The Effect of Foreign Productivity on FDI
Decisions
Jennifer L. Steele
May 12, 2006
Abstract
A firm’s decision of whether or not to undertake foreign direct investment (FDI) depends on their expected payoff. The more productive they
expect to be, relative to their home productivity, the more likely they
are to produce in the foreign country. When firms have varying levels of
productivity in the home country, the decision of whether or not to invest
in a foreign country will be based on the firm’s expectation of their productivity in that foreign country, and will differ from firm to firm. If their
foreign productivity is proportional to their home productivity, only the
relatively more productive firms choose to produce in the foreign country. Dependent on the distribution of foreign productivities, incomplete
information may yield an equilibrium where only the firms that are less
productive in the home country choose to enter the foreign country.
A firms’ choice of whether or not to serve a foreign market, and then whether
to serve it through export or FDI, is dependent on both their home and foreign
productivities. If firms know both their home and their foreign productivities
then as their foreign productivity increases their expected cost of producing in
the foreign country decreases.
Factor price equalization suggests that the large differences in wages across
countries are reflective of the differences in productivities between workers.
However, due to limited capital mobility, wage differentials may not accurately
1
reflect the differences in productivities, but instead reflect other factors, such as
level of industry operating in the country, barriers to entry, opportunity costs
for the workers and trade barriers.
Most trade literature treats FDI as a method of avoiding transport costs
when serving a new market. However, in developing economies a large portion of
FDI is an effort to cut the costs of production, rather than to serve a new market.
At times the firm may not even serve the market in the country where they
produce, choosing to export all production. This paper explores the motivations
behind FDI when the foreign demand may not be sufficient to warrant producing
in the foreign country just to serve the foreign market, and when there are
potential cost savings from producing in the foreign country over producing at
home.
Differences between a firm’s productivity in the home country and the foreign country can be divided into two main sources. First, the industry-specific
difference in productivity. This could be influenced by the educational system,
cultural practices, or institutions. Some countries may be more productive in
textile industries because of cultural emphasis on handicrafts, or becuase of the
group culture leading to more productive assembly lines, while other countries
may be more adept at work involving machines because of the emphasis on
machinery in the educational system. These are not to be thought of as innate
abilities, or characteristics of certain countries. Rather they should be thought
of as learned skills from social interaction in different cultural environments or
the educational system, which can vary dramatically from country to country
in terms of both degree of education and educational focus.
The second difference is derived from the individual firms’ technology, the
ability to adapt that technology to a foreign workplace. Firms are heterogeneous
with respect to labor productivity in the home market, and this is based on
their firm specific technology. Some technologies may be more productive than
others in a foreign workplace, and may not differ in the same way as in the
home country (i.e. more productive technologies in the home market are not
necessarily more productive in the foreign market).
2
Current theory suggests that FDI should be undertaken at much higher levels
than is seen in the data. In their 2002 paper Rodrik and Hausmann [3] suggest
that industrial investment is lower than expected due to incomplete information.
Entrepreneurs do not know what they are good at producing. In section 4 the
concept is expanded to evaluate firm’s FDI decisions when they do not know
their foreign productivity. Increases in uncertainty lead to the underprovision
of FDI under certain conditions.
Generally the spillovers from FDI make it a very appealing method to developing countries to promote economic growth. However, the means of encouraging FDI continue to elude policy makers. Traditionally countries have been
urged to follow free market policies in order to increase growth. Countries which
have resisted this open economy push in the short run, like South Korea and
Taiwan, saw strong growth, while countries making the requisite open economy
changes, like Brazil or Argentina, have not.
Assymetric information also suggests a reason why, when FDI does occur, it
often clusters in a few select industries, often quite different between countries.
For example, Bangladesh has a large concentration of hat and pant producers, Columbia cut flowers, and India software design. If firms learn about the
productivity of their industry from other firms’ investment decisions, then the
effects of assymetric information are decreased for that industry.
In section 1 I’ll look at what has been done in the FDI literature, and
where my paper differs. Then in section 2 the basic model is introduced and
the autarkic equilibrium is outlined. The model is extended to two countries,
and the complete information case is identified in section 3. In section 4 some
effects of incomplete information are outlined. Finally in section ?? the policy
implications of differing productivities and incomplete information are explored.
