Content uploaded by Nadine McCloud
Author content
All content in this area was uploaded by Nadine McCloud on Apr 19, 2021
Content may be subject to copyright.
Foreign Capital Inflows, Exchange Rates, and Government Stability
Nadine McCloud∗
Department of Economics
University of the West Indies at Mona
Michael S. Delgado†
Department of Agricultural Economics
Purdue University
Man Jin‡
Department of Economics
Oakland University
April 2, 2021
Abstract
In theory, changes in a host country exchange rate can be a cause or consequence of changes in
its level of foreign direct investment (FDI), and recent incidences suggest that government stability
may have sizable implications for the interactions between FDI and the exchange rate. This paper
uses a semiparametric system of simultaneous equations to empirically characterize the relationship
between FDI and the exchange rate, with each country’s level of government stability serving as a
moderator. The results suggest that across developed and developing economies the most prevalent
type of symbiosis between FDI and the exchange rate is a positive effect of FDI on the exchange
rate, but no effect of the exchange rate on FDI. This significant FDI effect is heterogeneous, with an
interquartile range of 1.241. At the median, a 10 percent increase in FDI inflows relative to GDP
causes approximately a 13.29 percent increase in the annual change in the exchange rate. Government
stability acts as a moderator variable by strengthening the relationship between FDI and the exchange
rate in some countries, but eliminates the relationship in other countries.
Keywords: Foreign direct investment; exchange rate; government stability; symbiosis; parameter het-
erogeneity; semiparametric system of equations model.
JEL Codes: C14; C26; E02; F21; F31; O19.
∗Nadine McCloud, Department of Economics, University of the West Indies at Mona, Kingston 7, Jamaica. Phone:
876-970-6282, Fax: 876-977-1483. Email:nadine.mccloud02@uwimona.edu.jm
†Michael S. Delgado, Department of Agricultural Economics, Purdue University, West Lafayette, IN 47907-2056. Phone:
765-494-4211, Fax: 765-494-9176, Email: delgado2@purdue.edu.
‡Man Jin, Department of Economics, Oakland University, Rochester, MI 48306. Phone: 248-370-4086. Email:
mjin@oakland.edu
Bryan Black and Rosan Reynolds-Salmon provided research assistance. We are grateful to participants at the 26th Silvaplana
Workshop in Political Economy 2017 for comments. A previous draft of this paper was titled “A mutualism analysis of capital
inflows and exchange rates: Does government stability matter?”.
1
1 Introduction
A recent spate of incidences in government instability across the globe appear to have had implications for
the interactions between exchange rates and FDI flows. For example, in early 2014, reports from different
sources indicated that Ukraine was on the brink of an economic disaster and that the currency plunged
amid political turmoil between Russia and the European Union. Between 2013 and 2014, estimates from
the World Bank database suggests that Ukraine’s local currency per US dollar depreciated by 48.7%, while
its FDI net inflows (as a % of GDP) decreased by 1.82 percentage points. In August 2018, according
to Bloomberg, the combination of political turmoil, hyperinflation, and years of disastrous polices in
Venezuela led to one of the greatest currency devaluations in history – a 95% depreciation of the Bol´ıvar
(Venezuelan dollar). Meanwhile, the World Bank database shows that FDI net inflows plunged from 2.956
billion dollars in 2015 to -68 million dollars in 2017.
Conversely, Angola represents a valuable case of a resource-rich country that experienced a protracted
government instability period and recovered well. After gaining political independence from Portugal in
1975, Angola erupted into a civil war that lasted for 27 years (1975-2002). Not surprisingly, the Angolan
Kwanza (AOA) was highly unstable during the civil war years. The currency became effectively worthless,
and the country introduced a new currency in the late 1990s. In the year 1999, the United States Dollar
($1.00) was worth AOA 2.80, and as of 2013, it was approximately US$1.00 to AOA 97.00. However, in the
first decade after the civil war, the exchange rate was relatively stable compared to the rapid depreciation
between 1999 and 2003. Angola became a major FDI player in its post-war period. UNCTAD’s data
records that Angola’s inward stock of FDI almost quadrupled from 2002 to 2011. According to UNCTAD’s
(2017) World Investment Report, Angola boasts other feats such as being the largest recipient of FDI
inflows in Africa in 2016, with a value of 14.4 billion dollars, and in the top 20 host economy in the world
of FDI inflows in 2015 and 2016. So, Angola experienced political instability for an extended period.
Subsequently, it moved towards a more stable political environment and achieved rapid growth in FDI
inflows and a more stable currency.1
Do these incidences reflect an empirical regularity that government stability matters for the inter-
action between a host country’s exchange rate and FDI inflows? This paper investigates empirically
the interaction between FDI inflows and the exchange rate as well as the effects of variation in gov-
ernment stability across host countries in this relationship. Conceptually, there are six theoretical
foundations of the FDI-exchange rate relationship – the traded and non-traded goods model (Bruce
& Purvis 1984), the portfolio model (Branson 1977), the monetary approach to balance of payments
model (Ffrench-Davis 1983, Corbo 1985), the strategic behavior of international firms model (Goldberg
& Kolstad 1995, Sung & Lapan 2000), the imperfect capital markets theory (Froot & Stein 1991, Klein
& Rosengren 1994), and the relative labor cost theory (Cushman 1985, Cushman 1987, Culem 1988).
These theories have been developed to explain different countries’ experiences with FDI flows and ex-
change rates, possibly stemming from differences between industrialized or non-industrialized nations or
whether a country maintains a fixed or floating exchange rate; we provide a more detailed summary of
these foundational theories in Section 2. And though few empirical papers examine the effect of FDI
flows on exchange rates, the literature is replete with empirical studies showing that the exchange rate
1Based on World Bank data, between 2003 and 2013, Angola’s economy grew at an annual average of about 10.3% relative
to all of Sub-Saharan Africa which grew by an average of approximately 5.7%.
2
is a determinant of FDI inflows.2Furthermore, the impact of government stability on the relationship
between the exchange rate and FDI inflows has not been analysed. Although anecdotal evidence strongly
suggests that government stability can impinge on the symbiotic relationship between a host country’s
exchange rate and FDI inflows (if such a relationship exists), it need not be the case that the magnitude
and role of government stability is likely to be homogeneous across countries.
We maintain that government instability can undermine the ability of national authorities to, in
accordance with domestic goals, efficiently and effectively use flows from FDI to impact changes in the
country’s exchange rate and vice versa. Angola, once again, is an appropriate illustration: according to
the 2004 IMF country report for Angola (IMF 2005), “... Directors commended the authorities for the
improvements in macroeconomic management since the peace agreement.” Additionally, it is challenging
for FDI inflows in some host countries to redound to their benefit because there is a huge disparity in
bargaining power between their governments and sender countries’ investors. This disparity in bargaining
power may be magnified in unstable host countries. Aleksynska & Havrylchyk (2013) lend credence to
this scenario in their discussion on Chinese firms’ choice of overseas investment location.
Government instability can also increase the costs borne by foreign investors. These costs need not
translate into lower or no FDI inflows in an unstable host country, as large corporations with high potential
tax revenues may parlay their local political capital to circumvent frictions in unstable host countries.
In fact, drawing on UNCTAD’s (2007) report on transnational corporations, Aleksynska & Havrylchyk
(2013) highlight that firms from the South are largely state-owned and are therefore less sensitive to poor
institutions in host countries than large private multinationals from the North; the presence of Chinese,
Indian and Malaysian firms in Sudan provides a fitting example of such phenomenon. In addition, foreign
investors may shift their preference from wholly-owned firms to joint ventures with local investors in an
attempt to mitigate the added risk induced by government instability. However, in the absence of such
risk-mitigation mechanisms, an increase in government instability may increase (or decrease) the strength
of the symbiotic effect of changes in the exchange rate on FDI inflows.
To undertake our empirical analysis, we adopt the semiparametric panel model of a system of simul-
taneous equations put forward by McCloud, Delgado & Kumbhakar (2018) – hereafter MDK – which
allows FDI and the exchange rate to be modeled jointly and uses instrumental variables to identify causal
effects. The semiparametric nature of the MDK model stems from the assumption that the coefficients
on all regressors in all equations are unknown smooth functions of a moderating index – here, government
stability – and unobserved country- and year-specific factors (fixed effects). The joint modelling of FDI
inflows and exchange rates allows us to discern whether the data are most consistent with some countries
experiencing, for example, (i) exchange-rate-FDI mutualism – a positive effect of FDI on the exchange
rate and a positive effect of the exchange rate on FDI – or (ii) FDI-commensalism – a positive effect
of FDI on the exchange rate but no effect of the exchange rate on FDI.3Our modelling framework can
be viewed as an empirical extension of the theoretical setting in Russ (2007) who argues that FDI and
exchanges rates are jointly determined by underlying macroeconomic factors, and consequently exchange
2See, e.g., Froot & Stein (1991), Broll & Wahl (1992), Blonigen (1997), Cushman (1985), Campa (1993), Aizenman (1992),
Goldberg & Kolstad (1994), Campa & Goldberg (1995), Kiyota & Urata (2004), Russ (2007), Udomkerdmongkol, Morrissey
& Gorg (2009), Yu & Walsh (2010), Alba, Wang & Park (2010).
3Here, we avail ourselves of the symbiosis taxanomy in the biological literature to characterize the interactions between
FDI inflows and changes in exchange rates.
3
rates are endogenous in FDI models. Russ (2007) develops a general equilibrium framework that links
demand and the bilateral exchange rate to common underlying fundamental variables, thus capturing the
multinational firm’s reaction to the net effect that macroeconomic shocks have on both the exchange rate
and sales overseas.
We fit the MDK apparatus to a panel dataset of 115 developed and developing countries over the
period 1984 to 2010. The index of government stability comes from the International Country Risk
Guide published by Political Risk Services, and is defined as “the government’s ability to carry out its
declared program(s), and its ability to stay in office”. The index is the sum of three subcomponents –
Government Unity, Legislative Strength and Popular Support – each with a maximum score of 4 points
and a minimum score of 0 points; a score of 4 points equates to very low risk and a score of 0 points to
very high risk. Consequently, this government stability index ranges from 0 to 12, with 0 representing
low government stability and 12 representing high government stability. It is important to emphasize
that a government must perform well in all three categories in order to have an overall high index value
of government stability. Or, a government that performs well in two categories, but poorly in the third,
will not be measured to have a high level of government stability. As we show in the data section of this
paper, developed countries are not guaranteed a high score in this government stability index.
We see several important empirical conclusions regarding the types of symbiotic interactions that exist
between FDI and the exchange rate, and the indirect effects of government stability on such interactions.
We observe that across developed and developing economies, causal, heterogeneous mutualism and FDI-
commensalism are the most dominant types of interaction between FDI and the exchange rate, indicating
that FDI generally has a significantly positive effect on exchange rates but the exchange rate has either
a significantly positive or insignificant effect on FDI inflows in most countries. We further find that
government stability is an important source of heterogeneity in the FDI-exchange rate relationship, with
approximately 8 points being an important threshold level of government stability. In light of Russ (2007),
our empirical findings suggest that simultaneous effects of underlying government stability on demand
and the exchange rate are important, and can dominate those effects from underlying macroeconomic
variables. Our contribution complements the broader literature on political-economy determinants of
FDI, and of exchange rates, their regimes, and policies.4Our modelling framework, however, accounts
for both direct and indirect effects of institutions on FDI and changes in exchange rates.
We begin in Section 2 with a review of literature. Section 3 introduces the MDK semiparametric
system of simultaneous equations model, which we use to examine the relationship between exchange
rate, FDI and government stability. We provide our empirical system model and a description of our
data, including the instrumental variables, in Section 4. We present our empirical results in Section 5,
and conclusions in Section 6. Further technical details regarding the MDK model are placed in a technical
appendix.
4Regarding the political economy determinants of FDI, see for example Schneider & Frey (1985), Hines (1995), Wei (2000),
Harms & Ursprung (2002), Egger & Winner (2005), Egger & Winner (2006), Hakkala, Norb¨ack & Svaleryd (2008), Kolstad
& Villanger (2008), Asiedu, Jin & Nandwa (2009), Faeth (2009), Javorcik & Wei (2009), Aleksynska & Havrylchyk (2013),
S´anchez-Mart´ın, de Arce & Escribano (2014), Akhtaruzzaman, Berg & Hajzler (2017) and the references cited therein; and
for exchange rates, see Bodea (2015), Beckmann, Ademmer, Belke & Schweickert (2017), Beckmann & Czudaj (2017) and
the references cited therein.
4
2 Literature Review
2.1 The Relationship between FDI and the Exchange Rate
As noted, the theoretical literature is undecided as to the nature of the relationship between FDI and
the exchange rate. There are at least six competing theoretical frameworks that seek to explain the FDI-
exchange rate relationship: the traded and non-traded goods model; the portfolio model; the monetary
approach to the balance of payments model; the strategic behavior of international firms model; the
imperfect capital markets theory; and the relative labor cost theory.
