Scale Economies, Product Di erentiation,

and the Pattern of Trade.

Paul Krugman, American Economic Review (1980).

1 Set up of the model:

1.1 Endowments

The model starts by assuming that there is full employment and that the number of workers
equals the number of consumers.
L: number of workers = number of consumers

1.2 Preferences

The best functional form to use in these types of model with imperfect competition is a produc-
tion function characterized by Constant Elasticity of Substitution (CES).
Krugman follows the "love for variety" approach of the Dixit and Stiglitz model (1977)1 .

n
X 1
U= ci (1) utility function (I)
i=1

Note how the utility of the individuals increases with the numbers of varieties ci they con-
sume: love for variety principle.
The term > 1 means that it exists a certain degree of sostituibility among di erent varieties.

1.3 Technology

It is assumed that there is only one production factor: labour. All goods will be produced
with the same cost function:

li = + xi (2) labour for variety i (II)

The market based on product di erentiation and with "free entry" condition brings to have
as many rms as product varieties ) n of varieties = n of rms.

The total cost function is then the following:

T Ci = w li = w( + xi ) = w + w xi

1
Roman numbers indicate the corrispondent equations in the paper by Krugman.

1
 where:

w =wage
w = xed cost
w = marginal cost

1.4 Equilibrium conditions

Product market:
The output of each good must equal the sum of individual consumptions. If we can identify
individuals with workers, output must equal consumption of a representative individual
times the labour force.

L ci = xi (3) product market equilibrium condition (III)

where:
L = labour stock
ci = consumption of variety i
xi = expenditure on variety i

Labour market:
Full employment condition: the total labour force must just be exhausted by labour used
in production.

m
X
li = L (4) labour market equilibrium condition (IV)
i=1

2 General equilibrium solution in closed economy

In order to nd the general equilibrium we have to nd the solution both on the demand side
(consumers) and on the producer side ( rms), and then match them.

2.1 The consumer problem

The consumers maximize (1) subject to the budget constraint

n
X
pi ci = m (5) budget constraint
i=1

where m represents both total income and total expenditure of a single consumer.
We solve the constrained optimization problem by using the Lagrangian

n
" n
#
X 1 X
L= ci + m pi ci
i=1 i=1

2
 (recall that because of the constant elasticity of substitution the varieties are symmetric)
Let's obtain the rst order condition (FOC) by deriving with respect to ci and setting the
result equal to 0:
1 1
rci : ci pi = 0 (6) FOC (V)

Then, solving for pi we obtain a form of the inverse demand function.

1 1 1
pi = ci (7)

2.1.1 The inverse demand function.

At this point we have to endogenize in order to obtain the nal form of inverse demand
function. First of all let's recall the budget constraint we used for the optimization problem:

n
X
: pi ci = m
i=1

Here are the algebraic passages to endogenize :

n
X
let's multiply both side of the FOC by ci and then add
i=1

n
X n
X
1 1
ci pi ci = 0
i=1 i=1

n
X
1
moving out of the summation and solving for (remembering that pi ci = m) we
i=1
obtain
n
X 1
ci
1 i=1
=
m
We can now substitue the obtained here in the price function (7) calculated above ob-
taining
1 1 1
pi = X
n ci
1
ci
1
* i=1 m

1
some trivial algebra ( eliminates, and m moves at the nominator) leds us to our nal
form of the Inverse Demand Function:
1
c
pi = n i m (8) inverse demand function
X 1
ci
i=1

3
Notation 1 Note that the price of the good i depends on the consumption of all the other
varieties. This is an important point as it means that there is also an interaction with the rest
of the market.

2.1.2 The demand for variety i.

Starting from the (inverse demand function) we want to obtain a demand function.

First of all we can rewrite the inverse demand function by raising both sides to the power
of :
1
ci
pi = ! m
n
X 1
ci
i=1

simply ng the nominator we obtain

ci
pi = ! m (9)
n
X 1
ci
i=1

n
X
we can rearrange it by mulplying both sides by pi and then add :
i=1

n
X
pi ci
n
X i=1
p1i = ! m
n
X 1
i=1
ci
i=1

n
X
Recalling the budget constraint we can replace pi ci with m (total income) and cross
i=1
multiply: !
n
X 1 m1
ci = n (10)
X
i=1 p1i
i=1

n
!
X 1
we can now replace ci as of eq. (10) in eq. (9)
i=1

ci
pi = m
m1
X
n

p1i
i=1

4
 now, again some trivial algebra (simply ng and solving for ci ) leds us to our nal form of
the demand function for the variety i.

pi
ci = n m (11) demand function for variety i
X
p1i
i=1

The demand function for variety i depends on its price, on the income level and (also) on
the prices of all the other varieties.

