Slides Dietzenbacher - World Input

This project is funded by the European Commission, Research Directorate General as part
of the 7th Framework Programme, Theme 8: Socio-Economic Sciences and Humanities.
Grant Agreement no: 225 281
WIOD
Why WIOD?
Erik Dietzenbacher
1. What is (in) WIOD
2. Factor intensity of trade
3. Vertical specialization
4. CO2 emissions: Consumer responsibilities and their
sensitivity
5. Does the Rest of the World matter?
6. Is there a life after WIOD?
WIOD
WORLD INPUT-OUTPUT DATABASE:
Construction and Applications
Background
Policies are designed at a detailed level of industries
and products
Production is characterized by interdependent
structures
Globalization increases the importance of crossborder interdependencies, which makes inclusion of
trade in analyses more essential than ever
Analyzing policy issues requires an all-encompassing
database with three dimensions:
time, industries/products and countries
Objectives of WIOD
To build a time series of global inter-country inputoutput tables;
To build socio-economic and environmental satellite
accounts;
To measure and analyze trends in trade, economic
growth, technological change and environmental
pressures;
To provide policy support to the European
Commission on socio-economic and environmental
issues.
Full database will become publicly
available in May 2012
WIOD: Data and Coverage
Time series in current and constant prices of:
 Harmonized national supply and use tables
 Harmonized IO tables
 Bilateral trade flows of goods and services
 Inter-country IO tables
 Socio-economic accounts and environmental accounts
The tables in the WIOD-database will cover:
 The period from 1995 to 2006
(and for some major countries back to 1980)
 27 EU countries and 13 other major countries
 More than 30 industries and at least 60 products
WIOD: Work Packages
WP1-3: Construction of harmonized supply and use tables,
national input-output tables, price deflators, trade flows
and intercountry input-output tables
WP4: Construction of environmental satellite accounts
(energy use, greenhouse gas emissions, etc.)
WP5: Construction of socio-economic satellite accounts
(skill levels, investment, accumulation of intangibles)
WP6: Methodological research
WP7-9: Development of new models and
extension/adaptation of models with track record within
EC
Who is in WIOD?
University of Groningen (The Netherlands)
Institute for Prospective Technological Studies (Sevilla, Spain)
Wiener Institut für Internationale Wirtschaftsvergleiche (Vienna, Austria)
Zentrum für Europäische Wirtschaftsforschung (Mannheim, Germany)
Österreichisches Institut für Wirtschaftsforschung (Vienna, Austria)
Konstanz University of Applied Sciences (Germany)
The Conference Board Europe (Brussels, Belgium)
CPB Netherlands Bureau for Economic Policy Analysis (The Hague, The
Netherlands)
Institute of Communication and Computer Systems (Athens, Greece)
Central Recherche SA (Paris, France)
* Organization for Economic Co-operation and Development (Paris
France)
My personal interest:
• What (types of) questions can be answered?
• What difference does it make?
Examples of preliminary studies
Factor intensity of trade
Leontief paradox:
labor content of 1 million $ of exports
versus labor content of 1 million $ of imports
Crucial:
same technology assumption
use the matrix of technical input coefficients
($ of steel per $ of US cars, no matter
whether US steel or German steel is used)
Factor intensity of trade
Problem 1:
labor content of US exports includes German workers
solution: use domestic input coefficients
Problem 2:
domestic input coefficients of the US
cannot be used for labor content of US imports
Countries that are “similar” in terms of technical input
coefficients, may have very different domestic input
coefficients
because their dependence on imported inputs differs
Factor intensity of trade
Exercise:
Intercountry IO tables for 6 European countries
GE, FR, IT, NL, BE, DK
1985, 1975
Factor intensity of trade
Exports as % of total output
1975
1985
GE
12
15
FR
IT
11
10
13
11
NL
28
28
BE
27
36
DK
16
17
Exports as % of total output
Factor intensity of trade
Exercise:
LAB(GE→FR) = GE labor embodied in GE exports to FR
LAB(FR→GE) = FR labor embodied in FR exports to GE
all exports amount to: 1 million ECU
K/L ratios for GE and FR
Factor intensity of trade
LAB(GE→FR) = GE labor embodied in GE exports to FR
LAB(FR→GE) = FR labor embodied in FR exports to GE
K/L ratios for GE and FR
K/LGE > K/LFR : according to HO, FR “exports” labor to GE
: LAB(FR→GE) > LAB(GE→FR)
Bilateral comparisons:
If yes: GREEN
If no: RED
1975
GE FR IT
NL BE DK
GE --FR
IT
NL
BE
DK
-----------
1975
GE FR IT
NL BE DK
GE --FR
IT
NL
BE
DK
-----------
Large countries behave according to HO
Small countries behave according to HO
1985
GE FR IT
NL BE DK
GE --FR
IT
NL
BE
DK
-----------
1985
GE FR IT
NL BE DK
GE --FR
IT
NL
BE
DK
-----------
Large countries behave almost according to HO
Small countries behave according to HO
Factor intensity of trade
6 out of 6 in 1975
and 5 out of 6 in 1985
is a wonderful score!
