Economic Systems Research

Vol. 17, No. 4, 365 –391, December 2005

A Consistent Input – Output Formulation

of Shared Producer and Consumer
Responsibility

BLANCA GALLEGO & MANFRED LENZEN

School of Physics, The University of Sydney, Australia

ABSTRACT Growing interest in environmental and socio-economic accounting at the sub-regional

and organisational level requires a consistent and comprehensive method for computing and
reporting responsibility for impacts of industrial production such as water use, pollution, or
employment. This work presents a formulation for allocating responsibility for production
impacts consistently amongst all agents such as consumers, producers, workers, and investors
throughout demand and supply chains, in a way that reflects their contribution to the production
process. Generalised input – output theory is used to re-trace the flow of past inter-industrial
transactions, and examine ex-post how, for example, inputs of resources or outputs of pollution
were associated with these transactions. Introducing the concept of a responsibility share we
enable the division of responsibility into mutually exclusive and collectively exhaustive portions
that are assigned to the different economic sectors, and that become consistently smaller as we
move away from the location of the impact within the supply or demand chain system.

KEY WORDS : Extended responsibility, producer, consumer, supply chain, environmental impact

1. Introduction and Aim

It is perhaps because of the tendency of economic policy in market-driven economies not
to interfere with final demand and consumers’ preferences that the producer-centric
representation is the dominant form of viewing the environmental or social impacts of
industrial production. Statistics on energy, labour, water etc are almost always presented
as attributes of industries rather than as attributes of the supply chains of products for
consumers (see for example ABS, 1996; ABARE, 1999). On a smaller scale, most existing
schemes for corporate sustainability report only on impacts that arise out of operations
controlled by the reporting company, and not on supply-chain impacts (WBCSD and
WRI, 2001). According to this world view, ‘upstream and downstream [environmental]
impacts are . . . allocated to their immediate producers. The institutional setting and the

Correspondence Address: Blanca Gallego, School of Physics, A28, The University of Sydney NSW 2006,
Australia. Email: [email protected]

0953-5314 Print=1469-5758 Online=05=040365–27 # 2005 The International Input–Output Association

DOI: 10.1080=09535310500283492
366 B. Gallego & M. Lenzen

different actors’ spheres of influence are not reflected’ (Spangenberg and Lorek, 2002,
p. 131).
On the other hand, a number of studies have highlighted that final consumption and
affluence, especially in the industrialised world, are the main drivers for the level and
growth of environmental pressure.1 Even though these studies provide a clear incentive
for complementing producer-focused environmental policy with some consideration for
consumption-related aspects, demand-side measures to environmental problems are
rarely exploited (Princen, 1999, p. 348).
When thinking about environmental impacts, producers and consumers, crucial
questions arise such as: who is responsible for what, or: how is the responsibility to be
shared, if at all? For example: should a firm have to improve the eco-friendliness of its
products, or is it up to the consumer to buy or not to buy? And further: should the firm
be held responsible for only the downstream consequences of the use of its products, or
– through its procurement decisions – also for the implications of its inputs from upstream
suppliers? And if so, how far should the downstream and upstream spheres of responsibil-
ity extend? Similar questions can be phrased for the problem of deciding who takes the
credit for successful abatement measures that involved producers and consumers.
As with many other allocative problems, a politically viable consensus probably lies
somewhere between producer and consumer responsibility. It is intuitively clear that the
responsibility for impacts associated with transactions in a productive system has to be
somehow divided between the supplier and the recipient of the respective delivered com-
modity. In order to assign responsibility to agents participating in these transactions, one
has to know the respective supply chains or inter-industry relations.
One method that deals with inter-industry relations is input – output analysis. This
technique is mostly applied in order to determine quantity and price impacts of exogen-
ously stated perturbations. It can also be used to re-trace the flow of past transactions,
and examine ex-post how, for example, inputs of resources or outputs of pollution were
associated with these transactions. In this view, input – output analysis is not used as a
causal (quantity or price) impact model, but rather as a descriptive tool. Therefore, in
this work, we apply input – output analysis in a descriptive, ex-post manner for distributing
responsibility for resource use and pollution caused during industrial production.
The novel idea in our work is that – having responsibility as a metric, rather than
quantities or monetary values – flows can be divided, and for each transaction be assigned
partly to the supplier, and partly to the recipient of the commodity. Accordingly, the
main part of this article is about how to describe this division and distribution with an
input –output model. Thus, our approach addresses the fact that ‘no scientifically
rigorous macro-level allocation of responsibilities for the impacts of consumption to
specific groups of actors . . . has been developed so far’ (Spangenberg and Lorek, 2002,
p. 128).
In the following sections, we first review the literature on the topic, and then define
terms and variables, and outline the basic equations of the relevant input – output
theory. We continue by describing a previous attempt for including both producers and
consumers into an input –output-type impact model. Highlighting the shortcomings of
this total flow model helps us to motivate our own approach. We illustrate how our
new formulation describes shared responsibility, discuss implications for practice, and
conclude. While our exposition of the topic is coined in terms of environmental
impacts, it is equally applicable to other types, such as social impacts.
 Shared Producer and Consumer Responsibility 367

2. Producer and Consumer Responsibility: a Brief Review of the Literature

The nexus created by the different views on impacts caused by industrial production is
exemplified by several contributions to the discussion about producer or consumer respon-
sibility for greenhouse gas emissions (Proops et al., 1993; Wyckoff and Roop, 1994; Imura
and Moriguchi, 1995; Subak, 1995; Korhonen, 2002; Lenzen et al., 2004; Bastianoni et al.,
2004). Emissions data are reported to the IPCC as contributions of producing industries
located in a particular country (Task Force on National Greenhouse Gas Inventories,
1996) rather than as embodiments in products consumed by a particular population,
irrespective of productive origin. However, especially for open economies, taking into
account the greenhouse gases embodied in internationally traded commodities can have
a considerable influence on national greenhouse gas balance sheets. Assuming consumer
responsibility, exports have to be subtracted from, and imports added to, national green-
house gas inventories. In Denmark for example, Munksgaard and Pedersen (2000) report
that a significant amount of power and other energy-intensive commodities are traded
across Danish borders, and that between 1966 and 1994 the Danish foreign trade
balance in terms of CO2 developed from a 7 Mt deficit to a 7 Mt surplus, compared
with total emissions of approximately 60 Mt. In particular, electricity traded between
Norway, Sweden and Denmark is subject to large annual fluctuations due to varying rain-
fall in Norway and Sweden. In wet years Denmark imports hydro-electricity whereas
electricity from coal-fired power plants is exported in dry years. The official Danish
emissions inventory includes a correction for electricity trade and thus applies the
consumer responsibility principle (Danish EPA, 1998).
An unusual perspective is also taken in Extended Producer Responsibility (EPR) frame-
works: ‘Producers of products should bear a significant degree of responsibility (physical
and/or financial) not only for the environmental impacts of their products downstream
from the treatment and disposal of their product, but also for their upstream activities
inherent in the selection of materials and in the design of products’ (OECD, 2001,
pp. 21 –22). ‘The major impetus for EPR came from northern European countries in the
late 1980s and early 1990s, as they were facing severe landfill shortages. . . . [As a
result,] EPR is generally applied to post-consumer wastes which place increasing physical
and financial demands on municipal waste management’ (EPA NSW, 2003, pp. 2 –4).
In trying to operationalise EPR, the Chartered Institute of Purchasing and Supply UK
(CIPS, 1993) launched voluntary guidelines for environmental purchasing (CIPS, 2000)
and ethical business practices (CIPS, 1999), and recommends achieving seven goals:
(1) establishment of a business case to make environmental purchasing viable and part
of day-to-day operations, (2) an understanding of the environmental issues affecting the
organisation and its supply chain, (3) the development of a purchasing policy which
addresses environmental issues, (4) environmental criteria for ranking suppliers, (5)
improved communication with suppliers, (6) suitable methods for collecting relevant
information, and (7) agreed targets for further environmental performance improvements.
In practice, many companies have implemented policies that are aimed at reducing CO2
emissions and packaging waste from upstream suppliers and increasing recyclability, or
supplier environmental awards.2 These environmental purchasing practices are not
limited to the domestic economy, but can include decisions about preferable production
locations (Maxie, 1994). At times, queries by users of products about the environmental
performance of the supplier have initiated knock-on effects that lead upstream
368 B. Gallego & M. Lenzen

organisations to begin conducting environmental audits and implementing environmental

management systems (Barry, 1996). At present, most of these initiatives proceed in a more
or less qualitative and ad-hoc, rather than quantitative and systematic way in selecting,
screening, ranking or influencing other agents in their supply chain. We agree with
Lloyd (1994) that ‘it will be impossible to produce a sufficiently credible ranking of
suppliers without quantitative rating’.