2.1.3 The aggregation problem

This model represent a microfounded economy with many goods and many prices. In order to
apply this microfoundation to the solution of aggregate economic issues it must be possible to
represent the microfounded economy on aggregate in terms of an aggregate PRICE INDEX (P )
and aggregate CONSUMPTION INDEX (C).
In order to do so, our price and consumption index have to respect the following condition:

n
X
P C=m= pi ci (12) aggregation property
i=1

Let's verify wheter or not this aggregation property holds in our case. The price index is
de ned as ! 1
n
X 1

P pi1 (13) price index

i=1

while the consumption index is de ned as

n
!
X 1 1

C ci (14) consumption index

i=1

Now we multiply P C in order to see if the aggregation property holds:

n
! 1
n
!
X 1 X 1 1

P C= pi1 ci
i=1 i=1

n
!
X 1
1
m
Recalling from (10) that ci =X
n , we can rewrite P C as
i=1 p1i
i=1

0 1 1
1

n
! 1
B C
X 1
B m1 C
p1i B C
BXn C
@ A
i=1 p1i
i=1

5
 1 1
Note that 1 = 1 , and thus simplifying we obtain
1
m1 1
that is equal to m
n
X
We can thus conclude that the aggregation property P C=m= pi ci is respected.
i=1
Finally we can write the demand function for variety i (equation 11) in terms of the price
index P :

pi
ci = m (15) demand function for variety i in terms of price index
P1

Note that consumption of variety i decreases when its price pi increases, while it increases if
the price of the other varieties P increases, the more so the more substitutable are the varieties
(large ), due to the competition among substitutable varieties. Of course, consumption also
depends on the income m. We note again that the interaction between di erent varieties in
shaping the demand function is fundamental in this model

2.2 The rm problem

In order to solve the rm problem we have to go through 3 steps: the pro t maximization by
the rm, the holding of the free entry (zero-pro t) condition of rms and the derivation of the
number of rms in the market.

2.2.1 First step: pro t maximization.

The rm maximizes the following pro t function (with respect to xi ):

max i = pi xi wli (16) pro t function

xi

subject to both
li = + xi (17) labour for variety i (II)

L ci = xi (18) product market equilibrium condition (III)

Using (17) the pro t function becomes

i = pi xi w w xi

We derive the rst order condition:

dpi
xi + pi w=0 (19) FOC
dxi

Collecting pi we obtain

6
 0 1
B C
B C
w = pi 1 + dpi xi
dxi pi = pi B
B1
1 C
C
@ dxi pi A
dpi xi
| {z }

then, solving by pi we nally get the optimum price for the variety/ rm i.

pi = w (20) price for variety i

| {z 1}

Notation 2 Since > 1 it follows that 1 > 1, hence the latter term can be interpreted as a
mark=up over marginal costs w. In a monopolistic competition model, each rm is a monopolist
on its own variety (one rm = one variety), and thus there is some market power, with rms
able to charge a positive mark-up to their prices. Note that if goods are perfect substitutes then
would tend to 1 and the mark-up term would tend to 1. In this case, we are close to perfect
competition, with rms not able to charge any mark-up to their prices. If instead tends to 1
the mark-up is maximum, since there is no competition among varieties.

2.2.2 Second step: free entry condition.

The model assumes that there is a free entry condition in the market. Firms will then
enter the market as long as the pro ts are positive.

The free entry condition holds when:

i = pi xi w w xi = 0 (21) free entry condition

By the pro t maximization price, equation (7) in the paper from Krugman, we know that
pi = w 1
we can then replace pi in the free entry condition, and after having made some simpli cation
we get
i = xi xi = 0
1
and, nally, solving for xi we get the optimal quantity produced under the free-entry (F E)
equilibriumt
( 1)
xFi E = (22) minimum operation scale

It exists a minimum scale of operation that permits to a rm to reach the "break-even"

point. If a rm produces less than xFi E its pro ts will be < 0, and the rm will be forced to
exit the market. As it was predictable the minimum scale of operation depends positively on
, that represents the xed cost of production. Another important notation is that w (wage)

7
does not appear in the equation. This means that the wage a ects the prices set by rms but
does not in uence the minimum scale of operation.

