This material has been published in the Journal of the Japanese and International Economies, 15: 178-198
(2001), the only definitive repository of the content that has been certified and accepted after peer review.
Copyright and all rights therein are retained by Academic Press. This material may not be copied or reposted
without explicit permission.
Exchange-rate Strategies in the Competition for Attracting Foreign Direct Investment*
Agnès BÉNASSY-QUÉRÉ
University of Paris X – Nanterre (THEMA) and CEPII, 200 avenue de la République F-92000 Nanterre,
a.benassy@cepii.fr
Lionel FONTAGNÉ
CEPII and TEAM, 9 rue Georges Pitard F-75015 Paris,
fontagne@cepii.fr
and
Amina LAHRÈCHE-RÉVIL
University of Amiens (CRIISEA), CEPII and TEAM, 9 rue Georges Pitard F-75015 Paris,
lahreche@cepii.fr
Journal of the Japanese and International Economies,15: 178-198 (2001).
* We are grateful to Bill Amis, Philippe Bacchetta, Benoît Coeuré, Daniel Cohen, Jean-Louis Guérin,
Daniel Laskar, Paulo Mauro, Hélène Raymond, André Sapir, Khalid Sekkat, Shinji Takagi, and to an
anonymous referee, for helpful comments on an earlier draft of this paper. All errors remain ours.
Abstract
Building on the needs for long-term capital inflows in developing countries, this paper
reconsiders the choice of an exchange-rate regime by integrating the determinants of
multinational firm’s locations. The trade-off between price competitiveness and a stable
nominal exchange rate is modeled. Empirical results show that exchange-rate volatility is
detrimental to foreign direct investment, and that its impact compares with that of
misalignments. One policy implication is that the building of currency blocks could be a way
of increasing FDI to emerging countries as a whole. The frontiers of monetary areas would
then be strongly influenced by geography, as FDI is.
JEL classification numbers: F21, F23, F31, F33.
Keywords: Exchange-rate regimes, foreign direct investment, currency blocks.
2
1. Introduction
Empirical studies have shown that the structure of capital inflows in a low-developed country
is not neutral for growth and for macroeconomic stability. In particular, foreign direct
investment (FDI) is a more stable source of financing than portfolio investment (Lipsey,
1999), and it raises the global factor productivity through technological spillovers
(Borenzstein and De Gregorio, 1995). The financial crises of 1997-1998 have reinforced that
point of view: short-run capital flows have been pointed out as one of the major causes of the
crises, both as inflows (through an excess of credit) and as outflows (triggering default).
In order to re-orient capital flows to longer-term investment, the needs for microeconomic
reforms are often stressed (see, for instance, Stiglitz, 1999). In addition, some economists
have been advocating capital controls (for instance, Cooper, 1999). This issue has generally
been separated from the debate on the choice of an exchange-rate regime which highlights the
difficulty of intermediate regimes in a world of free capital flows (see, for instance,
Eichengreen, 1999 versus Frankel, 1999). However, exchange-rate regimes themselves may
affect the composition of capital inflows: while portfolio investors should be indifferent to the
exchange-rate regime as long as derivative markets allow them to hedge, foreign direct
investors should conversely worry about the exchange-rate regime because they cannot hedge
at their horizon and are mainly interested in macroeconomic variables such as relative labor
costs or purchasing power.
This suggests reconsideration of the choice of an exchange-rate regime by integrating the
determinants of location choices by multinationals. Traditional determinants like comparative
costs of production are affected by the level of the real exchange rate. In addition, the new
trade theory stresses the so-called "proximity-concentration trade-off" (Brainard, 1993), a
combination of increasing returns to scale and transportation costs explaining location choices
3
of multinational companies. Lastly, a recent strand of papers looks at the location choices of
risk-averse firms facing exchange-rate variability: the latter affects the location of firms and
hence the degree of specialization of countries (Goldberg and Kolstad, 1994; Ricci, 1997;
Fontagné and Freudenberg, 1999). Hence it is worthwhile adopting an exchange-rate regime
capitalizing on micro-economic location strategies of risk-averse firms facing uncertainty, in
order to attract the stable part of international capital flows, namely FDI.
Here, we provide a framework for designing an exchange rate strategy aiming at attracting
FDI. We explicit the trade-off between price competitiveness and a stable nominal exchange
rate. We also show that exchange-rate regimes in other emerging countries affect FDI in each
specific country. Our findings justify the building of currency blocks as a way of increasing
the total amount of FDI towards developing countries and hence providing stability to
international capital flows.
Section 2 presents a theoretical model showing the impact of both the level and the volatility
of the exchange rate on FDI. The model also highlights the interaction between host countries
through their exchange-rate regimes. Empirical evidence of these effects is provided in
Section 3 on the basis of a panel of 42 developing countries receiving FDI from 17 OECD
countries. Section 4 draws the policy implications and Section 5 concludes.
2. The theoretical model
2.1 Background
In a developing country with a relatively high inflation, the exchange-rate regime results from
a trade-off between (a) nominal exchange-rate stability associated with cumulated real
appreciation of the currency, and (b) nominal volatility associated with little or no trend in the
4
real exchange rate. This trade-off can be made on the ground of various objectives and
constraints, attracting FDI being one possible objective in an emerging country.
As far as attracting FDI is concerned, both the level and the volatility of the exchange rate
have to be taken into account, since they affect FDI. The relationship is, however, ambiguous
and depends on the destination of the goods produced.
As far as the impact of the level of the exchange rate on FDI is concerned, two relationships
can theoretically be observed. First, if the investor aims at serving a local market where trade
or nontrade barriers are impediments to enter the market, FDI and trade are substitutes, and an
appreciation of the local currency in real terms lifts inward FDI, because the purchasing
power of consumers is increased, and also because barriers to trade usually tend to increase in
such a context. Alternatively, if the output from FDI is to be re-exported, trade and FDI are
complements; an appreciation of the local currency, because it brings competitiveness down
(higher labor and capital costs) and lowers the relative wealth of foreign investors, reduces
inward FDI.
