Unclassified ENV/WKP(2018)4

Organisation de Coopération et de Développement Économiques

Organisation for Economic Co-operation and Development 16-Apr-2018
___________________________________________________________________________________________
_____________ English - Or. English
ENVIRONMENT DIRECTORATE
Unclassified
ENV/WKP(2018)4

THE MACROECONOMICS OF THE CIRCULAR ECONOMY TRANSITION: A CRITICAL REVIEW

OF MODELLING APPROACHES - ENVIRONMENT WORKING PAPER No. 130

by Andrew McCarthy, Rob Dellink, and Ruben Bibas (OECD)

OECD Working Papers should not be reported as representing the official views of the OECD or of its member
countries. The opinions expressed and arguments employed are those of the author(s).

Authorised for publication by Anthony Cox, Acting Director, Environment Directorate.

JEL Classification: C68, O13, Q53

Keywords: Circular economy, resource efficiency, natural resources, raw materials, general equilibrium model

OECD Environment Working Papers are available at www.oecd.org/environment/workingpapers.htm

English - Or. English

JT03430211

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the
delimitation of international frontiers and boundaries and to the name of any territory, city or area.
ENV/WKP(2018)4

OECD ENVIRONMENT WORKING PAPERS

OECD Working Papers should not be reported as representing the official views of the OECD or of its
member countries. The opinions expressed and arguments employed are those of the author(s).
Working Papers describe preliminary results or research in progress by the author(s) and are published to
stimulate discussion on a broad range of issues on which the OECD works.

This series is designed to make available to a wider readership selected studies on environmental issues
prepared for use within the OECD. Authorship is usually collective, but principal author(s) are named. The
papers are generally available only in their original language –English or French- with a summary in the
other language.

Comments on Working Papers are welcomed, and may be sent to:

OECD Environment Directorate 2 rue André-Pascal, 75775 Paris Cedex 16, France
or by email: [email protected]

------------------------------------------------------------------------------------------
OECD Environment Working Papers are published on
www.oecd.org/environment/workingpapers.htm
------------------------------------------------------------------------------------------

Note by Turkey: The information in this document with reference to "Cyprus" relates to the southern part
of the Island. There is no single authority representing both Turkish and Greek Cypriot people on the
Island. Turkey recognises the Turkish Republic of Northern Cyprus (TRNC). Until a lasting and equitable
solution is found within the context of the United Nations, Turkey shall preserve its position concerning
the “Cyprus issue”.
Note by all the European Union Member States of the OECD and the European Union: The Republic of
Cyprus is recognised by all members of the United Nations with the exception of Turkey. The information
in this document relates to the area under the effective control of the Government of the Republic of
Cyprus.
The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities.
The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem
and Israeli settlements in the West Bank under the terms of international law.

© OECD (2018)

You can copy, download or print OECD content for your own use, and you can include excerpts from
OECD publications, databases and multimedia products in your own documents, presentations, blogs,
websites and teaching materials, provided that suitable acknowledgment of OECD as source and copyright
owner is given. [email protected].

2
 ENV/WKP(2018)4

ABSTRACT

This paper reviews the existing literature on modelling the macroeconomic consequences of the
transition to a circular economy. It provides insights into the current state of the art on modelling policies
to improve resource efficiency and the transition to a circular economy by examining 24 modelling-based
assessments of a circular economy transition. Four key conclusions emerge from this literature. First, most
models find that a transition to a more circular economy – with an associated reduction in resource
extraction and waste generation – could have an insignificant or even positive impact on aggregate
macroeconomic outcomes. Second, all models highlight the potential re-allocation effects – both between
sectors and regions – that the introduction of circular economy enabling policies could have. Third, certain
types of macroeconomic model are more appropriate for assessing the transition than others, notably due to
their accounting of interactions between sectors and macroeconomic feedbacks. Fourth, of the assumptions
that are fed into these models – those concerning future rates of productivity growth, the substitutability
between different material types, and future consumption patterns – are key determinants of model
outcomes.

Keywords: Circular economy, resource efficiency, natural resources, raw materials, general equilibrium
model.

JEL Classification: C68, O13, Q53.

