Journal Pre-proof

Assessing sustainability performance of global supply chains: An

input-output modeling approach

H. Wang , Chen Pan , Qunwei Wang , P. Zhou

PII: S0377-2217(20)30094-1
DOI: https://doi.org/10.1016/j.ejor.2020.01.057
Reference: EOR 16311

To appear in: European Journal of Operational Research

Received date: 27 November 2018

Accepted date: 28 January 2020

Please cite this article as: H. Wang , Chen Pan , Qunwei Wang , P. Zhou , Assessing sustainability
performance of global supply chains: An input-output modeling approach, European Journal of Oper-
ational Research (2020), doi: https://doi.org/10.1016/j.ejor.2020.01.057

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2020 Published by Elsevier B.V.

Highlights

 An input-output modeling approach to measuring global supply chain performance

 Both transnational and multidimensional features of supply chains are considered

 Static and dynamic sustainability performance of supply chains can be evaluated

 16 economies’ supply chain environmental performance rose by 20.6% in 2005-2014

1
Assessing sustainability performance of global supply chains: An
input-output modeling approach

H. Wanga,b, Chen Panc, Qunwei Wangd, P. Zhoua,b,d,*

a
School of Economics and Management, China University of Petroleum, Qingdao 266580, China
b
Institute for Energy Economics and Policy, China University of Petroleum, Qingdao 266580, China
c
School of Public Policy & Management, Tsinghua University, Beijing 100084, China
d
College of Economics and Management, Nanjing University of Aeronautics and Astronautics, 29 Jiangjun
Avenue, Nanjing 211106, China

Abstract: Measuring the sustainability performance of supply chains is fundamental to sustainable

supply chain management. Sustainability performance is usually evaluated from multiple aspects

within the triple bottom line framework. With globalization, supply chains have also been

characterized by the complex and global natures. Ignoring the multidimensional and transnational

features imposes challenges on the performance assessment of global supply chains (GSCs). To

resolve this issue, we propose an input-output modeling approach based on the multi-region

input-output (MRIO) model and the data envelopment analysis (DEA) technique, which is able to

account for the multidimensional characteristic of supply chains in a global context. Two indices are

introduced to measure the status and evolvement of environmental sustainability performance of

GSCs. We apply the proposed approach to empirically examine the environmental performance of

GSCs of the manufacturing sectors in 16 major economies during 2005-2014. The average

environmental inefficiency of the economies was considerable, and roughly 40% of the pollution

could potentially be reduced along GSCs. Overall the environmental performance of GSCs averagely

rose by 20.6% during the study period with fluctuations and regional/sectoral heterogeneities

observed.

Keywords: OR in environment and climate change; Global supply chains; Sustainability performance;
Multi-region input-output model; Data envelopment analysis

2
*
Corresponding author. Email: wh@upc.edu.cn (H. Wang), pzhou@upc.edu.cn (P. Zhou).

1. Introduction

Supply chains have become the generic organizational form of global production networks. With

global climate change, both producers and consumers have an increasing awareness of sustainability,

which urges production activities along supply chains to become greener and has induced the

transition of conventional supply chains towards sustainable ones (O’Rourke, 2014). Focusing on the

economic, environmental, and social impacts of products throughout their life cycle, sustainable

supply chain management (SSCM) aims to integrate the issue of sustainability into the management

of entire supply chains (Gupta & Palsule-Desai, 2011). Business practices can be improved via SSCM

to promote the sustainability of supply chains (Min & Kim, 2012). For example, Zhu, Sarkis, and Lai

(2007) examine the impacts of firms’ initiatives on supply chain performance, Dou, Zhu, and Sarkis

(2014) explore the inter-relationships between green supply chain management and supplier

performance, and Ding, Liu, and Zheng (2016) investigate strategies for balancing the economic and

environmental performances of firms while reducing environmental externalities. SSCM has also

attracted attention from policymakers and practitioners worldwide. An example is the United Nations

and BSR (2010) that propose practical guidelines for firms to continuously improve the sustainability

of supply chains.

A pillar of SSCM is the measurement of sustainability performance of supply chains and the

quantification of their determinants as well as consequences, according to which follow-up actions

could be taken to promote the performance. An important feature of sustainability performance

assessment is that multiple dimensions of supply chains and their impacts are involved (Genovese,

Morris, Piccolo, & Koh, 2017). Specifically, firms use multiple inputs that are sourced from multiple

suppliers along supply chains to produce goods and services that are distributed to multiple consumers.

In the meantime the economic activities along a supply chain generate a variety of social impacts and

environmental externalities, e.g. solid wastes and air emissions. Thus sustainability of supply chains is

commonly defined from these three facets, i.e. economic, environmental and social, which are also

known as the triple bottom line of supply chain performance (Elkington, 1997).

3
 Supply chains are increasingly characterized by complex and global natures. Goods and services

produced by a firm/sector/country flow to other firms/sectors/countries along a particular supply chain,

and all the individual supply chains intertwine together to form a complex network. With growing

vertical specialization, supply chains have become longer and more complex (Wang, Wei, Yu, & Zhu,

2017b). Meanwhile, supply chains have expanded transnationally with the emergence of

multinationals (Cohen & Mallik, 1997). To achieve economic efficiency, outsourcing and offshoring

have become the norm in production worldwide (Kim, Park, Jung, & Park, 2018). Cross-border

supply chains have given rise to production and consumption activities that often occur in different

countries and are linked via international trade. Cross-border economic activities have become an

important component of global production. For example, the ratio of international trade to global GDP

reached roughly 30% in 2011 (WTO, 2015), and this trend is projected to continue (Wang, Wei, Yu,

& Zhu, 2017c).

The preceding underlines the multidimensional and transnational features of supply chains

nowadays. Ignoring these salient features may lead to biased assessment of global supply chains

(GSCs) (Qorri, Mujkić, & Kraslawski, 2018). This study adds to the literature by proposing an

input-output modeling approach to account for the multidimensional and transnational features in

measuring sustainability performance of GSCs. We concentrate on the environmental aspect of

SSCM1 at the sectoral level.2 Specifically, the proposed approach is built upon the multi-region

input-output (MRIO) model and the data envelopment analysis (DEA) technique. By fully accounting

for the economic linkages among global economies, MRIO tables are able to holistically reflect GSC

dependences and interactions from an industrial lifecycle viewpoint (Acquaye et al., 2018). We first

adopt the MRIO model to measure the resources used, goods produced, and environmental impacts

related to an entire supply chain, which thereby captures the complex and transnational natures of

GSCs. With the MRIO estimates, the production processes of GSCs are modelled within the

1
The methodology introduced in the sections that follow can be adapted to evaluate the social performance of GSCs given that
relevant data are available.
2
Compared to the firm or product level, economic activities that add the most value usually take place at the sectoral level
(Gereffi, Humphrey, & Sturgeon, 2005). This study, therefore, focuses on the GSC performance at the sectoral level. As
discussed later in Section 3.4, the present sectoral analysis can be adapted to the firm, product or even process level by
adopting other background approaches.

4
total-factor production theory framework using DEA, which facilitates to capture the

multidimensional impacts of GSCs. Both the static and dynamic environmental sustainability

performances of GSCs, as well as their determinants, are studied. We apply the proposed approach to

empirically examine the environmental performances of GSCs in the manufacturing sectors of 16

major economies in 2005-2014. It is shown that the average environmental inefficiency of the 16

economies was considerable, and roughly 40% of the pollution could potentially be reduced along

GSCs. Overall the environmental performance of GSCs averagely rose by 20.6% during the study

period, which was driven primarily by the production technological improvement. The results are

expected to help policymakers monitor the progress of GSCs of sectors/economies towards

environmental sustainability.

The remainder of this paper is organized as follows. Section 2 reviews relevant literature. Section

3 introduces the methodology. Section 4 describes the data used in the empirical study. Section 5

presents the results with discussions. Section 6 concludes.

2. Literature review

A large number of studies have been devoted to the sustainability performance assessment of

supply chains (Barbosa-Póvoa, da Silva, & Carvalho, 2018). As the preceding discussed,

sustainability performance is usually assessed from the viewpoint of triple bottom line. As a result,

these three dimensions need to be integrated into a unified framework for performance assessment.

Besides, various factors surrounding supply chains and their management also affect the performance

assessment. For example, Qorri et al. (2018) conceptually specify the elements involved in the

performance measurement system and supply chain strategies, and Koberg and Longoni (2019)

qualitatively examine the relationship among configurations, governance mechanisms and

sustainability performance of supply chains.