Unfortunately, available data does not disentangle the various motivations of investors, and
existing studies have to rely on aggregate data. They generally show that a depreciation in the
exchange rate induces more FDI inflows (see Cushman, 1988; Froot and Stein, 1991; Ito et
alii, 1996; Goldberg and Klein, 1997; Barrel and Pain, 1998).
Note that the impact of local costs (and hence, of the real exchange rate) on FDI must be
qualified according to transportation costs and possible increasing returns (Horstman and
Markusen, 1992, Brainard, 1993, Markusen, 1995). When the multinational firm intends to
sell on the host market, transportation costs reinforce the incentive for producing locally.
Conversely, transportation costs should act as impediments to FDI if production is to be reexported, for instance to the home market of the firm. In both cases, high transportation costs
limit the benefits of concentrating the production in a small number of locations.
5
Turning to exchange rate volatility, its impact on FDI is ambiguous too. In the line of Dixit
and Pindyick (1994), Darby et al. (1999) emphasize the value of the option to wait in a
situation of uncertainty and sunk costs. Notwithstanding such option, exchange rate volatility
affects FDI in various ways. Cushman (1988) advocates that producing in the destination
market is a good substitute for exports if there is a strong uncertainty on exchange rates. But
this benefit vanishes if the production is partially re-exported. Hence, a foreign firm facing
large exchange rate volatility will produce in the local country if it intends to sell on the local
market, but refrain from doing so if it intends to re-export. From an empirical point of view,
Cushman (1988) finds a positive impact of volatility on outward FDI.
Lastly, the correlation (i.e. the relative behavior) of exchange rates can affect the location
strategies of risk-averse firms: risk aversion should lead firms to diversify across possible
locations. Indeed, according to Aizenmann (1992), investing abroad and bearing an exchangerate risk means, broadly speaking, buying the option to face alternative sets of production
costs if a shock occurs.1 In line with this approach, “countries whose exchange rates are
negatively correlated with global returns to capital (such as oil-exporting countries) may
actually benefit from their role as portfolio hedges. An increase in these countries’ exchange
rates may actually raise their FDI inflows on diversification grounds” (Ito et alii, 1996, p. 54).
Similarly, investing in countries whose exchange rates are negatively correlated to other
exchange rates should be a way of diversifying FDI. In this paper, we investigate this
possibility and highlight its role in the choice of an exchange-rate regime.
2.2 The model
We consider the case of a risk-averse multinational firm which contemplates locating in two
alternative foreign countries in order to re-export.2 Like Cushman (1988), we follow a twoperiod framework. In the first period, the firm decides where to invest. The decision to locate
abroad is already taken although the investor does not yet know where to do it. There is no
6
option to wait, as opposed to the strand of literature launched by Dixit and Pindyck (1994).
The capital cost and the interest rate are certain. In the second period, production is
undertaken by the firm, and sold on its own home market (in its own currency).
The firm faces labor costs, which depend on local wages (which are known and do not react
to exchange rate variations) and on the evolution of the exchange rate (which is unknown by
the time the firm invests). Of course, in reality, nominal wages react to exchange rate
variations, to an extent that depends on indexation mechanisms. Indeed, it would have been
preferable to consider the uncertainty on wages converted into the investor’s currency, or on
the real exchange rate, rather than the nominal exchange rate uncertainty. However, monthly
or even quarterly data is often missing for wages and prices in developing countries. As a
matter of fact, wage variations are smoother and delayed compared to exchange rate
fluctuations (see Appendix 1 for an illustration). The uncertainty surrounding the nominal
exchange rate therefore dominates that surrounding nominal wages, which justifies the
simplifying assumption of nominal wage rigidity when measuring cost uncertainty.
The firm also faces a capital cost, which is known by the time it invests. We assume that the
multinational firm gets financing from the world capital market; hence, the nominal interest
rate is the same for both locations.3 The capital has no residual market value after the
production period. The latter assumption can be justified by the high level of sunk costs.
In our model, the price is given in the investor’s currency. We also assume that inward FDI
has no impact in terms of real appreciation: they can be assimilated to a positive supply shock
and do not carry inflation pressures in the host economy (this point is discussed in Section 3).
Hence, we assume that, in a free-floating regime, the variations of the nominal exchange rate
against the investor’s currency are exogenous, and that the covariance with the alternative
host’s exchange rate is also exogenous. Noting Si the nominal exchange rate of country i
(i=1,2), mi its average value (which can be interpreted as the fundamental value), pi a variable
7
summarizing the exchange-rate regime (pi equals 0 in case of a free float and 1 in case of a
fixed peg against the investor’s currency), and εi a white noise on the corresponding foreign
exchange market (with variance σ i2 and covariance σ12),4 we have:
S i = mi + (1 − pi )ε i
(1)
In case of a free-floating regime in both countries (p1=p2=0), we have VarS i = σ i2 and
Co var(S1 , S 2 ) = σ 12 . In the case of a peg in one host country (say country 1) and a free float
in the other one (country 2), we have VarS1 = 0 , VarS 2 = σ 22 and Co var(S1 , S 2 ) = 0 . Finally,
in case of intermediate regimes (i.e. either soft pegs on the investor’s currency, or pegs on
third
currencies,
summarized
as
pi ≠ 0, 1),
we
have
VarS i = (1 − pi ) 2 σ i2
and
Co var(S1 , S 2 ) = (1 − p1 )(1 − p 2 )σ 12 . Note that the covariance is necessarily lower than in a
free float if the exchange rate policies consist in soft pegs against the investor’s currency
( 0 ≤ pi ≤ 1 ), whereas it can be higher or negative in case of a peg on a third currency (in the
latter case, pi can be negative or exceed unity, depending on the relationship between the
anchor currency and the investor’s currency).