3. Generalised Input–Output Theory: the Leontief and Ghosh Models

We now proceed by applying input –output analysis in a descriptive, ex-post manner for
distributing responsibility for resource use and pollution caused during industrial
production. Input – output theory is an analytical framework constructed from economic
data that connects gross sectoral3 outputs, or equivalently inputs, with final demand
(upstream or Leontief model ) or primary inputs (downstream or Ghosh model ). The mon-
etary flows in a particular time period (usually a financial year) from each producing to
each consuming industry are recorded in an inter-industrial transactions matrix T. The
gross output xi of a sector i is the sum of the sector’s final demand yi plus all its inter-
industry sales xi ¼ yi þ Sj tij . Analogously, the gross input xi of a sector (which – in a
static economy – is also its output) is the sum of the sector’s primary inputs vi plus all
its inter-industry purchases xi ¼ vi þ Sj tji .
In the demand-driven, upstream or Leontief model, a matrix of direct requirements A is
defined as the inter-industrial flows aij from an industry i to an industry j per gross output
of sector j, aij ¼ tij =xj . An element lij of the total requirements matrix, or Leontief inverse,
L ¼ (I
A)
1 represents the amount of gross output x ¼ (I
A)
1 y from sector i that
was produced to satisfy a unit of final demand y from sector j. I is the identity matrix.
On the other hand, in the downstream or Ghosh model, a matrix of direct sales Ā, is
defined as the inter-industrial flows a
ij ¼ tij =xi from an industry i to an industry j per
gross input of i. An element
lij of the total sales matrix, or Ghosh inverse
L̄ ¼ (I
Ā)
1 , represents the amount of gross input x0 ¼ v0 (I-Ā)21 into industry j that
absorbed, or utilised a unit of primary inputs v into industry i. The prime denotes
transposition.
There has been considerable debate on the correct economic interpretation of the Ghosh
model. The author himself (Ghosh, 1958) imagines a monopolistic or centrally planned
market with scarce resources, where allocation rather than production functions govern,
and where shortages lead to price increases and rationing. Although Ghosh perceived
the a
ij as value coefficients, these were subsequently employed as quantity coefficients.
The perception of the sales matrix as a price model to be used for analysing cost-push
inflationary processes was introduced by Dietzenbacher (1997). In this interpretation,
primary input prices change exogenously, are entirely passed on to price-taking purcha-
sers, and change only output values, while quantities are fixed. Ex ante, the Ghosh
model cannot be interpreted in a physical, causal sense: supply-side multipliers
lij do
not quantify the amount of output generated by an injection of primary inputs. Ex post,
however, the Ghosh model can be used descriptively for re-tracing embodiments of
primary inputs in downstream sales.
In order to trace social and environmental impacts, both input –output models can be
extended to account for so-called production factors (such as employment, resources,
pollution etc) associated with economic activity, and are then referred to as generalised
 Shared Producer and Consumer Responsibility 369

input –output models. A vector of coefficients f is defined, where each element fj represents
the impact (use, release, quantity) in terms of the production factor (in its own physical
units) associated with one unit of gross output xj of sector j.
Using the Leontief model, responsibility for factor impacts can be distributed across
final demand. The elements lij of the Leontief inverse can be understood as the amount
of gross output in industry i needed to satisfy final demand j. Final demanders are ulti-
mately responsible for the impacts fi lij that these outputs entailed during their production.
Using the Ghosh model, the responsibility for factor impacts can be distributed across
primary inputs. The elements
lij of the Ghosh inverse can be interpreted as the amount
of gross output of industry j generated by a unit primary input into industry i. Primary
input providers are ultimately responsible for the impacts
lij fj that these outputs entailed
to be produced.4
Accordingly, within generalised input – output analysis, total responsibility F for factor
impacts can be represented as the responsibility of producers, because of their productive
activities:
X X X (p)
F¼ F (p) ¼
i i i
f i x i ¼ f 0 x ¼ x0 f ¼ i
F
i (1)

or of final demanders (private households, recipients of capital goods, the government as

consumer, foreign importers), because without their demand, upstream sectors would have
impacted less:
X X
X 
F¼ F (c) ¼
i i i j
fj l ji yi ¼ f 0 Ly (2)

or of providers of primary inputs (workers, investors, the government as provider of

administration of industry, foreign exporters),5 because of them having supported
(worked for, invested in, or administered) production, since without their input, down-
stream sectors would have impacted less:6
X X
X 
(c)
F¼ F
¼ n
lij fj ¼ v0 L̄f (3)
i i i i j

Note that final demanders and providers of primary inputs represent the same segment of
society, since the entirety of citizens occupies simultaneously (i) the roles of private con-
sumers, consumers of capital, consumers in the public service, and consumers of imported
goods, and (ii) those of a employees, providers of savings for investment, public servants,
and producers of exports. Hence, equations (2) and (3) describe the responsibility for
impacts of citizens in their non-corporate roles. Equation (1) describes the responsibility
of those citizens that are owners of the means of production.
Neither the Leontief nor the Ghosh model is capable of assigning partial responsibility
for factor impacts to various agents along a supply chain (which is the aim of our work).
They either deal with consumer/worker-investor citizens (and then with entire supply
chains) or with producers (and then only on-site). The Leontief model for example ‘is
an unsatisfactory basis to determine the influence of households as state and business
do not show up as influencing environmental impacts: their activities are considered to
be totally demand driven. De facto, however, they influence the environmental impact
370 B. Gallego & M. Lenzen

by their own decisions (how to produce and provide goods and services) as well as through
their influence on consumer decisions’ (Spangenberg and Lorek, 2002, p. 131). Ideally, a
responsibility measure should include all actors that play a role in production and
consumption.
An early attempt to develop an impact measure that deals with producers and consumers
in a supply-chain formulation is Szyrmer (1992); based on Jeong’s (1982; 1984) total flow
concept, which we review in the following section.

4. Indirect Requirements of Producers: the Total Flow Concept

Szyrmer’s motivation for developing total flow arose out of the inability of Leontief’s
classical final-demand-driven inter-industry model to account for total industrial output,
but only for the fraction of output that is delivered into final demand. Taking this approach,
the responsibility for downstream impacts of some mining industries for example is
negligible, since most ores enter other industries rather than being absorbed by final
consumers, and all environmental consequences of mining would be passed on to metal
works and downstream manufacturers and their customers (compare Milana, 1985,
p. 284). Thus, Szyrmer sought to formulate a measure that would account for transactions
amongst producers as well as between producers and consumers, in other words for inter-
mediate as well as for final demand. Szyrmer thought the total flow concept to be particu-
larly applicable to firms, existing or new, and for identifying ‘key sectors’. While the
concept was introduced in terms of quantities, it applies equally to proxies such as
responsibility.
In its essence, the total flow concept results in requirement coefficients for gross output
x (as opposed to final demand y), computed as elements

@xi @xi =@yj lij

¼ ¼ (4)
@xj @xj =@yj l jj

of the Leontief inverse L, normalised by the diagonal elements ljj . These coefficients
‘translate the change in the jth gross output into a change in the jth final demand . . .
[, and] the jth final demand into ith gross output’ (Szyrmer, 1992, p. 925). Note that
instead of the causal relationship x ¼ Ly linking final demand (cause) with total output
(effect) in the simple Leontief model, the total flow model assumes a particular (non-
causal) association between xj and yj . The element l
1 jj determines the magnitude of
final demand yj that is necessary to absorb a total output xj . As a consequence, ‘the
total flow coefficients . . . do not express the changes in the economy created by Dxj ,
but those associated with Dxj ‘(Szyrmer, 1992, p. 928).
Szyrmer (1992, p. 928) correctly points out an important disadvantage of the total flow
concept, which is also at the heart of the argument pursued in this article:

A computational consequence of the non-causal nature of the total flow model is its
nonadditivity feature. As is well known, in the standard Leontief model, each unit of
final demand has its own ‘support network’, that is, its own direct and indirect inputs
that are perfectly separable from other inputs required by other final demand units.
Thus, the whole production system becomes a collection of mutually exclusive and
 Shared Producer and Consumer Responsibility 371

collectively exhaustive production inputs required by a given final demand mix. In

the total flow model the mutual exclusiveness of inputs is not present. The same
quantity of input i may be required at the same time by gross output of two (or
more) different sectors, say j and k independently. This non-additivity property
results in a number of computational (and conceptual) inconveniences. The total
flow coefficient matrix cannot be expressed as the sum of an infinite power series.
When using a total flow model for impact analysis, we should consider, in principle,
only one sector at a time.

