Markusen lectures-may2010

Lectures on International Trade and Investment:
Multinational Firms, Offshoring, Outsourcing, and Environment
James R. Markusen
University of Colorado, Boulder
University College Dublin
ETH Zurich May 25-28
Each lecture will be approximately two hours. You will get far more out of the lectures if you
work through the background reading ahead of time!
Required background reading is draft chapters (in order of use in lectures) from:
Markusen, James R. and Keith E. Maskus, “International Trade: Theory and Evidence”,
Wiley publishers, forthcoming early 2011.
Chapter 11: Imperfect Competition and Increasing Returns I: Oligopoly
Chapter 12: Imperfect Competition and Increasing Returns II: Monopolistic-
Competition
Chapter 16: Direct Foreign Investment and Multinational Firms
Chapter 17: Fragmentation, Offshoring, and Trade in Services
Chapter 10 Distortions and Externalities as Determinants of Trade
Tuesday, May 25
Lecture 1
Review of the Industrial Organization Approach to Trade
Oligopoly: pro-competitive gains from trade
Monopolistic Competition: gains through increased product and input variety
Lecture 2
Multinational Firms: Preliminaries, Partial-Equilibrium Models
Stylized facts
A general conceptual approach
Partial equilibrium analysis of location: horizontal and vertical firms
Wednesday, May 26
Lecture 3
Multinational Firms: General-equilibrium models, empirical evidence
The horizontal model
The knowledge-capital model
Export-platform strategies
Empirical evidence

Lecture 4
Multinational Firms: Outsourcing versus vertical integration
Asset specificity, holdup, and incentives
Physical versus knowledge capital
The Antrás holdup-model
The Markusen knowledge-capital and learning model
Thursday, May 27
Lecture 5
Fragmentation, Offshoring, and Trade and Investment in Services
The general gains-from-trade problem
The Grossman - Rossi-Hansberg trade in tasks version
The Markusen - Venables version
Trade in services and the four modes of supply
Lecture 6
Production and Consumption Distortions and Externalities in General-Equilibrium
The structure of a general-equilibrium model: the complementarity approach
Production distortions
Consumption distortions
Production externalities
Friday, May 28
Lecture 7
Trade in the presence of Environmental Externalities
Pollution from production affecting utility
Pollution from consumption affecting utility
Production externalities among firms and sectors
International spillovers
Lecture 8
Trade Policy in the Presence of Environmental Externalities
Policies and international spillovers
Cooperative policies and issue linkages
Tax competition and the race to the bottom

Chapter 11
IMPERFECT COMPETITION AND INCREASING RETURNS I: OLIGOPOLY
11.1 General discussion of increasing returns and non-comparative-advantage gains
from trade.
In the previous chapter and also in Chapter 2 (e.g., Figures 2.6 and 2.7), we noted that increasing
returns to scale, or scale economies for short, offer the opportunity for gains from trade even for identical
economies. Trade and gains from trade can occur without any comparative advantage basis for trade.
Trade economists believe that scale economies offer an important explanation for the observation of large
volumes of trade between similar economies that we observe in trade data.. In this Chapter and in the
next, we will analyze these idea in a more thorough fashion.
In this chapter, we will focus on a case where there are two industries, one producing a good
under conditions of perfect competition and constant returns to scale, and a second good under conditions
of increasing returns to scale. This IRS good is assumed to be homogeneous, meaning the goods
produced by different firms are identical from the consumers’ point of view. In the next chapter, we
assume that each firm in the IRS sector produces a good that is differentiated from those of its rival firms.
In order to avoid some possible confusion, we partly depart from our usual notation here and denote the
competitive good as Y and the IRS homogeneous good as X. This will prove useful in the next section
when we have different varieties of X.
We will assume that the X firms have technologies in which there is an initial fixed cost of
entering production and then a constant marginal cost of added output. We assume a single factor of
production L (call it labor) which must be divided between the Y and X sectors and among firms in the X
sector. Marginal cost is denoted by mc and total cost (tc) and average cost (ac) for an X firm are as
follows:
1
(11.1)
where mc and fc are constants (parameters), measured in units of labor. Figure 11.1 shows the cost
curves of a single firm. It is important to note that average cost is always falling with increased output
and it becomes close to marginal cost, but never quite equal to marginal cost. In mathematical
terminology, average cost is a rectangular hyperbola, going to infinity as X goes to zero and approaching
mc as X goes to infinity.
The consequences of this type of technology for competitive conditions, sometimes referred to as
market structure are rather profound. Specifically, this type of increasing-returns technology cannot
support perfect competition as a market outcome. This is typically explained by a proof by contradiction.
Suppose that there are many small firms such that each firm regards market price as exogenous. If price
equaled marginal cost as assumed in competitive theory, then each firm would be making losses since
average cost is greater than marginal cost, so this cannot be an equilibrium. Suppose then that the
competitive price exceeds marginal cost but again is viewed as a constant by each firm. Then each or any
firm knows that once it produces enough output, price will exceed average cost and so profits become
positive. But the firm should not stop there, it would maximize profits at the constant price by producing
an infinite output and so should expand without bounds. But this cannot be an equilibrium and
contradicts the assumption that firms are small. Thus the technology shown in (11.1) and Figure 11.1 is
inconsistent with perfect competition.
The equilibrium outcome (assuming an equilibrium exists) must be one in which only a small
number of firms are able to survive in equilibrium. Competitive will be imperfect, with each firm having

2
influence over market price. The consequences of the technology in Figure 11.1 are shown for a single
monopoly X firm in Figure 11.2. Also assume a very simple technology for Y, such that one unit of L
produces exactly one unit of Y: Y = Ly. Y and L must have the same price and we will use L and therefore
Y as numeraire, giving them a price of one. is the economy’s total endowment of L.
The production frontier for the economy, discussed briefly in Chapter 2, runs from to to
in Figure 1 where to is the fixed cost fc measure in units of Y which is of course also units of L
in this special case of Y = Ly. Suppose that the monopoly equilibrium is at point A in Figure 11.2, where
the output of Y and X are and . Then the average cost of producing X at point A is given by the
total amount of labor needed for X divided by the output of X.
(11.2)
But note that this is the slope of the dashed line in Figure 11.2, the line running from point A to the end of
the production frontier . The consequence of this is that, if the monopoly X firm is to at least break
even, then the price line at equilibrium A must be at least as steep as the dashed line giving average cost.
Such a price is shown at pa in Figure 11.2, where p gives the price of X in terms of Y (and therefore in
terms of L).
A crucial point is that the market equilibrium involves a market failure: the market outcome is not
optimal. The optimum would be a tangency between an indifference curve and the production frontier,
implying price equal to marginal cost, but we have shown that this is impossible because it involves the
firm making losses. With increasing returns to scale that are “internal” to individual firms as we are
assuming here, returns to scale are going to be inherently bound up with imperfect competition. Too little
X is produced at too high a price.
That is a rather long preliminary that now allows us to get to the heart of the issue: the
opportunity to trade is going to offer economies the possibility of pro-competitive gains from trade.
Countries will gain from the benefits of increased competition that makes firms produce more at lower
average costs (higher productivity) and lower prices. We will focus on putting together two identical
economies. Within this setting we consider two cases, one in which the number of firms is fixed before
and after trade opens up and one in which there is free entry or exit of firms in response to trading
opportunities.
11.2 Pro-competitive gains: the basics
Any discussion of imperfect competition must start with some assumption about how firms react
to one another, since each firm is going to be large relative to the market. In this chapter, we are going to
assume what is called Cournot-Nash (or Cournot for short) competition in which firms pick a quantity as
a best response to their rivals’ quantities. Cournot equilibrium occurs when each firm is making a best
response to its rivals’ outputs. Algebraically, this will be modeled as each firm picking an output quantity
holding rivals’ outputs constant.
Revenue for a Cournot firm i and selling in country j is given by the price in j times quantity of
the firm’s sales. Price is a function of all firms’ sales.
. where Xj is total sales in market j by all firms (11.3)

3
Cournot conjectures imply that ; that is, a one-unit increase in the firm’s own supply is a
one-unit increase in market supply. Marginal revenue is then given by the derivative of revenue in (11.3)
with respect to firm i’s output (sales) in j.
since (Cournot) (11.4)
Now multiple and divide the right-hand equation by total market supply and also by the price.
(11.5)
The term in square brackets in (11.5) is just the inverse of the price elasticity of demand, defined as the
proportional change in market demand in response to a given proportional change in price. This is
negative, but to help make the markup formula clearer we will denote minus the elasticity of demand,
now a positive number, by the Greek letter 0 > 0. We can then write (11.5) as
(11.6)
The term Xij/Xj in (11.6) is just firm i’s market share in market j, which we can denote by sij . Then
marginal revenue = marginal cost is given by:
(11.7)
Marginal revenue in Cournot competition turns out to have a fairly simple form as shown in
(11.7). . The term is referred to as the “markup”. As you will note, it looks something like the tax
formulas we had in the previous chapter and indeed it is possible to think of the monopolist as putting on
sort of a tax that raises price above marginal cost. If you refer back to equation (11.4) the markup relates
to how much the market price falls when the firm increases its sales. When the demand elasticity is high,
there is only a small fall in price when the firm expands output, and so the markup is small.
The role of the firm’s market share is more subtle, but crucial to understanding the whole idea of
pro-competitive gains. Suppose that we put two absolutely identical countries together in trade, each
having just a single X producer. The firms could just continue to do what they were doing in autarky, and
prices etc. would be preserved. But each firm will now understand that, if it increases output by one unit,
the increase is spread over twice as many consumers and thus the price will fall by only half as much as it
would if the monopolist increased supply by one unit in autarky. We sometime say that the firm now
perceives demand as more elastic or that the perceived elasticity of demand is higher. This is
reflected in (11.7) by the fact that sij goes from one in autarky to one-half when trade opens. The markup
up falls, perceived marginal revenue increases, and each firm has an incentive to increase output.

4
Figures 11.3 and 11.4 work through the consequences of this. The production frontier shown in
Figure 11.3 is the production frontier for each of the two identical countries. Point A is the autarky
monopoly equilibrium in each country. When the countries are put together in trade, each firm perceives
demand as more elastic according to (11.7) and expands output. Equilibrium is restored when the price of
X is forced down so that (11.7) is again an equality for both firms with market shares both equal to one
half. The new equilibrium must be at a point like T in Figure 11.3.
Note especially that this must constitute a welfare gain for both countries. Output was too low
and the price was too high initially in autarky, essentially a distortion like those of the previous chapter.
However, this is quite different in that it is an endogenous distortion and the opening of trade can be
thought of as reducing that distortion. There is a welfare gain from trade (Ua to U*)for the identical
economies through a lower X price (pa to p*), and higher X output (A to T), and more efficient
production: the average cost of a unit of X is lower, firms move down their average cost curves in Figure
11.1.
Figure 11.4 shown a somewhat different and equally important case. Suppose that there might be
more than one firm in each country initially, and that firms can enter or exit until profit are zero.1 Then,
referring back to equation (11.2) and Figure 11.2, the equilibrium price must be equal to average cost,
given by the line connecting the production point in Figure 11.4 with the endpoint of the production
frontier . Let point A in Figure 11.4 be the initial autarky equilibrium for each country and let pa denote
the autarky price: the price equals average cost reflecting the entry or exit of firms until profits are zero.
The distance in Figure 11.4 is now the combined or sum of the fixed costs of all firm active in the
market in equilibrium.
Now put the two identical economies together in free trade as before. Each firm individually has
an incentive to expand output: equation (11.7) continues to apply to pricing decisions. However, when
they all do this, profits will be driven negative. This will cause the exit of some firms in each country
until zero profits are re-established. This is represented by the outward shift of the linear segment of the
production frontier in Figure 11.4 from to : resources which were being used (uselessly) in fixed
costs are now freed up for actual production. The new equilibrium will be at a point such at T in Figure
11.4. Note that it is possible now for the consumption of both Y and X to increase due to the efficiency
gains of a larger world output being produced by fewer firms than the total number in the two countries in
autarky. Welfare in each country rises from Ua to U*.
A typical intelligent question at this point is how firm exit is consistent with more competition?
The answer is that there is some exit in each country individually, but more left in free trade in total
between the two countries than were in each individual country in autarky. For example, suppose that
each country has four firms in autarky. Suppose that trade forces the exit of one firm in each country,
leaving three in each country. But that means that there are now six firms in total competing for the
business of each consumer instead of four in autarky. Exit in each country individually is quite
compatible with increased competition.
The nice outcomes (for free trade advocates at least) shown in Figures 11.3 and 11.4 have not
been rigorously established, and then are a number of problems. First, the elasticity of demand which we
denoted by 0 is generally not a constant: it will change as prices and total sales change. While the basic
message is clear and some readers and professors will wish to move on, we now proceed to offer two
special cases that have been widely used in the literature to more rigorously solve for the effects of trade.
11.3 Special case I: quasi-linear preferences
The first special (or rather specific) case has been widely used in industrial organization as well
as in international trade. In the latter, it has been widely used to analyze what is known as “strategic trade
policy”, a topic treated later in the book. It will also be used later in analyzing multinational firms.

5
We can keep the notation simple at first by assuming just a single monopoly firm in each country,
so n = 1, and by normalizing the population L to the number one, so X give both total output of X, the
output per firm, and the consumption per capita. Assume that the preferences of a representative
consumer in each country are given by:
(11.8)
The crucial property of these preferences are constant marginal utility for good Y and the importance of
this will become clear in a minute. This linearity in Y has led some authors to call these preferences
“quasi-linear”. As above, one unit of factor L is need to produce one unit of good Y, and p denotes the
price of X in terms of Y or L. Let A denote profits of the firm and L (equal to one initially) the number of
workers/consumers. Profits are viewed as exogenous by individual consumers. Then the budget
constraint for the representative consumer is given by income (also measured in terms of L) equals
expenditure.
(11.9)
Substituting from the budget constraint for good Y in (11.8), we have the consumer’s choice problem,
where profits (i) are viewed as exogenous by an individual consumer.
(11.10)
The (inverse) demand function is given by the first-order condition, the derivative of (11.10) with respect
to X, and is linear in X.
(11.11)
The feature of quasi-linear preferences that makes them so attractive is that demand does not
depend on income and hence it does not depend on profits. This makes the model much easier to solve.
On the other hand, the unattractive feature of these preferences is that the demand for X does not depend
on income, there is a zero income-elasticity of demand for X, surely a totally unrealistic assumption. All
added income at a fixed price for X will be spent on Y.
Consider autarky first. Let  denote the profits of the domestic firm. These profits are revenues
minus marginal costs (mc will be denote by just c) and minus fixed costs, denoted simply as f.
Substituting in the demand function in (11.11) for p, we get
(11.12)
The first-order condition for profit maximization gives:
(11.13)
Now assume that two identical economies trade freely. Let the two countries and their firms be
denoted with subscripts i and j. There are now twice as many consumers in the integrated world economy
and the demand price depends on per capita consumption not on total world consumption. That is, world

6
price will be unchanged if there is twice as much output in the integrated world as there was in each
country individually in autarky. We now have L = 2 and so the world price of X will
(11.14)
Profits for firm i are now given by
(11.15)
Assume Cournot competition, so each firm makes a best response to the other firm’s output, maximizing
profits holding the other firm’s output fixed. The first-order condition for profit maximization is given
by
(11.16)
There is a corresponding equation for firm j, giving two equations in two unknowns. But since
we have assumed that the countries are absolutely identical, we know that the solution will be symmetric
with both firms producing the same amount in equilibrium. Exploiting this symmetry, we can solve
(11.16) for the Cournot output of the firm i (equal to the country j firm’s output) by setting Xi = Xj.
(11.17)
where the right-hand inequality is the autarky output of each firm given in (11.13). Output of each firm
expands by one-third ((2/3)/(1/2) = 1/3) when trade opens. This is the effect shown in Figure 11.3 and it
must imply mutual gains from trade for the two countries.
Now let’s turn to the case of free entry and exit. Again, let’s just think about one market and
think of trade between two countries as doubling the size of the market. This allows us to simplify the
notation.
Assume that there are L individuals (which can again be normalized to one in autarky as we did
above). We will also now need to keep track of the difference between output per firm and aggregate X
output. Summing over the number of firms n, per capita consumption is
(11.18)
The number of firms, n, is now endogenous.
Demand and profits for the ith firm are then given by
(11.19)
(11.20)

7
The firm’s first-order condition is the derivative of (11.20) with respect to Xi holding all of the other
firm’s outputs constant. Marginal revenue minus marginal cost for firm i is given by:
(11.21)
Now we again know that there will be symmetry in equilibrium: the output of any active firm will
be the same as that of every other firm. X will denote the output of an individual (not aggregate)
“representative firm”, and n the number of firms. All firms that are active in equilibrium will produce the
same amount.
(11.22)
The second equation we need for equilibrium is the free-entry condition that will be associated
with the number of firms. Then the zero profit condition is that the profits of the representative firm are
exactly zero.
(11.23)
Multiple (11.22) through by X.
(11.24)
We then have two equations (11.23 and 11.24) in two unknowns, n and X. Solving these two equations in
two unknowns we get
(11.25)
So output per firm increases with the size of the economy (L). But with price equal to average cost, this
must also mean that the equilibrium price of X falls. Finally, putting (11.25) into (11.22), we can solve
for n
(11.26)
The number of firms increases with the square root of the size of the world economy. A
restriction that the economies are sufficiently big such that n > 1 in autarky is sufficient to imply that
when L doubles the number of firms less than doubles. But this in turn means that each country
individually must have exit relative to the number of firms in autarky. It is much the same as our
numerical example in section 11.2 above: each country has some exit yet the integrated world economy
has more firms in total than the individual countries did in autarky. The situation is exactly like that
shown in Figure 11.4.
In summary then, the combination of increasing returns to scale with imperfect competition
means that there are gains from trade even for identical economies under the assumptions used here,
regardless of whether or not there is a fixed number of firms (same in each country) or there is free entry
and exit in response to the opening of trade. Note finally from Figures 11.3 and 11.4 that these gains
from trade for identical economies are not associated with any net trade between the economies. Both

8
countries are at point T in free trade. Yet (given our assumption of costless trade) there could be a lot of
two-way gross trade flows, with some consumers arbitrarily buying X from the producer in the other
country. This is referred to a intra-industry trade and the there is a lot of such trade among the high-income
developed countries.
11.4 Special case II: Cobb-Douglas preferences (can be skipped without loss of continuity)
One of the big limitations of the quasi-linear case is that it imposes the assumption that there is a
zero income elasticity of demand for X. But surely a lot of the manufacturing and service industries that
are characterized by increasing returns to scale and imperfect competition are producing goods with high
income elasticities of demand. Let X again denote the output of an individual X firm and assume that
there is a fixed number n of such firms. Y denotes the total output of Y as before. Suppose that
preferences are Cobb-Douglas and given by
with income (I) constraint (11.27)
We treated this exact case earlier in Chapter 2. Continue to let the price of Y be numeraire, equal
to one. We showed in Chapter 2 that the demand functions are
(11.28)
In this case, if we compute the elasticity of demand for X, we will find that 0 = 1. This is going
to greatly simplify our analysis, though note that we will need to be sure that there is more than one firm
producing in each country in equilibrium; that is, the market share of a firm must be less than one in
autarky in order for marginal revenue given by (11.7) to be positive. Subject to this restriction, the
markup is just the firm’s market share, which in turn is just 1/n. Marginal revenue equals marginal cost is
given quite simply by
(11.29)
So far, so good. But things get messy because income now matters for the demand for X and
profits are part of income. This is exactly the mess that quasi-linear preferences avoids! Since X refer to
the output of a single firm and, assuming all firms producing in equilibrium are identical, then nX gives
total X output. Substituting the budget constraint from (11.27) into the demand function (11.28) and
writing out the expression for profits, the aggregate demand for X is given by:
(11.30)
Rearranging the equation and making use of the second equation in (11.29) to eliminate the endogenous
variable p gives us
(11.31)
Dividing through by n(c) and multiplying one term by (n – 1)/(n – 1), we have