1
Background
There have been many models that determine how a firm decides on type and
level of FDI. However, few have explained the underinvestment in developing
3
countries, where differences in factor prices would suggest much higher levels of
investment. Models that have attempted to explain this gap used a variety of
econometric and analytic methods to create the asymmetries. Nocke and Yeaple
[7] set up an econometric test of a firm’s decision between FDI through greenfield
investment and FDI through cross-border acquisition, allowing countries to have
different levels of development, evidenced by their factor prices. They find that
both the efficiency of the firm and the level of development of the recipient
country affect the choice of FDI. Markusen and Wigle [5] look at ’explaining the
volume of North-South trade’. They use the hypotheses of high protectionism
in the south and economic size to explain the levels of trade, and conclude
that without these barriers the level of North-South trade would be higher than
North-North trade. Using a different angle, Grossman and Helpman [2] look
at the decision between outsourcing and FDI where the level of monitoring
in outsourcing contracts is country specific, and there is a wage differential
between the home and foreign countries. They find that countries with lower
levels of monitoring receive fewer outsourcing contracts. The hypothesis in this
paper is that although trade costs and country size do play important roles in
a firm’s FDI decision, the presence of incomplete information explains more of
the gap between expected and actual FDI, and also helps explain the existence
of industry-clusters within a country.
In terms of information asymmetries, there have been a variety of different
approaches to FDI and uncertainty. Rob and Vettas [8] look at the decision to
invest in FDI when there is demand uncertainty in the recipient country. As with
most FDI papers, they look at FDI as a choice between exporting to a foreign
market, and setting up a subsidiary to serve that foreign market. Horstmann
and Markusen [4] incorporate uncertainty as to the consumer’s perception of
the firm’s quality. By incorporating firm-specific reputation assets, they study
the firm’s decision between outsourcing, exporting and FDI. The uncertainty
leads to a higher than expected probability of choosing FDI to serve the foreign
market.
Helpman, Melitz and Yeaple [1] explore the choice between export and FDI
4
with heterogeneous firms. Every firm faces the decision of whether or not to
serve a foreign market, and whether to serve it through FDI or exporting. Their
framework suggests that heterogeneity in productivity determines the firms’ FDI
decisions, the least productive produce domestically and serve the domestic
market only, more productive firms export to the foreign market and the most
productive firms set up subsidiaries in the foreign market. However they do not
allow for differences in productivity between the two countries. The strength
of their paper lies in its empirics, where they use firm level data to show the
predicted link between firm-level heterogeneity and the ratio of exports to FDI
sales.
Although not an FDI paper, Rodrik and Hausmann [3] explain why Latin
America has not experienced the growth that Asia has, although Latin America
has made more attempts to follow the growth-creating policies suggested by
development economists. They use a two sector model with a traditional and
a modern sector, and allow uncertainty to enter in the production functions of
the goods in the modern sector. They conclude that uncertainty can lead to
too little investment and entrepreneurship, but once firms have invested there
is too much product variety in the market, as unproductive firms stay in once
the sunk costs are committed. Empirically they look at three ’building blocks’
for their argument, i) large element of uncertainty as to what a country will be
good at producing, ii) significant difficulties entailed in importing technology
off-the-shelf, with many changes required for local adaptation, iii) imitation follows quickly once the first two blocks are overcome.
2
Model Overview
Using a two country, general equilibrium model, the effects of differences in
factor prices, incomplete information, and heterogeneity in productivity levels
on the decisions of the firm is outlined. For this paper I concentrate on trade
between a developed country (the home country), and a less developed country
5
(the foreign country). The foreign country spends a smaller amount of money
on the heterogeneous good industries, and may have lower wages.
There are M industries in the world. In each mM industry a modern good
is produced, of which an infinite number of varieties are available for production.
For ease of analysis we assume each firm employs labour as the sole factor of
production. In each industry the set of varieties being sold in country c’s market
is denoted as V c . Each firm in the industry thus produces a variety vV c . Each
firm produces output by the following production function:
Yic = Lci ãci
(1)
where Yic is the output of firm i in country c, Lci is the labour hired by firm i in
country c, ãci is the productivity of firm i in country c and 1/ãci is the number
of workers required by firm i in country c to produce one unit of output. ãci
is then referred to as the firm’s productivity in terms of output units produced
per unit of labour. Letting wc indicate the wage in country c.
aci =
ãci
wc
(2)
Now aci represents the output units produced per dollar and is henceforth
referred to as the firm’s productivity in country c.
