Introduction
Model
Data and Summary Statistics
Empirical Analysis
Indirect Exports and Wholesalers: Evidence from
Interfirm Transaction Network Data
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
RIETI Workshop (2016/03/07)
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Section 1
Introduction
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Motivation
Wholesalers play an important role in facilitating international
trade
U.S: Wholesale and retail firms account for 11 and 24% of
exports and imports (Bernard et al. 2007)
Japan: Wholesalers account for 25% of exports in 2014
(Kikatsu)
Studies on a mechanism behind firms’ export mode choices
are important in understanding
factors that expand a market scope for firms/ role of
wholesalers
Still not many empirical evidence is accumulated other than
Bernard et al. (2012, 2014) and Ahn (2011)
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Goal of this paper
We examine manufacturers’ choice of export mode to understand
the mechanism behind their choice and wholesalers’ roles
Use Japanese interfirm transaction network data and analyze
the features of non-, indirect, and direct exporters
firms’ productivity (sales, sales per employment)
domestic transaction network (N of suppliers, N of customers)
Run multinomial logit regressions to see the determinants of
di↵erent export status
Build a simple Melitz-type model in which firms can also
export via wholesalers (indirect exporting)
Investigate industry heterogeneity
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Role of wholesalers
Wholesalers can reduce the fixed cost of exporting for
manufacturing firms
distribution (for sellers) and procurement (for buyers)
efficient matching
risk sharing
mitigating adverse selection: guarantee payment (for sellers)
and quality (for buyers)
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Literature review
Empirics of intermediated trade
Ahn et al. (2011), Bernard et al. (2012, 2014), Feenstra and
Hanson (2001)
Theory of intermediated trade
Antras and Costinot (2010, 2011), Ackerman (2014)
Value-added exports and input-output model
WIOD, Caliendo and Parro (2014)
Our study uses large Japanese firm-level data
We also incorporate information on domestic transaction
network
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Section 2
Model
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Summary
A simple trade model with heterogeneous firms, but firms can
also export indirectly via wholesalers
Direct export: a large fixed cost, but small variable cost
Indirect export: a small fixed cost (due to cost sharing through
wholesalers) but high variable cost (due to double
marginalization)
Implication –> Sorting of firm size/productivity into direct,
indirect and non-export status
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Export profits
Export profits and sorting
Ind
0
re
Di
c
x
te
po
rt
pr
ct
ire
ex
po
rt p
t
rofi
t
ofi
f^I
Domestic firms
Indirect exporters
Direct exporters
f^D
Productivity (size)
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Section 3
Data and Summary Statistics
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Data
TSR firm-level data in 2014
We focus on manufacturing and wholesaling sectors (they
account for more than 80% of exporters) –> 120,000
manufacturing firms
For each firm, we observe sales, employment, industry (4-digit
JSIC), address, and whehter a firm exports or not, etc.
Transaction partner information
We can identify firms that supply their goods to exporting
wholesalers
Export mode
Direct: manufacturing firms who report they are exporting
Indirect: manufacturing firms who supply their products to
exporting wholesalers
Domestic: other manufacturing firms
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Summary statistics
domestic
indirect
direct
Total
domestic
indirect
direct
Total
# of firms
91363
22368
6820
120551
Total sales
7.76E+10
7.98E+10
2.04E+11
3.61E+11
Mean of log sales
12.09
13.13
14.57
12.43
Sales share
0.22
0.22
0.56
1.00
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Median
11.99
13.02
14.36
12.25
Total Employment
2425321
1469724
2585678
6480723
SD
1.52
1.68
2.04
1.72
Emp share
0.37
0.23
0.40
1.00
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
CDF of sales
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
CDF of sales by industry
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