In addition to labor and capital costs, we include both a fixed cost in local labor (to be
interpreted as a training cost) and transportation costs. The former produces an incentive to
invest only in one country, in order to avoid the replication of fixed costs. Since transportation
costs tend to hinder trade in goods (Brainard, 1993), it could be advocated that transportation
costs boost FDI and local production serving the domestic market. However, in the model,
firms aim at re-exporting their production. Consequently, transportation costs are expected to
have a negative impact on re-exports and thus on FDI. To put it differently, we assume
complementarity between FDI and trade (Eaton and Tamura, 1994), not substitution. Here,
transportation costs will influence the sensitiveness of FDI to costs and to exchange rate
8
uncertainty. More generally, a preferential bilateral trade agreement, allowing products
exported from the host country to enter more freely in the investing country, could be
interpreted in the same way. Such discrimination would reduce transaction costs between the
two countries.
The investing firm chooses country 1 or 2 according to their respective level of wage, capital
price and exchange rate, according to the magnitude of transportation costs from the
subsidiary to the destination market of goods produced, and according to the prospects for
exchange-rate variations. Because it is risk-averse, it also takes exchange-rate uncertainty into
account. And because there are two possible locations, the covariance between the two
exchange rates will play a crucial role in the model.
The nominal profit of the investing firm is, in its own currency:
π = P(Q1 + Q2 ) − W1 ( F + L1 / τ 1 ) S1' − W2 ( F + L2 / τ 2 ) S 2' − r (R1 K 1 S1 / τ 1 + R 2 K 2 S 2 / τ 2 ) (2)
Where P is the price of the goods on the investor’s market, Qi, Li and Ki are output, labor and
capital used by the firm in country i; Wi and Ri are the nominal wage and the capital price in
country i, in local currency. Both Wi and Ri are assumed to be given for the investing firm (no
pressure on the local markets of labor or capital goods). The fixed cost F is assumed identical
in both locations, whereas the transportation cost 1 / τ i is different.5 r is one plus the world
nominal interest rate. It is assumed that firms creating an international set of operations are
able to borrow at home or in host countries at prevailing interest rates in the international
market. Si is the nominal exchange rate of currency i against the investor’s currency in the
first period (a rise in Si means an appreciation of currency i). Finally, S’i denotes the nominal
exchange rate in the second period. As in Cushman (1988), this is the only uncertainty
introduced in the model.
9
The technology is similar in both locations (it is specific to the multinational). We use a very
simple specification where capital and labor are complements:
Qi = Li =
Ki
k
(3)
Where k is the constant capital/labor ratio.
The firm maximizes its utility, which depends positively on expected profit Eπ and negatively
on its variance Varπ, φ being the risk-aversion coefficient:
MaxU = Eπ − φVarπ

Q1 ,Q2
(4)
2.3 Resolution
The two first order conditions lead to the following relationships:6
'


W2 τ 1Cov( S1' , S 2' )  Q2
2 P − (W1 / τ 1 ) ES1 − kr R1 S1 / τ 1

Q
=
τ
−
+ F  − τ 1 F
 1
1
2
'
'
W1
2φ W1 VarS1
VarS1
 τ2





P − (W2 / τ 2 ) ES 2' − kr R2 S 2 / τ 2 W1 τ 2 Cov( S1' , S 2' )  Q1
Q2 = τ 22
 + F  − τ 2 F
−
2
'
'
W2
2φ W2 VarS 2
VarS 2

 τ1

(5)
The production in country 1 depends on costs and uncertainty concerning country 1 (first
term), but also on the outcome in country 2 (through the second term). Three cases can be
characterized.
(
)
1st case: Cov S1' , S 2' = 0 . The productions carried out in the two countries are independent.
Each one depends on the (positive) difference between the price of the good and the expected
unit costs, and on the uncertainty surrounding these costs: a rise in VarS’i reduces Qi, other
things equal.
10
(
)
2nd case: Cov S1' , S 2' > 0 . The productions in the two countries are substitutes. This is because
when one currency appreciates (raising costs in this country), the other one also appreciates.
Hence, diversifying locations is useless: costs are expected to move in the same way in the
two locations. In this case, better conditions in one country lead to a transfer of FDI from the
other country.
(
)
3rd case: Cov S1' , S 2' < 0 . The productions in the two countries are complements: when costs
rise in one country – because its currency appreciates – they generally decrease in the other
one. In this case the firm raises (reduces) its production in both countries at the same time
when costs are reduced (increased) in one of them. Diversifying its locations reduces the
overall risk on its profit.
These three cases are represented in Figure 1, which depicts the impact of reduced costs in
country 1 (through a currency devaluation for instance). In the three cases, the shock increases
output in country 1. But in the independence case, production is unaffected in country 2,
whereas it declines in the substitutability case and rises in the complementary case.
Insert Figure 1
Hence, the relocation of FDI does not necessarily end in a zero-sum game: whenever
exchange rates are negatively correlated, both locations benefit from the reduction of costs in
one country. Conversely, in the substitutability case, a devaluation in country 1 will reduce
FDI to country 2 except if the latter also devaluates its currency or reduces local costs (for
instance through lowering capital prices).
Solving the model leads to:
Qi =
 τ i2

ρτ iτ j
1


C
−
C
i
j  − Fτ i
σ iσ jWiW j
2φ (1 − ρ 2 )  Wi 2σ i2

(6)
11
where ρ denotes the expected correlation between S1' and S 2' , σi is the expected standard

W
deviation of S i' , and C i = P −  ES i' i
τi


S
 − kr Ri i measures the competitiveness of country i.
