115

Improving the analysis of global value chains:

the UNCTAD-Eora Database

Bruno Casella, Richard Bolwijn, Daniel Moran and

Keiichiro Kanemoto*

The UNCTAD-Eora Global Value Chain (GVC) database offers global coverage
(189 countries and a “Rest of World” region) and a timeseries from 1990 to
2018, reporting on key GVC indicators. This paper explains the methodology for
compiling the UNCTAD-Eora GVC database, including nowcasting employed in
the estimation of recent years; second, it provides a comparison of the results
against other value-added trade databases, with a focus on the OECD Trade in
Value Added (TiVA) dataset; and lastly discusses the relevance of GVC data for
the analysis of globalisation patterns, particularly at the intersection between trade,
investment and development.
Keywords: trade in value added; MRIO; global value chains; complex value chains;
value added in export; input-output analysis

1. Introduction

A pivotal element in the analysis of international production are global value chains
(GVCs), which are fragmented and geographically dispersed production processes
where different stages are located across different countries. GVCs are coordinated
by multinational enterprises (MNEs) investing in productive assets worldwide and
trading inputs and outputs intra-firm, at arm’s length or through their network of
non-equity mode (NEM) partners. UNCTAD estimates that up to 80 per cent of
global trade involves MNEs (World Investment Report 2013). In this respect, the
analysis of GVCs is fully complementary to the analysis of FDI and international
production.

* Bruno Casella and Richard Bolwijn are at the United Conference on Trade and Development. Daniel
Moran and Keiichiro Kanemoto work at Eora. Correspondence with the authors may be addressed
jointly to Bruno Casella (Bruno.Casella@unctad.org) and the Eora MRIO maintainers (info@worldmrio.
com). The views expressed in this paper are solely those of the authors.
116 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

Recently, major analytical developments in the treatment of inter-country input-

output tables have opened new avenues for the empirical research on global value
chains. In particular, the availability of databases that break down trade according
to the origin of its value added (“value added trade” or “value added in exports”
data) enables the analysis of GVC patterns by countries and industries, at a level of
granularity that was unimaginable as recent as ten years ago. The most important
cross-regional value-added trade databases include the UNCTAD-Eora GVC
database, the World Input-Output Database (WIOD) and the OECD’s Trade in Value
Added Database (TiVA). Major regional initiatives include the Asian Multi-Region
Input-Output Database from the Asian Development Bank and the South-American
Input-Output Table from the Economic Commission for Latin America and the
Caribbean (ECLAC). Table 1 provides an account and a comparison of the different
and ongoing initiatives to map GVCs (see also Tukker and Dietzenbacker, 2013).
The UNCTAD-Eora GVC database was initially launched in the context of the analysis
conducted for the World Investment Report 2013 (WIR13), with its main theme
“Global Value Chains: Investment and Trade for Development” (UNCTAD, 2013).
Compared with alternative databases, its distinctive feature is broad geographical
coverage, including virtually all countries. Owing to this comprehensive coverage
the database has become the preferred reference source of value-added trade
data in analysis involving developing economies (AfDB, OECD, & UNDP 2014;
UNECA, 2015; UNIDO, 2016; IMF, 2015a; IMF 2015b; IMF 2016a; IMF 2016b).
Given the importance of GVC analysis in the context of globalization and
development and the high demand for value-added trade data, particularly for
developing countries, UNCTAD-Eora has upgraded its GVC database. This has led
not only to an update of the 2013 dataset to include GVC indicators up to 2015
but also a new improved version, featuring a “nowcast” methodology to project
value-added trade data from 2016 to 2018. This step addresses one of the main
weaknesses of available value-added trade databases (including the WIOD, TiVA
and the previous version of the UNCTAD-Eora GVC database), namely the time lag
of two to three years between the most recent data and the time of the analysis.
A further update of the UNCTAD-Eora GVC database, including GVC indicators
for 2016 and 2017 based on actual data, is in preparation and will be published in
conjunction with this paper.
The main outcome of the UNCTAD-Eora database is a set of basic GVC indicators,
including foreign value added (foreign value embedded in a country’s exports),
domestic value added (domestic value embedded in a country’s exports) and
domestic value added embedded in other countries’ exports. Other important
GVC indicators, such as GVC participation, can be easily computed from the
three basic indicators (Koopman et al., 2014). UNCTAD-Eora GVC indicators are
 Table 1. Efforts to map GVCs (status as of August 2019)
Project Institution Data sources Countries Industries Years Comments

UNCTAD/Eora National Supply-Use and I-O tables, 189 26-500 1990–2015 Meta database drawing together many sources
UNCTAD-Eora
and I-O tables from Eurostat, IDE- depending on (nowcast for 2016, and interpolating missing points to provide broad,
GVC Database
JETRO and OECD the country 2017 and 2018) consistent coverage

Trade in Value OECD National I-O tables 64 34 2005–2015 Information on all OECD countries, and 27 non-
Added (TiVA) (projections 2016) member economies (including all G20 countries)
dataset

World Input-Output Consortium of 11 National Supply-Use tables 43 56 2000–2014 Based on official national account statistics; uses
Database (WIOD): institutions, EU funded end-use classification to allocate flows across
2016 Release partners and countries

Other multi-region input-output databases

EU-based consortium, National supply-use tables 44+5 200 1995–2013 Covers 44 countries plus five rest-of-world
EXIOBASE
exiobase.eu regions

ADB Multi-Region Asian Development An extension of WIOD which 45 35 2000, 2005– The information for the 5 additional Asian
Input-Output Bank includes 5 additional Asian 2008, 2011 countries are estimates methodically produced to
Database economies (Bangladesh, Malaysia, assist research and analysis, not official statistics
(ADB MRIO) Philippines, Thailand and Viet Nam)
Improving the analysis of global value chains: the UNCTAD-Eora Database

Purdue University Contributions from individual 121 65 2004, 2007, Includes data on areas such as energy volumes,
Global Trade
researchers and organizations countries 2011, 2014 land use, carbon dioxide emissions and
Analysis Project
plus 20 international migration.
(GTAP)
regions

ECLAC and Institute National I-O tables 10 40 2005 Based on official information from National
South American of Applied Economic Accounts
Input-Output table Research (IPEA) from
Brazil

Source: UNCTAD.
117
118 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

publicly available at granular year-, country- and industry-level on the UNCTAD-

Eora webpage.1
The intention is to establish the UNCTAD-Eora project as a continuing project
for the update and improvement of GVC data and analysis, with annual updates
envisaged.
In this context, this paper has two objectives: First, it presents the analytic and
methodological construction of the UNCTAD-Eora database (sections 2 and
3). Second, it compares results with other available databases, particularly the
OECD TiVA, for data validation purposes (section 4). The concluding section puts
the UNCTAD-Eora database in the broad context of the analysis of the trade-
investment-development nexus: it shows how GVC data can provide an important
perspective on some relevant trends at the intersection between these three key
areas in modern globalization.