The traded and non-traded goods model maintains that an exogenous inflow of capital will lead a
country’s real exchange rate to either appreciate or depreciate, depending on whether the capital inflows
are used to finance domestic spending or capital accumulation (Bruce & Purvis 1984). If capital inflows
are used to finance domestic consumption, then FDI raises spending power and demand for both traded
and non-traded goods, leading to a real exchange rate appreciation. If, instead, capital inflows finance
capital accumulation in the non-traded sector in order to increase productivity, the real exchange rate
will depreciate as the prices of non-tradable goods fall. Importantly, this model more generally applies
to small countries that are price takers in the world market with normally low exchange rate volatility;
this model also only works with FDI inflows, not net FDI inflows. The model was particularly popular
among economists studying southern European countries in the late 1980s as characterized by large capital
inflows and real exchange rate appreciation.
The portfolio model suggests that an excess liquidity of capital due to foreign investment pushes
the domestic price level up, thereby increasing inflation that in turn leads to an appreciation in the real
exchange rate (Branson 1977). The model is based on a background of financial and capital liberalization,
and generally applies to countries characterized by low exchange rate volatility or to markets with fixed
exchange rate regimes. Empirical effort using this framework includes Harberger (1985), who investigates
massive capital inflows and exchange rate volatility in Chile.
The monetary approach to the balance of payments model, the strategic behavior of international
firms model, the imperfect capital markets theory, and the relative labor cost theory, focus on the effect
of the exchange rate on FDI, generally describing countries characterized by relatively high exchange
rate volatility or with freely floating currencies. The monetary approach to the balance of payments
framework holds that a real exchange rate appreciation associated with a current account deficit leads
to foreign capital inflows, assuming that the exchange rate is the main instrument of monetary policy,
the international capital market is liquid, and there are no barriers to capital inflows. The strategic
behavior of international firms model suggests that a domestic exchange rate appreciation results in a
large current account deficit. Under the fear of follow-up protectionism, such as an increase in tariffs,
risk-averse international firms abroad are more likely to invest in these markets with capital instead of
goods. In this case, an appreciation in the real exchange rate causes an increase in net FDI inflows.
Different from the these two theories, the imperfect capital markets model and the relative labor cost
model predict that a depreciation in the real exchange rate causes an increase in net FDI inflows. Froot &
Stein (1991) connect the exchange rate with FDI in the context of capital market imperfections that cause
external financing to be more expensive than internal financing, such that changes in wealth translate
into changes in the demand for direct investment. By developing a model characterized by capital market
5
imperfections, they find that a depreciation in the domestic currency (that causes the wealth of firms to
fall) increases firms’ demand for foreign capital. The relative labor cost theory focuses on the value of the
currency during the floating exchange rate period among industrialized countries, such as US, Germany,
or the UK. A depreciation of the domestic exchange rate is associated with relatively cheaper labor, in
turn attracting more FDI.
Efforts have been made to empirically test the above theories and provide supporting evidence. Con-
sistent with the theory of traded and non-traded goods, Giavazzi & Spaventa (1990) find that large foreign
capital inflows have led to the real exchange rate appreciation in the southern European countries; and
consistent with the portfolio model, evidence shows that FDI inflows cause a real exchange rate apprecia-
tion in financially liberalized countries, such as Spain (De Grauwe, Danthine, Katseli & Thygesen 1991),
Chile (Harberger 1985) and Argentina (Edwards 1985). In terms of the exchange rate effect on FDI, Corbo
(1985) finds that real exchange rate appreciation explains the huge foreign capital inflows into Chile during
the 1977-1982 period, which is consistent with the monetary approach. In line with the strategic behavior
model, Kogut & Chang (1996) find that Japanese multinational firms switch from support FDI outflows
to inflows following an appreciation in the domestic real exchange rate. Finally, Klein & Rosengren (1994)
provide empirical support of the imperfect capital markets model and Cushman (1985) provide evidence
of the relative labor cost model, both of which suggest a domestic exchange rate depreciation leads to
an increase in FDI among industrialized countries. This empirical backdrop provides a footing for us to
explore these different theoretical links more broadly across a large sample of developed and developing
countries, with government stability as a moderator.
2.2 Government Stability and Symbiotic Effects
Most researchers have focused on assessing the relationship between the exchange rate and FDI unidirec-
tionally (in one direction or the other). Empirical work focusing on the bidirectional relationship between
FDI and the exchange rate is limited, and most papers are either descriptive or are limited by taking a
regional focus.5It is clear from the FDI-exchange rate literature that the relationship between the two is
heterogeneous and complex, depending on a variety of factors, and in the introduction we have provided
ample anecdotal evidence that government stability may serve as a moderating feature in this relation-
ship. Additionally, our intuition for considering government stability as a factor governing the effect of
FDI on the exchange rate can be motivated by the portfolio model and several theoretical and empirical
monetary studies. According to Aisen & Veiga (2006), government stability is generally associated with
price stability. In a more politically-stable economy, the domestic aggregate price level in the short-run
tends to slowly adjust (or not adjust) in response to macroeconomic shocks. This in turn strengthens any
effect of an FDI shock on real household balances and therefore increases the scope of such a shock in
terms of the development of real effects on either domestic consumption or the real exchange rate. This
hypothesis also resonates with several studies in the field of international economics; for example, Hau
5For instance, Kosteletou & Liargovas (2000) empirically examine causality between the exchange rate and FDI for 12
EU countries using a simultaneous linear equation model where FDI and the exchange rate are jointly determined. They
find that in large countries with freely floating currencies causality runs from the exchange rate to FDI, but that causality
runs both ways in small countries with fixed or “quasi” fixed currencies. While we also focus on bidirectional links between
the exchange rate and FDI, in contrast to Kosteletou & Liargovas (2000), we do not restrict our attention to either a limited
sample of countries or to a linear regression setup.
6
(2002) finds that the effect of trade openness on exchange rate volatility depends on a country’s degree
of government stability.
When it comes to FDI, a recent perception is that FDI is not only driven by macroeconomic stability
(which includes inflation and the exchange rate), but also government stability in host countries. Empirical
evidence on the effect of government stability on FDI has been mixed: for example, using cross-country
data, Busse & Hefeker (2007), Asiedu (2006) and Wei (2000) find that government stability is a strong,
positive determinant of FDI, though in a study of foreign investment in U.S. firms, Wheeler & Mody
(1992) fail to find a significant link between these two factors. Therefore, it is important to account
for government stability in our study of FDI, and more importantly, to treat it as a moderator via
the coefficient functions rather than an additional regressor. Such a specification might provide a new
explanation to these mixed findings of FDI.
3 The Semiparametric System of Simultaneous Equations Modelling
Framework
We use the very general semiparametric simultaneous system of equations model put forward by MDK
to extract empirical evidence on the types of symbiosis between exchange rate and FDI, and the extent
to which government stability moderates this relationship. MDK develop a novel class of semiparametric
estimators suited for obtaining consistent estimates from different formulations of the semiparametric
simultaneous system of equations model. Details on the MDK estimators are included in the technical
appendix. In what follows, we use the term vector to mean a column vector, unless otherwise stated.
To begin, consider in general form a bivariate semiparametric system of simultaneous equations
y1,it
y2,it
=
=
Y0
−1,itλ1(Z1,it ) + X0
1,itγ1(Z1,it ) + 1,it
Y0
−2,itλ2(Z2,it ) + X0
2,itγ2(Z2,it ) + 2,it
(3.1)
for i= 1, . . . , N , and t= 1, . . . , T. In equation j= 1,2,for cross-sectional unit iin time period t,yj,it is
a scalar response variable, Y−j,it is a pj-dimensional vector of endogenous variables that includes at least
one yj1,it with j16=j; hence, the presence of Yj,it in each equation renders the system non-triangular.
In addition, Xj,it is a kj-dimensional vector of exogenous variables in which the first entry is equal to 1,
Zj,it ∈Rdjis a vector of exogenous variables, λj(·) and γj(·) are unknown Borel measurable functions of
conformable dimensions, and j,it is the idiosyncratic error term. Notice that, for the general derivation,
we assume that the elements of Zj,it are continuously distributed; in practice, this assumption is easily
relaxed to accommodate mixed categorical and continuous data using the important tools developed by
Racine & Li (2004).
The maintained conditional moment assumptions for model 3.1 are that
E[it|Zit ]=0, E[it|e
Xit]6= 0 and E[j,itk,it|Z0
it,e
X0
it]6= 0,(3.2)
where it = (1,it, 2,it )0,Zit = (Z0
1,it, Z 0
2,it)0, and e
Xit = ( e
X0
1,it,e
X0
2,it)0.
Our main interest is in the set of unknown coefficient functions {λj(·)}, which clearly captures the
types of interactions between the pairs yjand yj1with j16=j. To characterize all interactions between
7
any pair yjand yj1with j16=j, we therefore follow MDK and borrow the ensuing taxonomy from the
biological literature:
Definition 3.1. Let lj∈ {1,2}and λj(·) = {λj,lj(·) : Rdj→R, lj6=j}. Assume that for cross-sectional
unit iin time period tthe effect of yjand yj1with j16=jcan be positive,negative or zero, and vice versa.
Between the pair of variables (yj,it, yj1,it ) we say there exists:
(a) mutualism if λj,j1(·), λj1,j (·)>0;
(b) yj1,it-commensalism if λj,j1(·)>0 and λj1,j (·) = 0;
(c) synnercrosis if λj,j1(·), λj1,j (·)<0;
(d) yj1,it-antagonistic symbiosis if λj,j1(·)>0 and λj1,j (·)<0;
(e) yj,it-ammensalism if λj,j1(·)<0 and λj1,j (·) = 0;
(f) non-symbiosis if λj,j1(·) = λj1,j (·) = 0.
As mentioned in the preamble of this paper, plausible theoretical predictions are that the effect of
FDI on the exchange rate and the effect of the exchange rate on FDI can be positive,negative, or zero.
Moreover, within a country there can be mutualism between FDI and exchange rate in one time period, but
exchange rate-commensalism in another time period as a result of, say, certain country-specific policies.
Thus, this general taxonomy seems quite fitting for characterizing all possible theoretical interactions
between FDI and exchange rate.
4 Empirical Model and the Data
4.1 An Empirical Simultaneous Model of Exchange Rate and FDI
We let i= 1,2, . . . , N denote country index, and t= 1,2, . . . , T denote the time period. Our empirical
bivariate semiparametric system of equations model allows for the change in exchange rate, ∆EXit, and
FDI inflows, F DIit, to be modeled simultaneously in the following way:
∆EXit = ∆EXi,t−1λ0,1(Zit ) + F DIit λ1(Zit) + X0
1,itγ1(Zit ) + 1,it (4.1)
F DIit = ∆EXitλ2(Zit ) + X0
2,itγ2(Zit ) + 2,it.(4.2)
We use ∆EXit rather than the level of the exchange rate, EXit, as foreign investors are more concerned
with fluctuations in the value of the host-country currency than with its level. In our empirical model
(equations (4.1) and (4.2)), Xj,it is a kj-dimensioned vector of control variables for equations j= 1,2,
such that the first entry in the vector is equal to one; Xj,it may share common elements across j.γj(·)
and λj(·) are unknown smooth coefficient functions of conformable dimensions. We presume that Zit is a
d-dimensioned vector that may include a mix of continuous and discrete regressors (Racine & Li 2004, Li
& Racine 2010); this vector includes our moderator as well as country- and year-specific categorical
variables. We assume that Zit is constant across both equations and across each of the mjcoefficient
functions. That is, we maintain the hypothesis of the same sources of parameter heterogeneities in the
8
exchange rate and FDI equations. As in our general model in Section 3, the errors j,it are assumed
to be mean zero disturbances that are correlated across equations, and all other model assumptions are
assumed to be satisfied.
4.2 Data Overview
We garner most of our data from the 2012 World Development Indicators (WDI) database published by
the World Bank. Our sample is an unbalanced panel of 115 developed and developing countries over the
period 1984-2010. We average the data into a 3-year panel. A 3-year averaged panel, in lieu of an annual
panel, should capture relationships between FDI and the exchange rate over longer horizons, and using
time-averaged rather than contemporaneous panels reduces the serial correlation within the data, thereby
smoothing the effects of outliers induced by business cycles and other annual fluctuations on our results.
Averaging the panel over 3-year periods provides a sample size of 727 observations. The effective sample
size used for each regression, however, is further reduced because of data limitations for different sets of
control variables. Our use of a wild bootstrap to compute the standard errors mitigates, among other
things, the lack of precision of the estimates that is usually a by-product of small sample estimation.
4.3 Explanatory Variables in the Exchange Rate Equation
The dependent variable in this equation is the 3-year average of the annual change in the exchange rate,
where the exchange rate is the domestic currency per US Dollar. The main explanatory variable, FDI
inflows, is defined as the net inflows of FDI as a percentage of GDP. Our supplementary explanatory
variables control for additional factors that are correlated with both FDI and the exchange rate, and in so
doing mitigate against omitted variable bias in the estimated effects of FDI. We consequently draw on the
literature on exchange rate determination, which is well established. Specifically, we consider openness
to trade and GDP growth as control variables (see, e.g., Branson 1981, Hacche 1983, Pearce 1983).