τi

We find the standard trade-off between fixed costs and transportation costs referred to above:7
if the choice is between exporting and investing abroad, high specific fixed costs (F) reduce
the incentive to produce abroad. However, since we are interested in the choice between two
foreign competing locations devoted to re-export, a rise in fixed costs reduces output in both
locations, but especially in the host country facing low transportation costs with the investing
country ( ∂Qi ∂F = −τ i ). Indeed, for distant locations, fixed costs are largely offset by
transportation costs as a deterrent to foreign inward investment. In contrast, close or adjacent
countries are highly sensitive to changes in fixed costs, transportation costs remaining
negligible in any case.
Turning to the arbitrage between the two locations according to expected costs and to the
uncertainty pattern, the results can be summarized as follows:
- Lower expected costs in country 1 (higher C1) induce more output in that country, especially
if transportation costs are low (τ1 high).
- Lower uncertainty in country 1 (lower VarS’1) increases the sensitivity of output in
country 1 to local costs.
- Lower expected costs in country 2 reduce, increase or leave unchanged output in country 1
depending on whether the two exchange rates are positively, negatively or not correlated
respectively.
- Output in country 1 is more sensitive to country 2 costs if transportation costs are low
between both countries and the investing country (the product τ 1τ 2 is high): the firm will
mainly arbitrate between close locations.
12
3. Econometric Analysis
3.1 Methodology
We test the theoretical model on a panel of 42 developing countries receiving FDI from 17
OECD countries over 1984-1996 (see the list of countries in Appendix 2). The dependent
variable is the logarithm of the stock of FDI received by the emerging country i from the
OECD country k. It is expressed in US dollars at constant world price and elaborated using
the declarations of the OECD reporting countries (we therefore consider mirror data). This
variable is highly unstable and does not exhibit a particular trend (Figure 2). Besides, it should
be noted that, because we work on bilateral data, the cross-section dimension of the sample is
much more important than its time-series dimension.
In order to get an alternative location for each country i, emerging countries are aggregated
into an emerging area indexed j.8
Two equations are estimated. In the first one (Eq. 7), the impact of distance is considered
separately from that of other variables. Still, the theoretical model shows that firms are likely
to arbitrate more between close locations. The second estimation is closer to the theoretical
model since competitiveness, volatility and correlation in exchange rates are weighted by the
proximity of country i (and of the aggregate j in the case of the correlation) to the investing
country k (Eq. 8).
log FDI itk = a1 log C itk + a 2Vol S itk + a3 ρ itk log C kjt + a 4 log DISTi k + a5 OPEN it
(
)
+ OILi a6 log C itk + a 7Vol S itk + a8 log DISTi k + e k + et + u itk
(
)
(7)
log FDI itk = a1τ~i k log C itk + a 2τ~i k Vol S itk + a3 τ~i k τ~ jk ρ itk log C kjt + a 4 log DISTi k
(
+ a5 OPEN it + OILi a 6 log C itk + a 7Vol S itk + a8 log DISTi k
)
(8)
+ e k + et + u itk
13
The definition of the variables and the data sources used are detailed in Appendix 3. The
competitiveness of country i (Ci in the theoretical model) is proxied by the real exchange rate
of i against each investing country k (Citk), calculated with consumer price levels. A rise in
C itk points to a real depreciation of the currency of country i. Consistently, the
competitiveness of the alternative location (Cj in the theoretical model) is the real exchange
rate of the emerging countries aggregate j against each investor k (noted Cjtk).
The uncertainty on the nominal exchange rate is given by the volatility of the quarterly
nominal exchange rate of country i against k, defined as the coefficient of variation of this
exchange rate over the past three years ( Vol S itk ). The correlation between i/k and j/k exchange
rates is computed over the same period ( ρ itk ).9
Transportation costs are proxied by a measure of distance between i and k ( DISTi k ).10 The
distance between alternative locations and the investing country is constant throughout the
time sample.
τ~i k is a relative proximity variable, which ranges from 0 (lowest proximity) to 1 (highest
proximity), and which is calculated using distance data. This variable is meant to capture the
fact that the investing firm is all the more sensitive to a depreciation in the host country, or to
an increase in volatility in the exchange rate, when transportation costs are low. In the same
vein, τ~jk measures the relative proximity of the aggregate j to the investing country k. The
higher the double proximity τ~iτ~j , the higher the impact of interdependence on the host
country. Note that the distance variable is also introduced as a separate variable in Eq. 8,
consistent with the theoretical model, where the impact of fixed costs is weaker for distant
locations.11
14
In addition to the theoretical model referred to above, we introduce an openness variable
(OPENit), which is designed to control for the nature of foreign direct investment: if FDI aims
at re-export, then it should translate into a large openness ratio (the ratio of exports plus
imports to GDP), since entering the small domestic markets of host economies can hardly be
the investor's motivation.
A dummy (OILi) is also introduced to control for the particular behavior of oil-exporting
countries. In these countries, FDI is mostly related to the energy sector, which is itself linked
to the real exchange rate through a Dutch-disease effect: when the energy sector booms, the
real exchange rate tends to appreciate, but at the same time FDI is attracted because its
profitability in this sector is increased.12 Hence, we should expect a positive link between real
appreciation and FDI, in contradiction with the theoretical model, which mainly describes
manufacturing FDI. Along the same line, the effect of exchange-rate volatility is likely to be
important in oil-exporting countries, since the law of one price applies to oil exports: a large
volatility in the nominal exchange rate means a large uncertainty on local operating costs and
thus on profits. Lastly, transportation costs are expected to have little role in the location
strategy since the location of primary commodities is determined by the nature. This contrasts
with the aim of our model, which is directed at footloose industries.