2. The analytical background of the new UNCTAD-Eora

database

In this section we briefly retrace the steps that lead to the establishment of the
new UNCTAD-Eora database. The first step (section 2.1) – the construction of a
multiregional input-output (MRIO) dataset – is the most technically complex and
computationally intensive. We present it only qualitatively; for more detail the
existing literature is referenced. Once an MRIO is available, some straightforward
algebraic steps allow to fit the relevant information contained in the MRIO into the
framework of value-added trade and derive the key GVC indicators (section 2.2).
Finally, a nowcasting procedure is implemented to project value-added trade data
from the last available year onward (section 2.3). Unlike section 2.1 and section 2.2
which are essentially summaries of existing material, the treatment of nowcasting in
section 2.3 is new, hence its analytical elaboration here is more detailed.

2.1. The construction of the Eora MRIO dataset

This section provides an overview of how the Eora MRIO is constructed. For a more
comprehensive explanation, the primary reference paper is Lenzen et al. (2012).
Some more approachable summary papers are Lenzen et al. (2013); Moran and

1
http://worldmrio.com/unctadgvc/. For references to the UNCTAD-Eora database, cite this method
paper as follows:
Casella, B. et al. (2019). Improving the analysis of global value chains: the UNCTAD-Eora Database,
Transnational Corporations Journal 26(3). New York and Geneva: United Nations.
Improving the analysis of global value chains: the UNCTAD-Eora Database 119

Geschke (2013); and Moran (2013). The documentation section of the Eora website
(at http://worldmrio.com) also provides several papers and reports that present the
main elements of I/O analysis.
The Eora dataset provides a multi-region input-output table at the global level to
estimate value added in trade. The construction of the Eora MRIO table follows
several steps.
a. The starting points are the national IO tables or supply/use tables (SUTs).
National SUTs are recommended over input-output tables because they provide
information on both products and industries. However, the national statistics
bureaus in some countries still provide only input-output tables. A supply table
provides information on products produced by each domestic industry and a
“use” table indicates the use of products by industries or final users. As SUTs
are only available for a limited number of countries, the remaining countries
are hence represented by input-output (I/O) tables, which can be sourced
from available data or compiled according to a range of assumptions. In order
to avoid departures from the original raw data, EORA preserves the sectoral
classification from each data provider. The complete list of raw data sources
involved in preparing the IO table for each country in Eora is available at the
Quality Report section of the Eora website and in the Supplementary Information
of Lenzen et al. (2012).
b. National SUTs and I/O tables are linked through international trade statistics
using import tables to obtain a multi-region input-output table. At this step, an
estimation procedure is used to construct so-called “off-diagonal” trade blocks,
estimating flows from each export sector in each origin country (rows) to each
importing sector in each destination country (columns). Trade data is most often
reported by product and by producer and consumer country. However, an off-
diagonal trade block in an IO table requires knowing how goods from each
exporting sector are absorbed into each importing sector. Put another way, the
raw data is three-dimensional, but the IO table requires four dimensions. Thus,
creating the trade blocks involves several assumptions and estimation steps.
The challenges and procedures used to estimate trade are presented in full in
Lenzen et al. (2012).
c. After obtaining a first estimate of an MRIO table, the resulting trade data are
balanced through an industry-level balancing condition: the total output
produced by each sector must equal the sum of the inputs used by that sector.
This has been achieved via “constraints data”: i) Input-output tables and main
aggregates data from national statistics offices; ii) Input-output compendia from
Eurostat, IDE-JETRO and OECD; iii) The UN National Accounts Main Aggregates
Database and official country data; iv) The UN COMTRADE and UN Service
Trade international trade databases. An optimization procedure (a variant of the
120 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

RAS algorithm that can handle multiple conflicting constraints; see Lenzen et
al., 2014) is set up so that the solution should be some compromise table that
respects the initial estimates and also satisfies constraints with as little deviation
as possible. For the optimization exercise, a standard error is estimated for each
data point based on the reliability of the data. In general, larger values are taken
to be more reliable than smaller values, in relative terms. Data from national
statistics agencies are assumed to be more reliable than other sources. The
ordering of data sources listed above largely corresponds to the data reliability
assumed in assigning standard errors.
d. The time series is constructed iteratively, by starting with an initial year estimate
(year 2000), balancing it with all the starting year constraints, and taking the
solution as the initial estimate for the following year, and so on. In each year, all
available data for that year (GDP totals, trade data, new I/O tables, interpolated
I/O table estimates, and so on) are overlaid onto the initial estimate of that year,
and the table is rebalanced. The practice of using the previous-year solution as
the initial estimate for the subsequent year has an effect to “smooth” timeseries
data, though other constraints that introduce “jumps” will also be considered in
the solution table for each year.
Figure 1 shows a simplified MRIO table, considering only one industry for two
countries.
Figure 1. Structure of an MRIO Table

Exports from A to B Exports from A to B

of Intermediates of Final Products

Intermediate use Final demand

Country A Country B Country A Country B Gross
Industry Industry Industry Industry output

Intermediate Intermediate use

Final use of Final use by B of
Country A Industry use of domestic by B of exports
domestic output exports from A XA
output from A

Intermediate use Intermediate

Final use by A of Final use of
Country B Industry by A of exports use of domestic
exports from B domestic output XB
from B output

Value added VA VB

Gross input XA XB
Improving the analysis of global value chains: the UNCTAD-Eora Database 121

The rows in an MRIO table indicate the use of gross output from a particular sector
in a country. The gross output X produced in country A (first row) can be used by
country A itself as intermediate or as final consumption, or by country B, again as
an intermediate input or final product. From here, we can retrieve a measure of
gross exports from A to B, summing the intermediate and final output produced in
country A and used in country B (the grey blocks in the example above).
The columns of an MRIO table provide information on the technology of production,
as they indicate the amounts of intermediates needed for the production of the
gross output whose use is then decomposed along the row. Hence, each column
provides the domestic and foreign share of intermediates in the production of one
unit of output. The first column thus shows how much domestic inputs contribute
to the production of the gross output of country A (first cell, “Intermediate use of
domestic output”), and how many inputs are sourced from abroad through imports
(second cell, “Intermediate use by A of exports from B”). The difference between
the gross output produced in each country and the sum of the (domestic and
foreign) inputs necessary for production yields the value added generated in each
country (V).

2.2. Deriving value-added trade from Eora MRIO

The derivation of value-added trade from the MRIO table follows the standard
approach proposed by Koopman et al. (2010; 2014). Here we provide a concise
description and we refer to Koopman’s paper and other reviews such as the
OECD’s De Backer and Miroudot (2013) and the IMF’s Aslam et al. (2017) for the
details. The IMF paper in particular explicitly uses the Eora MRIO computational
framework to derive value-added trade indicators. Some other important papers
addressing issues in the computation of value-added trade include Hummels et al.
(2001), Johnson and Noguera (2012), Stehrer et al. (2012), Timmer at al. (2012),
Wang et al. (2013, 2017a, 2017b), Los et al. (2016), Johnson and Noguera (2016),
Timmer et al. (2016), Antras and de Gortari (2017), and Los and Timmer (2018).
We first establish standard IO analysis identities for an MRIO table with N countries
and H industries:

(1)
122 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

where x is the (NH x 1)2 vector of gross outputs by countries and by industries, T
is the corresponding vector of intermediate uses, y is final demand. From (1), we
introduce the (NH x NH) key matrices of the GVC construction: the technological
coefficient matrix A and the Leontief inverse L (Leontief, 1970).
The fundamental relationships in (1) can be applied to the “value-added trade”
framework. After introducing the (NH x NH) diagonal matrices V and E, reporting
respectively value-added share and exports by countries and industries, we define
the matrix (NH x NH) of embodied value-added flows F as follows:

(2)

rs
where F is a (H x H) matrix showing inter-sector flows between country r and
country s (domestic flows in the case that r and s are the same country). The
matrix F is the key matrix of our analysis (figure 2). The matrix essentially describes
how the value added contained in the exports of each country (and industry) is
generated (by column) and distributed (by row) across countries. Henceforth, in
order to facilitate the intuition, we will describe the elements of F (2) as if they were
scalar (this is equivalent to considering an economy with only one product) rather
than (H x H) matrices as in the general case. Thus, the first column of the matrix
describes the value added contained in the export of country 1. This is composed
of two parts:
11 11 1 11 1
• the term F (in the matrix multiplication we have that F = V L E ) denotes
the Domestic Value Added (DVA) content of exports of country 1;
r1 r1 r r1 1
• the generic term F (in matrix notation F = V L E ) denotes the Foreign
Value Added (FVA) content of exports of country s generated by country r
(with r ≠ 1). Recall that the production of output by country s (part of which is
exported) requires inputs from other countries. In producing these inputs, the
other countries also generate value added. Hence, this term represents the
r
share of value added that has been generated in country r (V ) and that has
r1 1
been imported by country 1 (L ) in order to produce its exports (E ).
The (column) sum of domestic and foreign value added, by construction, will yield
the total exports of country 1. The other columns of the F matrix replicate the

2
The notation (NH x 1) refers to the dimensions of a matrix with NH (i.e. N times H) rows and 1 column
(a column-vector). The same type of notation is used throughout the paper to provide the dimensions
of any matrix when relevant.
Improving the analysis of global value chains: the UNCTAD-Eora Database 123

exercise for the other countries. Therefore, in column 2 of the matrix we will find the
22
term F , which denotes the DVA content of exports of country 2, as well as the
r2
generic term F , which denotes the FVA content of exports of country 2 generated
by country r, and so on. Hence, the DVA can be read on the diagonal of the matrix
rr
as the generic term F for any country r in the dataset.
Finally, by reading the matrix along the row rather than along the column (and
rr
excluding the diagonal terms F ), we have an indication of how much of each
country’s domestic value added enters as an intermediate input in the value added
exported by other countries. The latter terms are what Koopman et al. (2014) call
“indirect value-added exports” (DVX). Clearly, by constructing what each country
contributes to all the others in terms of indirect value-added exports has to be
equal at the world level to what each country sources from all the others in terms of
foreign value added, that is at the world level FVA = DVX. The latter gives a rough,
though not perfect, proxy of the double counting embedded in the gross (official)
trade figures.
Figure 2. The matrix of the value-added content of trade

DVA DVX

Country 1 Country 2 Country 3 ... Country K ... Country N

11 12 13 1K 1N
Country 1 F F F ... F ... F
21 22 23 2K 2N
Country 2 F F F ... F ... F
31 32 33 3K 3N
Country 3 F F F ... F ... F

... ... ... ... ... ... ... ...

FVA K1 K2 K3 KK KN
Country K F F F ... F ... F

... ... ... ... ... ... ... ...

N1 N2 N3 NK NN
Country N F F F ... F ... F

2.3. Nowcasting value added trade for more recent years

Compared to the original 2013 version, the current version of the UNCTAD-Eora
database (as of August 2019) includes a nowcasting procedure to extend the time
horizon covered by the GVC time-series to the most recent years. Specifically, the
UNCTAD-Eora GVC results are based on reported data for the years from 1990
to 2015, and are nowcasted to estimate results for 2016, 2017 and 2018. The full
MRIO elaboration is available only until 2015 because of the time lag (2-3 years) of
the underlying macroeconomic data.
124 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

The nowcasting is based on estimates from the IMF’s World Economic Outlook
(WEO), December 2017 edition (IMF, 2017). The WEO provides estimates of the
annual change of GDP, imports and exports, in each country. These estimates are
provided as nowcasts for recent years (2016, 2017 and 2018 for the 2017 edition)
and with 2+ year predictions for selected indicators.
The UNCTAD-Eora nowcasting of GVC indicators is performed in two stages. First,
the value-added contribution from each origin country is adjusted according to their
(nowcasted) change in GDP. Second, for each exporting country, resulting value-
added contributions are then rescaled and normalized in order to sum the WEO
nowcasted values for gross exports. In other words, the WEO GDP nowcasting
determines the changes in the distribution of a country’s export among its value-
added contributors, while export nowcasting affects the change in the level of
value-added trade. In this way, nowcasting essentially provides a simple and
transparent way to project GVC indicators from actual year t to a following year
t+1, by incorporating the macroeconomic estimates from the IMF’s WEO into the
standard GVC setting of section 2.2.
We may provide a formal elaboration of the procedure. The mathematical treatment
presented below will be more detailed than for the standard GVC calculations
illustrated in the previous section (noting that the basic computation of value added
in trade is presented in a number of papers already, cited in the previous section).
To this end, we also develop the formulas in the most general case of N countries
and H industries.
Let F then be the final GVC matrix (2) at time t containing data from the latest
observed period. For each country r = 1, 2, …, N, let and be
diagonal (H x H) matrices, reporting on the diagonal the sum between the unit and
the (WEO-nowcasted) annual growth rate of GDP and export respectively, say
and . In principle, of course, each industry would have its own growth rates, i.e.
the elements in the diagonal of the matrices should be different. However, this is
not possible in the nowcasting setting as the WEO estimates are provided only at
the aggregate level.
First, we define the adjusted matrix .
Step 1. Value-added adjustment:

(3)

where is a (NH x NH) block diagonal matrix with matrices

(r = 1, 2,..., N) on the diagonal, while F is the (NH x NH) block diagonal matrix
Improving the analysis of global value chains: the UNCTAD-Eora Database 125

defined by (2). The generic element of the (H x H) matrices (r, s = 1, 2, …, N)

in (3) is then given by with i, j = 1, 2, …, H.
In this context, consistent with the matrix notation introduced in (2), is the value-
added share of country r in the production of product i; is the element of the
Leontief inverse matrix corresponding to the countries’ pair (r, s) and industries’ pair
(i, j), is the export of product j by country s and the GDP growth of country r.3
At time t+1, the value added extracted by country r at time t, represented by the
generic elements (s=1, 2, …, N) of the matrix F in (3), is therefore adjusted
to account for economic performance of country r between t and t+1, as reflected
by the country’s GDP growth, .

The matrix in (3) potentially defines a new structure of the countries’ export at
time t+1; this is denoted by a (NH x NH) block diagonal matrix where each
component (s= 1, 2,…, N) is a (H x H) matrix reporting the exports of country s
as implied by (3). Otherwise stated, the diagonal elements of correspond to the
sums of the NH columns of .4 These elements are determined by the structure
of the exports at time t, by the existing production technology at time t and by the
economic growth between t and t+1.

In the second step of the nowcasting we incorporate in the GVC estimation the
WEO information on the export’s growth rates by country, (s = 1, 2, …, N). Let
be the (NH x NH) export matrix, as resulting by the application of the WEO
nowcast of export growth to export at time t, i.e. = x where is the
matrix of exports at time t and is a diagonal block matrix with components
(s = 1, 2, …, N).