RÉSUMÉ

Ce rapport passe en revue les travaux existants sur la modélisation des conséquences
macroéconomiques de la transition vers une économie circulaire. Quatre conclusions clefs émergent de
l’examen de 24 études, qui constituent l’état de l’art de la modélisation des politiques d’efficacité des
ressources et de transition vers une économie circulaire. Premièrement, la transition vers une économie
plus circulaire, en l’associant avec une réduction de l’extraction de ressources and de la production de
déchets, aurait pour la plupart des modèles un impact macroéconomique négligeable voire positif.
Deuxièmement, tous les modèles soulignent les potentiels effets de réaffectation entre secteurs et entre
régions que l’introduction de ces politiques pourrait engendrer. Troisièmement, certains types de modèles
macroéconomiques sont plus appropriés à l’évaluation de la transition que d’autres, en particulier du fait de
leur prise en compte des interactions entre secteurs et des rétroactions macroéconomiques. Quatrièmement,
les hypothèses intégrées par ces modèles concernant l’évolution future des taux de croissance de la
productivité, de la substituabilité entre différents matériaux, et des modes de consommations sont des
déterminants majeurs des résultats des modèles.

Mots clés : économie circulaire, efficacité des ressources, ressources naturelles, matières premières,
modèle d’équilibre général.

Classification JEL: C68, O13, Q53.

3
ENV/WKP(2018)4

ACKNOWLEDGEMENTS

This report has been authored by Andrew McCarthy, Rob Dellink, and Ruben Bibas of the
OECD Environment Directorate. The authors are grateful to delegates of the Working Party on Resource
Productivity and Waste and the Working Party on Integrating Environmental and Economic Policies for
helpful comments on earlier drafts of this paper. They would also like to thank Peter Börkey, Jean Chateau
and Elisa Lanzi (all OECD Environment Directorate) as well as Alvaro Calzadilla, Martin Distelkamp,
Mark Meyer, and Matthew Winning for their advice and feedback. Stephanie Halimi and Katjusha Boffa
provided editorial assistance. The authors are responsible for any remaining omissions or errors.

Work on this paper was conducted under the overall supervision of Shardul Agrawala, Head of
the Environment and Economy Integration Division of the OECD’s Environment Directorate.

4
 ENV/WKP(2018)4

TABLE OF CONTENTS

ABSTRACT ....................................................................................................................................................3
RÉSUMÉ .........................................................................................................................................................3
ACKNOWLEDGEMENTS ............................................................................................................................4
EXECUTIVE SUMMARY .............................................................................................................................7
LIST OF ABBREVIATIONS .........................................................................................................................9
1. INTRODUCTION .....................................................................................................................................10
1.1 Context ............................................................................................................................................10
1.2 The scope of this review .................................................................................................................14
2. MODELLING THE TRANSITION TO A CIRCULAR ECONOMY: FOUR KEY DIMENSIONS ......18
2.1 The geographic coverage in existing studies ..................................................................................19
2.2 The sectoral coverage in existing studies ........................................................................................19
2.3 Material coverage and linking of economic and physical flows .....................................................23
2.4 Decoupling mechanisms in existing studies ...................................................................................25
3. ASSESSING THE ASSUMPTIONS AND RESULTS OF EXISTING STUDIES ..................................31
3.1 Business as usual projections ..........................................................................................................31
3.2 Policy coverage ...............................................................................................................................32
3.3 Economic consequences projected by the existing studies .............................................................38
4. CONSIDERATIONS FOR FUTURE MACROECONOMIC ASSESSMENTS......................................44
4.1 Baseline development .....................................................................................................................45
4.2 Modelling technological change .....................................................................................................46
4.3 Representing substitution ................................................................................................................46
4.4 Modelling “soft” enabling policies .................................................................................................47
BIBLIOGRAPHY .........................................................................................................................................49

Tables

Table 1. The studies and models considered in this review .......................................................................18

Table 2. Elasticities of substitution between primary and secondary materials in existing modelling ......29
Table 3. Summary of policy coverage in selected studies .........................................................................34
Table 4. Assumptions regarding average material productivity improvements in selected studies...........36

Figures

Figure 1. Differing definitions of the circular economy ...........................................................................15

Figure 2. Decoupling mechanisms: material circularity vs material efficiency in production vs
consumption ..............................................................................................................................16

5
ENV/WKP(2018)4

Figure 3. The models considered in this review by regional and sectoral coverage ................................20
Figure 4. Evolution of material productivity at the global level for major material categories ...............32
Figure 5. Headline modelling results in the studies considered in this review: GDP with respect to
baseline .....................................................................................................................................40
Figure 6. Headline modelling results in the studies considered in this review: GDP and resource
extraction with respect to baseline ............................................................................................41
Figure 7. Effects of a materials tax in the ICES, MEMO II, and MEWA models ...................................42