Finally, we add fixed effects for time (et) and for investing countries (ek). The latter control
inter alia for the size of the investing country, whereas the former catch trends in the world
economy such as deregulation. The size of the host country is partially captured by the
openness variable, although other host characteristics not accounted for by our theoretical
model such as human capital are not included in the estimation.13
We cannot a priori rule out the possibility that the real exchange rate is influenced by inward
FDI, although causality tests can hardly be implemented with annual data. One can argue that
inward FDI induces aggregate supply to rise in the host country, which acts against an
15
appreciation in the exchange rate. Empirically, Artus (1999) shows that there is little
correlation between the average FDI net flows (as a percentage of GDP) and the average
growth of the real exchange rate, for a set of 18 developing countries over the 1992-1996
period. Finally, there is no reason why bilateral exchange rates should react to bilateral FDI:
since forex markets are typically small in emerging host countries and large in OECD investor
countries, a bilateral capital inflow will typically impact on the exchange rate of the host
against all other currencies rather than against the sole investor’s currency, leaving exchange
rates between key currencies unchanged. For all these reasons, the reverse causality problem
should not be very damaging in our estimations, although it cannot be properly tested for.
3.2. Results
The results of the econometric estimations are reported in Table I. In both equations, all the
coefficients are significant at the 1% level, except that on exchange rate volatility in the oilexporting countries, which is significant at the 5% level in Eq. (7).
Insert Table I
Let us first consider Eq. (7). As expected, a depreciation (rise in the real exchange rate) of i
against the investing country increases inward FDI (competitiveness effect), whereas an
increase in the nominal exchange rate volatility tends to reduce FDI (volatility effect). The
coefficient associated to the multiplicative variable ρ itk C kjt (interdependence effect thereafter)
also bears the negative, expected sign: when the exchange rate of other emerging countries14
is positively correlated to that of country i, improved competitiveness in other emerging
countries reduces FDI to country i (through a substitution effect); conversely, in the case of a
negative correlation, an improved competitiveness in other emerging countries raises FDI to
country i (through a diversification effect).
16
When the proximity between the investing country and the potential recipient countries is
controlled for, through the use of the weighting coefficient τ (Eq. 8), the conclusions of the
model are further validated. First, the signs of the estimated coefficients are not affected by
the correction for distance; their interpretation is the same as above. Second, the
competitiveness and interdependence effects are not highly affected, but the significance of
the estimated coefficients is increased. Third, and more importantly, the volatility effect is
magnified since the estimated coefficient is more than four times higher. This means that the
impact of volatility is more important when host countries are close to potential investors,
suggesting that for distant countries, the costs associated to distance are so large that they
override the disutility stemming from exchange-rate risk.
Whatever the estimated equation, geographic distance by itself accounts for a significant part
of the investing behavior of industrialized countries, consistent with the literature on
economic geography. In addition, openness has a significant impact on FDI, which confirms
that the re-export orientation accounts for a significant proportion of FDI in these countries.
According to Eq. (7), the effect of a real appreciation is opposite in oil-producing countries
compared to other emerging countries, since it raises inward FDI. This is consistent with the
Dutch-disease effect referred to above. The impact of volatility is larger than for other
countries, whereas the coefficient on economic distance is almost zero (0.352-0.310=0.042).
The latter outcome can be explained primarily by the fact that oil-exporting countries are all
far from potential investors (mainly the United States). In addition, as already noted, the
location of primary commodities is determined by geological factors, and location strategies
are consequently highly constrained.
4. Policy implications
The policy implications of our results are threefold.
17
First, our estimations highlight the trade-off to be made between price competitiveness and
nominal exchange rate stability when designing an exchange-rate strategy oriented towards
FDI attraction. More precisely, Eq. (7) shows that a 1% appreciation in the real exchange rate
reduces the FDI stock by 0.22%, whereas a 1-point increase in exchange rate volatility
reduces it by 0.60%. The trade-off is more acute for close locations: Eq. (8) shows that when
a host country is very close (or even adjacent) to a potential investor ( τ~i k = 1 ), a 1%
appreciation in the real exchange rate still reduces inward FDI by 0.21%; but now, a 1-point
increase in exchange rate volatility reduces it by 2.25%.
Such trade-off arises because emerging countries typically suffer from a positive inflation
differential with investing countries. Hence, preventing the real exchange rate from
appreciating means allowing the nominal exchange rate to depreciate periodically, which
induces some volatility. Our analysis shows that, even if free floating prevents the real
exchange rate from appreciating (which should not be taken for granted), the induced
volatility has a sizeable impact on FDI which compares with the impact of price
competitiveness, whether or not the proximity is controlled for. This is not to say that
currency boards should be adopted everywhere. But in evaluating exchange-rate regimes, one
should account for the detrimental impact of exchange rate volatility, an impact that is
magnified by proximity.
A second implication of our estimates concerns the choice of a specific anchor: since potential
anchor currencies fluctuate to a large extent between themselves, it is not possible to monitor
price competitiveness and reduce exchange-rate volatility against all of them at the same time.
Hence a second trade-off must be considered across possible anchors. One possibility would
be to peg the local currency to a depreciating currency. This would produce both some
stability against the anchor and an effective depreciation against third currencies. There is
some evidence that Asian emerging countries had such a strategy in the past (Takagi, 1996),
18
and we therefore can interpret the failure of the yen to become an anchor currency as a
consequence of the “ever rising yen” (McKinnon and Ohno, 1997). However, pegging to a
depreciating anchor would entail switching from one anchor to another over time, which is
not possible in a pre-announced regime. In addition, the benefits in terms of price stability
would be reduced. The alternative solution consists in stabilizing the local currency against
the currency of the main investor country, or against a basket of foreign currencies, which
would minimize the volatility of FDI. For those countries which receive FDI from one major
area (the USA for Latin America, the EU for the CEECs and some Mediterranean countries),
this would mean a polarization of exchange-rate regimes consistent with economic geography
(given that the impact of exchange rate variables increases with proximity), hence a step
towards monetary regionalism. For countries experiencing large intra-regional flows of FDI
(essentially in Asia), basket pegs could be more appealing before a genuine regional monetary
framework can be implemented.
Finally, our analysis shows that the choice of a monetary anchor is made more complex by the
fact that recipient countries must take the behavior of other emerging countries into account.