Consumer Preferences
Consumers in both countries have CES preferences, represented by the log
utility function:
u = (1 −
M
X
βm ) log(z) +
m=1
Z
M
c
X
βm
log(
xm (v)αm dv)
α
c
m
vVm
m=1
(3)
Where xm (i) is the consumption of variety i (from firm i) in sector m, and
Vmc
is the set of all firms in sector m selling their good in the country c. The
elasticity of substitution is ε =
1
1−α .
It is assumed that ε is greater than one,
and in section 4 it is assumed that ε lies between one and two. The proportion
6
of income spent on goods in sector m is fixed at βm and Σm βm = 1.
From these preferences we find that the demand for good i in sector m is
xcm (i) = R V c
0
βm E c
pc (v)1−ε dv
pc (i)−ε
(4)
and the monopolistic firm producing in country c will offer their good in country
c at price pcc (i) = (aci α)−1 . A firm producing in country c exporting their good
for sale in country ¬c will offer their good at price pc¬c (i) = τ c¬c (aci α)−1 . E c
is the income of country c and τ c¬c represents the iceberg transport costs of
shipping a good from country c to country ¬c, and τ c¬c = τ ¬cc = τ > 1.
This gives us the following profit for a firm producing in country c and serving
country c:
π cc = (1 − α)(ac α)ε−1 Ac
(5)
And for a firm producing in country c and serving country ¬c
π c¬c = (1 − α)(ac α)ε−1 A¬c τ 1−ε
(6)
Each firm faces fixed costs from three different sources. First, they must
pay fe to enter the industry, after which they draw their productivity ac from
the distribution G(a). Then, if they choose to produce, they bear fixed costs
fp for each plant location. Finally, they must pay distribution costs fx for each
market they wish to sell their good in.
Autarkic Equilibrium
We begin by looking at the equilibrium in the home country, assuming the
home country is currently in an autarkic equilibrium, with no firms exporting to
or operating in the foreign country. Following Melitz (2003) [6] the firms are heterogeneous with respect to their productivity in the home country. Firms that
are currently producing goods know their productivity in the home country, and
firms that are not currently producing know the distribution of productivities
they could draw from, were they to enter the market. Each industry consists
7
of an equilibrium determined number of heterogeneous firms, differentiated by
their productivity levels. Firm i producing at home has productivity ahi
Because the firm’s only option is to produce for the domestic market or not
produce, we only have one cutoff point ah∗
i
1
above which it will be profitable
to produce, and below which the firm will choose not to produce. Given the
cutoff we can determine whether or not the firm will choose to pay fe and draw
a productivity.
Because we are summing over the prices of all firms in the sector, an increase
in the number of firms decreases the expected profits of a potential entrant, so
firms will continue to enter until the expected profits reach zero, noting that the
cutoff point, ah∗ is increasing in the number of firms in the market. From there,
given the fixed entry costs fe we can determine how many firms will serve the
home market under autarky (na ).
Export Equilibrium
If we open up the foreign country, and allow the firms producing in the autarkic equilibrium in the home country to start exporting to the foreign country,
firms now have three options: a1 do not produce, a2 produce in the home country and serve the home market only, a3 produce in the home country and serve
both the home and foreign market. It is assumed that the foreign country is
sufficiently small
2
so it will never be optimal to produce in the home country
and serve only the foreign market.
There are a number of reasons why at some point in time a foreign country
may ’open up’. It may be that previously it was simply too expensive to export
to that market (τ was too large). This encompasses both trade restrictions
(tariffs) or transport costs. A decrease in transport costs or tariffs could lead
to the foreign country suddenly being attractive to home firms. Another reason
could be the size of the foreign market. If the foreign economy is growing,
1 (ah∗ )ε−1
=
fx +fp
(1−α)Ah αε−1
where Ah = R V h
βE h
ph (v)1−ε dv
2 β f E f is sufficiently small, so it could mean that they have a lower GDP or that they
spend a smaller amount of their income in that specific industry
8
at some point home firms will choose to export to the foreign market to take
advantage of the growth in income.
Using the firm’s payoffs from each possible action we get the following productivity cutoff points:
ah1 =
ah2 =
fp + fx
(1 − α)Ah αε−1
fx
(1 − α)Af τ 1−ε αε−1
Once a firm has drawn a productivity level, they will not produce if ah < ah1 ,
will produce and only serve the home market if ah1 < ah < ah2 and will produce
and serve both the home and foreign markets if ah > ah2 .