Factor intensity of trade
6 out of 6 in 1975
and 5 out of 6 in 1985
is a wonderful score!
But (admittedly) the “ sample size” is rather small
Use WIOD:
to include more countries
to include refinements (types of labor, capital)
Vertical Specialization
Production processes more and more split up
in subsequent phases, carried out in different countries
→ Trade in intermediate goods and services becomes
more and more important
→ increase in interconnectedness of industries across
countries
→ intercountry IO tables reflect exactly that
→ measure vertical specialization in intercountry IO tables
Vertical Specialization
Measuring vertical specialization
Hummels, Ishii & Yi (JIE, 2001):
import content of the exports
Vertical Specialization
Z
M
v´
x´
c
e
x
Z = intermediate deliveries matrix
c = domestic consumption vector
e = gross exports vector
x = gross output vector
M = imports matrix
v´ = value added vector
Vertical Specialization
Z
M
v´
x´
c
e
x
Z = intermediate deliveries
→ A = input coefficients
→ (I – A)-1 = Leontief inverse
Vertical Specialization
Z
M
v´
x´
c
e
x
Z = intermediate deliveries
→ A = input coefficients
→ (I – A)-1 = Leontief inverse
M(I – A)-1e = imports necessary for exports
s´M(I – A)-1e = total imports necessary for
exports, s´ = summation row vector = (1,…,1)
sM (I  A )1e
VS 
se
Z11
Z12
c1
e1
x1
Z21
Z22
c2
e2
x2
ZR1
ZR2
(v1)´
(v2)´
(x1)´
(x2)´
Z12 = intermediate deliveries from 1 to 2
c1 = domestic consumption in country 1
e1 = exports to consumers in country 2 plus all
exports to the Rest of the World (= R)
ZR1 = imports from R to country 1
Z11
Z12
c1
e1
x1
Z21
Z22
c2
e2
x2
ZR1
ZR2
(v1)´
(v2)´
(x1)´
(x2)´
Three types of final demand that drive the model
Exports
e1
Imports from 2
Imports from R
Exports
Imports from 2
e1
(production in 1)
Imports from R
Exports
Imports from 2
e1
(production in 1)
inputs
Imports from R
inputs
Exports
Imports from 2
Imports from R
e1
(production in 1)
inputs
inputs
(production in 2)
Exports
Imports from 2
Imports from R
e1
(production in 1)
inputs
inputs
(production in 2)
inputs
(production in 1)
Exports
Imports from 2
Imports from R
e1
(production in 1)
inputs
inputs
(production in 2)
inputs
(production in 1)
inputs
(production in 2)
inputs
Exports
Imports from 2
Imports from R
e1
(production in 1)
inputs
inputs
(production in 2)
inputs
(production in 1)
inputs
(production in 1)
inputs
(production in 2)
inputs
inputs
(production in 2)
inputs
Consumption
c1
(production in 1)
Exports
inputs
(production in 1)
inputs
(production in 1)
Imports from 2
Imports from R
inputs
(production in 2)
inputs
inputs
(production in 2)
inputs
inputs
(production in 2)
inputs
Final demand 2
c2 + e 2
(production in 2)
Exports
inputs
(production in 1)
inputs
(production in 1)
Imports from 2
Imports from R
inputs
(production in 2)
inputs
inputs
(production in 2)
inputs
Vertical Specialization
Collect all exports and all imports (from 2 and from R)
all imports from 2 + all imports from R
VS =
all exports by 1
Use some matrix algebra, then:
it is exactly the same as for the single country case
Conclusion:
to measure vertical specialization of a country
it is not necessary to use an intercountry IO table
Two exercises on CO2
Central question: does it matter whether we use an
intercountry IO table (and how much)?