Namely, it is important for a recipient country to differentiate its exchange-rate regime from
those of other emerging countries in order to benefit from the diversification of investments
from risk-averse multinational firms. This last feature is captured by the negative coefficient
on the correlation variable. Because all real exchange rates are taken as an index (on a 100
basis for the United States), the order of magnitude of their decimal logarithm is 2. Hence,
according to Eq. 7, a drop in the correlation between i/k and j/k (k being the investor’s
currency) from 1 to zero increases FDI to country i by 8.8%, which is equivalent to a 40%
depreciation (in country i) or a 15-point drop in volatility. However weighting exchange-rate
variables by geographic proximity (Eq. 8) once again highlights the prominent role of the
bilateral exchange-rate volatility against the investing country, this time compared to
19
interdependence effects: for a country either very close or adjacent to an investor, the same
drop in the correlation of the exchange rate with that of other emerging countries would be
fully compensated by a 4.3-point increase in volatility.
In brief, host countries which are close to one investing country, like Central and Eastern
European or Mediterranean countries (close to the EU) or Latin American countries (close to
the US), and which want to attract FDI, face a high incentive to stabilize their currencies
against that of this specific investor. This is consistent with the building of currency blocks
around the main FDI providers. However, such strategies raise competition among the host
countries of a given block, because FDI from the anchor country becomes more sensitive to
the relative competitiveness in the various locations of the block.
Conversely, in host countries which are relatively far from all potential investors, exchangerate regimes have little impact on FDI. However, these countries produce an externality on
countries which are relatively close to one investor in the sense that the latter would benefit
from seeing the former adopt differentiated exchange-rate policies. Hence, in terms of the
international monetary architecture, it can be concluded that (soft) pegs against one key
currency should be limited to those countries which are relatively close to the corresponding
anchor country.
5. Conclusion
The financial crises of 1997-1998 have given birth to a debate over the reform of the so-called
international financial architecture. Exchange-rate regimes in emerging countries undoubtedly
constitute one pillar of this new architecture.
Our contribution can be viewed as a demonstration that exchange-rate volatility does matter
for foreign direct investment, and hence for a stable financing of growth in emerging
countries, especially for those countries which are close to one main investing country. In
20
addition, we show that exchange-rate regimes in emerging countries should be defined in a
global framework given the externalities they encompass. More precisely, our analysis shows
that monetary regionalism can be a way of increasing FDI to emerging countries as a whole,
although it would likely increase competition within each region. The frontiers of monetary
areas would then be strongly influenced by geography, as FDI is.
Accordingly, we point out two complementary patterns of the future financial architecture: in
the framework of the regionalization of the world economy, the building of currency blocks is
shown here to be beneficial as far as inward FDI and the related benefits of it for emerging
countries (stability, technological progress) are concerned.
21
Appendix 1: Nominal wage sluggishness in some emerging countries
Insert Figure 3
22
Appendix 2: List of the countries included in the sample
17 investing countries
Austria, Australia, Canada, Denmark, Finland, France, Germany, Iceland, Italy, Japan,
Netherlands, New Zealand, Norway, Switzerland, Sweden, United Kingdom, United States.
42 recipient countries
Algeria, Argentina, Brazil, Bulgaria, Chile, China, Colombia, Costa Rica, Czech Republic,
Egypt, Greece, Hong Kong, Hungary, India, Indonesia, Iran, Ireland, Israel, Kuwait, Libya,
Malaysia, Morocco, Mexico, Netherlands Antilles, Panama, Philippines, Poland, Portugal,
Romania, Russia, Saudi Arabia, Singapore, Slovenia, Slovakia, South Africa, South Korea,
Taiwan, Thailand, Turkey, Ukraine, United Arab Emirates, Venezuela.
23
Appendix 3: The regression variables
1 – Inward FDI stocks are tabulated using mirror data statistics, i.e. outward FDI from OECD
countries (source: OECD). Real FDI stocks are constructed as follows, with the world
consumer price index coming from IMF, International Financial Statistics, line 64 (100 in
1991):
 FDI stock in millions USD 

log FDI = log 
 World consumer price index 
2 – The competitiveness indicator C ith is the real exchange rate of each host country i against
each investing country k (units of consumer goods in i needed to buy a unit of consumer
goods in k) in level (hence, real exchange rates can be compared across countries, and not just
across time). It is taken from the CHELEM data-base (CEPII).
3 – The volatility VolS itk is the coefficient of variation of the quarterly nominal exchange rate
of country i against country k, during the three years preceding year t. Source: IMF,
International Financial Statistics, line rf.
VolS itk = σ S k S itk , with σ the standard deviation operator, and S itk the mean of the nominal
it
exchange rate of country i against country k over the three years preceding year t.
4 – The coefficient of correlation between i/k and the rest of emerging countries/k exchange
rates is calculated over the three years preceding year t.
The nominal exchange rate of the group of emerging countries against k is
42
S kjt =
∑ GDP
mt
m =1
k
⋅ S mt
42
∑ GDP
m =1
mt
24
ρ itk = corr (S itk , S kjt ), S itk being the nominal exchange rate of country i against country k, and
corr being the correlation operator, used over the three years preceding t.
Source: IMF, International Financial Statistics, line rf for nominal exchange rates and line 99
for GDPs.
5 – The competitiveness of emerging countries against k is computed using the CHELEM
data-base on real exchange rates. It is a real, effective, annual exchange rate, where each
emerging country exchange rate is weighted by the GDP of this country in the total GDP of
emerging countries. GDPs come from IMF, International Financial Statistics, line 99. Real
exchange rates in levels come from the CEPII-CHELEM data base.
42
C kjt =
∑ GDP
mt
m =1
⋅ C mtk
42
∑ GDP
m =1
mt
6 – Distance is constructed as follows : DISTi k is the distance between i and k when i and k
have no common border. DISTi k is set to one when i and k share a common border. This data
set was provided by M. Pajot.
42
The distance of the host aggregate is defined as DIST =
k
j
∑ GDP
1995
m
m =1
⋅ DISTm
42
∑ GDP
m =1
.