2.2.3 Third step: number of rms in the market

We want now to derive the number of rms that operate in the economy in terms of labour
stock (L), xed cost ( ) and .

In order to derive the number of rms we use the only condition we have not used yet,
that is the labour market clearing condition: labour force = labour demand.

n
X n
X
L= li = ( + xi ) (23) labour market clearing condition
i=1 i=1

Recalling equation (17) the aggregate labour demand is equal to the summation over all
the varieties of the labour demand for variety i:

We can now replace xi with its value found in the minimum operation scale equation:

n
X n
X
( 1)
L= + = =n
i=1 i=1

solving for n we can nally get the equation of the size of the market (number of rms/varieties
in the country)
L
n= (24) number of rms/varieties (X)

We can draw some conclusions on the variables that determine the number of varieties/ rms
in a country.
- Large economies, that have large labour supply (L), have more rms and produce more
varieties and, as a consequence, achieve higher welfare levels.
- High xed costs ( ) make the minimum operation scale rise and thus lead to less rms
(characterized by bigger dimensions) / varieties in the country.
- A relatively high elasticity of substitution ( ) would lower the number of operating rms
(varieties), as consumers value less the consumption of di erent varieties (recall that if !1
varieties are perfectly substitutable). However, prices would tend to perfect competition levels:
this is a feature of monopolistic competition!

8
 Appendix: the elasticity of demand to the price.

dci pi
Let's calculate now the elasticity of demand " = dpi ci by deriving the demand function
pi
obtained for pi and multipling it by ci :

1
dci pi pi (1 )pi pi
"= = n + pi !2 m
dpi ci X n
X ci
pi1 pi1
i=1 i=1

1
p pi
we can collect the following term: X
n
i
ci and obtain
p1i
i=1

0 1
B C
pi B pi
B
1
pi C
"= (1 ) n i BX mC
X @
n
ci C
A
pi1 p1i
i=1 i=1

0 1
B C
p B pi 1
pi C ci
as from (11) we know that ci = X
n
i
m we can assume that B X
n ci mC = ci =1
@ A
p1i p1i
i=1 i=1
. So the previous function becomes

pii
"= (1 ) n 1 (25)
X
pi1
i=1

recalling equation (9) and mutil plying both sides of it by pi we get

ci
p1i = pi ! m
n
X 1
ci
i=1

we can now replace p1i with its function in (25) obtaining.

9
 ci
pi 0 n 1 m
X 1
@ ci A
i=1
"= (1 ) 2 3 1
n
X 6 7
6 ci 7
pi 6 0 n 1 7m
4 X 1 5
i=1 @ ci A
i=1

simply ng equal terms we get

pi ci
"= (1 ) n
X
pi ci
i=1

pi ci
It is easy to note that X
n represents the market share of the variety i: i.e. the expenditure
pi ci
i=1
on variety i / total expenditure.
We can use the notation Si for the market share. Finally our elasticity of demand is the
following:

"=j (1 ) Si j (26) elasticity of demand

The fact that the market share of each variety enter into the elasticity of the demand is an
important characteristic of this generation of models.
It is the formalization of the strategic interaction between varieties/ rms in the model.
1
We also know that, from the moment that varieties are symmetric, Si = n
Then the elasticity of demand depends on the number of varieties/ rms.
In the case of perfect competition n ! 1:In that case the u :From the moment that >1
by assumption (because of substitutability among varieties), in the case of perfect competition
the demand is elastic.
In the case of monopoly n = 1 . In that case the elasticity is unitarily elastic " = 1:This
means that when only one variety is produced in the market a one percent change in the price
of that variety generates a one percent change in quantity.

Comparison with the original Dixit-Stiglitz continuum version.

The utility function in the Dixit- Stiglitz model is

Zn
1
U= c(i) di
0

10
 From the optimization of this utility function we obtain the following demand:

pi
c(i) = Zn p(i) m instead of X
p1i
p(i)1 di
o

Obviously the di erence is that in the continuum version we have the integral insted of the
summation.
The elasticity in the continuum case is
Zn
d p(i)1 di

0
"= dp(i) =0

This result signi es that in the original Dixit-Stiglitz model there would be no strategic
interaction between varieties/ rms.