9
(11.32)
and then an explicit solution for X in the case when n is fixed.
(11.33)
Suppose now that we put two identical economies together, each with a fixed number of firms n.
doubles in (11.33) and so does n relative to autarky. Then the first bracketed term on the right-hand
side of (11.33) does not change: both the numerator and denominator double. But the second bracketed
term increases given " < 1. For example, let n = 2 in autarky and let " = 0.5. Then the introduction of
trade (n = 2 to n = 4) increases the second term from 2/3 to 6/7, an increase in output per firm of 29
percent. Once again, this is exactly the situation shown in Figure 11.3. Note from (11.29) that the price
of X also falls as shown in the Figure.
Now let us consider the free entry and exit version of the Cobb-Douglas case. The marginal
revenue, marginal cost equation is unchanged.
(11.34)
The free entry or zero-profit equation is given by
(11.35)
and, with no profit income, the demand for X is given by
(11.36)
Multiple (11.34) through by X and then divide (11.35) by (11.34)
(11.37)
which gives us output per firm.
(11.38)
Multiple both sides of (11.34) by X, and substitute for pX from (11.36)..
(11.39)

10
Now substitute the expression for X in (11.38) to give us a solution for n, the endogenous number
of firms.
(11.40)
Take the square root of the right-hand equation to get n and then substitute this into (11.39) to get X.
(11.41)
With free entry and exit, both the number of firms and the output per firm increase with the
square root of the size of the market. Once again, thinking of trade as a doubling of market size, this
means that there is some exit in each country individually with the opening of trade. The effects of trade
for two identical economies are exactly those shown in Figure 11.4.
11.5 Summary
To this point in the book, we have concentrated on determinants of trade that involve differences
between countries. Gains from trade involve exploiting these differences, such as producing and
exporting according to comparative advantage. Now we come to a situation in which there can exist
gains from trade even between identical economies and indeed we concentrate here on precisely this case.
We did touch on the role of increasing returns to scale in the previous chapter, but now we turn to a fuller
analysis and assume that scale economies occur at the level of the individual firm, sometimes termed
“internal” economies of scale to distinguish them from the external economies of the previous chapter.
The situation quickly becomes complicated because internal or firm-level economies of scale are
not compatible with perfect competition and hence the simple tools needed to analyze competitive models
need to be extended. Increasing returns to scale and imperfect competition are inherently related to one
another. This require introducing new methods and new tools to take into consideration firms with
market power and the strategic interaction between such firms. In doing so, we simplified our theoretical
economies in other ways, in particular on the factor market side, assuming only a single factor as in the
Ricardian model of trade.
In this chapter, we focused on an industry in which firms produce a homogeneous good or
alternatively the goods of the different firms are perfect substitutes. Firms compete according to the
Cournot-Nash model, choosing quantities that are best responses to the outputs of other firms. Autarky
equilibrium for a country is distorted, and we showed that this distortion looks a lot like a production tax
as modeled in the previous chapter. Price exceeds marginal cost in market equilibrium, and too little is
produced at too high a price to maximize social welfare.
We then open up two identical economies to trade and show that this generates a pro-competitive
effect. In technical terms, firms perceive demand as more elastic and hence expand outputs in response to
the opening of trade. But when all firms do this, industry output expands, the price and markup falls,
average cost falls (or productivity improves) and social welfare increases.
We considered two versions of the model, one in which the number of firms is fixed before and
after the opening of trade and one in which there is free entry and exit of firms in response to trade. The
effects just mentioned in the previous paragraph are present in both cases, but the free entry/exit case is
particularly interesting in that it may be possible for the identical economies to end up consuming more of

11
both goods, the competitive good as well as the increasing-returns good. The important lesson is that
trade does not offer welfare gains just based on difference between countries, it also offers gains to very
similar countries in terms of more efficient production, lower prices, and high consumption quantities.

12
REFERENCES
Brander, James A. (1981), “Intra-industry trade in identical commodities”, Journal of International
Economics 11, 1-14.
Helpman, Elhanan. and Paul A. Krugman (1985). Market Structure and Foreign Trade, Cambridge: MIT
press.
Horstmann, Ignatius J. and James R. Markusen (1986), “Up the average cost curve: inefficient entry and
the new protectionism”, Journal of International Economics 20, 225-228.
Markusen, James R. (1981), “Trade and the gains from trade with imperfect competition”, Journal of
International Economics 11, 531-551.
Melitz, Mark J. and Gianmarco I.P. Ottaviano (2008), “Market size, trade, and productivity”, Review of
Economic Studies 75, 295-298.
Venables, Anthony J. (1985), “Trade and trade-policy with imperfect competition - the case of indentical
products and free entry”, Journal of International Economics 19, 1-19.
Endnotes
1. The assumption that firms enter or exit until profits are exactly zero means that we are
allowing the number of firms to be a continuous variable and not restricted to integer values. This can be
puzzling at first, but it is a common trick in economic theory. It allows the problem to be formulated as
equations rather than as a difficult (or impossible) to solve integer programming problem.

Figure 11.1
Cost
mc
fc
ac mc
X
= +
X
Figure 11.2
`
Y
Ua
ac pa
Y1
fc
Y
X
A
X0 X1
Y0
Figure 11.3
`
Y
A
Ua
pa
p*
U*
Y
X
T
Figure 11.4
`
Y
p*=ac*
Y1
Y0
exit
T
U*
Y
X
Ua
A
p=ac

Chapter 12
INCREASING RETURNS AND IMPERFECT COMPETITION II:
MONOPOLISTIC COMPETITION
9.1 Trade and gains from trade through increased product diversity
The previous chapter introduced economies of scale and imperfect competition as a determinant
of trade and gains from trade. We concentrated on a homogeneous-good industry (firms produce
identical products) and the idea that scale economies limit the number of firms in the market. This
limitation leads to imperfect competition in equilibrium, with products marked up above marginal costs.
Firms interact strategically with one another and the consequence of this is that the effective enlargement
of the market following the opening of trade leads to pro-competitive gains from trade. Firms move down
their average cost curves and consumers benefit from lower prices for the same goods.
In this Chapter, we are again going to look at increasing return to scale and imperfect competition
but from a different point of view. We are going to assume that firms produce differentiated products but
that the market can support a relatively large number of firms such that there is minimal strategic
interaction among the firms. Instead of trade resulting in pro-competitive gains, the same products at
lower prices, trade will result in a greater variety of products at the same prices and this will raise
consumers’ welfare.
These assumptions, that products are different but that there are many firms and minimal strategic
interaction, are typically referred to as monopolistic competition. Monopoly refers to the fact that each
firm produces a somewhat different product and hence will have influence over its market price even if
there are literally hundreds of firms. Competition refers to the fact that there are sufficiently many firms
such that there is little strategic interdependence among them.
The following preferences (utility function) are typically used to introduce monopolistic
competition and are generally referred to as Dixit-Stiglitz (1977) preferences. This approach is also
sometimes referred to as “love of variety” for a reason that should become clear. Let Xi refer to one good
in a set of n goods. Ignoring any other sectors for the moment, utility is given by
1
(12.1)
Many of you will recognize this as a CES function that we introduced earlier in the book, where F gives
the elasticity of substitution among the varieties. It is in fact a special case in which all goods carry the
same weighting in producing utility, and in which the elasticity of substitution is restricted to values
greater than one. The latter assumption means that indifference curves intersect the axis and so positive
utility can be derived from a subset of goods. Indeed, the whole point of this approach is that only a
subset of goods gets produced in equilibrium and that the number of goods available is endogenous.
The individual goods in (12.1) are said to be symmetric but imperfect substitutes. Symmetric
means that they all have the same weight in producing utility as just noted and hence a consumer is
indifferent between one apple and one orange. However, they are also imperfect substitutes meaning that
variety is valuable: a consumer would rather have one apple and one orange than either two apples or two
oranges. To see this last point, assume that each good that is produced is produced in the same amount,
so the that the summation in (12.1) is just n, the number of goods, time X, which we will term the quantity
of a “representative good”.

2
(12.2)
We see from (12.2) that utility has constant returns in the amount of each good consumer: double
each X and utility doubles. But we also see that utility has increasing returns to scale in the range of
product variety (henceforth just “variety”). Let n0 and X0 denote the number of goods and the
representative quantity produced initially, and U0 the initial level of utility. Then suppose that the number
of goods doubles to 2n0 but the quantity consumed of each falls in half to X0/2. Utility is then given by
(12.3)
Equation (12.3) shown that welfare improves when the consumer has half as much of each of twice as
many goods. Hence the term “love of variety”.
Some intuition is provided in Figures 12.1 and 12.2, where we assume that each of two varieties
are produced with increasing returns: a fixed cost and a constant marginal cost as in the previous Chapter.
Consider first Figure 12.1, where the production frontier . It may be that in autarky, a country
may wish have variety even though it is expensive in terms of having to pay the fixed costs for both
goods, and hence prefers the autarky outcome shown as Xa = Da in Figure 12.1. Trade with the second
identical country can then allow each country to specialize in one of the goods, trading half of its output
for half of the output of the other country’s good. Then both countries can consumer at point D* in
Figure 12.1.
Figure 12.1
On the other hand, the high fixed costs and the sacrifice of scale economies may mean that it is
better to produce and consumer just one good in autarky, which is the situation shown in Figure 12.2.
Here the country achieves utility level Ua in autarky by producing either X1 or X2 but not both. This is a
higher level of welfare than producing both goods: the added variety is not worth the sacrifice of quantity.
Now let the two identical countries get together, with each specializing in one of the goods. Now they
can trade to the common consumption point D* in Figure 12.2, achieving a gain from trade.
Figure 12.2
Note the difference in the source of gains from trade between Figure 12.1 and 12.2. In the former
case, the consumer gets more of the same goods, the source of gains in the previous Chapter. In the case
of Figure 12.2, the consumer actually gets less quantity of a given good, but enjoys more variety. This is
in fact exactly the outcome explored in equation (12.3). It is crucial to note in Figure 12.2 that trade does
not result in increased output of any good that is produced initially. No firm moves down it average cost
curve, there is no increase in firm scale. Nevertheless it is indeed scale economies that are responsible for
the welfare gains in that scale economies limit the number of goods produced initially.
To press this last point a little further, note that if there were no fixed costs and all goods were
produced with constant returns to scale, that the optimum under Dixit-Stiglitz preferences would be to
have infinitely many goods produced in infinitely small quantities. This is just a logical extension of the
argument behind equation (12.2). Increasing returns to scale make diversity costly and hence limit the
range of goods in equilibrium.
Figures 12.1 and 12.2 illustrate what can happen, but they by no means prove that this is what
will happen in a market environment characterized by imperfect competition.

3
12.2 A more formal approach to Dixit-Stiglitz and love of variety
We will assume that there are two sectors: sector X is composited of firms producing
differentiated goods as above, and sector Y produces a homogeneous good with constant returns to scale.
We will assume a very simple factor market structure much the same as in the previous chapter. There is
only one factor of production which we will call labor, and we will use this as numeraire assigning a
value of one to the wage rate, w = 1. We will assume that the consumer has Cobb-Douglas preferences
between Y and X, and CES preferences among the X varieties. Much of what follows is based in
Krugman (1979, 1981) and Helpman and Krugman (1985).
Total income is given by L when the wage is chosen as numeraire. We are also going to assume
that all potential X varieties have the same cost function. This is a common assumption that, when
combined with symmetry in demand, gives us the result that any good that is produced is produced in the
same amount and sells for the same price. Henceforth X and px will denote the price of a representative
good which are the same for all goods actually produced. The utility function and the budget constraint
for the economy are given by:
(12.4)
If you solve the optimization problem, the consumer’s demands for X varieties and Y are
(12.5)
The demand response for a given variety in response to a change in its own price is a bit complex,
since the variety’s own price appears both as the first term on the right-hand side of the second equation
of (12.5) but also appears in the summation term inside the square brackets. Thus the derivative of the
demand for X with respect to its own prices must be found by using the differentiation of a product rule.
However, it can be shown (and we will not do so here) that the effect of a change in a firm’s price on the
summation term in square brackets become extremely small as the number of varieties (firms) n becomes
large. As a consequence, most work in this area assumes that an individual firm is too small to affect the
summation term in (12.5), an assumption known as “large-group monopolistic competition. Assuming
that the term in brackets in (12.5) is viewed as a constant by an individual firm, the price elasticity of
demand for an individual goods is given simply by F, the elasticity of substitution among the X goods.
Referring back to our derivation of marginal revenue in the previous chapter, the markup takes on the
very simple formula 1/F. The elasticity of demand and marginal revenue are then given as follows.
(12.6)
Turning to production, marginal cost for Y, marginal cost for X, and fixed costs of an X variety
are denoted by mcy, mcx, and fcx respectively. The full general equilibrium model for a single economy is
given by a set of inequalities with associated variables as described back in Chapter 4. First, there is are
pricing equations for the Y industry and for each X variety. Second, there is a zero profit condition for
each X variety, which is typically written as markup revenues equal fixed costs instead of the longer
equation for revenues equal total costs. It is useful to think of fixed costs as a produced good, such as
factor and equipment, hence there is a pricing equation for factories (fixed costs). These three pricing
inequalities are given as follows:
Inequality Definition Complementary
Variable

4
pricing for Y Y (12.7)
pricing for X X (12.8)
pricing for n (free entry) n (12.9)
Then there are three market-clearing conditions, which require that supply equal demand (strictly
speaking supply is greater than or equal to demand). There is demand and supply for Y, for total X
production, and for labor. These equations follow from (12.5) and are as follows.
demand/supply Y py (12.10)
demand/supply X varieties px (12.11)
demand/supply L w (12.12)
This model can be solved analytically due to the powerful advantages of the large-group monopolistic-competition
assumption. Equations (12.7) and (12.8) can be solved for both X and px. Then these
solution values can be used in (12.11) to get n. The solution values are:
(12.13)
Note from the first equation of (12.13) that the output of any good that is produced is a constant
and that from the second equation that any expansion in the economy creates a proportional increases in
variety n. Let X/L, the consumption of a representative variety per capita, be given by C. Then note from
the last equation of (12.13) that nC is a constant:
(12.14)
Figure 12.3 plots n against C, and equation (12.14) is shown as a negatively-sloped curve in this
Figure. Next, note from the second equation in (12.13) that n depends only on L and fixed parameters.
Thus we show a second curve which is just a horizontal line in Figure 12.3 which gives the fixed value of
n for a given value of L. The intersection of these two curves gives the number of varieties and
consumption of a representative variety per capita. For a single economy, the outcome could be shown
by variety level n0 and variety consumption per capita by C0.
Figure 12.3
Now repeat our usual experiment: put two identical countries together in free trade. This is
represented by simply letting L double. The nC curve in Figure 12.3 does not shift as shown by (12.14),
but the n curve shift up in proportion to L, doubling in value. If n doubles, then from (12.14) C must be
cut in half. The new values in the open economy are C1 and n1 in Figure 12.3. Note that this is exactly
what is analyzed in (12.3) and suggested in Figure 12.2 above. Equations (12.7) and (12.10) will show
that consumption per capita of Y remains unchanged after trade, and hence welfare increases in each
country due to the variety effect.

5
12.3 Monopolistic competition in specialized intermediate inputs (basics idea, then
algebra after (12.20) can be skipped without loss of continuity)
The basic idea behind love of variety has also been applied to intermediate inputs, starting with
Ethier (1982). This has in turn been applied in a number of different context including endogenous
growth models, technology transfer through trade, and trade in producer services. Here we will analyze a
much simplified Ethier model, in particular as extended to consider trade in final goods only versus trade
in intermediates in Markusen (1989).
Suppose that there are two final consumption goods, X and Y, which are homogeneous and
produced with constant returns to scale by competitive firms. Utility of the representative consumer is
given by
(12.15)
There is a factor of production labor, L, which is in fixed supply. In addition, we assume a
sector-specific factor K in the Y sector. The purpose of K is to generate a concave (bowed out) production
fontier as we will discuss shortly. Good X is assumed to be costlessly assembled from differentiated or
specialized intermediate goods Si in a Dixit-Stiglitz fashion. The two production functions are given as
follows, where F is the elasticity of substitution as derived earlier in the book.
(12.16)
Each Si is produced with increasing returns to scale, consisting of the constant marginal cost and
fixed-cost technology that we have now used many times. To reduce notation, one unit of S requires a
single unit of labor. Labor requirements in S goods and the total labor supply constraint are given as
follows, where n is the (endogenous) number of intermediates.
(12.17)
Since each S enters (12.16) symmetrically and each has an identical technology, we can anticipate
the result that any S that is produced is produced in the same amount and sells for the same price as any
other S. Let superscript “a” denote a situation in which only the final X and Y goods can be traded and na
the number of and Sa the amount of each S good in the “a” equilibrium. The X technology reduces to
(12.18)
Now again do our standard experiment where we put two identical economies together in trade.
In order to illustrate the main idea, hold the amount of each S good produced constant and assume the
number produced in each country constant. The number of intermediate goods used in each country in
free trade in intermediates is double the number in autarky (n* = na), with the total output of each shared
evenly between the two countries. Output of X in each country is now given by
(12.19)
Simplifying the right-hand side, we can compare X output with intermediates trade to output under trade
in final goods only given in (12.18).