Due to the multidimensional nature of supply chain performance assessment, multi-criteria

decision making (MCDM) and multi-objective mathematical programming approaches have been

most often chosen (Brandenburg, Govindan, Sarkis, & Seuring, 2014). Of the commonly used tools,

DEA has gained popularity in the past decades (Zhou, Yang, Chen, & Zhu, 2018). Established within

5
a multi-input multi-output framework, DEA is able to construct composite indicators that can

simultaneously and holistically reflect multiple impacts of supply chains. Another desirable property

of DEA is that it does not require advance specification of functional form due to its non-parametric

nature, which helps to avoid misspecification in modelling. In addition, DEA can further identify

peers with the best practice which are located on the production frontier for benchmarking purposes.

Given these strengthens, DEA has been increasingly applied to evaluate supply chain performance

(Kao, 2014). For example, with the network DEA technique that captures the operation structure of

production systems, Kao (2009, 2017) measures static and dynamic performances of a parallel

production system, Chen and Yan (2011) examine supply chain performance under different

organization mechanisms, and Mirhedayatian, Azadi, and Farzipoor Saen (2014) evaluate green

supply chain management considering dual-role factors and uncertainties.

With the emergence of globalization and multinationals, studying supply chain performance in a

global context becomes relevant. Cohen and Mallik (1997) discuss the management of supply chain

globalization, Fransoo and Lee (2013) and Pourakbar and Zuidwijk (2018) respectively examine the

roles of ocean container transport and customs in GSC performance, and Wang and Huang (2018)

investigate operational and financial strategies to improve performance of a capital-constrained GSC.

However, few studies have been reported to quantitatively measure sustainability performance of

GSCs (Koberg & Longoni, 2019). A main difficulty is the complex structure and transnational nature

of GSC activities. A common practice for resolving this issue is to adopt the life cycle assessment

(LCA) approaches (Boons, Baumann, & Hall, 2012). On the basis of MRIO tables that detail supply

chain linkages at the sectoral level, Costantini, Crespi, Marin, and Paglialunga (2017) examine the

impacts of international spillovers on the supply chain performances of EU countries, Acquaye et al.

(2017) and Acquaye et al. (2018) estimate the footprints and consumption-based intensities of

pollutants in order to measure the environmental performances of GSCs. By fully accounting for the

linkages among firms/sectors/economies, these studies are able to measure both the direct and indirect

environmental impacts induced by activities of an entire supply chain.

Despite the achievements aforementioned, some research gaps exist in the sustainability

performance assessment of GSCs. First, single factor indicators (e.g. consumption-based CO2

6
emissions and emission intensity) have been the norm in measuring sustainability performance of

GSCs, which fail to account for the multidimensional nature of GSCs and thus may bias the

performance assessment. Second, quantifying the determinants of GSC performance is largely ignored

in the present literature. A good understanding of the sources of changes in performance helps

stakeholders to identify priorities in GSC management and policymaking. Third, benchmarking GSCs

has rarely been studied in the literature, which can further reveal directions regarding how to improve

GSC performance. It is the purpose of this study to fill the gaps mentioned above. As the preceding

discussed, DEA is useful in modelling the multi-input multi-output supply chain framework and can

well serve the benchmarking purpose. However, conventional DEA (e.g. the network DEA models)

becomes impractical in the GSC context since numerous production stages and complex structure are

involved. To account for both the multidimensional and transnational features of GSCs, we propose

an input-output modelling approach on the basis of MRIO and DEA in the sections that follow. To

illustrate the usefulness of the proposed approach, we further apply it to evaluate the environmental

sustainability performance of GSCs.

3. Methodology

3.1 MRIO model

Various materials and resources are used by firms to produce goods and services along GSCs

while generating environmental pollution as the by-product. To characterize the general production

processes in supply chains, we follow the conventional production theory and consider capital, labor,

and energy to be inputs. Value added is used to measure the economic output.3 Production activities

affect the environment by producing a variety of pollutants, e.g. solid wastes, wastewater, and air

pollutant emissions. Here, we use the emissions of CO2, SO2, and NOx to represent the undesirable

outputs.4

3
Another commonly used indicator to measure economic activity is total output. Here we select value added since it is able
to overcome the well-known “double counting” problem in trade. The methodology introduced in the sections that follow
can be similarly applied if the total output indicator is chosen.
4
CO2 emissions are the primary cause of global warming, and SO2 and NOx are key air pollutants. These three pollutants have
been selected to indicate the environmental impacts of economic activities in a number of studies, e.g. Pasurka (2006) and Du,
Chen, and Huang (2018).

7
 As an example, we apply the MRIO model to estimate the flow of energy input that is embodied

in goods and services transferred among sectors and economies via GSCs. Suppose the global

economy is divided into Q economies, each of which is further disaggregated into M economic sectors.

The total output of a sector in an economy is consumed as either intermediate inputs or final goods by

all sectors in the Q economies. The equilibrium relationship is illustrated by Fig.1 and can be

formulated as:

X = Z1 + Y (1)

where X and Y are QM 1 vectors denoting the total output and final demand, respectively.5 Z is a

QM  QM matrix with entries Z ji denoting the intermediate inputs of sector i in region r supplied
sr

by sector j in region s, and 1 is a summation vector with proper length. Define a direct intermediate

input coefficient matrix A with entries Aji  Z ji X i

sr sr r
such that Z1 = AX . Following the

conventional Leontief I-O model, Eq. (1) can be rewritten as:

X = (I - A)-1 Y =  I + A1 + A2 + ... + A  Y = LY (2)

where I is an identity matrix and L = (I - A)-1 = I + A1 + A2 + ... + A is the Leontief inverse matrix.

The term I + A 1
+ A2 + ... + A  Y reflects the iterative production processes within the global

production network. For example, IY denotes the direct production of final demand, A1 Y denotes

the first tier intermediate inputs that are required to produce the final demand, and A2 Y denotes the

intermediates that are needed to produce the intermediate goods A1 Y , which therefore is the second

tier intermediate inputs required to produce the final demand. With the iterations IY , A1 Y , A2 Y

and so forth, all production activities along the GSCs are completely captured by Eq. (2).

Table 1. Nomenclature

Notation Description Notation Description

Q Number of economies E Energy input
M Number of sectors K Capital input

5
It should be noted that here we adopt the conventional notations used in I-O studies, which differ from the convention in the
DEA literature. All the notations appeared in this article are summarized in Table 1.

8
 r, s Index of economies H Labor input
i, j Index of sectors V Value added
X Total output C CO2 emissions
Y Final demand S SO2 emissions
Z Intermediate inputs N NOx emissions
A Direct input coefficient  Intensity variable
I Identity matrix SEPI Static environmental performance index
L Leontief inverse matrix DEPI Dynamic environmental performance index
f Energy intensity

Outputs
Intermediate transactions
Final Total
demand outputs
R1 RQ
Inputs

R1 Z1,1 Z1,Q Y1 X1

Intermediate inputs

RQ ZQ ,1 ZQ ,Q YQ XQ

Value added V1 VQ

Total inputs X1 XQ

Note: the letter R denotes regions.

Figure 1. Structure of MRIO table

Combined with sectoral energy intensity, Eq. (2) can be extended to measure the energy flows

ss
among sectors and regions via the supply chains. Define a diagonal matrix, f , with entries f jj

denoting the ratio of energy consumption to total output for sector j in region s. The energy inputs of

ˆ ˆ , where
the sectors and regions along the GSCs can be modeled as E = fLY indicates a diagonal

matrix. The term

 E11 E12 ... E1N 

 21 
E E22 ... E2N 
E=
 ... ... ... ... 
 N1 
E EN2 ... ENN 

9
is a block matrix where the block E sr is an M  M matrix with entries E ji denoting the energy
sr

input embodied in the intermediates that are needed by sector i in region r and supplied by sector j in

region s. Summing the matrix along the columns yields the total energy input required by a sector in

an economy through the GSC network. For example, the overall energy requirement of sector i in

region r to fulfill its production is Eir   E srji . Compared to the conventionally used
s, j

production-based accounting measures that quantify the energy input consumed in the territory of a

particular sector/economy (e.g. E

s, j
rs
ij for sector i in region r), Eir accounts for all energy inputs

that are necessary to the production of sector i in region r, regardless of where the production occurs

in the GSC. Similarly, other inputs as well as the desirable and undesirable outputs of sectors and

economies along the GSCs can be calculated. Taking sector i in region r as an example, its production

activity needs capital input K ir , labor input H ir , and energy input Eir which are supplied via the

global production network. Meanwhile, the associated economic output along the supply chain of this

sector is Vi r and the environmental impacts are CO2 emissions Cir , SO2 emissions Sir , and NOx

emissions N ir . As such, the production process along the supply chain of sector i in region r is

characterized from an industrial lifecycle perspective in the global context.