The export structure resulting from (3) does not coincide with the one implied by
the WEO nowcasting, i.e. ≠ . Thus, we need to normalize and rescale (3) to
make sure that the resulting export at time t+1 is consistent with nowcast provided
by the WEO.

3
More specifically, when r = s, the element indicates the domestic value added extracted by
country r, related to the intermediate use of domestic output i necessary to meet export levels of
product j; if s ≠ r, it indicates the foreign value added generated by country r, related to the provision
of the intermediate input i necessary to meet export of product j from country s.
4
Formally, for each exporting country s, the (H x 1) vector of exports implied by (3), say (the vector
of the diagonal element of the matrices , is defined by where i is a unit vector (1 x
NH) and is a (NH x H) representing the value-added structure of export of country s.
126 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

Analytically, this is equivalent to calculate a new matrix as follows:

Step 2. Normalization and rescaling:

(4)

The first product in (4) normalizes value-added exports resulting from (3), the second
rescales them in order to sum aggregate exports implied by the WEO nowcasting.
The generic element of the block matrices (r,s = 1, 2, …, N) is then given by

. It is easy to verify that the

value-added shares implied by the matrix is the same as for , i.e.

for any i, j, r, s. At the same time, for each exporting country s and each industry
j, the sum of value added contributed by all other countries (domestic and foreign)
equals the export implied by the WEO nowcasting:

This nowcasting approach is simplified compared to the full procedure used to

compute value added in trade for years with observed data. In particular, the
lack of timely information on the sectoral composition of the economy and the
corresponding disaggregation between intermediate and final use, as provided by
national I/O tables, does not allow constructing a sectoral detailed MRIO such as
in figure 2. The most computationally intensive steps, illustrated in section 2.1,
are not possible in the nowcasting setting. Instead, the inter-country, inter-sectoral
structure of the economy is fully inherited from the last year, say t, for which full
macroeconomic data are available. What nowcasting does is to adjust the GVC
indicators at the national level at time t to account for the changes in the (relative)
economic performance of countries and the expected trend in exports, assuming no
change in the underlying economic structure. We also note that there is no explicit
balancing step in the nowcasting procedure, since the WEO provides balanced
forecasts (e.g. growth in exports from one country is 100 per cent absorbed by
growth in imports from other countries).
Improving the analysis of global value chains: the UNCTAD-Eora Database 127

Table 2 provides an example of the nowcasting approach using three countries with
one industry.

Table 2. Numerical example illustrating the nowcasting method

GVC indicators at GVC indicators at
time t (actual) Intermediate step time t+1 (nowcasted)
VA VA Nowcasted Adjusted Adjusted VA VA
Value added (VA) embodied added growth (WEO value value added embodied added
originating in: in exports shares input)* added shares in exports shares

Country A 700 70% 9% 763 71% 750 71%

Country B 100 10% 3% 103 10% 101 10%
Country C 200 20% 1% 202 19% 199 19%
Exports from A 1 000 100% 5% 1 068 100% 1 050 100%
* Nowcasted growth (shaded column) refers to GDP growth for the first three rows and to export growth for the last row.

3. Limitations and areas for further development

There are two main sources of uncertainty in the estimation of value-added trade
data and GVC indicators. The major one, discussed in section 3.1, is related
to the original construction of an MRIO, requiring modelling assumptions and
computational steps. This uncertainty is common to all MRIO approaches and it
stems from the complexity of the estimation problem inherent to the construction
of an MRIO, i.e. reconstructing the global network of bilateral trade flows across
sectors and countries in the most comprehensive and granular way. A second set
of uncertainties involves more specific data issues affecting the interpretation of
value-added trade data and GVC indicators (section 3.2).

3.1. Common limitations related to the construction of an MRIO

The topic of MRIO construction and reliability has been extensively discussed
in the MRIO literature (Wiedman et al., 2011, Tukker and Dietzenbacher, 2013,
Dietzenbacher, 2013, Tukker et al., 2018). This section is a brief, non-technical
overview only.
All MRIO databases are to some degree modelled. Some portions of the
databases are overdetermined, with multiple, conflicting reports, while others
are underdetermined and need assumptions or modelling to fill in portions of the
dataset not covered by official sources. In particular the trade blocks of an MRIO
are underdetermined. Trade statistics provide data as [good/service – country of
origin – absorbing country] tuples, while the MRIO database structure reports data
128 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

as [good/sector of export – country of origin – absorbing good/sector – absorbing

country]. This results in trade flows at the sector-to-sector level being inferred or
estimated.
The modelling and interpolation approach ranges in complexity from simple linear
interpolation to more complex proxy or statistical methods. In this context, there
is no “correct” global MRIO table. Rather, there is variety of models that differ in
how comprehensive and detailed they are and in how they treat conflicting and
missing data. In view of this uncertainty, every single data point in the Eora MRIO is
accompanied by an estimate of its standard deviation, reporting the extent to which
it was contested, interpolated, estimated or adjusted away from its original value
in order to assemble a balanced global I/O table. A large number of reliability and
confidence reports are made available on the Eora website.
The several available MRIO databases are constructed by independent research
teams. It should not be expected that they agree perfectly. They generally obey
similar macroeconomic constraints at the national level (total GDP, total exports,
imports, consumption, and value added created), though even on these basic
macroeconomic totals the MRIOs do not perfectly match. There are multiple data
providers for these macro statistics (the United Nations, the World Bank, national
statistics agencies) and the values are not always identical across providers. The
Eora website has a page that offers a comparison of the various MRIOs in terms
of their reported values for these key macroeconomic totals (http://worldmrio.com/
comparison/).
Even if the MRIOs were constructed using identical macroeconomic constraints,
there remains considerable room for variation across the independent models at
the sector level. The level of aggregation/disaggregation chosen is one major cause
of variation. The OECD database opts to aggregate national IO tables to a relatively
high degree (to 34 sectors). The WIOD database offers higher resolution (56
sectors). This means that the national IO tables have to be reclassified, aggregated
or disaggregated, in order to adjust the source national tables to match the 56
sector classification. The Eora database preserves each country’s national IO
table in its native classification scheme. Eora’s heterogenous classifications make
inter-country comparison difficult and makes the MRIO slightly more complex to
assemble and use, but, as major advantage, it minimally disturbs each original
national IO table. The details of how the sector-level results are constructed vary
substantially across the MRIOs. The effects of sectoral aggregation are well studied
(Steen-Olsen et al. 2014, de Koning et al. 2015).
A significant body of work has investigated the reliability of MRIO databases using
side-by-side comparison, sensitivity analysis, and using decomposition analysis
to isolate sources of divergence (Lenzen et al., 2010; Wilting, 2012; Geschke et
al., 2014; Moran and Wood, 2014; Wood et al., 2014; Inomata and Owen, 2014;
Improving the analysis of global value chains: the UNCTAD-Eora Database 129

Owen et al., 2014; Owen et al., 2016; Steen-Olsen et al., 2016; Owen, 2017;
Tukker et al., 2018; Rodrigues et al., 2018). Together, these studies indicate that
the major MRIOs agree to within +/-10 per cent for most values for most larger and
structurally central economies, and to within +/-30 per cent for smaller economies
or economies with less comprehensive or reliable data.