6
(12.20)
Allowing trade in intermediates increases productivity in X production as X producers now have access to
a larger range of specialized intermediates, a greater division of labor.
Results are shown in Figure 12.4, where we assume that the diminishing returns to the fixed
factor K in the Y sector outweigh the increasing returns to scale in the X sector, so that the production
frontier is concave. The frontier through A gives the frontier when only final goods can be traded. Both
countries are identical by assumption, and so there are no gains from trade: countries could trade but there
is no benefit from doing so. Point A could represent the equilibrium production and consumption point
for each country under goods trade only (the non-tangency of the price ratio with the frontier will be
treated shortly).
Figure 12.4
If we do allow trade in intermediate goods (which could include specialized services), then the
production frontier shifts to the one passing though F in Figure 12.4. Each country exports half of each of
its inputs for half of each of the other country’s inputs. With free trade in intermediatates, production
and consumption for each country could be at a point like F in Figure 12.4.
As noted earlier, this formulation of the monopolistic competition model in a manner reminiscent
of Adam Smith’s division of labor has been used in a number of contexts including endogneous growth
theory and the liberalization of trade in producer services. We now look at the issue of optimality of the
market outcomes in this model; this is a somewhat more esoteric issue and the remainder of the section
may be skipped by some readers. Results of this are equally applicable to the more standard final-goods
model treated in the previous two sections.
Suppose that the economy faces a fixed price of X relative to Y, denoted p. Then the optimal
number of intermediates and the output level of each can be found by maximizing the value of final
output of X minus input costs (representing the opportunity cost of labor in producing Y.
with respect to n and Si, for all i (12.21)
The first-order condition with respect to Si is given by applying the chain rule.
(12.22)
The first-order condition for the number of goods is given by the effect of adding one more n.
(12.23)
If we solve these two equations in two unknowns, we get the optimal output of any intermediate
that is produced. Note that w is just the marginal product of labor in the Y sector, Gl. From (12.22), we
can also get the optimality condition for a tangency between the price ratio and the marginal rate of
transformation along the production frontier.

7
(12.24)
Now turn to the outcome under a market solution. The price of an individual S is the value of its
marginal product in producing X. This is in fact given in the first term on the right-hand side of the first
equation in (12.22). Let r denote the price of an individual S.
(12.25)
In the tradition of large-group monopolistic competition discussed above, assume that each individual S
producer views q in (12.25) as fixed. Then each S producer maximizes the following expression with
resepect to Si viewing q as fixed.
(12.26)
The first-order condition is given by
(12.27)
Secondly, the free entry condition to determine n is that each S producer’s profits are zero.
(12.28)
Solving (12.27) and (12.28), we get the market equilibrium amount of any S that is produced.
Second, substituting the expression for q in (12.25) back into (12.27), we can get the relationship between
the competitive price ratio and the marginal rate of transformation given in (12.24). These are given by
(12.29)
Comparing the optimum in (12.24) to the market outcome in (12.29) we see that any S that is
produced is produced in the optimal amount. But we also see that the market outcome involves a
distortion between the price ratio and the marginal rate of transformation, much as in the case of a tax, or
the external-economies case of Chapter 10, or the oligopoly outcomes in Chapter 12. This is the
distortion between the price ratio and the slope of the frontier shown in Figure 12.5.
The conclusion of this exercise is that the market outcome is not first best: it produces the optimal
output of any good that is produced, but too few goods are produced. The intuition behind this is
essentially an externality argument. (12.28) gives the private profits of entry. But when one firm enters,
it increases the productivity of every other firm holding prices constant. This is seen in (12.25): the value
of the marginal product of an additional unit of S in producing X (r) is increasing in n, the division of
labor. This effect is not considered by an individual firm in its entry decision and hence there is a positive
externality in the X sector. Note that when (12.28) holds with equality (the private profits from entry are
zero), the “social” marginal product of an addition S given in (12.23) is strictly positive. There is thus a
close analogy here between the present result and that in Chapter 10 in the section on production
externalities.

8
12.4 The ideal variety approach to product diversity
Gains from trade through product differentiation can be looked at in a second way as well. While
consumers may prefer diversity as just noted, consumers themselves may also have different tastes.
Consumers may, for example, buy only one automobile each, but they have different views as to what is
the "ideal" automobile for their tastes and income level. This approach to product diversity is thus
labelled the "ideal variety" approach (Lancaster, 1980). Due to scale economies, no country can afford to
produce a unique automobile for each consumer. Germany produces Volkswagens and Mercedes, and
France produces Renaults and Peugeots, all of which have somewhat different characteristics from
consumers' points of view. Trade in automobiles then occurs between France and Germany due to the fact
that some Germans prefer Renaults or Peugeots and some Frenchmen prefer Volkswagens or Mercedes.
This situation is shown in Figure 12.5. Suppose that automobiles have only two characteristics:
C1 and C2 (e.g., size and speed). There is a trade-off between these two characteristics such that if one
wants a bigger car he or she must sacrifice some speed if the two are going to cost the same (e.g.,
Mercedes versus Porsche). Figure 12.5 shows three possible combinations of C1 and C2, denoted X1, X2,
and X3, corresponding to three different types of cars. Suppose that all three models could be produced at
the same average cost for the same volume of production and that at this common cost, the amounts of C1
and C2 embodied in the cars are given by points in Figure 12.5. The straight line through
these points represents a sort of iso-cost line at a common scale of production.
Figure 12.5
Now suppose that we have two groups of consumers with distinct tastes (identical within each
group). Consumer type-1 has a relative preference for characteristic C1 and consumer type-2 has a
relative preference for characteristic C2. On the isocost line shown in Figure 12.5, an indifference curve
for consumer type-2 is tangent to the isocost line at point and an indfference curve for consumer type-
1 is tangent at point . X2 is then referred to as consumer type-1's ideal variety and X1 as consumer
type-1's ideal variety.
Variety X3 could be referred to as a compromise variety. Note from Figure 12.5 that for X3 to be
equally attractive to our two consumer types, more “stuff” (e.g., stereo, air conditioning) would have to be
added to it: to make the consumers indifferent between the compromise variety and their ideal varieties,
we would have to offer , which lies outside the isocost line. The proportional difference
in Figure 12.5 is sometimes called the compensation ratio: the proportional amount of
extra “stuff” needed to make the compromise variety as attractive as the ideal varieties.
If there were constant returns to scale, the problem would be trivial: each consumer would get
there ideal variety, no one would even buy the compromise variety because it would cost more.
However, the problem is far from trivial with increasing returns to scale. The production scale in
producing the compromise variety is twice as large as in giving each consumer their ideal variety. Thus,
adjusting for the benefits of larger production scale, variety , , may in fact cost less than variety or
.
Suppose that there is only a single factor labor, L, with a wage one. We assume, as is typical in
this approach, that each consumer wants only one unit of the good but of course prefers their ideal
variety. There are fixed and marginal costs for each variety. mc0 is the labor needed for on unit
of and mc1 is the labor need for one unit of . Labor requirements are then

9
(12.30)
The relevant question is whether or not two ideal varieties (two times the first equation of (12.29)) costs
more than two units of the compromise variety. The compromise variety is preferred if
(12.31)
The important point is that the compromise variety means incurring the fixed costs only once versus twice
if each group gets its ideal variety.
Suppose that the inequality in (12.30) “marginally” holds, meaning that the left-hand side is less
than the right-hand side by a very small amount. Then the compromise variety is preferred. Consider
then our now-familiar experiment of putting two identical countries together. To give every consumer
one unit, the total world labor requirements for ideal varieties are now twice the marginal costs shown in
(12.30) and similarly for the compromise variety: no additional resources are needed for fixed costs.
Total requirements (costs) and the condition for the compromise variety to be preferred are now
(12.32)
If (12.30) holds marginally, the inequality in (12.31) will not hold. It will be reversed and the ideal
varieties will be preferred.
As in the love-of-variety approach, we see here that the ideal-variety approach means that there
are gains from trade through increased product variety. The nature of those gains are different here. The
idea is that trade allows each consumer type to get a product closer to their ideal variety embodying their
ideal characteristic combination. In more general models with many consumer types and goods, the
welfare statement in these models is typically that “on average” consumers get varieties closer to their
ideal type.
In a world of imperfect competition that accompanies scale economies, we have to be cautious
and note that just because something is preferred doesn’t mean it is going to happen. These ideal-variety
models get quite mathematically complicated, much more so than the simple love-of-variety model. For
example, there may be a continuum of consumers who preferences are “located” at difference points of a
line or a circle. Such models are sometimes referred to as “location” or “address” models of product
differentiation, but these embody the same basic ideas: a consumer’s location or address on the circle is
the consumer’s ideal variety (see Helpman, 1981). These models have been perhaps more used in
industrial-organization economics. The problem for international trade is that they quickly get very
difficult in general-equilibrium settings.
12.5 Some useful algebra for Dixit-Stiglitz (may be skipped without loss of continuity
but useful later for analyzing trade costs)
This section lays out the algebra required to solve for the demand function and price index
(defined and discussed shortly) for the Dixit-Stiglitz utility function. In order to simplify it a bit, let’s
assume that there is just a single section producing differentiated X varieties. But there is a
straightforward extension to a two-sector model in which the other sector, Y, has constant returns to scale
and perfect competition, and there is Cobb-Douglas substitution between the X goods and good Y. In this
situation, the consumer always devotes constant share of income to each sector, so we will start with an
allocation of income Mx to the X sector.
Let Xc denote the utility derived from the X varieties; Xc is sometimes referred to as a composite

10
commodity. Xc does have a price associated with it. This is a price index, the minimum expenditure
necessary to buy one unit of composite good Xc. We will denote this price or rather price index as ex.
The value of Xc (the utility from X consumption) is defined as follows.
The consumer maximizes utility subject to a budget constraint. The maximization problem is written as a
Lagrangean function and the first-order condition for an arbitrary good Xi are given as follows.
(12.33)
Divide the first-order condition in (12.33) for an arbitrary good i by the corresponding condition
for good j.
(12.34)
Now perform several steps. (1) write the second equation of (12.34) with Xj on the left. (2)
multiply both sides of this equation by pj. (3) sum this equation over all goods j. These three
steps are
(12.35)
Inverting this last equation, we have the demand for an individual variety i:
(12.36)
Now we can use Xi to construct Xc and then solve for ex, the price index, noting the
relationship between  and . First, raise (12.36) to the power $.
(12.37)
Now sum over all of the i varieties of X.
(12.38)

11
Now raise both sides of this equation to the power 1/$ to get the composite commodity demand
for Xc.
(12.39)
The demand for Xc must be the expenditure Mx on Xc, divided by the price index ex. Examining
the last equation of (12.39), this means that the price index is given by:
(12.40)
Again, the price index ex, also called the expenditure function, is the minimum cost or
expenditure necessary to buy one unit of the composite commodity Xc. If all of the X varieties
sold for the same price, the expenditure function (price index) in (12.40) simplifies as follows.
(12.41)
Note that the price index is homogeneous of degree one in the prices of the individual
goods: if we double all prices we double to cost of buying one unit of Xc. However note that the
price index is decreasing in the number of varieties available (1 - F < 0). The is another way of
thinking about the love-of-variety effect. The same utility derived from two apples or two
oranges might be derived from (for example) 0.8 apples and 0.9 oranges. When more variety is
available, the consumer can achieve the same utility (same value of Xc) by actually reducing total
expenditure.
Finally, having derived e, we can then use equation (13) in (9) to get the demand for an
individual variety.
(12.42)
We now move onto a chapter introducing the existence of trade costs. The price index in (12.40)
and the demand function in (12.42) will prove very useful in discussing trade costs in the context of
monopolistic-competition models in the next chapter.
12.6 Summary
This is the second of two chapters on trade with increasing returns to scale and imperfect
competition. The first (Chapter 11) focused on a pure case in which firms produced identical goods but
scale economies limited the number of firms such that individual firms took into account their strategic
interactions with other firms in their output decisions. A principal result is that the larger economy
created through trade offer welfare gains through pro-competitive and production-scale effects. This
Chapter focused also on a pure case in which firms produce somewhat different goods and there are a
large number of firms such that each views the decision of rival firms as exogenous. A principal result is
that the larger economy created through trade offers welfare gains through increased product or

12
intermediate-good diversity: more products at the same prices rather than the same products at lower
prices in Chapter 11. Of course, the two approaches can be combined, but the analysis becomes more
difficult and is left to more advanced treatments (e.g., Melitz and Ottaviano (2008)) and indeed the ideal
variety approach does involve variable markups and pro-competitive effects.
The two main approaches to product diversity are considered. The “love of variety” approach
assume an endogenous set of symmetric but imperfectly substitutable products. Consumers are rewarded
by a more diverse consumption bundle through trade. The “ideal variety” approach works rather
differently. Consumers are assumed to differ in their views as to the ideal product, a product being a
bundle of characteristics. Often in this approach, the consumer is assumed to buy just a single unit or
nothing. As in the love-of-variety approach, diversity, in this case giving each consumer their ideal
product, is costly in the presence of scale economies and so consumers accept compromise varieties in
small autarky markets. The rewards to trade are that consumers, on average, get products closer to their
ideal varieties.
While much of the literature has focused on final goods, an important variation of these models
considers the differentiated products to be intermediate goods used in producing homogeneous final
goods. Allowing trade in intermediate goods offers final producers higher productivity by increasing the
division of labor. This ideal is a cornerstone of what is referred to as endogenous growth theory (e.g.,
Romer 1987) and has also been applied to trade in components and in producer services.

13
REFERENCES
Dixit, Avinash K. and Joseph Stiglitz (1977), “Monopolistic Competition and Optimum Product
Diversity”, American Economic Review 67, 297-308.
Ethier, Wifred J., (1982), "National and International Returns to Scale in the Modern Theory of
International Trade", American Economic Review 72, 389-406.
Helpman, Elhanan, (1981). "International Trade in the Presence of Product Differentiation, Economies of
Scale and Monopolistic Competition. A Chamberlinian, Heckscher-Ohlin Approach." Journal of
Helpman, E. and P. Krugman (1985). Market Structure and Foreign Trade, Cambridge: MIT press.
Krugman, P. (1979). "Increasing Returns, Monopolistic Competition, and International Trade." Journal of
Krugman, P. (1981), "Intraindustry Specialization and the Gains from Trade", Journal of Political
Economy 89, 469-479.
Lancaster, K. (1980). "Intra-Industry Trade Under Perfect Monopolistic Competition." Journal of
Markusen, J. R. (1989). Trade in Producer Services and in Other Specialized Intermediate Inputs."
American Economic Review 79, 85-95.
Melitz, Mark J. and Gianmarco I.P. Ottaviano (2008), “Market size, trade, and productivity”, Review of
Economic Studies 75, 295-298.
Romer, Paul M. (1987), “Growth Based on Increasing Returns to Scale due to specialization”, American
Economic Review 77, 56-62.

Figure 12.1
X2
U*
Ua
D*
2 X
2 X ¢
X1
Xa=Da
1 X
1 X ¢
Figure 12.2
X2
a a X = D
U*
Ua
D*
2 X
2 X ¢
2 2
a a X = D
X1
1 X
1 X ¢
2 2
Figure 12.3
n
n1
n0
1 0 L = 2L
C
C
C1 C0
Figure 12.4
Y
A
c
p > pB = MRT
F
X
p*
pa

Figure 12.5
C2
U2
U1
0
2 X
0
1 X
0
3 X
1
3 X
C1

Chapter 16
MULTINATIONAL FIRMS AND FOREIGN DIRECT INVESTMENT
1
16.1 Stylized facts, basic concepts
Multinational firms have become a crucial element in the modern world economy, and a few
statistics are presented in Table 16.1 to document this. The first number documents that fact that sales of
foreign affiliates of multinational corporations are double the total value of world trade. Other number
we have seen range up to a factor of five. The second number notes that value added in multinational
affiliates is about 11 percent of world GDP (itself a value-added measure). The third number notes that
one-third of all world exports originate in the foreign affiliates of multinational firms. The final number
actually surprises some folks as being rather small: foreign affiliate exports are only 18 percent of their
total sales. Most of the output of foreign affiliates is sold locally, a point that we very much need to make
sense of.
Table 16.1
Several decades ago, foreign direct investment (FDI) was viewed as simply a capital movement.
Consistent with the dominant theory of the day, Heckscher-Ohlin theory, capital should flow from
capital-rich high-income countries to poor capital-scarce countries. During the 1980s, it became very
obvious that the old view was at best inadequate and at worst simply wrong
(A) FDI flows primarily from high-income developed countries to other high-income
countries, not from capital-rich to capital-poor countries.
Table 6.2 gives some evidence on this. The top panel gives flows of new FDI for 2007,
expressed as the developed countries’ share of total flows, in both inward and outward directions. These
figures are followed by the corresponding shares for the stock of FDI in 1990 and 2007. Clearly, the
developed countries are the major source of FDI, but what is less appreciated is that they are also the
major recipients of FDI.
Table 16.2
(B) Affiliate production is primarily for local sale and not for export back to the parent
country.
Another myth that persists in the popular press though I hope not in the economics literature is
that multinationals move production to poor countries to pay low wages and export the output back home.
Table 6.3 breaks down sales of foreign manufacturing affiliates of US, Japanese, and Swedish parents into
local sale and export sales, and imports are also listed. Note that two-thirds or more of foreign output of
affiliates is sold locally. Table 16.4 has data from a more restricted sample for the US and Sweden. In
2003, local sales account for 60 percent of all sales, but that still leaves a large portion for export. But
columns 2 and 3 reveal that most of the exports do not go back to the parent country, they go to third
countries. The point is that foreign affiliates are not primarily in the business of producing cheaply
abroad for shipment back to the parent: they are primary in the business of producing abroad for local and

2
regional markets.
Table 16.3 Table 16.4
Several other stylized facts will be important.
(C) FDI is attracted to large markets and high-income markets
This again fits well with the idea that production is largely for local or regional sale and, if there
are fixed costs of setting up or acquiring new plants, then such investments are more likely to be observed
in large and high-income markets. If multinationals were simply in search of low wages to produce for
export back home, then we should not observe firms to choose high-income markets and there should be
little relationship to host-country size.
(D) There are high levels of intra-industry cross-investment, particularly among the high-income
countries
Much has been written about intra-industry trade or “cross-hauling” over the last several decades.
It seems to be less well known that exactly the same phenomenon is observed for affiliate sales.
Combined with earlier statistics, these numbers emphasize that a satisfactory theory of multinational firms
must be consistent with large volumes of cross-investment (intra-industry affiliate production) among
similar large, high-income countries. But it must also of course, be able to explain the fact that firms
from the high-income countries are net investors in developing countries.
The weight of these statistics also suggests that much if not most FDI is “horizontal” or “market
seeking”; that is, foreign affiliates of multinational firms are doing much the same things in foreign
countries as they are at home. The same products and services are produced, generally for local and
regional sale. “Vertical” or “resource seeking” (e.g., low cost labor) investments involve the
fragmentation of the production process into stages, with stages located where the factors of production
they use intensively are relatively cheap. While not the dominant motivation for FDI, vertical
investments may nevertheless be quite important in developing countries.
16.2 A basic organizing framework
Modern theory often begins with the premise that firms incur significant costs of doing business
abroad relative to domestic firms. Therefore, for a firm to become a multinational, it must have offsetting
advantages. A limited but very useful organizing framework for inquiring into the nature of these
advantages was proposed by John Dunning (1977). Dunning proposed that there are three conditions
needed for firms to have a strong incentive to undertake direct foreign investments.
Ownership Advantage: the firm must have a product or a production process such that the firm
enjoys some market power advantage in foreign markets.
Location Advantage: the firm must have a reason to want to locate production abroad rather than
concentrate it in the home country, especially if there are scale economies at the plant level.
Internalization Advantage: the firm must have a reason to want to exploit its ownership

