Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Dixit Stiglitz Explained PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

The Dixit-Stiglitz demand system, monpolistic competition and trade.

Economics students are generally well trained in perfectly competitive markets. Such
markets are often thought to be characterized by well defined utility functions and
homogenous production functions with constant and decreasing marginal productivities
and constant returns to scale. Such models are very useful for understanding a wide set of
economic mechanisms. Increasing returns, however, are known to be feature in many real
world production processes. Often such production technologies are said to generate
natural monopolies since an implication of increasing returns is that large production
entities are more productive than smaller ones. Still, we know that many markets are
characterized by many producers producing different varieties of the same products under
increasing returns to scale technologies. A stylized description of such markets is the
market type monopolistic competition. Such markets are characterized by many
producers who each enjoy some market power but by free entry so that profit
opportunities are limited.
Markets with monopolistic competition captures can easily be modeled with increasing
returns to scale production technologies. This has shown very useful for many
applications. Below we will consider international trade. We will demonstrate that such
markets generate trade between similar countries and trade within the same industries, so
called intra industry trade. These are known to be of great importance for real world trade
flows. Since such trade can hardly be explained by comparative advantages, trade theory
is fruitfully supplemented with monopolistic market approaches. However, monopolistic
markets models have also had wide applications in many other economics topics, as e.g.
growth theory, environmental economics, macroeconomics and microeconomics. The
modeling framework presented below therefore have many applications.
The demand side
Monopolistic competitive markets must be characterized by a demand side that captures
many product varieties. One specific approach for modeling this is to introduce a
representative consumer who always demands the existing varieties. This can be
interpreted literally so that one assumes that every consumer prefers variety in the
consumption basket. Consequently the demand system we will describe is sometimes
referred to as the love of variety approach. This may be misleading however since the
underlying utility function implies less love of variety than standard utility functions of
the Cobb-Douglas type (see below). Another interpretation is that the demand system is
for a representative consumer who is an aggregate of many consumers with distinct
individual preferences for each variety. Under some assumptions (which we will not
discuss), it can be showed that such an aggregation may give rise to the preference
structure for a representative consumer that we will introduce. The specific utility
function introduced below is one with constant elasticity of substitution and it is referred
to as a CES utility function.
The representative consumers preferences are described with the utility function:


U (q1 , q 2 ,..q N ) = q
i =1
N

Above, qi, denotes quantity of consumption good i and is the elasticity of substitution
among varieties. Generally we will assume that >1.
If =, the utility function is just the sum of consumed quantities of each variety:

U = q
i =1
N

N
= qi ,
i =1

when =

The reason for this is that if =, both the exponents (within and outside the bracket)
goes to one. Therefore the case when = describes the case when the goods are perfect
substitutes. In this case, consumers do not care if one good is substituted for an equal
quantity of another.
If =1, the above utility function goes to infinity. A variety of the CES utility function
were one imposes equal weights on all varieties converges to a Cobb-Douglas utility
function for N goods when =1. We will rewrite the utility function with the use of the
parameter =(-1)/. Now if =1, =0 (and if =, =1)

U = q
i =1
N

N
= qi
i =1

Taking the limit of the above when approaches 0 gives:


1

N
lim U = qi =
i =1

Now consider the function V below. In V, each element within the summation term are
given equal weights.

N 1
V = qi
i =1 N

Take the natural logarithm of the above expression to obtain:

N 1
ln qi
i =1 N
N 1

ln V = ln qi
=
i =1 N

Letting approaching zero gives rise to an expression of the type (0/0). This can be
handled with lHopitals rule which allows derivation of the nominator and the
denominator with respect to . This produces the expression:

N 1
ln qi
i =1 N
lim ln V = lim
0

qi ln qi

1
=" 0 " = lim N
= ln qi
0 0
N
1
N qi

lim ln V =
0

1
ln qi
N

N ln qi
1

V =e

The above expression is a Cobb-Douglas function for consumption of N goods where


each has an expenditure share equal to (1/N). We have dropped the subscripts and
superscripts from the summation sign above and will only reintroduce it when necessary
henceforth. This proves that a CES function of the type V approaches a Cobb-Douglas
function when goes to zero. For many maximization purposes, one uses the ratios of
marginal utilities. These are the same for the two functions U and V.
The preference for variety can easily be seen from the utility function if we assume that
all goods have the same price so that pi=pj=p, and are consumed in the same amount so
that qi=qj=q. In this case, the utility function can be written:

N 1 1 1 1
1

= Nq
=N q
U = q
i =1

Writing consumers income as Y, we know that Npq=Y and therefore that q=Y/Np.
Substituting this into the above gives:

U=N

q=N

1 Y
Y
Y

= N 1 = N 1
Np
p
p
1

Since >1, this implies that U increases with the available number of varieties. The
higher is , the less does utility depends on the number of varieties. This is in line with
intuition.

Now assume that incomes are generated from a mass of labour L that earns w per unit.
This will generalize to total income in an economy populated with L workers. Therefore
the budget constraint is wL. Consumers want to maximize their utility given this budget
constraint. The corresponding Lagrange function is:

1 1

L = qi ( pi qi wL )

First order conditions for utility maximization are:


1


L
=
qi
qi 1
1

qi

1
1

1
qi

p i = 0

qi pi = 0

1 1 1

qi q j p j = 0

In the second line above the first order condition was somewhat simplified. In the third
line the first order condition is repeated for another good, j. Maximization requires that
similar conditions are fulfilled for all goods.
From the first order conditions we obtain:

qi

qj

q
= i
qj

p
q
i = i
q j p j
p
qi = q j i
p
j

pi
pj

p
pi qi = q j pi i
p
j

= q j p j pi1

p
pi qi = q j pi p i
j

= q j p j pi1 = q j p j pi1
i =1

The second line is simply a reorganization of the first. In the third line we have solved for
qi. In the fourth line we multiply with pi. In the fifth we sum over all qi. The second
equality reminds us that this summation is over i. Therefore elements in the sum with
subscripts j can be multiplied outside the summation sign. From this we can solve for qj:
pi qi

qj pj =
pi1
p j pi qi

qj =

1
i

p j wL

1
i

since

pq
i

= wL

The second line is the demand function for good j. It depends on the price, pj, on total
income, wL and on an expression that is a function of all prices in the economy. This is a
function of a price index for all goods given by:
P=

( p )

1
1 1
i

Therefore the demand function can be written as:

qj =

p j wL
P 1

Note that elasticity of demand is just :


dq j p j
dp j q j

p j 1 wL p j
P 1

qj

p j wL

P 1

P 1

p j wL

The above implies that one term is neglected. That is the influence of firmss pricing
decisions on the price index P. This is a valid approximation when there are many firms.
When there are few firms however, the neglecting of each firms influence on the price
index is not valid. We will continue to ignore this term.
Why is P a price index for all prices? It can be shown that this is the cost for obtaining
one unit of utility. Form the first order conditions we have:
p
q
i = i
q j p j
p
qi = q j i
p
j

Insert this expression into the utility function:

N 1
U = qi
i =1

N p
= q j i
p
i =1
j

N
1
= q j p j 1 pi1
i =1

N 1

= q j p j 1 pi1
i =1

N
1
1

=
q
p
p

j
j
i

i =1

Again we have set the elements containing subscript j outside the summation since
summation runs over i. From the above, solve for qj:

N
1
q j = p j pi1 U
i =1

N
1
p j q j = p 1j pi1 U
i =1

N
1
N 1 1
1
1
p
q
p
p
U
=
=

pi U

i
j j
j
j =1
j =1
i =1
i =1

N
1
E = pi1
=P
i =1

In the second equation we multiplied the solution for qj with pj. In the third line we
summed over all js. Since the sum over all js and over all is involves summing over all
prices the sums can be expressed as in the last equality in third line. In the second line we
set U=1 and write the corresponding sum as E. These are the expenditures for one unit of
utility and therefore the price index for all goods.

