International Trade

Lecture 5: endogenous mark-ups

Thomas Chaney

Sciences Po

Thomas Chaney (Sciences Po) International Trade 1 / 21

Exogenous or endogenous mark-ups?

Krugman’s assumptions:
CES + many firms ) constant demand elasticity.
Constant demand elasticity ) constant desired mark-up.
Monopolistic competition ) desired mark-up = actual mark-up.
Relaxing Krugman’s assumptions:
1 CES + few firms ) variable demand elasticity (Atkeson & Burstein
AER 2008).
2 "Non monopolistic" competition ) desired mark-up 6= actual
mark-up (BEJK AER 2003).
3 Non CES preferences ) variable demand elasticity (Melitz &
Ottaviano REStud 2005).

Thomas Chaney (Sciences Po) International Trade 2 / 21

CES with few firms (e.g. Atkeson & Burstein)

Nested CES preferences.

Many differentiated sectors (i ), but few differentiated firms (n ) within
each sector, 8 ⇣´ h 1 ⌘ hh1
>
< U= 1
0 ( yi ) di
h

> ⇣ ⌘ r
: y = ÂN (q ) r r 1 r 1
i n =1 in

with r > h (more differentiation across than within sectors).

Thomas Chaney (Sciences Po) International Trade 3 / 21

Optimal pricing (Bertrand versus Cournot)

Simple solution to optimal pricing,

# (s ) W
pin (s ) =
# (s ) 1 jin
⇢
# Bertrand (s ) = sh + (1 s ) r
with 1
# Cournot (s ) = sh 1 + (1 s ) r 1
✓ ◆1
pin qin (p )1 r pin r
and sin = N = N in 1 r
=
Âm=1 pim qim Âm=1 (pim ) Pi

Thomas Chaney (Sciences Po) International Trade 4 / 21

Endogenous mark-ups in Dixit-Stiglitz

Polar cases: small (s ⇡ 0) versus large (s ⇡ 1) firms,

( r
p (s = 0) = r 1 W j r h
h W with <
p (s = 1) = h 1 j r 1 h 1

Intermediate cases: each firm takes into account its impact on

sector’s price index.
1 Big firms have a larger impact on the price index than small firms.
2 Big firms lose fewer marker shares by raising prices than small firms.
3 Big firms charge higher mark-ups than small firms.
Endogenous mark-ups: a change in market share induces a change
in mark-up.

Thomas Chaney (Sciences Po) International Trade 5 / 21

Bernard, Eaton, Jensen and Kortum (AER 2003)

N asymmetric countries.
Fixed set of sectors w 2 [0, 1] , CES preferences,
✓ˆ 1 ◆ ss1
s 1
Un ⌘ qn ( w ) s dw
0

Simple iso-elastic demand,

✓ ◆
pn ( w ) 1 s
xn ( w ) = Xn
Pn
✓ˆ 1 ◆ 11s
1 s
with Pn = pn ( w ) dw
0

Thomas Chaney (Sciences Po) International Trade 6 / 21

Within sectors

CRS technology of production.

Many firms within sector w differ in their labor productivity.
Within sector w, all firms produce the same variety.
1 only the best survives.
2 the best charges a price  the second best marginal cost.
3 Only the joint distribution of best and 2nd best productivity matter.

Thomas Chaney (Sciences Po) International Trade 7 / 21

Productivity distributions

Best productivity draws are Fréchet distributed,

⇥ ⇤
Fi (z1 ) = Pr zi1 (w )  z1
⇣ ⌘
= exp Ti z1 q

Best and 2nd best draws are jointly Fréchet distributed,

⇥ ⇤
Fi (z1 , z2 ) = Pr zi1 (w )  z1 ; zi2 (w )  z2 | 0  z2  z1
h ⇣ ⌘i ⇣ ⌘
= 1 + Ti z2 q z1 q exp Ti z2 q

1 Ti indexes overall productivity in country i.

2 q (same across countries) indexes (inverse of) dispersion of
productivities.

Thomas Chaney (Sciences Po) International Trade 8 / 21

Why the Fréchet?