3
advantage internally, rather than license or sell its product/process to a foreign firm.
Internalization is often nowadays referred to as vertical integration and the choice between and
owned subsidiary versus licensing is often now referred to as the outsourcing decision, which is the same
thing put the other way around.
The basic idea is shown in Figure 6.1. A firm wanting to produce in a foreign country for local
sale has a location decision between producing there and exporting. We can refer to this as the
“offshoring” decision. If the multinational chooses foreign production, it has the further choice between
an owned subsidiary, perhaps some sort of joint venture, and an arm’s length licensing contract. We can
refer to this as the internalization or vertical integration versus outsourcing decision, whether or not to
produce outside the ownership boundaries of the firm.
Figure 16.1
Consider first ownership advantages. Evidence indicates that multinationals are related to R&D,
marketing, scientific and technical workers, product newness and complexity, and product differentiation
(Caves 2007, Markusen 2002). This suggests that multinationals are firms which are intensive in the use
of knowledge capital. This is a broad term which includes the human capital of the employees, patents,
blueprints, procedures, and other proprietary knowledge, and finally marketing assets such as trademarks,
reputations, and brand names.
There are several reasons why the association of multinationals with knowledge-based assets
rather than physical capital is appealing. First, the services of these assets may be easily used in distant
plants, such as managers and engineers visiting those plants. Second and more subtlety, knowledge
capital often has a joint-input or non-rivaled property within the firm. Blueprints, chemical formulae, or
even reputation capital may be very costly to produce, but once they are created, they can be supplied at
relatively low cost to foreign production facilities without reducing the value or productivity of those
assets in existing facilities.
The sources of location advantages are varied, primarily because they can differ between
horizontal and vertical firms. Consider horizontal firms that produce the same goods and services in each
of several locations. Given the existence of plant-level scale economies, there are two principal sources
of location advantages in a particular market. The first is the existence of trade costs between that market
and the MNEs home country, in the form of transport costs (both distance and time), tariffs and quotas,
and more intangible proximity advantages. The second source of location advantage, again following
from the existence of plant-level scale economies, is a large market in the potential host country. If that
market is very small, it will not pay for a firm to establish a local production facility and the firm will
instead service that market by exports.
The sources of location advantage for vertical multinationals are somewhat different. This type
of investment is likely to be encouraged by low trade costs rather than by high trade costs and by factor-price
differences across countries. Low trade costs facilitate the intra-firm trade of intermediate and final
goods and thus facilitate the geographic breaking up of the production chain.
Internalization advantages (or outsourcing disadvantages) are the most abstract of the three. The
topic quickly gets into fundamental issues such as what is a firm, and why and how agency problems
might be better solved within a firm rather than through an arm's-length arrangement with a licensee or

4
contractor. Basically, it is our view that internalization advantages often arise from the same joint-input,
non-rivaled property of knowledge that create ownership advantages.
16.3 A simple monopoly model of location choice
In this section, we explore a basic model very similar to that used in Chapters 11 and 13 in order
to get the crucial intuition as to the optimal international organization of a firm. Questions of outsourcing
versus vertical integration (internalization) are left to a later section. There are two countries, home and
foreign and one monopoly firm in country h. There is a linear inverse demand for the product where the
intercept is " and slope is (1/L), L = market size. The price (pi), quantity (Xi) and market size (Li) in
market i = h,f are related as follows, where the second equation is firm revenues (Ri) in market i.
(16.1)
There is a constant marginal cost ci in market i and a specific trade cost t between markets. Profits before
fixed costs for a plant producing in market i and selling in i and a plant producing in i and selling in j are
given by
(16.2)
Taking the first-order conditions for profit maximization given the optimal levels of domestic and export
supply.
(16.3)
The firm is headquartered in country h but may chooses between three (discrete) alternatives. It
can choose a single plant at home in country h, exporting to country f and we refer to this as a national or
domestic firm (d). It can have a single plant in country f, exporting back to h, a vertical (v) structure. Or
it can be a horizontal multinational with plants in both countries (m). There is a firm-specific fixed cost F
and plant specific fixed cost G where the latter must be incurred for each plant. This model thus displays
firm-level scale economies: the total fixed costs of a two-plant firm are (F + 2G) which is less that the
total fixed costs of two independent single-plant firms, (2F + 2G). F thus represents the cost of creating a
product or process, and this knowledge is a joint or non-rivaled input across plants.
If you substitute (16.3) into the two equations of (6.2) and add in fixed costs, you will find that
the profits for each of the multinational firms three choice are given as follows (we did a derivation of this
in Chapter 13).
Profits of a national firm: one plant at home (h) exporting to f
(16.4)
Profits of a vertical firm: one plant in j exporting back to i

5
(16.5)
Profits of a horizontal firm: plants in both countries.
(16.6)
This is really a surprisingly rich little model. Figure 16.2 shows profits from the three possible
choices holding total world demand constant, but varying the size of the two markets along the horizontal
axis with the parent country h small on the left and big on the right. If you stare at these three equations
long enough and consider how each curve shifts (or does not shift) in response to some parameter change,
you will conclude that a two-plant horizontal structure is more likely as:
Both markets are large characteristic of markets
Markets of similar size characteristic of markets
Marginal costs are similar characteristic of markets
Firm fixed costs > plant fixed costs characteristic of industry
Transport/tariff costs are large geography/policy
Figure 16.2
Large markets mean that the added fixed costs of a second plant outweigh the higher variable
costs of exporting. The intuition behind the second and third results is that if one market is much larger
and/or production costs much smaller, then it pays to put a single plant there and export to the
smaller/costlier market. For the third result, note that if we raise F and lower G in the same amounts, this
increases the profits of a type-m firm while leaving the profits of a type-d or type-v firm unchanged (the
horizontal curve shifts up). Trade costs reduce the profits of a type-d or type-v firm but leave the profits
of a type-m firm unchanged (the national and vertical curves shift down).
Despite its simplicity, this model fits the data well: horizontal firms will be important between
similar, large (rich) markets in industries where knowledge capital is important (F is large relative to G).
A vertical structure is preferred to a national structure as:
Foreign market is larger
Foreign marginal cost is low
Low trade costs: vertical structure if cf < ch even if country f is very small
As noted above, if you are going to have a single plant, put it in the large and/or low-cost market.
Note however that the importance of the local market size disappears as trade costs go to zero. As t
converges to zero, the relationship between (16.4) an (16.5) is determined entirely by which is the low-cost
location, and note that the horizontal structure will never be chosen at t = 0 (G > 0). As t goes to
zero, we can say that the vertical structure is chosen if and only if country f is the low-cost location.
16.4 Monopolistic competition and the choice of exporting versus horizontal production

6
A crucial feature of the modern theory of the multinational as noted earlier is the notion of joint
or non-rivaled knowledge-based assets of firms (these are sometimes also called firm-level or multi-plant
economies of scale). Once a firm develops a product, process, or even brand identification and a
reputation for quality, it can add additional plants abroad for significantly less additional fixed costs than
is incurred for setting up the firm and first plant at home. In this section, we will consider a firm wanting
to sell abroad as well as at home, and limit its choices to exporting to the foreign market or becoming a
two-plant horizontal multinational. We will use the monopolistic-competition model developed in
Chapters 12 and 13.
To keep things simple, we will consider identical countries. Assuming in addition that marginal
costs of production is the same for both domestic and multinational firms, the pricing equation in the
model says that all varieties will have the same (domestic) prices in equilibrium, pi = pj, regardless of
whether they are produced by a locally-owned firm or by a branch plant of a foreign multinational.
Similarly, imported goods will have the same prices in each country. Let superscript d denote a domestic
(one plant) firm and superscript m denote a two-plant horizontal multinational. As in Chapter 13, N = t1 - F
is the “phi-ness” of trade where t is the (gross) iceberg trade cost: N = 1 is free trade and N < 1 means
positive trade costs. Let the first subscript on X denote country of firm ownership and the second the
country of sale. The demand functions for the various X varieties (domestic production and imports) are
given by:
(16.7)
Zero-profit conditions for d and m firms located in country i are markup revenues equal fixed
costs. Let fcd denote the fixed costs of a first plant and $fcd denote the fixed costs of a two-plant
multinational. The non-rivaled or multi-plant economies idea implies that 1 < $ < 2: a two-plant firm has
less than double the fixed costs of a one-plant firm. The zero-profit conditions for one and two-plant
firms are then
(16.8)
(16.9)
Using the demand functions for Xii and Xij above, these are:
(16.10)
(16.11)
Suppose that we pick values of parameters such that national and multinational firms can both
just break even in the two identical countries: (16.10) and (16.11) both hold with equality. Then dividing
the first equation by the second give us the critical relationship between trade costs and fixed costs for
indifference.
(16.12)

7
Lower trade costs (higher N) must be balanced against higher firm-level scale economies ( lower $) for
firms to be indifferent as to type. To put it differently, higher trade costs encourage firms to be
multinationals and higher firm-level scale economies do the same. In line with our earlier discussions,
we can think of high firm-level scale economies as being associated with knowledge and R&D intensive
industries. Our results clearly have empirical predictions about what sort of firms should be
multinationals and which industries should be dominated by multinationals. Tests of these ideas have
confirmed the basic theory and we will discuss these later.
16.5 The knowledge-capital model
A recent development that integrates both horizontal and vertical motives for multinationals is
Markusen’s knowledge-capital model. The full definition of the model is found in Markusen (2002). The
knowledge-capital model is a general-equilibrium approach that incorporates both horizontal and vertical
motives for multinationals. The configuration of firms that arises in equilibrium depends on country
characteristics (size, relative size, and relative endowments), industry characteristics (firm versus plant-level
fixed costs or scale economies) and trade costs.
There are two goods, X and Y and two factors of production, skilled and unskilled labor, S and L
There are two countries i and j. Y is produced with constant returns by a competitive industry and
unskilled-labor intensive. X is produced with increasing returns by imperfectly competitive firms. There
are both firm-level and plant-level fixed costs and trade costs and firm-level fixed costs result in the
creation of “knowledge-based assets”.
There are three defining assumptions for the knowledge-capital model.
(A) Fragmentation: the location of knowledge-based assets may be fragmented from production.
Any incremental cost of supplying services of the asset to a single foreign plant versus the cost to
a single domestic plant is small.
(B) Skilled-labor intensity: knowledge-based assets are skilled-labor intensive relative to final
production.
(C) Jointness: the services of knowledge-based assets are (at least partially) joint (non-rivaled) inputs
into multiple production facilities. The added cost of a second plant is small compared to the cost
of establishing a firm with a single plant.
There are three possible firm “types” that can exist in equilibrium in either country (so six firm
types in all), and there is free entry an exist into and out of firm types.
Type m - horizontal multinationals which maintain plants in both countries, headquarters is located
in country i or j.
Type d - national firms that maintain a single plant and headquarters in country i or j. Type di
firms may or may not export to the other country.
Type v - vertical multinationals that maintain a single plant in one country, and headquarters in
the other country. Type vi firms may or may not export back to their headquarters

8
country.
Various assumptions can be make about factor intensities and they do make some quantitative
difference to the results. The ones used to generate the diagrams attached below assume that the skilled-labor
intensity of activities are
[headquarters only] > [integrated X] > [plant only] > [Y]
When countries are similar in size and in relative factor endowments, the model predicts
that horizontal multinationals will be important. Firms will build plants in the foreign country to serve
the local market instead of incurring trade costs on exports. We discussed this in the previous section for
identical countries, but now need to comment on this issue of size and endowment similarity. The
intuition about the role of these characteristics is understood by considering what happens when countries
are dissimilar in size or in relative endowments.
To a good degree, the intuition is seen in Figure 16.2 above. When one country is quite large
and the other quite small, a national firm locating in the large country will have an advantage since is
spends little in trade costs whereas a horizontal firm still has to spend a lot on a second plant regardless of
the size of the market. Similarly, if the countries are similar in size but one country is skilled-labor
abundant, a vertical firm locating in that country has an advantage over the other two types: it can locate
the headquarters in the skilled-labor-abundant country and the single plant in the unskilled-labor-abundant
country. Note that this advantage for the vertical firm is reinforced if the skilled-labor-abundant
country is also small. The vertical firm has the added advantage that locating its single plant in the large
market abroad economizes on trade costs.
Figure 16.3 plots the results of a simulation solving for the number, types, and production of
firms over a world Edgeworth box. The world skilled-labor endowment is on the axis running toward the
northwest and the unskilled-labor endowment on the axis running to the northeast. The total equilibrium
volume of affiliate production is shown on the vertical axis, where affiliate production is defined as the
output of foreign plants of horizontal and vertical multinationals (i.e., the output of a horizontal firm’s
home plant is not counted). In the center of the box where the two countries are identical, exactly half of
all world X output is affiliate output: all firms are horizontal multinationals and their home and foreign
plants are identical, so half of each firm’s output is affiliate output.
Figure 16.3
But affiliate sales can be even higher and this occurs in Figure 16.3 when one country is both
small and skilled-labor abundant. In this case most or even all firms are vertical multinationals with their
headquarters in the small, skilled-labor-abundant country. But this in turn means that all (or virtually all)
plants and all production are in the other (large) country, and so all of world output is classified as
affiliate output and sales.
Figures 16.4 and 16.5 help sharpen our intuition by showing two restricted versions of the model.
Figure 16.4 eliminates the possibility of vertical firms (the first plant must be located with the
headquarters) and so only national firms and horizontal multinationals can exist. The consequence is we
should observe multinationals when the countries are similar in size and in relative endowments. Figure
16.5 does the opposite, ruling out horizontal multinationals by eliminating firm-level scale economies: the
fixed costs of a two-plant firm are double the fixed costs of a single-plant firm. In this case, the is no

9
multinational activity between identical countries and activity is maximized when one country is both
small and skilled-labor abundant.
Figure 16.4 Figure 16.5
A nice thing about Figures 16.3-16.5 is that they provide clearly testable hypotheses, and further
hypotheses follow from the size of trade costs relative to firm and plant-level scale economies. We will
discuss this later in the chapter, but the consistent empirical finding is that the data give far more support
to the horizontal model. The purely vertical case in Figure 16.5 is overwhelming rejected and the world
looks much more like Figure 16.4 than 16.5: multinational activity is concentrated among similar, high-income
countries. Multinational activity and investment from high income to low income countries is of
much less importance, especially relative to what seems to be the popular impression.
16.6 Outsourcing versus internalization (vertical integration)
The final node in the decision tree shown in Figure 16.1 concerns the choice to maintain
ownership of a foreign production facility or to outsource / license a foreign firm to produce for the
multinational. While this is an old question in the international business literature, it has been only in the
last two decades that it has attracted interests from international trade economists. There are two principal
approaches and we will try to present simple versions here.
Some of the first formal models of the internalization decision were published in the late 1980's
and 1990's, and draw their empirical motivation from the strong association of multinationality with
knowledge-based assets such as those described in the previous paragraph (see Caves 2007 and Markusen
2002). On the one hand, these assets (or the services thereof) are easily transferred overseas, such as
providing a blueprint, chemical formula, or procedure to a foreign plant. On the other hand, the same
characteristics that make it easy to transfer these assets makes them easily learned by foreign managers,
agents, or licensees. Once the agent sees the blueprint or formula he or she could defect to produce the
product in a new firm. Knowledge is non-excludable, at least after some period of time.
About the same time as this last set of papers appeared, an important advancement to the
internalization question was being developed by several authors, and we will review and explain a key
paper by Antrás (2003). All of these authors substitute the term outsourcing for the converse term
internalization. Their approach is sometimes termed the “property-rights” approach to the firm. The new
literature combines a number of separate elements that together produce a coherent model that offers clear
empirical predictions. The first element (in no particular order) is the assumption that production requires
“relation-specific investments”, meaning that a multinational and a foreign individual or firm must incur
sunk investments prior to production that have no outside value if the relationship breaks down. The
second element is the assumption of incomplete contracting: certain things are simply not contractible or
alternatively any contract on these items is not enforceable. The assumptions of sunk investments and
non-contractibility lead to a third problem, which is ex-post “hold up”. What happens after production
occurs cannot be contracted ex ante, and so each party has some ability to negotiate ex post and to prevent
the other party from fully utilizing the output.
The first set of papers, focusing on the non-excludability of knowledge which is learned by a
local agent or manager, can be explained by a simple version of Markusen (2002). To economize on
things, we will deal only on the issue of foreign production versus exports in order to focus clearly on the

10
role of non-excludability of knowledge.
(1) The MNE introduces (or attempts to introduce) a new product every second time periods.
Two periods are referred to as a "product cycle". A product is economically obsolete at the end of the
second period (end of the product cycle).
(2) The probability of the MNE successfully developing a new product in the next cycle is
1/(1+r) if there is a product in the current cycle, zero otherwise (i.e., once the firm fails to develop a new
product, it is out of the game). The probability of having a product in the third cycle is 1/(1+r)2 etc.
Ignore discounting.
(3) The MNE can serve a foreign market by exporting, or by creating a subsidiary to produce in
the foreign market. Because of the costs of exporting, producing in the foreign country generates the
most potential rents.
(5) But any local manager learns the technology in the first period of a cycle and can quit
(defect) to start a rival firm in the second period. Similarly, the MNE can defect, dismissing the manager
and hiring a new one in the second period. The (defecting) manager can only imitate, but cannot innovate
and thus cannot compete in the next product cycle.
(6) No binding contracts can be written to prevent either partner from undertaking such a
defection. I will assume that the MNE either offers a self-enforcing contract or exports.
R- Total per period licensing rents from the foreign country.
E- Total per period exporting rents (E < R).
F- Fixed cost of transferring the technology to a foreign partner. These include physical
capital costs, training of the local manager, etc.
T- Training costs of a new manager that the MNE incurs if it dismisses the first one (i.e., if
the MNE defects).
G- Fixed cost that the manager must incur if he/she defects. This could include costs of
physical capital, etc.
Li- Licensing or royalty fee charged to the subsidiary in period i (i = 1,2).
V Rents earned by the manager in one product cycle: V = (R-L1) + (R-L2).
V/r- Present value of rents to the manager of maintaining the relationship.
The manager ("a" for agent) has an individual rationality constraint (IR): the manager must earn
non-negative rents. The manager also has an incentive-compatibility constraint (IC): the manager must
not want to defect in the second period: second-period earnings plus the present value of earning from
future products (if any) must exceed the single-period one-shot return from defecting.
(R - L1) + (R - L2) $ 0 IR a (16.13)

11
(R- L2) + V/r $ (R - G) IC a (16.14)
where V = (R- L1) + (R - L2) is the present value to the manager of the future rents, if there are any. (R -
G) is the payoff to unilaterally defecting.
The MNE similarly has an individual rationality constraint (IR): the MNE must earn profits at
least equal to the profits from exporting. The MNE also has an incentive-compatibility constraint: the
MNE must not want to defect (fire the manager) in the second period.
L1 + L2 - F $ 2E IR m (16.15)
L2 $ R - T IC m (16.16)
Combine the IC constraints.
R - T # L2 # G + V/r (16.17)
Firm's objective is to minimize V subject to this incentive compatibility. Making V as small as possible
subject to (16.17) holding gives us:
2R - L1 - L2 = V = r(R - T - G) $ 0 (rent share to the manager) (16.18)
Our first results is then that, if R # G + T, the MNE captures all rents in a product cycle,
henceforth referred to as a rent-capture (RC) contract, and the agent’s IRa constraint holds with equality.
This occurs when
(1) The market is relatively small.
(2) Defection costs for the MNE (T) are high.
(3) Defection costs for the manager (G) are high.
If R > T + G, there is no single-product fee schedule that will not cause one party to defect. In
this case, the manager's IRa constraint does not hold as a strict equality: that is, the MNE shares rents with
the manager and the amount of rent sharing is given in (16.18). This is a credible commitment to a long-term
relationship that we could think of as a subsidiary. However, it is costly for the multinational and if
it gets too costly then the multinational will choose exporting instead: dissipating some rent is preferable
to sharing a larger total. This is the inefficiency caused by the lack of contractibility of knowledge, and
may lead (will lead form many parameter values) the firm to make a welfare-inefficient choice to export
and dissipate total surplus rather than share a larger surplus with the local agent.
As noted above, the “property-right” approach works rather differently. Here I present a much
simplified version on Antrás (2003). The idea is that the firm and the local agent must make ex ante
investments that are not contractible. The multinational invests capital K and the agent invests labor L.
Ex post, they divide the surplus via the Nash bargaining solution, with the firm getting share s and the
agent the share (1-s). We cannot go through an analysis of Nash bargaining here, but many readers may
know from industrial-organization or game theory that the equilibrium share is a function of each party’s
“bargaining power” and its outside option. Antrás assumes the firm has a bargaining power parameter of
at least 1/2. Ownership is defined as a property right to anything left (e.g., an intermediate input) in the
event of bargaining breakdown.