A further disadvantage of the total flow concept is that the coefficients are highly depen-
dent on the diagonal entries ljj of the Leontief inverse, which in turn are somewhat arbi-
trary because they depend critically on the degree and nature of aggregation over
enterprises and products in the respective input –output table. In contrast, Leontief-type
inverse coefficients are independent of the magnitude of the diagonal entries of the
direct requirements matrix (for a proof see Weber, 1998).
Another approach seeking to describe the effect of producing entities on production
factors or total output is the hypothetical extraction method (Strassert, 1968), which oper-
ates by eliminating a sector (or a company) from an input – output table (by setting to zero
the corresponding row and column), and then examining the effect of this elimination on
multipliers. Hypothetical extractions have often been used for calculating linkages and
identifying ‘key sectors’ (see for example Dietzenbacher et al., 1993). Szyrmer and
Walker (1983) introduced the notion of total flow, using a hypothetical extraction
approach instead of equation (4). In fact, both methods are equivalent as first pointed
out by Szyrmer (1992), and proven in Appendix 1. As the total flow method, hypothetical
extractions are affected by non-additivity, since when adding up all extraction effects, each
transaction between two producers is eliminated twice.
Non-additivity was already recognised as a problem for determining indirect require-
ments for gross output by Milana (1985). Based on Miyazawa’s (1966) formalism for
partitioned input – output matrices, Milana and later Heimler (1991) constructed multi-
pliers for the gross output of an industry sector or a company by separating that sector
or company from the rest of the economy.7 Indexing a company with p and the remainder
of the economy with q,8 a partitioned formulation of the Leontief system is

      
xp a pp a0 pq xp yp
¼ þ (5)
xq aqp Aqq xq yq

which can be solved for xq according to

xq ¼ (I
Aqq )
1 aqp xp þ (I
Aqq )
1 yq (6)

The term aqp xp represents the company’s purchased inputs, and (I
Aqq )
1 aqp xp are the
total output requirements for producing these inputs. In equation (6), the company is effec-
tively treated as part of final demand, and feedback a0 pq from the company into the rest of
the economy is excluded. As with the total flow and hypothetical extraction methods, if the
above procedure was repeated for each company, intermediate transactions would appear
372 B. Gallego & M. Lenzen

multiply in different companies’ total output requirements, and thus lead to double
counting.9
Finally, Ferng’s (2003) production-benefit principle involves double counting for
similar reasons. Even his shared-responsibility formulation FA þ (1 2 F)B, where A
and B are allocations of responsibility according to the consumer- and producer-benefit
principles and 0 , F , 1, does not solve the problem, since double counting is always
inherent in the second term.
We have dealt with the total flow concept in more detail because it provides us with the
main motivation for our work: to find a responsibility measure that combines all actors in a
supply-chain context, but that is not affected by non-additivity/double counting. Since this
measure has to avoid assigning supply chains fully to more than one producing entity, our
formulation automatically leads to the feature of shared responsibility.

5. Allocating Shared Responsibility for Supply-chain Impacts in Production

Systems
As an example, let us consider the problem of allocating responsibility for the impacts of
(a) CO2 emissions (assumed as associated only with the generation of electricity), and (b)
waste (assumed as associated only with car manufacturing) in a simplified economy with
only three industrial sectors: the electricity sector e, a steel industry s that purchases from
e, and a car manufacturing sector c that purchases from both e and s. We will further
assume that only the electricity and the car manufacturing sectors receive primary
inputs and deliver to final consumers. For the sake of brevity we will refer to sectors
and industries, but the concept applies equally to companies. The transactions in this
hypothetical economy are illustrated in Figure 1, and a corresponding input – output
table is given in Table 1.
We will start with the responsibility for upstream impacts resulting from CO2
emissions.10 As pointed out in equation (1), one possibility would be to allocate all the
responsibility to the producer, in this case the (thermal) electricity sector. In this frame-
work, the responsibility F for total emissions of CO2 lies only with the electricity sector
(indexed e):

F ¼ Fe(p) ¼ fe xe (7)

The other possibility contemplated in the generalised Leontief model in equation (2), con-
sists of re-allocating these emissions as an upstream responsibility to final consumers. The

Figure 1. Supply-chain relations in a hypothetical three-sector economy (e ¼ electricity generation,

s ¼ steel making, c ¼ car manufacturing, v ¼ primary input, y ¼ final demand)
 Shared Producer and Consumer Responsibility 373

Table 1. Generalised input – output table of a hypothetical three-sector economy

Electricity Steel Cars Final demand Waste

Electricity 0 tes tec ye 0

Steel 0 0 tsc 0 0
Cars 0 0 0 yc Fc

Primary inputs ve 0 vc
CO2 emissions Fe 0 0

responsibility for the power sector’s CO2 emissions would then be distributed among final
demands as:

Fj(c) ¼ fe lej yj (8)

where j is any sector the final demand yj of which requires some output from the electricity
sector e, in any layer of production. For example, the final consumers of cars yc would be
assigned the amount fe lec yc , which divides into fe aec yc for the car manufacturing industry’s
direct use of thermal electricity, as well as the amount fe aes asc yc from the use of electricity
in the steel industry s that provides its steel to car manufacturers. Since car manufacturers
are assumed to produce no CO2, there is no direct CO2 responsibility associated with
buying a car.
The producer responsibility approach (equation (7)) ignores using industries and final
consumers that are located downstream in the demand chain, and that ultimately drive pro-
duction, and – one could argue – should therefore be partly responsible. On the other
hand, the consumer responsibility approach (equation (8)) fails to consider industrial
sectors that are playing a part in the production process but that deliver only to other indus-
trial sectors and not into final demand. Returning to our simple example, even though
being a part of the industrial metabolism by driving electricity demand and therefore
emissions, the steel sector is not responsible for any amount of CO2, neither under produ-
cer nor under consumer responsibility.
Similar arguments hold in the case of downstream responsibility for waste.11 Under
producer responsibility only the car manufacturers are responsible:

(p)
F ¼ F
c ¼ xc fc (9)

Under worker/investor responsibility in equation (3), the responsibility for emissions

would be distributed among primary inputs as:

(c)
F
j ¼ nj
l jc fc (10)

where j is any sector with primary input nj driving some output from the car manufacturing
sector c into final demand, via any number of production layers. For example, working or
investing in a power plant (value ne ) would be assigned the responsibility ne
lec fc , which
divides into ne a
ec fc for the power plant’s involvement in the car industry c, as well as
374 B. Gallego & M. Lenzen

ne a
es a
sc fc for the power plant’s involvement in the steel industry s that provides its steel to
car manufacturers.
Once again, in this simplified example, the producer responsibility approach (equation
(9)) ignores supplying industries, workers and investors who are located upstream in the
supply chain, and who are ultimately supporting and facilitating production, and – one
could argue – should therefore be partly responsible. On the other hand, the worker/inves-
tor responsibility approach (equation (10)) fails to account for sectors that form part of the
industrial metabolism but require no (or realistically, few) primary inputs. In fact, even
though enabling car manufacturing, the steel sector is not responsible for any waste at
all, neither under producer nor under worker/investor responsibility.

5.1. Upstream Responsibility Shared between Consumers and Producers

Now we introduce a new formulation that divides the responsibility for impacts amongst
all agents throughout demand and supply chains in a way that reflects their contribution to
the production process. We make a given sector k accountable for its final demand plus a
fraction (1
a) of its intermediate output. We call this fraction of xk the accountable
output, y(ka) ,

y(ka) ¼ yk þ (1
a)(xk
yk ) (11)

The responsibility for the remaining fraction a of the sector’s intermediate output is
assigned to its intermediate downstream users. Furthermore, the accountable output of a
given sector k includes the parts of xk associated with both the industry k as well as the
final consumers of sector k. Here, the final consumers are made responsible for a fraction
b of the final demand byk , while the rest of the accountable output is assigned to industry k.
In this way the responsibility for the output of sector k, is distributed as follows:
8
>
> byk ! assigned to final consumers of sector k
>þ
>
<
xk ¼ (1
b)yk þ (1
a)(xk
yk ) ! assigned to industry k (12)
>
>
>þ
>
: P
a(xk
yk ) ¼ a j akj xj ! assigned to sectors j downstream from k

The same responsibility distribution is applied to the gross output xj of each sector j. That
is, each industry j buying from k is made responsible for k’s impact due to a fraction
aakj ½(1
b)yj þ (1
a)(xj
yj ) of xk ; the final consumers purchasing from k are
assigned the fraction aakj byj , and the rest is distributed amongst the sectors that purchase
from j:
8
>
> aakj byj ! assigned to final consumers of sector j
>
>
<þ  
aakj xj ¼ aakj (1
b)yj þ (1
a)(xj
yj ) ! assigned to industry j
>
>
>
> þ P
:
aakj a(xj
yj ) ¼ a2 akj i a ji xi ! assigned to sectors i downstream from j
(13)
 Shared Producer and Consumer Responsibility 375

The same distribution process repeats itself as we move down the demand chain. As a
result, the responsibility of a given sector i for the production of an upstream sector k is
given by

X 
dki þ aaki þ a2 j
a kj a ji þ . . . y(i a) ¼ l(kia) y(i a) (14)

where dki ¼ 1 if k ¼ i and 0 otherwise,12 and where L(a) ¼ (I
aA)
1 . The quantity in
equation (14) includes the responsibility allocated to industry i,

l(kia) ½(1
b)yi þ (1
a)(xi
yi ) (15)

and that allocated to its final consumers,

l(kia) byi (16)