Trade openness or high GDP growth could mean higher exports, and by extension larger foreign currency
earnings, which in turn diminish the chance of an exchange rate depreciation. This aligns with the
traditional balance of payments approach (see Branson 1981). We also include the one-period lag of the
change in the exchange rate to capture inertia. These variables constitute the explanatory variables in the
benchmark exchange rate equation, where FDI inflows affect the exchange rate directly. To echo other
theories in the literature (see Section 2) where FDI inflows affects the exchange rate indirectly, we add
additional regressors in the two alternative exchange rate equations. Specifically, in light of the traded
and non-traded goods model, we include the CPI in one alternative model, where FDI inflows affects the
exchange rate through consumer prices. As this theory applies primarily to the case of a small, price-
taking country, we include GDP as an additional control when empirically assessing this model. In light
of the portfolio model, we add the net foreign investment (FDI inflows minus outflows) and the inflation
rate as additional regressors. With them, we are able to test whether FDI inflows affects changes in the
exchange rate indirectly through the channel of foreign investment and inflation. See Table 1 for the
description of the regressors.
9
Table 1: List of data and definitions.
Variable Definition
Outcomes
Exchange Rate Exchange rate (domestic currency per US$)
FDI Net FDI inflows as a percentage of GDP in constant 2002 dollars.
This series shows net inflows (new investment inflows less disin-
vestment) in the reporting economy from foreign investors, and is
divided by GDP.
Controls in Exchange
Rate Equation
Lag Exchange Rate Exchange rate in the previous period
Trade Openness Ratio of exports plus imports as % of GDP
Growth Rate Growth rate of real GDP per capita
CPI Consumer price index
GDP Logarithm of GDP (constant 2000 US$)
Net Foreign Investment Net FDI inflows minus net FDI outflows (% of GDP)
Inflation Rate Inflation, consumer prices (annual %)
Controls in FDI Equation
Lag FDI Net FDI inflows as a % of GDP in constant 2002 dollars, in the
previous period
Trade Openness Ratio of exports plus imports as a % of GDP
Growth Rate Growth rate of real GDP per capita
Inflation Rate Inflation, consumer prices (annual %)
Interest Rate Interest rate of lending
Current Account Current account balance (% of GDP)
OECD Dummy, one for OECD countries and zero otherwise
Market Capitalization Market capitalization of listed domestic companies (% of GDP)
Coefficient Variable
Government Stability 12 point index of government stability which is defined as “ the
government’s ability to carry out its declared program(s), and its
ability to stay in office”. The index is the sum of three subcompo-
nents - Government Unity, Legislative Strength and Popular Sup-
port - each with a maximum score of 4 points and a minimum score
of 0 points; a score of 4 points equates to very low risk and a score
of 0 points to very high risk
Note: All variables, except government stability, come from the WDI 2012. Government stability
comes from the International Country Risk Guide database.
4.4 Explanatory Variables in the FDI Equation
Various theories on the correlates of FDI have been developed and empirically analysed; see, for example,
Blonigen (2005), Russ (2007), Faeth (2009), Blonigen & Piger (2014), S´anchez-Mart´ın et al. (2014) for
in-depth reviews. Foreign investors are guided by, among other things, profitability, complementarity with
investors’ global strategy, and host country environment. Given the aggregate nature of our hypothesis,
10
we focus on the latter of these factors. Of the variables which have been associated with FDI inflows,
the exchange rate is our primary variable of interest. For the control set, we choose trade openness and
the economic growth rate as well. Trade openness likely has a direct impact on FDI flows as a country
that is more integrated into the world economy through trade is more likely to be an attractive place for
FDI. Economic growth, of course, is generally regarded as having a positive impact on FDI inflows (see
Borensztein, Gregorio & Lee 1998), as a faster growing economy is more attractive to an investor than
one that grows more slowly.
The exchange rate, trade openness, and the growth rate of real GDP make up the set of benchmark
regressors in the FDI model, where the exchange rate has a direct effect on FDI inflows. Meanwhile,
we set up three alternative FDI equations with additional regressors to examine indirect effects. First,
in light of the monetary approach to the balance of payments theory, we include the inflation rate and
the interest rate as additional regressors. When the exchange rate is the main monetary instrument of
stabilization, an exchange rate appreciation can lead to a higher domestic inflation rate and a higher
interest rate. Widening gaps between domestic inflation or interest, relative to worldwide levels, could
increase the demand of foreign borrowing or financing and therefore increase FDI inflows. By including
the inflation rate and the interest rate as additional regressors in the FDI model, we are able to test the
above mechanism wherein the exchange rate has an indirect influence on FDI. Second, in light of the
theory of strategic behavior of international firms, we consider the current account as a means through
which the exchange rate affects FDI inflows. Third, in light of the theories of imperfect capital markets
and relative labor cost, we consider the indirect effect of the exchange rate on FDI through the channel
of firms’ wealth by including the market capitalization of listed firms. The theories particularly apply for
the industrialized countries, thus we include a dummy variable that equals one if a country is a member
of the OECD and zero otherwise.
4.5 Coefficient Variables
Recall that the coefficient variables, Zit, are introduced as sources of heterogeneity in the effects of all the
explanatory variables on the outcome variable. Our maintained assumption is that government stability
and unobservable country and time effects are important factors in the relationship between the exchange
rate and FDI inflows. The index of government stability comes from the International Country Risk
Guide published by Political Risk Services, and is defined as “the government’s ability to carry out its
declared program(s), and its ability to stay in office”. The index is the sum of three subcomponents –
Government Unity, Legislative Strength and Popular Support – each with a maximum score of 4 points
and a minimum score of 0 points; a score of 4 points equates to very low risk and a score of 0 points to very
high risk.6Consequently, this government stability index ranges from 0 to 12, with 0 representing low
government stability and 12 representing high government stability. See Table 8 for a granular description
of countries in the government-stability distribution.
Our model (Equations 4.1 and 4.2) also allows for government stability to exert direct influences
on both FDI and the exchange rate via the intercept coefficient functions. Indeed, empirical evidence
suggests that changes in government stability are directly associated with changes in the foreign exchange
6This government stability index is available for 1984 and onwards, which inevitably fixes the beginning year of our
dataset.
11
or currency market (see, e.g., Cosset & Rianderie 1985, Fielding & Shortland 2005). Empirical evidence
of a direct association between government instability and FDI inflows can be inferred from, for example,
Schneider & Frey (1985), Asiedu (2006), Busse & Hefeker (2007), and S´anchez-Mart´ın et al. (2014).7
We use an unordered categorical country variable and ordered categorical year variable to control for
unobserved country- and time-specific heterogeneity in all the coefficient functions. Thus, in contrast to
the orthodox case of a homogeneous coefficient fixed effects panel specification, we do not assume that
the country and year effects have a neutral effect on the model by influencing the intercept only.8That
is, we control for country and time-varying effects – i.e., fixed effects – in a non-neutral fashion. Our
country and year indicators are therefore capable of capturing any country and time-varying factors that
induce heterogeneity in the intercept and slope coefficients across countries and time.9Consequently,
innumerable time-invariant measures of country-specific government policies are subsumed in our fixed
effects in Zit.10
5 Empirical Results
We now turn to the results from applying our semiparametric instrumental variables systems estimator to
our novel empirical model of the exchange rate and FDI in (4.1) and (4.2).11 We note that our underlying
sample excludes the United States. We first discuss our estimated coefficient functions for our bivariate
exchange rate-FDI system, and then analyze the marginal effects on these coefficients from an increase in
government stability. Since our semiparametric systems estimator provides observation-specific estimates
and standard errors, we summarize these estimates using kernel density plots and 45 degree gradient plots
that are depicted by Figure 1, and the 25th, 50th (median), and 75th percentiles that are in Table 2. The
45 degree gradient plots found in the lower panels of the figures show the observation specific function
estimates plotted on the 45 degree line, with 95 percent observation specific confidence intervals plotted
above and below each point estimate. If the horizontal dotted line at zero lies outside of each observation
specific confidence interval, then that point estimate is statistically significant.
5.1 Benchmark Model Estimates
We first present the coefficient function estimates from the benchmark model, where the control sets
for equations (4.1) and (4.2) are X0
1,it := (Openness0
it, Growth0
it) and X0
2,it := (Openness0
it, Growth0
it),
respectively.
7Barro (1991) and Brunetti & Weder (1998) find empirical evidence that government instability is significantly related to
cross-country differences in private investment.
8Moreover, the Racine & Li (2004) kernels allow for interaction of unknown form between both fixed effects and variables
that vary both across country and over time, so that our control of country- and year-specific effects is not restricted to
additively separable effects.
9With exception of the country and year indicators, we require the variables in Zit to vary across both iand t, thus
allowing our country and year indicators to absorb all country or time-varying factors that may lead to heterogeneity in the
coefficient functions.
10Our use of non-neutral fixed effects – in lieu of their neutral counterparts - circumvents the need to remove the fixed
effects via some type of weighting or first difference transformation prior to estimation, to avoid biased and inconsistent
estimates of, in particular, the marginal effects. Further, our use of generalized product kernels (Racine & Li 2004) allows
us to avoid the incidental parameters problem associated with many parametric panel models that include dummy variables
to account for unobserved effects.
11We consider estimates of (A.9) that use first stage estimates of Aas our weighting matrix.
12
Exchange Rate Equation The top-left panel of Figure 1 presents the kernel density of the FDI
coefficient estimates. It is clear from the kernel density that the effect of FDI on the exchange rate is
heterogeneous, though largely positive. The corresponding 45 degree gradient plot (bottom-left of Figure
1) shows that the majority of these estimates are statistically significant at the 5 percent level; this is
evident from the clustering of estimates in the first quadrant with point-wise confidence bounds that
exclude zero.
In Table 2, we find that the effect of FDI on the exchange rate (see the exchange rate equation) are
unanimously positive across the 25th, 50th and 75th percentiles. Evidently, the effect is heterogeneous,
which has an interquartile range of 1.241, with a median effect of 1.329. Our point estimates at the
quartiles are statistically significant at all three reported quartiles suggesting that for most countries,
FDI has a significantly positive, heterogeneous, and causal effect on the exchange rate. At the median, a
10 percent increase in FDI inflows relative to GDP causes approximately a 13.29 percent increase in the
annual change in the exchange rate.
Table 2 also reports summaries of the estimates of the other coefficients in the exchange rate equation.
The reported quartiles for the lagged exchange rate coefficient are all significantly positive, and satisfy the
stability condition for the exchange rate equation (a coefficient estimate less than one in absolute value).
Trade openness has a negative and significant effect at the lower quartile, and economic growth has a
negative and significant effect on the exchange rate at the lower and median quartiles. Thus, an increase
in trade openness reduces the annual change in the exchange rate in a few countries, and an increase in
the economic growth rate reduces the annual change in the exchange rate in many countries. For trade
openness and growth, there is a clear absence of statistical parity among the corresponding quartiles of
estimates. These results therefore complement the existing exchange rate literature by providing added
evidence of sizable heterogeneities in the effects of each of these variables on the annual change in the
exchange rate across countries.
Table 2: Summary of system NPGMM coefficient estimates: benchmark model
∆Exchange Model FDI Model
Q1 Q2 Q3 Q1 Q2 Q3
FDI 0.534∗∗∗ 1.329∗∗∗ 1.775∗∗∗ ∆Exchange 0.002∗∗ 0.003∗∗∗ 0.006∗∗∗
0.213 0.328 0.529 0.001 0.002 0.002
Lag ∆Exchange 0.112∗∗ 0.513∗∗∗ 0.803∗∗∗ Growth 0.131∗∗∗ 0.277∗∗∗ 0.391∗∗∗
0.050 0.076 0.144 0.049 0.088 0.141
Openness −0.081∗∗ −0.001 0.105 Openness −0.003 0.020∗∗∗ 0.030∗∗∗
0.041 0.066 0.085 0.005 0.008 0.011
Growth −4.900∗∗∗ −1.864∗∗ −0.196
0.494 1.020 2.067
Intercept 0.016 0.124∗∗ 0.242∗∗ Intercept −0.008∗∗ −0.001 0.013
0.033 0.068 0.117 0.004 0.007 0.010
Note: Q1, Q2, Q3 refer to percentiles in the distribution of coefficient estimates. Estimate specific standard
errors are obtained via wild bootstrap and are reported below each estimate. The level of statistical significance
of estimates are denoted by ∗p < 0.1;∗∗ p < 0.05;∗∗∗ p < 0.01. The dependent variables are first difference of the
exchange rate and FDI inflows. See Table 1 for definitions of other variables.
13
Figure 1: Coefficient estimates-benchmark model.
FDI Equation The top-right panel of Figure 1 reveals that the distribution of the exchange rate
coefficient estimates is generally positive, and the corresponding 45 degree plot (bottom right of Figure
1) reveals that most of these positive estimates are statistically significant. In Table 2, we see that the
change in the exchange rate has a statistically significant causal effect on FDI inflows at each reported
quartile. These coefficients have an interquartile range of 0.004 with a median effect of 0.003. These
estimates imply that at the median, a 10 percent increase in ∆Exchange causes about a 0.03 percent
increase in the ratio of FDI inflows to GDP. Overall, we find strong evidence that for most countries the
change in the exchange rate has a significantly positive, heterogeneous, and causal effect on FDI. This
empirical evidence of heterogeneity in the effect of exchange rate on FDI represents a salient distinction
between our work and the existing FDI literature. Considering the effects of other control variables in the
FDI equation, we see from Table 2 that economic growth has a statistically significant positive effect at all
quartiles and with a large interquartile range. Trade openness has a positive and significant effect on FDI
inflows at the median and upper quartiles, but an insignificant effect at the lower quartile. These findings
mark important empirical results for research investigating factors related to macroeconomic drivers of
FDI inflows. In particular, the nature of the heterogeneous effects in the FDI equation partially explains
why the previous results from linear models were unable to unearth statistically robust correlates of FDI
inflows.