Take T independent draws from a Pareto distribution,

q
Pr [z (w )  z ] = z

The best and second best draws from T draws are jointly distributed
Fréchet.
1 Note: non Pareto fat tailed distributions converge asympotically to
Fréchet.
2 Note: the best of many Fréchet’s is Fréchet.

Thomas Chaney (Sciences Po) International Trade 9 / 21

Trade costs

Iceberg transportation costs, tin 1, tin  tik tkn .

1 Perfectly substitutable goods within sectors ) not all firms survive in a
market.
2 Positive trade costs ) some firms survive at home but not abroad.

Thomas Chaney (Sciences Po) International Trade 10 / 21

Entry into markets

Best producer in sector w from i can deliver goods in n at a cost,

1 tin wi
cin (w ) =
zi1 (w )

Consumers in n buy goods from the best supplier,

cn1 (w ) = min cin

1
(w )
i

Thomas Chaney (Sciences Po) International Trade 11 / 21

Distribution of marginal costs

Distribution of lowest marginal costs in country n is Fréchet,

⇥ ⇤
Gn1 (c1 ) = Pr cn1 (w )  c1

= Pr min cin1 (w )  c1
i

= 1 Pr min cin1 (w ) > c1
i
⇥ ⇤
= 1 ’ 1 Gin1 (c1 )
i
⇣ ⌘
=1 exp Fn c1q

with Fn = Â Ti (tin wi ) q

Thomas Chaney (Sciences Po) International Trade 12 / 21

Pricing

Bertrand competition with homogenous goods ) limit pricing.

Bertrand competition with differentiated goods ) constant
Dixit-Stiglitz pricing.
1 Only the best firm in w survives.
2 Limit pricing: it charges a price not higher than 2nd lowest marginal
cost.
3 Dixit-Stiglitz: it charges a mark-up not higher than Dixit-Stiglitz
mark-up.
n o
pn (w ) = µ (w ) ⇥ cn1 (w ) = min cn2 (w ) ; µ̄cn1 (w )
⇢ s
with µ̄ = s 1 , s >1
• , s1

Thomas Chaney (Sciences Po) International Trade 13 / 21

Lowest and second lowest costs

Either two firms from the same country are best and 2nd best, or the
best is from one country, the 2nd best from another.
⇢
2 1
cn (w ) = min min cjn (w ) , cin2 (w )
i j 6 =i

Joint distribution of the lowest and the 2nd lowest costs from country i
in country n.
⇥ 1 ⇤
Ginc (c1 , c2 ) = Pr cin (w ) c1 , cin2 (w ) c2

t w t w
= Pr zin1 (w )  in i , zin2 (w )  in i
c1 c2
h ⇣ ⌘i ⇣ ⌘
q
= 1 + Ti (tin wi ) c2q c1q exp Ti (tin wi ) q c2q

Thomas Chaney (Sciences Po) International Trade 14 / 21

Joint distribution of best and second best costs

⇥ ⇤
Pr cn1 (w ) c1 , cn2 (w ) c2

"both the lowest and the second lowest draws are above c2 "
= Pr or
"the lowest cost is in [c1 , c2 ] and the second lowest above c2 "
2 1 w
3
for⇢all i 0 s, "cin ( ) c2 and cin2 (w ) c2 "
= Pr 4or "c1  cin 1 w < c and c 2 w c2 in i" 5
( ) 2 in ( )
for some i, 1 2 w
and "ckn (w ) c2 and ckn ( ) c 2 in k 6 = i"
= ...
⇣ ⌘ ⇣ ⌘ ⇣ ⌘
= exp Fn c2q + Fn c2q c1q exp Fn c2q

Thomas Chaney (Sciences Po) International Trade 15 / 21

Technical steps

⇥ ⇤
Pr cn1 (w ) c1 , cn2 (w ) c2
= ’ [1 Gin (c2 , c2 )]
i
( )
+ Â ([1 Gin (c1 , c2 )] [1 Gin (c2 , c2 )]) ’ [1 Gkn (c2 , c2 )]
i k 6 =i
q q
= ’e Ti (tin wi ) c2
i
( )
⇣ ⌘ q q q q
+ Â Ti (tin wi ) q
c2q c1q e Ti (tin wi ) c2
’e Tk (tkn wk ) c2
i k 6 =i
⇣ ⌘ ⇣ ⌘ ⇣ ⌘
= exp Fn c2q + Fn c2q c1q exp Fn c2q

