The Impact of Exchange-Rate Volatility on Export

Growth: Some Theoretical Considerations and

Empirical Results

Martin J. Bailey, University of Maryland and U.S.

Department of State, George S. Tavlas, International
Monetary Fund, and Michael Ulan U.S. Department of State

This paper (i) examines the theoretical relationship between exchange-rate volatility
and export growth and (ii) tests for the empirical impact of such volatility on real
export growth of 11 OECD countries. We argue that, theoretically, exchange-rate
volatility can have an impact on trade in either a positive or negative direction.
Empirical results are provided for the managed-rate and flexible-rate periods. Both
nominal and real measures of exchange rates are used in two specifications of volatility:
absolute percentage changes and standard deviations. Of 33 regressions presented,
only three support the hypothesis that exchange-rate volatility impedes export
performance.

1. INTRODUCTION
The issue of whether exchange-rate volatility hampers trade flows
has been actively investigated in the literature. Earlier studies (e.g.,
Makin 1976; Hooper and Kohlhagen 1978) failed to find a relationship,
but these studies were based primarily on data prior to the move to
generalized floating of exchange rates in 1973. In recent years, a
substantial body of evidence dealing with the post-1973 period has
been produced, most of it supporting the proposition that exchange-
rate volatility does indeed impede trade (Coes 1981; Cushman 1983;
Akhtar and Hilton 1984; Thursby and Thursby 1986; Kenen and Rodrik
1986; Maskus 1986; De Grauwe 1986). The coverage of these recent
studies has been impressive. They have encompassed both total and
bilateral trade flows, differences in sampling data (i.e., time-series and
pooled times-series cross-sectional), alternative measures of exchange-

Address correspondence to George S. Tavlas, Room 9-210, International Monetary Fund,

Washington, D.C. 20431.
We are grateful to William G. Dewald for helpful comments. This research was initiated while
George Tavlas was Senior Economist, U.S. Department of State. The views expressed are the
authors' own and are not to be interpreted as those of their respective organizations.

Journal of Policy Modeling 9(1):225-243 (1987) 225

© Society for Policy Modeling, 1987 0161-8938/87/$3.50
226 M . J . Bailey, G. Tavlas, and M. Ulan

rate volatility, real and nominal exchange rates, and a range of in-
dustrial countries.
In contrast, only several recent studies have rejected the hypothesis
that exchange-rate volatility has had a systematically adverse impact
on trade. Gotur (1985) tests the robustness of Akhtar and Hilton's
empirical results and finds their methodology to be flawed in several
respects; the correction of the flaws has the effect of weakening their
conclusions. The present authors have recently investigated the effects
of nominal exchange-rate volatility----defined as the absolute percentage
change in the trade weighted exchange rate---on real exports of the
Big Seven OECD countries (Bailey, Tavlas, and Ulan 1986). We used
quarterly data over the interval 1973:1-1984:3 and employed our meas-
ure of exchange-rate volatility in both current-period form and
distributed-lag (eight-quarter) form. Testing a number of alternative
equation specifications, in no instance (out of a total of 28 equations)
were we able to find a negative and significant impact of exchange-
rate volatility on trade. Similar results were obtained in another recent
study, which included two of the current authors (see Aschheim,
Bailey, and Tavlas 1987); the latter study used an alternative measure
of relative export prices in regression equations.
This paper is an extension of our previous work. In Section 2 we
address the theoretical relationship between exchange-rate volatility
and trade. Previous studies have given only perfunctory treatment to
the theoretical aspect, implicitly assuming that a negative association
between volatility and trade exists in theory. In fact, as we demonstrate,
in theory, exchange-rate volatility may either deter trade or stimulate
it. Section 3 presents our model and describes our data. Empirical
results are provided in Section 4. In particular, we expand our earlier
work in several directions. Thus, for the Big Seven OECD countries,
we test for the effects of both nominal and real exchange-rate volatility,
using two measures of volatility. We also add observations to the post-
1973 sample data and present separate results for the pre-1973 period
in order to examine whether the relationship between exchange-rate
volatility and trade has changed between the prefloating and floating-

Iln particular, Aschheim, Bailey, and Tavlas (1987) used a ratio of export-price indices as a
measure of relative prices whereas Bailey, Tavlas~ and Uian (1986) used a relative-price variable
which also allowed for the effects of nominal-exchange-rate changes on relative export prices.
The latter procedure is the one used in this study; it is clearly preferable as it captures the effects
of changes in real terms of trade among countries. Thus, Bailey, Tavlas, and Ulan (1986) found
significantly higher relative,price coefficients than did Aschheim, Bailey, and Tavlas (t987).
IMPACT OF EXCHANGE RATE ON EXPORT GROWTH 227

rate periods. Finally, we test the exchange-rate-volatility-trade hy-

pothesis for several other OECD countries: Australia, the Netherlands,
New Zealand, and Switzerland.