3.2. Other specific issues affecting value-added trade data

In this section we discuss some issues that proved to be particularly relevant in the
economic applications of GVC data and indicators, according to our experience
with UNCTAD-Eora database and feedbacks received from UNCTAD-Eora users.

Re-exports / re-imports
Re-exports refers to goods imported and then re-exported with null or negligible
transformation (e.g. goods that land, are warehoused and are then shipped
out). The accounting of re-exports can be problematic. Different countries may
account for re-exports differently. Additionally, the value of re-exports is sometimes
estimated. The estimated value of re-exports can form a significant portion of trade,
in particular for trade-intensive economies such as Belgium and the Netherlands.
When re-exports form a large share of imports or exports, inconsistencies in how
re-exports are reported in MRIOs or, whether they are excluded entirely, can drive
large divergences in the calculation of value added in trade. Eora preserves re-
exports. Other databases may handle re-exports differently. In the benchmark
provided in the next section, we shall see that such differences in the treatment
of re-exports is a major cause of divergence in results between GVC indicators as
estimated by the UNCTAD Eora and the OECD TiVA databases, in particular for
trade-exposed countries such as Belgium and the Netherlands.

Processing trade
Most MRIO databases and published Chinese IO tables treat export processing as
structurally identical to domestic production. They do not differentiate the technical
coefficients between production for exports and production for domestic use.
However, in reality, production for exports often uses more foreign imports than
does production intended for domestic consumption (Dietzenbacher et al. 2012).
Processing exports account for 35-50 per cent of total Chinese merchandise
exports (varying by year) so this homogeneity assumption affects a substantial
share of the total economic activity in China. Mexico, and likely other countries, face
a similar situation whereby export-led firms operate with a different mix of inputs
than their peers selling to the domestic market. It is important to differentiate export
processing. Chen et al. (2018) empirically studied the importance of distinguishing
130 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

export processing zones. They conclude, “[I]f China’s processing trade is

undistinguished… China’s bilateral net trade in value added with some economies,
such as Japan, Korea and Taiwan, would be significantly underestimated, while it
would be significantly overestimated for some other economies, such as the United
States”. However, official public data on processing trade for China are currently
not available, making undifferentiated treatment a necessary choice.

Re-imported domestic value added

In complex value chains it is possible that value is added in a domestic sector,
the intermediary good is then exported, value is added in one or more foreign
countries, and the final good is then imported back into the originating country. This
is called re-imported or “feedback” value added.
Investigation of “re-imported DVA”, e.g. by Koopman (2012), shows that the latter
is relatively small at the world level (though it might be slightly more significant
for some countries or industries than others). Koopman et al. (2012) estimate
the domestic content of foreign exports that finally return home at 4 per cent of
gross exports in 2004. The results computed by Stehrer (2012), using the WIOD
database, indicate at the world level a range from a minimum share of 2.6 per
cent in 1995 to a maximum of 3.3 per cent in 2008, with the figure for 2009 at 2.9
per cent. The OECD/WTO initiative, in turn, estimates that the re-imported DVA
equals to just 0.6 per cent of world gross exports in 2009. The magnitude of these
feedback effects was also investigated by Moran et al. (2017). The study concludes
that re-imported value added usually comprises 2-6 per cent of value added in
imports for most countries and sectors.

***

Following the discussion of the issues above, it is possible to identify three areas
where future development would help improve the data accuracy and reliability of
the database. This list is not intended as a fully-fledged research agenda for future
work but rather as a partial list of issues that merit priority.
i. Improve results agreements across MRIO databases. Other fields have inter-
comparison projects or model suite projects that help implementors identify
errors and improve alignment across models.
ii. Improve sectoral detail that will offer high sector and product level resolution in
the results.
iii. Provide more consistent treatment of re-exports and processing trade.
Improving the analysis of global value chains: the UNCTAD-Eora Database 131

4. C
 omparison between UNCTAD-Eora and other GVC
databases

While there are several studies providing comparison and cross-validation of

the Eora MRIO against other MRIO databases (see section 3.1), less effort has
been made to directly compare the key GVC indicators across different value-
added databases. Our goal in this section is to contribute to covering this gap
by investigating the consistency between the UNCTAD-Eora GVC estimates and
results from other creditable GVC databases, particularly the OECD TiVA. In section
4.1 we present the results of a novel comparison between the UNCTAD-Eora GVC
database (version 2018) and the latest OECD TiVA database (December 2018).
In section 4.2, we briefly recall the findings from two previous studies that have
performed similar cross-validation, IMF’s paper by Aslam et al. (2017) and UNCTAD
(2013b).
Overall, all these efforts confirm a general alignment of UNCTAD-Eora GVC results
with the OECD TiVA at the countries’ level. This is an important, and not at all
obvious, achievement given that the coverage of the UNCTAD-Eora database is
higher than that of the other databases (see table 1).

4.1. UNCTAD-Eora GVC Database and the OECD’s TiVA (2018 versions)
We compare results from the new UNCTAD-Eora database and the OECD TiVA
(2018 versions). To run the comparison, we selected one key GVC indicator, the
foreign value-added share or FVA share, i.e. the share of foreign value added in
total export. This, and the corresponding domestic value-added shares, is the most
basic and fundamental GVC indicator. The comparison involves those years for
which both datasets report actual values, a time horizon between 2005 and 2015.
The reference year for most analysis is 2015, the most recent year of comparison.
Country perimeter includes all 64 countries covered by the OECD TiVA, a subset of
the 189 countries covered by UNCTAD-Eora.
Figure 3 shows the correlation between FVA share from UNCTAD-Eora and the
OECD TiVA for 2015. High correlation (linear correlation coefficient Rho= 0.75)
indicates an overall consistency between the results. A slope of the linear regression
line close to 1 (0.85) suggests that values of FVA shares are generally similar
between the two databases. The consistency between the results is substantially
preserved over time, as confirmed by figure 4a plotting FVA share across countries
and years. The correlation coefficient between the two sets of data is consistently
above 0.7 in all years considered (figure 4b).
132 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

Figure 3. FVA shares of UNCTAD-Eora and the OECD TiVA, 2015

70

60
FVA share UNCTAD-Eora

50

40

30

20

10 Slope = 0.85;
Linear correlation ρ = 0.75

0 10 20 30 40 50 60 70
FVA share OECD TiVA

Figure 4a. FVA shares of UNCTAD-Eora and the OECD TiVA, 2005–2015
70

60
FVA share UNCTAD-Eora

50

40

30

20

10

0 10 20 30 40 50 60 70

FVA share OECD TiVA

Figure 4.b. Linear correlation coefficient between UNCTAD-Eora and OECD TiVA
across countries by year, 2005–2015
Coefficient (ρ)

0.77
0.76
0.75
0.74
0.73
0.72
0.71
0.70
0.69
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Improving the analysis of global value chains: the UNCTAD-Eora Database 133

Figure 5 summarizes the results of the by-country comparison along the two critical
dimensions: comparison of values (x-axis) and of trends (y-axis). Almost 60 per cent
of the countries (36 out of 64) show highly consistent trends of FVA shares in the
period of interest 2005–2015 ( >0.6) and more than a third (23 countries) display
similar values ( I FVA share I < 5pp).