Summing over all the supply chain, the upstream shared responsibility of a given sector
i is:

X 
Fi(a) ¼ k
fk l(kia) y(i a) (17)

and the total upstream responsibility for production impacts F can be written as:
X X
X 
(a )
F¼ F (a ) ¼
i i i j
fj l(jia) y(i a) ¼ f 0 L y(a) (18)

The shape of this formulation is similar to that in equation (2), with L and y
replaced by

L(a) ¼ (I
aA)
1 and y(a) ¼ by þ (1
b)y þ (1
a)(x
y) (19)

The parameters a (the producer responsibility share) and b (the consumer respon-
sibility share) are numbers between 0 and 1. At the limit a ¼ 0, where L(a) ¼ I
and y(a) ¼ x, we recover the producer responsibility formulation F ¼ f 0 x. Meanwhile,
at the other extreme where a ¼ 1, we have L(a) ¼ L and y(a) ¼ y. Allocating the
sector’s responsibility solely to the final consumers by setting b ¼ 1, we recover
the consumer responsibility formulation F ¼ f 0 Ly.
The introduction of the responsibility share a allows for two important properties: first,
the upstream responsibility for a given impact decreases by a factor a each time we move
downstream in the demand chain. This is an interesting feature since it seems logical to
assume that the further a receiving sector is located from the producer of the impact,
the less influence it has over that impact. A discussion on the choices of a and b, and
the implications for policy can be found in the next section.
Another important implication of the responsibility share a is that it allows for additi-
vity, that is the upstream responsibilities for a given impact are shared amongst all down-
stream agents in the demand chain without incurring double counting: the sum over all
376 B. Gallego & M. Lenzen

contributions of sectors k to the responsibility for the impact of an industry i is equal to the
total impact of that industry:

X
Fi(a) ¼ f l (a ) ½ y k
k i ik
þ (1
a)(xk
yk )
X
¼ f l (a ) ½ x k
k i ik

a(xk
yk )
X
X 
(a )
¼ k
f i l ik x k
a j
a kj x j (20)
X (a )
X
¼ fl
k i ik j
½I
aAkj xj
X
¼ f d x ¼ fi xi ¼ Fi(p)
j i ij j

Returning to our three-sector economy in Figure 1, and using the shared responsibility
framework, the CO2 emissions F ¼ fe xe from the generation of thermal electricity are
distributed amongst agents of the economy according to (see Figure 2)

F ¼ Fe(a) þ Fs(a) þ Fc(a) (21)

Figure 2. Downstream re-allocation of responsibility for an initial impact fexe as upstream

responsibilities to intermediate and final consumers (FD ¼ final demand)
 Shared Producer and Consumer Responsibility 377

The responsibility Fe(a) assigned to the electricity sector is proportional to its accountable
output:

Fe(a) ¼ fe l(eea) y(ea) ¼ fe ½ye þ (1
a)(xe
ye ) (22)

Of this responsibility, bfe ye rests with the final consumers of electricity, and the remainder
with the power plant operators.
The CO2 responsibility Fs(a) assigned to the steel industry for its demand of electricity is
proportional to a fraction aaes of its accountable output. In the absence of final demand,
the accountable output is just a fraction (1
a) of total output xs :

Fs(a) ¼ fe l(esa) y(sa) ¼ fe aaes (1
a)xs (23)

Note that, since ys ¼ 0, this sector is completely ignored in both the producer responsibil-
ity (a ¼ 0) as well as in the consumer responsibility (a ¼ 1) approach. Finally, the respon-
sibility Fc(a) assigned to the car manufacturing sector consists of two parts: one part is
associated with its direct purchase of electricity and is proportional to a fraction aaec of
its accountable output (see the second layer in Figure 2). The other part is associated
with the purchase of electricity by the steel sector, which then delivers steel to the car
industry. It is therefore proportional to a fraction a2 aes asc of the accountable output
(see the third layer in Figure 2):
 
Fc(a) ¼ fe l(eca) y(ca) ¼ fe aaec þ a2 aes asc yc (24)

Once again, portions b of the responsibility for final demands yc are assigned to final con-
sumers, with the remainder resting with the car manufacturers.
Thus, upstream responsibility is consistently and consecutively either allocated to
industries and consumers, or passed on downstream to the following layer of intermediate
demand. In our simplified example, this process stops after two rounds, with no intermedi-
ate demand left to be passed on, i.e. a2 fe aes asc (xc
yc ) ¼ 0 in Figure 2. The total impact
fe xe has been exhaustively distributed to producing and consuming agents.

5.2. Downstream Responsibility Shared between Workers-investors and Producers

The same logic can be applied to responsibility for impacts being re-allocated to providers
of primary inputs and industries, by using the Ghosh model. We make a given sector k
accountable for its primary inputs plus a fraction (1
a
) of its intermediate input. We
call this fraction of xk the accountable input, n(ka
) ,

v(ka
) ¼ vk þ (1
a
)(xk
vk ) (25)

Of this quantity, workers-investors are made responsible for a fraction b
of primary inputs
b
vk , while the rest of the accountable input is assigned to industry k. The responsibility for
the remaining fraction a
of the sector’s intermediate input is assigned to its intermediate
upstream suppliers.
378 B. Gallego & M. Lenzen

Applying this responsibility distribution to all the gross inputs throughout the supply
chain in a similar fashion as in the upstream formulation, we obtain that the responsibility
of a given sector i for (its association with) the production of a downstream sector k is
given by

(a
)
v(i a
)
lik (26)
(a
)
where L̄ ¼ (I
a
Ā)
1 . The quantity in equation (26) includes the responsibility
allocated to industry i,

(a
)
½(1
b
)vi þ (1
a
)(xi
vi )
lik (27)

and that allocated to the workers-investors to industry i,

(a
)
b
vi
lik (28)

Summing over all the demand chain, the downstream shared responsibility of a given
sector i is:

X 
(a
)
F
i ¼ n(i a
)
l(a
) f (29)
k ik k

and the total downstream responsibility for production impacts F can be written as:
X X
X 
F¼
(a
)
F
¼ v(a
)
l(a
) fj ¼ (v(a
) )0 L̄(a
) f (30)
i i i i j ij

The shape of this formulation is similar to that in equation (3), with L̄ and v replaced by

(a
)
L̄ ¼ (I
a
Ā)
1 and v(a
) ¼ b
v þ (1
b
)v þ (1
a
)(x
v) (31)

Once again, the parameters a
(the producer responsibility share) and b
(the worker/
investor responsibility share) are numbers between 0 and 1. At the limit a
¼ 0, where
(a
)
L̄ ¼ I and v(a
) ¼ x, we recover the producer responsibility formulation F ¼ x0 f. If a
¼
(a
)
1 we have L̄ ¼ L̄ and v(a
) ¼ v. Allocating the sector’s responsibility solely to the
primary inputs by setting b
¼ 1, we recover the worker/investor responsibility formu-
lation F ¼ v0 L̄f.
As before, the downstream responsibility for a given impact decreases by a factor a

each time we move upstream in the supply chain, and it is shared amongst all upstream
agents in the supply chain without incurring double counting. Returning to our three-
sector economy in Figure 1, and using the shared responsibility framework, the waste
impact F
¼ xc fc resulting from manufacturing cars is distributed amongst agents of the
economy according to (see Figure 3)

(a
) (a
) (a
)

F
¼ F
e þ F
s þ F
c (32)
 Shared Producer and Consumer Responsibility 379

Figure 3. Upstream re-allocation of responsibility for an initial impact fcxc as downstream

responsibilities to intermediate and workers-investors (PI ¼ primary inputs)
(a
)
The responsibility F
c assigned to the car manufacturing sector is proportional to its
(a
)
accountable input nc :

(a
)
F
c ¼ ½vc þ (1
a
)(xc
vc ) fc (33)

Of this responsibility, b
nc fc rests with workers-investors of car manufacturing, and the

(a
)
remainder with the manufacturing plant operators. The responsibility F
s assigned to
the steel industry for its supply to the car industry is proportional to a fraction a
a
sc of
its accountable input n(sa
) . In the absence of primary inputs, the accountable input is just
a fraction (1
a
) of total input xs :

(a
)
F
s ¼ (1
a
)xs a
a
sc fc (34)

Note that, since ns ¼ 0, this sector is completely ignored in both the producer respon-
sibility (a
¼ 0) as well as in the worker/investor responsibility (a
¼ 1) approach.
(a
)
Finally, the responsibility F
e assigned to the electricity sector consists of two parts
One part is associated with its direct delivery to car manufacturers and is proportional
380 B. Gallego & M. Lenzen

to a fraction a
a
ec of its accountable input n(ea
) (see the second layer in Figure 3). The other
part is associated with the purchase of electricity by the steel sector, which then delivers
steel to the car industry. It is therefore proportional to a fraction a
2 a
es a
sc of the accoun-
table input n(ea
) (see the third layer in Figure 3):

(a
)  
F
e ¼ ve a
a
ec þ a
2 a
es a
sc fc (35)

Once again, portions b
of the responsibility for primary inputs ne are assigned to workers-
investors, with the remainder resting with power plant operators.
Thus, downstream responsibility is consistently and consecutively either allocated to
industries and workers-investors, or passed on upstream to the following layer of inter-
mediate supply. In our simplified example, this process stops after two rounds, with no
intermediate supply left to be passed on, i.e. (xe
ne )a
2 a
es a
sc fc ¼ 0 in Figure 3. The
total impact xc fc has been exhaustively distributed to producing and working/investing
agents.
The formulations and additivity features of all frameworks dealt with in this article,
using both the Leontief and the Ghosh model, are summarised in Table 2.