2. THEORETICAL CONSIDERATIONS
The foregoing authors, and others, have addressed two questions
about the view that exchange-rate volatility hampers trade: Is it true
in theory, and is it true in practice? However, with the exception of
Farrell et al. (1983), the treatment of the theory is generally perfunctory
and seriously incomplete. It seems self-evident that exchange risk is
implicitly a cost, or is avoidable at an explicit cost, and that this risk
in a given transaction is proportional to the prospective variability of
the bilateral exchange rate applicable to that transaction. Neither part
of this pair of apparently self-evident propositons, however, fares well
under close examination. The theoretical case is, in fact, mixed and
uncertain--like the empirical evidence.
The usual analysis goes as follows. In a two-country context, con-
sider a firm located in country A that sells its product in country B.
For simplicity, suppose that the firm sells in a forward market in each
country so that the firm knows the future price of its product at the
time it incurs its costs of production. However, if there is no futures
or forward market for foreign exchange, the firm has an exchange risk
for the future conversion of its sales revenues from country B into the
currency of country A. If the firm is risk-averse, it would be willing
to incur an added cost to avoid this risk, so that the risk, if not hedged,
is an implicit cost. In its presence, this reasoning suggests that the
firm's supply price at each quantity of export sales is higher than in
the absence of the risk. For such firms in the aggregate, the quantity
of exports supplied at a given price is smaller with this risk than without
it. The same reasoning applies to firms in country B. If the risk is
present for firms in both countries, the supply curve of exports from
each country to the other is shifted to the left, compared to that in a
case of no risk. Trade is reduced in a way similar to the reduction
following an increase of transportation costs.
Where there is a forward market for foreign exchange, a discount
of the forward exchange rate in one direction, below the expected
future rate, is a premium in the other direction. Thus, if expectations
are similar in the two countries, such a discount cannot be a deterrent
to trade in both directions. However, the brokerage cost (spread) for
forward transactions is generally greater than that for spot transactions
in foreign exchange, and the spread is an increasing function of the
228 M.J. Bailey, G. Tavlas, and M. Ulan

variability of the exchange rate (see Akhtar and Hilton 1984.) Hence.
the risk can be hedged only at a cost; the existence of forward or futures
markets for foreign exchange does not change the thrust of the above
argument, though it reduces its quantitative significance.
Certain other considerations, however, call into question the entire
effect of variability on trade, either in principle or as measured. There
are two reasons to doubt that variability of a bilateral exchange rate
measures the risk to the firm connected with its bilateral trade. Further
offsetting this risk axe opportunities for profit that vary directly with
exchange-rate variability.
First, for a given set of influences on the balance of payments, if
the authorities intervene to reduce exchange-rate variability, there will
be a tendency to resort to exchange controls and other quantitative
restrictions. Such restrictions can be more costly to an exporting finn
than the exchange-rate variability they replace (IMF, 1984). Conse-
quently, there may be no correlation between measured exchange-rate
variability and the relevant cost to the firm.
Second, variability of an exchange rate does not measure the effect
added amounts of that foreign currency have on the overall riskiness
of the firm's asset portfolio. The latter risk effect depends on the
covariance of an exchange rate with the prices of the finn's other assets
as well as the own variance of the exchange rate. In particular, the
firm may hold a portfolio of several foreign currencies, thereby di-
versifying the risk. If variations inone currency's exchange rate against
the home currency axe negatively correlated with the variations in
others, its variability reduces portfolio risk rather than increasing it
when that currency is added to the portfolio. In general, variance by
itself does not measure the exchange risk (Farrell et al. 1983).
In addition, exchange-rate variability in any form other than a ran-
dom walk (where next period's rate is this period's rate plus a random
number with zero mean, or the same thing in logarithms,) presents
opportunities for profit. One canaotsay a priori how such opportunities
interact with the risks of foreign trade. Two worV~ng hypotheses have
become standard, or almost standard, assumptions in these matters.
First, it is standard to assume that "speculators" are different people
from traders, and that the latter tr&~,,sfer risk to the former for a fee.
Second, it is now wideay accepted that competition among speculators
eliminates all opportunities for profit, so that markets are "efficient";
if so, the exchange rate would in fact be a random walk. However.
each of these points can be overdrawn. Consider first the second point,
that relating to efficient markets. If speculators competed so effectively
that there was never any profit for them. due to zero incomes they
IMPACT OF EXCHANGERATE ON EXPORT GROWTH 229