3.2 Environmental production technology and distance functions

To account for sectoral heterogeneity in environmental performance (Wang, Ang, Wang, & Zhou,

2017a; Wang, Ang, & Zhou, 2018), we model the production processes of sectors separately. As

production units along supply chains use inputs (i.e. capital, labor, and energy) to produce desirable

outputs (value added) and undesirable outputs (i.e. CO2, SO2, and NOx), the production possibility set

of the supply chains in sector i can be formulated as:

Ti   K i , Hi , Ei , Vi , Ci , Si , Ni  :  K i , Hi , Ei  can produce  Vi , Ci , Si , Ni  (3)

Production processes in reality often exhibit two features. First, undesirable outputs will not be

eliminated unless the production process completely stops. Second, the abatement of undesirable

outputs is not free. To reflect these two characteristics, we follow Färe, Grosskopf, Lovell, and

10
Pasurka (1989) and Färe, Grosskopf, and Hernandez-Sancho (2004) to impose two assumptions on Eq.

(3).

i. Null-jointness assumption: if  K i , Hi , Ei , Vi , Ci , Si , Ni  Ti and  Ci , Si , Ni   0 , then Vi  0 .

ii. Weakly disposable assumption: if  Ki , Hi , Ei , Vi , Ci , Si , Ni  Ti and 0    1 , then

 Ki , Hi , Ei , Vi , Ci , Si , Ni  Ti .

With the two assumptions, Eq. (3) is then able to represent the production technology of sector i

taking into account the impacts of pollution. Hence, this technology is referred to as environmental

production technology. To operationalize the conceptual model above, we adopt the non-parametric

frontier approach. The DEA model exhibiting the constant returns to scale can be formulated as6:

Ti   K i , H i , Ei , Vi , Ci , S i , N i  :  r K ir  K i
r

 H
r
r i
r
 Hi

 E
r
r i
r
 Ei

 V r i
r
 Vi
r (4)
 C
r
r i
r
 Ci

 S
r
r i
r
 Si

 N
r
r i
r
 Ni

r  0, r  1,..., Q

where  is the intensity variable. By incorporating the observations for the Q economies, Eq. (4)

constructs a best practice frontier for sector i. The economic and environmental performances of the

supply chains of sector i in a particular economy can be assessed by measuring the distance of an

observation from the efficient frontier. In the literature, a variety of efficiency measures have been

proposed and applied for this purpose. Some commonly used ones include the Russell measure,

6
A number of assumptions on the returns to scale, e.g. non-increasing returns to scale (NIRS) and variant returns to scale (VRS)
have been introduced and adopted in the DEA literature. Of the common assumptions, CRS has been the most widely used. In
the present paper, we follow this common practice. The DEA model specified in Eq. (2) can be modified if other assumptions
on the returns to scale are needed. Interested readers may refer to Zhou, Ang, and Poh (2008) for such extensions.

11
slacks-based measure (Chen & Delmas, 2012; Tone, 2001), directional distance function (Färe &

Grosskopf, 2000) and non-radial directional distance function (Zhou, Ang, & Wang, 2012). Here, we

adopt the non-radial directional distance function because of its additive nature, which can

exhaustively explore slacks in inputs and outputs (Chen, 2013, 2014). In accordance with the purpose

of this study and following Zhou et al. (2012), an undesirable output-oriented non-radial directional

distance function is defined as follows:

Dio  K io ,H io ,Eio ,Vi o ,Cio ,Sio ,N io   max  io   io   io  3

 r K ir  K io
 r
  Hr  Ho
 r i i

 r
 r Eir  Eio
 r
  V r V o
 r i i (5)
s.t.  r
 r Cir  Cio  io Cio
 r

 r Si  Si   i Si
r o o o

 r
 r N ir  N io   io N io
 r
  0, r  1,..., Q
 r

where o denotes the region under evaluation. The three terms, io Cio ,  io Sio , and  io Nio , respectively

denote the slacks in the three pollutants, i.e. CO2, SO2, and NOx, for sector i in region o. As there is no

preference for these three pollutants in the present study, equal weights are specified for their slacks

coefficients in the objective function. By maximizing the slacks coefficients, Eq. (5) captures the

technical inefficiency in the three undesirable outputs as compared to the best practice frontier. A


higher value of Dio Kio ,H io ,Eio ,Vi o ,Cio ,Sio ,Nio  implies that more slacks in the undesirable outputs

are identified, and hence the supply chain of sector i in region o operates with less environmental

efficiency.

3.3 Environmental performance indices

We study both the status and evolvement of GSC environmental performances. The former

implies the static performance in a particular year, whereas the latter refers to the change in

12
performance over a time period. Accordingly, we introduce two indices to quantify the static and

dynamic environmental performances of supply chains.

Based on Eq. (5), the static environmental performance of sector i in region o in year t can be

measured as:

SEPIiot  1  Diot  Kio,t ,Hio,t ,Eio,t ,Vi o,t ,Cio,t ,Sio,t ,Nio,t  (6)

where the superscript t denotes the time period. The term SEPI iot falls in the interval (0,1] . A higher

value of SEPI iot indicates better environmental performance, since fewer slacks exist in the

undesirable outputs. The term SEPI iot can therefore be viewed as an environmental performance

index from an efficiency viewpoint. Following Wang et al. (2017a) and using the sectoral index in Eq.

(6), an aggregate supply chain environmental performance index for region o in year t can be

formulated as follows:

1  Cio,t o Sio,t o Nio,t o 

SEPI ot  1     i  S o ,t  i  N o ,t  i 
3 i  C o ,t
(7)


The aggregate index in Eq. (7) is a weighted average of the sectoral indices in Eq. (6), with the

sectoral share of pollutant emissions being the weight. The value of SEPI ot is also between 0 and

unity, and the interpretation of the aggregate index is similar to that at the sectoral level.

Further to the static environmental performance in a particular year, the evolvement in a supply

chain’s environmental performance during a time period can be measured as the ratio of the

performance in the final year to that in the initial year. Taking sector i in region o as an example, the

change in the environmental performance of its supply chain can be quantified as:

1  Diot  Kio,t 1 ,H io,t 1 ,Eio,t 1 ,Vi o,t 1 ,Cio,t 1 ,Sio,t 1 ,N io,t 1 

1  Diot  Kio,t ,H io,t ,Eio,t ,Vi o,t ,Cio,t ,Sio ,t ,N io ,t 

where the time period outside the brackets in the distance functions is for the reference production

technology, and that inside the brackets is for the observations. The measure above compares all the

observations with respect to the best practice frontier in year t. Alternatively, the efficient frontier in

13
year t+1 can also be used as the benchmark, which may yield different results. To avoid arbitrary

selection between the two benchmarks, we take a geometric mean of the two measures, thus the

dynamic environmental performance index for the supply chain of sector i in region o between years t

and t+1 eventually becomes:7

1  Diot  K io ,t 1 ,H io ,t 1 ,Eio ,t 1 ,Vi o ,t 1 ,Cio ,t 1 ,Sio,t 1 ,N io,t 1  

1/ 2

 
 1  Diot  K io ,t ,H io ,t ,Eio ,t ,Vi o ,t ,Cio ,t ,Sio ,t ,N io ,t  
t ,t 1
DEPI io   
1  Diot 1  K io ,t 1 ,H io ,t 1 ,Eio ,t 1 ,Vi o ,t 1 ,Cio,t 1 ,Sio,t 1 ,N io,t 1  
 
 1  Diot 1  K io ,t ,H io ,t ,Eio ,t ,Vi o ,t ,Cio ,t ,Sio ,t ,N io ,t  
1  Diot 1  K io ,t 1 ,H io ,t 1 ,Eio ,t 1 ,Vi o ,t 1 ,Cio,t 1 ,Sio,t 1 ,N io,t 1 
 
1  Diot  K io ,t ,H io ,t ,Eio ,t ,Vi o ,t ,Cio ,t ,Sio ,t ,N io ,t  (8)

 1  Diot  K io ,t 1 ,H io ,t 1 ,Eio,t 1 ,Vi o,t 1 ,Cio ,t 1 ,Sio ,t 1 ,N io ,t 1  

1/ 2

 
1  Dio  K i ,H i ,Ei ,Vi ,Ci ,Si ,N i  
t 1 o ,t 1 o ,t 1 o , t 1 o , t 1 o , t 1 o , t 1 o , t 1

 
 1  Diot  K io ,t ,H io ,t ,Eio ,t ,Vi o ,t ,Cio ,t ,Sio ,t ,N io ,t  
 
1  Dio  K i ,H i ,Ei ,Vi ,Ci ,Si ,N i 
t 1 o ,t o ,t o ,t o ,t o ,t o ,t o ,t

 ECiot ,t 1TCiot ,t 1

where EC denotes the efficiency change, and TC denotes the technological change. The term