12
Under outsourcing, denoted with the subscript “o”, any intermediate output produced is worthless
to both in the event of a breakdown. Thus the outside option of both the firm and the agent is zero even
though the agent owns what is left. Under FDI, denoted with the subscript “v” for vertical integration, the
multinational has some use for what is left and so has an outside option. If N > 1/2 denotes the
multinational’s bargaining power and * denotes the share of the total potential rent (under a successful
contract) the Nash solution gives the multinational the following shares.
(16.19)
The second equation is a common result in the bargaining literature: the firm gets its outside option, plus
its bargaining share of the total minus the sum of the outside options (the agent’s outside option is zero).
Let the prices of K and L equal one. Profits from the project are given by
(16.20)
Knowing that there will be holdup and ex post bargaining with equilibrium share s, the firm and the agent
respectively maximize the following in choosing the input they control.
Firm chooses K (16.21)
Agent chooses L (16.22)
The first-order conditions for capital (chose by the firm) and labor (chosen by the agent) are respectively
given by
(16.23)
(16.24)
The last equality in each line is just to emphasize that the agent problem here is much like having a tax on
capital of tk = (1-s)/s and a tax on labor of tl = s/(1-s). Indeed, the first-order conditions are exactly those
of a single integrated firm maximizing profits subject to these input taxes. Ex post holdup is like each
party being able to tax the other. Given (16.19), vertical integration is effectively a lower tax on capital
than outsourcing, and vertical integration is effectively a higher tax on labor. A first-best outcome (i.e.,
by an integrated single decision maker) would be to set both equations to one (the true prices of capital
and labor) rather than to something greater than one. Rearranging (16.23) and (16.24) we get
firm’s choice of K for a given L (16.25)

13
agent’s choice of L for a given K (16.26)
where the quantities on the far right of each expression are the first-best outcomes.
Antrás notes that (16.25) and (16.26) can be thought of as reaction or best-response functions for
the firm and agent respectively and the situation is shown in Figure 16.6.1 When s = 1 and (1-s) = 1 in the
two equations respectively, we have the first-best outcome in which the two reaction functions F* (firm)
and S* (agent) cross at point A. Both outsourcing and vertical integration shift in both reaction functions.
However, since sv > so, vertical integration shifts in F less (to Fv) than outsourcing does (to Fo) in Figure
16.6. Conversely, outsourcing shifts in S less (to So) than vertical integration does (to Sv) in Figure 16.6.
The vertical integration equilibrium is at point B in Figure 16.6 while outsourcing leaves us at C. Both
outcomes are inefficient, but vertical integration has a relatively less under utilization of capital while
outsourcing has a relatively less under utilization of labor.
Figure 16.6
It may therefore be intuitive that a firm in a capital-intensive industry will choose vertical
integration while a firm in a labor-intensive industry will choose outsourcing, but this requires a fair bit
more algebra to prove. The two first-order conditions can be solved to yield
(16.27)
These can be substituted back into the first order conditions to give the equilibrium inputs
(16.28)
(16.29)
The profit level for the firm in (16.20) is then given by:
(16.30)
The choice between vertical integration or outsourcing then reduces to evaluating the ratio
(16.31)
Given the assumption that , it is true that : the function
reaches a maximum value of 1/4 at s = 1/2 and falls off as s grows larger (or shrinks for that
matter). As the share on labor becomes small (the industry is very labor intensive), (6.35) reduces to

14
(16.32)
and so vertical integration is chosen by a capital-intensive firm. At a value of ( = 1/2, implying a value
of $ < 1/2, we have
(16.33)
Thus outsourcing is chosen by a labor-intensive firm. Antrás then presents empirical evidence, using
intra-firm versus arm’s-length trade, that gives good support to his model.
16.7 Summary
A number of empirical observations, particularly the fact that the high-income countries are both
the major sources but also the major recipients of multinational investment has led economists to re-think
the old idea of foreign direct investment as simply capital flows, moving from where capital is abundant
to where it is scarce (developing countries). A cornerstone of the new theory is the existence of firm-level
(or multi-plant) economies of scale arising from the joint input or non-rivaled property of
knowledge based assets. Blueprints, processes, managerial techniques and so forth are costly to create
but once created can be applied to multiple plants at very low additional cost.
Combined with trade costs and differing factor intensities across activities (e.g., skilled-labor
intensive headquarter services versus less skilled-labor-intensive production), knowledge-based firm scale
economies generate rich models and predictions about what sort of firms will arise between what sorts of
country pairs. Further, the predictions of the models are empirically testable and have generated good
support in formal econometric estimation.
In parallel with this new learning is a literature on outsourcing versus vertical integration,
previously termed internalization (the converse of outsourcing). In one strand of literature, the same
properties of knowledge-based assets that lead firms to expand abroad in the first place create difficulties
in contracting and in securing property rights over the knowledge. This can lead firm to choose vertical
integration (subsidiaries) or exporting when they would in principal like to choose outsourcing (licensing)
on the basis of cost. A more recent approach focuses on what is referred to as the property-rights theory
of the firm and to the existence of hold-up problems arising from relation-specific investments. In the
approach reviewed here (Antrás), capital-intensive firms choose vertical integration while labor-intensive
firms choose outsourcing. Empirical work is in its early stages, but results are supportive to the theory.

15
REFERENCES
Antrás, Pol (2003), “Firms, Contracts, and Trade Structure”, Quarterly Journal of Economics 118(4),
1375-1418.
Braconier, Henrik, Pehr-Johan Norbäck, and Dieter Urban (2005), “Reconciling the Evidence on
the Knowledge-Capital Model.” Review of International Economics 13(4): 770-86.
Brainard, S. Lael (1997), "An Empirical Assessment of the Proximity-Concentration Tradeoff between
Multinational Sales and Trade", American Economic Review 87(4): 520-544.
Carr, David L., James R. Markusen, and Keith E. Maskus (2001), “Estimating the Knowledge-Capital
Model of the Multinational Enterprise.” American Economic Review 91(3): 693-708.
Caves, Richard E. (2007). Multinational Enterprise and Economic Analysis. Cambridge: Cambridge
University Press, third edition.
Davies, Ronald B (2008), “Hunting High and Low for Vertical FDI”. Review of International
Economics, 16(2), 250–267, 2008
Dunning, John H. (1973). “The Determinants of International Production”. Oxford Economic
Papers 25: 289-336.
Ethier, Wilfred J. and James R. Markusen (1996), “Multinational Firms, Technology Diffusion and
Trade,” Journal of International Economics 41, 1-28.
Helpman, Elhanan (1984), "A Simple Theory of Trade with Multinational Corporations", Journal of
Political Economy 92(3): 451-471.
Markusen, James R (1984), “Multinationals, Multi-Plant Economies, and the Gains from Trade”, Journal
of International Economics 16(3-4): 205-226.
Markusen, James R. (2002), Multinational Firms and the Theory of International Trade. Cambridge:
MIT Press.
Markusen, James R. and Anthony J. Venables. (2000) "The Theory of Endowment, Intra-Industry, and
Multinational Trade", Journal of International Economics 52(2): 209-234.

16
ENDNOTES
1. The curvature of the two reaction functions follow from the assumption that $ + ( < 1, which in turn
implies that (/(1-$) < 1 and $/(1-() < 1.

Table 16.1: World statistics, 2007
Affiliates sales as a share of world exports 1.82
Value added of affiliates as a share of world GDP 0.11
Affiliate exports as a share of world exports 0.33
Affiliate exports as a share of affiliate sales 0.18
Source: UNCTAD World Investment Report
Table 16.2: Developed countries as source and destination for FDI:
developed countries' share of world totals
FDI inflows FDI outflows
2007 0.66 0.85
FDI inward stock FDI outward stock
1990 0.73 0.92
2007 0.69 0.84
Source: UNCTAD World Investment Report
Table 16.3: Local sales, export sales, and imports of
foreign affiliates, 2007
Affiliate local sales Affiliate exports Affiliate imports
as as share of as a share of as a share of
total sales total sales affiliate sales
United States 0.72 0.28 0.06
Japan 0.65 0.35 0.43
Sweden 0.78 0.22 0.16
UNCTAD World Investment Report 2008, Annex Tables B.12, B.15, B16
Table 16.4 Sales by US and Swedish manufacturing affiliates:
shares in total, 2003
local sales export sales export sales
back to parent to third
country countries
USA
2003 0.60 0.13 0.26
Sweden
1998 0.65 0.08 0.27
Source: Markusen (2002), Davies, Norbaeck, Tekin-Koru (2009)
high
degree of
independence
from MNE:
outsourcing?
low
high
low
MNE
offshoring
decision
vertical integration /
outsourcing decision
license arm's
length
joint
venture
foreign direct
investment
produce in
host country
export to
host country
degree of
foreign value
added:
offshoring?
Figure 16.1: Decision tree for FDI

130
120
110
100
90
80
70
60
50
V M D
0.03 0.08 0.13 0.18 0.23 0.28 0.33 0.38 0.43 0.48 0.53 0.58 0.63 0.68 0.73 0.78 0.83 0.88 0.93 0.98
Share of demand in country i
Figure 16.2: Relative size differences and choice of regime,
Firm profits
horizontal
vertical national
0.05
0.15
0.25
0.35 0.45 0.55 0.65 0.75 0.85 0.95
0.95
0.85
0.75
0.65
0.55
0.45
0.35
0.25
0.15
0.05
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
World endowment of the
composite factor
World
endowment
of skilled labor Origin for country i
Volume of
affiliate
production
Figure 16.3: Affliate sales in the knowledge-capital model
0.05
0.15
0.25
0.35
0.45 0.55 0.65 0.75 0.85 0.95
0.95
0.85
0.75
0.65
0.55
0.45
0.35
0.25
0.15
0.05
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Volume of
affiliate
production
World endowment of
the composite factor
World
endowment
of skilled labor
Origin for country i
Figure 16.4: Affliate sales in the knowledge-capital model,
restricted to horizontal firms
0.05
0.15
0.25
0.35
0.45
0.55 0.65 0.75 0.85 0.95
0.95
0.85
0.75
0.65
0.55
0.45
0.35
0.25
0.15
0.05
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Volume of
affiliate production
World endowment of the
composite factor
World
endowment
of skilled
labor Origin for country i
Figure 16.5: Affliate sales in the knowledge-capital model,
restricted to vertical firms

L
K
A
B
C
F*
S*
Fv
Fo
Sv So
Lv Lo L*
K*
Kv
K0
A - first best, B - vertical integration, C - outsourcing
Figure 16.6: vertical integration versus outsourcing

Chapter 17
FRAGMENTATION, OFFSHORING, AND TRADE IN SERVICES
1
17.1 Stylized facts, basic concepts
There are a few topics that have attracted a good deal of attention but which do not fit neatly into
any simple model of trade. One of these is the fact that more things seem to have become traded in the
last decade or two, including intermediate goods and various types of services. Indeed, twenty years ago
many trade economists more or less defined services as non-traded.
A couple of terminology points first. There is a distinction between new goods and newly-traded
goods. The former of course refer to things that simply did not exist before while the latter refer to things
that existed as non-traded goods or services which have now become traded. The latter is the topic of this
chapter. The breaking up of the production process for a final good into different stages which are
physically located in different countries has been referred to by a number of names, but we will tend to
use the term “fragmentation”. Vertical specialization is another name, though this has been used in a
more narrow sense. The introduction of newly traded goods is sometimes referred to as the expansion of
trade at the “extensive” margin whereas more trade in the same goods is often called expansion at the
“intensive” margin. By whatever name, trade economists have argued persuasively that much of the
great increase in trade over the last decades has come by expansion at the extensive margin (Hummels,
Rapoport, and Yi 1998; Hummels, Ishii, and Yi 2001; Yi 2003).
Often intermediate goods are produced in one or several countries while final assembly occurs in
another country. Of course, final assembly may also occur in many countries which is the stuff of
horizontal multinationals discussed in the previous chapter. In many cases such as in electronics, the
“upstream” intermediates or components are more capital and skilled-labor intensive while “downstream”
assembly is less-skilled-labor intensive but there are surely examples which run the other way. Finally,
we again draw the distinction between “outsourcing”, which we reserve for production outside instead of
inside the ownership boundaries of the firm and “offshoring” which we reserve for geographical location
of production outside the home country.
There are obvious advantages that fragmentation offer to the world economy. Think back to
Chapter 15 where we used the concept of the factor-price-equalization set. If the countries’ endowments
are outside this set, then trade in goods alone cannot exhaust all gains from trade. In that Chapter, we
discussed how added gains from trade can be captured by allowing trade in factors. The same is true in
the presence of increasing returns to scale as we discussed in that Chapter. Trading intermediate goods
and services accomplishes the same thing and in the previous chapter we advanced the view that
multinational firms could be thought of primarily as vehicles for the intra-firm trade in services such as
management, technology, marketing, and finance.
There is no general theory of fragmentation at this time and it is probably fair to say that the
literature consists of a lot of fairly specific situations. Along with this, there few if any general welfare
results about fragmentation and the expansion of trade at the extensive margin. Specifically, it is
impossible to predict that all countries will be made better off by added traded goods, never mind the
further complication of factor-price changes within economies. The fundamental problem is that newly-

2
traded goods and services introduce general-equilibrium effects that mean that some individual countries
could be made worse off through adverse price changes. We can give an example of this which is taken
from Markusen and Venables (2007). Suppose that our country is really good at making shirts, a product
produced in two stages: fabric (textiles) and cut-and-sew. Our country is not great at either of those
individually, but good enough at each to be very competitive on the world market. Now fragmentation of
the industry is made possible and suppose that we find that some countries are really good at textiles and
some others really good at cut-and-sew: neither group was good at the other activity initially which is
why we were so competitive. With trade in intermediates allowed, the world price of shirts is going to
fall and our country is going to experience an adverse terms-of-trade change. In spite of our best efforts
to adapt, we could be worse off.
The next two sections of the Chapter will look at quite specific examples of fragmentation from
two recent papers. The first is from Markusen and Venables (2007) and Markusen (2010) while the
second is from Grossman and Rossi-Hansberg (2008). The two provide an interesting contrast. Section
4 provides a somewhat general gains-from-trade theorem with respect to the effects of newly-traded
goods and services and we do indeed see the crucial role of changes in the prices of existing traded goods
for establishing welfare results. The final theory section looks at the issue of trade and FDI in services
and the so-called four modes of trade.
17.2 Fragmentation and newly-traded intermediate goods
The first special case we will consider might be termed “conventional”, with no normative
connotation that this is good or bad, in that it follows quite a number of more general formulations in
the literature. Examples are Deardorff (2001, 2008), Jones and Kierzkowski, (2001), and a series of
articles in Arndt and Kierzkowski (2001). Empirical analysis is found in Feenstra and Hanson (1996,
1997). Initially-traded final goods are assembled from intermediate goods and services or “components”
and may also use primary factors directly. Innovations in transportation or management allow some of
the components to be directly traded.
An example is shown in Figure 17.1 which sticks to the Heckscher-Ohlin tradition for familiarity:
there are two final goods, X1 and X2 and two primary factors K and L (or these could be skilled and
unskilled labor)There are three intermediate goods, denoted A, B, and C which are produced from primary
factors L and K. A is the most capital intensive, B is in the middle, and C is the most labor intensive. X1
is costlessly assembled (no added primary factor requirements) from A and B and X2 is costlessly
assembled from B and C. All production functions as assumed to be Cobb-Douglas so that the value
shares of factors in intermediates and value shares of intermediates in final production are constants.
Figure 17.1 gives these shares. Including intermediate use, X1 is capital intensive overall with factor
shares 65 for capital and 35 for labor in the simulation we present shortly. X2 is the mirror imagine, with
share 35 for capital and 65 for labor.
Figure 17.1
We cannot present a great many cases here and will stick with just one specific example in order
to illustrate some general ideas. We assume three countries, labeled 1,2 and 3. Country 1 is capital
(skilled labor) abundant with an endowment ratio K/L = 85/15 and country 3 is labor abundant with an
endowment ratio K/L = 15/85. Country 2 is in the middle, with an endowment ratio 50/50. The model is
very symmetric, both with respect to production and with respect to countries. Note that the country 1