would disappear. Their competition squeezes most of the rents out of

their incomes and forces out less skillful speculators. In that circum-
stance, exchange-rate movements will differ enough from a random
walk to allow them to earn their equilibrium incomes. On the first
point, there is no basis in evidence for the suggestion that businessmen
who trade in real goods continuously hedge their forward commitments
precisely, transferring all risk to speculators. In any fast-changing
business environment, price-affecting information is scarce and val-
uable, and these businessmen are likely to have proprietary access to
some of it. If so, they can earn income by using this knowledge in
the foreign exchange market.
A trader in real commodities can most conveniently use his spe-
cialized knowledge to speculate on future foreign exchange rate
changes by varying the extent to which he hedges his forward com-
mitments. If he has information that leads him to expect that the price
of a foreign currency will fall, he can sell futures in anticipation of
forward commitments, not yet made, to sell goods. If he has infor-
mation that leads him to expect the opposite, he can reduce his sales
of futures in that foreign currency below the amount on his order books
in that currency. He can also advance or delay his spot currency
transactions on the basis of such information. Of course, he could also
be more adventurous and speculate on a larger scale, if he is prepared
to venture outside the range of what most businessmen regard as pru-
dent business practice.
These considerations lead to no clear a priori conclusion about the
profit opportunities that may offset risk in foreign trade, when there
is the prospect of variability of the foreign exchange rate. Opportunities
to profit on specialized trade knowledge of fundamentals affecting
foreign exchange rates would tend to offset the trade-volume effects
of the costs of variability only if two assumptions, indicated by the
above discussion, are true:
1. The trader in real goods has specialized information on which
he can profit by varying his transactions in spot or future foreign
exchange, or by varying his rates of production and sales.
2. The size of his profit form such activity is positively correlated
with the volume of his business in real goods transacted with
that foreign exchange as the medium of exchange.
The second assumption would hold either if a larger volume of
business generates more information, or if portfolio considerations (in
the guise of "prudent business practice") lead to such a correlation.
It would also have to be true that doing business in real goods generates
230 M.J. Bailey, G. Tavlas, and M. Ulan

enough valuable information to place the businessman in the group

that decides to profit on such information. In all cases on valid test of
our hypothesis would answer the question whether this type of busi-
nessman obtains a more favorable exchange rate on his transactions
than a weighted average exchange rate, with the weights derived from
the flows of real goods over time. A second more complicated test
would look for a correlation between goods shipments and variations
of the exchange rate around its trend or moving average. Such tests,
however, are outside the scope of the present article. Instead, we shall
follow the more conventional path of testing for the correlations be-
tween trade volumes and variability of exchange rates, the broad hy-
pothesis under review.
In conclusion, variability of an exchange rate may either deter bi-
lateral trade or may stimulate it. The question is empirical, and two-
tailed tests of significance are necessary.

3. THE MODEL AND THE DATA

In Bailey, Tavlas, and Ulan (1986), we tested the impact of a single
measure of nominal exchange-rate volatility on exports of the Big
Seven OECD countries over the period 1973:1-1984:3. In this study,
we examine the effects of two measures of nominal exchange-rate
volatility on the real exports of each of these countries over two time
periods, 1962:2-1974:4 and 1975:1-1985:3 and of two measures of
real-exchange-rate change from 1975:1-1985:3 for these countries. For
three of our four new countries, Australia, the Netherlands, and Switz-
erland, we examined the effect of variability in real and nominal ef-
fective rates between 1975"1 and 1985:3. (For New Zealand, we tested
only nominal-exchange-rate variability; no real-effective-exchange-
rate series was available for that country.) The choice of 1975:1 as
the first observation of our second time period permitted us to employ
an eight-quarter lag on some exogenous variables without having any
of the lagged observations from the second subperiod extend into the
time of the fixed-rate regime. (Clearly, it is impossible to have a clean
break between subperiods if any of the variables in the regression are
lagged except by sacrificing scarce degrees of freedom.)
The estimated regressions are of the form:
log(Xi) = log a, + a2 log Yi + a3 log RPi (I)
+ a4 log OP + asVi + log ei
where X, is the volume of exports of country i, Y, is real GDP/GNP
of industrial trading-partner nations, RP~ is a measure of relative prices
of exports of country i to those of its trading partners, OP represents
IMPACT OF EXCHANGE RATE ON EXPORT GROWTH 231

real export earnings of oil-producing countries, and V is exchange-rate

variability. The residual e is assumed to be a random error term with
Gauss-Markov properties.
There are problems involved in devising proxies for the independent
variables. Theory tells us that income in trading-partner nations may
affect a country's exports. Quarterly GDP (GNP) data for eight OECD
nations (Australia, Austria, Canada, West Germany, Italy, Japan, the
United Kingdom, and the United States) are available for the period
under study. Similar data for LDC trading partners are not available.
In lieu of LDC GDP, we have employed another variable (described
below) which could affect the foreign purchases of oil producers, whose
transactions account for a very important segment of LDC's interna-
tional intercourse.
For each of the nations whose trade we examined here, we con-
structed a series for aggregate GDP (GNP) of the trading-partner in-
dustrial countries for which quarterly data are available over the 1962-
1985 period. These series were employed as our industrial-country-
income variable Yi. In order to aggregate national GDP series, it was
necessary to convert them to a common currency; we chose the U.S.
dollar. However, we wanted to ensure that our income variables were
affected by only changes in real incomes in trading-partner nations;
we did not want the variables to be affected by the changing foreign-
exchange value of the dollar. Thus we converted all GDP (GNP) data
to dollars at a set of fixed exchange rates. We valued trading-partner
income in dollars at first-quarter-1985 exchange rates.
A priori, our choice of the set of exchange rates at which to convert
national-currency income data to numeraire terms could affect the
calculated regression results. However, in our previous study we con-
structed two separate series for industrial-trading-partner income for
each of the Big Seven countries--the United States, the United King-
dom, Canada, France, West Germany, Japan, and Italy. We chose the
dollar as numeraire and converted foreign GDPs (GNPs) at 1980:3 and
1985:1 exchange rates, periods of a " l o w " and a " h i g h " dollar re-
spectively. The differences in the income parameter estimates calcu-
lated on the basis of the two series were negligible.
Since 1973, oil-exporting nations have provided important markets
for exports of OECD countries. However, unlike industrial countries,
oil-exporters' real GDPs are not good proxies for their national in-
comes. Hence, we posit that their ability to buy other nations' exports
with their own export earnings (rather than their real GDPs) is the
relevant " i n c o m e " variable to be included in a demand function for
other countries' exports. We used the dollar values of the oil-exporters'
232 M . J . Bailey, G. Tavlas, and M. Ulan