DEPI iot ,t 1 has a similar form with the Malmquist-Luenberger index. A value of DEPI iot ,t 1 larger than

unity implies that the environmental performance of the supply chain of sector i in region o has

improved, and vice versa. To quantify the determinants of changes in environmental performance,

DEPI iot ,t 1 can be decomposed into two components, i.e. efficiency change (EC) and technological

change (TC), as shown in Eq. (8). The former captures the change in technical efficiency of pollutant

emissions, and the latter quantifies the improvement/deterioration in general production technology of

the supply chain. For example, a value of ECiot ,t 1 larger than unity implies an improvement in the

7
Mixed-period linear programming problems, e.g. Dio K i
t
 o , t 1 o , t 1
, Hi
o , t 1
, Ei ,Vi
o , t 1 o , t 1
,Ci
o , t 1
, Si
o , t 1
, Ni  , exist in Eq. (8), which
may not have feasible solutions. A number of approaches have been proposed to deal with this issue. Some commonly used
methods include treating the observation with infeasible solution as on the efficient frontier (Wang, Zhou, & Zhou, 2013),
using the window analysis (Pulina, Detotto, & Paba, 2010), and modifying the constraints on inputs and undesirable outputs
(Du et al., 2018). In the empirical study that follows we adopt the first approach, i.e. locating the DMUs with an infeasible
solution on the frontier. However, this practice may introduce additional uncertainty to the results, which should be borne in
mind when interpreting the empirical results.

14
emission efficiency during the time period, while a value of TCiot ,t 1 larger than unity implies a shift

of the supply chain’s production technology toward the best practice frontier.

Similar to the static performance index, the dynamic performance index at the sectoral level

given in Eq. (8) can also be aggregated to form an economy-wide index. Following Wang et al.

(2017a), the change in aggregate environmental performance for economy o between year t and year

t+1 can be measured as:

1/ 2
 1  Cio ,t 1 o ,t 1 Sio ,t 1 o ,t 1 N io ,t 1 o ,t 1  
 1    o ,t 1  i  o ,t 1  i  o ,t 1  i  
 3 i C S N  
 1  C o ,t o ,t 1( t ) S o ,t o ,t 1( t ) N o ,t o ,t 1( t )  
1    io ,t i  io ,t  i  io ,t  i  
t ,t 1  3 i C S N  
DEPI o   
1  1  Ci o ,t ( t 1)  
o ,t 1 o ,t 1 o ,t 1
Si Ni
 3  o ,t 1  i
o ,t ( t 1)
 o ,t 1  i o ,t ( t 1)
 o ,t 1  i 
i C S N 
 
 1  C o ,t
S o ,t
N o ,t
 
1    o ,t  i  o ,t  i  o ,t  i 
i o ,t i o ,t i o ,t
 3 i  C S N  
 
1  Cio ,t 1 o ,t 1 Sio ,t 1 o ,t 1 N io ,t 1 o ,t 1 
1 
3 i  C o ,t 1
i  o,t 1  i  o,t 1  i 

S N  
1  C o ,t S o ,t N o ,t 
1    io ,t  io ,t  io ,t  io ,t  io ,t  io ,t 
3 i C S N 
1/ 2
 1  C o ,t S o ,t N o ,t  
 1    io ,t  io ,t  io ,t  io ,t  io ,t  io ,t  
 3 i C S N   
 1  C o ,t o ,t 1( t ) S o ,t o ,t 1( t ) N o ,t o ,t 1( t )  
1    io ,t  i  io ,t  i  io ,t  i  
 3 i C S N  
 
1  1  Ci o ,t ( t 1)  
o ,t 1 o ,t 1 o , t 1
Si Ni
 3  o ,t 1  i
o ,t ( t 1)
 o ,t 1  i o ,t ( t 1)
 o,t 1  i 
i C S N 
  (9)
 1  o ,t 1 o ,t 1 o ,t 1

i  C o,t 1 i  S o,t 1  i  N o,t 1  i  
1 C i o ,t 1 S i o ,t 1 N i o ,t 1
 3   

 ECot ,t 1TCot ,t 1

where the superscript t(t+1) denotes quantifying slacks in the observations in year t+1 by using the

production frontier for year t. The superscript t+1(t) is similarly defined. The interpretation of the

aggregate environmental performance index is similar to that at the sectoral level. Also, the overall

performance index can be decomposed into efficiency change and technological change, as shown in

Eq. (9).

15
3.4 Comparisons and discussions

Figure 2 illustrates the framework of the proposed input-output modeling approach. As shown in

the preceding sections, the proposed approach consists of two techniques, i.e. the MRIO model and

DEA, both of which take an input-output perspective. The MRIO model is first adopted to measure

the flow of inputs and outputs, as well as environmental externalities, along GSCs. On the basis of the

accounting of production factors induced by an entire GSC, DEA is further used to model supply

chains’ production activities of transforming inputs into outputs. Two indices, i.e. SEPI and DEPI, are

then introduced to capture the static and dynamic environmental performances of GSCs. Essentially

the input-output modeling approach can be viewed as a combination of the MRIO model and the DEA

technique.

Performance
Initial data MRIO DEA measurements

Estimate inputs & outputs Model the transformation Static: SEPI

Input & output of an entire global supply of inputs into outputs of
data for local
chain global supply chains Dynamic: DEPI
production
r
Vi
K ir H ir Eir K ir Efficiency
r
Cir change
H i DEPI
MRIO tables Sir
r r r r r
Vi C i S i N i E i
Technological
N ir change

Figure 2. Framework of the input-output modeling approach

Compared to existing methodologies, the proposed approach possesses some advantages. First,

the input-output modeling approach is able to measure the multidimensional performance of supply

chains in a multi-input multi-output production framework while accounting for the complex and

transnational natures of GSCs. As a result, the proposed approach extends DEA to performance

assessment in the GSC context and improves the conventional single-factor GSC performance

measurements by adopting a multi-factor framework. Second, built upon the DEA technique, the

input-output modeling approach benchmarks GSCs from an efficiency perspective. This allows the

optimal levels of resource inputs, goods outputs and environmental/social externalities of inefficient

GSCs to be determined, which helps to indicate possible directions where management could be

strengthened. Third, similar to the productivity indices in the literature, the dynamic performance

16
index (DEPI) introduced can be decomposed into two components, which respectively quantify the

catch-up effect in terms of technical efficiency change and the technological innovation of GSCs with

respect to the best practice frontier.

The proposed approach can be adapted or extended in a number of ways to examine supply chain

performance in greater depth. For example, by adopting LCA as the background approach, the present

methodology can be similarly applied to supply chain performance assessment at the product and

plant levels. Another example is to slice GSCs further into individual production paths using the

structural path analysis technique (Wood & Lenzen, 2009) and examine the environmental

performances of key production paths separately. Production stages in a particular path that lead to

environmental inefficiency can then be identified.

4. Data

We apply the proposed methodology to evaluate the environmental performances of GSCs in

manufacturing sectors during the decade 2005-2014. This period is relevant in assessing GSCs’

environmental performances. Manufacturing sectors worldwide flourished with the significant growth

of global economies since 2000. Nevertheless, the 2008 financial crisis largely affected global

demand and international fragmentation of production. Global economy and manufacturing sectors

recovered slowly after 2009 with structural transitions in GSCs (McKinsey Global Institute, 2019).

Hence, with the latest available data up to 2014, the empirical study is expected to provide timely

results on the recent development of GSCs in a changing global economy.

We collect global MRIO tables from the World Input-Output Database (WIOD)8, which covers

43 economies9 and a Rest of World (RoW) region representing the remaining economies (Timmer,

Dietzenbacher, Los, Stehrer, & de Vries, 2015). A single economy is further disaggregated into 56

sectors, 18 of which are manufacturing industries and can be classified into three categories, i.e.

labor-intensive, capital-intensive, and technology-intensive sectors. The list and classification of

8
The WIOD data used in this study was released in 2016, which can be retrieved from http://www.wiod.org/release16.
9
The 43 economies covered the majority of global production and pollutant emissions in the study period. For example, the 43
economies accounted for 86% of global GDP and 82% of global CO2 emissions in 2014.