Figure 5. Summary of the comparison between UNCTAD-Eora and the OECD

TiVA, 2005–2015
Number of countries

23 (36%) 35 (55%) 6 (9%)

(Time series correlation 2005–2015, linear correlation ρ)

12 10 11 3 36 (57%)

0,6
Trend comparison

4 4 4 1 13 (20%)

0,3

7 4 2 2 15 (23%)

5 10 15
Value comparison
( FVA share 2015, percent points, absolute value)

There are six countries (Hong Kong, the Netherlands, Singapore, Belgium,
Lithuania and Malta) that present substantial divergence between the estimates
( I FVA share I > 15pp). These economies, particularly Hong Kong, Netherlands,
Belgium and Singapore have a large amount of imports and exports relative to their
total GDP, so the challenges discussed above relating to the macro constraints
of total imports and total exports, and the sector-wise attribution of value added,
become especially acute. Additionally, for these countries, the difference in the
treatment of re-exports between UNCTAD-Eora and the OECD TiVA (see section
3.2) may heavily affect the final estimation as high level of re-exports would amplify
UNCTAD-Eora FVA share relative to the OECD TiVA. Figure 6 tests this hypothesis
by comparing the two databases, both in their original form (left-hand side) and
134 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

after removing the re-export component from UNCTAD-Eora estimate. In half of the
cases (the Netherlands, Belgium and Lithuania), the values of UNCTAD-Eora and
the OECD TiVA substantially realign after removing re-export from the UNCTAD-
Eora estimate. Hong-Kong and Singapore are somehow surprising cases as we
would expect the level of re-exports to be high and relevant. These cases warrant
further consideration and analysis.
More generally removing re-export from the comparison further improves the
overall consistency between the UNCTAD-Eora and the OECD TiVA. For exemple
in figure 5, the share of countries with absolute delta less than 5 percent points
would increase from current 36 per cent to 58 per cent after removing the re-export
component.

Figure 6. Comparison between FVA shares of UNCTAD-Eora and the OECD TiVA
for selected (problematic) countries, with and without re-exports,
2015

Comparison with re-exports Comparison without re-exports

Hong Kong 27% 27%

31 38
57% 65%

Netherlands 28% 28%

26 3
54% 25%

Singapore 41% 41%

21 22
61% 63%

Belgium 34% 34%

20 2
55% 33%

Lithuania 32% 32%

51% 20 4
35%

Malta 59% 59%

19 15
40% 44%

OECD TiVA UNCTAD-Eora Delta pp, abs. value

4.2. Previous comparisons

The numerical comparison presented in the previous section is the most detailed
cross-check of UNCTAD-Eora GVC indicators but not the only one. Here we briefly
recall other two comparative analysis which generally confirm the consistency of
UNCTAD-Eora with the other available GVC databases.
 Figure 8. FVA shares of UNCTAD-Eora and the OECD TiVA, selected years (from Aslam et al, 2017)

Year 1995 Year 2000 Year 2005

50 60 60

40

40 40
30

20
20 20
10

Foreign Value Added, WTO OECD

Foreign Value Added, WTO OECD
Foreign Value Added, WTO OECD
0 0 0
0 20 40 60 80 0 20 40 60 80 10 20 30 40 50 60
Foreign Value Added, Eora MRIO Foreign Value Added, Eora MRIO Foreign Value Added, Eora MRIO

Year 2009 Year 2010 Year 2011

60 60 60

40 40 40
Improving the analysis of global value chains: the UNCTAD-Eora Database

20 20 20

Foreign Value Added, WTO OECD

Foreign Value Added, WTO OECD
Foreign Value Added, WTO OECD

0 0 0
10 20 30 40 50 60 0 20 40 60 80 0 20 40 60 80
Foreign Value Added, Eora MRIO Foreign Value Added, Eora MRIO Foreign Value Added, Eora MRIO
135
136 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

Aslam et al (2017) compare for different years the FVA shares of UNCTAD-Eora and
the OECD TiVA, essentially the same exercise as figure 3 but replicated on several
years. For illustrative purposes figure 8 reports some of their scatterplots, showing
a substantial alignment between the two datasets similar to what we found. The
authors conclude that “Overall, the scatterplots reassure us that Eora and the
OECD-WTO TiVA statistics are generally consistent with one another. Given this,
we can feel somewhat more comfortable using Eora for countries for which the
OECD-WTO data are not available. However, the researcher should be aware of
possible problems, given the method by which the input-output table have been
constructed for countries where no official supply-use tables are available. Some
important country examples, such as China, Hong Kong etc. … depending on the
year, have Eora data points that are not aligned with those of the OECD-WTO”
(page 19).
Comparison in UNCTAD (2013b), while quite limited in scope, is interesting
because it uses the WIOD instead of OECD TiVA. The UNCTAD report shows that
global average FVA shares estimated by UNCTAD-Eora and the WIOD are close,
both in values and trends, and the difference is narrowing over time (figure 9a).
Furthermore, the comparison of FVA shares at the country-level for 2009 reveals
a strong correlation between data reported by UNCTAD-Eora and by the WIOD,
close to 0.9, and a slope of the regression line at around 1 (figure 9b).

Figure 9. FVA shares of UNCTAD-Eora and WIOD (from UNCTAD, 2013b)

a. FVA share in exports, comparison between UNCTAD-Eora and WIOD

30%

25%

20%

Eora
WIOD
15%
1990 1995 2000 2005 2010

b. FVA share in exports by country, WIOD vs. UNCTAD-Eora, 2009

60%

50%
 20%

Eora
Improving the analysis of global value chains: the UNCTAD-Eora Database 137
WIOD
15%
1990 1995 2000 2005 2010

b. FVA share in exports by country, WIOD vs. UNCTAD-Eora, 2009

60%

50%
WIOD

40%

30%

20%

10%
10% 20% 30% 40% 50% 60%
Eora

5. C
 oncluding remarks: the importance of GVC data in the
analysis of globalization

The analysis of GVCs has long occupied a central place in the analysis of trade
and development. The concept gave development economists, in particular, an
essential tool to examine the role of countries in the global production system and
to identify opportunities for investment and growth in specific industries and value
chain segments.
GVC analysis received a significant boost when data on value added in trade
became available in the early part of this decade. The new data yielded many policy
insights. For example, it was helpful in explaining the link between economies’
openness to imports and export success; it showed the importance of services in
GVCs; and it shed light on relative levels of GVC participation of, and integration
between, countries and regions in the world.
The slowdown of trade growth relative to GDP growth after the global financial
crisis again showed the utility of the new data as they helped to explain the factors
behind the trend. At the time, GVC data could not provide all the answers, mainly
because of the significant time lag inherent in most datasets. With the UNCTAD-
Eora database now covering the full timespan since the financial crisis, the data
show that GVCs reached an inflection point at about 2010-2012. Since then, foreign
value added in exports has been stagnating after a lengthy period of continuous
growth that started in 1990 (see UNCTAD, 2018, p. 22).
138 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