export earnings (as given by the IMF) deflated by the dollar-

denominated export-unit-value index of the industrial nations taken as
a whole to represent the real purchasing power Of the oil exporters as
it relates to industrial-country exports. We include both the industrial-
nation and oil-exporter "income" variables in our equations since the
"income" elasticities of demand for imports may not be similar for
the two groups of countries.
In theory, the relative-price variables in the equations should be the
ratio of export prices of country i to those of its major trading partners.
For our relative-price variables, the RP's, we have used the ratios of
the dollar-denominated export unit values for each country relative to
the dollar export unit values for the IMF's "industrial-country" ag-
gregate. 2 Since they are dollar-denominated, changes in the ratios are
changes in real terms of trade, reflecting the impacts of changes in
nominal exchange rates, differing rates of inflation among countries,
and changes in relative prices in each country between its nontraded
goods and its exports. Hence, the ratios reflect real exchange rates in
terms of traded goods.
Exchange-rate volatility can be measured by using either nominal
or real rates. Since persuasive arguments can be made for either choice,
we have tested both measures when both were available. For the period
beginning in 1970, real and nominal effective exchange rates for all
of the countries under study here except New Zealand have been
constructed and published by Morgan Guaranty Bank. Hence, it is this
set of exchange rates on which the exchange-rate volatility variables
we tested are based. Using the same weights as Morgan Guaranty (total
bilateral trade---imports plus exports -in manufactures with each of
15 industrial countries in 1980), we constructed nominal effective-
exchange-rate series for the Big Seven countries for the period 1960-
1969. For New Zealand, we constructed a "Morgan-weighted" nom-
inal effective exchange-rate series for the period 1973-1985:3.
For each country, Morgan constructs real effective exchange rates
by adjusting the couna-y's nominal effective exchange rate by the
differential movemeats between the mamffactures' segments of trading
partners" domestic wholesale price indices and that of the manufactures
segment of the subject nation's WPI. Since we do not have information
on the ~ components of foreign WPIs, we were not able

~l'he exception is Ca_n_~ For that counlry, we used its dollar-denominated export unit value
divided by the U.S. dollar-denominated export unit value as the relative price term. The reason
for doing so is the strong dependence of Canadian exports upon the U,S. market. In 1984, for
example, almost three-fourths of Canadian exports were to the United States.
IMPACT OF EXCHANGERATE ON EXPORT GROWTH 233

to extend the Morgan real exchange rates back in time or to construct

a real effective exchange rate for New Zealand.
Other studies have tended to use one of two measures of volatility
in the empirical literature on exchange-rate volatility and trade. In one
previous study we (see Bailey, Tavlas, and Ulan 1986) used a poly-
nomial distributed lag of the absolute value of the quarter-to-quarter
percentage change in the exporting nation's effective exchange rate:
Vi., = I(Ei.,- E,.,_,) Ei, _, I.
In this expression, E is the effective exchange rate (nominal or real)
of the currency of exporting nation i. Others have employed volatility
measures based on the moving standard deviations of exchange rates.
The percentage-change variable tests for a stable and significant
response of exports per 1% change in the exchange rate. The standard-
deviation variable tests for a stable and significant response of exports
per one-percentage-point change in the exchange rate. Hence, an au-
thor' s choice of volatility measure depends essentially on his perception
of the behavior of agents in the market.
Since the literature is divided with respect to the choice of a vari-
ability variable, we tested several measures of exchange-rate volatility
for each exporting nation---one measure per equation. One variable
we tested was the variable cited in Equation (2) above. We tested this
absolute-percentage-change measure for both the nominal and real
effective exchange rates. In the results reported below, this percentage-
change volatility variable enters the equations as an eight-period
second-degree polynomial distributed lag over quarters t through t-7.
We used this specification in order to capture delayed responses of
exports to exchange-rate volatility.
In addition, we tested the logarithms of the moving standard devia-
tions of nations' nominal and real effective exchange rates over eight
quarters as "volatility" measures. Because they are calculated over
an eight-quarter period, these variables appear in the equations in
unlagged form.
With regard to the other regressors in our equations, we sought to
avoid excessive manipulation of lag structures which could have the
effect of employing search procedures. Hence, we tested the Y variable
in current-period form, the OP variable in current-period form and
with a one-quarter lag, and the RP variable with an eight-period,
second-degree polynomial distributed lag as well as in current-period
and one-quarter-lag form. Departures from the foregoing guidelines
were as follows: For the UK equation, the OP variable is an eight-
period second-degree polynomial lag, and for Australia, the relative-
price series is a two-period moving average.
234 M . J . Bailey, G. Tavlas, and M. Ulan