6
 ENV/WKP(2018)4

EXECUTIVE SUMMARY

Natural resources, and the materials derived from them, represent the physical basis for economic
growth. Recent decades have witnessed an unprecedented growth in demand for resources. This has
sparked increased interest from policymakers in a transition to a more circular economy. Three main
reasons are often highlighted for promoting a circular economy transition. First, reduced extraction,
processing, and disposal of natural resources may have significant environmental synergies; more efficient
resource use could represent an important tool for achieving climate and other environmental goals.
Second, the reduced reliance on critical resource and material inputs and improved security of materials
access that result from expanded domestic secondary material supply is important; supply risks associated
with future geopolitical shocks could be mitigated in importing countries. Third, the activities that will
drive any circular economy transition could also become significant drivers of job creation and economic
growth. New opportunities will emerge in various sectors, including secondary material production, repair
and remanufacture, the services sector, and the sharing economy.

The focus of this paper is to review the literature on the third point, i.e. the macroeconomic
implications of the transition to a more circular economy and increased efficiency in the use of material
resources. Addressing this issue is complex. Any such transition will involve multiple interactions between
different sectors and countries, and will take place in parallel with other trends like digitalisation and
automation. Ex ante, economy-wide quantitative models appear to be best suited to analysing this
transition as they capture the major drivers of the economic consequences. Furthermore, there is
insufficient ex-post data on the consequences of circular economy enabling policies for a robust empirical
assessment. As such, the analysis in this paper is restricted to studies that have used macroeconomic
models. However, in this context, such models are also only recently emerging; 15 of the 24 studies
considered are either currently in progress or were published since 2015.

There is no single commonly accepted definition of the term “circular economy”, but different
definitions share the basic concept of decoupling of natural resource extraction and use from economic
output, i.e. increased resource efficiency as outcome. One core view of the circular economy is that it can
be defined relative to a traditional linear economic system, i.e. one that focuses on closing resource loops.
A second, slightly broader, view of the circular economy stresses the importance of slower material flows,
either within an economy with some degree of material circularity, or within one that is more linear. The
third, and broadest, view of the circular economy is that it involves a more efficient use of natural
resources, materials, and products within an existing linear system. This broad view of the circular
economy affects potentially all economic activities, not only those that have a high material use profile,
and is the one applied in most modelling assessments and in this review.

Four key conclusions emerge from the existing literature. First, most economic models find these
shifts will have an insignificant or even positive impact on aggregate macroeconomic outcomes. In other
words, the current literature indicates that a transition to a (broadly defined) circular economy – with the
associated reductions in resource extraction and waste generation – could take place with potentially
significant positive (or at least without negative) consequences for economic growth or overall
employment. Second, all models highlight the potential re-allocation effects that the introduction of
circular economy enabling policies could have. The competitiveness of material intensive sectors – natural
resource extraction and certain types of manufacturing – will probably decline; workers, regions, and

7
ENV/WKP(2018)4

countries specialising in these activities may be made worse off in any circular economy transition. Other
sectors – waste management and recycling, remanufacturing and repair, and services more generally – will
probably expand as their output becomes relatively affordable. Third, dynamic multi-region models are
well suited to capturing the transition in the economy, as well as the socioeconomic trends and trade
impacts that will accompany any transition. In contrast, (static) single region models may be better suited
to representing material circularity in more detail. Fourth, three key sets of assumptions that drive
modelling outcomes, and the quality of the policy advice that emerges from them, are identified: (i)
assumptions on future efficiency improvements (e.g. future rates of material productivity growth, cost of
the underlying drivers, and role of policies), (ii) assumptions on the degree of substitutability between
primary and secondary materials, both for different materials, and in different applications, and (iii)
assumptions on the changes in the future structure of the economy and consumption patterns, and to what
extent will these take place in the absence of policy drivers.