17
economies and sectors are given in Appendix A. The annual data for capital stock, labor input and

value added at the same disaggregation levels (i.e. regions and sectors) are provided in the

Socio-Economic Accounts of WIOD. The data for energy use and pollutant emissions are collected

from the Eora database (Lenzen, Kanemoto, Moran, & Geschke, 2012).10 As the regional and sectoral

disaggregation levels in Eora and WIOD are different, however, we map the Eora data to the regional

and sectoral levels of WIOD. Details on the data treatment are provided in Appendix B. We deflate all

the monetary data in our dataset into constant 2010 USD using the price index and exchange rates

published by WIOD. The summary statistics of the data are given in Table 2.

Table 2. Descriptive statistics of data

Capital Labor Energy Value added CO2 SO2 NOx

(Billion USD) (Million) (Mtoe) (Billion USD) (Giga Kg) (Giga g) (Giga g)
Min 0.1 0.0 0.0 0.1 0.0 1.0 0.8
Max 1820.4 67.1 133.7 733.1 413.2 2397.7 2092.9
2005 Mean 77.6 1.8 10.0 33.5 26.0 143.5 112.7
Median 16.2 0.2 2.6 6.9 6.0 40.3 28.1
Std.dev 180.3 6.6 20.3 77.2 57.4 308.6 242.6
Min 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Max 1838.4 73.3 167.3 698.5 591.2 2641.6 1834.3
2014 Mean 109.2 2.1 11.0 41.4 30.8 144.8 109.9
Median 19.7 0.2 2.4 8.1 5.5 33.7 23.5
Std.dev 237.4 7.5 24.0 92.6 76.7 345.8 245.8
Note: Std.dev stands for standard deviation.

With the data collected and following the MRIO model presented in Section 3.1, the inputs and

outputs of GSCs can be calculated. For ease of interpretation, we aggregate the 28 European Union

member countries into a single EU-28 region.11 Figure 3 shows the foreign share in the overall

production activity and emissions induced by GSCs of manufacturing sectors for the remaining 16

economies. It is found that the 16 economies integrated into GSCs at varying degrees. Generally, the

developed economies and economies with limited territories, e.g. Switzerland, Japan, Korea and

EU-28, tended to import more raw materials and goods that contain high levels of pollution but low

10
The Eora data can be retrieved from https://www.worldmrio.com.
11
Readers should bear in mind that the aggregation may mask possible heterogeneity among the 28 European Union member
countries in terms of GSC performance. We thank an anonymous reviewer for pointing this issue out.

18
economic value. In contrast, large producers in the emerging economies, e.g. China and India, mainly

imported intermediates that are often high in economic value but low in levels of pollution along

GSCs to support domestic production. Comparing Fig. 3(a) and Fig. 3(b) reveals that the developed

economies’ participation in GSCs had been marginally deepened. On the other hand, China, India and

Indonesia witnessed a slight decline in the participation degree. This was accompanied by the

maturing industrial capabilities and growing demand in these emerging economies (McKinsey Global

Institute, 2019). Figure 4 shows that the overall emission intensities, particularly for SO2 and NOx, of

GSC production activities significantly differed from those of domestic production activities, and the

disparity generally expanded during the study period. These facts stress the necessity for measuring

supply chain environmental performance in a global context and from the industrial lifecycle

perspective. Next, we study the environmental performances of GSCs for the 16 economies’

manufacturing sectors in the decade from 2005 to 2014.

80%

60%

40%

20%

0%

CO2 SO2 NOx Value added

(a) 2005

19
 80%

60%

40%

20%

0%

CO2 SO2 NOx Value added

(b) 2014
Figure 3. Foreign share in the total emissions and total economic output of supply chains of
manufacturing sectors of 16 economies, 2005 and 2014

35
30
25
20
15
10
5
0

CO2-direct SO2-direct NOx-direct

CO2-gsc SO2-gsc NOx-gsc

(a) 2005
35
30
25
20
15
10
5
0

CO2-direct SO2-direct NOx-direct

CO2-gsc SO2-gsc NOx-gsc

(b) 2014

20
 Figure 4. Air emission intensities of 16 economies, 2005 and 2014 (units: Kg/USD for CO2, Kg/103
USD for SO2 and NOx,)

Note: Emission intensity is defined as the ratio of emissions to value added. The suffix ‘direct’ implies the
conventional production-based emission intensity, which accounts for the emissions and economic
activities that occur in a particular territory. On the other hand, the suffix ‘gsc’ denotes the global supply
chains and implies that the emission intensity accounts for all the emissions and economic output induced
by an entire supply chain.

5. Results and discussions

5.1 Static environmental performance of global supply chains

Following Eq. (7), we first estimate the static environmental performances of GSCs in individual

years. Here, we only present the results for 2005 and 2014 for illustration purposes. Figure 5 shows

that the 16 economies’ average GSC environmental performances were 0.61 and 0.63 for the two

years, respectively. The results imply that the overall environmental inefficiency was not negligible.

On average, roughly 40% of the pollutant emissions could have been reduced along GSCs if the most

efficient available production technology had been adopted. The results also suggest that the average

GSC environmental performance of the 16 economies had slightly shifted to the best practice frontier

during the study period. This shift is an indication of improvement in the technical efficiency of

emissions for the GSCs as a whole. Figure 6 further shows that the developed economies’

manufacturing sectors had a better average performance than those in the emerging economies in both

2005 and 2014.12 However, the gap between developed economies and emerging economies shrank,

due mainly to the progress made by the emerging economies. A possible reason is the increasingly

stringent environmental regulation applied in the emerging economies with the transfer of global

production in the past decades. Both the emerging economies and developed economies had top

performers, e.g. Mexico and Switzerland, respectively. On the other hand, only the emerging

economies, i.e. China and India, were always located at the lower end for both years.

12
According to the International Monetary Fund (IMF), seven countries in our dataset, i.e. Brazil, China, Indonesia, India,
Mexico, Russia, and Turkey, are emerging economies, and the remaining economies are classified as developed economies.
The IMF classification of emerging and developed economies is available at https://www.imf.org/en/Data.

21
 1.0

0.8

0.6

0.4

0.2

0.0

2005 2014

Figure 5. GSC environmental performances of 16 economies in 2005 and 2014

Figure 6. Box plots for SEPI of developed and emerging economies

Tables C1 and C2 in Appendix C reveals that the sources of the 16 economies’ GSC

environmental inefficiencies were fairly diverse from the sectoral perspective. As a top player in the

emerging economies, Mexico was located on the best practice frontier for the textile sector (C13-C15)

and coke and petroleum, chemical and pharmaceutical industries (C19-C21). On the other hand, the

country’s inefficiency in 2005 was sourced mainly from the manufacturing of rubber and metal

products (i.e. C22 and C24-C25). In 2014, Mexico’s overall technical efficiency in supply chain

emissions increased, arising mainly from the improvement in the capital-intensive industries. The

development coincided with the rapid expansion of the capital-intensive and technology-intensive

22
sectors in Mexico and the increasing participation of the country in GSCs by integrating in the North

American Free Trade Agreement (NAFTA) free trade zone since the 1990s. Indonesia performed

relatively well compared to other emerging economies. Its inefficiency was sourced primarily from

the capital-intensive and technology-intensive sectors, e.g. the pharmaceutical (C21), metal (C26),

and electronic (C28) industries. Nonetheless Indonesia registered an improvement in these two

categories, particularly in the coke and petroleum sector (C19) and automobile industry (C29). This

development was accompanied by the country’s rise as a main automobile production base in the

Southeast Asia.

The two largest emerging economies, China and India, recorded extremely low GSC

performance throughout the study period. The primary sources of environmental inefficiency for the

two economies were the capital-intensive and technology-intensive sectors. Tables D1 and D2 in

Appendix D further show that the inefficiency was induced mainly by CO2 and SO2 emissions.

Compared to the global best practice, China had a potential to reduce CO2 emissions by 2,845.0

million tons, SO2 emissions by 11.6 million tons, and NOx emissions by 5.2 million tons, which

respectively correspond to 77.4%, 71.9%, and 56.0% of the observed GSC emissions in 2014. The

largest abatement potential for these three pollutants came from the machinery sector (C28), the

electronic sector (C26) and the automobile industry (C29). These results identify the priorities for

further environmental protection measures. Another observation is that both China and India

experienced noteworthy and comprehensive improvements, particularly the former. The development

was accompanied by structural changes in the GSC participation patterns of these economies. With

deeper integration into GSCs, producers in China and India tended to conform with the stringent

environmental standards in developed economies in order to be qualified as suppliers, which is likely

to promote the overall technical efficiency in the two economies via spillover effects (Lin, Moon, &

Yin, 2014). This may partly explain the improvement in China and India.