14
Government Stability In all, we interpret these significant effects in both equations to be robust
evidence that our model is flexible enough to significantly capture heterogeneity across countries that
might arise because of nonlinearities from interactions of these variables with respect to government
stability or unobserved country and time effects. Looking at the kernel density plots in Figure 2, it
is clear that the FDI-exchange rate interaction is stronger for OECD countries as the distribution of
FDI inflows coefficients (or DExchange coefficients) is to the right of that for the non-OECD countries.
Testing for density equality via the Li, Maasoumi & Racine (2009) nonparametric test confirms our visual
inspection: we reject the null hypothesis of density equality between the OECD and non-OECD sets
for the FDI inflows and DExchange coefficients at the 1% significance level. Our data shows that, on
average, OECD countries have better government stability as the index average is 8.10 compared with
the 7.79 for the non-OECD countries, though the average difference is not particularly stark. Figure 2
indirectly suggests an increase in government stability may strengthen the effect of FDI on the change in
the exchange rate and vice versa.
Figure 2: Kernel densities of the estimated FDI inflows coefficients and exchange rate coefficients across
OECD and non-OECD country delineations.
To examine the direct impact of government stability on the FDI-exchange rate interaction, we use
nonparametric methods to regress the government stability index on the estimated FDI inflows coeffi-
cients and the estimated exchange rate coefficients. We use nonparametric methods so as to avoid any
specification concerns; given that these are univariate regressions, we can plot the estimated regression
functions and confidence bands. We present plots of these estimated regression functions in Figure 3.
From the left side of the panel, we see a positive relationship between government stability and the effect
of FDI inflows on the exchange rate (i.e., the estimated FDI inflows coefficients), particularly at higher
levels of government stability (roughly above 6 on the 0-12 scale). This suggests that for many coun-
tries, particularly those with more stable governments, an increase in government stability strengthens
the effect of FDI on the exchange rate. The finding is consistent with our hypothesis that government
stability can enhance the ability of national authorities to efficiently and effectively use FDI inflows to
15
impact changes in the exchange rate. From the right side of the panel of Figure 3, we find there is a non-
monotonic relationship between government stability and the exchange rate coefficients. For relatively
unstable countries with a stability index value less than 8, an increase in government stability is generally
associated with a weakening of the impact of the exchange rate on FDI inflows. This finding is expected,
as an increase in government stability may undermine foreign firms’ political capital and therefore re-
duce the sensitivity of FDI to the exchange rate. For countries that have relatively stable governments
(stability index value greater than 8), an increase in government stability is generally associated with a
strengthening of the impact of the exchange rate on FDI inflows. For those countries that already have
relatively strong institutions, a further increase in government stability reduces political/market risk and
therefore increases the effect of a change in the exchange rate on FDI inflows. Overall, our results from
Figure 3 suggest that government stability is an important source of parameter heterogeneity in both the
exchange rate and FDI equations.
Figure 3: Plots from nonparametric regressions of estimated FDI and exchange rate coefficients on the
index of government stability.
5.2 Alternative Theories: Building on the Benchmark Model
We empirically investigate the salience of each of the six theories described previously via systematically
augmenting the benchmark model to incorporate the tenets of each theory. An important advantage of
our nonparametric approach is that we can directly assess the relevance of each particular theory for
each country specifically; that is, the heterogeneity garnered from our empirical approach allows us to
determine which theories are salient for different countries or subsets of countries, rather than the all-or-
nothing assumption explicit in traditional models that assume parameter homogeneity. This extension not
only allows us to relate the conventional theory about FDI-exchange rate interactions, but also to enhance
16
our understandings about the mechanism of FDI’s influence on the exchange rate and the exchange rate’s
influence on FDI.
The Traded and Non-Traded Goods Theory The first two theories, the traded and non-traded
goods and portfolio theories, suggest that capital inflows lead to changes in the exchange rate. In light of
the traded and non-traded goods theory, we include CPI and GDP as additional regressors in the exchange
rate equation. Thus, the control sets of equation (4.1) is X0
1,it := (Openness0
it, Growth0
it, CP I 0
it, GDP 0
it).
Equation (4.2) is unchanged that X0
2,it := (Openness0
it, Growth0
it) because the theory focuses on the
effect of FDI on the exchange rate. Figure 4 shows the density and 45 degree plots for the FDI and
exchange rate coefficient estimates, and Table 3 reports the estimates of coefficients in the 25th, 50th
and 75th percentiles. We see in the figure that for both FDI inflows (the exchange rate equation) and
exchange rate coefficients (the FDI equation), there is a large proportion of estimates that are positive
and statistically significant. These estimates appear fairly similar to the estimates from the baseline
model. Table 3 shows that these similarities are generally borne out, though the FDI coefficient in the
exchange rate equation is no longer significant at the 25th percentile. Further, the coefficients at the
median and upper quartiles, while still significant, have smaller magnitudes compared to those in Table
2. Clearly the significance of CPI and GDP, particularly at the lower quartiles, have accounted for part
of the significance of FDI on changes in the exchange rate as found in the baseline model, lending some
credence to the traded and non-traded goods theory.
Table 3: Summary of system NPGMM coefficient estimates: traded and non-traded goods theory
∆Exchange Model FDI Model
Q1 Q2 Q3 Q1 Q2 Q3
FDI −0.027 0.562∗∗ 1.052∗∗ ∆Exchange 0.002∗∗ 0.003∗∗ 0.006∗∗∗
0.042 0.291 0.547 0.001 0.002 0.002
Lag ∆Exchange 0.086∗∗ 0.480∗∗∗ 0.801∗∗∗ Growth 0.131∗∗∗ 0.276∗∗∗ 0.390∗∗∗
0.037 0.063 0.121 0.048 0.088 0.140
Openness −0.073∗∗ 0.061 0.219∗∗ Openness −0.003 0.020∗∗∗ 0.030∗∗∗
0.031 0.065 0.129 0.005 0.008 0.011
Growth −5.518∗∗∗ −2.331∗∗∗ −0.110
0.042 0.866 1.609
CPI −0.011∗∗∗ −0.002∗∗ 0.000
0.001 0.001 0.004
GDP −0.037∗∗∗ −0.011 0.057∗
0.007 0.015 0.036
Intercept −0.669∗∗∗ 0.463 1.179∗Intercept −0.008∗∗ −0.001 0.013
0.043 0.369 0.795 0.004 0.007 0.010
Note: Q1, Q2, Q3 refer to percentiles in the distribution of coefficient estimates. Estimate specific standard
errors are obtained via wild bootstrap and are reported below each estimate. The level of statistical significance
of estimates are denoted by ∗p < 0.1;∗∗ p < 0.05;∗∗∗ p < 0.01. The dependent variables are first difference of
exchange rate and FDI inflows. See Table 1 for definitions of other variables.
The Portfolio Theory In line with the portfolio theory, we include net foreign investment and inflation
as additional regressors in the exchange rate equation. Figure 5 and Table 4 report estimation summaries.
The distributions of coefficient estimates are still generally positive, though we can see from the 45 degree
17
Figure 4: Coefficient estimates for M1: traded and non-traded goods theory
plot of the FDI inflows coefficient estimates that there is no longer a large and significant cluster of positive
coefficients. As expected, we still see the positive and significant clustering of estimates in the FDI equation
for the exchange rate coefficient estimates. Table 4 shows that at the 25th, 50th, and 75th percentiles,
the FDI inflows coefficient estimates are no longer significant and instead net foreign investment (at the
25th percentile) and the inflation rate (at all reported percentiles) are significant. These results bear out
the portfolio theory: FDI inflows create excessive liquidity in the domestic economy, which then pushes
up the domestic price level and worsens the current account deficit that leads to a depreciation of the
nominal exchange rate. Another interpretation of our results is that the FDI effect on the exchange rate
identified in the baseline model operates through the inflation rate.
Monetary Approach to the Balance of Payments Theory Next, we link our model with the
monetary approach to the balance of payments by including the inflation and interest rates as additional
regressors in the FDI equation. The control set of equation (4.1) reverts back to the baseline specification
while that of equation (4.2) becomes X0
2,it := (Openness0
it, Growth0
it, I nflation0
it, I nterest0
it). The esti-
mates are presented in Figure 6 and Table 5. At first glance, the kernel densities of the estimates reveal
that a large portion of the estimated coefficients are positive, though the 45 degree plots reveal less sta-
tistical significance of these estimates compared to the baseline model. Table 5 confirms this observation:
18
Figure 5: Coefficient estimates for M2: portfolio theory
the exchange rate becomes largely insignificant in the FDI equation, except at the 75th percentile and
at the 10 percent significance level, while other control variables are largely significant. And, although
there is no change in the exchange rate model specification, the coefficients on FDI are smaller and even
become insignificant at the lower quartile. This finding suggests that there are systematic interactions
between FDI and the exchange rate, which further validates our approach in estimating these models
simultaneously instead of separately. Thus, as suggested by the monetary approach theory, when the
exchange rate is the main monetary instrument of stabilization, a rise in the domestic interest rate leads
to an increase in demand for foreign capital.
The Strategic Behavior Theory Following the strategic behavior theory, we include the current
account as the additional regressor in the FDI model while keeping the exchange rate model unchanged
(from the baseline). The estimation results of the nonparametric system of equations are presented in
Figure 7 and Table 6. The density plots and 45 degree plots reveal a generally positive and significant
set of estimated FDI inflows coefficients in the exchange rate model, and a heterogeneous (in sign and
significance) set of estimates on the exchange rate coefficients in the FDI equation. Looking at the
point estimates in the table, the estimates of the exchange rate coefficients in the FDI model become
insignificant at the lower and median quartiles, and are significant at the 10% level at the upper quartile.
19
Table 4: Summary of system NPGMM coefficient estimates: portfolio theory
∆Exchange Model FDI Model
Q1 Q2 Q3 Q1 Q2 Q3
FDI −0.045 0.511 1.442 ∆Exchange 0.002∗∗ 0.004∗∗∗ 0.006∗∗∗
0.232 0.415 1.141 0.001 0.001 0.002
Lag ∆Exchange 0.295∗∗∗ 0.537∗∗∗ 0.973∗∗∗ Growth 0.139∗∗ 0.267∗∗ 0.407∗∗∗
0.055 0.103 0.221 0.064 0.104 0.162
Openness −0.015 0.056 0.141∗∗ Openness −0.011∗∗ 0.015∗∗ 0.027∗∗
0.023 0.052 0.073 0.005 0.008 0.012
Growth −5.254∗∗∗ −2.312∗∗∗ −0.520
0.217 0.830 1.394
Net Foreign Invest −0.006∗∗ 0.002 0.013
0.003 0.007 0.015
Inflation 0.090∗∗ 1.112∗∗∗ 2.420∗∗∗
0.047 0.243 0.516
Intercept −0.076∗∗∗ 0.005 0.093 Intercept −0.008∗∗ 0.001 0.017∗
0.021 0.045 0.096 0.004 0.007 0.011
Note: Q1, Q2, Q3 refer to percentiles in the distribution of coefficient estimates. Estimate specific standard
errors are obtained via wild bootstrap and are reported below each estimate. The level of statistical significance
of estimates are denoted by ∗p < 0.1;∗∗ p < 0.05;∗∗∗ p < 0.01. The dependent variables are first difference of
exchange rate and FDI inflows. See Table 1 for definitions of other variables.
Table 5: Summary of system NPGMM coefficient estimates: monetary approach to the balance of payment
∆Exchange Model FDI Model
Q1 Q2 Q3 Q1 Q2 Q3
FDI 0.011 0.989∗∗∗ 1.561∗∗ ∆Exchange 0.000 0.002 0.003∗
0.143 0.295 0.631 0.001 0.002 0.002
Lag ∆Exchange 0.028 0.487∗∗∗ 0.800∗∗∗ Growth 0.076∗0.171∗∗ 0.259∗∗∗
0.046 0.080 0.152 0.048 0.080 0.106
Openness −0.104∗∗∗ −0.009 0.075 Openness 0.019∗∗∗ 0.031∗∗∗ 0.042∗∗∗
0.036 0.071 0.091 0.003 0.005 0.007
Growth −4.792∗∗∗ −1.734∗−0.012 Inflation −0.014∗∗∗ −0.002 0.009
0.316 1.055 2.343 0.002 0.004 0.007
Interest 0.016∗∗∗ 0.047∗∗∗ 0.129∗∗∗
0.007 0.012 0.025
Intercept 0.024 0.101∗∗ 0.255∗∗ Intercept −0.030∗∗∗ −0.015∗∗∗ −0.002
0.033 0.058 0.116 0.004 0.006 0.008
Note: Q1, Q2, Q3 refer to percentiles in the distribution of coefficient estimates. Estimate specific standard
errors are obtained via wild bootstrap and are reported below each estimate. The level of statistical significance
of estimates are denoted by ∗p < 0.1;∗∗ p < 0.05;∗∗∗ p < 0.01. The dependent variables are first difference of
exchange rate and FDI inflows. See Table 1 for definitions of other variables.