GVC data confirmed some important intuitions right away. A key insight was that
GVCs have created an inextricable link between trade and investment. With the
exchange of goods and services within the international production networks of
MNEs comprising such a large part of global trade, it meant that the slowdown
in global FDI flows – which today are still well below their peak level in 2007 –
was a major factor behind the deceleration of global trade. The reverse is equally
true; the current suite of policies designed to slow cross-border trade will have
consequences for FDI. Trade and investment are two sides of the same coin – the
very coin that ultimately pays for development.
The importance of GVC data as a barometer of trends in international production
means the accuracy, universality and contemporaneity of the data are crucial. For
these reasons, the efforts to renew and improve the UNCTAD-Eora dataset, as
described in this paper, were undertaken.
The requests UNCTAD receives for GVC data are growing in number. This is in part
owing to the realization among researchers that the dataset is reliable while the
coverage has been expanded. The growing reliance on GVC data is also in large
part the result of the current turbulence in the global policy environment for trade
and investment. GVC analysis is critical to enable a serious assessment of the
consequences of trade wars, including the shifting of supply chains, the effects on
intra-firm trade and the potential relocation of production stages. It is also important
for understanding other major global policy trends, such as the increasing reliance
on regional economic cooperation, which is explained by the relatively greater
importance of regional, over global, value chains.
GVC analysis is also relevant for understanding the impact of technology
development on global trade and investment patterns. The digital economy and
the new industrial revolution will cause important shifts in value chain-related
sourcing patterns across geographies, industries and value chain segments. For
policymakers, especially those in the 100+ countries that are actively pursuing
industrial policies (cf. UNCTAD 2018), anticipating potential changes and identifying
future opportunities for economic growth and development will be paramount.
Improving the analysis of global value chains: the UNCTAD-Eora Database 139

References
AfDB, OECD and UNDP (2014). African Economic Outlook 2014: Global Value Chains and
Africa’s Industrialization. Paris: OECD Publishing.
Antràs, P., A. de Gortari (2017). On the Geography of Global Value Chains. National Bureau
for Economic Research, Working Paper Series No. 23456, doi:10.3386/w23456.
Aslam, A., N. Novta, and F. Rodrigues-Bastos (2017). Calculating Trade in Value Added, IMF
Working Paper No. 17/178. Available noline at https://www.imf.org/en/Publications/WP/
Issues/2017/07/31/Calculating-Trade-in-Value-Added-45114
Chen, Z.M., M. Lenzen, S. Ohshita, T. Wiedmann (2018). Consumption-based greenhouse
gas emissions accounting with capital stock change highlights dynamics of fast-
developing countries. Nature Communications 9, 3581.
Chen Q., K. Zhu, P. Liu, X. Chen, K. Tian , L. Yang, C. Yang (2018). Distinguishing China’s
processing trade in the world input-output table and quantifying its effects. Economic
Systems Research, 1–21, doi: 10.1080/09535314.2018.1534225.
De Backer, K. and S. Miroudot (2013). "Mapping Global Value Chains", OECD Trade Policy
Papers, No. 159, OECD Publishing, Paris, doi:10.1787/5k3v1trgnbr4-en.
De Koning A, M. Bruckner, S. Lutter, R. Wood, K. Stadler and A. Tukker (2015) Effect of
aggregation and disaggregation on embodied material use of products in input–output
analysis Ecological Economics 116 pp. 289–99, doi: 10.1016/j.ecolecon.2015.05.008.
Dietzenbacher, E., M. Lenzen, B. Los, D. Guan, M. Lahr, F. Sancho, S. Suh, C. Yang (2013).
Input-Output Analysis: The Next 25 Years. Economic Systems Research 25, pp. 369–
389.
Dietzenbacher, E., J. Pei, C. Yang (2012). Trade, Production, Fragmentation, and China’s
Carbon Dioxide Emissions. Journal of Environmental Economics and Management 64(1).
Doi: 10.1016/j.jeem.2011.12.003
Edens B., R. Hoekstra, D. Zult, O. Lemmers, H. Wilting and R. Wu (2015) A Method to Create
Carbon Footprint Estimates Consistent with National Accounts. Economic Systems
Research pp. 1–18, doi:10.1080/09535314.2015.1048428.
Geschke A., R. Wood, K. Kanemoto, M. Lenzen and D. Moran (2014) Investigating Alternative
Approaches to Harmonise Multi-Regional Input-Output Data. Economic Systems
Research 26 pp. 354–85, doi:10.1080/09535314.2014.937069.
Hummels, D., J. Ishii, K.-M. Yi (2001). The Nature and Growth of Vertical Specialization in
World Trade. Journal of International Economics 54, pp. 75–96.
IMF (International Monetary Fund) (2018). World Economic Outlook. IMF, Washington, DC.
Inomata S. and A. Owen (2014) Comparative Evaluation of MRIO Databases Economic
Systems Research 26 pp. 239–44, doi:10.1080/09535314.2014.940856.
Johnson R.C. and G. Noguera (2016). A Portrait of Trade in Value Added over Four Decades.
National Bureau for Economic Research, Working Paper Series No. 22974, doi:10.3386/
w22974.
140 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

Johnson R.C. and G. Noguera (2012). Fragmentation and Trade in Value Added over Four
Decades. National Bureau for Economic Research, Working Paper Series No. 18186,
doi:10.3386/w18186.
Kitzes, J. (2013). An Introduction to Environmentally-Extended Input-Output Analysis.
Resources 2 489–503, doi:10.3390/resources2040502.
Koopman, R., W. Powers, Z. Wang and S.-J. Wei (2010). Give credit where credit is due:
tracing value added in global production chains, National Bureau for Economic Research,
Working Paper Series, No. 16426, doi:10.3386/w16426.
Koopman, R., Z. Wang and S.-J. Wei (2012). Tracing Value-added and Double Counting
in Gross Exports. National Bureau for Economic Research, Working Paper Series No.
18579, doi:10.3386/w18579.
Koopman, R., Z. Wang, S.-J. Wei (2012). Tracing Value-added and Double Counting in Gross
Exports. National Bureau for Economic Research, Working Paper Series No. 18579,
doi:10.3386/w18579.
Koopman, R., Z. Wang, S.-J. Wei (2008). How Much of Chinese Exports is Really Made in
China? Assessing Domestic Value-Added When Processing Trade is Pervasive. National
Bureau for Economic Research, Working Paper Series No. 14109, doi:10.3386/w14109.
Lenzen, M., R. Wood, T. Wiedmann (2010). Uncertainty analysis for Multi-Region Input-
Output models - a case study of the UK’s carbon footprint. Economic Systems Research
22, pp. 43–63, doi: 10.1080/09535311003661226.
Lenzen, M. and J. M. Rueda-Cantuche (2012). A note on the use of supply-use tables in
impact analyses, Statistics and Operations Research Transactions, 36(2): pp. 139–152.
Lenzen, M., D. Moran, A. Geschke, K. Kanemoto (2014). A Non-Sign-Preserving RAS
Variant. Economic Systems Research 26, pp. 197–208.
Lenzen, M., K. Kanemoto, D. Moran and A. Geschke (2012). Mapping the Structure of the
World Economy, Environmental Science & Technology, 46(15): pp. 8374–8381.
Lenzen, M., D. Moran, K. Kanemoto, A. Geschke (2013). Building Eora: A Global Multi-
Regional Input-Output Database at High Country and Sector Resolution. Economic
Systems Research, 25:1, pp. 20–49, doi:10.1080/09535314.2013.769938.
Los, B., & M. Timmer (2018). Measuring Bilateral Exports of Value Added: A Unified
Framework. Cambridge MA: NBER.
Los, B., M.P. Timmer and G.J. de Vries (2016). Tracing Value-Added and Double Counting
in Gross Exports: Comment. American Economic Review, 106(7), pp. 1958–1966. doi:
10.1257/aer.20140883.
Moran, D., R. Wood, and J. F.D. Rodrigues. (2017). A Note on the Magnitude of the
Feedback Effect in Multi-Region Input-Output Tables Journal of Industrial Ecology 22 (3).
doi: 10.1111/jiec.12658.
Moran D. and R. Wood (2014). Convergence Between the Eora, WIOD, EXIOBASE, and
OpenEU’s Consumption-Based Carbon Accounts Economic Systems Research 26 245–
61 doi:10.1080/09535314.2014.935298.
Improving the analysis of global value chains: the UNCTAD-Eora Database 141