4. EMPIRICAL RKSULTS
Our estimation procedure was as follows: First, we estimated a set
of preferred equations, based on theoretical considerations, for the Big
Seven countries (one equation for each country) over the period
1975:1-1985:3. The regressors were Y, RP, and OP. Next, we ex-
amined in each preferred equation separately the impact of: (i) nominal
exchange-rate volatility in terms of absolute percentage changes; (ii)
nominal exchange-rate volatility as measured by the moving standard
deviation; (iii) real exchange-rate volatility in terms of absolute per-
centage changes; and (iv) real exchange-rate volatility as measured by
the moving standard deviations. We then tested the impact of each of
the measures of nominal volatility in the same (except for France)
preferred specifications over the period 1962:2-1974:4. Finally, we
examined the effects of nominal and real exchange-rate volatility as
measured by absolute percentage changes on exports of the four smaller
OECD economics over the period 1975:1-1985:3.
Where necessary, a serial-correlation correction was performed on
the equations using a maximum likelihood procedure. A nonparametric
test for the presence of seasonality (assuming stability) using the
Kruskal-Wallis statistic was performed on real e ~ for each country
and for real oil revenues. All the export series displayed evidence of
seasonality at the 0.1% level so that these data were seasonally adjusted
using the X-11 ARIMA technique. The data used in constructing the
OECD GDP variable had already been seasonally adjusted.
The regression results are reported in Tables 1-3. Table 1 presents
the equations which include the effects of both nominal and real
exchange-rate volatility--as measured by absolute percentage
changes--on real exports of the Big Seven countries. Equations with
the suffix (a) for each country refer to the prefloating period. Equations
with the suffixes (b) and (c) refer to the floating-rate period; Equations
(b) include the effects of nominal exchange-rate variability and Equa-
tions (c) incorporate the impacts of variability in real exchange rates.
The equations demonstrate that some structural change occurred
between the two estimation periods, an inference which is verified by
stability tests performed by the authors.3 This situation was particularly
apparent in equations for France, where several relative-price variables

3Using the Chow ~ t o test for structural change. These results are not reported but are
available from the authors upon request.
IMPACT OF EXCHANGE RATE ON EXPORT GROWTH 235

tested in the second estimation period proved to be insignificant. When

we tried these relative-price terms in the earlier period, they were
negative and very significant. Hence, we opted to include a relative-
price term for France in the earlier period though it was not part of
the preferred equation derived for the second period.
The equations show that the dominant determinant of Big Seven real
exports over both estimation periods is real economic activity in the
remainder of the OECD. For the most part, the coefficients on this
variable are quite stable, although they do fall significantly in the
second period in the equations for Canada, Germany, and Italy. Rel-
ative prices are more important in the first period in equations for
France, Japan, and the United Kingdom; they are more important in
the second period in equations for Canada, Italy, and the United States;
they show little change in the case of Germany. Hence, the results on
the relative-price variables entail mixed results at best for the price-
elasticity-pessimism thesis put forth by Niehans (1975) and McKinnon
(1978), which posits that price elasticities of demand for traded goods
will be lower under floating rates. 4 The variable representing oil ex-
porters' export earnings contributes less to the explained variances of
the equations than do the variables representing OECD economic ac-
tivity or relative prices. In the equation for Canada, it was insignificant
and dropped. In equation (1 b) for Germany it has a t statistic of only
1.4.
Nominal exchange-rate volatility--as measured by the absolute per-
centage change----does not impact negatively and significantly on real
exports in any of the Big Seven countries over the period 1975:1-
1985:3 (Table 1). This finding is consistent with that of our previous
study which encompassed the period 1973:1-1984:3. Nor does it show
a negative and significant impact on real exports in any of these coun-
tries over the earlier period. Nominal exchange-rate volatility does
show a significant and positive coefficient in several equations covering
the earlier period (i.e., in the equations for France, Germany, and
Japan). The results are not so overwhelming in the case of absolute
percentage changes in real exchange rates [Equations (c) in Table 1].

4For example, McKinnon (1978, p. 4) argues that: "with the advent of floating, the future
direction of exchange rate movements has proved highly uncertain . . . . And it may not be in the
interests of merchants to engage in active arbitrage in industrial commodities if tomorrow's
exchange rate is unknown. Hence, the quantities of goods traded respond sluggishly to exchange
rate fluctuations giving response to a modem version of elasticity pessimism."
Table 1: Effects of Exchange-Rate Variability on Export Volumes of Big Seven OECD Countries