11
3 The open economy model
We now characterize an open-economy version of the model. As our goal is to use the model to
explore the issue of regional disparities, we proceed as Krugman itself did, that is we integrate the
1980 open-economy version of the model with some new features which allow for an interpretation
of the results in terms of regional market shares. In so doing, we incorporate in particular features
from Krugman (1991 - Journal of Political Economy)
We suppose two countries identical in size, tastes and technology that open up to trade.
As before, in each country the economy produces a di erentiated good, with varieties being
country-speci c (i.e. varieties are di erent across countries: a green short-sleeved French shirt is
di erent from a green short-sleeved American shirt2 ). Trade will occur because, in the presence
of increasing returns, each good (i.e., each di erentiated product) will be produced in only one
country - for the same reasons that each good is produced by only one rm. Gains from trade will
occur because the world economy will produce a greater diversity of goods than either country
alone would, thus o ering each individual a wider range of choice.

3.1 Endowments

The total factor endowments of the two countries in terms of labour (the only production factor
considered in the model) is

Lw = L + L

that can be split into:

the labour endowment of the home country: L = SL Lw

and the labour endowment of the foreign country: L = (1 SL )Lw

where S denotes the size of the home country and indicates from here on the foreign
country variables.

3.2 Preferences

The aggregate utility function (for the home country) is

U = C A1

where A represents the consumption of an aggregate homogeneous good, C is the consump-

tion index previously de ned and is the fraction of income devoted to the consumption of the
di erentiated good.
In particular, recalling the utility function derived above we have:

2
The French green shirt is likely to have a crocodile etched on the right chest, the American one a polo player,
and you are happier if you own both of them...

12
 "n+n #
X 1
1

C= ci
i=1

where the fact that the summation now goes from 1 to n + n indicates that individuals can
(and are willing to) consume both the goods produced in the home country (n) and the goods
produced in the foreign country (n ).

3.3 Technology

The technology of production is the same as in the closed economy case.

However, we introduce some transport costs in the model. Transport costs are assumed
to be of the of the "iceberg" type > 1 . The idea is that during the transportation some
1
part of the product "melts" away so that only a percentage of its value e ectively arrives at
destination. In other words, if you write a contract stating that a quantity q has to reach the
foreign consumers, you have to ship (and thus produce) q, with > 1.

3.4 Equilibrium conditions in open economy

In the open economy context, with both the homogeneous and the di erentiated goods present,
the labour market equilibrium condition becomes

n
X
L = LA + li (27) labour market condition
i=1

that is, labour in the home country can be used in the production of homogeneous goods (for
Xn
which you need a quantity LA of labour) or in the production of di erentiated goods ( li ),
i=1
and symmetrically in the foreign country. Note that labour is not mobile across countries.
This model does not consider migration.

4 General equilibrium solution in open economy

In order to derive the general equilibrium solution in the case of free trade, we have to go through
all the passages we made in the case of the closed economy: solution on both the demand side
(consumers) and the producer side ( rms) of the model.

4.1 The consumer problem

Having already discussed the aggregation problem we are able to solve the consumer problem
in the aggregate context.
The aggregate income E can be allocated across the two aggregate goods A (the aggregate
homogeneous good) and C (the aggregate di erentiated good).
The consumer maximize his utility subject to two budget constraints:

13
 1. The rst one refers to the share of income devoted to the aggregate di erentiated good
C:
n+n
X
E = PC = pi ci (28) budget constraint I
i=1

As before, note that the consumer in Home has also in his budget constraint the consump-
tion of foreign varieties (n ).