20
Figure 6: Coefficient estimates for M3: monetary approach to the balance of payments
Meanwhile, the newly-added current account is significant and negative at all presented quartiles. The
finding is consistent with the strategic behavior of multinational firms’ theory in that an appreciation of
the exchange rate at home results in a large current account deficit. Under a fear of bilateral protectionism,
multinational firms increase capital investment abroad, which is less sensitive to trade policy, such as tariff
jumping.
Imperfect Capital Markets Theory Lastly, based on the theory of imperfect capital markets, we
include an OECD dummy and market capitalization as the additional regressors in the FDI model.
Results are reported in Figure 8 and Table 7. Due to the data availability of market capitalization, the
new sample size for this model is only half of the original, baseline sample. In the figure, we can see
generally positive and significant FDI inflows coefficients in the exchange rate model, but heterogeneous
(in sign and significance) estimates of the exchange rate coefficients in the FDI model. Particularly, in the
FDI model, the exchange rate coefficients become insignificant at the median quartile and become negative
in the lower quartile. Meanwhile, the newly-added market capitalization is significantly negative across
all quartiles. The finding is consistent with the proposed theory that a decrease in firms’ value/wealth will
increase the demand of foreign investment. A depreciation of the exchange rate reduces the market value
of domestic firms relative to their foreign counterparts, which stimulates an increased demand of foreign
21
Figure 7: Coefficient estimates for M4: strategic behavior
Table 6: Summary of system NPGMM coefficient estimates: strategic behavior theory
∆Exchange Model FDI Model
Q1 Q2 Q3 Q1 Q2 Q3
FDI 0.030∗∗∗ 0.815∗∗∗ 1.192∗∗∗ ∆Exchange 0.000 0.001 0.003∗
0.001 0.001 0.003 0.001 0.001 0.002
Lag ∆Exchange 0.113∗∗∗ 0.528∗∗∗ 0.791∗∗∗ Growth 0.136∗∗∗ 0.239∗∗∗ 0.351∗∗∗
0.022 0.062 0.089 0.045 0.067 0.102
Openness −0.103∗∗∗ −0.013 0.091∗∗∗ Openness 0.003 0.024∗∗∗ 0.034∗∗∗
0.012 0.020 0.032 0.004 0.007 0.010
Growth −1.175∗∗∗ −0.292∗∗∗ −0.017∗∗∗ Current Account −0.004∗∗∗ −0.004∗∗∗ −0.003∗∗∗
0.001 0.001 0.002 0.000 0.000 0.000
Intercept 0.009 0.085∗∗∗ 0.173∗∗∗ Intercept −0.017∗∗∗ −0.009∗0.006
0.016 0.034 0.056 0.003 0.006 0.009
Note: Q1, Q2, Q3 refer to percentiles in the distribution of coefficient estimates. Estimate specific standard
errors are obtained via wild bootstrap and are reported below each estimate. The level of statistical significance
of estimates are denoted by ∗p < 0.1;∗∗ p < 0.05;∗∗∗ p < 0.01. The dependent variables are first difference of
exchange rate and FDI inflows. See Table 1 for definitions of other variables.
22
investment to offset their losses in the capital market. In addition, the significantly negative coefficients
on the OECD dummy variable indicates that the effect of the exchange rate on FDI via the channel
of capital markets is stronger in non-OECD countries. In this sense, our finding extends the previous
empirical work of Klein & Rosengren (1994) that uses only data on industrial countries to test the theory
of imperfect capital markets.
Figure 8: Coefficient estimates for M5: imperfect capital market theory
23
Table 7: Summary of system NPGMM coefficient estimates: imperfect capital market theory
∆Exchange Model FDI Model
Q1 Q2 Q3 Q1 Q2 Q3
FDI 0.001∗0.005∗∗∗ 0.013∗∗∗ DExchange −0.004∗∗∗ 0.000 0.006∗∗∗
0.000 0.001 0.002 0.000 0.001 0.002
LDExchange 0.019∗∗∗ 0.068∗∗∗ 0.547∗∗∗ Growth 0.128∗∗∗ 0.261∗∗∗ 0.401∗∗∗
0.003 0.011 0.028 0.038 0.055 0.077
Openness −0.029∗∗∗ 0.012 0.049∗∗ Openness −0.019∗∗∗ 0.014∗∗ 0.025∗∗∗
0.005 0.011 0.022 0.004 0.006 0.008
Growth 0.000 0.001∗∗ 0.004∗∗∗ OECD −0.052∗∗∗ −0.031∗∗∗ −0.020∗∗∗
0.000 0.001 0.002 0.004 0.005 0.007
Mkt Cap −0.0003∗∗∗ −0.0002∗∗∗ −0.0002∗∗∗
0.000 0.000 0.000
Intercept −0.008 0.035∗∗ 0.109∗∗∗ Intercept 0.017∗∗∗ 0.029∗∗∗ 0.055∗∗∗
0.008 0.016 0.035 0.004 0.006 0.010
Note: Q1, Q2, Q3 refer to percentiles in the distribution of coefficient estimates. Estimate specific standard
errors are obtained via wild bootstrap and are reported below each estimate. The level of statistical significance
of estimates are denoted by ∗p < 0.1;∗∗ p < 0.05;∗∗∗ p < 0.01. The dependent variables are first difference of
exchange rate and FDI inflows. See Table 1 for definitions of other variables.
24
5.3 An Increase in Government Stability
Figure 9: Index of government stability
We now turn towards taking a deeper look into the estimated FDI-exchange rate relationship by in-
vestigating the extent to which our estimates depend on government stability. Recall that the index of
government stability ranges from 0 to 12, and is made up of three subcomponents: government unity,
legislative strength, and popular support. Looking deeper into the index, we find a maximum of gov-
ernment stability at 12 points coming from China in the year 2000 and Kazakhstan in the year 1999.
For the United States, the minimum index value is 5.6 in the year 1995 and the maximum value is 10.8
in the year 2000. To provide a slightly broader perspective, in Figure 9 we plot the time series of the
government stability index for the entire world (averaged), the United States, the United Kingdom, Mex-
ico and Pakistan jointly. This sampling of countries is ad hoc; we merely select both developed and
developing countries as a means of comparing their indices over time. We see from the figure that these
countries – and the world as an average – follow a similar time path. Perhaps more importantly, we
see that developed nations do not have a particular advantage in government stability over developing
countries. This is divergent from other common measures of institutional strength, e.g., corruption, in
which there is a marked advantage of developed nations over developing nations. Yet, in thinking about
the government stability index components, it is sensible that developed nations must also actively work
to promote improved government stability. For instance, a country might have strong legislative strength
through well guarded legal institutions, but have relatively little popular support. It is also not clear
that democratic governments necessarily have a better chance at a higher level of government stability,
as evidenced by China in 2000.
25
(a) Benchmark model (b) M1: traded and non-traded goods theory
(c) M2: portfolio theory
(d) M3: monetary approach to the balance of pay-
ments
(e) M4: strategic behavior (f) M5: imperfect capital market theory
Figure 10: Nonparametric regressions of government stability on FDI and exchange rate coefficients
26
To investigate the effect of a marginal change in the government stability index on the FDI and
∆Exchange coefficients, we once again use nonparametric methods to regress the FDI and ∆Exchange
rate coefficients on the index of government stability (separately). These plots are shown in Figure 10. It
is clear from the figure that across most of the estimated models, government stability has a significantly
positive relationship with the FDI inflows coefficients, indicating that the effect of FDI inflows on the
exchange rate is stronger at higher levels of government stability. It is also clear that there is a kind of
threshold effect in this relationship given that the relationship between government stability and the FDI
inflows coefficient is not significant at relatively low levels of government stability. In other words, there
seems to be a level of government stability – at approximately 8 points on the 12 point scale – above
which increases in government stability strengthens the FDI inflows effect on changes in the exchange
rates. It is worth noting that the upward trend in the government stability-FDI coefficient relationship
turns downwards in the portfolio theory model, and is downward trending at low levels and flat at high
levels in the imperfect capital market theory model (recall that this last model has a substantially smaller
sample size).
Looking at the effect of government stability on the ∆Exchange rate coefficients, we find a generally V-
shaped relationship, with the only exception being the portfolio theory model. According to the majority
of these models, an increase in government stability weakens the exchange rate effect on FDI when the
existing level of government stability is low, but strengthens this relationship when the existing level
of government stability is above (approximately) 8 points. In the portfolio theory model, government
stability does not significantly influence the effect of the exchange rate on FDI flows at low levels of
stability, but has a significantly positive effect at higher levels (again, above approximately 8 points).
Returning to the empirical reality that all governments – in both developed and developing nations
– appear to struggle in maintaining a relatively high level of government stability, it is apparent that
over recent decades most governments in the world have experienced periods of interaction between
government stability, FDI, and exchange rates, as well as periods of non-interaction. Further, our estimates
provide clear understanding of the approximate levels of government stability necessary for facilitating the
bivariate FDI-exchange rate relationship, regardless of which particular channel through these bivariate
relationships occur (i.e., the six theories we consider beyond our baseline model). Moreover, from Figure 9,
the average nation’s level of government stability fluctuates over time around 8 points on the stability index
– the exact “threshold” that emerges from our empirical estimates; clearly, small changes in government
stability can have dramatic effects on these important variables in the macroeconomy.
5.4 Characterizing the Relationship Between FDI and the Exchange Rate
We can now turn towards characterization of our bi-directional FDI-exchange rate effects via the mutual-
ism classifications presented in Section 3. For each of our semiparametric models, including the baseline
specification, we classify each country according to its modal (across years) estimate. Take as a simple
example, a country with 5 time series observations. If in a majority of the 5 years the country has a sig-
nificantly positive FDI coefficient but insignificant exchange rate coefficient, this country will be classified
as FDI-commensalism, meaning that for this country FDI significantly affects the exchange rate but the
exchange rate does not affect FDI. In some cases, a country does not have a unique modal classification,
and so the country is listed multiple times in the table; we mark these countries in the table. We report
27
classification matrices listing each country’s placement in Tables 9 – 14.
It is clear from the classification of the baseline model in Table 9 that the majority of countries are
classified mutualism, meaning that both FDI and the exchange rate have a significantly positive effect on
each other. A relatively small number of countries are classified as FDI-commensalism, meaning that FDI
positively affects the exchange rate, but the exchange rate does not affect FDI inflows. Only a handful
of other countries exhibit other classifications. Looking across the classification matrices corresponding
to the auxiliary models, it is clear that many country’s FDI-exchange rate relationships are contingent
on the set of control variables included in each specification. For instance, moving from the benchmark
model classifications in Table 9 to the traded and non-traded goods theory classification, Table 10, the
predominant classification switches from mutualism to FDI-commensalism and FDI-antagonistic symbiosis
(positive FDI effect on the exchange rate but negative exchange rate effect on FDI). Taking a close second
look across all classification tables, it is clear that the predominant classifications – regardless of model
specification or sample size – is either mutualism or FDI-commensalism, with certain models (particularly
traded and non-traded goods, Table 10, and imperfect capital markets, Table 14) showing a relative large
number of countries classified as FDI-antagonistic symbiosis.
From these classifications, we gather that (i) these different theories generally reflect significant chan-
nels through which FDI and the exchange rate interact, at least for particular countries according to each
theory, and (ii) overall, FDI has a significantly positive effect on the exchange rate but that the exchange
rate has a more sensitive effect on FDI flows.
5.5 Model Assessment
One important way we glean additional insight from our model about the nature of parameter hetero-
geneity is to examine the cross-validated bandwidths used for regression estimation. It is widely accepted
(e.g., Li & Racine 2007) that if a continuous regressor’s cross-validated bandwidth does not exceed its
upper bound in a local linear regression, then that variable is chosen by the cross-validation procedure to
enter nonlinearly into the regression model. For discrete variables, a less than unitary bandwidth implies
nonlinear, nontrivial interactions in the regression. An examination of the cross-validated bandwidths in
our model shows that government stability has nonlinear interactive effects. That is, we find that our
government stability bandwidths are less than their upper bounds, which is a signal that the data do
not justify any ad-hoc parametric linear restriction. Further, the existence of nonlinear interactions does
not provide insight into the correct parametric specification. Hence, our bandwidth analysis signals that
parametric restrictions on the functional form of heterogeneity within our model should be carefully con-
sidered and supported by appropriate model specification tests. Also, the country and year fixed-effects
enter the FDI model in a nonlinear manner. We finally note that since the degree of smoothing varies
across equations for each regressor, we conclude that the nature of these nontrivial interactions differs
across equations as well.
6 Conclusion
In theory, FDI inflows can have positive, negative, or no effect on the exchange rate, and vice versa. If,
for example, within a country FDI has a positive effect on the exchange rate and the exchange rate has a
28
positive effect on FDI – our concept of mutualism – then FDI-promoting strategies for stabilizing changes
in the exchange rate have added and direct multiplier benefits. Further, recent incidences suggest that
government stability may have sizable implications for the interactions between FDI inflows and exchange
rates. To date, however, no study has analyzed empirically the types of interactions between the exchange
rate and FDI that may exist within and across countries and the effect of government stability on such
interactions.