Of the developed economies, Switzerland was ranked top in terms of GSC environmental

performance, followed by Norway. In particular, Switzerland’s overall GSC environmental efficiency

was unity in both 2005 and 2014, implying that the economy achieved the most efficient available

practice in all the 18 manufacturing sectors in the two years. As to the East Asia, Japan remained a

23
frontrunner, and Korea witnessed remarkable progress in the decade, especially in the

technology-intensive sectors. It is noteworthy that Taiwan stayed at the lower end among the

developed economies, though marginal improvement was achieved. The results are consistent with

the region’s stagnancy in economic development and industrial progress in the past decade. The

United States outperformed EU-28 in 2005 and 2014, but the two economies shared certain

similarities in the pattern of GSC performance. Their inefficiencies came primarily from the

capital-intensive sectors. Moreover, the United States registered a higher improvement. In 2014, only

three capital-intensive sectors (i.e. the pharmaceutical, rubber and mineral industries) and the

machinery industry were found environmentally inefficient, while the remaining 14 sectors of the

United States were located on the best practice frontier. On the other hand, EU-28 performed fairly

stable in this period.

5.2 Dynamic environmental performance of global supply chains

Further to the static performance in separate years, this section investigates the changes in GSC

environmental performance of the 16 economies and quantifies the underlying determinants. Figure

7(a) shows that the 16 economies’ environmental performance had increased by 20.6% on average in

2005-2014. Fluctuations in the development path of GSC performance were observed. Treating the 16

economies as a whole, the average environmental performance rose rapidly in 2006-2008 after a

stagnancy in 2005-2006. A sudden drop was found in 2008-2009, which was likely caused by the

global financial crisis occurred in that year. The average environmental performance then steadily

improved until 2013, followed by a slight fall of 1.5% in 2014. The findings are consistent with that

reported by Acquaye et al. (2018). At the global level, both the technical efficiency change and

production technological change improved the overall GSC performance, with the latter being the

primary contributor throughout the period studied, as shown in Fig.7(a). A notable observation is that

during the 2008/09 financial crisis, the driver of the falling GSC environmental performance was the

technological change. The results imply that the depressed market in that period hindered

technological innovation and the deployment of more efficient technologies. Besides, technical

efficiency in emissions declined in 2005-2007, due primarily to the changes in Australia, Turkey and

Canada. Although recovered in subsequent periods, the technical efficiency effect contributed to the

24
overall improvement at a fairly limited degree. Thus the results suggest that strengthening the supply

chain management to promote technical efficiency in emissions is critical to further improvement in

GSC performance.

2.00 2.00 2.00

1.75 1.75 1.75

1.50 1.50 1.50

1.25 1.25 1.25

1.00 1.00 1.00

0.75 0.75 0.75

0.50 0.50 0.50

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

DEPI Efficiency ch an ge Technological change DEPI Efficiency ch an ge Techn ological change DEPI Efficiency ch an ge Techn ological change

(a) Global average (b) Australia (c) Brazil

2.00 2.00 2.00

1.75 1.75 1.75

1.50 1.50 1.50

1.25 1.25 1.25

1.00 1.00 1.00

0.75 0.75 0.75

0.50 0.50 0.50

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

DEPI Efficiency ch an ge Techn ological change DEPI Efficiency ch an ge Techn ological change DEPI Efficiency ch an ge Techn ological change

(d) Canada (e) China (f) Indonesia

2.00 2.00 2.00

1.75 1.75 1.75

1.50 1.50 1.50

1.25 1.25 1.25

1.00 1.00 1.00

0.75 0.75 0.75

0.50 0.50 0.50

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

DEPI Efficiency ch an ge Techn ological change DEPI Efficiency ch an ge Techn ological change DEPI Efficiency ch an ge Techn ological change

(g) India (h) Japan (i) Korea

2.00 2.00 2.00

1.75 1.75 1.75

1.50 1.50 1.50

1.25 1.25 1.25

1.00 1.00 1.00

0.75 0.75 0.75

0.50 0.50 0.50

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

DEPI Efficiency ch an ge Technological change DEPI Efficiency ch an ge Technological change DEPI Efficiency ch an ge Technological change

(j) Mexico (k) Norway (l) Russia

2.00 2.00 2.00

1.75 1.75 1.75

1.50 1.50 1.50

1.25 1.25 1.25

1.00 1.00 1.00

0.75 0.75 0.75

0.50 0.50 0.50

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

DEPI Efficiency ch an ge Technological change DEPI Efficiency ch an ge Technological change DEPI Efficiency ch an ge Technological change

(m) Switzerland (n) Turkey (o) Taiwan

2.00 2.00

1.75 1.75

1.50 1.50

1.25 1.25

1.00 1.00

0.75 0.75

0.50 0.50
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

DEPI Efficiency ch an ge Technological change DEPI Efficiency ch an ge Techn ological change

(p) United States (q) EU-28

25
Figure 7. Cumulative aggregated DEPI and its determinants for manufacturing sectors in 16 economies,
2005–2014

Generally, the emerging economies witnessed dramatic growth in GSC environmental

performance by 33.9% during 2005-2014, while the progress made by the developed economies was

relatively moderate (11.2%). The disparity stemmed mainly from the GSC technical efficiency

changes occurred in the two groups of countries. In recent years, the emerging economies, e.g. China

and India, have implemented a lot of measures to improve environmental quality. Constrained by

strict regulations, manufacturing plants tended to improve their management and technical efficiency

to reduce emissions. Besides, the developed economies often have more stringent standards on

industrial production and emissions given their intensive and widespread awareness of environmental

protection. The stringent requirements apply not only to production plants located in the developed

economies but also often extend to suppliers that are located in the emerging economies. These

internal and external requirements together urge suppliers in the emerging economies to catch up in

terms of technique efficiency. It should be noted that, however, the emerging economies still shown

worse performance than the developed economies at the end of the period studied after substantial

growth in the decade. As to technological change, on the other hand, the two groups of economies

experienced similar progress. This may signal the trend of convergence among global economies in

terms of production technology with the transfer of manufacturing industries.

Figures 7(b)–7(q) show the results at the national and regional levels, and reveal that the sources

of changes in DEPI for various economies were fairly different. Of the nine developed economies,

only Canada and Australia are found to have deteriorated GSC performance, which was driven mainly

by the technical efficiency change. Improvements in production technology and technical efficiency

performance are observed for the rest seven developed economies except for a slightly falling

technical efficiency occurred in Japan. Switzerland remained on the best practice frontier throughout

the decade, and thus no change in its technical efficiency performance. All the seven emerging

economies experienced growth in GSC environmental performance except Brazil. The changes were

in line with the dynamics in emission intensities of these economies, as shown in Fig.4. For Indonesia

and Turkey, the rising GSC performance were driven primarily by the technological change, whereas

26
the key source of performance growth in China, India, Mexico and Russia was the improvement in

technical efficiency.

A closer look at the results reveals more sectoral and temporal features of the dynamics in GSC

environmental performance of the 16 economies. Here we present the results for only five

representative economies, i.e. China, India, Indonesia, the United States, and EU-28. As a turning

point of the GSC performances was in 2009, we divide the entire study period into two sub-periods,

i.e. period I (2005–2009) and period II (2009–2014). Table E1 in Appendix E shows that the growth

in China’s GSC environmental performance accelerated in period II. China’s manufacturing sectors

improved technical efficiency continuously in the decade and achieved similar progress in period I

and period II. Thus the acceleration in the overall GSC performance arose from the boosting

technological change after the global financial crisis. From the sectoral perspective, all the 18

manufacturing sectors in China improved in the two sub-periods and the only exception was the

deterioration in coke and refined petroleum industry (C19) during period I. After 2009, the

technology-intensive and capital-intensive sectors outperformed the labor-intensive sectors in terms of

GSC performance improvement. In particular, the ICT sector (C26) and the automobile industry (C29)

registered the most significant growth. A possible reason is that these two sectors usually involve in

GSCs most deeply, and thus face increasingly severe pressure on environmental standard of products

from overseas. The results are also consistent with the developments in China during the past decade

as the country had imposed stricter environmental regulations on heavy industries.