2. The second b.c. refers to the share of income (1 ) devoted to the aggregate homogeneous
good A
(1 )E = pA yA (29) budget constraint II

Knowing the fraction of income devoted to the consumption of the di erentiated good in
Home, substituting the b.c. in the Lagrangian we can easily derive the domestic demand
function for any variety i, as we did for the closed economy (equation 11)3 :

pi pi
di = E= E (30) demand function for variety i in the home country
n+n
X P1
p1i
1=i

In particular, given the fact that we are now dealing with domestic and foreign varieties, we
have replaced the sum with a (new de nition of) price index:
n+n
X
P 1 = p1i = np1i + n pi 1
i=1

Also note that, due to the de nition of P which encompasses n , the entire income of domestic
consumers E is not spent on the domestic variety, part of it is spent on the foreign variety.
Simmetrically, however, foreign consumers will spend part of their income E to demand our
variety i with a demand function:

pi
di = E (31) demand function for variety i in the foreign country
P 1

n+n
X
where P 1 = pi 1 = npi 1 + n pi1
i=1

However we already know that production of a given variety, wherever it is consumed, is

made by only one rm in H or in F (due to increasing returns) for both markets. Hence, when
a rm has to decide how much to produce of her variety, she will need to take into account both
demand functions, a problem to which we now turn.

3
Note that di , the demand for variety i arising in Home, is di erent from ci , as the latter is the total demand
of variety i (we are in a closed economy), while the former is only part of the total demand, since also foreign
consumers can demand to consume the variety i.

14
4.2 The rms problem

As in the closed economy case, we split the rm problem in three steps:

4.2.1 First step: pro t maximization

The rm's problem for the homogeneous aggregate good is trivial. Being the good homo-
geneous, the market is characterized by perfect competition. As a consequence the price
must equal the marginal cost of production in both countries: pA = w and pA = w

Assuming that factor price equalization holds in equilibrium, salaries are equal across
countries and we can assume that:

w = pA = pA = w = 1 ((32) Homogeneous good price)

With respect to the di erentiated good, rms now produce for both the local market
and the foreign market. In particular, rms maximize their pro t function subject to the
following constraints:

1.
xH
i = di (33) constraint I

that is, the production of the good i for the domestic market equals local demand. There
is no transport cost for the home market.

2.
xFi = di (34) constraint II

that is, the production of good i for the foreign market (xFi ) is larger than the quantity
that e ectively arrives to the consumers in the foreign country (remember that > 1).
This is because of the transport costs. Due to transport costs, rms in the home country
have to produce more than what they e ectively sell in the foreign country.

3.
xH F
i + xi = xi

This last constraint simply states that the sum of the productions for the domestic and for
the foreign market must equal the total production for a given variety i.
Taking into account the constraints above we can rewrite the pro t function as follows:

xFi
i = pi xH
i + pi w( + xH + xF ) (36)
| {z } | {z } | i {z i}

In this open economy setting rms have to decide the optimum production level for both
goods allocated in the domestic market xH F
i and goods allocated in the foreign markets xi .

15
 Maximizing with respect to xH
i we obtain the FOC for the home market production. The
result is of course equal to the result obtained for the closed economy case (equation 20):

pi = w (37) FOC I home market production

1

The solution of the maximization problem for the production allocated in the foreign country
is similar, the only di erence is the presence of the transport cost :

pi = w (38) FOC II foreign market production

1

Comparing the two optimal prices for the variety i we thus have that:

pi = p i

The solution of the pro t maximization problem of the rms leads us to the conclusion that
for the same variety i the optimum price in the foreign market is higher than the optimum price
in the home market (where the variety i is physically produced). This result is generated by the
fact that trade is costly (transport costs).

4.2.2 Second step: free entry condition

Also in open economy the model assumes that there is a free entry condition in the market.
Firms will then enter the market as long as the pro ts are positive. If we substitute pi
and pi into (36) the free entry (FE) condition appears as the following:

xFi
i = w xH
i + w w( + xH F
i + xi ) = 0 (39) FE condition
1 1

Reorganizing the terms and applying some trivial algebra yields:

( 1)
xFi E = xH F
i + xi = (40) minimum scale of operation

We obtained the minimum scale of operation of rms under free entry (F E) in the open
economy setting. What is interesting of this result is that this equation is the same that we
found in the case of closed economy (equation: minimum operation scale (22)). The openess to
trade and the transport costs do not in uence the minimum scale of operation that a rm needs
to have in order to break-even. This is a peculiar result of the model, with gains from trade
accruing only from the higher number of varieties, as explained in the following section.