3
and 3 endowment ratios are most extremely (differ more from one) than the factor intensities of A and C
which are 80/20 and 20/80 respectively. This is important for some results, but we will not comment
more on this in more detail.
Table 17.1 shows some simulations of this model. The model is benchmarked (parameters
picked) so that the integrated world equilibrium we have discussed before gives each country a welfare
level of one and all factor and commodity prices are one. The first column of results gives results when
X1 and X2 are freely traded but no trade in intermediates is allowed. Because country’s 1 and 3 have
extremely endowment ratios, they are outside the factor-price equalization set, specialized in X1 and X2
respectively. Each of these countries has a high real price for its scarce factor and a low real price for its
abundant factor, and each country has a welfare level (0.885) which is lower than in the integrated
equilibrium. Because of the symmetry in the model, relative prices of X and Y are one, and so country 2
is indifferent to trade and enjoys the same welfare level as in the integrated world equilibrium.
Table 17.1
The second column of results in Table 17.1 allows free trade in A and C, what we could call a
“symmetric” fragmentation. Countries 1 and 3 become the sole producers of A and C respectively which
are better suited to their factor endowments, and country 2 becomes specialized in producing B only.
Countries 1 and 3 have a welfare gain, but have the usual Stolper-Samuelson effect which raises the real
income of the abundant factor and lowers the real income of the scarce factor. Perhaps the most
unanticipated result of Table 17.1 is the large welfare gain for country 2: it is not the case that the country
with the average world endowment is unaffected by the symmetric fragmentation.
The intuition for the large gain for country 2 can be explained in relatively simple supply and
demand terms. In the base case, we noted that country 2 is indifferent to trading goods only and, when
trade in A and C are opened up, it remains indifferent as long as prices all remain at one. But countries 1
and 3 are not at all indifferent. They are inefficient producers of B, and so at prices of unity they want to
export A and C to country 2 and import finished X1 and X2 back in exchange. This drives down the prices
of A and C relative to B, which creates a welfare gain for country 2 in the middle: country 2 imports A and
C cheaply from countries 1 and 2 respectively, assembles these with local B, and exports finished X1 and
X2 back to 1 and 3 respectively. In this simulation, country 2 imports A and C at the price 0.91 while the
domestic price of B, its specialty, is 1.10. Both factors share fully in this gain for country 2.
The third column of results in Table 17.1 shows the results for an asymmetric fragmentation in
which only intermediate good C, the most labor intensive, is traded. A and B are non-traded. Here we see
a couple of interesting results. First, note that the relative prices of X1 and X2 are going to change: the
relative prices of X1 rises and that of X2 falls. Country 3 specializes more in C and this drives down the
price of C and hence of X2 in general equilibrium. In this particular example, this negative terms-of-trade
effect is actually strong enough to reduce the welfare of country 3 relative to the benchmark. The
Stolper-Samuelson theorem applies here, and the scarce factor K in country 3 suffers a loss and also does
so in country 2 as the theorem would predict. But the opening of trade in C is pure trade creation for
country 2 (it did not trade initially), so the gains-from-trade theorem fully applies and we see that its
welfare rises.
Perhaps the most interesting result in the third column of Table 17.1 is that not only does country
1's welfare increase, but both factors gain. The reason for this is that country 1 does not produce or
import C either before or after fragmentation. Country 1 produces X1 from domestically produced A and

4
B: its relative factor prices are pinned down by its domestic specialization in X1. So its welfare level and
the real prices of both factors rise by the same proportion.
Table 17.1 is a very specific and special case, but it shows both the opportunities and the pitfalls
that are involved in expanding trade at the extensive margin. All countries can gain, but it is possible
that some countries might lose due to an adverse terms-of-trade effect. We will shown in section 17.4
that this is in fact a very general proposition: a country can only lose if it suffers a deterioration in the
price(s) of its initially exported good(s), otherwise it is assured of gain. Table 17.1 also shows that it is
possible though not inevitable that all primary factors in a country might gain: country 2 in the second
column and country 1 in the third column.
17.3 Fragmentation and trade in “tasks”
A recent paper by Grossman and Rossi-Hansberg (2008) has proposed a rather different structure
which they term “tasks”. A simplified version of their formulation is shown in in Figure 17.2. There are
again two final goods, X1 and X2 and two primary factors, L and K. A task uses just a single factor, either
capital or labor. But there are many labor tasks and many capital tasks. In our simplified version here,
we assume that there are just two labor tasks, L1 and L2, and two capital tasks, K1 and K2. Each final good
requires all labor tasks in equal amounts and it is further assumed that there is no substitution possible
among these tasks; e.g., X1 might require exactly one unit of both L1 and L2 with no substitution possible.
The same is true for capital tasks.
Figure 17.2
Where the final goods differ is in the ratio of labor to capital tasks and hence there remains a
Heckscher-Ohlin type of factor intensity. A numerical example of this is given in Figure 17.2. X1 is
assumed to be capital intensive, requiring 65 units of each of tasks K1 and K2 and 35 units of each of tasks
L1 and L2 to produce 400 units of X1 if all prices are initially normalized at unity. X2 is the labor-intensive
mirror image: 65 units of each of tasks L1 and L2 are required and 35 units of each of K1 and K2.
There are two important differences between this and the formulation in the previous section.
First, a country cannot, by definition, have a factor-endowment ratio that is more extreme than some
intermediate (task) that is potentially traded. Tasks use only a single primary factor and hence that there
are thing that are potentially tradeable that have a more extreme factor intensity than any country (or at
worse a tie if there is a country that has only one factor). Second, if only a subset of labor or capital tasks
are tradeable, then free trade in the subset is quite different from simply allowing labor or capital
migration as in Chapter 15. A trabeable L task must, for example, be combined in fixed proportions with
a non-tradeable domestic L tasks produced by a domestic worker.
This last property is crucial in certain results that Grossman and Rossi-Hansberg emphasize in
their paper and so some intuition is warranted. Suppose we have a small country h facing fixed world
prices, but it is capital abundant and specialized in X1 and so has a wage rate higher than the rest of the
world. If task L1 becomes tradeable, our country will of course import L1. But each unit of imported L1
must be combined with a unit of L2 produced by a domestic worker. Let’s do the thought experiment of
holding domestic factor prices constant in country h. Production of X1 is now strictly profitable since the
labor “composite” input (tasks L1 and L2 in fixed proportions) is now cheaper: a combination of the lower

5
cost of importing L1 and the existing wage for task L2. Capital is in fixed supply, so firms will demand
more labor until the imported L2 is equal to the domestic labor force, all of whom now do task L2.
This expansion in X1 output will raise the relative and real price of the composite capital input (in
fixed supply) and lower the price of the composite labor input. However, although the final price of the
composite labor input may be below its price before task trade, the former is a combination of the lower
import price of L1 and the higher domestic wage paid for task L2. Thus it may well be that the final price
of domestic labor is higher than before task trade and this is indeed a principal finding that Grossman and
Rossi-Hansberg highlight. When a limited number of tasks can be traded, the domestic factor performing
those tasks (competing with the imports) may be made better off. Grossman and Rossi-Hansberg make
an interesting analogy between this trade in a limited number of tasks and technical change. In our simple
case just discussed, the composite L task becomes cheaper at a given domestic wage rate, akin to a
technical improvement in the production of the composite L task.
In order to provide some concreteness with the minimum possible algebra, suppose that the
composite capital task (K1 and K2 in fixed proportions, perfect complements) and the composite labor task
(L1 and L2 in fixed proportions, perfect complements) are perfect substitutes in producing final good X1.
This is equivalent to saying that there are two alternative ways of producing X1:
(17.1)
Let r and w be the domestic prices of capital and labor, and t the price of imported L1. The fixed world
price of X1 is denoted p1 and superscript n denotes no trade in tasks and superscript t denotes trade in L1
permitted. Then the cost functions corresponding to (17.1) and the price-equals-marginal-cost conditions
for country h are as follows:
(17.2)
The second equation applies in the absence of trade in tasks and the third equation with imports of task L1.
In this very special case, it is clear that allowing trade in L1 does not affect the return to domestic capital.
However, with t < w, it must raise the price of w, the wage of the domestic labor that (superficially)
competes with imported task L1: wt > wn.
While Grossman and Rossi-Hansberg put a great emphasis on this interesting result, it is of
course a special case. Table 17.2 presents some simulations of our simplified version of their model. In
the top panel of the Table the countries are the same size and have the same factor endowments as
countries 1 and 3 of the previous section. X1 and X2 are assumed Cobb-Douglas as in the previous section
while the two labor tasks are required in fixed proportion and the same for capital tasks. Factor shares
etc. are chosen to be the same as in the previous section so that the benchmark when the countries trade in
X1 and X2 only produces exactly the same outcome as in Table 17.1. The two models are complete
equivalent when only X1 and X2 are traded.
Table 17.2
The first counterfactual simulation in Table 17.2 (second column of results) allows trade in tasks
L1 and K2, a symmetric fragmentation similar to that in the previous section where we allowed A and C to
both be traded. Both countries gain and the Stolper-Samuelson property again appears: there are real
gains to each country’s abundant factor and real losses for each country’s scarce factor. The second

6
counterfactual allows only trade in L1, with the results reported in the third column of Table 17.2. Both
countries gain, but country h, the importer of task L1 gains less. Lower down we see the reason: the
added production capacity in country h and lower capacity in country f moves the terms of trade against
country h: the relative price of X1, its export good, declines. The direction of factor-price changes is the
same as in the symmetric fragmentation though the magnitudes are different.
Panel B in the lower part of Table 17.2 conducts the experiment we discussed above: country h is
small (everything is the same except the factor endowment of country f is multiplied by 100). Country
h’s welfare and the prices of both factors are higher in the benchmark now because the relative price of
X1, country h’s export good, is higher in the benchmark. The first counterfactual reported in the second
column of Panel B is again the symmetric fragmentation of allowing trade in both L1 and K1. Country h
gets a huge welfare increase, but yet again we see that the scarce factor labor suffers a real income loss.
Capital exports from country h reduce the capital available for local production and so lower the marginal
product of domestic labor in spite of being able to import L1. The third column of Table 17.2 shows the
result emphasized above in which allowing trade in L1 only actually raises the prices of both factors of
production in country h. It also demonstrates that the result does not rely on the perfect substitutes
assumption used to explain the result above (Table 17.2 assumes that L and K composites are Cobb-
Douglas substitutes exactly the same as Table 17.1).
As a final point, both Tables 17.1 and 17.2 report the proportional change in the volume of trade
in the counterfactuals. In all cases, these numbers are positive which certainly concurs with intuition:
more things traded means more trade. For exactly this reason, we might note that Markusen and
Venables (2007) and Markusen (2010) emphasize that this is not a general result and that it is easy to
construct cases in which some countries will trade less as a consequence of fragmentation. A simple
example should suffice. Suppose that there are two goods, wheat and cars. Suppose that country h is
generally really good at cars but simply cannot produce tires. Then if only final goods are traded, country
h will be forced to export wheat in exchange for cars. But now allow tires to be traded. Country h can
now just import the tires and pay for this by exporting a few cars and/or a little wheat. The volume of
trade falls. Markusen and Venables show that this is by no means pathological and it occurs for
significant subsets of countries in their multi-country simulations.
17.4 A gains-from-trade theorem
General gains-from-trade results in trade theory are hard to obtain. The basic textbook result only
proves that free trade is better for a country than autarky or, somewhat more generally, any level of
restricted (but not subsidized!) trade is better than autarky. However, there is no general result that says
that more trade is better than less (but positive) trade. This is due to the possibility of adverse price
changes; e.g., further liberalization by a large country can make it worse off. But this is exactly the type
of result we are seeking here: under what conditions will the introduction of an additional set of trade
possibilities to an already-existing set of traded goods improve welfare? Our approach follows the classic
“revealed preference” methodology that we have used several times in the book. The theorem itself is a
much simplified version of a more general proof in Markusen (2010).
X denotes a vector of final goods quantities, with the vector p denoting their prices. Similarly, Z
denotes a vector of intermediate goods with prices q. In our explicit model above, X (now denoting X1,
X2) and Z (now denoting A, B, and C) were disjoint sets. The important distinction is that (a) only X
goods enter into welfare-producing consumption and (b) we will define the absence of fragmentation to

7
mean that goods cannot be traded for intermediate use. That is, the Z goods are initially non-traded.
Superscript f denotes prices with fragmentation (trade in intermediates allowed), while subscript
n denotes prices in the initial no-fragmentation equilibrium. Thus vectors of world final goods prices
with and without fragmentation are pf and pn respectively. qf denotes the world prices of the Z goods
when they can be traded and qn denotes their domestic prices when they are non-traded. In the simplified
proof here, we assume that the X goods are all freely and costlessly traded (again, see Markusen (2010)
for a more complete proof with trade costs). Thus the domestic prices of the X goods must equal world
prices.
The notation that we have been using, X for production and D for consumption is now a little
awkward so we make a modification in order to avoid confusion. Just X and Z will be used: a subscript
“o” on an X or Z quantity denotes production (output) of the good and a subscript “c” denotes
consumption. For an intermediate good, “consumption” means its use as a domestic input in a final good.
Thus for both final and intermediate goods, a positive value for production minus consumption of the
good indicates it is an export good, and so forth. We will also simplify the notation a bit: rather than
keep writing lots of summations, we will use a shorthand such as
We assume a competitive undistorted economy, and so production efficiency applies and the
value-added of the economy in regime j (j = f, n) must be maximized at that regime’s prices. Specifically,
the value added in regime f must be greater than or equal to regime-n value added when the latter is
evaluated at regime-f prices. Total value added for the economy is given by the value of final goods
production minus intermediate usage (“consumption” of intermediates) plus the value of intermediate
goods production. The production efficiency condition is then given by
(17.3)
This rearranges to an inequality on final production plus net exports of intermediates.
(17.4)
The last term in brackets on the right-hand side is zero: intermediates are not traded in the n-regime, so
production and consumption are the same for each good.
(17.5)
Now we introduce the balance of trade constraints for each regime, which require that the value
added in production equals the value of final consumption.
(17.6)
(17.7)

8
Substitute the right-hand side of (17.6) for the left-hand side of (17.5)
(17.8)
Now add the right-hand side of (17.7) and subtract the left-hand side, and also add and subtract the term
. (17.8) then becomes
(17.9)
(17.10)
(17.11)
The classic revealed-preference criterion for gains from fragmentation requires that the left-hand
side of (17.11) is greater than or equal to the first term on the right-hand side: the value of regime-f
consumption is revealed preferred to regime-n consumption valued at regime-f prices. So a sufficient
condition for this to be true is that the second additive term on the right-hand side of (17.11) is non-negative.
Inequality (17.11) tell us that a sufficient condition for the added trade in the intermediate Z
goods to improve welfare is that world prices of the initially-traded X goods don’t
change . Actually all that is needed is that the world relative prices of the X goods don’t
change ( a numeraire rule could be that the X prices sum to one, so then the price vectors pf and pn will be
the same before and after Z trade). The absence of terms-of-trade changes for the X goods could be
referred to as a “neutral” or “symmetric” fragmentation, emphasizing that these are definitions (e.g., a
symmetric fragmentation is one that leaves relative final-goods prices unchanged).
The absence of relative price changes for initially-traded goods rules out the sort of example we
posed about: a country good at shirts and the result of fragmentation is that efficient fabric and cut-and-sew
producers lead to a fall in the world relative price of shirts. This welfare loss cannot happen if the
relative price of shirts to other initially-traded goods don’t change and in fact the country will generally
benefit by exploiting some differences between its (old) domestic prices for fabric and tailoring and the
new world prices.
The condition that world prices do not change is very restrictive and “overly” sufficient.
Consider again the last term in (17.11): if this is non-negative, then this is sufficient for gains from
fragmentation. Let denoted the vector of net exports in regime n (no trade in
intermediates), also referred to as the “initial” net export vector. The last term in (17.11) can be written as
either of the following
by trade balance (17.12)
The first expression is just a simple correlation between the changes in domestic final-goods
prices and the initial net export vector. We will have gains from fragmentation if “on average” the prices

9
of initially-exported goods rise and the prices of initially-imported goods fall; that is, there is an
“aggregate” terms-of-trade improvement. Noting that the value of the initial trade vector at initial prices
is zero (trade balance: equation (7)), we have the second expression. If, after fragmentation, the country
were to retain its initial net export vector, it would run a trade surplus if the inequality is strict. It could
then improve its welfare by cutting some exports and/or increasing some imports to restore trade balance.
Unfortunately, the inequality in (17.12) is unlikely to hold for all countries. If there are only two
countries, then it cannot hold for both, because the elements of their net export vectors are equal and
opposite in sign. One country must have an aggregate terms-of-trade deterioration. However, we must
emphasize again that (17.12) is a sufficient condition for which is itself a sufficient
condition for gains. It is easy to produce numerical examples where a country benefits substantially from
fragmentation and outsourcing in spite of a terms-of-trade loss and all countries gain.
17.5 Trade and foreign direct investment in business services
Services cover a very wide range of items, and we have made a decision here to restrict our
analysis to business services which are generally intermediate goods (services supplied to other
businesses), consistent with the overall focus of this chapter. It is indeed the case that trade and FDI in
business service has increased greatly in the last decade or two. Statistics are presented in Table 17.3 for
the United States and a very thorough analysis is found in Francois and Hoekman (2008). It is also clear
that much of the increased activity is mediated by multinational firms, whether that is actual FDI (we
measure sales of foreign affiliates) or intra-firm trade between parent and affiliate. Table 17.3 shows that
both trade and FDI have increased somewhat faster than all trade and that they now account for a very
significant proportion of US international economics activity.
One reason to study trade and FDI in business services, in addition to the rate of growth, is that
they are often subject to very different restrictions than trade in goods. Barriers to trade and investment in
services can be roughly broken down into what we could call “natural” economic costs and “policy-imposed”
costs. For the former, we are thinking of things like communications and transport costs
(workers flying between countries), language, customs, time zones, the need for face-to-face interaction
and so forth.
Policy-induced barriers, henceforth “barriers” for this section, to trade in services take diverse
forms and therefore affect service suppliers’ cost functions differently. Regulatory policies, in addition to
explicit and implicit barriers to trade in services, generally fall into one of five basic categories. First,
there can be quantity-based restrictions imposed on services suppliers that explicitly restrict the volume of
services imported, similar to a quota. The use of a fixed number of licenses available or access to only
certain firms or sectors also falls into this category. If a “quota” type policy is only applied to imported
services, then we would expect to see more multinationals establishing affiliates in the market (all other
things equal). This is similar to a “tariff jumping” activity discussed in traditional theories of the
multinational firm.
Thirdly, there are numerous barriers to establishment that restrict foreign supply of services due
to the high costs of establishing a commercial presence. Policies regarding licensing procedures,
requirements and fees can be prohibitive. Bureaucratic red tape, requirements for local management, or
lack of transparency all have detrimental effects on the fixed costs of establishing commercial presence
for multinationals. These may create a substitution effect with multinationals supplying the service from

10
abroad rather than establish an affiliate.
Fourthly, barriers to trade and establishing commercial presence in services may take the form of
restricting the use of inputs. This category can include restrictions on workers, required percentages of
locally produced material inputs, as well as barriers or limits on the use of networks or media for
promotion and/or marketing purposes. These policies can greatly increase the costs of operations for
foreign suppliers and may be prohibitive to entering the market.
The last category of restrictions encompasses the various domestic regulatory barriers that take
many forms and are often overlooked when discussing impediments to trade and investment. These
include policies regulating professional qualification, residency and citizenship restrictions, obligatory
membership in local professional association, juridical requirements, and limitations of inter-professional
cooperation. While the policies and regulations may not explicitly target foreign firms they often have
this effect in practice. Regulations on professional qualifications are important domestic policies to have
so as to guarantee a level of skill and professionalism to consumers. However, when these policies require
residency, citizenship, or involve re-certification for professionals with comparable certifications from
another country, they become costly. While they are not explicit barriers to trade these policies severely
increase the fixed costs (time and money) for a firm establishing commercial presence and may be
prohibitive as well.
It is beyond the scope of this chapter to present a detailed analysis of these issues, but we would
like readers to at least be aware of the basic ideas involved. Figure 17.3 presents a schematic of the
general principals. Final consumption is assumed to be composed of manufacturing and agriculture.
Business services are an intermediate input into both of these industries which also use primary factors of
production skilled and unskilled labor. Following closely on our modeling of multinational firms in the
previous Chapter, business services can be decomposed into a headquarters activity and offices. Banks,
finance, and insurance companies, for example, may have a headquarters in New York or London, and
have offices throughout the world. Similar to horizontal multi-national (which is what many services
multinationals are), a firm has one headquarters but possibly many offices.
Figure 17.3
Figure 17.3 notes the differences in trade in services and foreign direct investment in services. In
the upper part of the chart, we note that trade in services occurs when there is an international geographic
separation between the service firm and the service purchaser in agriculture or manufacturing. In the
lower part of the Figure, we note that FDI in services occurs when there is a geographic separation
between headquarters and one or more offices. Either type of separation can occur without the other; for
example, a horizontal firm tends to have geographic separation between headquarters and some offices,
but not between the offices and service buyers.
As briefly discussed above, quite a range of natural and policy restrictions impact on whether or
not it is easy and/or legal to fragment headquarters from offices and fragment the service provider from
final user. Figure 17.4 characterizes these in terms of what are know as the four modes of service trade,
which comes from the World Trade Organization’s GATS: general agreement on trade in services. The
schematic in Figure 17.4 has two countries, north and south, and the headquarters of a service firm
headquartered in the north is represented by the central box, and a possible northern office and southern
office are represented by the two boxes to either side.

