(0.64)
0.042
(0.25)
-0.204
(-1.24)
0.021
(0.14)
-0.146
(-0.53)
-0.405
(-1.52)
0.197
(1.47)
-0.182
(-0.82)
-0.159
(-1.04)
-0.049
(-0.40)
-0.150
(-1.26)
CHOI & PRASAD / EXCHANGE RISK SENSITIVITY AND ITS DETERMINANTS
87
characteristics or in their financial strategies.'3 Further
study is needed to explain the exchange risk exposure
characteristics of various industry groups in a greater detail.
The robustness of these results is examined by estimating
a three-factor model that includes interest rate risk (Table 5).
Once again, we find that results are consistent with the
two-factor model; the same two industry portfolios (mining
and other retail) are significantly exposed. The intertemporal
behavior of exchange rate exposures for the industry groups
is also similar to that of individual firms. We find only two
industry groups with significant exchange risk exposures
during each of the subperiods. During the strong-dollar
period mining and finance and real estate display positive
exchange exposure coefficients that are significant at the
0.10 level. Financial firms, part of the finance and real estate
industry, appear to gain when the dollar is strong, perhaps
due to the increased international capital inflow into
dollar-denominated securities. However, subperiod results
cannot be generalized because of the limited number of
observations in each subperiod.
VII. Summary
In this paper, we estimated a model of firm valuation to
examine the exchange risk sensitivity of firm value. The
model was estimated for 409 U.S. multinational firms
during the 1978-1989 period and for 20 industry portfolios.
Results for individual firms indicated that approximately
60% of the firms with significant exchange risk exposure
benefited, and 40% lost, with a depreciation of the dollar.
The evaluation of exchange rate sensitivity during different
dollar regimes revealed a higher percentage of firms with
significant exchange exposures during the weak-dollar
regime. Downward price rigidity in the short run may
be one reason for this exposure pattern.
When we examined exchange risk exposures at the
industry level by grouping the firms into 20 portfolios based
on two-digit SIC codes, we found limited support for the
importance of the exchange rate factor. This may be
explained by the fact that although firms in a given industry
are in the same primary line of business, they are still
heterogeneous in terms of their operational and financial
characteristics. Since industry groups include firms with
positive and negative exchange risk exposure, aggregating
across such firms will result in finding an insignificant
exposure coefficient for the industry group.
We also found that the cross-sectional variation in
exchange risk sensitivity of individual firms is related to
firm-specific operational variables. Our estimations revealed
a positive relationship between the scope of the foreign
operations of a firmmeasured by foreign sales, assets, and
operating profitsand its exchange risk sensitivity. Future
studies that include additional operational and managerial
information collected from survey data should provide
further insights into the complex relationship between
exchange rate fluctuations and firm value.
References
Adier, M. and B. Dumas, 1984, "Exposure to Currency Risk: Definitions
and Measurements," Financial Management (Summer), 41-50.
Choi, J.J., 1986, "Model of Firm Valuation Under Exchange Exposure,"
Journal of International Business Studies (Summer). 153-159.
Bartov, E. and G.M. Bodnar, 1994, "Firm Valuation, Earnings Expectations,
and the Exchange-Rate Exposure Effect," Journal of Finance
(December), 1755-1786.
Choi, J.J., E. Elyasiani, and K.J. Kopecky, 1992, "The Joint Sensitivity of
Bank Stock Retums to Market Interest and Exchange Rate Risks,"
Journal of Banking and Finance (September), 983-1009.
Bodnar, G.M. and W.M. Gentry, 1993, "Exchange Rate Exposure and
Industry Characteristics: Evidence from Canada, Japan and U.S.,"
Journal of International Money and Finance (February), 29-45.
Choi, J.J. and A. M. Prasad, 1992, " Market Recognition of Exchange Risk:
Multinational versus Non-Multinational Firms," Working Paper,
Presented at the Annual Meeting of the Financial Management
Association, San Francisco.
We experimented with various grouping procedures based on SIC codes.
For example, we combined industry groups with similar characteristics (e.g.,
textile and apparel, department stores, and other retail), and we also further
decomposed an industry group into sub-groups using all four-digits of SIC
codes (e.g., electrical equipment (3600-3699) into electrical (3600-3669)
and electronic & computers (3670-3699)), in each case without significantly
different results. This supports our thesis that exchange risk sensitivity may
be more of a firm-specific phenomenon than an industry-specific one.
Dunning, J.H., 1973, "The Determinants of International Production,"
Oxford Economic Papers (November), 289-336.
Eaker, M.E., 1980, "Denomination Decision for Multinational
Transactions," Financial Management (Autumn), 23-29.
Elton, E.J. and M. Gruber, 1991, Modern Portfolio Theory and Investment
Analysis, 4th ed.. New York, NY, John Wiley.
88 FINANCIAL MAN/!,GEMENT/ AUTUMN 1995
Errunza, V.R. and L.W. Senbet, 1981, "The Effects of International
Operations on the Market Value of the Firm," Journal of Finance
(May), 401-418.
Eun, C.S. and B.G. Resnick, 1988, "Exchange Rate Uncertainty, Forward
Contracts, and International Portfolio Selection," Journal of Finance
(March), 197-216.
Fama, E. and K.R. French, 1992, "The Cross-Section of Expected Stock
Returns," Journal of Finance (June), 427-465.
Federal Reserve System, 1978, "Index of Weighted-Average Exchange
Value of the U.S. Dollar: Revision," Federal Reserve Bulletin
(August), 700.
Grammatikos, T., A. Saunders, and I. Swary, 1986, "Retums and Risks of
U.S. Bank Foreign Currency Activities," Journal of Finance (July),
671-682.
Hamada, R.S., 1972, "The Effect of Firm's Capital Structure on the
Systematic Risk of Common Stock," Journal of Finance (June),
435-452.
Hodder, J.E., 1982," Exposure to Exchange- Rate Movements," Journal of
tnternational Economics (November), 375-386.
Jorion, P., 1990, "The Exchange Rate Exposure of U.S. Multinationals,"
Journal of Business (July), 331-345.
Jorion, P., 1991, "The Pricing of Exchange Rate Risk in Stock Market,"
Journal of Financial and Quantitative A.nalysis (September), 363-376.
Ma, C.K. and G.W. Kao, 1990, "On Exchange Rate Changes and Stock
Price Reactions," Journal of Business Finance and Accounting
(Summer), 441-449.
Prasad, A.M. and M. Rajan, 1995," The Role of Exchange and Interest Risk
in Equity Valuation: A Comparative Study of International Stock
Markets," Journal of Economics and Business (forthcoming).
Sweeny, R. and A. Warga, 1986, "The F'ricing of Interest-Rate Risk:
Evidence from the Stock Market," Journal of Finance (June),
393-410.