The production side


It is important the demand functions that we derived above has the price elasticity .
Since >1 this implies that firms facing these demand functions have some market
power. It is well known that when there are increasing returns in production, marginal
costs are lower than average costs. Therefore perfectly competitively pricing where price
equals marginal costs cannot sustain production when there are increasing returns to
production. We will allow firms to price monopolistically in our model. Therefore prices
are set above marginal costs and there are operating surpluses. This allows increasing
returns since the operating surplus can cover also fixed costs in production. The

monopolistic competition market form also assumes that there is free entry to the market.
Therefore profits are assumed to be squeezed away.
Firms are assumed to produce one variety each. Since there are increasing returns to
scale, production costs are lower when they concentrate on one product variety only.
Since producers of single varieties have market power it does not make sense to produce
the same variety as other firms. Thus firms produce distinct varieties.
We will assume that labour is the only factor of production. Production takes place under
increasing returns to scale. In order to operate firms have to use the fixed amount of
labour f. The cost of this, wf, is therefore a fixed cost component. In addition production
also requires b units per product. The use of labour in production of variety j is therefore
given in the first equation below. Total costs are given in the second equation.

l j = f + bq j

C j = w( f + bq j )
Note that firms are assumed to have similar cost structures so that w, b and f are similar
across firms.
Profits are therefore:

j = p j q j w(bq j + f )
Remember that demand is given as
qj =

p j wL
P 1

Profit maximization therefore requires setting the derivative of the profit function with
respect to qj equal to zero:

d j
dq j

= pj +

dp j
dq j

q j wb = 0

dp j q j
= wb
p j 1 +
dq p
j
j

1
1

p j 1 = p j
= wb


pj =

wb

j = p j q j w(bq j + f ) =

( 1)
wbq j w(bq j + f ) = w
bq j f = w
bq j f
1

Above we also solved for the optimal price in the fourth line. Note that prices are set as a
markup over marginal costs (/(-1)>1). In the fifth line we inserted the resulting price
into the profit equation and found the resulting expression for profits. Note that prices are
equal for all firms.
The assumption of free entry implies that profit opportunities are eliminated. We can
solve for each firms produced quantity by setting profits equal to zero:

1
bq j f = 0

j = w

1
bq j = f
1
1
qj =
f
b

The above equation depends only on constants that are common for all firms. We have
therefore established that all firms produce the same quantities.
We have determined both common prices and common produced quantities. It remains to
determine the number of firms and thus the number of product varieties available in the
market. This can be done using the labour market equilibrium:

Nl = L
N (bq + f ) = L

1
Nb
f + f = N (f f + f ) = Nf = L
b

L
N=
f

Welfare
The above model can be interpreted as a general equilibrium model for a closed
economy. Consumers demand consumption goods according to their demand functions.
The demand functions are derived from the underlying utility function. Consumers earn
their income from being part of the labour force. Firms employ workers for production of
their individual varieties, produced under increasing returns to scale. The firms price
monopolistically but free entry ensures that profits are squeezed away. Since firms are
symmetrical, their price and produced quantities are similar. The number of firms is
determined by the available labour force.
We noted when we introduced the utility function that when pi=pj=p and qi=qj=q the
utility function can be written:

N 1 1 1 1
= Nq
= N 1 q
U = q
i =1

Consumers income is wL, and we know that Npq=wL and therefore that q=wL/Np.
Substituting this into the above gives:
U=N

q=N

1 wL
wL
wL
= N 1
= N 1

p
p
Np
1

Inserting for the number of firms and prices give:

U=N

1
1

wL

= N
p
1

L 1 1 f 1

1
1

wL L

=
p f

wL

wb

( 1)
b

Since >1, this implies that U increases with the available labour force. The higher is ,
the less does utility depends on the labour force. In this model, increased population
implies that a larger number of product varieties is on the market. Each firm produces the
same amount. This is a consequence of the specific utility function that is used. In

Krugman (1979) and in Feenstra (2004) a similar model to the one above is used, but
with a more general utility function. In that case, the elasticity of substitution depends on
the consumption of each good. Therefore also the demand elasticity depends on
substation. When demand elasticity increases with consumed quantity, increased
population results both in more varieties and increased production of each variety and
therefore lower costs (because of increasing returns) and lower prices.