11
Figure 17.4
The Figure shows four possible ways in which services can be provided by the offices to north or
south manufacturing and agricultural firms. The channel on the upper left has the headquarters and
office both in the north and providing services to northern firms. We mark this as “always feasible” since
it is a strictly domestic activity. Mode 1 of the GATS is defined as cross-border trade in services, which
is the channel on the lower left of Figure 17.4: the northern office of the northern firm supplies services to
southern firms. The northern firm may also have a southern office, which is by definition FDI in services.
This is known as Mode 3 trade in services in the GATS terminology. This southern office can supply
services to local firms, which is represented by the channel on the lower right of Figure 17.3. Finally, the
southern office may be able to supply services to northern firms which would require both Modes 1 and 3
being feasible. An example of this last type of supply is call centers which are located in developing
countries such as India. A customer in the north needing service from a northern firm finds their phone
call re-routed to a call center in India.
For completeness, GATS mode 2 is when the buyer travels to the seller’s country of origin.
Traditionally, this category has been dominated by tourism. But the extensive margin of Mode 2 is
expanding as in “medical tourism”. Patients in high-income countries are finding that they can get high-quality
surgery or dental work done in places as diverse as Hungary, India, and Mexico for a fraction of
the cost of having it done at home. We have seen websites in Ireland, for example, that advertises dental
holidays in Hungary, in which you get round-trip plane tickets, two days in a resort, and all your dental
work done for less than just the latter would cost in Dublin. GATS Mode 4 is the international movement
of person to work abroad, typically for limited periods of time rather than permanent migration. There is
obviously some overlap and fuzziness between this and Mode 3. If Gregorz the Polish plumber goes to
work in Denmark, that is Mode 4; if Gregorz Plumbing Incorporated opens a one-man office in
Copenhagen, that is apparently Mode 3.
As suggested earlier, both natural and policy-imposed restrictions can make it costly or even
prohibitive for a firm to engage in one or more of the modes shown in Figure 17.4. If face-to-face contact
is required, for example, then Mode 1 can be prohibitively costly. If countries restrict the right of foreign
firms to establish or acquire local subsidiaries (known as right of establishment) then Mode 3 is difficult.
The principle of national treatment means that foreign firms are suppose to be treated exactly the same as
domestic firms, but this is often violated in practice (de facto) by rules such as requiring managers to be
local citizens and so forth.
Finally, arguments about whether or not something is a barrier and how significant is it abound
between countries. Part of the difficulty is that many service industries are by their very nature highly
regulated. Consider banking, finance, insurance, architecture and construction, telecommunications, legal
services and medical services. These tend to be regulated in all countries and while foreign firms see such
regulations as barriers, the local firms and governments always tend to claim that they are non-discriminatory
against foreign firms.
17.6 Summary
It has been documented that a good deal of the expansion of trade over the last decade or two has
been through the extensive margin, meaning trade in goods and services which were not previously
traded, rather than through the intensive margin: trading larger quantities of the same things. Much of

12
this new trade seems to be in intermediate goods and services as firms fragment the production process to
take advantage of favorable factor prices or other considerations across geographic locations.
The theoretical literature on fragmentation and outsourcing is now extensive, but there is no
general model we can rely on. We presented two special cases from recent literature that posit different
mechanisms as to how we might think about fragmentation and newly-traded goods. These are
interesting but even in these rather restrictive special cases results still depend on choices of parameters
such as country sizes and factor intensities.
A simplified gains-from-trade theorem pinpoints one difficult in trying to obtain general results.
This difficulty is that, when new goods and services become traded, the prices of existing goods change in
general-equilibrium which leads to terms-of-trade effects for the countries involved. A country could lose
through fragmentation because the components and stages of production for something the country was
very good at can now be individually done better by several other countries. It is easy to create specific
examples in which everyone gains but unfortunately also not impossible to create examples in which
someone loses.
We closed with a discussion of trade and FDI in intermediate business services, a fast-growing
component of total world economic activity. The theoretical issues here are much the same as in the
previous Chapter on multinational firms and the choice between exporting and FDI. What make trade
and FDI in business services of particular interest is that many of the natural and policy barriers to trade
and investment differ substantially from those restricting trade in goods.

13
REFERENCES
Arndt, Sven W. and Henryk Kierzkowski (editors) (2001), Fragmentation: New Production Patterns in
the World Economy. Oxford University Press, Oxford.
Brainard, S. Lael. and Susan. Collins (eds.) (2006), Brookings Trade Forum 2005: Offshoring White-collarWork
(Washington, DC: Brookings Institution).
Deardorff, Alan V. (2001), “Fragmentation across Cones,” in Sven W. Arndt and Henryk Kierzkowski,
eds., Fragmentation: New Production Patterns in the World Economy, Oxford: Oxford
University Press, 35-51.
Deardorff, Alan V. (2008), “Gains from Trade and Fragmentation”, in Steven Brakman and Harry
Garretsen, eds., Foreign Direct Investment and the Multinational Enterprise, CESifo Seminar
Series, Cambridge: MIT Press, 155-169.
Feenstra, Robert C. and Gordon H. Hanson (1996), “Globalization, Outsourcing, and Wage Inequality”,
American Economic Review, 86, 240–45.
Feenstra, Robert C. and Gordon H. Hanson (1997), “Foreign Direct Investment and Relative Wages:
Evidencefrom Mexico’s Maquiladoras”, Journal of International Economics, 42, 371–93.
Francois, Joseph and Bernard Hoekman (2008), “Liberalization of Trade in Services: A Survey”, World
Bank working paper (under revision for the Journal of Economic Literature).
Grossman, Gene. M., Estaban Rossi-Hansberg (2008), “Trading Tasks: A Simple Theory of Offshoring”,
American Economic Review 98, 1978-1997.
Hanson, Gordon. H., Raymond J. Mataloni, and Matthew J. Slaughter (2005), “Vertical production
networks in multinational firms”, Review of Economics and Statistics 87, 664-678.
Hummels, D., Rapoport, D., Yi, K-M. (1998), Vertical Specialization and the Changing Nature of World
Trade. Federal Reserve Bank of New York Economic Policy Review 4, 79-99.
Hummels, David, Ishii, J. and K.-M. Yi (2001), “The Nature and Growth of Vertical Specialization in
World Trade”, Journal of International Economics 54, 75-96.
Jones, R. W., Kierzkowski, H. (2001), A Framework for Fragmentation, in: Arndt, S.W.. Kierzkowski,
H., (Eds), Fragmentation: New Production Patterns in the World Economy, Oxford University
Press, Oxford..
Leamer, Edward and James Levinsohn (1995), “International trade theory; the evidence”, in: Gene
Grossman and Kenneth Rogoff (editors), Handbook of International Economics, Vol. 3. North-
Holland, Amsterdam, 1339–1394.
Markusen, James R. (2010), “Fragmentation and offshoring: a general gains-from-trade theorem and
some specific cases”, working paper.

14
Markusen, James R. and Anthony J. Venables (2007), “Interacting factor endowments and trade costs: a
multi-country, multi-good approach to trade theory”, Journal of International Economics 73,
333-354.
Markusen, James R. and Bridget Strand (2009), “Adapting the Knowledge-Capital Model of the
Multinational Enterprise to Trade and Investment in Business Services”, World Economy 32, 6-
29.
Ng, F., Yeats, A. (1999), “Production sharing in East Asia; who does what for whom and why”, World
Bank Policy Research Working Paper 2197, Washington DC.
Venables, Anthony J. (1999), “Fragmentation and multinational production”, European Economic
Review,43, 935-945.
Yeats, A. (1998), “Just how big is global production sharing?”, World Bank Policy Research Working
Paper 1871, Washington DC.
Yi, K-M. (2003), “Can Vertical Specialization Explain the Growth of World Trade?”, Journal of
Political Economy 111, 52-102.

Figure 17.1: Structure of production
Markusen / Venables
FINAL
GOODS
(50/50 shares in U)
INTERMEDIATE
GOODS
(50/50 shares in X,Y)
PRIMARY
FACTORS
Utility
X1 X2
A B C
L K L K L K
(20) (80) (50) (50) (80) (20)
FINAL GOODS ALWAYS TRADABLE AT COUNTRY-SPECIFIC TRADE COST
INTERMEDIATE GOODS NONE / SOME / ALL TRADABLE AT COUNTRY-SPECIFIC TRADE COST
PRIMARY FACTORS NOT TRADABLE

Table 17.1: Simulation results
Markusen (2010)
Benchmark Trade in Trade in
No trade in A (most K-int) C (most L-int)
Intermediates C (most L-int) No trade in A, B
Level Level Prop change Level Prop change
from bench from bench
Welfare country 1 (K abundant) 0.885 0.903 0.020 0.944 0.067
real price of labor 2.064 1.560 -0.244 2.202 0.067
real price of capital 0.676 0.789 0.167 0.722 0.067
Welfare country 2 1.000 1.103 0.103 1.035 0.035
real price of labor 1.000 1.103 0.103 0.894 -0.106
Welfare country 3 (L abundant) 0.885 0.903 0.020 0.849 -0.041
real price of labor 0.676 0.789 0.167 0.750 0.109
real price of capital 2.064 1.560 -0.244 1.420 -0.312
Price of X 1.000 1.000 1.068
Price of Y 1.000 1.000 0.937
Proportional change in trade volume 1.492 0.861
3.096 3.050

Grossman / Rossi-Hansberg (2008)
FINAL
GOODS
(50/50 shares in U)
INTERMEDIATE
GOODS
(AKA TASKS)
PRIMARY
FACTORS
Utility
X1 X2
65 65
K2
35 35
K1 L1 L2
100 100 100 100
K L
FINAL GOODS ALWAYS TRADABLE
INTERMEDIATE GOODS NONE / SOME / ALL TRADABLE
PRIMARY FACTORS NOT TRADABLE

Table 17.2: Simulation results
Grossman / Rossi-Hansberg
Benchmark Trade in Trade in
No trade in L1 and K1 L1
Intermediates
Panel A: Countries symmetric
Level Level Prop change Level Prop change
from bench from bench
real price of labor 2.065 1.395 -0.324 1.838 -0.110
Welfare country 2 (L abundant) 0.885 0.994 0.123 0.964 0.089
real price of labor 0.677 0.923 0.363 0.786 0.161
real price of capital 2.065 1.395 -0.324 1.974 -0.044
Price of X 1.000 1.000 0.922
Price of Y 1.000 1.000 1.085
Panel B: Country h small Level Level Prop change Level Prop change
(same relative endowments as panel A:) from bench from bench
real price of labor 2.674 0.433 -0.838 2.983 0.116
3.049 2.673

TABLE 17.3: US Foreign Affiliate Sales and Cross Border Trade
Outward US Foreign Affiliate Sales-all countries
1999 2005 % change
Total Sales-All Industries 2316654.8 3276024.4 41.41%
Total Private Services 353200.0 528000.0 49.49%
Information 63236.5 97069.9 53.50%
Finance & Insurance 86337.1 140341.6 62.55%
Finance 32330.4 43847.0 35.62%
Insurance 54006.6 96495.7 78.67%
PST 65290.2 97490.9 49.32%
Sales in Total Private Services = 15.25% and 16.12% of All Industries Sales in
1999 and 2005 respectively
All data are in millions of 2000 US dollars
PST- Professional, Scientific, and Technical Services
Inward Foreign Affiliate Sales-all countries
1999 2005 % change
Total Sales-All Industries 1831561.1 2213172.5 20.84%
Information 46440.3 48138.5 3.66%
Finance & Insurance 95840.7 104308.3 8.84%
Finance 15651.8 25458.9 62.66%
Insurance 80188.9 78849.4 -1.67%
PST 15757.0 49648.7 215.09%
Sales in Total Private Services = 16.02% and 17.57% of All Industries Sales in
1999 and 2005 respectively
PST- Professional, Scientific, and Technical Services
All data are in millions of 2000 US dollars
Cross-Border Trade- All countries-Exports
1999 2005 % change
All Industries 1287247.6 1586285.5 23.23%
All Industries-Affiliated 194697.1 186463.5 -4.23%
All Industries-Unaffiliated 1092550.5 1399822.0 28.12%
Total Private Services-Affiliated 32952.4 44441.2 34.86%
Total Private Services-Unaffiliated 73245.5 101278.7 38.27%
Financial Total 17789.3 31626.3 77.78%
Financial-Affiliated 4087.2 4312.0 5.50%
Financial-Unaffiliated 13702.2 27314.3 99.34%
Insurance Total* 3119.2 6011.5 92.73%
BPT Total 54683.1 73911.2 35.16%
BPT-Affiliated 26379.5 37062.1 40.50%
BPT-Unaffiliated 28303.5 36849.1 30.19%
Exports of total private services = 20.59% and 23.19% of trade in all industries
in 1999 and 2005 respectively
*Insurance transactions are considered unaffiliated by BEA
BPT-Business, Professional, and Technical Services
Cross-Border Trade- All countries-Imports
1999 2005 % change
All Industries 1543920.8 2177244.7 41.02%
All Industries-Affiliated 186215.3 231946.0 24.56%
All Industries-Unaffiliated 1357705.5 1945298.8 43.28%
Total Private Services-Affiliated 25670.2 35340.6 37.67%
Total Private Services-Unaffiliated 31048.8 53285.4 71.62%
Financial Total 9623.2 11105.6 15.40%
Financial-Affiliated 6130.7 5192.0 -15.31%
Financial-Unaffiliated 3492.5 5913.6 69.32%
Insurance Total* 9593.9 25064.2 161.25%
BPT Total 28238.2 42913.2 51.97%
BPT-Affiliated 19462.0 29868.1 53.47%
BPT-Unaffiliated 8776.1 13045.1 48.64%
Imports of total private services = 11.85% and 12.95% of trade in all industries
in 1999 and 2005 respectively
*Insurance transactions are considered unaffiliated by BEA
BPT-Business, Professional, and Technical Services

Trade and FDI in intermediate business services
final consumption
agriculture manufacturing
business
services
headquarters office
value added by skilled and unskilled labor
geographically
separation
= trade in
services
geographical
separation
= foreign investment
in services
geographically
separation
= trade in
services
Figure 17.4: Modes of trade in services for a northern
service firm
Services provided to North MAN and AG industries
Headquarters of a service
firm headquartered in
North
always
feasible
mode 1, 3
trade in S
North office South office
Services provided to South MAN and AG industries
mode 1
trade in
S
mode 3
trade in S

Chapter 10
DISTORTIONS AND EXTERNALITIES AS DETERMINANTS OF TRADE
1
10.1 Departures from our stylized world
Chapters 7, 8, and 9 discussed in some detail two of the five determinants of trade described in
Chapter 6, namely, production function (technology) differences, and endowment differences. In this
chapter we will discuss domestic distortions as a determinant of trade. The term distortion refers to
departures from our highly stylized world which is characterized by (1) all agents are perfectly
competitive and have no market power, which in turn generally implies that firms have constant returns to
scale as we will discuss in the next chapter; (2) prices and quantities are freely determined by the forces
of supply and demand; and (3) all interactions are though markets, meaning that there are no externalities.
An externality in turn is an effect one agent (firm or consumer) has on another which is not priced
through a market. Pollution is a classic example of a negative externalities, also called a spillover, in
which the actions of one agent harm another and the latter is not compensated for this. A knowledge
spillover in which the innovation of one firm is copied by other firms without the former being
compensated is a classic example of a positive externality. Externalities, spillovers and other distortions
are also often called “market failures”.
The principal focus of the chapter is to understand how distortions and externalities by
themselves can be causes of trade and whether or not that distortion-induced trade leads to gains from
trade. The chapter retains the focus on the positive theory of trade and does not focus on normative issues
such as what is optimal government policy in the presence of market failures; some of that will be
considered later in the book. Words like distortion and market failure inevitably sound like bad things
and they are relative to a best possible outcome. But keep in mind that the terms refer both to failures in
which too much of something is produced such as pollution and in which too little of something is
produced, such as knowledge.
There are two distinct parts of the chapter. The first part focuses on taxes and subsidies. The
examination of taxes and subsidies is intended as an example of the effects that a wide range of
government policies can have on trade. For example, environmental policies and regulations impact on
firms costs, and hence have effects on outputs and trade similar to those produced by taxes. We are not
asserting that commodity and factor taxes rank with factor endowments as a cause of trade, but we do
believe that collectively, government policies have a much more profound impact on trade than is
suggested in most international trade textbooks. One theme of the chapter is that government policies can
generate trade, but it not necessarily beneficial trade.
As before, the approach will be to neutralize other factors so that a clear understanding of the
specific effects of each can be obtained. Thus throughout the tax analysis, we will assume that either (A)
we have a single country facing fixed world prices, or (B) there are two countries that are identical in all
respects. They have identical technologies, identical factor endowments, identical homogeneous utility
functions, and constant returns and perfect competition in production. In the absence of the distortions
that we introduce, the two countries would have no incentive to trade, or alternatively, the free trade
equilibrium would be identical to autarky.
10.2 Distinguishing among consumer, producer and world prices
When we introduce taxes and subsidies into the analysis, it becomes important to distinguish
prices paid by consumers from prices received by producers. Once we introduce trade, consumer and
producer prices must be distinguished from world prices, the prices at which the country can trade.
Throughout this Chapter, we will use the notation q to represent consumer prices, p to represent producer
prices, and p* to represent world prices. This notation will also be used in later Chapters, such as the