4.2.3 Third step: number of rms in the market.

The nal step in deriving the equilibrium solution of the model is to nd the number of
rms/varieties in the two countries after countries opened up to trade. In order to derive

16
the number of rms/varieties in equilibrium we apply the market clearing condition for the
di erentiated good C: the supply of good C must equal demand.

xi = di + di (41) market clearing condition for good C

We need to use three equations we found before:

( 1)
-The equation (40): minimum scale of operation: xi =
pi
-The equation (30): demand function for variety i in the home country: di = P1
E
pi
-The equation (31): demand function for variety i in the foreign country: di = P1
E

We also know that total expenditure E must equal income L : E = wL = L because w = 1:

Than for symmetry in the foreign country we have E = L

Recalling the de nition of the aggregate price indexes for the home and for the foreign
country, we can rewrite the market clearing condition (41) using (30) and (31) and substituing
P1 and P 1 with their functions:

( 1) pi pi
xi = = L+ L = di + d i
np1i + n pi 1
npi 1
+ n pi1
| {z } | {z }

Now let's multiply everything by pi and substitute pi with pi :

( 1) pi1 p1i
xi = pi = L+ L (42)
np1i +n 1 pi1 n 1 p1i + n pi1

( 1)
Recalling from FOC I (home market production (37)) that pi = 1 our pi
becomes : We can now, after some simpli cations, rewrite (42) for both domestic and foreign
varieties, considering that 1 = 4:

L L
= + (43) domestic variety
n+n n +n

L L
= + (44) foreign variety
n +n n+n
Now we solve rst for the total number of rms and then for domestic and foreign number
of rms.
The total number of rms is oviously the sum of the rms in the two countries: nw = n + n :
At this point we multiply the condition for domestic varieties for n , and the condition for
foreign varieties for n and nally sum up the two conditions. After the summation many terms
eliminate and we get the following equation for the total number of rms:

(n + n ) = (L + L ) leading to:

4
For this reason, trade models often refer to the "phi-ness" or " ness" of trade. Note that it ranges from 0,
where and are high enough to eliminate all trade, to 1, where trade costs are negligible (recall that 1 < 0)

17
 Lw
nw = (45) total numer of rms/varieties
L
Recalling the equation for the total number of rms in the closed economy (n = ) we
observe that the latter is really similar (apart from the fact that now only a fraction of the
income is spent in di erentiated goods). Also in the open economy the number of rms/varieties
is proportional to the factor endowments (labour force). Nevertheless in open economy the
number of varieties/ rms is bigger because Lw > L.

Gains from trade in the Krugman's model.

The economic intuition of the model is the following: the gain from trade in this setting
derives exclusively from a "variety" e ect. When countries open up to trade consumers
can buy goods from both countries. As consumer utility increases with the number of
varieties (love for variety) the welfare with trade increases too. Consumers prefer to spread
consumption across varieties, and they can do that better in open economy because they
can choose among more varieties available. Another important conclusion of the model is
that there is no "selection" e ect neither "scale" e ect on rms due to trade. All the rms
producing before liberalization produce also after and maintain the same size. In other
words, trade grows through an "extensive" (n. of trading rms) rather than an "intensive"
(average trade per rm) margin.

Notation 3 These conclusions di er by the ones suggested by Krugman in his previous

paper (Increasing returns, monopolistic competition, and international trade; Journal of
International Economics, 1979). In the 1979's model, indeed, trade is fully driven by
economies of scale lowering prices of the goods: rms after opening to trade are able
to expand their output and thus to reduce unitary production costs by better exploiting
economies of scale. The gains from trade are generated by a \scale" e ect that is not
present in the 1980's model. Trade thus grows along the intensive rather than the extensive
margin.

4.3 The pattern of trade

The last step of the analysis is to derive the distribution of rms between countries and subse-
quently countries' specialization.
Let's recall that:

- Lw = L + L

L
- that the "size" of the country is described as SL = L+L

n
- that the share of home varieties on the total is Sn = n+n

18
 Taking equations (43) and (44) and rewriting them using the equations of Lw and SL recalled
above we get:

SL Lw (1 SL )Lw
+ (46) home country
[Sn + (1 Sn )] nw [ Sn + (1 Sn )] nw

(1 SL )Lw SL Lw
+ (47) foreign country
[ Sn + (1 Sn )] nw [Sn + (1 Sn )] nw
Lw and nw eliminate. After some sempli cations we can solve for Sn and we get:

Sn = 1 (48) home country specialization

1 + 1 SL

than, subtracting 1/2 from both sides we obtain a measure of "deviation" in terms of pro-
duction:
1 1+ 1
Sn = (SL ) (49) deviation in terms of production
2 1 2
1
where SL 2 represents a measure of the size of the country, in particular the deviation in
terms of endowments.