Trade
The model sketched above was for one economy only. One can analyze trade within this
model by assuming two countries described as above. Transition from trade to autarky
within that model involves expanding the size of the two economies. From the welfare
considerations above it seems clear that this increases welfare. More product varieties
become available. For a more general model, the effects of trade may potentially be
several:

Trade increases market share and competition. This could imply both increased
production per firm so that average costs decrease and reduced prices.
Trade increases the number of firms so that more product varieties become
available. With the CES utility function we have used, effects of trade is limited
to this mechanism. In the expressions for firms output, produced quantity does
not depend on market size. In this model therefore, firms production is the same
with and without trade, and trade merely involves more products being available.

Also, the model we have described is a one-sector model. The product varieties are
imperfect substitutes but are of the same types. Consumers consume some of all varieties
available. Trade between two countries is therefore intra-industry trade, i.e. trade within
the same industries.
Suppose that there are two countries characterized as in the model above. Assume that
their technologies and their preferences are identical. The two countries are only allowed
to differ in size which are LH and LF, respectively. Since firms and prices are identical,
we drop the firm subscripts from now. Subscripts will instead be used to indicate
counties. Since prices are the same also wages are the same, so that w=wH=wF.
Prices, quantities and the number of firms are given by the equations:
p=

q=

wb

b
L
Ni = i
f

i = H,F

As we have noted, consumers consume some of all varieties available, also of varieties
produced abroad. Thus exports from country i to country n is:

X in = N i pin qin = N i pi

pin wLn
Pn1

pin wLn
qin =
Pn1
The first line above gives exports as the value of sales per firm from country i to country
n, (pinqin), times the number of firms in country i. In the second equation in the first line
we have inserted the demand function. This is given in the second line. Demand for a
good produced in country i in country n is given by the same demand function as derived
above, but with due notification that income and the price index is for country n. The
price index in country n is given by:

Pn = N i pin1 + N n p 1nn

1
1

The fraction of total income in country n spent on goods from country is

in =

N i pin qin
Li
X in
=
=
N i pin qin + N n p nn q nn Li + Ln
Xn

The first equation simply defines income in country n as Xn in the denominator. Since all
income is spent it is the sum of what is spent in country i and what is spent in country n.
This is the denominator in the second equation. The nominator simply uses the definition
of exports from above. In the third equation we have used the fact that prices and per firm
quantities are the same in the two countries. These can therefore be cancelled from the
equation. Furthermore, the number of firms in both countries is proportional to the two
countries. The result follows. From the above we also obtain total exports from country i
to country n:
X in =

Li
wLn
Li + Ln

This follows since total income in country n is wage income. It follows that trade is
balanced (you can try to show this). You should also note that exports from i to n depends
on the two countries, sizes.
With trade we also find welfare by using the expression from above:
U i = (L H + L

1
F

1
1

( 1)

b
Gains from trade come from increased product variety. Also note that the welfare
increase is largest for the smallest country.