In this paper, we characterize empirically the types of interactions between FDI and the exchange rate,
guided by the theoretical literature that provides six specific predictions about the FDI-exchange rate
relationship. We also analyze the effect of government stability on the FDI-exchange rate interactions.
To do so, we use a recently developed semiparametric system of simultaneous equations model that
accommodates the exchange rate and FDI as a bivariate response. This simultaneous equations model
allows us to coalesce several important aspects of the empirical and theoretical exchange rate and FDI
literatures, including (i) the joint determination of the exchange rate and FDI, (ii) nonlinear and nontrivial
interactions of government stability with each of the conditioning variables, (iii) an instrumental variables
approach for identification, (iv) unobserved heterogeneity (country- and time-specific effects) of unknown
and non-neutral form, and (v) correlations in errors across equations. Only a few existing papers have
explored even a subset of these important model structures.
We find several important interactions in the exchange rate-FDI-government stability nexus that have
not been documented by germane existing literatures. Specifically, our proposed semiparametric system
of equations model, and associated specification tools, suggests that across developed and developing
economies, causal, heterogeneous mutualism and FDI-commensalism are the most dominant types of in-
teractions between FDI and the exchange rate; this suggests that in most countries, FDI has a significantly
positive effect on the exchange rate, and in some countries the exchange rate has a significantly positive
effect on FDI while in others the exchange rate does not effect FDI inflows. We find that government sta-
bility is an important source of heterogeneity in these effects, particularly with approximately 6-8 points
being an important government stability threshold for which (i) above this threshold the FDI effect on
exchange rates substantially increases, and (ii) an inflection point in the exchange rate effect on FDI such
that the negative effect at low levels of government stability switches to become positive. We note that
the average world economy fluctuates right around this important threshold – including both developed
and developing nations – indicating that small changes in government stability can have dramatic effects
on relationships in the macroeconomy.
These findings are strong evidence in support of research advocating a more tailored, country-specific
set of macroeconomic policies for the relationship between the exchange rate and FDI. It is well-known
that neglected heterogeneity can lead to misleading inferences on the parameters of interest. Thus, our
findings underscore the importance of accounting for different sources of heterogeneities in a flexible –
rather than the traditionally ad hoc parametric – manner to obtain consistent and generally reliable
results. Our semiparametric system of simultaneous equations model coupled with its instrument-based
estimator seems appropriate for assessing empirically the types of interactions between the exchange rate
and FDI.
29
References
Aisen, A. & Veiga, F. J. (2006), ‘Does political instability lead to higher inflation? A panel data analysis’,
Journal of Money, Credit and Banking 37(5), 1379–1389.
Aizenman, J. (1992), ‘Exchange rate flexibility, volatility, and domestic and foreign direct investment’,
IMF Staff Papers 39(4), 890–922.
Akhtaruzzaman, M., Berg, N. & Hajzler, C. (2017), ‘Expropriation risk and FDI in developing countries:
does return of capital dominate return on capital?’, European Journal of Political Economy 49, 84–
107.
Alba, J. D., Wang, P. & Park, D. (2010), ‘The impact of exchange rate on FDI and the interdependence
of FDI over time’, Singapore Economic Review 55(04), 733–747.
Aleksynska, M. & Havrylchyk, O. (2013), ‘FDI from the south: the role of institutional distance and
natural resources’, European Journal of Political Economy (29), 38–53.
Asiedu, E. (2006), ‘Foreign direct investment in Africa: the role of natural resources, market size, gov-
ernment policy, institutions and political instability’, The World Economy 29(1), 63–77.
Asiedu, E., Jin, Y. & Nandwa, B. (2009), ‘Does foreign aid mitigate the adverse effect of expropriation
risk on foreign direct investment?’, Journal of International Economics 78(2), 268 – 275.
Barro, R. J. (1991), ‘Economic growth in a cross section of countries’, The Quarterly Journal of Economics
106(2), 407–443.
Beckmann, J., Ademmer, E., Belke, A. & Schweickert, R. (2017), ‘The political economy of the impossible
trinity’, European Journal of Political Economy 47, 103 – 123.
Beckmann, J. & Czudaj, R. (2017), ‘Exchange rate expectations and economic policy uncertainty’, Eu-
ropean Journal of Political Economy 47, 148 – 162.
Blonigen, B. A. (1997), ‘Firm-specific assets and the link between exchange rate and foreign direct in-
vestment’, The American Economic Review 87(3), 447–465.
Blonigen, B. A. (2005), ‘A review of the empirical literature on FDI determinants’, Atlantic Economic
Journal 33(4), 383–403.
Blonigen, B. A. & Piger, J. (2014), ‘Determinants of foreign direct investment’, Canadian Journal of
Economics 47(3), 775–812.
Bodea, C. (2015), ‘Fixed exchange rates with escape clauses: the political determinants of the European
monetary system realignments’, European Journal of Political Economy 39, 25 – 40.
Borensztein, E., Gregorio, J. D. & Lee, J.-W. (1998), ‘How does foreign direct investment affect economic
growth?’, Journal of International Economics 45(1), 115 – 135.
Branson, W. (1977), ‘The specification of goods and factor markets in open-economy macroeconomic
models’, Handook of International Economics 2. Dordecht: North-Holland.
Branson, W. H. (1981), Macroeconomic determinants of real exchange rates, NBER Working Papers 0801,
National Bureau of Economic Research, Inc.
Broll, U. & Wahl, J. E. (1992), ‘International investments and exchange rate risk’, European Journal of
Political Economy 8(1), 31 – 40.
30
Bruce, N. & Purvis, D. (1984), ‘The specification of goods and factor markets in open-economy macroe-
conomic models’, Handbook of International Economics 2, Dordrecht: North–Holland.
Brunetti, A. & Weder, B. (1998), ‘Investment and institutional uncertainty: a comparative study of
different uncertainty measures’, Review of World Economics 134(3), 513–533.
Busse, M. & Hefeker, C. (2007), ‘Political risk, institutions and foreign direct investment’, European
Journal of Political Economy 23, 397–415.
Cai, Z. & Li, Q. (2008), ‘Nonparametric estimation of varying coefficient dynamic panel data models’,
Econometric Theory 24, 1321–1342.
Campa, J. M. (1993), ‘Entry by foreign firms in the United States under exchange rate uncertainty’, The
Review of Economics and Statistics 75(4), 614–22.
Campa, J. M. & Goldberg, L. S. (1995), ‘Investment in manufacturing, exchange rates and external
exposure’, Journal of International Economics 38(3-4), 297–320.
Corbo, V. (1985), ‘Reforms and macroeconomic adjustments in Chile during 1974–1984’, World Develop-
ment 13(8), 893–916.
Cosset, J.-C. & Rianderie, B. D. (1985), ‘Political risk and foreign exchange rates: an efficient-market
approach’, Journal of International Business Studies 16(3), 21–55.
Culem, C. G. (1988), ‘The locational determinants of direct investments among industrialized countries’,
European Economic Review 32(4), 885–904.
Cushman, D. O. (1985), ‘Real exchange rate risk, expectations, and the level of direct investment’, The
Review of Economics and Statistics 67(2), 297–308.
Cushman, D. O. (1987), ‘The effects of real wages and labor productivity on foreign direct investment’,
Southern Economic Journal 54(1), 174–185.
De Grauwe, P., Danthine, J., Katseli, L. & Thygesen, N. (1991), North-south in the EMS - convergence
and divergence in inflation and real exchange rates, CEPS paper no. 50.
Delgado, M. S., McCloud, N. & Kumbhakar, S. C. (2014), ‘A generalized empirical model of corruption,
foreign direct investment, and growth’, Journal of Macroeconomics 42, 298–316.
Edwards, S. (1985), The behavior of interest rates and real exchange rates during a liberalization episode:
the case of Chile 1973-83, NBER Working Papers 1702, National Bureau of Economic Research, Inc.
Egger, P. & Winner, H. (2005), ‘Evidence on corruption as an incentive for foreign direct investment’,
European Journal of Political Economy 21(4), 932 – 952.
Egger, P. & Winner, H. (2006), ‘How corruption influences foreign direct investment: a panel data study’,
Economic Development and Cultural Change 54(2), 459–86.
Faeth, I. (2009), ‘Determinants of foreign direct investment – a tale of nine theoretical models’, Journal
of Economic Surveys 23(1), 165–196.
Fan, J. & Gijbels, I. (1996), Local Polynomial Modeling and its Applications, New York: Chapman and
Hall.
Ffrench-Davis, R. (1983), ‘The monetarist experiment in Chile: a critical survey’, World Development
11(11), 905–926.
31
Fielding, D. & Shortland, A. K. (2005), ‘How does political violence affect confidence in a local currency?
Evidence from Egypt’, Journal of International Development 17(7), 841–866.
Froot, K. & Stein, J. (1991), ‘Exchange rates and foreign direct investment: an imperfect capital markets
approach’, Quarterly Journal of Economics 106(4), 1191–1217.
Giavazzi, F. & Spaventa, L. (1990), The “new” EMS, Quaderni-working paper DSE.
Goldberg, L. S. & Kolstad, C. D. (1994), Foreign direct investment, exchange rate variability and demand
uncertainty, NBER Working Papers 4815, National Bureau of Economic Research, Inc.
Goldberg, L. S. & Kolstad, C. D. (1995), ‘Foreign direct investment, exchange rate variability and demand
uncertainty’, International Economic Review 36(4), 855–873.
Hacche, G. (1983), The determinants of exchange rate movements, OECD Economics Department Work-
ing Papers 7, OECD Publishing.
Hakkala, K. N., Norb¨ack, P. & Svaleryd, H. (2008), ‘Asymmetric effects of corruption on FDI: evidence
from Swedish multinational firms’, The Review of Economics and Statistics 90(4), 627–642.
Hall, P., Li, Q. & Racine, J. S. (2007), ‘Nonparametric estimation of regression functions in the presence
of irrelevant regressors’, Review of Economics and Statistics 89(6), 784–789.
Harberger, A. (1985), ‘Observations on the Chilean economy 1973-83’, Economic Development and Cul-
tural Change 33(3), 451–462.
Harms, P. & Ursprung, H. W. (2002), ‘Do civil and political repression really boost foreign direct invest-
ments?’, Economic Inquiry 40(4), 651–663.
Hau, H. (2002), ‘Real exchange rate volatility and economic openness: theory and evidence’, Journal of
Money, Credit and Banking 34(3), 611–630.
Hines, J. R. (1995), Forbidden payment: foreign bribery and American business after 1977, NBER Work-
ing Papers 5266, National Bureau of Economic Research.
IMF (2005), Angola: 2004 Article IV Consultation: Staff Report; Staff Statement; Public Information
Notice on the Executive Board Discussion; and Statement by the Executive Director for Angola, IMF
Staff Country Reports, International Monetary Fund.
Javorcik, B. & Wei, S. (2009), ‘Corruption and cross-border investment in emerging markets: firm-level
evidence’, Journal of International Money and Finance 28(4), 605–624.
Kiyota, K. & Urata, S. (2004), ‘Exchange rate, exchange rate volatility and foreign direct investment’,
The World Economy 27(10), 1501–1536.
Klein, M. W. & Rosengren, E. (1994), ‘The real exchange rate and foreign direct investment in the United
States: relative wealth vs. relative wage effects’, Journal of International Economics 36(3-4), 373–
389.
Kogut, B. & Chang, S. J. (1996), ‘Platform investments and volatile exchange rates: direct investment in
the US by Japanese electronic companies’, The Review of Economics and Statistics 78(2), 221–231.
Kolstad, I. & Villanger, E. (2008), ‘Determinants of foreign direct investment in services’, European
Journal of Political Economy 24(2), 518 – 533.
Kosteletou, N. & Liargovas, P. (2000), ‘Foreign direct investment and real exchange rate interlinkages’,
Open Economies Review 11, 135–148.
32
Li, Q., Maasoumi, E. & Racine, J. S. (2009), ‘A nonparametric test for equality of distributions with
mixed categorical and continuous data’, Journal of Econometrics 148(2), 186–200.
Li, Q. & Racine, J. S. (2004), ‘Cross-validated local linear nonparametric regression’, Statistica Sinica
14, 485–512.
Li, Q. & Racine, J. S. (2007), Nonparametric Econometrics: Theory and Practice, Princeton University
Press.
Li, Q. & Racine, J. S. (2010), ‘Smooth varying coefficient estimation and inference for qualitative and
quantitative data’, Econometric Theory 26, 1607–1637.
McCloud, N., Delgado, M. & Kumbhakar, S. (2018), ‘Foreign direct investment and growth symbiosis: a
semiparametric system of simultaneous equations analysis’, Journal of Econometric Methods 7(1), 1–
31.
Pearce, D. K. (1983), ‘Alternative views of exchange-rate determination’, Economic Review (68), 16–30.
Racine, J. S. & Li, Q. (2004), ‘Nonparametric estimation of regression functions with both categorical
and continuous data’, Journal of Econometrics 119(1), 99–130.