Data from Nori Yamano
37 countries, 16 sectors, 80% of world-GDP
Two exercises on CO2
Abuse the data:
exports to RoW become part of domestic consumption
imports from RoW become part of value added
Why?
we want to work with a perfect world-IO table
if you cannot construct the table, adapt your world
Hence:
our world consists of 37 countries
that is, USA ≠ USA, USA = “USA”
Consumer responsibility CO2
Consumer responsibility (CR) of country 1=
all CO2 emissions (all over the world) that are necessary
for producing the “consumption” in country 1
Crucial element of the carbon footprint of country 1
Calculate the CR of country 1:
using the full world-table yields “true” answer
various cases with limited information
measure the percentage error for country 1
Do this for country 1, …, country 37
→ (unweighted) average % error
A11 A12 A13
A21 A22 A23
f11
f21
x1
x2
A31 A32 A33
↓
A
f31
↓
f•1
x3
↓
x
Z → input coefficients A
emission coefficients, row vectors (w1)´, (w2)´, (w3)´
“true” consumer responsibility for country 1:
[(w1)´, (w2)´, (w3)´](I – A)-1f•1
A11
+A21
f11
+f21
x1
x2
+A31
↓
A1
+f31
↓
f1
x3
↓
x
Case 1: only technical coefficients available for country 1
consumer responsibility (w1)´(I – A1)-1f1
Average error: -37.5%
i.e. reported CR is (on average) only 62.5% of “true” CR
A11
A21
0
0
0
0
f11
f21
x1
x2
A31
0
↓
A
0
f31
↓
f•1
x3
↓
x
Case 2: information for imports from RoW
• use true emission coefficients: (w1)´, (w2)´, (w3)´
[(w1)´, (w2)´, (w3)´](I – A)-1f•1
average error: -27.6%
• use emission coefficients of country 1 only
[(w1)´, (w1)´, (w1)´](I – A)-1f•1
average error: -31.0%
A11 0
A21 A2
0
0
f11
f21
x1
x2
A31
A3
f31
↓
f•1
x3
↓
x
0
↓
A
Case 3: technical coefficients for all other countries
• use true emission coefficients: (w1)´, (w2)´, (w3)´
[(w1)´, (w2)´, (w3)´](I – A)-1f•1
average error: +0.3%
• use emission coefficients of country 1 only
[(w1)´, (w1)´, (w1)´](I – A)-1f•1
average error: -7.9%
A11
0
A21+A31 0
↓
A
f11
f21+f31
x1
x2+x3
↓
f•1
↓
x
Case 4: aggregated RoW
• use true emission coefficients: (w1)´, (wRoW)´
[(w1)´, (wRoW)´](I – A)-1f•1
average error: -29.0%
• use emission coefficients of country 1 only
[(w1)´, (w1)´](I – A)-1f•1
average error: -31.0%
A11
0
A21+A31 A1
↓
A
f11
f21+f31
x1
x2+x3
↓
f•1
↓
x
Case 5: aggregated RoW, estimate RoW using country 1
• use true emission coefficients: (w1)´, (wRoW)´
[(w1)´, (wRoW)´](I – A)-1f•1
average error: -16.1%
• use emission coefficients of country 1 only
[(w1)´, (w1)´](I – A)-1f•1
average error: -20.9%
Consumer responsibility CO2
Conclusion:
underestimation is (on average) substantial
unless a lot of information is available (Case 3)
Estimating RoW effects
We will never be able to cover all countries
there will always remain a RoW
How does this affect our findings, and what can we do?
Our world covers only 37 countries
“delete” one of them (which plays RoW)
and consider the effects on consumer responsibility CO2
large effects: neglecting RoW matters
small effects: who cares about RoW?
A11 A12 0
A21 A22 0
0
0
0
↓
A
f11 f12
f21 f22
f31 f32
x1
x2
x3
↓ ↓
f•1 f•2
↓
x
Delete country 3
Case 1:
CR country 1: [(w1)´, (w2)´, (wav)´](I – A)-1f•1
→ % error country 1
CR country 2: [(w1)´, (w2)´, (wav)´](I – A)-1f•2
→ % error country 2
wav = average emission coefficients (of countries 1 and 2)
A11 A12 0
A21 A22 0
0
0
0
↓
A
f11 f12
f21 f22
f31 f32
x1
x2
x3
↓ ↓
f•1 f•2
↓
x
Delete country 3
% error country 1, % error country 2
finally: weighted average % error
Example: delete China (i.e. treat China as RoW)
% error in any other country is on average -5%
A11 A12 0
A21 A22 0
A31 A32 0
↓
A
f11 f12
f21 f22
f31 f32
x1
x2
x3
↓ ↓
f•1 f•2
↓
x
Delete country 3
Case 2: extra information on imports from RoW
CR country 1: [(w1)´, (w2)´, (wav)´](I – A)-1f•1
→ % error country 1
CR country 2: [(w1)´, (w2)´, (wav)´](I – A)-1f•2
→ % error country 2
A11 A12 0
A21 A22 0
A31 A32 Aav
f11 f12
f21 f22
f31 f32
x1
x2
x3
↓
A
↓ ↓
f•1 f•2
↓
x
Delete country 3
Case 3: extra information on imports from RoW and
estimate technical coefficients of RoW
CR country 1: [(w1)´, (w2)´, (wav)´](I – A)-1f•1
→ % error country 1