Thomas Chaney (Sciences Po) International Trade 16 / 21

Distribution of mark-ups

For all µ0 s such that 1  µ  µ̄,


⇥ ⇤ c2
Pr µn (w )  µ|cn2 (w ) = c2 = Pr  cn1 (w )  c2 |cn2 (w ) = c2
µ
´ c2 ∂ 2 G n
c /µ ∂c ∂c |c1 ,c2 dc1
= ´2c2 ∂2 G1 2
0 ∂c1 ∂c2 |c1 ,c2 dc1
n

c2q (c2 /µ)q

=
c2q
q
=1 µ

Allowing the possibility of Dixit-Stiglitz mark-ups,

⇢
1 µ q , 1  µ  µ̄,
H (µ) = Pr [µn (w )  µ] =
1 , µ > µ̄ = s s 1

Thomas Chaney (Sciences Po) International Trade 17 / 21

Average versus individual mark-ups

Distribution of mark-ups is invariant (knife-edge but interesting case).

Mark-ups depend on q and s:
1 Lower q, more dispersed productivity, higher distance between best and
2nd best, higher mark-ups.
2 Lower s, less elastic demand (across sectors), higher mark-ups.
Reduction in trade barriers (lower t ):
1 If 2 best firms remain domestic, mark-up unchanged.
2 If best firm domestic, better foreign competitors, mark-up goes down.
3 If best firm becomes foreign, mark-up may go up or down.
4 If best firm foreign, mark-up goes up.
5 If 2 best firms remain foreign, mark-up unchanged.
6 With Fréchet distributed productivities, all those effects cancel out.

Thomas Chaney (Sciences Po) International Trade 18 / 21

Measured productivity

Simple measured productivity, output per worker, is simply mark-up,

output (z1 )
= µ ( z1 )
workers (z1 )
Observed distribution of productivity,
⇥ ⇤
Hin (µ | z1 ) = Pr µin (w )  µ | zin1 (w ) = zn1 (w ) = z1
( h i
1 exp Fn µq 1 (tin wi )q z1 q , 1  µ  µ̄,
=
1 , µ > µ̄ = s s 1

1 More productive firms (higher z1 ) charge higher mark-ups on average.

2 The easier it is to export (tij lower or wi lower), the higher the
mark-up, on average.
3 The tougher the market (lower Fn ), the lower the mark-up, on average.
4 The tougher the market structure (q or s higher), the lower the
mark-up (first order stochastic dominance).
Thomas Chaney (Sciences Po) International Trade 19 / 21
Efficiency and export status

Not all firms export, and exporters are larger and more productive on
average.
1 To produce domestically, must be the best at home,
wi
zi1 (w ) zk1 (w ) , 8k 6 = i
tki wk
2 To export, must be the best abroad (after incurring the trade barrier
tin ),
tin wi
zi1 (w ) zk1 (w ) , 8k 6 = i
tkn wk
3 Triangular inequality, t1  ttin , it is harder to export than to sell
ki kn
domestically,
wi t w
zk1 (w )  zk1 (w ) in i
tki wk tkn wk
Note: since mark-ups abroad are lower than at home, firms "lose"
productivity ( global output
total workers ) when they "become" exporters, as in Melitz.

Thomas Chaney (Sciences Po) International Trade 20 / 21

Aggregate exports

Aggregate exports,

Xin T (t w )q
= pin = i in i
Xn Fn
1 Aggregate exports unaffected by mark-ups (Eaton-Kortum).
2 Aggregate exports as in Krugman or Chaney.
Welfare,
8
< Pn 1 = g ⇥ F1/qn
h i1/(s 1)
1+q s +(s 1)µ̄ q 1+2q s
: g= 1+ q s G q

1 Imperfect competition distortions summarized in g.

⇣ ⌘1/ ( s 1)
2 g|Perfect competition > g|Bertrand competition = G 1+2qq s
.
⇣ ⌘ ⇣ ⌘
3 ∂ ggPerfect /∂s, ∂g ggPerfect /∂q > 0.
Bertrand Bertrand

Thomas Chaney (Sciences Po) International Trade 21 / 21