1995
m
7 – The proximity variable τ~i k is built as follows:
k
k
τ~i k = τ ik τ iMAX
, with τ ik = 1 DISTi k and τ iMAX
standing for the highest τ ik between k and the
host countries of the sample. Hence, τ~i k depends on both i and k, and it ranges from 0 (large
distance) to 1 (high proximity). Using τ~i k hence tends to underweight the impact of
competitiveness or volatility when recipient countries are far from investing countries.
25
The proximity of the host aggregate is τ~jk = τ kj τ kMAX
, with τ kj = 1 DIST jk , and τ kMAX
j
j
standing for the highest τ kj between the aggregate j and investor countries k. τ~jk depends on k
and ranges from 0 (large distance) to 1 (maximum proximity).
8 – Openness. Openness is the ratio of the sum of total exports and imports of country i, to
the GDP of country i (source : IMF, International Financial Statistics for the GDP, Direction
of Trade for exports and imports).
OPEN it = 100 ⋅ (X it + M it ) GDPit , where Xit and Mit are exports and imports, and GDPit is the
GDP, all measures in current dollars each year t.
26
References
Aizenmann, J., (1992). “Exchange Rate Flexibility, Volatility and the Patterns of Domestic
and Foreign Direct Investment”, NBER Working Paper, 3853.
Artus, P., (1999). “The Role of Foreign Direct Investment in Financing Growth and in
Stabilizing the Exchange Rate”, CDC Working paper, September.
Barrel, R., and Pain, N. (1998). “Real Exchange Rates, Agglomeration and Irreversibilities:
Macroeconomic Policy and FDI in EMU”, Oxford Review of Economic Policy, 14, 152-167.
Borenzstein, E., and De Gregorio, J. (1995). “How Does Foreign Direct Investment Affect
Economic Growth”, NBER Working Paper, N°5057.
Brainard, L. S., (1993). “An Empirical Assessment of the Proximity-Concentration Trade off
Between Multinational Sales and Trade”, NBER Working Paper, 4580.
Cooper, R., (1999). “A Tour of International Financial Reform”, interview in Challenge, 42,
5-28.
Corden, W. M., and Neary, J. P. (1982). “Booming Sector and De-Industrialisation in a Small
Open Economy”, The Economic Journal, 92, 825-48.
Cushman, D. O., (1988). “Exchange-Rate Uncertainty and Foreign Direct Investment in the
United States”, Weltwirtschaftliches Archiv, 124, 322-336.
Darby J., Hughes Hallett, A. H., Ireland, J., and Piscitelli, L. (1999). “The Impact of Exchange
Rate Uncertainty on the Level of Investment”, The Economic Journal, 105, C55-C67.
Dixit, A., and R. Pindyck, (1994). Investment under Uncertainty, Princeton University Press.
Goldberg, L. S., and Kolstad, C. D. (1994). “Foreign Direct Investment, Exchange Rate
Variability and Demand Uncertainty”, NBER Working Paper, 4815
27
Eichengreen, B., (1999). Toward a New International Financial Architecture: A Practical
Post-Asia Agenda, Washington, D.C.: Institute for International Economics.
Eaton, J., and Tamura, A. (1994). “Bilateralism and regionalism in Japanese and US trade and
direct foreign investment patterns”, NBER working paper 4758.
Fontagné, L., and Freudenberg, M. (1999). “Endogenous Symmetry of Shocks in a Monetary
Union”, Open Economies Review, 10, 263-287.
Frankel, J. A., (1999). “No Single Currency Regime is Right for All Countries or at All
Times”, NBER Working Paper, 7338.
Froot, K. A., and Stein, J. C. (1991). “Exchange Rates and Foreign Direct Investment: an
Imperfect Capital Markets Approach”, Quarterly Journal of Economics, 106, 1191-1217.
Goldberg, L. S., and Klein, M. W. (1997). “Foreign Direct Investment, Trade and Real
Exchange Rate Linkages in Southeast Asia and Latin America”, NBER working paper, 6344.
Horstman, I. J., and Markusen, J. (1992). “Endogenous Market Structures in International
Trade”, Journal of International Economics, 32, 109-129.
Ito, T., Isard, P., Symansky, S., and Bayoumi, T. (1996). “Exchange Rate Movements and
Their Impact on Trade and Investment in the APEC Region”, chapter IV, IMF Occasional
Paper No 145, December.
Klein, W. M., and Rosengren, E. S. (1994). “The Real Exchange Rate and Foreign Direct
Investment in the United States: Relative Wealth vs Relative Wage Effects”, Journal of
International Economics, 36, 373-389.
Lipsey, R. E., (1999). “The Role of Foreign Direct Investment in International Capital
Flows”, NBER Working Paper 7094, April.
28
McKinnon, R., and Ohno, K. (1997). Dollar and Yen: Resolving Economic Conflict Between
the United States and Japan, MIT Press.
Markusen, J., (1995). “The Boundaries of Multinational Enterprises and the Theory of
International Trade”, Journal of Economic Perspectives, 9, 169-189.
Ricci, L. A., (1997). “Exchange-rate regimes and Location”, IMF working paper, WP-97/69.
Stiglitz, J., (1999). “Reforming the Global Economic Architecture: Lessons from Recent
Crises”, The Journal of Finance, 54, 1508-1521.
Swenson, D., (1999). “Foreign Investment and International Trade Linkages”, mimeograph,
University of California Davis.
Takagi, S., (1996). “The Yen and its East Asian Neighbors, 1980-95: Cooperation or
Competition?”, NBER working paper, 5720.
29
Footnotes
1. In fact, the option reduces the risk related to the flow of production, but not that related to
the capital value.
2. Available data do not disentangle greenfield investments from mergers and acquisitions.
Our approach matches better the former since mergers and acquisitions are less
complementary to trade (Swenson, 1999).