OECD Relative Real Oil Exchan~e Rate Estimation

Country Equation Constant GDP Prices Revenues Variability p DW Pert~
Canada (la) -9.14 1.97 -0.85 3,16 0.40 0.989 1.98 62:2-74:4
(1i9.6) (28.0) (0.9) (1.4) (2.6)
(Ib) -5.13 1~38 -t.12 1,28 0.38 0.935 2, t8 75:1-85:3
(2.2) (4.1) (2.,3) (0;5) (2.2)
(lc) -5.01 !,36 -1.04 -0.09 0.43 0.938 2.23 75" 1-85:3
(2, l) (4.0) (2.1) (0.03) (2.6)
France (la) -6,51 1.55 -3.03 0.12 6~82 0.987 2.41 62:2-74:4
(10.8) (!8.1) (4.5) (4.1) (3.6)
(lb) -6.24 1..53 0.11 0.72 0.985 1.54 75; 1-85:3
(21.3) (37,7) (7.8) (1.3)
(lc) -6.52 1.57 O. 12 1:42 0.984 1.58 75:1-85:3
(17.2) (29.8) (7.3) (1.7)
Germany (la) -8.46 1,85 -0.55 0.05 4.16 0.34 0.992 2.09 62:2-74:4
(11.9) (19,5) (2.1) (0.9) (2.5) (2.3)
(Ib) -5,09 1.37 -0.54 0.06 -0,96 0.6t 0.976 2.02 75:1-85:3
(3. 3) (6.4) (2.2) (1.4) (0.4) (3.5)
(lc) -3.67 1.18 -0.41 0.06 -6.60 0.36 0.981 1.93 75:1-85:3
(4.4) (10.2) (2.5) (1~9) (3.4) (2.0)
Italy (la) -12.94 2.49 -0.52 -0.03 -0.77 0.986 1.84 62:2-74:4
(10,1) (14.1) (0.7) (0.5) (0.3)
(lb) -4.80 1.34 -0.90 0.13 -2.24 0.896 2.04 75:1-85:3
(3.8) (7.7) (1.9) (1,9) (1.6)
(lc) -2.30 0.99 -0.86 0.13 =6.83 0.909 2.34 75:1-85:3
(1.4) (4.3) (2.0) ~2.2) (2.8)
Japan (la) -4.86 1.48 -4.76 0.20 4.91 0.80 0.994 1.95 62:2-74:4
(1.0) (2.1) (3.6) (3.9) (4.3) (8.2)
(lb) -6.90 1.66 -1.47 0.11 -0.93 0.96 0.985 2.31 75:1-85:3
(2.6) (4.4) (3.0) (2.2) (0.7) (21.0)
(lc) -7.15 1.70 -1.39 0.09 -2.26 0.95 0.986 2.44 75:1-85:3
(2.9) (4.9) (2.9) (1.8) (1.3) (17.8)
U.K. (la) -4.88 1.22 -3.64 -0.06 0.31 0.955 2.43 62:2-74:4
(1.9) (4.0) (4.9) (0.3) (0.1)
(lb) -7.18 1.69 -0.70 0.25 -1.77 0.927 2.48 75:1-85:3
(5.6) (8.9) (3.0) (2.4) (1.13)
(lc) -5.91 1.49 -0.46 0.11 0.50 0.921 2.30 75:1-85:3
(4.5) (7.7) (2.4) (1.1) (0.3)
U.S. (la) -3.24 1.09 -0.37 0.11 3.35 0.946 1.67 62:2-74:4
(4.2) (10.2) (2.1) (2.1) (0.9)
(lb) -2.62 1.02 -0.63 0.14 0.78 0.57 0.901 2.01 75:1-85:3
(2.3) (6.3) (3.8) (2.8) (0.5) (3.5)
(lc) -3.03 1.07 -0.66 0.14 1.20 0.54 0.908 1.94 75:1-85:3
(2.7) (6.8) (4.1) (2.7) (0.9) (3.4)

Real OECD GDP is real GDP (current-period) in national-currency terms for 12 countries, converted to U.S. dollars at 1985:1 exchange rates. The
relative-price variable for all countries except Canada is the nation's dollar-denominated export-unit-value index divided by the IMF's "industrial-country"
export-unit-value series. For Canada, it is that nation's U.S.-dollar-denominated export-unit-value index divided by the U.S. export-unit-value index
(denominated in U.S. dollars). For Canada, France [Equation (la)], Germany, Italy, Japan, the United Kingdom, New Zealand, and the Netherlands, the
relative-price variables were constructed using eight-period (t through t-7) second-degree polynomial distributed lags. For the U.S., relative prices are
entered with a one-period lag; for Australia, they enter with a two-period moving average. All export-unit-value series are quarterly averages. Real oil
revenues are in current-period form for all countries except Japan (lagged one period), the United States (lagged one period), and the United Kingdom
(eight-period t through t-7--second-degree Almon lag). Exchange-rate variability in Tables 1 and 3 is the absolute value of the quarterly percentage change
in the country's nominal effective exchange rate; its impact is estimated by using an eight-period (t through t-7) second-degree Almon lag. In Tables 1
and 3, exchange-rate variability is the absolute value of the quarterly percentage change in the country's nominal and real (except for New Zealand)
exchange rates; its impact is estimated by using an eight-period (t through t-7), second-degree Almon lag. In Table 2, exchange-rate variability is the
logarithm of the standard deviation of the exchange rate over eight quarters. For Australia, a shift dummy variable (reported in "Other" column in Table
3) was" used to account for policy shift from managed exchange rate system to floating rates in December 1983; the shift dummy equals unit, in 1983:4- t~
t.,o
1985:3, and zero in earlier period.
Sources: IMF, International Financial Statistics; Morgan Guaranty Bank; and authors' calculations.
238 M.J. Bailey, G. Tavlas, and M. Ulan