Table 2. A summary of frameworks that can be used for assigning producer and consumer
responsibility

Approach Upstream or Leontief model Downstream or Ghosh model

Producer Assigns responsibility to industries Assigns responsibility to industries

responsibility according P to their own total output. according toPtheir own total input.
Fj ¼ fj xj ; j Fj ¼ F Fj ¼ xj fj ; j Fj ¼ F
Consumer/ Assigns responsibility to consumers Assigns responsibility to workers-
worker- according
P to theirPfinal demand. investors according to their
investor Fj ¼ k fk lkj yj ; j Fj ¼ F primaryP inputs. P
responsibility Fj ¼ vj k
ljk fk ; j Fj ¼ F
Total flow Assigns responsibility to industries Assigns responsibility to industries
according to the total upstream according to the total downstream
production requirement of their total utilisation requirement of their
outputP(non-additive).P total input
P (non-additive). P
Fj ¼ k fk (lkj =ljj )xj ; j Fj . F Fj ¼ xj k (
ljk =
ljj )fk ; j Fj . F
Hypothetical Assigns responsibility to industries Assigns responsibility to industries
extraction according to the residual total output according to the residual total
after hypothetically extracting them output after hypothetically
from Pthe economy (non-additive).
P extracting them from the economy
Fj ¼ k fk (xk
x½j
k ); j Fj . F (non-additive).
P P
Fj ¼ k (xk
x
½j
k )fk ; j Fj . F

Shared Assigns responsibility to industries Assigns responsibility to industries

responsibility and consumers according to a and workers-investors according
prescribed combination of to a prescribed combination of
intermediate
P (aoutput and final demand. intermediate Pinput and its primary (a
)
F
P j ¼ k f k l kj
)
½y j þ (1
a)(xj
yj ) inputs.
P Fj ¼
)(xj
vj )
ljk fk
k ½vj þ (1
a
j Fj ¼ F F
j j ¼ F
 Shared Producer and Consumer Responsibility 381

5.3. Transaction-specific Responsibility Shares

So far it has been assumed that the responsibility passed from a sector i to a sector j in the
next production layer is a fixed fraction a (or a
) of the given inter-industrial transaction,
aaij xj (or a
xj a
ji ). Analogously, the fraction of final demand (or primary input) of a sector
that is assigned to the final consumers (or workers/investors) is the same amount b (or b
)
regardless of the economic sector. A natural extension of the formulation presented here is
to consider a different producer responsibility share aij (or a
ij ) for each transaction
between industries i and j, as well as a different consumer/worker-investor responsibility
share bi (or b
i ) for each sector i. If implemented in practice, this would mean that each
transaction featured a different responsibility share specific to the pair of supplier and
recipient of the respective commodity, based on the nature of that transaction. It is even
conceivable for the tensor a to depend on both the transaction as well as the type of
impact, although this might prove to lead to a framework too complex to be administered.
In the tensorial formulation of upstream shared responsibility, the Leontief inverse L(a)
and accountable output y(a) are given by:

L(a) ¼ (I
a#A)
1 and y(a) ¼ b#y þ (s
b)#y þ ½(s
a)#Ts (36)

where a is the producer responsibility share tensor of rank n, b is an n-dimensional vector

of consumer/worker-investor responsibility shares, and s is the n-dimensional summation
vector with all elements equal to one. The symbol # represents the tensorial product, that is
{A#B}ij ¼ aij bij . Each ith element of the accountable output vector contains the final
demand of sector i, plus the sum over the accountable fractions of all the intermediate
outputs of industry i,
X
y(i a) ¼ yi þ {½(s
a)#Ts}i ¼ yi þ (1
aij )aij x j (37)
j

(a
)
Similarly, the tensorial forms of the Ghosh inverse L
and accountable input v(a
) in the
downstream shared responsibility framework are:

(a
)


1 and v(a
) ¼ b
#v þ (s
b
)#v þ ½(s
a
)#T0 s
L
¼ (I
a
#A) (38)

Note that, as in the case of constant responsibility shares, the tensorial formulation by con-
struction avoids the problem of double counting, and the proof of additivity in equation
(20) still holds.

5.4. Applications and Policy Relevance

Having developed a quantitative method for consistently dividing and assigning responsi-
bility for impacts to both producers and consumers, it is of course necessary to consider
how this method might be applied in real life. One example application would be for
Corporate Sustainability Reporting (CSR), where as a result of compliance obligations,
or pressure from consumers, insurers, employees, stakeholders or the media, organisations
disclose their overall impacts on the environment, society and economy. Similarly, inves-
tors’ requests for transparency on ethical standards have driven socially responsible
382 B. Gallego & M. Lenzen

investment (SRI) in dealing with corporate performance, and managing bodies of SRI
portfolios screen companies not only as to their own products, but also as to their upstream
and downstream involvement in harmful or unethical production practices. In the opposite
direction, companies themselves, especially those with a high proportion of purchased
inputs, are the drivers behind supply-chain thinking in terms of financial risk from
upstream and downstream environmental impacts.
Within CSR, the need for assigning responsibility for impacts in a standardised and
accepted way has been expressed recently.13 One problem within the Sustainability
Reporting Guidelines set out by the Global Reporting Initiative (2002) is how to define
appropriate system boundaries for sustainability assessments (Dey et al., 2002). In
particular,

reported information should be complete in relation to the operational boundaries of

the reporting organisation, in other words, the range of entities for which the report-
ing organisation gathers data. These boundaries should be selected with consider-
ation of the economic, environmental, and social impacts of the organisation. . . .
In some cases, the most appropriate boundaries for meeting the expectations out-
lined by other reporting principles may extend beyond traditional financial reporting
boundaries. . . . The process for determining a complete scope may include, for
example, the results of lifecycle analysis of products or services and assessment
of the full range of direct and indirect social or ecological impacts of the reporting
organisation. (Global Reporting Initiative, 2002, p. 26)

Input –output analysis has been suggested as a suitable and standard technique to deal with
these boundary problems (Global Environmental Management Initiative, 2002; Suh et al.,
2004).
Using the technique outlined in the previous sections, a sustainability report for an
organisation could be compiled using equations (18) and (31), based on the organisation’s
financial expenditure and sales accounts (providing the data for intermediate demand x – y
and supply x – v, see equations (17) and (30)), sales to final demand y, and wage, tax and
capital bills v. Given that this information is readily available in financial accounting
systems, enumerating equations (18) and (31) should not pose unsurmountable barriers.
Unlike a total-flow-type method, using a shared responsibility formulation for producer
impacts such as developed in this work avoids double counting.
Extending responsibility beyond producers’ and consumers’ premises into the
supply-chain network leads to decision-makers not only targeting on-site impacts, but
also indirect impacts. Abatement would be achieved both through on-site measures and
procurement policies, and could be streamlined and prioritised using input –
output-related techniques such as structural path analysis (for a corporate example see
Lenzen et al., 2003). These techniques would assist in answering vital management
questions such as ‘Which of my suppliers has the greatest impact on my environmental
performance?’ [and] “How far down the supply chain do I go?’’ (CIPS, 1993, p. 19).
A problem is of course determining the values for a, b, a
and b
. It is possible for these
values to result from a legislative process, involving dialogue between the government,
consumers, organisations participating in various supply chains, and other stakeholders.
It is difficult to say whether there would be ‘optimal’ values. However, given that ‘differ-
ent (groups of) economic agents occupy overlapping spheres of social, economic and
 Shared Producer and Consumer Responsibility 383

political influence . . ., a quantification of [these] influences . . . would be a precondition for

an allocation of environmental responsibilities to specific actors’ (Spangenberg and Lorek,
2002, p. 128). Accordingly, choices of responsibility shares should ideally reflect suppli-
ers’ and recipients’ financial control, innovation potential and business relations, as well as
their influence over production processes and their options to substitute suppliers or buyers
(compare Global Reporting Initiative, 2002, p. 26).
An interesting approach that can be used to define criteria for setting responsibility
shares is that of Cerin and Karlson (2002), who suggest rearranging corporate legal own-
ership of environmental impact property rights in order to provide incentives for extended
producer responsibility. These authors argue that even if major environmental impacts
occur outside producers’ legal boundary, they often possess the best competence to
reduce those impacts, but they are neither obliged nor stimulated to abate because they
do not own property rights over upstream and downstream stages of the products’ life-
cycle (stewardship). In this case, if other agents (for example consumers) wanted to
abate they would first have to acquire the necessary information, thus incurring a trans-
action cost that could be avoided if the agent with the best abatement competence (the
producer) was given the property rights over the respective part of the product life cycle.
A rearrangement of legal rights that minimises transaction cost would lead to a higher
degree of vertical integration, but it would also alter the structure of risk from supply-
chain environmental impacts. Note that a tensorial formulation of shared responsibility
would be able to describe re-arrangements of property rights across product life cycles.
More concrete suggestions on setting the weights – possibly involving values – would
of course be desirable. However, considering the fact that concepts such as extended
producer responsibility or shared responsibility are still in their infancy, it is not surprising
that more quantitative suggestions for the weights cannot be made at this time.