8
 ENV/WKP(2018)4

LIST OF ABBREVIATIONS

CES Constant elasticity of substitution

CGE Computerised general equilibrium
DMC Domestic material consumption
EEMRIO Environmentally extended multi-region input output
EPR Extended producer responsibility
ME Macro-econometric
MRIO Multi-region input output
REE Rare earth element
RMC Raw material consumption
SAM Social accounting matrix
SSP Shared socioeconomic pathways

9
ENV/WKP(2018)4

1. INTRODUCTION

1.1 Context

Natural resources, and the materials derived from them, represent the physical basis for economic
growth. Land, water, and a variety of mineral-based fertilisers are critical inputs in our food production
system. Coal, oil, and natural gas, despite receiving a diminishing share of new energy investment,
continue to dominate the energy mix in many countries. Iron transformed into steel, and non-metallic
minerals transformed into cement, are central to infrastructure development. Bauxite transformed into
aluminium is important to the transport sector, while the group of rare earth elements are vital inputs across
a range of low carbon technologies.
Recent decades have witnessed an unprecedented growth in demand for resources. This has been
driven by the rapid industrialisation of emerging economies and continued high levels of material
consumption in developed countries. As a result, the weight of materials consumed worldwide has more
than doubled since 1980, and increased ten-fold since 1900. In 2013, global extraction of minerals, fossil
fuels, and biomass reached 84 billion tonnes (SERI, 2017), or around 30 kilograms per person per day.
There is also considerable variation across regions; material consumption in developed countries tends to
be several times higher than that in developing ones (Wiedmann et al., 2015).1
By 2050, the world population is expected to increase from about 7 billion to more than 9 billion, and
the per capita income of the world’s population to roughly triple (OECD, 2012). This will substantially
increase demand for natural resources, especially if global production and consumption patterns converge
with those of OECD countries. Robust projections of future global resource consumption are scarce
(OECD, 2012; OECD, 2016a), however UNEP’s International Resource Panel (UNEP, 2017) has projected
that total resource use may more than double by 2050 if existing trends continue. Unless the efficiency
with which resources are used is significantly improved, this is likely to lead to increasing input costs and,
for some resources, a growing risk of supply shortages (e.g. Coulomb et al. 2015).
Business as usual resource use will also increase the environmental impacts that are associated with
harvesting resources, processing and using them, and disposing of the resulting waste. Areas of particular
concern include the local environmental damages and greenhouse gas emissions associated with material
extraction and processing (e.g., Nuss and Eckelman (2014). Interestingly, not many of the studies surveyed
in this paper make the environmental benefits associated with improved resource efficiency and a
transition to a circular economy explicit. These benefits include the energy savings while recycling energy-
intensive materials. For instance, every ton of steel scrap made into new steel, over 1,400 kg of iron ore,
740 kg of coal, and 120 kg of limestone are saved (World Steel Association, 2012). Similarly, the
aluminium production from scrap requires around 10% of the energy input of primary aluminium
production (IEA, 2015). Clearly, environmental impacts associated with material extraction differs widely
between materials and extraction and processing technologies, and advanced extraction methods may not
be more harmful than dirty recycling techniques. Nonetheless, many papers have an implicit assumption
that a transition from primary to secondary materials will reduce environmental pressures. A detailed
global analysis of the net environmental impacts of such a transition is however beyond the scope of the
economic assessment in this paper.

1
On a raw material consumption (RMC) basis, i.e. where the materials embodied in imports are accounted for.

10
 ENV/WKP(2018)4

1.1.1 The growing interest in a transition to a more circular economy

These issues have led to increasing interest in the circular economy and improved resource efficiency
in recent years (see Box 1 for an introduction to the key terminology used in this paper). At the
international level, a succession of multilateral initiatives and frameworks have been introduced. An
OECD Council Recommendation issued in 2008 encouraged member countries to “take appropriate
actions to improve resource productivity and reduce negative environmental impacts of materials and
product use”. In the same year, G8 environment ministers signed the Kobe 3R Action Plan, in which
countries agreed to prioritise implementation of 3Rs2 policy in order to improve resource productivity.
There have also been several important recent developments. The creation of the G7 Alliance on Resource
Efficiency at Schloss Elmau in 2015, and the subsequent adoption of the Toyama Framework on Material
Cycles3, signalled increasing interest from G7 countries. The inclusion of specific goals4 related to resource
efficiency in the 2030 Agenda for Sustainable Development also represented a major landmark. Finally,
the introduction of resource efficiency into the G20 agenda in 2017 was notable, particularly given the
presence in that forum of various countries with large resource endowments.