3. Due to restrictions on profit repatriation in some countries, the cost of re-investing profits
varies across countries. Relaxing the simplifying assumption of a common interest rate
does not affect our results.
4. For the sake of simplicity, we neglect the possibility that the exchange rate follows a
random walk or is submitted to a constant drift. The same reasoning would apply
however, with a crawling peg instead of a fixed peg and a variance calculated on logvariations instead of on levels.
5. We assume standard iceberg transportation costs for tractability. This transportation cost
is proportional to output: it does not affect the fixed cost which can be interpreted as a
training cost. Hence, the fixed cost is more important compared to variable costs when the
host country is close to the investing country.
6. If the variance is zero (in the case of a currency board for instance), the firm maximises its
expected profit. The first order condition then leads to P = (W1ES '1 + rR1kS1) / τ1 , which is a
form of purchasing power parity.
7. Fixed costs enter the first order conditions due to the calculation of profit variance.
8. For simplicity, each emerging country is assumed to be small in the aggregate, which
therefore always includes the 42 countries of the sample. Weights are the shares of each
host country in the GDP of the 42 countries emerging area. See Appendix 3.
30
9. The use of past volatility and correlation can be justified by the auto-regressive pattern of
the volatility: except for institutional breaks like the settlement of a currency board,
looking at past volatility helps to forecast future volatility.
10. We are grateful to Michaël Pajot for kindly providing this measure.
11. Dropping this variable does not change the results (sign, level and significance of
estimated coefficients) in Eq. (8), the main consequence being a slight reduction in the
adjusted R².
12. See Corden and Neary (1982).
13. In order to keep enough degrees of freedom and to avoid multicolinearity with the
distance variable, a choice must be made between fixed effects on investors or on host
countries. Since we are interested in explaining inward FDI, we must control for factors
which are unrelated to host countries, such as the size of investing countries. Given that
we have 17 investing countries and 42 host countries, the loss in degrees of freedom is
smaller with fixed effects on investing countries rather than host countries. Incidentally,
estimations run with host country fixed effects basically lead to the same results, although
the interdependence effect loses significance and the explanatory power of the regression
is lowered.
14. Against the currency of country k, not written here for the sake of readability.
15. S. Dees kindly provided Chinese nominal exchange rates.
31
Table I. Econometric resultsa
Eq. (7)
Eq. (8)
Nb. obs.
1,757
1,757
log C itk
0.222
-
[0.009]
-0.597
VolS itk
-
[0.002]
ρ itk ⋅ log C kjt
-0.044
-
[0.011]
τ~i k log C itk
-
0.206
[0.000]
τ~i k VolS itk
-
-2.248
[0.000]
(τ~ τ~ )⋅ ρ
k
i
k
j
k
it
⋅ log C kjt
-
-0.048
[0.002]
-0.310
-0.161
[0.000]
[0.019]
0.891
0.878
[0.000]
[0.000]
-0.454
-0.333
[0.000]
[0.004]
-0.742
-0.983
[0.049]
[0.005]
0.352
0.281
[0.001]
[0.000]
θ
0.004
0.004
Hausman test
χ²(17)
χ²(19)
27.984
28.698
[0.045]
[0.071]
0.604
0.607
log DISTi k
OPENit
OILi ⋅ log C itk
OILi ⋅ VolS itk
OILi ⋅ log DISTi k
R²
a
Dependent variable : logarithm of the real inward FDI stock in country i from country k. p-
values in parentheses.
32
FIGURE 1 : the relationship between production in the two locations, Impact of a reduction
in country 1 costs.
1a: independence
Q2
Q 1 (Q 2 )
Q 2 * =Q 2 *’
E
E’
Q 1*
Q 2 (Q 1 )
Q 1 *’
Q1
1b: substitutability
Q2
*’
Q2
Q1(Q2)
E
E’
*’
Q2
*
Q2(Q1)
*’
Q1
Q1
Q1
1c: complementarity
Q2
Q1(Q2)
*
Q2
E’
*’
Q2
Q2(Q1)
E
*
Q1
*’
Q1
Q1
33
4000
3500
3000
2500
2000
1500
1000
500
CEECs
Latin America
MENA countries
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
0
1984
Million USD x 100, at 1991 world price
FIGURE 2 the real FDI stock in various emerging areas
Asia
Notes: MENA = Middle East and North African.
CEECs= Central and Eastern European Countries.
Years 1996 to 1998 were removed from the graph due to missing data in some
countries.
Source: OECD.
34
FIGURE 3: exchange rate pass-through
Growth rate (in % per year) of hourly earnings in manufacturing and of nominal exchange
rates (against USD) in selected emerging countries.
Greece
0,25
0,20
0,15
0,10
0,05
0,00
1985
1987
1989
1991
1993
1995
1997
-0,05
-0,10
Hourly Earnings in Manufacturing
Nominal exchange rate
35
Portugal
0,20
0,15
0,10
0,05
0,00
1985
1987
1989
1991
1993
1995
1997
-0,05
-0,10
Hourly Earnings in Manufacturing
Nominal exchange rate
-0,15
Mexico
0,90
Hourly Earnings in Manufacturing
Nominal exchange rate
0,80
0,70
0,60
0,50
0,40
0,30
0,20
0,10
0,00
1985
1987
1989
1991
1993
1995
1997
36
South Korea
0,50
Hourly Earnings in Manufacturing
Nominal exchange rate
0,40
0,30
0,20
0,10
0,00
1985
1987
1989
1991
1993
1995
1997
-0,10
-0,20
The Czech Republic
0,20
0,15
0,10
0,05
0,00
1994
1995
1996
1997
1998
-0,05
-0,10
Hourly Earnings in Manufacturing
Nominal exchange rate
37
Poland
2,50
Hourly Earnings in Manufacturing
Nominal exchange rate
2,00
1,50
1,00
0,50
0,00
1985
1987
1989
1991
1993
1995
1997
Source: OECD.
38