The "home market" e ect

The equation of the deviation in terms of production leads us to one of the most important
prevision of this model on the pattern of trade: the so called "home market" e ect.

A country will specialize in the production of those goods for which it has the larger home
market (domestic demand). This result is in contrast with the predictions of classical trade
models. In the traditional trade models indeed a country that has a bias in the consumption
of a good becomes a net importer of that good. On the contrary in the framework of
Krugman (1980) the prediction is exactely the contrary: a country with relatively larger
home market for a given good will be a net exporter of that good. The extremely important
results obtained by Krugman are driven by two fundamental assumptions: the existence
of increasing return to scale and of transport costs. In a world as the one described by
Krugman there is a clear incentive to concentrate production of a good near its largest
market, in order to exploit scale economies and minimize transport costs.

Depending on the size of the home market of the two countries and on transport costs, we
will have di erent results concerning the pattern of trade. For a given level of transport
costs, if the two countries are identical, then SL = 21 , and thus we get the classical result
where the number of rms/varieties is equal in the two countries (Sn = 12 ). If transport
costs are very high to lead to autarky, then ! 0, and thus the share of production
(number of rms/varieties) in each country is proportionally to its size (SL = Sn ), i.e. n =
L
which not by chance is our closed-economy result. Nevertheless, in case of asymmetry
among countries and for given levels of transport costs, a di erent number of varieties will
be produced and the bigger country will specialize and produce proportionally more.

19
In fact, this model not only demonstrates that di erences in endowments lead to di erences
in production (each country will be a net exporter in the industry for whose products it
has the relatively larger demand). An even more important insight of the model is that the
di erence in production (number of rms/varieties) will be more than proportional than
the di erences in endowments between countries. Since ((1 + ')=(1 ')) > 1 , indeed, we
have that for positive transport costs Sn > SL . In other words, the larger country will be
disproportionately specialized in the production of the di erentiated good: the latter is,
formally, the "home market" e ect.

Complete vs incomplete specialization

Another fundamental result of this model, again in contrast with the outcomes of classical
trade models, is that it formally justi es the (empirically relevant) fact that specialization
is not complete (both countries produce for both markets), thus allowing for intra-industry
trade of varieties. In the Krugman's model there are two forces, one driving in favor of
complete specialization, the increasing return to scale, and another one that goes in the
opposite direction, transport costs. The possibility of incomplete specialization is greater,
the greater are transport costs and the less important are economies of scale.

We can see this e ect analytically by observing the e ect of di erent trade costs on the
home market e ect. From (49) we can infer that when ' (the inverse measure of trade
barriers) increases, Sn SL increases too. Reducing trade barriers thus makes the "home
market" e ect stronger. The larger is the openness to trade, the stronger will be the
"home market" e ect i.e. the specialization. Thus, one of the most important messages
of this paper is that it manages to justify the existence of trade between countries even
after controlling for technological asymmetries and factor endowments, thus leaving only
di erences in the \home market".

Theoretical implication: Krugman (1991) and the \New Trade Theory".

Now, start from the latter nding: imagine Germany and Spain trading varieties at a
given point in time. For a given level of transport costs, Germany (larger than Spain) will
produce proportionally more varieties than Spain. Now, imagine that the EU Commission
proposes to create a Single Market, removing most barriers to trade, i.e. increasing in
our model, with all the rest being the same. Will Spain accept ? What will happen to
territorial disparities in the EU ?

Working out this example, you can clearly see that the introduction of imperfect competi-
tion, economies of scale and transport costs in trade models is the rst step of a revolution
in international trade theory that will take to the birth of the \New Economic Geography".
"New economic geography" builds on the concepts of economies of scales and increasing
returns and leads to a world where the economic activity agglomerates in certain regions
and temporary di erences can generate long run e ects. Still, in Krugman (1980)'s model,
temporary di erences across countries lead only to temporary di erences in the allocation

20
of production. Following works, as Krugman (1991), will open the doors to long run e ects
of temporary di erences (or shocks). Through the process of cumulative causation, even
temporary di erences can lead to long run and persistent di erences between countries.
This represents a "Copernican" revolution in International Economics.

21