Russ, K. N. (2007), ‘The endogeneity of the exchange rate as a determinant of FDI: a model of entry and
multinational firms’, Journal of International Economics 71(2), 344 – 372.
S´anchez-Mart´ın, M. E., de Arce, R. & Escribano, G. (2014), ‘Do changes in the rules of the game affect
FDI flows in Latin America? A look at the macroeconomic, institutional and regional integration
determinants of FDI’, European Journal of Political Economy 34, 279 – 299.
Schneider, F. G. & Frey, B. S. (1985), ‘Economic and political determinants of foreign direct investment’,
World Development 13(2), 161–175.
Sung, H. & Lapan, H. E. (2000), ‘Strategic foreign direct investment and exchange-rate uncertainty’,
International Economic Review 41(2), 411–423.
Udomkerdmongkol, M., Morrissey, O. & Gorg, H. (2009), ‘Exchange rates and outward foreign direct
investment: US FDI in emerging economies’, Review of Development Economics 13(4), 754–764.
UNCTAD (2007), Transnational corporations, extractive industries and development, World Investment
Report 9789211127188, United Nations: New York and Geneva.
Wei, S.-J. (2000), ‘How taxing is corruption on international investors?’, The Review of Economics and
Statistics 82(1), 1–11.
Wheeler, D. & Mody, A. (1992), ‘International investment location decisions: the case of U.S. firms’,
Journal of International Economics 33(1-2), 57–76.
World Bank (2012), ‘World Development Indicators [CD-ROM]’, Washington, D.C.
Yu, J. & Walsh, J. P. (2010), Determinants of foreign direct investment, IMF Working Papers 10/187,
International Monetary Fund.
33
Technical Appendix
A.1 Estimation and Identification
To understand the estimation technique, let us first rewrite each equation jin (3.1) as
yj,it =e
X0
j,itgj(Zj,it ) + j,it,(A.1)
where e
X0
j,it := (Y0
−j,it, X 0
j,it) and gj(Zj,it ) := (λ0
j(Zj,it), γ 0
j(Zj,it))0and mj:= pj+kj.12 Given the general
unspecified form of our coefficient functions gj(·) in (A.1) and the absence of a distributional assumption
on j,it, MDK choose the generalized method of moments (GMM) approach to estimate the system of
equations. Heuristically, MDK first linearize the gj(·) in each equation using local-linear approximation
(Fan & Gijbels 1996); MDK apply this method to the system of equations as follows. Assume each gjis
sufficiently smooth and consider a first-order Taylor series expansion of gj(Zj,it) around a fixed point zj
in a neighborhood of {Zj,it }, so that the sth component of this expansion is
gs
j(Zj,it)≈as
j+ (bs
j)0(Zj,it −zj), s = 1, . . . , mj,(A.2)
where bs
j:= ∂gs
j(zj)/∂zj, a dj×1 vector of first-order derivatives. Note that for the j-th equation, the
remainder term of the second-order Taylor series expansion of the s-th component of gj(Zj,it), gs
j(Zj,it),
is
Rs
j(Zj,it, zj) = gs
j(Zj,it)−as
j−(bs
j)0(Zj,it −zj)−1
2(Zj,it −zj)0∇2gs
j(zj)(Zj,it −zj),(A.3)
and Rj(Zj,it, zj)=(R1
j(Zj,it, zj), R2
j(Zj,it, zj), . . . , Rmj
j(Zj,it, zj))0is a mj-dimensional vector. Define
¯
Rs
j(Zj,it, zj) := 1
2(Zj,it −zj)0∇2gs
j(zj)(Zj,it −zj) to be the second order term in the expansion, and
¯
Rj(Zj,it, zj) = ( ¯
R1
j(Zj,it, zj),¯
R2
j(Zj,it, zj),..., ¯
Rmj
j(Zj,it, zj))0.
Combining (A.1) and the first-order approximation in (A.2) we obtain
yj,it ≈U0
j,itαj+j,it ,(A.4)
where Uj,it := e
Xj,it
e
Xj,it ⊗(Zj,it −zj)!is a vector of dimension mj(dj+ 1), ⊗is the Kronecker product
operator, and the corresponding coefficient vector is αj:= (a1
j, . . . , amj
j,(b1
j)0,...,(bmj
j)0)0. Now stacking
observations by T, then by N, and then by Jgives the compact system formulation
y≈Uα +, (A.5)
where y= (y0
1, y0
2)0,U=block diag(U1, U2) so that for each j,Ujis a matrix of N T ×mj(dj+ 1)
observations on all right-hand side variables, α= (α0
1, α0
2)0, and = (0
1, 0
2)0with
j= (j,11, . . . , j,1T, . . . , j,21 , . . . , j,2T, . . . , j,N1, . . . , j,N T )0.
Assume the existence of additional information in the form of instruments, W, to ensure the identi-
12In general, Zj,it is required to be the same across gj(·) for any jbecause of substantial econometric difficulties that arise
in estimation of a semiparametric varying coefficient model in which the coefficient variables differ across coefficients.
34
fication of the αparameter in the system in (A.5). For the population moment conditions, let Vj,it :=
(W0
j,it, Z 0
j,it)0and Vit = (V0
1,it, V 0
2,it)0. Thus,
E(it|Vit )=0.(A.6)
In light of the moment equality in (A.6), for any measurable function Q(Vit),
E(it|Vit )=0 ⇐⇒ E(Q(Vit)it|Vit )=0.(A.7)
In the spirit of Cai & Li (2008), MDK choose for each equation j,
Qj,it := Q(Vj,it) = Wj,it
Wj,it ⊗(Zj,it −zj)/hj!,
which is a low-order polynomial vector of dimension lj(dj+ 1) in Wj,it and Zj,it,ljis the dimension of
Wj,it, and lj≥mjfor identification. In addition, the first entry of the vector Wj,it is equal to one.
For ease of exposition, let Q=block diag(Q1, Q2) so that for each j,Qjis a matrix of N T ×lj(dj+ 1)
observations on the variables in Qj,it. Also, let the system kernel matrix K=block diag(K1, K2) where
Kj=diag(Khj(Zj,11 −zj), . . . , Khj(Zj,N T −zj)) with Khj(·) := h−dj
jKj(·/hj), a kernel function in Rdj
for equation j. Define ˜mj:= mj(dj+ 1), ˜m:= ˜m1+ ˜m2, and similarly, ˜
lj:= lj(dj+ 1), ˜
l:= ˜
l1+˜
l2.
One specific local-linear GMM system MDK estimator that is relevant to our system analysis assumes
Z1,it =Z2,it =Zit,h1=h2=h, and K1=K2=K. This MDK local-linear GMM system estimator,
bαGM M , is defined as
bαGM M = arg min
α
(y−Uα)0e
KQΓ−1Q0e
K(y−Uα),(A.8)
where e
K=K⊗IJ, and Qand Uare as previously defined but with Zit in lieu of Zj,it,∀j, and Γ−1is a
known ˜
lט
lpositive definite weighting matrix. Then
bαGM M =hU0e
KQΓ−1Q0e
KUi−1hU0e
KQΓ−1Q0e
Kyi.(A.9)
The MDK estimator bαGM M is very easy to implement. Under the assumptions enumerated in MDK,
bαGM M is consistent and asymptotically normal – which renders easy statistical inference. We, however,
obtain standard errors for our estimate of αusing a wild-bootstrap predicated on 399 replications.
A.2 Selection and Interpretation of Bandwidth
As in Delgado, McCloud & Kumbhakar (2014), we select the optimal smoothing parameters, {hc, λu, λo}
using the method of least squares cross validation. The method of least squares cross validation selects
{hc, λu, λo}by minimizing the following criterion function
min
{hc,λu,λo}(nT J )−1
J
X
j=1
n
X
i=1
T
X
t=1 hyj,it −e
X0
−j,itbgj(Z−it)i2
,(A.10)
35
in which e
X0
−j,itbgj(Z−it) is the leave-one-out nonparametric GMM estimate of e
X0
j,itgj(Zit ). It is common
to employ a cross validation procedure to select bandwidths in applied research as regression estimates are
typically sensitive to choice of bandwidth parameter. Further, the least squares cross validation procedure
has been shown to asymptotically select the optimal bandwidth, and has been shown to have impressive
finite sample performance that includes the ability to detect nonlinearities in the data (Li & Racine 2007).
In the local-linear least squares approach, a continuous nonparametric variable that has nonlinear
interactions with other variables is assigned a relatively small bandwidth when chosen with the least
squares cross validation criterion. Li & Racine (2004) and Hall, Li & Racine (2007) show that an effective
finite sample threshold for interpretation of nonlinear effects is approximately two standard deviations of
the data. Variables whose cross validated bandwidth exceeds this threshold are interpreted to have linear
interactions with the other nonparametric variables. Hence, examination of the cross validated bandwidths
yields important insight into the data driven specification of heterogeneity within the model. Formal
statistical tests can be used to choose between linear parametric and potentially nonlinear semiparametric
models.
36
Table 8: List of countries grouped by percentile according to time-averaged government stability
Below 25th 25th to Below 50th 50th to Below 75th 75th and Above
G < 7.225 7.225 ≤G < 7.924 7.924 ≤G < 8.512 G≥8.512
Argentina Albania Angola Algeria
Bangladesh Bolivia Armenia Australia
Colombia Brazil Austria Azerbaijan
Costa Rica Bulgaria Bahrain Belarus
Czech Republic Dominican Republic Belgium Botswana
Ecuador Gabon Burkina Faso China
El Salvador Greece Cameroon Cyprus
Guatemala Guinea Canada Estonia
Haiti Guinea-Bissau Chile Finland
Honduras Guyana Croatia Iceland
Iraq Hungary Denmark Ireland
Israel India Ethiopia Jordan
Italy Indonesia France Kuwait
Kenya Jamaica Germany Libya
Liberia Japan Ghana Luxembourg
Malawi Lebanon Latvia Malta
Nicaragua Madagascar Lithuania Moldova
Niger Mali Malaysia Morocco
Nigeria Mexico Mongolia Namibia
Pakistan New Zealand Mozambique Oman
Papua New Guinea Norway Netherlands Saudi Arabia
Paraguay Panama Portugal Singapore
Peru Romania Senegal Slovenia
Philippines Sierra Leone South Africa Switzerland
Poland Thailand Spain Tanzania
Sri Lanka Togo Sudan Tunisia
Suriname Turkey Sweden Uganda
Zambia Ukraine United Kingdom United States
Zimbabwe Uruguay Vietnam
Note: The index of government stability comes from the International Country Risk Guide published by Political
Risk Services, and is defined as “the government’s ability to carry out its declared program(s), and its ability
to stay in office”. The index is the sum of three subcomponents – government unity, legislative strength and
popular support – each with a maximum score of 4 points and a minimum score of 0 points; a score of 4 points
equates to very low risk and a score of 0 points to very high risk. This table indicates the groups of countries
that fall into the quartile ranges of the distribution of the time-averaged level of government stability, denoted
by G, for the period 1984 to 2010.
37
Table 9: Countries and their types of interactions between ∆ exchange rate and FDI – benchmark model
∆ Exchange Rate Effect
Positive Negative Zero
FDI Effect
Positive Mutualism:
Algeria, Argentina*, Australia, Austria, Azer-
baijan, Bahrain, Bangladesh, Botswana*, Brazil,
Burkina Faso, Cameroon, Canada, China, Colom-
bia*, Cyprus, Denmark, Dominican Repub-
lic,Ecuador, El Salvador, Estonia*, Finland,
France, Gabon*, Germany, Ghana, Greece,
Guinea, Haiti, Honduras, Hungary, India, Ire-
land, Israel*+, Italy, Jamaica, Japan, Kenya,
Kuwait*, Latvia, Libya, Luxembourg, Mada-
gascar, Malaysia, Mali, Morocco, Mozambique,
Namibia, Netherlands, Nicaragua, Niger*, Nor-
way, Oman, Pakistan, Panama, Peru, Philippines,
Poland, Portugal, Saudi Arabia, Senegal*, Sierra
Leone, Singapore, Slovenia, South Africa*, Spain,
Sri Lanka, Sudan*, Suriname, Sweden, Switzer-
land, Tanzania, Thailand, Togo, Tunisia, Turkey,
United Kingdom, United States, Uruguay, Viet-
nam, Zimbabwe
FDI-Antagonistic
Symbiosis:
FDI-Commensalism:
Albania, Angola, Argentina*, Armenia, Belarus,
Belgium, Botswana*, Bulgaria, Chile, Colombia*,
Croatia, Estonia*, Ethiopia, Gabon*, Guatemala,
Guyana, Iceland, Israel*+, Jordan, Kuwait*,
Lebanon, Liberia+, Lithuania, Malawi+, Malta,
Mexico, Moldova, Mongolia, New Zealand,
Niger*, Nigeria, Papua New Guinea, Senegal*,
South Africa*, Sudan*, Zambia+
Negative ∆ Exchange Rate -Antagonistic Symbiosis:
Costa Rica+
Synnercrosis: ∆ Exchange Rate -Ammensalism:
Malawi+, Paraguay
Zero ∆Exchange Rate - Commensalism:
Costa Rica+, Israel*+, Romania, Uganda
FDI-
Ammensalism:
non-Symbiosis:
Bolivia, Czech Republic, Guinea-Bissau, Indone-
sia, Iraq, Israel*+, Liberia+, Ukraine, Zambia+
Repeated countries in the same row are marked with *. Repeated countries in the same column are marked with +. Repeated countries across rows
and columns are marked with *+.