Box 1. Definitions: circular economy, resource efficiency, secondary materials and decoupling
There is no single accepted definition of a circular economy. The precise meaning of a “transition to a circular
economy” varies across the current literature, but tends to involve reduced demand for certain natural resources, and
the materials that are derived from them. The resources usually emphasised are minerals (both metallic and non-
metallic), fossil fuels, and various biotic resources such as forestry, fish, or other biomass. Relatively little attention
tends to be given to other resources: land and water are the most obvious examples.
Three main mechanisms for reduced demand are often highlighted (e.g., Bocken et al., 2016). Creating material
loops involves the substitution of secondary materials (i.e. those that have been used already in production processes
and are derived from the recycling of industrial or household waste) and second-hand, repaired, or remanufactured
products for their virgin or new equivalents. Slowing material flows involves the emergence of products which remain
in the economy for longer, usually due to more durable product design. Narrowing material flows involves the more
efficient use of natural resources, materials, and products, either through the development and diffusion of new
production technologies, the increased utilisation of existing assets, or shifts in consumption behavior away from
material intensive goods and services. In sum, a “transition to a circular economy” could therefore be seen as
involving any process that might lead to lower rates of natural resource extraction and use. This is the definition that is
used in this review (see Section 1.2.1 for further details).
Resource productivity is more easily defined. It refers to “the effectiveness with which an economy or a
production process is using natural resources” OECD (2015b). Resource efficiency is generally used in a broader
sense. It is used by UNEP (2017) to refer to a set of ideas including: (i) the technical efficiency of resource use, (ii)
resource productivity, or the extent to which economic value is added to a given quantity of resources, and (iii) the
extent to which resource extraction or use has negative impacts on the environment. In concrete terms, resource
efficiency, or more precisely resource intensity, can be calculated as the ratio between the value of economic output
from a particular sector or economy, and the amount of resources (typically in terms of weight) used to produce it. This
is the definition used in this paper. An improvement in resource efficiency therefore describes a situation where more
economic value is being produced with a particular amount of resources (or one where fewer resources are being used
to produce a particular level of economic value).
Decoupling is used to describe an improvement in resource efficiency, usually at the aggregate level of an
economy. Relative decoupling refers to a situation where the value of economic output and the amount of resource
inputs are growing, but with the former at a higher rate than the latter. This process has been well documented at the
level of the global economy during the last 30 years (OECD, 2016). Absolute decoupling refers to a situation where
the value of economic output is growing while the amount of resource inputs used is shrinking. There is little evidence
for absolute decoupling in any country once the materials embodied in intermediate imports are taken into account
(e.g. OECD, 2016, Wiedmann et al., 2015).

2
3Rs = reduce, reuse, recycle.
3
In which G7 countries committed to taking ambitious action on resource efficiency.
4
For example, goal 8 is to “promote inclusive and sustainable economic growth, employment and decent work
for all” while goal 12 seeks to “ensure sustainable consumption and production patterns”.

11
ENV/WKP(2018)4

A circular economy transition, to the extent that it results in lower resource extraction without an associated
reduction of economic output, can result in improved resource efficiency and decoupling. Whether this is possible is
the main question addressed in the studies considered in this review.

There has also been tangible policy action by a number of governments. Circular economy roadmaps
were introduced in the People’s Republic of China (hereafter China) in 2013, in the European Union in
2015, and in Finland, France, and the Netherlands in 2016. Several of these roadmaps include specific
quantitative targets on resource efficiency, recycling rates, or disposal quotas. For example, China has a
stated objective of reusing 72% of industrial solid waste (Mathews and Tan, 2016) while the Netherlands is
aiming for a 50% reduction in the use of virgin resource inputs by 2030 (MIE, 2016). Other countries have
introduced national policy frameworks related to resource efficiency or materials management. Japan’s
Fundamental Law for Establishing a Sound Material-Cycle Society is supported by regulations on the
management of specific waste streams, and targets a cyclical use rate of 17% by 2020 (MoE, 2013). The
Sustainable Materials Management Program Strategic Plan in the United States focuses on tracking and
reducing the overall amount of materials disposed of, reducing lifecycle environmental impacts of
materials, and increasing socio-economic benefits from materials. It includes a national target of a 50%
reduction in food waste by 2030 (US EPA, 2015).