Compared to the autarkic equilibrium, the expected profit from entering the
market is higher because the firms retain the payoff from serving the home market, plus get profits from serving the foreign market. As a result, more firms
will now enter the market. Thus if firms are allowed to export their products
to a foreign market, the number of firms, ne , producing in the home market in
equilibrium will be higher.
3
Benchmark Model
In the benchmark model we look at two countries. The home country is currently
operating in an export equilibrium in industry m, and there are (ne ) firms
producing and selling their good in the home market. There are no firms in
industry m producing in the foreign country although they may be exporting
their goods to the foreign market.
Now firms have six possible actions (given their home and foreign productivities) 3 .
3 assuming that the foreign market is sufficiently small, producing at home only for the
foreign market, and producing only in the foreign country for only the foreign market will
f +f
never be optimal. Algebraically this requires Af < τ ε−1 Ah xf p
x
9
• a1 = do not produce
• a2 = produce in the home country and serve the home market only
• a3 = produce in the home country and serve both the home and foreign
markets
• a5 = produce in the foreign country and serve the home market
• a6 = produce in the foreign country and serve both the home and foreign
markets
• a8 = produce in the home country and the foreign country to serve their
respective markets
Given these actions and the firm’s profit function (equations (5) and (6)),
we can determine the optimal actions for any set of productivities {ah , af }.
If the foreign market is sufficiently small4 , it may be profitable to produce in
the foreign market and not serve the foreign market. For example, Banana
Republic makes many of its items in the Philippines, but does not sell its goods
4 Af
fx
x +fp
< Ah τ 1−ε f
10
in the Philippines. This is even more apparent in input goods, where a firm
that produces inputs for another industry may choose to produce some of their
components in a foreign country, even though there are no firms that demand
their good in the foreign market.
Looking at figure 1, the optimal actions for each possible set of productivities
are outlined. If the firm is relatively unproductive ({ah , af } are sufficiently
close to zero) then the firm will choose not to produce in either country. As ah
increases the firm chooses to produce in the home country, serving the home
market, and as ah continues to increase, they will produce in the home market
and export to the foreign market. If af is increasing, and the foreign market is
sufficiently small, the firm may go from not producing to producing in the foreign
market and only serving the home market, to producing in the foreign market
and serving both market. If both home productivity and foreign productivity
are sufficiently large the firm may choose to produce in both the home and
foreign markets, serving their respective markets.
If we allow the firm’s foreign productivity to be a linear transformation of
their home productivity (for all home productivity levels), we can see the effects
of differences in wage, or in human capital levels. If the number of workers
required to make one unit of output is the same in both the home and foreign
countries, then the only differentiating factor would be the wage. Allowing
(af )ε−1 = c(ah )ε−1 , if the productivity levels were the same in both countries,
but the wages were different, c would simply equal
wh
.
wf
If we had factor price
equalization, where wages counteracted any difference in productivity, c would
equal one.
The concept of comparative advantage suggests that relative productivities
across industries may differ from country to country. We could then denote
the industry specific productivity difference as cm . If the home country had a
comparative advantage in industry i and the foreign country had a comparative
advantage in industry j then cj > ci .
Looking at a firm in industry i (but dropping the subscript) with both the
large foreign market and the small foreign market, as c grows the firm becomes
11
increasingly likely to engage in FDI. However, because we assume that fixed
costs are the same in both countries the firm’s actions depend largely on shipping
costs. If c < τ 1−ε no firms will produce in the foreign country. This is because
the increased costs of production are greater than the savings from no longer
shipping to the foreign market, even for the most productive firms. For any
values of c between τ 1−ε and τ ε−1 the firm will decide between producing at
home, and producing in the foreign country based on their productivity level
at home. As their home productivity level increases they are more likely to
produce in the foreign country. Finally, if c > τ ε−1 all firms will produce in the
foreign country. In this case the lower cost of production in the foreign country
outweighs the cost of shipping the products from the foreign country to the
home market even for the least productive firms.