6. Conclusions and Outlook

Using two input – output models, modified from the original Leontief and Ghosh models,
we have shown that it is possible to ex-post divide impacts into mutually exclusive and
collectively exhaustive portions that can be assigned to producers and consumers as
shared responsibilities. This method has applications in (corporate) sustainability
reporting (CSR) where it can reflect in a quantitative way the joint responsibility of
interdependent actors in a complex supply-chain system.
Quantitative CSR is still in its early stages, and certainly needs to be further dissemi-
nated, discussed and operationalised before techniques such as the shared responsibility
approach outlined in this article can enter practice. Nevertheless, supply-chain issues
for CSR are becoming increasingly topical, which is reflected for example in a quote
by the World Business Council on Sustainable Development (2002, p. 55): ‘Current
reporting practices are often performed within the boundaries of the reporting organiz-
ation. In the coming years, it is likely that companies will increasingly report across the
value chain. This will represent a new challenge in terms of reporting on the upstream
(supplier related) issues linked to human rights, environmental and societal impacts,
and also of coping with the wider downstream (consumer related) impact of products
and services’.
Once the need for comprehensive and quantitative sustainability reporting reaches
policy and legislation, organisations will ask for guidelines and standards to prescribe
384 B. Gallego & M. Lenzen

consistent and fair procedures – such as the one proposed in this work – in order to assess
corporate responsibility. One aspect of consistency and fairness will be that supply-chain
relations are appropriately represented in reporting procedures, to reflect the integration
and interdependence of producers and consumers in a modern economy, and hence
their shared responsibility for impacts that their activities entail.

Notes
1
Wolvén (1991); Mélanie et al. (1994); Parikh and Painuly (1994); Parikh (1996); Hamilton and Turton
(1999, 2002); Lenzen and Smith (2000); Wier et al. (2001).
2
See CIPS (1993); Barry (1996); Carillion (2001); Ford Motor Company of Australia Limited (2003,
pp. 9-10); Norwich Union Central Services (2003); Toyota Motor Corporation (2003, p. 5).
3
In the following, we refer to intermediate demand as between industries, while gross output – comprising
intermediate and final demand – is that of sectors. We will denote scalars in italic not bold, vectors in italic
and bold, and matrices in bold not italic.
4
Since the Ghosh model is usually interpreted as a price model, and not as a quantity model, a qualification
is needed as to how changes in the direct sales matrix A
have to be interpreted. Let as assume an increase
(from one year to another) in the primary inputs into the industry i that generates an increase in the price of
final demands without any changes in physical flows. In our interpretation, more responsibility would be
assigned to primary inputs into i in the second year than in the first, on the basis of higher monetary inputs
utilised in downstream production. However, in the absence of input– output tables in physical units, there
is simply no other means of distributing this responsibility.
5
For the sake of simplicity, we will refer to these citizens in the following as ‘consumers’ and ‘workers-
investors’ (i.e. knowing that we have neglected the public sector and foreign countries).
6
Compare Miller and Blair’s (1985, p. 320) interpretation of reductions in primary inputs leading to
reductions in gross output.
7
See also Tiebout (1967) and Billings and Katz (1982) on the separation technique.
8
Instead of using one row and one column, a company can also be represented as an entire subsystem of
intra-company processes (see Farag, 1967; Polenske, 1997; Lin and Polenske, 1998; Albino et al.,
2002; Heijungs and Suh, 2002).
9
Milana (1985, p. 289) writes that ‘the sum of total output and primary input requirements for gross output
. . . leads to a double counting procedure because some indirect requirements of one industry are indirect
requirements of other industries which use part of that output of that industry as an intermediate input.’
Similarly, Heimler (1991, p. 263) acknowledges that ‘gross output of different industries cannot be
summed because of double counting of intermediate consumption.’
10
Since in our example, car manufacturing does not supply any other industry, there is no upstream respon-
sibility for any agent in terms of waste, only a sharing (b) between car manufacturers and buyers of cars.
11
Since in our example, the electricity industry has no intermediate suppliers, there are no downstream
impacts for any agent in terms of CO2 emissions, only a sharing (b) between power plant operators and
electricity consumers.
12
Note that dki is the element (k, i) of the identity matrix I.
13
The Global Reporting Initiative (2002, p. 4) states: ‘Inconsistent reporting approaches developed by
business, government, and civil society continue to appear. At the same time, many other organisations
wonder how best to engage in reporting. As diverse groups seek information, the multiplicity of infor-
mation requests gives rise to redundancy, inefficiency, and frustration.’ On the standardisation problem,
see also Olsthoorn et al. (2001).

References
ABARE (1999) Australian energy consumption and production, Research Report, Catalogue No: 99.4 Common-
wealth of Australia. (Canberra: Australian Bureau of Agricultural and Resource Economics).
ABS (1996) Labour force survey, Australia, Revisions to historical ANZSIC industry data, Information Paper,
Catalogue No: 6259.0 (Canberra: Australian Bureau of Statistics).
 Shared Producer and Consumer Responsibility 385

Albino, V., Izzo, C. and Kühtz, S. (2002) Input–output models for the analysis of a local/global supply chain,
International Journal of Production Economics, 78, pp. 119– 131.
Barry, A. (1996) Buyers start to spread the ‘green’ message, Purchasing & Supply Management, February, p. 21.
Bastianoni, S., Pulselli, F.M. and Tiezzi, E. (2004) The problem of assigning responsibility for greenhouse gas
emissions, Ecological Economics, 49, pp. 253 –257.
Billings, R.B. and Katz, J.L. (1982) A technique to obtain accurate impact multipliers for individual firms by
means of existing input–output models, Environment and Planning A, 14, pp. 739–744.
Carillion (2001) Sustainable supply chain management, Sustainability Report (Wolverhampton: Carillion).
[Available at http://www.carillionplc.com/sustain2001/sustainable_supply_chain_managem.htm.]
Cerin, P. and Karlson, L. (2002) Business incentives for sustainability: a property rights approach, Ecological
Economics, 40, pp. 13 –22.
CIPS (1993) Green purchasing: are you buying into the environment?, Report, Catalogue No: 03097242 (Easton
on the Hill, Stamford, UK: Chartered Institute of Purchasing and Supply).
CIPS (1999) Ethical business practices in purchasing and supply, Positions on Practice (Easton on the Hill,
Stamford, UK: Chartered Institute of Purchasing and Supply). [Available at http://www.epolitix.com/
data/companies/images/Companies/Chartered-Institute-of-Purchasing-and-Supply/ethics.pdf.]
CIPS (2000) Environmental purchasing and supply management summary, Policies and Positions (Easton on
the Hill, Stamford, UK: Chartered Institute of Purchasing and Supply). [Available at http://www.epolitix.
com/data/companies/images/Companies/Chartered-Institute-of-Purchasing-and-Supply/environmental.
pdf.]
Danish EPA (1998) Denmark’s second national communication on climate change submitted under the
UN FCCC (Copenhagen: Danish Environmental Protection Agency). [Available at http://www.mst.dk/
homepage/.]
Dey, C., Lenzen, M., Foran, B. and Bilek, M. (2002) Addressing boundary issues in the Global Reporting
Initiative, Comments on the Draft 2002 Sustainability Reporting Guidelines (Sydney: The University
of Sydney, CSIRO Sustainable Ecosystems). [Available at http://www.globalreporting.org/feedback/
PublicComments2002/UniversityofSydneyCSIRO.pdf.]
Dietzenbacher, E. (1997) In vindication of the Ghosh model: a reinterpretation as a price model, Journal of
Regional Science, 37, pp. 629– 651.
Dietzenbacher, E., van der Linden, J.A. and Steenge, A.E. (1993) The regional extraction method: Applications to
the European Community, Economic Systems Research, 5, pp. 185–206.
EPA NSW (2003) Extended producer responsibility priority statement, Consultation Paper, Catalogue No: 2003/
10 (Sydney: Environment Protection Authority of New South Wales).
Farag, S.M. (1967) Input–Output Analysis: Applications to Business and Accounting (Urbana: Center for
International Education and Research in Accounting).
Ferng, J.-J. (2003) Allocating the responsibility of CO2 over-emissions from the perspectives of benefit principle
and ecological deficit, Ecological Economics, 46, pp. 121–141.
Ford Motor Company of Australia Limited (2003) Policies and organisation, Broadmeadows Public Environment
Report 2002/2003 (Campbellfield, Australia: Ford Motor Company of Australia Limited). [Available at
www.ford.com.au/resources/pdf/environment/7.pdf.]
Ghosh, A. (1958) Input–output approach in an allocation system, Economica, 25, pp. 58 –64.
Global Environmental Management Initiative (2002) How to Tally Emissions (Washington, DC: Global
Environmental Management Initiative).
Global Reporting Initiative (2002) Sustainability Reporting Guidelines (Boston: Global Reporting Initiative).
Hamilton, C. and Turton, H. (1999) Population policy and environmental degradation: sources and trends in
greenhouse gas emissions, People and Place, 7, pp. 42–62.
Hamilton, C. and Turton, H. (2002) Determinants of emissions growth in OECD countries, Energy Policy, 30,
pp. 63 –71.
Heijungs, R. and Suh, S. (2002) The Computational Structure of Life Cycle Assessment (Dordrecht: Kluwer).
Heimler, A. (1991) Linkages and vertical integration in the Chinese economy, Review of Economics and
Statistics, 73, pp. 261 –267.
Imura, H. and Moriguchi, Y. (1995). Economic interdependence and eco–balance: accounting for the flow of
environmental loads associated with trade, in: S. Murai (Ed.) Toward Global Planning of Sustainable
Use of the Earth, pp. 189 –208 (Amsterdam: Elsevier).
Jeong, K. (1982) Direct and indirect requirements: a correct economic interpretation of the Hawkins–Simons
conditions, Journal of Macreconomics, 4, pp. 349–356.
386 B. Gallego & M. Lenzen