CDF of sales by industry
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
CDF of other variables
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Section 4
Empirical Analysis
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Multinomial logit model
Firms’ discrete choice problem –> latent utility model
Type I extreme value distribution –> multinomial logit model
Probability of direct exporting is
Pr [i=direct] =
⇣ 0
exp Xi
0
1 + exp Xi
D
⌘
0
+ exp Xi
I
D
Probability of indirect exporting is
Pr [i=indirect] =
⇣ 0
exp Xi
0
1 + exp Xi
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
I
I
⌘
0
+ exp Xi
D
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Baseline estimates
(1)
(2)
(3)
(4)
export type
indirect
direct
indirect
direct
indirect
direct
indirect
direct
log sales
0.404***
(0.00484)
0.805***
(0.00773)
0.190***
(0.00715)
0.486***
(0.0122)
0.452***
(0.0135)
0.729***
(0.0223)
0.133***
(0.00626)
0.536***
(0.0101)
-6.488***
(0.0626)
-13.22***
(0.109)
-4.536***
(0.0782)
-9.883***
(0.149)
0.781***
(0.0112)
-4.420***
(0.0699)
0.781***
(0.0174)
-11.16***
(0.121)
0.104***
(0.00754)
0.102***
(0.0138)
0.739***
(0.0125)
-4.131***
(0.0812)
0.403***
(0.0138)
0.361***
(0.0253)
0.647***
(0.0196)
-9.806***
(0.153)
log in-degree
log outdegree
Intercept
Observations
2-digit JSIC FE
prefecture FE
Pseudo R2
120,551
No
No
0.104
120,551
No
No
0.118
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
120,551
No
No
0.142
120,551
No
No
0.144
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Estimates with fixed e↵ects
(1)
(2)
(3)
(4)
export type
indirect
direct
indirect
direct
indirect
direct
indirect
direct
log sales
0.408***
(0.00512)
0.826***
(0.00859)
0.147***
(0.00751)
0.593***
(0.0129)
0.441***
(0.0146)
0.793***
(0.0236)
0.110***
(0.00661)
0.530***
(0.0111)
-6.477***
(0.0767)
-15.21***
(0.158)
-4.081***
(0.0910)
-11.52***
(0.193)
0.867***
(0.0120)
-4.056***
(0.0846)
0.871***
(0.0193)
-12.83***
(0.168)
0.0539***
(0.00794)
0.186***
(0.0145)
0.791***
(0.0133)
-3.519***
(0.0945)
0.386***
(0.0149)
0.398***
(0.0269)
0.715***
(0.0219)
-11.35***
(0.196)
log in-degree
log outdegree
Intercept
Observations
2-digit JSIC FE
prefecture FE
Pseudo R2
120,549
Yes
Yes
0.157
120,549
Yes
Yes
0.175
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
120,549
Yes
Yes
0.199
120,549
Yes
Yes
0.201
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Results: productivity measures
Intercepts for “Direct” and “Indirect” are lower than
“Doemstic”
Existence of fixed export-costs
Intercept for “Direct” is lower than “Indirect”
Fixed export-costs are lower for indirect than direct trades =>
Shared by other firms using the same wholesaler)
Sales slope are positive: higher variable costs
Transportation costs, tari↵ etc.)
Sales slope for “Indirect” is steeper than “Direct”
Marginal costs might be higher for “Indirect” trade
Results are robust and consistent with the theoretical model
and robust
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Intercept and slope by industry
Indirect vs Direct (slope)
-5
1.2
Indirect vs Direct (intercept)
1
IRON, STEEL
RUBBER PRODUCT
ICT EQUIP
CHEMICAL PRODUCT
BUSINESS MACHINE
ELECTRONIC PARTS
PETROLEUM
ELECTRICAL MACHINE
PRINTING
MISCCERAMIC
MANU
TRANSPORT EQUIP
NON-FERROUS METALS
PULP,PRODUCTION
PAPER
LEATHER PRODUCT
TEXTILE PRODUCT
MACHINE
GENERAL
FOOD MACHINE
PLASTIC PRODUCT
RUBBER PRODUCT
FABRICATED METAL
LUMBER, WOOD
Slope (direct)
.6
.8
BEVERAGES, TOBACCO
PRODUCTION
MACHINE
GENERAL
MACHINE
FABRICATED
METAL
FURNITURE
PLASTIC
PRODUCT
TEXTILE PRODUCT
MISC MANU
NON-FERROUS METALS
FOOD PULP, PAPER
CERAMIC
TRANSPORT EQUIP
ELECTRICAL MACHINE
LUMBER, WOOD
BUSINESS MACHINE
ELECTRONIC PARTS
PETROLEUM
PRINTING
ICT EQUIPCHEMICAL PRODUCT
BEVERAGES, TOBACCO
.4
Intercept (direct)
-15
-10
LEATHER PRODUCT
FURNITURE
.2
-20
IRON, STEEL
-10
-8
-6
-4
Intercept (indirect)
Both 5% significant
45-degree line
Other
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
.2
.3
.4
.5
Slope (indirect)
Both 5% significant
45-degree line
.6
.7
Other
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Results: domestic transaction network
Higher in-degree raises the probability of both indirect and
direct exporting but more for direct exporting
A possible explanation is the cost sharing
Higher out-degree raises the both probability for about the
same magnitude
higher out-degree promotes export in general; demand side
explanation?