In two instances--the German and Italian equations--a negative and

significant coefficient was found. These latter results do not seem to
be influenced by the possible existence of coltinearity between the
relative-price variables and the real-exchange-rate-volatility variables
(even though the denominator of the latter incorporates relative prices).
As evidence of the absence of a multicollinearity problem, note that
the coefficients on the relative price terms in the German and Italian
Equations (lc) showed little change from their values in Equations
(lb). Hence, the emergence of negative and significant coefficients on
the real-exchange-rate-volatility terms in these equations does not ap-
pear to be due to collinearityAnduced t ratios which would be unre-
liable. However, Equations (c) for Germany and Italy do show sharp
falls in the coefficients on the OECD income variables compared to
Equations (b) for these countries.
Table 2 reports results for the Big Seven countries based on the
standard-deviation measure of exchange rate volatility; both nominal
[equations denoted as (a)] and real [equations denoted as (b)]. The
equations were estimated over 1975:1-1985: 3. In only one instance--
that of Italian Equation (2b)---did the exchange-rate-volatility measure
show a negative and significant coefficient. However, in Italian Equa-
tion (2b), while the t statistic on the volatility term was significant, it
was (at 1.9) only marginally so. In the United Kingdom and the United
States Equations (2b), the volatility term was positive and significant
(though marginally so for the United Kingdom).
Table 3 presents results for Australia, New Zealand, the Netherlands,
and Switzerland, using nominal and real (except for New Zealand)
measures of absolute percentage changes of exchange rates over
1975:1-1985:3. Exchange-rate volatility did not impact negatively and
significantly on real exports of these countries in any of the equations.

5. CONCLUDING REMARKS
This paper tested for the impact of exchange-rate volatility on real
exports of 11 OECD eomatries. For the Big Seven, we used two meas-
ures of volatility for both real and nominal exchange rates. Our results
run contrary to those found in much of the recent empirical literature.
The importance of the possible exchange-rate-volatility-trade linkage
for the future of the flexible-exchange-rate regime is self-evident. Over
the flexible-rate period, we present 33 regression equations. Consistent
with theoretical conside,rations, in several instances we found a positive
and significant association between exchange-rate variability and real
exports. In only three instances---one of which was marginal--did we
T a b l e 2: Effects of Exchange-Rate Variability as Measured By M o v i n g Standard Deviations ( 1 9 7 5 : 1 - 1 9 8 5 : 3 )

Real
OECD Relative Real Oil Exchange Rate
Country Equation Constant GDP Prices Revenues Variability p ~2 DW

Canada (2a) -5.95 1.52 -0.96 0.03 0.31 0.941 2.11

(2.7) (4.7) (2.1) (1.7) (1.9)
(2b) -5.75 1.46 -1.06 0.03 0.35 0.939 2.16
(2.4) (4.3) (2.3) (1.2) (2.1)
France (2a) - 5.95 1.49 0.09 0.01 0.41 0.985 1.89
(15.5) (26.6) (4.4) (0.6) (2.7)
(2b) -5.57 1.44 0.08 -0.01 0.39 0.985 1.91
(10.6) (20.4) (4.2) (0.7) (2.5)
Germany (2a) -5.14 1.40 -0.51 0.06 -0.0007 0.60 0.976 1.97
(4.9) (9.2) (2.1) (1.5) (0.1) (3.9)
(2b) -4.75 1.35 -0.52 0.06 -0.02 0.56 0.976 1.99
(5.0) (10.2) (2.4) (1.5) (0.8) (3.1)

t,~
q~
Table 2: Effects of Exchange-Rate Variability as Measured By Moving Standard Deviations (1975:1-1985:3)

Real
OECD Relative Real Oil Exchange Rate
Country Equation Constant GDP Prices Revenues Variability p ~2 DW

Italy (2a) -6.11 1.52 -0.62 0.07 -0.01 0.892 1.76

(4.6) (7.6) (1.3) (1.0) (0,3)
(2b) -4.17 1.27 -0.49 0.10 -0.04 0.902 2.02
(3.0) (6.8) (1.1) (1.8) (1.9)
Japan (2a) -4.99 1.43 -1.45 0.12 -0.01 0.98 0.982 2.28
(i.7) (3.3) (2.4) (2.2) (0,5) (38.4)
(2b) --4.93 1.44 -1.18 0.11 -0.03 0.98 0;983 2.33
(1.7) (3.4) (2.2) (2.1) (1.4) (35.6)
U.K. (2a) -6.25 1.54 -0.51 0;13 -0.01 0.920 2.22
(6.3) (11.0) (2.8) (1.6) (0.3)
(2b) -5.30 1.37 -0.42 0.01 0.05 0.926 2.37
(4.9) (8.4) (2.6) (0.1) (1,8)
U.S. (2a) -1,74 1.02 -0.65 0i13 0.01 0.62 0.902 1.88
(1.6) (6.1) (3.5) (2.4) (0.2) (4.3)
(2b) -1.98 1.04 -0.80 0.13 0.03 0.53 0.911 1.95
(2.3) (7.6) (5.6) (2.8) (2.1) (3.5)
Table 3" Exchange-Rate Variability and Export V o l u m e s - - S m a l l e r OECD Countries (1975:1-1985:3)