In the benchmark model there will be more weakly more firms producing in
equilibrium than in the export equilibrium. Assuming that all firms know the
transformation from ah to af then they just have to pay fe once, to determine
their productivity levels in both countries. Allowing c to vary across productivities (firms then know c(ah ) for all possible ah ), if c(ah ) < τ 1−ε for all ah then
the equilibrium will be the same as the export equilibrium. If c(ah ) > τ ε−1 for
all values of ah once the foreign country is open to FDI all firms will choose
to produce in the foreign country, profits then must be higher than under the
export equilibrium (given the distribution of ah ), so the expected profits would
be higher with only ne firms in the market, so there must be more than nE
firms in the market. More generally, if c(ah ) > τ 1−ε for some value of ah then
the expected profits are weakly greater than in the export equilibrium for all
firms, and there will be weakly more firms producing.
In terms of comparative advantage and gains from trade, if the productivity
gain is great enough, all firms in an industry will produce in the foreign country.
Likewise, if there is no productivity gain all firms will produce in the home
country, and only the most productive firms will produce in the foreign country
for the foreign market (in order to save the shipping costs). Opening up the
foreign economy to imports from the home country will always result in more
12
firms producing in the home country. Further opening up the foreign economy
to FDI may result in additional firms producing, but where the firms choose to
produce depends on the relative productivity of the foreign country to the home
country. The size of the comparative advantage will determine the equilibrium
structure.
4
Incomplete Information
In many cases a firm may not be able to infer their foreign productivity from
their home productivity. Instead they may have some beliefs about the distribution of the productivity, where the cdf of their beliefs is G(af |ah ) and
G0 (af |ah ) = g(af |ah ). We assume that this cdf is common knowledge.
Assumption If ahi > ahj then G(af |ahi ) < G(af |ahj ).
If firm i’s productivity at home is higher than firm j’s productivity at home,
then firm i’s expected productivity in the foreign country must first order stochastically dominate firm j’s. So if a firm is less productive in the home country,
ex-ante they expect to be less productive in the foreign country as well.
Now firms must choose whether or not to invest in the foreign country based
on the expectation of their productivity in the foreign country and their beliefs
about the distribution of that productivity. They must also choose whether
or not to serve the foreign market when producing in the foreign country, and
whether or not to close their home country production facilities when producing
abroad. If a firm is already operating in the home country, their expected profits
for each action are now a function of the expected distribution of productivity
levels in the foreign country, and can be written as follows:
(a1 ) Do not produce at home or in the foreign market:
π=0
13
(7)
(a2 ) Produce at home and serve the home market:
π hh − fx − fp = Ah (1 − α)(ah )ε−1 αε−1 − fx − fp
(8)
(a3 ) Produce at home and serve both the home and foreign markets:
π hh + π hf − 2fx − fp = Ah (1 − α)(ah )ε−1 (Ah + Af τ 1− )αε−1 − 2fx − fp (9)
(a5 ) Produce in the foreign country and serve the home market (Only optimal
if home productivity is sufficiently low, and the foreign market is sufficiently
small):
R af2
π
fh
h 1−
− fx − fp = (1 − α)(A τ
)α
f
ε−1 a1
aε−1 g(a|ah )da
G(af2 ) − G(af1 )
− fx − fp
(10)
(a6 ) Produce in the foreign country and serve both the home and foreign
markets:
R∞
π
fh
+π
ff
− 2fx − fp = (1 − α)α
ε−1
h 1−
(A τ
f
+A )
af2
aε−1 g(a|ah )da
1 − G(af2 )
− 2fx − fp
(11)
(a8 ) Produce at home and serve the home market and produce in the foreign
country to serve the foreign market (Only optimal if home productivity is
sufficiently high):
hh
ff
ε−1
π +π −2(fx +fp ) = (1−α)α
Ah (ah )ε−1 +Af
R af2
aε−1 g(a|ah )da −2(fx +fp )
G(af2 ) − G(af0 )
(12)
af0
Choosing between a1 , a2 and a3 depend solely on the firm’s home productivity (ah ) and exogenous variables, specifically the size of the home and foreign
markets, and the fixed costs. The payoffs from actions a4 , a6 and a8 depend on
the distribution of foreign productivities, G(af |ah ).
14
Looking back at figure 1, we can identify three ranges of home productivity
which involve firms putting positive weights on one or two of the foreign production actions {a5 , a6 or a8 }. The first range is if the foreign market is sufficiently
small, and home productivity is sufficiently low
((ah )ε−1 <
fx τ 1−ε
)
Af (1−α)αε−1
x
Af < Ah τ 1−ε fxf+f
p
In this case the firm will not produce in the foreign country if af < af1 ,
where af1 is a function of the firm’s home productivity. They will produce in the
foreign country, only serving the home market if af1 < af < af2 and will produce
in the foreign country serving both markets if af > af2 .