2
Chapter on tariffs.
In order to focus on trade issues, we will also make the assumption throughout the Chapter that
there is no government sector per se; the government returns all tax collections to consumers in lump-sum
fashion and/or raises all subsidies by lump-sum taxation. We implicitly assume a very large number of
consumers, with each consumer getting a check or a bill which gives to (or takes from) the consumer his
or her share of taxes (subsidies). Each consumer regards their bill or check as being unaffected by their
own purchases. For examples, if a consumer pays $1 in sales tax, the consumer gets a refund of only
$1/N of that amount where N is the number of consumers (consider the refund if there are 100 million
consumers). Thus each consumer does indeed regard the tax as raising prices, even though the tax is
return to all consumers collectively. Similar comments apply to subsidies.
We will specify taxes and subsidies in an ad valorem (percentage of value) form throughout the
Chapter rather than in specific form. t will denote a tax and s a subsidy. Ad valorem taxes are quoted as
rates. A sales tax of 5%, for example, would mean a tax rate of t = .05 in this context. (Specific taxes, on
the other hand, are quoted in monetary units per unit of the good: the US gasoline tax is quoted in cents
per gallon.) Thus the relationship between consumer and producer prices with a tax or subsidy is given as
follows.
(10.1)
A tax raises the consumer price above the producer price while a subsidy lowers the consumer price
below the producer price. A tax rate t = .05 raises the consumer price 5% above the producer price: q =
p(1.05). When there are only two goods, the effects of a tax on one good are equivalent to a subsidy on
the other good. In order to see this, consider the commodity price ratios resulting from a tax on X1 versus
a subsidy to X2.
(10.2)
We see from (10.2) that a subsidy to X2 and a tax on X1 induce the same "wedge" between the consumer
and producer price ratios.
Figure 10.1 gives autarky equilibrium at point E when there is either a tax on X1 or a subsidy on
X2. In order to focus on the distortions per se, we will assume throughout this Chapter that the tax
revenue is redistributed lump sum and the subsidy is raised by a lump sum tax. These latter assumptions
are reflected in the fact the consumption and production bundles are the same even though, in the case of
a tax, for example, the consumption bundle costs more than the value of those goods at producer prices.
The consumers pay more than the producers receive because of the tax, but then they receive an income
in excess of the value of production due to the fact that they receive the tax refund. Let D denote
consumption (demand) and X denote production quantities. For a tax on X1,
(10.3)

3
The left-hand side of (10.3) is consumer expenditure at consumer prices. The first bracketed term on the
right-hand side is income received from production (payments to factors of production), while the second
term on the right-hand side is redistributed tax revenue. Thus consumer expenditure equals consumer
income. A similar analysis of a subsidy simply requires us to change the sign of t.
Figure 10.1
Figure 10.1 illustrates the distortionary effect of the tax on X1 or subsidy on X2. Welfare is lower
at E that at the undistorted competitive equilibrium at A. The producer price ratio p is tangent to the
production frontier TT' while the consumer price ratio q is tangent to the indifference curve through E.
The tax causes the consumers to perceive Y as more costly than it actual is to produce, or a subsidy to X
causes consumers to perceive X as relatively cheaper than what it actually costs to produce.
The previous paragraph should not be taken to suggest that all taxes or subsidies are bad. First,
governments usually raise revenues in order to provide public goods that are not or cannot be provided by
markets. No account is taken of public goods in this analysis. Second, not all taxes are distortionary or as
distortionary as the commodity tax shown here. For example, in the present model, an equal ad valorem
tax on both goods would leave the relative consumer and producer prices equal. Such a set of taxes is
non-distortionary.
Finally, some government policies are imposed to correct an existing distortion in the economy,
such as an environmental externality. In such a situation, Figure 10.1 might accurately depict the effects
of a pollution tax (on X1) on production and trade, but the indifference curves no longer indicate welfare
change. Welfare may be improving due to lower pollution (i.e., there is actually a third good,
environmental quality, not shown in the diagram). More will be said about taxes in the presence of
existing distortions later in the book.
10.3 Taxes and subsidies as determinants of trade: a small open economy
Suppose that the home country faces fixed world prices. Assume also that these prices just
happen to be equal to home's autarky price ratio such that home does not choose to trade at these prices.
This is completely unlikely, but we are simply following the strategy outlined in Chapter 6: "neutralize"
all causes of trade except the one which we wish to examine. The situation is shown in Figures 10.2
where the autarky equilibrium A is also the free-trade equilibrium at price ratio p*.
Once we introduce trade, we not only have to keep track of consumer and producer prices, but
world prices as well. This in turn means that we have to specify whether a tax or subsidy is assessed on
consumption or production. In the closed economy, it does not matter since production and consumption
of each good are equal. But with trade, consumption and production are in general not equal, so it matters
which one we are taxing. In this section, we will limit ourselves to looking only at production taxes and
subsidies. These are not necessarily more common than consumption taxes, but space constraints limit
the range of distortions we can deal with here. It is also true that focusing on production taxes helps build
some intuition for the analysis of imperfect competition which begins in the next Chapter.
Consider a tax on the production of X1 or a subsidy on the production of X2. In this case,
consumers face world prices, not producers. Consumer prices and world prices will be equal to one
another, but not to world prices. The relationships among the three price ratios are given by
(10.4)
The relationships in (10.4) are shown in Figure 10.2. The producer price ratio is now greater than the

4
consumer and world price ratios, so production is shown as taking place at point Xt in Figure 10.2.
Consumption must take place along the world price ratio through Xt, and consumers now face world
prices. Thus the consumption point is given by the tangency between an indifference curve and the price
line p* through point Xt. We show the consumption point as Dt in Figure 10.3. The production tax
discourages production of X1 and leads to a substitution in production toward good X2.
Figure 10.2
Several important results are shown in Figure 10.2. First, it clearly demonstrates that government
policies such as taxes and subsidies can generate trade. However, it shows equally clearly that trade
induced by the introduction of distortions is not beneficial trade. In Figure 10.2, the country receives a
welfare loss as a consequence of distortion-induced trade. Point X* in Figure 10.2 would be the
undistorted equilibrium yielding a utility level of U*. The distorted equilibrium yields a utility level of
Ut.
This is a very important results insofar as governments sometimes decide that it would be a good
thing if the country produced and exported more of a certain good (e.g., "high tech" goods). We could
think of Figure 10.2 as a situation generated by a government deciding that it must be good to produce
and export X2. By putting on a subsidy to the production of X2 we do indeed get exports of X2 and the
government congratulates itself on the success of its project. However, exports generated by distortions
are welfare reducing (put differently, the initial level of exports, zero, is optimal).
Welfare is reduced by this distortion, because producers make privately efficient choices given
the prices they face, but they do not make efficient social choices when the do not face the true costs of
producing the commodities. But again, we should separate this welfare result from the results concerning
consumption and trade, since not all commodity taxes need be welfare reducing. Most countries have
gasoline (petrol) taxes, for example, which have the beneficial effects of reducing pollution and traffic
congestion.
10.4 Taxes and subsidies as determinants of trade: two identical countries.
Now we introduce a second country and explicitly return to our concept of the two- country, no-trade
model. Suppose that we have two identical countries, both with production frontiers as in Figure
10.1, such that the point A in Figure 10.1 represents both the free-trade and autarky equilibria for both of
the countries. Now country h imposes a production tax on X1 or a production subsidy X2. At the initial
free-trade price ratio p*, country f will wish to continue to produce and consume at A in Figure 10.3 (the
same as A in Figure 10.1) while country h will wish to shift production away from X1 toward X2.
This cannot be an equilibrium because there will be excess demand for X1 and excess supply of X2
at the initial prices. The price ratio p* must rise, giving us a new equilibrium as shown in Figure 10.3.
The fall in p* induces country h to export X2 and import X1, producing at Xh and consuming at Dh in Figure
10.3. Country f produces at Xf and consumes at Df. We see that a production tax on X1 or subsidy on X2
can indeed generate exports of X2, but this is not welfare improving trade. Country h creates trade by its
tax or subsidy, but it is not beneficial trade.
The interesting additional result that we get from Figure 10.3 is that country f is made better off
be h's tax or subsidy. Recall from Chapter 5 that the ability of a country to trade at any prices other than
its autarky prices can make it better off (and will make it better off if it has no distortions). The institution
of the tax or subsidy in country h now allows country f an opportunity to trade at prices different than its
autarky prices. This might also help us understand why country h has to be worse off. With the countries
absolutely identical, there are no opportunities for mutual gains from trade. If the distortion makes
country f better off, it must make country h worse off.
Figure 10.3

5
The implication here is that country f should be happy when it’s trading partner subsidies its
exports to h. Intuitively, the subsidizing country h is selling for less than the cost of production to the
benefit of the passive country f. Happiness in f is rarely the reaction in practice however to a trading
partner’s subsidy. In some cases, a government may simply misunderstand this gift. But our result here
is not general. In a Heckscher-Ohlin world of multiple factors of production, someone in country f is
surely worse off and will understandably make a political fuss. Second, suppose that there are three
countries, two of which export X2 to the third. Then the first two, call them h and f are competitors, and a
subsidy to X2 by country h is going to drive down the price of X2 for both countries, making country f
worse off as well.
Again, we will not provide a detailed normative discussion of this here. Our purpose is simply to
show that a distortion can serve as a basis for trade but also that trade generated by a distortion is not
necessarily good trade.
10.5 Production externalities
As noted above, externalities are another source of “market failure” yet at the same time can
imply an additional source of gains from trade. Such externalities come up in quite a number of contexts,
ranging from pollution to intellectual property. In this section, we will look at positive production
externalities among firms in an industry. As suggested earlier, this could result from knowledge
spillovers in which the innovations of one firm are quickly copied by other firms without compensating
the innovating firm. Many other cases are discussed in the literature, including the increases in the range
of intermediate goods as an industry or country grows, an idea going back to Adam Smith’s “the division
of labor is limited by the extent of the market”. One interpretation of division of labor is increases in the
number of specialized intermediate goods, a topic we will return to in Chapter 12.
Suppose that there is just a single factor of production labor, L, which is in fixed supply. Good X2
is produced with constant returns to scale by a competitive industry. Good X1 is produced by competitive
firms who perceive themselves as having constant to scale, but the productivity of their labor inputs is
positively related to the overall output of the industry. These competitive firms view total industry output
as constant, much as we assume that competitive firms view the industry price as constant and unaffected
by their own decisions. Let Xi1 denote the output of an individual firm in industry 1 and let X1 denote
total industry output, the sum over all i firms. The production side of our economy is given as follows:
(10.5)
where 0 #" < 1 is an externality parameter: " = 0 is the special case of no externality, in which case the
model reduces to the Ricardian model of Chapter 7. As just noted, each individual firm i in industry X1 views total industry output as constant. In competitive equilibrium, each firm equate the value of the
marginal product of labor to the wage rate, denoted w as in the Ricardian model. Competitive equilibrium
is then described by
(10.6)
Total industry output in X1 is given by summing the first equation in (10.5) over all i firms. We
do this and then rearrange the equation to given total industry output X1 as follows.
(10.7)

6
Since " < 1, the exponent on the right-hand equation of (10.7) is greater than one: total industry output
exhibits increasing returns to scale in its total labor input. Differentiate the middle equation in (10.7)
along with the equation for X2 output, making use of the total labor supply constraint.
(10.8)
Divide the first equation of (10.8) by the second and rearrange.
(10.9)
which is the slope of the production frontier, the marginal rate of transformation. As we noted back in
Chapter 2, the production frontier is a convex function (the production set is non-convex) reflecting the
increasing returns to scale in X1: the denominator of (10.9) gets smaller as X1 gets larger. The production
frontier for our economy is shown as in Figure 10.4.
Figure 10.4
Now combine (10.9) with the competitive pricing condition in (10.6). This gives us a
relationship between the marginal rate of transformation and the equilibrium price ratio.
(10.10)
Now we discover a second issue, in addition to the convexity issue, connected with a positive production
externality. The competitive-equilibrium price ratio is not tangent to the production frontier, but rather
cuts it as shown in Figure 10.4. The intuition behind this result is the fact that when an individual firm
expands output a little, it confers a positive productivity effect on all other firms taken together. Thus the
true or “social” marginal product of an additional worker hired is greater than the “private” marginal
product of an individual firm. Or to put it the other way around, the private cost of an addition worker
hired is more than the true social cost. The slope of the production frontier depends on the true social cost
and so it is flatter than the price ratio, equal to private marginal cost. Competitive equilibrium is at a
point like A in Figure 10.4, giving a welfare level of Ua.
While we don’t want to get into a detailed normative analysis here, we should note that this is a
case where an offsetting distortion could increase welfare. The first-best outcome in Figure 10.4 is at
point S yielding welfare Us. This could be achieved by a subsidy to X1 in effect compensating for or
“internalizing” the externality. If the externality is due to imperfect protection of intellectual property
(the innovating firms is not compensated for benefits conferred on other firms), then added intellectual
property protection will act to offset the distortion and move the country toward point S in Figure 10.4.
This is an application of what is known as the theorem of the second best:
Theorem of the second best: in the presence of one distortion, the imposition of an additional and
offsetting distortion can improve welfare.
We will return to this idea later in the book.
10.6 Trade and gains from trade in the presence of production externalities

7
Let us abstract from the issue of the price line cutting the production frontier by assuming that
there are positive externalities in both X1 and X2 and in the same degree ". Then there will be a term (1 -
") in both the numerator and denominator of (10.10) and these will cancel out, leaving the marginal rate
of transformation equal to the price ratio. In this special case, we will have an autarky equilibrium at
point Xa = Da in Figure 10.5, giving an autarky welfare level of Ua.
Figure 10.5
Now assume that there are two absolutely identical economies and let them trade. It is beyond
the scope of this chapter to given a detailed analysis of the adjustment process but in short, the outcome
shown in Figure 10.5 at Xa continues to be an equilibrium but it is unstable. A small perturbation can
send the two countries off to corners, each specializing in only one of the two goods. One country could
specialize in good X2 and the other in X1 and they could each trade half of their output for half of the other
country’s good, leading both counties to share a common consumption point at D* in Figure 10.5. Here
is our first instance of how there can exist gains from trade even between identical countries arising from
increasing returns to scale.
The outcome shown in Figure 10.5 in which the equilibrium price ratio is exactly the cord
connecting the two endpoint of the production frontier a special case requiring strong symmetry
assumptions. Suppose at this price ratio, the countries do not demand X1 and X2 in the proportions
produced, but each country wants to consume a lot of X1 and not so much X2. Then at the price ratio
shown in Figure 10.5, there will be excess demand for X1 and excess supply of X2. The outcome is going
to have to be as shown in Figure 10.6: the relative price of X1 will rise to clear the market. The country
specializing in X2, call that country h, consumes at Dh and country f specializing in X1 consumes at Df in
Figure 10.6.
Figure 10.6
Note that the outcome shown in Figure 10.6 is not the only possible one. Reversing which
country is which is also an equilibrium. Thus situations in which there are production externalities can be
characterized by multiple equilibria. In such a situation, we want to note that this phenomenon has many
implications, for development economics in particular. In the presence of multiple equilibria, a country
doesn’t want to end up in the “wrong” outcome: it wants to be country f in Figure 10.6, not country h. A
more detailed analysis is beyond the scope of this chapter. But in closing, refer back to Figure 10.4.
Suppose one country has a free market and the other one “internalizes” the externality via intellectual
property protection for example. The former country is at A while the latter one is at S in Figure 10.4. If
trade opens up, the country at S has a lower price for X1 and hence will tend to end up specializing in X1:
this country will be country f shown in Figure 10.6. Internalizing positive production externalities can
give a country an advantage in a situation of multiple equilibria.
10.7 Summary: what you should know
This chapter has turned our attention away from underlying production differences
between countries, principally differences in technologies and in factor endowments. Instead, we
neutralize these differences by assuming that technologies, factor endowments and also demand are
identical across two countries. This continues the methodology outlined in Chapter 6. Here we look at
distortions and externalities (which are themselves a class of distortions) to see how asymmetries in
distortions across countries can generate trade and may or may not generate gains from trade for each of
two countries.
One class of distortions is represented here by simple production taxes and subsidies. In such a
situation, it is important to keep track of different sets of prices, in particular consumer, producer, and
world prices. The principal result of this analysis is that production taxes or subsidies can generate trade
between otherwise identical countries, but it is a welfare-worsening trade for the taxing/subsidizing
country. This carries a very important lesson for policy makers, which is that exports should never be

8
confused with welfare. A production subsidy to a favored sector may indeed generate exports from that
sector, but these exports are being sold abroad for less than the cost of production and hence are welfare
worsening.
The second class of distortions addressed here are positive production externalities or spillovers,
arising due to some failure on the part of firms to be able to capture returns for benefits they confer on
rival firms. Examples include the lack of protection for intellectual property and innovations, and
increases in productivity arising from the finer “division of labor” in a larger market. These issues will
arise again in this book.
These positive production externalities imply aggregate increasing returns to scale even though
individual firms have constant returns technologies. As discussed back in Chapter 2, this can imply that
the production frontier is convex (the production set non-convex) or “bowed in”. This in turn can imply
positive gains from trade for each of two identical countries: with each country specializing in only one
sector, productivity rises and then can exist mutual gains from trade. The situation is not simple however,
and we touch briefly on issues like the existence of multiple equilibria, with the alternative equilibria
having very different welfare implications for the trading partners.
The next two chapters continue to look at similar, even identical economies, and analyze how
imperfect competition and increasing returns to scale offer added sources of gains beyond those arising
from comparative advantage linked to differences between countries.

9
REFERENCES
Bhagwati, J. and T.N. Srinivasan (1983), Lectures on International Trade, Cambridge: MIT Press,
Chapters 20-23.
Grossman, Gene M. and Estaban Rossi-Hansberg (2009), “External economies and international trade
redux”, Princeton University working paper.
Ethier, Wilfred J. (1979), “Internationally decreasing costs and world trade”, Journal of International
Economics 9, 1-24.
Kemp, Murray C. (1969), The pure theory of international trade and investment, New Jersey: Prentice
Hall.
Markusen, James R. (1900), “Micro-Foundations of External Economies”, Canadian Journal of
Economics 23, 495-508.
Markusen, James R. and James R. Melvin (1981), “Trade, Factor Prices, and the Gains from Trade with
Increasing Returns to Scale," Canadian Journal of Economics 14, 450-469.
Melvin, J. R. (1970). "Commodity Taxation as a Determinant of Trade." Canadian Journal of Economics
3, 62-78.
Melvin, J. R. (1969). “Increasing Returns to Scale as a Determinant of Trade”, Canadian Journal of
Economics 2, 389-402.

Figure 10.1
X2
Ua
Ue
E
A
T
mc
mc
1
2
X1
T'
q
q
1
2
Figure 10.2
X2
U*
X*=D*
Ut
Dt A
Xt
X1
*
1
*
2
p
p
*
1
*
2
p
p
Figure 10.3
X2
Uf
Dh
Xh
Df
A
X1
Uh
Xf
p* p*
Figure 10.4
X2
Ua
A
p
p
1
2
2 X
X1
Us
S
1 X

Figure 10.5
X2
U*
U D*
a
p*
2 X
X1
Xa=Da
Ua
1 X
Figure 10.6
X2
Uf
p*
Df
Dh
p*
2 X
X1
Uh
1 X

Markusen lectures-may2010

More Related Content

What's hot (20)

Similar to Markusen lectures-may2010 (20)

More from Thanh Thanh (20)

Recently uploaded (20)

Markusen lectures-may2010