Moran, D., A. Geschke (2013). Tracing Embodied CO2 in Trade Using High-Resolution
Input-Output Tables, chapter in Computational Intelligent Data Analysis for Sustainable
Development, T. Yu, N. Chawla, S. Simoff, Eds. (Chapman and Hall/CRC, 2013).
Moran, D. (2013). The Eora MRIO chapter in The Sustainability Practitioner’s Guide to
Multi-Regional Input-Output Analysis, J. Murray, M. Lenzen, Eds. (Common Ground,
Champagne, Illinois, 2013), pp. 66–71.
Owen A, K. Steen-Olsen, J. Barrett, T. Wiedman and M. Lenzen (2014). A Structural
Decomposition Approach to Comparing Input-Output Databases Economic Systems
Research 26 pp. 262–83.
Owen, A., R. Wood, J. Barrett, A. Evans. (2016). Explaining value chain differences in MRIO
databases through structural path decomposition. Economic Systems Research 28, pp.
243–272.
Owen, A. (2017). Techniques for Evaluating the Differences in Multiregional Input-Output
Databases A Comparative Evaluation of CO2 Consumption-Based Accounts Calculated
Using Eora, GTAP and WIOD Cham, Switzerland, Springer International Publishing;
Rodrigues J. F. D., D. Moran, R. Wood and P. Behrens (2018). The uncertainty of consumption-
based carbon accounts. Environmental Science & Technology, doi:10.1021/acs.
est.8b00632.
Schaffartzik, A., M. Sachs, D. Wiedenhofer, N. Eisenmenger (2014). Environmentally Extended
Input-Output Analysis.. ISSN 1726-3816. Available online at https://worldmrio.com/pdf/
Schaffartzik_IntroToEEMRIO.pdf.
Steen-Olsen K, A. Owen, E.G. Hertwich and M. Lenzen (2014). Effects of Sector Aggregation
on CO2 Multipliers in Multiregional Input-Output Analyses. Economic Systems Research
26.
Steen-Olsen K., A. Owen, J. Barret, D. Guan, E. Hertwich, M. Lenzen, T. Wiedmann (2016).
Accounting for value added embodied in trade and consumption: an intercomparison of
global multiregional input–output databases. Economic Systems Research 28, doi:10.1
080/09535314.2016.1141751.
Stehrer, R., N. Foster and G. de Vries (2012). Value Added and Factors in Trade: A
Comprehensive Approach, wiiw Working paper 80, Vienna.
Södersten, C.J., R. Wood, E. G. Hertwich (2018). Endogenizing Capital in MRIO Models: The
Implications for Consumption-Based Accounting. Environmental Science & Technology,
doi:10.1021/acs.est.8b02791.
Timmer, M., B. Los, and G. de Vries, G. (2016). The Rise of Global Manufacturing Value
Chains: A New Perspective Based on the World Input-Output Database. In D. W.
Jorgenson, K. Fukao, & M. P. Timmer (Eds.), The World Economy: Growth or Stagnation?
(pp. 535-563). Cambridge, United Kingdom, Cambridge University Press.
Timmer, M., B. Los, R. Stehrer, G. de Vries (2012). Fragmentation, Incomes and Jobs. An
analysis of European competitiveness, WIOD Working Paper 9, Groningen.
Tukker A. and E. Dietzenbacher (2013). Global Multiregional Input-Output Frameworks: An
introduction and Outlook. Economic Systems Research 25(1) pp. 1–19.
142 TRANSNATIONAL CORPORATIONS Volume 26, 2019, Number 3

Tukker A., A. de Koning, A. Owen, S. Lutter, M. Bruckner, S. Giljum, K. Stadler, R. Wood, R.

Hoekstra (2018). Towards Robust, Authoritative Assessments of Environmental Impacts
Embodied in Trade: Current State and Recommendations. Journal of Industrial Ecology
22, doi:10.1111/jiec.12716.
UNCTAD (2013a). World Investment Report 2013: Global Value Chains: Investment and
Trade for Development. New York and Geneva: United Nations.
UNCTAD (2013b). Global Value Chain and Development. Working paper. https://unctad.org/
en/PublicationsLibrary/diae2013d1_en.pdf.
UNCTAD Secretariat (2015). UNCTAD-EORA Global Value Chain Database: methodology
and further research agenda. Transnational Corporations 21(3). https://unctad.org/en/
PublicationsLibrary/diaeia2014d1_en.pdf.
UNECA (2015). Economic Report on Africa 2015: Industrializing Through Trade. Addis Ababa:
United Nations Economic Commission for Africa.
UNIDO (2016). Industrialization in Africa and Least Developed Countries: Boosting Growth,
Creating Jobs, Promoting Inclusiveness and Sustainability. Vienna: United Nations
Industrial Development Organization.
Wang, Z., S.-J. Wei and K. Zhu (2013). Quantifying International Production Sharing at the
Bilateral and Sector Levels. National Bureau for Economic Research, Working Paper
Series No. 19677, doi:10.3386/w19677.
Wang, Z., S.-J. Wei, X. Yu and K. Zhu (2017a). Measures of Participation in Global Value
Chains and Global Business Cycles. National Bureau for Economic Research, Working
Paper Series No. 23222 (2017a), doi:10.3386/w23222.
Wang, Z., S.-J. Wei, X. Yu and K. Zhu (2017b). Characterizing Global Value Chains: Production
Length and Upstreamness. National Bureau for Economic Research, Working Paper
Series No. 23261, doi:10.3386/w23261.
Wiedmann, D., H. C. Wilting, M. Lenzen, S. Lutter and V. Palm (2011). Quo Vadis MRIO?
Methodological, data and institutional requirements for multi-region input-output analysis.
Ecological Economics 70, pp. 1937–1945.
Wilting, H. C. (2012). Sensitivity and uncertainty analysis in MRIO modelling; Some empirical
results with regard to the Dutch Carbon footprint. Economic Systems Research 24, pp.
141–171.
Wood R., T. Hawkins, E. Hertwich and A. Tukker (2014). Harmonizing National Input-Output
Tables for Consumption Based Accounting – Experiences in EXIOPOL. Economic
Systems Research 26.
Xing, Y. (2014). Measuring Value Added in the People’s Republic of China’s Exports: A Direct
Approach. ADBI Working Paper 493. Tokyo: Asian Development Bank Institute. Available
at: http://www.adbi.org/working-paper/2014/08/06/6372.measuring.value.added.prc.
exports/.