Table E2 in Appendix E shows that the development in India’s environmental performance had a

similar pattern with that in China despite different sources and degree of changes. India’s

improvement in period I was hindered by the marginal technological change, which arose mainly

from the deterioration in labor-intensive sectors, e.g. the lumber sector (C16) and the food and

beverage sector (C10-C12). In period II, the growth in environmental performance was remarkable,

and the primary driving forces were the efficiency improvement in labor-intensive sectors and the

technological change in technology-intensive sectors. As a recently emerging production base in the

global economy, Indonesia deserves special attention. A leap in the country’s GSC environmental

performance occurred in period I, followed by a relatively less growth after the global crisis, as shown

27
in Table E3 in Appendix E. Technological change remained the primary driving force behind the GSC

performance improvement in the two sub-periods, which was possibly due to the aggressive

development of Indonesia’s manufacturing sectors in GSCs. During the past decade, global

production gradually shifted from China to Southeast Asia, primarily to Indonesia, because of the

comparative advantage in lower labor costs. Advanced technologies and knowledge were diffused and

introduced to Indonesia with the industrial transfer, and thus helps the country to upgrade its industry.

For Indonesia’s performance improvement in period I, the automobile industry (C29) and the rubber

sector (C22) contributed the most. The former was largely dominated by several foreign

multinationals (e.g. Toyota and Mitsubishi), while the latter has been a pillar of Indonesia’s economy.

After the global crisis in 2008, the slowing growth was caused mainly by the lagging efficiency

improvement. Nonetheless an encouraging observation is that the capital-intensive and

technology-intensive sectors generally exhibited larger improvement in period II.

The United States and EU-28 exhibited some similarities and differences in their patterns of

changes in environmental performance, as shown in Tables E4 and E5 in Appendix E. Both

economies witnessed considerable improvements in their overall performances. The growth of the

United States occurred mainly in period I, while the improvement slowed down after the financial

crisis. On the contrary, EU-28 exhibited a very opposite pattern. The improvements in both economies

were largely driven by the technological change. It should be noted that the slowing performance

improvement in the United States in period II arose from the weakening efficiency change.

Meanwhile the technological change accelerated in the United States, particularly for the

technology-intensive and capital-intensive sectors, which nonetheless cannot compensate the

deterioration caused by the efficiency change. For EU-28, on the other hand, the acceleration in

period II stemmed mainly from the technological change in the electrical equipment, machinery and

automobile industries (C27-C30).

6. Conclusion

The objective of this study is to evaluate the environmental performance of GSCs. Our

contribution to the literature is twofold. First, we develop an input-output modeling approach to

28
measuring the performance of GSCs. The proposed approach is built upon the MRIO model and DEA.

The MRIO model is first adopted to fully account for all the production activities induced by a supply

chain from an industrial lifecycle perspective. With the estimates of inputs and outputs involved in

GSCs, we further model supply chains’ production processes within the total-factor production

framework using the DEA technique. Two indices, i.e. SEPI and DEPI, are introduced to respectively

measure the static and dynamic environmental performances of GSCs from an efficiency viewpoint.

Moreover, the sources of changes in performance are quantified by decomposing DEPI into two

components. Compared to other performance measurement tools in the existing literature, the key

advantage of our proposed input-output modeling approach lies in its capacity to capture the

multidimensional impacts of supply chain activities while accounting for the transnational and

complex natures of GSCs.

Second, we apply the proposed approach to study the environmental performances of GSCs in 16

major economies for 2005–2014 with a focus on the manufacturing sectors. The empirical results

show that the environmental inefficiency in the supply chains was not negligible during the decade.

On average, roughly 40% of CO2, SO2, and NOx emissions could have been reduced along GSCs if

the most efficient available production technology had been adopted. Driven primarily by the

technological improvement in the decade, the overall environmental performance of GSCs averagely

rose by 20.6%. At the country level, the developed economies outperformed the emerging economies

in terms of GSC performance during the period studied. A possible reason is the accelerating shift of

emission-intensive industries from the developed economies to the emerging economies during the

past decades. Moreover, the emerging economies experienced larger improvement than the developed

economies in terms of environmental performance. The results at the sectoral level will help

policymakers to monitor and scrutinize the environmental performance of GSCs in greater depth.

Inevitably, the present study has some limitations, which deserve further research in the future.

First, MRIO model formulates the supply chain dependencies and interactions using a system of linear

equations with the assumptions of constant returns to scale and no factor substitution (Jensen, 1980),

which may not be able to fully reflect the reality. Second, on the basis of the assessment of

environmental performances of GSCs, it could be meaningful to further study the impact of

29
participation in GSCs on environmental performance and reveal the mechanism behind the impact.

The results may suggest targeted policy measures that could better promote environmental

performance in the process of globalization, particularly for the fast-developing economies.

Acknowledgements

The authors are grateful to the anonymous reviewers for their constructive comments on an

earlier version of the paper. We also acknowledge the financial support provided by the National

Natural Science Foundation of China (nos. 71804189, 71934007, 71625005, 71573119 & 71573186),

the Humanities and Social Sciences Project of Ministry of Education in China (no. 18YJC630176),

the Shandong Provincial Natural Science Foundation (no. ZR2019QG001), and the Fundamental

Research Funds for the Central Universities (no. 19CX07005A).

References

Acquaye, A., Feng, K., Oppon, E., Salhi, S., Ibn-Mohammed, T., Genovese, A., & Hubacek, K. (2017).
Measuring the environmental sustainability performance of global supply chains: A multi-regional
input-output analysis for carbon, sulphur oxide and water footprints. Journal of Environmental Management,
187, 571-585. doi: https://doi.org/10.1016/j.jenvman.2016.10.059
Acquaye, A., Ibn-Mohammed, T., Genovese, A., Afrifa, G. A., Yamoah, F. A., & Oppon, E. (2018). A
quantitative model for environmentally sustainable supply chain performance measurement. European
Journal of Operational Research, 269(1), 188-205. doi: https://doi.org/10.1016/j.ejor.2017.10.057
Barbosa-Póvoa, A. P., da Silva, C., & Carvalho, A. (2018). Opportunities and challenges in sustainable supply
chain: An operations research perspective. European Journal of Operational Research, 268(2), 399-431. doi:
https://doi.org/10.1016/j.ejor.2017.10.036
Boons, F., Baumann, H., & Hall, J. (2012). Conceptualizing sustainable development and global supply chains.
Ecological Economics, 83, 134-143. doi: https://doi.org/10.1016/j.ecolecon.2012.05.012
Brandenburg, M., Govindan, K., Sarkis, J., & Seuring, S. (2014). Quantitative models for sustainable supply chain
management: Developments and directions. European Journal of Operational Research, 233(2), 299-312.
doi: https://doi.org/10.1016/j.ejor.2013.09.032
Chen, C.-M. (2013). A critique of non-parametric efficiency analysis in energy economics studies. Energy
Economics, 38, 146-152. doi: http://dx.doi.org/10.1016/j.eneco.2013.03.009
Chen, C.-M. (2014). Evaluating eco-efficiency with data envelopment analysis: an analytical reexamination.
Annals of Operations Research, 214(1), 49-71. doi: 10.1007/s10479-013-1488-z