CR country 2: [(w1)´, (w2)´, (wav)´](I – A)-1f•2
→ % error country 2
Aav = average technical coeffients
of countries 1 and 2
Estimating RoW effects
Average % errors
Germany
Japan
USA
China
Russia
Average
Case 1
-2.3
Case 2
-1.5
Case 3
+1.4
-1.3
-4.8
-5.3
-0.8
-3.4
-4.9
+2.0
+2.2
-3.0
-4.8
-0.9
-4.5
-0.7
-4.0
+0.2
Case 1
Case 2
Case 3
Germany
-2.3
-1.5
+1.4
Japan
-1.3
-0.8
+2.0
USA
-4.8
-3.4
+2.2
China
-5.3
-4.9
-3.0
Russia
-4.8
-4.5
-4.0
Average
-0.9
-0.7
+0.2
• Move from left to right: add more information
average underestimation decreases
• Delete GE/USA/Japan (“clean and efficient” countries)
estimate them by the average
(which is less “clean and efficient”)
NL depends on USA, but US emissions are overestimated
hence consumer responsibility of NL is overestimated
Case 1
Case 2
Case 3
Germany
-2.3
-1.5
+1.4
Japan
-1.3
-0.8
+2.0
USA
-4.8
-3.4
+2.2
China
-5.3
-4.9
-3.0
Russia
-4.8
-4.5
-4.0
Average
-0.9
-0.7
+0.2
Tentative conclusion for WIOD:
• GDP of RoW in WIOD is approximately 15%
• omission of USA from “37 world”
has limited effects (less than 5%)
• suggests that WIOD will generate reliable results
• but: these are average results
• for specific countries the errors may be substantial
• oil-producing countries are not included in WIOD
Is there a life after WIOD?
Consider the effects involved in off-shoring
or processing trade in general
Lot of imports and exports, little value added
little activity, little CO2 emissions
Recent debate about China’s CO2 emissions
claim: a substantial part is in China’s exports
Use an IO analysis
Is there a life after WIOD?
In 2002:
of all CO2 emitted in production activities
20.3% due to exports
(and thus 79.7% due to domestic “consumption”)
20.6% for SO2, 20.7% for Nox
Strongly increasing over time:
Weber et al (2008), for CO2
21% in 2002
33% in 2005
Is there a life after WIOD?
Claim: these numbers are seriously overestimated
Why?
Production activities related to processing trade
requires: essentially imported inputs
almost no domestically produced inputs
involves little emissions at home
Ordinary production:
large dependence on domestically produced inputs
little dependence on imported inputs
Is there a life after WIOD?
Input-output tables reflect the “average” production structure
processing trade 50% of exports, but
a minor share of production
(>1 billion Chinese consumers)
production related to processing trade
receives a small weight in the “average” production
Conclusion: ordinary IO tables cannot do the job
we need special IO tables
better reflecting the input structure of processing trade
Tripartite IO table (Chen et al., 2006)
ZDD
0
ZDP
0
ZDN
0
fD
0
0
eP
xD
xP
ZND ZNP ZNN
M D MP MN
(vD)’ (vP)’ (vN)’
fN
eN
xN
(xD)’ (xP)’ (xN)’
Distinguish between three types of production
 D = for domestic use only
 P = for processing exports
 N = for non-processing exports
and other production of foreign-invested enterprises
(high dependence on imports and many indirect exports)
Is there a life after WIOD?
In 2002:
of all CO2 emitted in production activities
20.3% due to exports
Using the tripartite IO table:
of all CO2 emitted in production activities
12.6% due to exports
CO2 emissions due to exports are overestimated by 61%!!
Is there a life after WIOD?
In 2002:
of all CO2 emitted in production activities
12.6% due to exports
CO2 emissions due to exports are overestimated by 61%!!
Is this to be viewed as a ‘measurement error’?
The ‘measurement error’ has the size of
the total CO2 emissions of the Netherlands, or Poland
or 1.5 times Turkey
Is there a life after WIOD?
Conclusion:
• using “ordinary” IO tables grossly overestimates the CO2
emissions involved in exports when processing trade is
present
• this applies to countries for which:
 processing trade is a major part of exports
 large domestic market
 characteristic production structure is not reflected in
the average production structure (which is
dominated by the structure for domestic production)
Is there a life after WIOD?
Yes, there is a life after WIOD!
Is there a life after WIOD?
Yes, there is a life after WIOD!
Thank you for your attention.