A General Method to Assess Domestic Value-added
in Exports When Processing Trade is Pervasive,
with an Application to China
Robert Koopman and Zhi Wang
US International Trade Commission
Shang-Jin Wei
Columbia University, NBER, and CEPR



The paper is a recipe with an illustration
Growing up in a centrally planned economy, I
was force-taught input-output tables and
Leontief inverse as central planners love them
as useful tools
After coming to the United States, I eagerly
embraced market economy and had resolved to
never touch the input-output tables again…
Imported inputs reduce value added in exports

The story of an iPod/iPad
…has been told too many
times
In trade statistics

China’s export value =$150/unit
Chinese value added = $4




iPods/iPads are assembled
through processing trade
How general is the iPod story?
Presentation Outline



Conceptual Framework
 Problems with Existing Method
 A new method
Application to China
 Domestic content in China’s total
exports
 By sector
Conclusions and Future Research
How to compute foreign content in exports?
Existing Literature

International Trade: “vertical specialization”







Hummels, Ishii and Yi (JIE, 2001) - HIY for short
Yi (JPE, 2003)
Goh and Olivier (HEC France wp, 2004)
Chinn (NBER wp, 2005)
U.S. National Research Council (Nat Acad Press, 2006)
Dean, Fung and Wang (ITC wp, 2007)
Input-output models: “domestic/foreign content”


Chen, X., L. Cheng, K.C. Fung and L. J. Lau. 2004
Lau L.J., X. Chen, L. K. Cheng, K. C. Fung, Y. Sung, C. Yang,
K. Zhu, J. Pei and Z. Tang. 2007
How to compute foreign (and domestic) content in
exports?

The HIY approach assumes the same intensity in the use of
imported inputs between all exports and domestic sales.

Might not be appropriate for most countries


“Duty drawback” -> use of more imported inputs in exports
Especially inappropriate for developing countries with
aggressive export promotion programs


Mexico: Maquiladora
China: pervasive processing trade
How to compute foreign/domestic content in
exports?

Step 1: Recognize processing exports explicitly



Tracking input-output coefficients and import coefficients
separately for processing exports vs for normal exports and
domestic sales
Challenge: the new input-output coefficient matrices not
collected by authorities
Step 2: Estimate the new I/O coefficients

Combine info from trade statistics with existing I/O tables
The I/O model when processing exports are not explicitly
recognized - the implicit approach by HIY
(1)
A X Y
D
(2)
A X Y
M
D
M
(3)
X
M
uA  uA  Av  u
D
M
Total production = dom sale
+exports
Total imports = final sale
+ use as intermediates
Total cost of production =
direct VA(cost of factors)
+cost of intermediates
AD = [aDij] = matrix of direct input coefficients of domestic products;
AM = [aMij] is an matrix of direct input coefficients of imported goods;







AD = [aDij] = matrix of direct input coefficients of domestic products;
AM = [ aMij] ] is an matrix of direct input coefficients of imported
goods;
YD = vector of final demands for domestically produced products,
including usage in gross capital formation, private and public
consumption, and gross exports;
YM = vector of final demands for imported products, including
usages in gross capital formation, private and public final
consumption;
X is a vector of gross output;
M is a vector of imports;
Av = [avj] is a vector of each sector j’s ratio of value-added to gross
output;
The I/O model when processing exports are not explicitly
recognized - the implicit approach by HIY
The solution:
D 1
X  (I  A ) Y
D
(M  Y M )  AM ( I  AD ) 1 Y D


DVS  Av X / Y  Av ( I  AD ) 1
D
YD is a vector of final demands for domestic products, which includes
domestic products used in gross capital formation, private and public final
consumption, and gross exports;
Define share of domestic value in exports

Define share of domestic content (domestic value
added) in final demand:


DVS  V / Y D  Av X / Y D  Av ( I  AD ) 1


Share of foreign content:
FVA = 1- DVA
Assessing DVA/FVA when processing trade is
pervasive: New approach

Existing approach:


Domestic final demand and all exports are
assumed to have the same input-output matrix
New approach


Recognize processing exports
Let processing exports have a potentially different
I/O matrix (while still letting normal exports and
domestic final demand to have the same matrix).
This is not a trivial extension!
Re-work the model when processing trade is
recognized
(1*)
ADD ( X  E P )  ADP E P  Y D  X
(2*)
AMD ( X  E P )  AMP E P  Y M  M
 P
(3*)
( ADP  AMP )' E P  A
v
EP  EP
 D
(4*)
( ADD  AMD )' ( X  E P )  A
(5*)
uA  uA  Av  u,
Dk
Mk
v
(X  EP)  X  EP
k  D, P
Imported VA Share/Foreign content: generalize HIY
Total imported intermediate inputs request:
M  Y M  A MD ( I  A DD ) 1 (Y D  E N )  AMD (1  A DD ) 1 A DP E P  A MP E p
Total FVA(VS) share in a country’s exports by industry
D T
VSS
VSS 
VSS P
DD 1
uA ( I  A )
 MD
uA (1  ADD ) 1 ADP  uAMP
MD
T
Total FVA (VS) share in a country’s aggregate exports
N
P
E
E
TVSS  uA MD ( I  A DD ) 1
 u ( A MD (1  A DD ) 1 A DP  A MP )
te
te


This is a generalization of HIY(2001)
When ADP=ADD, and AMP=AMD,


TVSS reduces to the HIY formula
When EP/te = 0,

TVSS also reduces to the HIY formula
Domestic content/DVA: generalizes the HIY formula
Total DVA share in a country’s exports by industry
D T
DVS
DVS 
DVS P
DD 1
AV ( I  A )
 D
P
AV (1  ADD ) 1 ADP  AV
D
T
Total DVA share in a country’s aggregate exports
N
P
E
E
D
P
TDVS  AV ( I  ADD ) 1
 ( AV (1  ADD ) 1 ADP  AV )
te
te
D
Estimation Challenge



Challenge: ADD, ADP, AMD, AMP, AVD, and AVP are not directly
observable.
Approach: Combine input-output tables (from NBS) with trade
statistics
To minimize squared errors in separating the processing exports input –
output structure from a given national input-output account.