Jeong, K. (1984) The relation between different notions of direct and indirect input requirements, Journal of
Macreconomics, 6, pp. 473–476.
Korhonen, J. (2002) Two paths to industrial ecology: applying the product-based and geographical approaches,
Journal of Environmental Planning and Management, 45, pp. 39–57.
Lenzen, M., Lundie, S., Bransgrove, G., Charet, L. and Sack, F. (2003) Assessing the ecological footprint of a
large metropolitan water supplier – lessons for water management and planning towards sustainability,
Journal of Environmental Planning and Management, 46, pp. 113– 141.
Lenzen, M., Pade, L.-L. and Munksgaard, J. (2004) CO2 multipliers in multi-region input–output models,
Economic Systems Research, 16, pp. 391 –412.
Lenzen, M. and Smith, S. (2000) Teaching responsibility for climate change: three neglected issues, Australian
Journal of Environmental Education, 15/16, pp. 69 –78.
Lin, X. and Polenske, K.R. (1998) Input–output modeling of production processes for business management,
Structural Change and Economic Dynamics, 9, pp. 205–226.
Lloyd, M. (1994) How green are my suppliers? – Buying environmental risk, Purchasing & Supply Management,
October, p. 36.
Maxie, E. (1994) Supplier performance and the environment. Paper at the IEEE International Symposium on
Electronics & the Environment, San Francisco.
Mélanie, J., Phillips, B. and Tormey, B. (1994) An international comparison of factors affecting carbon dioxide
emissions, Australian Commodities, 1, pp. 468–483.
Milana, C. (1985) Direct and indirect requirements for gross output in input–output systems, Metroeconomica,
37, pp. 283–292.
Miller, R.E. and Blair, P.D. (1985) Input–Output Analysis: Foundations and Extensions (Englewood Cliffs, NJ:
Prentice-Hall).
Miyazawa, K. (1966) Internal and external matrix multipliers in the input–output model, Hitotsubashi Journal of
Economics, 7, pp. 38– 55.
Munksgaard, J. and Pedersen, K.A. (2000) CO2 accounts for open economies: producer or consumer responsibil-
ity?, Energy Policy, 29, pp. 327–334.
Norwich Union Central Services (2003) Suppliers, Corporate Social Responsibility Report (Norwich: Norwich
Union Central Services). [Available at http://www.aviva.com/responsibility/2003report/reports/
nu_cent_suppliers.htm.]
OECD (2001) Extended Producer Responsibility: A Guidance Manual for Governments (Paris: Organisation for
Economic Co-operation and Development).
Olsthoorn, X., Tyteca, D., Wehrmeyer, W. and Wagner, M. (2001) Environmental indicators for business:
a review of the literature and standardisation methods, Journal of Cleaner Production, 9, pp.
453– 463.
Parikh, J. (1996) Consumption patterns: the driving force of environmental stress, in: P.H. May and R. Serôa da
Motta (Eds) Pricing the Planet, pp. 39– 48 (New York: Columbia University Press).
Parikh, J.K. and Painuly, J.P. (1994) Population, consumption patterns and climate change: a socioeconomic
perspective from the south, Ambio, 23, pp. 434 –437.
Polenske, K.R. (1997) Linked system of enterprise, regional and national input–output accounts for policy
analysis, in: M. Chatterji (Ed.) Regional Science: Perspectives for the Future, pp. 26–42 (Basingstoke,
UK: Macmillan).
Princen, T. (1999) Consumption and environment: some conceptual issues, Ecological Economics, 31, pp.
347– 363.
Proops, J.L.R., Faber, M. and Wagenhals, G. (1993) Reducing CO2 Emissions (Berlin: Springer-Verlag).
Spangenberg, J.H. and Lorek, S. (2002) Environmentally sustainable household consumption: from aggregate
environmental pressures to priority fields of action, Ecological Economics, 43, pp. 127 –140.
Strassert, G. (1968) Zur Bestimmung strategischer Sektoren mit Hilfe von Input–Output Tabellen, Jahrbücher
für Nationalökonomie und Statistik, 182, pp. 211–215.
Subak, S. (1995) Methane embodied in the international trade of commodities, Global Environmental Change, 5,
pp. 433 –446.
Suh, S., Lenzen, M., Treloar, G.J., Hondo, H., Horvath, A., Huppes, G., Jolliet, O., Klann, U., Krewitt, W.,
Moriguchi, Y., Munksgaard, J. and Norris, G. (2004) System boundary selection in life–cycle inventories,
Environmental Science & Technology, 38, pp. 657–664.
Szyrmer, J.M. (1992) Input–output coefficients and multipliers from a total-flow perspective, Environment and
Planning A, 24, pp. 921–937.
 Shared Producer and Consumer Responsibility 387

Szyrmer, J.M. and Walker, R.T. (1983) Interregional total flow: a concept and application to a U.S. input–output
model, Review of Regional Studies, 13, pp. 12–21.
Task Force on National Greenhouse Gas Inventories (1996) Revised 1996 IPCC Guidelines for National Green-
house Gas Inventories – Reporting Instructions, Volume 1 (Tokyo: Institute for Global Environmental
Strategies, Intergovernmental Panel on Climate Change – IPCC National Greenhouse Gas Inventories
Programme). [Available at http://www.ipcc-nggip.iges.or.jp/public/gl/invs4.htm.]
Tiebout, C.M. (1967) Input–output and the firm: a technique for using national and regional tables, Review of
Economics and Statistics, 49, pp. 260– 262.
Toyota Motor Corporation (2003) 2003 Environmental Report (Port Melbourne: Toyota Motor Corporation).
[Available at http://environment.toyota.com.au/TWP/Upload/Media/144.pdf.]
Weber, C. (1998) Zerlegung kumulierter Energieaufwendungen als Instrument der umweltorientierten Verflech-
tungsanalyse, Jahrbücher für Nationalökonomie und Statistik, 217, pp. 273–291.
Wier, M., Lenzen, M., Munksgaard, J. and Smed, S. (2001) Environmental effects of household consumption
pattern and lifestyle, Economic Systems Research, 13, pp. 259 –274.
Wolvén, L.-E. (1991) Life–styles and energy consumption, Energy, 16, pp. 959–963.
WBCSD (2002) Sustainable development reporting: striking the balance, Report (Conches–Geneva: World
Business Council on Sustainable Development). [Available at http://www.wbcsd.ch/DocRoot/
GGFpsq8dGngT5K56sAur/20030106_sdrepot.pdf.]
WBCSD and WRI, World Resources Institute (2001) The greenhouse gas protocol, Report (Conches–Geneva:
World Business Council on Sustainable Development). [Available at http://www.ghgprotocol.org/stan
dard/ghg.pdf, http://www.ghgprotocol.org/.]
Wyckoff, A.W. and Roop, J.M. (1994) The embodiment of carbon in imports of manufactured products –
Implications for international agreements on greenhouse gas emissions, Energy Policy, 22, pp. 187 –194.