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Conclusion
Sorting of size into three export status (consistent with the
model)
Lower intercept and steeper slope of sales for Direct than
Indirect Exporting
Consistent with our model setup: Fixed costs are lower, and
marginal costs are higher for Indirect than Direct exporting
In-degree: promotes direct exporting
Sharing of fixed costs with other firms using the same
wholesaler
Out-degree: promotes trade in general
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
20
Exporters share by industry
Indirect/Direct ratio (in log scale)
5
10
15
LUMBER, WOOD
FURNITURE
FOOD
PRINTING
TEXTILE PRODUCT
PULP, PAPER
FABRICATED METAL
IRON, STEEL
BEVERAGES,
PLASTIC TOBACCO
PRODUCT
LEATHER PRODUCT
RUBBER PRODUCT
NON-FERROUS METALS
CERAMIC
MISC MANU GENERAL MACHINE
ELECTRICAL MACHINE
PETROLEUM
TRANSPORT EQUIP
PRODUCTION MACHINE
ICT EQUIP
ELECTRONIC PARTS
CHEMICAL PRODUCT
BUSINESS MACHINE
.1
.2
.3
.4
Total (direct + indirect) exporters share
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
.5
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Multinomial logit by industry
Slope vs Intercept (direct)
-10
-4
Slope vs Intercept (indirect)
ICT EQUIP
BEVERAGES, TOBACCO
ICT EQUIP
CHEMICAL
PRODUCT
CHEMICAL
PRODUCT
LEATHER
PRODUCT
ELECTRONIC
PARTS
Intercept (direct)
-16
-14
-12
BUSINESSPARTS
MACHINE
ELECTRONIC
Intercept (indirect)
-8
-6
ELECTRICAL
MACHINE
RUBBER PRODUCT
TEXTILE
PRODUCT
BUSINESS MACHINE
TRANSPORT EQUIP
PETROLEUM
TOBACCO
CERAMIC BEVERAGES,
GENERAL MACHINE
MISC
MANU PRODUCT
PLASTIC
FOOD
PRODUCTION MACHINE
PRINTING
NON-FERROUS METALS
PULP, PAPER
FABRICATED METAL
PETROLEUM
ELECTRICAL MACHINE
PRINTING
MISC MANU
TRANSPORT
EQUIP
CERAMIC
NON-FERROUS METALS
PULP, PAPER
LEATHER
PRODUCTMACHINE
TEXTILE
PRODUCT
PRODUCTION
GENERAL MACHINE
FOOD
PRODUCT
LUMBER, WOODPLASTIC
RUBBER PRODUCT
FABRICATED METAL
FURNITURE
FURNITURE
-18
-10
IRON, STEEL
LUMBER, WOOD
.2
.3
.4
.5
Slope (indirect)
Both 5% significant
.6
Other
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
.7
IRON, STEEL
.4
.6
.8
Slope (direct)
Both 5% significant
1
1.2
Other
Indirect Exports and Wholesalers: Evidence from Interfirm Transa
Introduction
Model
Data and Summary Statistics
Empirical Analysis
Scatter plots by industry
Scatter Matrix of Industry Characteristics
.1
.2
.3
12
12.5
13
13.5
.2
direct
exporters
share
.1
0
.3
indirect
exporters
share
.2
.1
10
log of #
of firms
8
6
13.5
mean of
log
sales
13
12.5
12
2.5
SD of
log
sales
2
1.5
0
.1
.2
6
Daisuke Fujii, Yukako Ono, Yukiko Umeno Saito
8
10
1.5
2
2.5
Indirect Exports and Wholesalers: Evidence from Interfirm Transa