Real
OECD Relative Real Oil Exchange Rate
Other
~2 DW
Country Equation Constant GDP Prices Revenues Variability P

0.06 0.57 0.850 1.59

Australia (3a) -3.01 1.07 -0.25 1.47
(1.8) (4.6) (0.8) (0.9) (1.2) (4.0)
0.09 0.51 0.848 1.59
(3b) -2.46 0.99 -0.38 0.58
(1.6) (4.6) (0.7) (0.3) (2.1) (3.3)
0.902 1.89
New Zealand (3a) -10.50 2.12 -0.60 1.96
(7.9) (11.3) (1.6) (1.9)
0.35 0.954 1.93
Netherlands (3a) -3.54 1.15 -0.80 0.07 0.51
(6.1) (14.2) (2.5) (2.0) (0.3) (2.0)
0.14 0.956 1.91
(3b) -3.43 1.13 -0.98 0.11 1.44
(7.6) (17.9) (4.0) (3.2) (1.5) (0.8)
0.25 0.927 2.06
Switzerland (3a) -6.27 1.53 0.02
(8.6) (15.2) (0.02) (1.5)
0.32 0.923 2.03
(3b) -6.11 1.51 -0.35
(8.5) (14.8) (0.5) (2.0)

4~
242 M . J . Bailey, G. Tavlas, and M. Ulan

find a significant and negative relationship between exchange-rate

volatility and real trade. These three regressions all included real-
exchange-rate volatility variables. So perhaps it is real volatility that
is at issue. Considering only those equations with real-exchange-rate-
volatility variables [equation sets (lc), (2b), and (3b)], we find only
three of 17 instances where exchange-rate volatility negatively and
significantly affects real exports.
Given the divergence between our results and those of other aggre-
gate-trade studies, we find the state of the empirical evidence trouble-
some. Perhaps exchange-rate volatility affects bilateral trade flows.
Certainly, some of the recent empirical literature persuasively supports
this hypothesis. Further work is surely needed. Consequently, we will
make available our data files to any researchers who wish to replicate,
correct (if need be), or improve upon our work. We hope authors of
other recent studies will be able to do the same.

REFERENCES
Akhtar, M.. and Hilton, R. S. (1984) Effects of Exchange Rate Uncertainty on German and
U.S. Trade, Federal Reserve Bank of New York Quarterly Review 9: 7-16.
Aschheim, J., Bailey, M. J,, and Tavlas, G. S. (1987) Dollar Variability, the New Protectionism,
Trade and Financial Performance. In The New Protectionist Threat to Worm Welfare (D.
Salvatore, Ed.), Amsterdam: North-Holland.
Bailey, M. J., Tavlas, G. S,, and Ulan, M. (1986) Exchange-Rate Variability and Trade Per-
formance: Evidence for the Big Seven Industrial Countries, WeltwirtschaftlichesArchly 122:
466~t77.
Coes, D. (1981) The Crawling Peg and Exchange Rate Uncertainty. In Exchange Rate Rules:
The Theory, Performance and Prospects of the Crawling Peg (J. Williamson, Ed.). New
York: St. Martin's Press.
Cushman, D. V. (1983) The Effects of Real Exchange Rate Risk on International Trade. Journal
of International Economics 15: 45-63.
DeGrauwe, P. (1986) Exchange Rate Variability and the Slowdown of International Trade,
preliminary draft, August.
Farrell, V. S., DeRosa, D. A., and McCowan, T. A. (1983) Effects of Exchange Rate Variability
on International Trade and Other Economic Variables: A Review of the Literature, Federal
Reserve Board Staff Studies, No. 130.
Gotur, P. (1985) Effects of Exchange Rate Volatility on Trade: Some Further Evidence, IMF
Staff Papers 32: 475-512.
Hooper, P., and Kohlhagen, S. W. (1978) The Effects of Exchange Rate Uncertainty on the
Prices and Volumes of International Trade, Journal of International Economics 8: 483-
511.
International Monetary Fund (1984) Exchange Rate Volatility and World Trade, Occassional
Paper No. 28, July.
Kenen, P. B., and Rodrik, D. (1986) Measuring and Analyzing the Effects of Short-Terna
Volatility in Real Exchange Rates, The Review of Economics and Statistics 68:311-15.
Makin, J. (1976) Eurocurrencies and the Evolution of the International Monetary System. In
IMPACT OF E X C H A N G E RATE ON EXPORT G R O W T H 243

Eurocurrencies and the International Monetary System (J. Makin and D. Logue, Eds.).
Washington, D.C.
Maskus, K. E. (1986) Exchange Rate Risk and U.S. Trade: A Sectoral Analysis, Federal Reserve
Bank of Kansas City Economic Review 7 I: 16-23.
McKinnon, R. I. (1978) Exchange Rate Instability, Trade Imbalances, and Monetary Policies
in Japan and the United States, mimeo, Stanford University.
Niehans, J. (1975) Some Doubts About the Efficacy of Monetary Policy Under Flexible Exchange
Rates, Journal of International Economics 5: 274-281.
Thursby, J., and Thursby, M. C. (1986) Bilateral Trade Flows, the Linder Hypothesis, and
Exchange Risk, University of Western Ontario Working Paper 8612C, April.