Objective function:
K
K
M in S = 
( z ijdn  z 0 ijdn ) 2
z0
i 1 j 1
K
K
+
i=1 i=1
dn
ij
( z ijmp  z 0 ijmp ) 2
z0
mp
ij
K
K
 
( z ijdp  z 0 ijdp ) 2
i=1 j 1
K
+
j=1
(v nj  v0 nj ) 2
v0
n
j
z0
dp
ij
K
 
i 1 j 1
K
(v jp  v0 pj )
j 1
v0 pj

K
( z ijmn  z 0 ijmn )
z ijmn
Constraints
Four “balance” conditions:
K
(z
j 1
dn
ij
 z ) y
dp
ij
e = xi
 (z
p
 i
e
j 1
 (z
mn
ij
 zijmp )  yim = mi
K
K
j 1
K
d
n

i
i
dn
ij
 z )  v = xj
mn
ij
n
j

e
dp
mp
p
p
(
z

z
)

v
=
e
 ij ij
j
j
p
j
i 1
Six adding up constraints
K
z
j 1
mn
ij
K
 (z
j 1
K
mp
z
 ij
=m
n
i
j 1
K
dn
ij
=m
p
i
 z ) =  z ij  (min  mip )
dp
ij
j 1
v nj  v jp = v j
yid  yim = yi
zijdn  zijdp  zijmn  zijmp = zij
Parameters and their data sources
•
•
•
•
•
•
•
•
•
•
•
x = Gross output by sector; (from IO table)
z= Goods used as intermediate inputs in sector ; (from IO table)
v= Value-added by sector ; (from IO table)
en= Normal exports by sector ; (from IO table and trade statistics)
ep = Processing exports by sector ; (from IO table and trade statistics)
m= Total imports by sector ; (from IO table)
mp= Imports by sector used as intermediate inputs to produce processing exports;
(share from trade statistics)
mn = Imports by sector used as intermediate inputs for domestic \production and normal
exports; (share from trade statistics)
y = Total final demand by sector; (includes consumption and investment, from IO
table)
ym = Final demand from imports by sector (residuals of m- mn - mp )
yd= Final demand by sector provided by domestic production (residual of y- ym);
Let’s look at an application

Aggregate

By industry
Chinese domestic content in its merchandise
exports is increasing overtime
100
Percent of total exports
90
80
70
60
50
40
30
20
10
0
1997
Direct domestic value-added
Direct foreign value-added
2002
2007
Indirect domestic value-added
Indirect Foreign value-added
HIY’s level of FVA (VS) share is under-estimated
Upward trend in FVA is most likely incorrect
50
Percent of total exports
45
40
35
30
25
20
15
10
5
0
Total Foreign value-added
HIY-1997
HIY-2002
HIY-2007
KWW-1997
KWW-2002
KWW-2007
Chinese domestic content is much lower in
processing exports than in normal exports
100
90
Percent of total exports
80
70
60
50
40
30
20
10
0
Normal-1997
Processing 1997
Normal-2002
Total Foreign value-added
Processing-2002
Normal-2007
Total Domestic Value-added
Processing-2007
Should there be any trend in DVA share?

Two opposing forces


Domestic input producers get better over time
But the cost of using imported inputs may get
lower over time
Decomposing Chinese Total Manufacturing Exports
Differences between HIY and redefined measures
100
Percent of total exports
90
80
70
60
50
40
30
20
10
0
HIY-1997
HIY-2002
HIY-2006
Direct domestic value-added
Direct foreign value-added
KWW-1997
KWW-2002
KWW-2006
Indirect domestic value-added
Indirect Foreign value-added
Domestic Value-added Share in Chinese
Merchandise Exports to its Major Trading Partners
Low DVA share destinations, in percent, 2007
Hong Kong
Singapore
Processing exports
as share of
Chinese exports
United States
EU10
DVA as share of
Chinese exports
Malaysia
Taiwan province
Share as China’s
exports to the
World
EU15
Japan
0
10
20
30
40
50
60
70
80
90
100
Domestic Value-added Share in Chinese
Merchandise Exports to its Major Trading Partners
High DVA share destinations, in percent, 2007
Russia
Eastern Europe/Central Asia
Processing exports
as share of
Chinese exports
Sub-Saharan Africa
Middle East/North Africa
DVA as share of
Chinese exports
Rest Asia
Rest of Latin Amer/Caribbean
Share as China’s
exports to the
World
Indonesia
India
0
10
20
30
40
50
60
70
80
90
100
Domestic Value-added Share in Chinese
Merchandise Exports by Firm Ownership
in percent, 2007
Private Firms
High DVA share destinations, in percent, 2007
Processing exports
as share of
Chinese exports
Collectively Owned Firms
State Owned Firms
DVA as share of
Chinese exports
Joint Venture Firms
Share as China’s
exports to the
World
Wholly Foreign Owned
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
80.0
90.0
100.0
Conclusions

Across all products, the average share of
imported value added is about 50-60% for
China.

The FVA share varies across products


Relatively more sophisticated sectors are more
likely to have a high FVA share (e.g., consumer
electronics, 65%, and computers, 83%)
One could combine it with regional inputoutput matrices, or incorporate it in a global
analysis such as the KPWW framework

For products with a high foreign content, Japan,
Korea, Taiwan, Hong Kong and the U.S. are the
primary contributors to foreign content

Thank you.