38
Table 10: Countries and their types of interactions between ∆ exchange rate and FDI – M1
∆ Exchange Rate Effect
Positive Negative Zero
FDI Effect
Positive Mutualism:
Albania, Australia, Croatia*, Cyprus,
Ecuador*, Ghana, Kuwait, Latvia,
Lebanon, Luxembourg, Mali*, Malta*,
Nicaragua+, Niger
FDI-Antagonistic Symbiosis:
Angola, Armenia, Brazil,
Cameroon*, Canada, Chile,
China*, Colombia*, Croatia*,
Ecuador*, El Salvador*, Estonia+,
Ethiopia, Guatemala*+, Guinea,
Guyana, Haiti*, India, Indonesia+,
Israel+, Italy, Japan*, Kenya*,
Lithuania, Mali*, Mexico*+, New
Zealand*, Oman, Papua New
Guinea, Philippines, Poland,
Sierra Leone, Singapore, Spain+,
Sri Lanka, Thailand*+, Tunisia,
Turkey*, United Kingdom*, Viet-
nam, Zambia
FDI-Commensalism:
Armenia, Austria, Azerbaijan+, Bahrain,
Bangladesh, Belgium, Botswana, Bul-
garia, Burkina Faso, Cameroon*, China*,
Colombia*, Denmark+, Dominican
Republic,El Salvador*, Finland, Ger-
many, Greece, Guatemala*+, Haiti*,
Honduras, Hungary, Ireland, Jamaica,
Japan*, Kenya*, Libya, Madagascar,
Malawi+, Malaysia, Malta*, Mexico*+,
Moldova, Mongolia, Morocco, Mozam-
bique, Namibia, Netherlands, New
Zealand*, Nigeria, Peru, Portugal, Saudi
Arabia, Senegal, Slovenia+, Sudan, Swe-
den, Thailand*+, Togo, Turkey*, United
Kingdom*, United States, Uruguay
Negative ∆ Exchange Rate -Antagonistic
Symbiosis:
Argentina*, Gabon, Nicaragua+,
Paraguay*
Synnercrosis:
Argentina*+, Costa Rica,
Guatemala*+, Indonesia+, Israel+,
Mexico*+, Paraguay*
∆Exchange Rate -Ammensalism:
Guatemala*+, Iraq*+, Malawi+,
Mexico*+
Zero ∆Exchange Rate - Commensalism:
Algeria, Argentina*+, Belarus*, Guinea-
Bissau, Latvia, Slovenia*, Uganda*,
Ukraine
FDI-Ammensalism:
Argentina*+, Czech Republic*,
Estonia+, Iceland*, Liberia*,
Spain+, Thailand*+
non-Symbiosis:
Azerbaijan+, Belarus*, Bolivia, Czech
Republic*, Denmark+, France, Iceland*,
Iraq*+, Ireland*, Jordan, Liberia*,
Romania, Slovenia*+, South Africa,
Thailand*+, Uganda*
Repeated countries in the same row are marked with *. Repeated countries in the same column are marked with +. Repeated countries across rows
and columns are marked with *+.
39
Table 11: Countries and their types of interactions between ∆ exchange rate and FDI – M2
∆ Exchange Rate Effect
Positive Negative Zero
FDI Effect
Positive Mutualism:
Albania, Angola+, Argentina, Austria, Bahrain,
Bangladesh, Belgium, Bolivia+, Brazil*, Burk-
ina Faso, Cameroon, Canada, China, Croa-
tia, Denmark+,El Salvador, Estonia*, France,
Germany, Ghana, Guatemala, Guinea, Hon-
duras, Hungary, Iceland+, India, Indonesia, Ire-
land, Israel, Italy, Jamaica, Japan, Jordan*,
Kenya, Kuwait+, Latvia, Lebanon, Libya, Lithua-
nia, Luxembourg, Malaysia+, Mexico+, Mo-
rocco, Mozambique, Netherlands, Niger*, Nige-
ria, Norway, Oman, Pakistan+, Papua New
Guinea, Poland, Portugal, Senegal+, Sierra
Leone, Singapore, Slovenia+, South Africa,
Spain+,Sri Lanka+, Sudan, Sweden+, Thailand,
Tunisia, Turkey, United Kingdom, United States,
Uruguay, Vietnam, Zambia, Zimbabwe*
FDI-Antagonistic
Symbiosis:
FDI-Commensalism:
Armenia, Belarus, Brazil*, Bulgaria, Colom-
bia, Costa Rica, Estonia*, Gabon+, Guinea-
Bissau, Jordan*, Mali, Moldova+, Mongolia+,
New Zealand, Niger*, Paraguay, Peru, Togo, Zim-
babwe*
Negative ∆ Exchange Rate -Antagonistic Symbiosis:
Algeria, Angola+, Australia+, Kuwait+,
Pakistan+, Slovenia+, Sri Lanka+
Synnercrosis:
Haiti
∆Exchange Rate -Ammensalism:
Zero ∆Exchange Rate - Commensalism:
Australia+, Azerbaijan, Bolivia+, Botswana,
Cyprus, Denmark+, Finland, Greece*, Iceland+,
Kuwait+, Malaysia+, Malta*, Mexico+,
Mozambique+, Namibia, Philippines, Saudi
Arabia, Senegal+, Spain+, Sri Lanka+, Sweden+,
Switzerland, Uganda
FDI-
Ammensalism:
Romania
non-Symbiosis:
Czech Republic, Gabon+, Greece*, Guinea-
Bissau+, Malawi, Malta*, Moldova+, Mongolia+,
Ukraine
Repeated countries in the same row are marked with *. Repeated countries in the same column are marked with +. Repeated countries across rows
and columns are marked with *+.
40
Table 12: Countries and their types of interactions between ∆ exchange rate and FDI – M3
∆ Exchange Rate Effect
Positive Negative Zero
FDI Effect
Positive Mutualism:
Angola, Australia, Austria*, Azerbaijan*,
Bahrain*, Burkina Faso, Cameroon+, China,
Colombia*, Denmark*, Finland*, France, Greece,
Haiti*, Hungary, Ireland, Japan, Kenya*,
Kuwait*, Libya, Malaysia*, Mali, Morocco,
Netherlands*, Niger*+, Oman, Peru*, Portugal,
Sierra Leone, Singapore, Slovenia, Spain, Suri-
name, Switzerland, Tanzania, United States*,
Vietnam
FDI-Antagonistic
Symbiosis:
Liberia, Niger*+
FDI-Commensalism:
Albania, Armenia, Austria*, Azerbaijan*,
Bahrain*, Bangladesh, Belarus, Belgium,
Botswana, Brazil, Bulgaria, Canada, Colombia*,
Croatia, Cyprus, Denmark*, Dominican Repub-
lic, Ecuador, Estonia, Ethiopia, Finland*, Gabon,
Germany, Guatemala, Guyana, Haiti*, Honduras,
Iceland, India, Indonesia, Italy, Jamaica, Kenya*,
Kuwait*, Latvia, Lebanon, Lithuania, Mada-
gascar, Malaysia*, Malta, Mexico, Moldova,
Mongolia, Mozambique, Namibia, Netherlands*,
New Zealand, Nicaragua*, Nigeria, Norway,
Panama, Papua New Guinea, Peru*, Philippines,
South Africa, Sri Lanka, Sweden, Togo, United
Kingdom, United States*, Uruguay, Zambia
Negative ∆ Exchange Rate -Antagonistic Symbiosis:
Costa Rica+, Niger*+
Synnercrosis:
Niger*+
∆Exchange Rate -Ammensalism:
Malawi, Nicaragua*, Paraguay
Zero ∆Exchange Rate - Commensalism:
Algeria, Argentina*, Cameroon+, Costa Rica+,
Ghana, Poland, Romania, Thailand*, Tunisia,
Zimbabwe
FDI-
Ammensalism:
Ukraine
non-Symbiosis:
Argentina*, Bolivia, Botswana, Bulgaria, Czech
Republic, Israel, Jordan, Romania, Senegal, Thai-
land*, Uganda
Repeated countries in the same row are marked with *. Repeated countries in the same column are marked with +. Repeated countries across rows
and columns are marked with *+.
41
Table 13: Countries and their types of interactions between ∆ exchange rate and FDI – M4
∆ Exchange Rate Effect
Positive Negative Zero
FDI Effect
Positive Mutualism:
Algeria*, Australia, Austria*,
Azerbaijan*, Bahrain, Cameroon,
Canada*, China*, Colombia*, Den-
mark*, Finland, France*, Ghana,
Greece*, Hungary, Japan, Kenya,
Kuwait*, Libya, Luxembourg, Mo-
rocco*, Mozambique, Nicaragua,
Oman, Pakistan, Portugal*, Saudi
Arabia, Senegal*, Slovenia, Spain*,
Sri Lanka*, Switzerland, Tanzania,
Thailand, Togo, Tunisia, United
States*, Vietnam, Zimbabwe
FDI-Antagonistic
Symbiosis:
Jordan
FDI-Commensalism:
Albania, Algeria*, Angola, Argentina, Arme-
nia,Austria*,Azerbaijan*,Bangladesh, Belarus, Belgium,
Bolivia, Botswana, Brazil, Bulgaria, Burkina Faso, Canada*,
Chile, China*, Colombia*, Croatia, Cyprus, Denmark*, Do-
minican Republic, Ecuador, El Salvador, Estonia, Ethiopia,
France*, Gabon, Germany, Greece*, Guatemala, Guinea,
Guyana, Haiti, Honduras, Iceland, India, Indonesia+, Ireland,
Israel+, Italy, Jamaica, Kuwait*, Latvia, Lebanon, Liberia,
Lithuania, Madagascar, Malawi, Malaysia, Mali, Malta,
Mexico+, Moldova, Mongolia, Morocco*, Namibia, Nether-
lands, New Zealand, Niger, Nigeria, Norway, Panama, Papua
New Guinea, Peru+, Philippines, Poland, Portugal*, Romania,
Senegal*, Sierra Leone, Singapore, South Africa, Spain*, Sri
Lanka*, Sudan, Suriname, Sweden, Thailand, Turkey, Uganda,
United Kingdom, United States*, Uruguay
Negative ∆ Exchange Rate -
Antagonistic Symbiosis:
Costa Rica*
Synnercrosis:
Guinea-Bissau*,
Iraq, Ukraine
∆Exchange Rate -Ammensalism:
Costa Rica*, Czech Republic, Guinea-Bissau*, Indonesia+,
Israel+, Mexico+, Paraguay, Peru*, Zambia
Zero ∆Exchange Rate - Commen-
salism:
FDI-
Ammensalism:
non-Symbiosis:
Repeated countries in the same row are marked with *. Repeated countries in the same column are marked with +. Repeated countries across rows
and columns are marked with *+.
42
Table 14: Countries and their types of interactions between ∆ exchange rate and FDI – M5
∆ Exchange Rate Effect
Positive Negative Zero
FDI Effect
Positive Mutualism:
Argentina, Australia, Bahrain,
Bangladesh, Botswana*, Bulgaria*,
Canada, Costa Rica*, Finland,
France*, Ghana+, Hungary*, Ireland+,
Japan,Jordan, Kenya*, Kuwait, Luxem-
bourg, Malaysia, Malta, Namibia, New
Zealand, Pakistan+, Peru, Philippines,
Saudi Arabia, Slovenia+, Tanzania, Thai-
land, United Kingdom*, United States,
Vietnam, Zimbabwe
FDI-Antagonistic Sym-
biosis:
Bangladesh, Belgium,
Botswana, Brazil, Bul-
garia*, Canada, Costa
Rica*, Croatia, Cyprus,
Denmark, Greece, Hun-
gary*, Italy, Malta, Mexico,
Netherlands*, New Zealand,
Nigeria+, Norway, Papua
New Guinea+, Poland,
Portugal, Romania+, Sri
Lanka, Turkey, Uruguay*,
Zambia+
FDI-Commensalism:
Austria, Botswana*, Bulgaria*, Costa Rica*,
Czech Republic, Ecuador, France*, Germany,
Hungary*, Indonesia, Israel, Jamaica, Kenya*,
Morocco, Netherlands*, Romania*, South Africa,
Spain,Sweden, United Kingdom*, Uruguay*
Negative ∆ Exchange Rate -Antagonistic
Symbiosis:
China, Ireland+, Oman, Singapore,
Tunisia
Synnercrosis:
Colombia*, India, Lebanon,
Nigeria+, Papua New
Guinea+, Paraguay*,
Romania+,Zambia+
∆Exchange Rate -Ammensalism:
Colombia*, Romania*
Zero ∆Exchange Rate - Commensalism:
Algeria, Azerbaijan, Ghana+, Pakistan+,
Panama, Slovenia+, Switzerland
FDI-Ammensalism:
Chile, Papua New Guinea+,
Paraguay+, Ukraine
non-Symbiosis:
Repeated countries in the same row are marked with *. Repeated countries in the same column are marked with +. Repeated countries across rows
and columns are marked with *+.
43