30
Chen, C.-M., & Delmas, M. A. (2012). Measuring Eco-Inefficiency: A New Frontier Approach. Operations
Research, 60(5), 1064-1079. doi: 10.1287/opre.1120.1094
Chen, C., & Yan, H. (2011). Network DEA model for supply chain performance evaluation. European Journal of
Operational Research, 213(1), 147-155. doi: https://doi.org/10.1016/j.ejor.2011.03.010
Cohen, M., & Mallik, S. (1997). Global supply chains: Research and applications. Production and Operations
Management, 6(3), 193-210. doi: 10.1111/j.1937-5956.1997.tb00426.x
Costantini, V., Crespi, F., Marin, G., & Paglialunga, E. (2017). Eco-innovation, sustainable supply chains and
environmental performance in European industries. Journal of Cleaner Production, 155, 141-154. doi:
https://doi.org/10.1016/j.jclepro.2016.09.038
Ding, H., Liu, Q., & Zheng, L. (2016). Assessing the economic performance of an environmental sustainable
supply chain in reducing environmental externalities. European Journal of Operational Research, 255(2),
463-480. doi: https://doi.org/10.1016/j.ejor.2016.05.003
Dou, Y., Zhu, Q., & Sarkis, J. (2014). Evaluating green supplier development programs with a grey-analytical
network process-based methodology. European Journal of Operational Research, 233(2), 420-431. doi:
https://doi.org/10.1016/j.ejor.2013.03.004
Du, J., Chen, Y., & Huang, Y. (2018). A Modified Malmquist-Luenberger Productivity Index: Assessing
Environmental Productivity Performance in China. European Journal of Operational Research, 269(1),
171-187. doi: https://doi.org/10.1016/j.ejor.2017.01.006
Elkington, J. (1997). Cannibals with Forks: The Triple Bottom Line of 21st Century Business. John Wiley Press.
Färe, R., & Grosskopf, S. (2000). Theory and Application of Directional Distance Functions. Journal of
Productivity Analysis, 13(2), 93-103. doi: 10.1023/A:1007844628920
Färe, R., Grosskopf, S., & Hernandez-Sancho, F. (2004). Environmental performance: an index number approach.
Resource and Energy Economics, 26(4), 343-352. doi: http://dx.doi.org/10.1016/j.reseneeco.2003.10.003
Färe, R., Grosskopf, S., Lovell, C. A. K., & Pasurka, C. A. (1989). Multilateral Productivity Comparisons When
Some Outputs are Undesirable: A Nonparametric Approach. The Review of Economics and Statistics, 71(1),
90-98. doi: 10.2307/1928055
Fransoo, J. C., & Lee, C.-Y. (2013). The Critical Role of Ocean Container Transport in Global Supply Chain
Performance. Production and Operations Management, 22(2), 253-268. doi:
10.1111/j.1937-5956.2011.01310.x
Genovese, A., Morris, J., Piccolo, C., & Koh, S. C. L. (2017). Assessing redundancies in environmental
performance measures for supply chains. Journal of Cleaner Production, 167, 1290-1302. doi:
https://doi.org/10.1016/j.jclepro.2017.05.186
Gereffi, G., Humphrey, J., & Sturgeon, T. (2005). The governance of global value chains. Review of International
Political Economy, 12(1), 78-104. doi: 10.1080/09692290500049805
Gupta, S., & Palsule-Desai, O. D. (2011). Sustainable supply chain management: Review and research
opportunities. IIMB Management Review, 23(4), 234-245. doi: https://doi.org/10.1016/j.iimb.2011.09.002
Jensen, R. C. (1980). The Concept of Accuracy in Regional Input-Output Models. International Regional Science
Review, 5(2), 139-154. doi: 10.1177/016001768000500203
Kao, C. (2009). Efficiency measurement for parallel production systems. European Journal of Operational
Research, 196(3), 1107-1112. doi: https://doi.org/10.1016/j.ejor.2008.04.020
Kao, C. (2014). Network data envelopment analysis: A review. European Journal of Operational Research,
239(1), 1-16. doi: https://doi.org/10.1016/j.ejor.2014.02.039

31
Kao, C. (2017). Measurement and decomposition of the Malmquist productivity index for parallel production
systems. Omega, 67, 54-59. doi: https://doi.org/10.1016/j.omega.2016.04.001
Kim, B., Park, K. S., Jung, S.-Y., & Park, S. H. (2018). Offshoring and outsourcing in a global supply chain:
Impact of the arm’s length regulation on transfer pricing. European Journal of Operational Research, 266(1),
88-98. doi: https://doi.org/10.1016/j.ejor.2017.09.004
Koberg, E., & Longoni, A. (2019). A systematic review of sustainable supply chain management in global supply
chains. Journal of Cleaner Production, 207, 1084-1098. doi: https://doi.org/10.1016/j.jclepro.2018.10.033
Lenzen, M., Kanemoto, K., Moran, D., & Geschke, A. (2012). Mapping the Structure of the World Economy.
Environmental Science & Technology, 46(15), 8374-8381. doi: 10.1021/es300171x
Lin, L., Moon, J., & Yin, H. (2014). Does International Economic Integration Lead to a Cleaner Production in
China? Production and Operations Management, 23(4), 525-536. doi: doi:10.1111/poms.12087
McKinsey Global Institute. (2019). Globalization in transition: The future of trade and value chains.
McKinsey&Company.
Min, H., & Kim, I. (2012). Green supply chain research: past, present, and future. Logistics Research, 4(1), 39-47.
doi: 10.1007/s12159-012-0071-3
Mirhedayatian, S. M., Azadi, M., & Farzipoor Saen, R. (2014). A novel network data envelopment analysis model
for evaluating green supply chain management. International Journal of Production Economics, 147,
544-554. doi: https://doi.org/10.1016/j.ijpe.2013.02.009
O’Rourke, D. (2014). The science of sustainable supply chains. Science, 344(6188), 1124-1127. doi:
10.1126/science.1248526
Pasurka, C. A. (2006). Decomposing electric power plant emissions within a joint production framework. Energy
Economics, 28, 26-43. doi: 10.1016/j.eneco.2005.08.002
Pourakbar, M., & Zuidwijk, R. A. (2018). The role of customs in securing containerized global supply chains.
European Journal of Operational Research, 271(1), 331-340. doi: https://doi.org/10.1016/j.ejor.2018.05.012
Pulina, M., Detotto, C., & Paba, A. (2010). An investigation into the relationship between size and efficiency of
the Italian hospitality sector: A window DEA approach. European Journal of Operational Research, 204(3),
613-620. doi: https://doi.org/10.1016/j.ejor.2009.11.006
Qorri, A., Mujkić, Z., & Kraslawski, A. (2018). A conceptual framework for measuring sustainability
performance of supply chains. Journal of Cleaner Production, 189, 570-584. doi:
https://doi.org/10.1016/j.jclepro.2018.04.073
Timmer, M. P., Dietzenbacher, E., Los, B., Stehrer, R., & de Vries, G. J. (2015). An illustrated user guide to the
world input–output database: The case of global automotive production. Review of International Economics,
23, 575–605.
Tone, K. (2001). A slacks-based measure of efficiency in data envelopment analysis. European Journal of
Operational Research, 130(3), 498-509. doi: http://dx.doi.org/10.1016/S0377-2217(99)00407-5
United Nations, & BSR. (2010). Supply Chain Sustainability: A Practical Guide for Continuous Improvement.
United Nations Global Compact and Business for Social Responsibility.
Wang, H., Ang, B. W., Wang, Q. W., & Zhou, P. (2017a). Measuring energy performance with sectoral
heterogeneity: A non-parametric frontier approach. Energy Economics, 62, 70-78. doi:
http://dx.doi.org/10.1016/j.eneco.2016.12.005
Wang, H., Ang, B. W., & Zhou, P. (2018). Decomposing aggregate CO2 emission changes with heterogeneity: An
extended production-theoretical approach. Energy Journal, 39(1), 59-79. doi: 10.5547/01956574.39.1.hwan

32
Wang, H., Zhou, P., & Zhou, D. Q. (2013). Scenario-based energy efficiency and productivity in China: A
non-radial directional distance function analysis. Energy Economics, 40, 795-803. doi:
10.1016/j.eneco.2013.09.030
Wang, M., & Huang, H. (2018). The design of a flexible capital-constrained global supply chain by integrating
operational and financial strategies. Omega. doi: https://doi.org/10.1016/j.omega.2018.11.016
Wang, Z., Wei, S.-J., Yu, X., & Zhu, K. (2017b). Characterizing Global Value Chains: Production Length and
Upstreamness. National Bureau of Economic Research Working Paper Series, No. 23261. doi:
10.3386/w23261
Wang, Z., Wei, S.-J., Yu, X., & Zhu, K. (2017c). Measures of Participation in Global Value Chains and Global
Business Cycles. National Bureau of Economic Research Working Paper Series, No. 23222. doi:
10.3386/w23222
Wood, R., & Lenzen, M. (2009). Structural path decomposition. Energy Economics, 31(3), 335-341. doi:
10.1016/j.eneco.2008.11.003
WTO. (2015). World Trade Organization: International Trade Statistics 2015.
Zhou, H., Yang, Y., Chen, Y., & Zhu, J. (2018). Data envelopment analysis application in sustainability: The
origins, development and future directions. European Journal of Operational Research, 264(1), 1-16. doi:
https://doi.org/10.1016/j.ejor.2017.06.023
Zhou, P., Ang, B. W., & Poh, K. L. (2008). Measuring environmental performance under different environmental
DEA technologies. Energy Economics, 30, 1-14. doi: http://dx.doi.org/10.1016/j.eneco.2006.05.001
Zhou, P., Ang, B. W., & Wang, H. (2012). Energy and CO 2 emission performance in electricity generation: A
non-radial directional distance function approach. European Journal of Operational Research, 221(3),
625-635.
Zhu, Q., Sarkis, J., & Lai, K.-h. (2007). Green supply chain management: pressures, practices and performance
within the Chinese automobile industry. Journal of Cleaner Production, 15(11), 1041-1052. doi:
https://doi.org/10.1016/j.jclepro.2006.05.021

33