Appendix 1: Equivalence between the hypothetical extraction and the total flow
formulations
In the following, the hypothetical extraction and the total flow formulations are shown to
be equivalent for the Leontief model, but equivalence applies equally to the Ghosh model.
In the total flow method, the requirement of gross output xk of a given industry k associated
with the gross output of another sector j is given by:

lkj
TFkj ¼ xj (A1)
l jj

Without loss of generality, we can partition the direct requirements matrix A into nine
blocks in order to separate row and column j from the rest of the matrix as indicated in
equation (A2). The diagonal blocks of this partitioned matrix are square matrices of
dimensions (j
1)  (j
1), 1  1 and (n
j)  (n
j), where n represents the number
of industrial sectors.

1!(j
1) j (jþ1)!n
z}|{ z}|{ z}|{
2 3
A(11) A(12) A(13)
6 (21) 7
A ¼ 4A A(22) A(23) 5 (A2)
A(31) A(32) A(33)
388 B. Gallego & M. Lenzen

Using basic matrix algebra, we can compute the inverse of the partitioned matrix (I
A):

(I
A)
1 ¼
2
1
1
1
1
1
3
D(123)
C(113) C(123) D(231)
C(112) C(132) D(312) (A3)
6
1
1
1
1
1 7
4
C(223) C(213) D(123) D(231)
C(221) C(231) D(312) 5

1
1
1
1
1

C(332) C(312) D(123)
C(331) C(321) D(231) D(312)

where:

1
C(pqr) ¼ (I
A)(pq)
(I
A)(pr) (I
A)(rr) (I
A)(rq)

1
(A4)
D(pqr) ¼ C(ppr)
C(pqr) C(qqr) C(qpr)

with p, q and r representing the indices 1, 2 or 3, and (I
A)(pq) representing the block
( pq) of (I
A). Expression (A3) can be validated by proving that
(I
A)(I
A)
1 ¼ I, which implies

1
1
1
½(I
A)(pp) þ A(pq) C(qqr) C(qpr) þ A(pr) C(rrq) C(rpq) D(pqr) ¼I

1
1
1
(A5)
½
A(qp)
(I
A)(qq) C(qqr) C(qpr) þ A(qr) C(rrq) C(rpq) D(pqr) ¼0

for any p, q = p and r = p, q and with 0 being a matrix with all elements being zero. In
order to prove equation (A5), the following identities will be used:

1
1
C(rrq) ¼ {(I
A)(rr)
A(rq) (I
A)(qq) A(qr) }
1

1
1
¼ {A(rq) (I
A)(qq) ½(I
A)(qq) A(rq) (I
A)(rr)
A(qr) }
1

1
¼ {A(rq) (I
A)(qq)

1
1
 ½(I
A)(qq)
A(qr) (I
A)(rr) A(rq) A(rq) (I
A)(rr) }
1

1
1
¼ {A(rq) (I
A)(qq) C(qqr) A(rq) (I
A)(rr) }
1

1
1
1
¼ (I
A)(rr) A(rq) C(qqr) (I
A)(qq) A(rq) (A6)
 Shared Producer and Consumer Responsibility 389

1
C(rpq) ¼
½A(rp) þ A(rq) (I
A)(qq) A(qp) 

1
1
¼
A(rq) (I
A)(qq) ½(I
A)(qq) A(rq) A(rp) þ A(qp) 

1
1
¼
A(rq) (I
A)(qq) ½{(I
A)(qq)
A(qr) (I
A)(rr) A(rq)

1
1
þ A(qr) (I
A)(rr) A(rq) }A(rq) A(rp) þ A(qp) 

1
1
1
¼
A(rq) (I
A)(qq) ½{(I
A)(qq)
A(qr) (I
A)(rr) A(rq) }A(rq) A(rp)

1
þ {A(qp) þ A(qr) (I
A)(rr) A(rp) }

1
1
¼
A(rq) (I
A)(qq) ½C(qqr) A(rq) A(rp)
C(qpr)  (A7)

1
Substituting C(rrq) and C(rpq) by expressions (A6) and (A7) in equations (A5) we
obtain

1
½(I
A)(pp) þ A(pq) C(qqr) C(qpr)

1
1
1
1
þA(pr) {
(I
A)(rr) A(rp) þ (I
A)(rr) A(rq) C(qqr) C(qpr) }D(pqr)

1
1
1
¼ ½C(ppr)
C(pqr) C(qqr) C(qpr) D(pqr) ¼ D(pqr) D(pqr) ¼I
(A8)

1
½
A(qp)
(I
A)(qq) C(qqr) C(qpr)

1
1
1
1
þA(qr) {
(I
A)(rr) A(rp) þ (I
A)(rr) A(rq) C(qqr) C(qpr) }D(pqr)

1
1
1
¼ ½C(qpr)
C(qqr) C(qqr) C(qpr) D(pqr) ¼ ½C(qpr)
C(qpr) D(pqr) ¼0

Using equation (A3), each element TFkj can be written as follows:

8  (113) (123) 
lkj <
C C x for k , j
kj j
TFkj ¼ xj ¼  xj for k ¼ j (A9)
l jj :
C(331) C(321)  x for k . j
kj j

On the other hand, in the hypothetical extraction method, the fraction of gross output xk of
a given industry k assigned to an industrial sector j is given by

HEkj ¼ xk
x½j
k (A10)

where x½j
k is the hypothetical gross output of sector k in an economy without sector j.
Sector j is hypothetically extracted by zeroing row and column j in the direct require-
ments matrix, forming A½j , and by zeroing the sector’s final demand yj ¼ 0. In the
390 B. Gallego & M. Lenzen

trivial case k ¼ j, in which x½j

j ¼ 0, HEjj ¼ TFjj ¼ xj . In the case k = j, the hypothetical
gross output x½j
k can be rewritten as a function of the gross output vector x as follows:

X XX
x½j
k (k=j) ¼ (I
A½j )
1
ki yi ¼ (I
A½j )
1
ki (I
A)il xl : (A11)
i i l

We can re-express matrix A as the sum of A½j plus a matrix A½
j whose only non-zero
elements are the direct requirement coefficients in row and column j. Equation (A11)
then becomes

XX XX
x½j
k (k=j) ¼ (I
A½j )
1 ½j
ki (I
A )il xl
(I
A½j )
1 ½
j
ki ail xl
i l i l

¼ xk
((I
A½j )
1 A½
j )kj xj (A12)

Here we have made use of the fact that a½
j

il ¼ 0 for all l = j. Substituting expression
(A12) in the definition of HEkj we obtain

HEkj(k=j) ¼ ((I
A½j )
1 A½
j )kj xj (A13)

Using the same partitioning as in equation (A2), the inverse of I
A½j can be expressed as
the partitioned matrix

2
1
1 3
C(113) 0 C(113) A(13) (I
A(33) )
1
(I
A½j )
1 ¼ 4 0 1 0 5 (A14)

1
(331)
1
C(331) A(31) (I
A(11) )
1 0 C

Here, 0 represents a matrix with all elements being zero. It can be easily shown that
(I
A½j )
1 (I
A½j ) ¼ I, since:

1
1
1
C(ppr) ½(I
A)(pp)
A(pr) (I
A)(rr) A(rp)  ¼ C(ppr) C(ppr) ¼ I

1
1
1
(A15)
C(ppr) ½
A(pr) þ A(pr) (I
A)(rr) (I
A)(rr)  ¼ C(ppr) ½
A(pr) þ A(pr)  ¼ 0
 Shared Producer and Consumer Responsibility 391

Multiplying (I
A½j )
1 by the matrix A½
j results in:

(I
A½j )
1 A½
j

2
1
1 3
C(113) 0 C(113) A(13) (I
A(33) )
1
6 7
¼4 0 1 0 5

1
1
C(331) A(31) (I
A(11) )
1 0 C(331)
2 3
0 A(12) 0
6 (21) (22) 7
4A A A(23) 5
0 A(32) 0
2
1 3
0 C(113) ½A(12) þ A(13) (I
A(33) )
1 A(32)  0
6 (21) 7
¼ 4A A(22) A(23) 5

1
0 C(331) ½A(31) (I
A(11) )
1 A(12) þ A(32)  0
2
1 3
0
C(113) C(123) 0
6 7
¼ 4 A(21) A(22) A(23) 5 (A16)
(331)
1 (321)
0
C C 0

Therefore, each element HEkj can be written as follows:

8  (113) (123) 
( C x for k , j
<
C
>
xj for k ¼ j kj j
HEkj ¼   ¼ xj for k ¼ j
(I
A½j )
1 A½
j kj xj for k = j >: 
C(331) C(321)  x for k . j
kj j

(A17)

Comparing equations (A9) and (A17) we can conclude that HEkj ¼ TFkj .