The second range involves moderate levels of home productivity. Here the
firm will not produce in the foreign country if af < af2 and will produce in the
foreign country, serving both markets (action a6 ), otherwise. In this case af2 is
a function of the firm’s home productivity.
The final range is sufficiently high levels of (ah )5 . In this case the firm will
choose action a8 , produce in both countries, serving both markets, if af0 < af <
af2 , and will choose action a6 , produce only in the foreign country, serving both
markets, if af > af2 . Here both af0 and af2 are functions of the firm’s home
productivity.
If the foreign country is sufficiently large, then only the second and final
ranges will be applicable (whenever the firm is producing in the foreign market
they will also want to serve the foreign market).
Assumption ε[1, 2]
The elasticity of substitution, ε, is assumed to be between one and two. This
is a common macroeconomic assumption, and in this case it implies that profits
are concave with respect to productivity.
5 (ah )ε−1
>
fp
Ah τ 1−ε +Af
(1−α)αε−1 Ah Af (τ ε−1 −τ 1−ε )
15
Assuming that the fixed cost of drawing your productivity in the foreign
country is the same as in the home country (fe ), a firm currently producing in
the home market has two options: draw (and pay fe ) or don’t draw and limit
their action set to {a1 , a2 , a3 }. If the firm draws they can then choose from
actions {a1 , a2 , a3 , a5 , a6 , a8 }. When deciding whether or not to draw the firm
looks at the expected payoff from drawing (the probability of a5 , a6 or a8 being
optimal multiplied by the payoff less the optimal action had the firm chosen not
to draw the foreign productivity)), and if it’s higher than fe the firm will draw.
As the number of firms who draw their foreign productivity grows, the number
of firms producing in the foreign country will also grow, and expected profits
will decrease until the market is in equilibrium.
However, it’s not necessarily true that firms with higher home productivity
are more likely to draw their foreign productivity. If they are more productive at
home the gains from drawing the foreign productivity may be lower, dependent
upon the distribution G(af |ah ) If the first moment of the distribution increases
less than proportionally with home productivity (E(af |ah ) is a concave function
of ah ) then the more productive a firm is, the lower their expected profit from
drawing a foreign productivity (because their opportunity cost is higher). As a
result, the equilibrium could consist of only the firms that are less productive
at home producing in the foreign country.
Using backwards induction, if a firm is not currently producing in the home
country, the expected profit from drawing a home productivity level (given the
set of all actions) must be weakly greater than that in the export equilibrium.
As a result, with incomplete information we would expect to see more firms
active in equilibrium than in the export equilibrium. Given certain distributions,
G, we could see fewer firms than in the complete information model, as the
uncertainty may decrease the expected profits, leading fewer firms to draw their
foreign productivity.
Some distributions G yield an equilibrium where only the more unproductive
home firms will draw their foreign productivity, yielding a weakly less productive
foreign market than a complete information equilibrium.
16
Looking at this model from a development perspective, it is unclear whether
incomplete information decreases FDI. However, it is clear that certain distributions yield unsavory results, and there are equilibriums where only the less
productive firms produce in the foreign market.
17
References
[1] M. J. M. Elhanan Helpman and S. R. Yeaple. Export versus fdi with heterogeneous firms. American Economic Review, 94(1):300, 2004.
[2] G. M. Grossman and E. Helpman. Outsourcing versus fdi in industry equilibrium. Journal of the European Economic Associaton, 1(2), 2003.
[3] R. Hausmann and D. Rodrik.
Economic development as self-discovery.
NBER Working Paper, 8952, May 2002.
[4] I. J. Horstmann and J. R. Markusen. Licensing versus direct investment: A
model of internalization by the multinational enterprise. Canadian Journal
of Economics, 20(3):464, 1987.
[5] J. R. Markusen and R. M. Wigle. Explaining the volume of north-south
trade. Economic Journal, 100(403):1206, 1990.
[6] M. J. Melitz. The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica, 71(6):1695, 2003.
[7] V. Nocke and S. Yeaple. An assignment theory of foreign direct investment.
NBER Working Paper, 11003, 2004.
[8] R. Rob and N. Vettas. Foreign direct investment and exports with growing
demand. Review of Economic Studies, 70(3):629, 2003.
18