More than one industry


Sometimes the Dixit Stiglitz demand system is used in combination with demand for
other goods. This is particularly convenient in models for more than one industry. One
might, for instance, assume that there is one standard industry producing a homogenous
good under constant returns to scale and one sector producing varieties of a differentiated
good under increasing returns to scale. A common example where this approach is used
is for models with an agriculture sector producing a homogenous good under constant
returns to scale and a manufacturing sector producing differentiated goods under
increasing returns to scale.
Below we will formulate this as a two stage optimization problem. The first stage is to
choose between consumption of the homogenous good and an aggregate of the
differentiated good. This first stage therefore determines the expenditures on the
homogenous good and the aggregate of the differentiated good, respectively. In the
second stage the quantities of each variety of the differentiated good, given the
expenditures on these goods, are determined.
Consumers utility function is written
U = C a1 C m

N 1 1
C m = qi
i =1

The first equation is the overall utility function for consumption of Ca and the aggregate
of differentiated good Cm. The utility function is a standard Cobb-Douglas utility
function. The second equation is a CES function of the type we introduced above. This
CES function has the same characteristics as we have already described.
We will use the Ca good as numeraire and set its price level equal to 1. Prices on the
differentiated good are described with the price index above; written P. Consumers
income are their wage income. We therefore have the budget constraint:
C a + PC m = wL
The above constraint simply expresses that total expenditures on Ca and Cm equal
income. The optimization problem is constrained by the budget constrain and can
therefore be solved with use of the Lagrange method. The Lagrangian is:
L = C a1 C m (C a + PC m wL )

First order conditions are:

dL
= (1 )C a C m = 0
dC a
dL
= C a1 C m 1 P = 0
dCm

C m1C = P
1
C a = (1 )C m P
(1 ) C P
Ca =
m

From the first order conditions we obtained the three last equations above. The last
equation gives consumption of Ca as a function of prices and the aggregate consumption
Cm. From the budget equation we obtain:

(1 ) PC

+ PC m = wL

PC m = wL

C a = (1 )wL

The above demonstrates that and (1-) are the expenditures for Cm and Ca respectively.
This was the first stage budgeting. The second stage involves choosing each qi given
aggregate expenditures on the aggregate good Cm. We have established that aggregate
expenditures on Cm is wL. The corresponding Lagrangian for this constrained
maximization problem is:

1 1

L* = qi ( pi qi wL )

The * simply indicates that the above Lagrangian is different from the one above. The
first order conditions for optimisation good i and good j are:
1


L
=
qi
qi 1
*

qi

qi

1
1

qi pi = 0
1

1
qi

q j p j = 0

p i = 0

It is obvious that the structure of this problem is similar to that of maximizing U subject
to the budget constraint we described in the one sector case. Also in this case we obtain:

qi

qj

q
= i
q
j

p
q
i = i
q j p j
p
qi = q j i
p
j

pi
pj

The last equation above gives demand for good i as a function of demand for good j and
their relative price. Insert this expression into the sub utility function Cm to obtain:

N 1
C m = qi
i =1

1 1 N 1
q j p j pi

i =1

N p
= q j i
p
i =1
j

N 1 1 1

= q j p j pi

i =1

N
1
1

=
q
p
p

j
j
i

i =1

In the second equation the expression was inserted into the utility function. In the third
equation the expression was simplified somewhat. In the fourth equation we used the fact
that summation runs over i and not over j. The resulting equation can be solved for the
demand for good j:

qj =

p j C m

( q )

1 1
i

Note that total expenditures on the Cm goods are E:

j =1

j =1

E = p jq j =

p 1j C m

pi1
i =1

= Cm p
j =1

1
j

N 1 1
=
pi
i =1

N
1
C m pi1
i =1

Above, the last equation is valid since summing over all js and all is involves summing
over all goods and therefore over the same goods. Note that the last equation contain the
same expression as we have defined as the price index before, P:
1

N
1
P = pi1
i =1

From the first stage budgeting we also have the results that
Cm =

wL
P

The demand function can therefore be written:


qj =

p j C m

( q )

1 1
i

p j wL

( q )
1
i

p j wL

( q )

1 1
i

p j wL

( q ) ( q )
1 1
i

1
1 1
i

p j wL
P 1

The above is the same demand function that we arrived at in the one sector case except
for the fact that only a fraction, , of total income is used for the differentiated goods Cm.

You might also like