The transition to a more circular economy, and to improved resource efficiency, is not usually
considered to be a policy goal in itself. Rather, it is the economic, environmental, and social gains that
might accompany such a transition that seem to be of interest for governments. Four specific sets of
benefits tend to be highlighted in discussions of a circular economy transition:

 Increased demand for natural resources will increase the wastes and emissions generated in
extraction, processing, consumption, and disposal activities. The resulting deterioration in
environmental quality could become a bottleneck for continued improvements in living
standards. In this context, policymakers see synergies between natural resource decoupling and
achieving various environmental objectives. In particular, there is increasing awareness that more
efficient material management can be a useful tool for meeting national level climate
commitments. Around 50% of industrial CO2 emissions can be attributed to the production and
processing of five basic materials (FMEAE, 2015) – steel, cement, paper, plastic, and aluminium
– most of which have secondary equivalents that are considerably less energy intensive to
produce.
 It is also pointed out that increased domestic secondary material production to reduce imports and
production of virgin material resources can reduce supply risks associated with future
geopolitical issues. This is especially relevant for metallic mineral resources which, among other
things, are highly geographically concentrated, amenable to recycling, and critical inputs in an
increasing number of applications. The oil shocks of the 1970s and the recent Chinese export
quotas on rare earth elements (REEs) represent historic examples of such supply risks.
 It is often suggested that the activities that will drive any circular economy transition could also
become significant drivers of re-industrialisation, job creation, and economic growth. New
opportunities will emerge in various sectors, including secondary material production, repair and
remanufacture, the services sector, and the sharing economy. Further, early adopting countries
could realise additional benefits by becoming exporters of circular economy expertise and
technology. In this view of the circular economy, there is a win-win proposition that is not being
realised.
 International fora emphasise that large increases in demand for natural resources will be
associated with a growing and increasingly affluent global population (G7, 2016; UNEP, 2017).
Given the finite nature of many natural resources, it is often concluded that future resource
scarcity could become a drag on long run economic growth. In this context, decoupling of

12
 ENV/WKP(2018)4

economic output from natural resource use is seen as a vital ingredient for sustainable
development. From an economic perspective, it may make more sense to look at resource
criticality, i.e. a combination of risk of supply disruptions and economic importance, rather than
absolute scarcity (Coulomb et al., 2015).

1.1.2 The macroeconomic impacts of a circular economy transition are not well understood

The focus of this review paper is on the third point: the macroeconomic implications of a transition to
a more circular economy and increased efficiency in the use of material resources. It is clear that the
policies required to drive any such transition will result in structural shifts involving the decline of certain
sectors and the rise of others, with a reallocation of capital and labour along the way. What is less clear is
the possible magnitude of these shifts, and what their overall impact on aggregate economic outcomes
might be. A recent review of the circular economy (Rizos et al., 2017) concludes that there is “little
specific analysis or data on how different sectors will be affected” and “there is also a need to understand
the indirect effects on the economy (e.g. impacts on the value chain and/or changes in consumption
spending patterns) in order to estimate the overall impacts”.

There is an emerging body of work that employs ex ante quantitative models to address these
questions. Ex ante because many aspects of a circular economy transition are “out of sample”; there is no
historical experience that can be drawn upon for empirical analysis.5 Quantitative because any such
transition is likely to be highly complex; it will affect many types of resources and materials, involve
multiple sectors and countries, with spill-over and interaction effects between each, and take place in
parallel with other emerging trends such as digitalisation and automation. These models, which are the
main focus of this review, are only very recently being employed more widely; well over half of the known
literature has been published since 2015. Although most assessments find that circular economy enabling
policies will have a positive impact on aggregate economic outcomes, there is considerable uncertainty in a
number of the underlying modelling assumptions, and therefore in the reliability of these results.

1.1.3 The objectives of this review

The primary aim of this paper is to critically review the existing assessments of the macroeconomic
consequences of a transition to a circular economy, and to improved resource efficiency. It builds on an
earlier OECD assessment (Dubois, 2015), which concluded that the “current quantification methodologies
for the circular economy give a first estimation of the benefits but are not robust enough to serve as
trustworthy policy tools”. In addition to reviewing the methodologies and results of existing modelling,
this paper therefore also offers a set of recommendations on how future work can better assess the
efficiency and effectiveness of frequently proposed circular economy enabling policies.

A full meta-analysis of the economic implications of the transition to the circular economy is well
beyond reach given the current state of the literature. Currently, the economic literature is still scarce on
the topic, especially concerning macroeconomic impacts rather than sectoral impact, and therefore any
statistical analysis on these results would have very large error margins and not provide very robust
insights. In addition, a comprehensive meta-analysis would require detailed insights into the comparability
of parameters and input data, which for many sectors and materials is incomplete, in particular at the global
level. Therefore, this paper focuses on a more qualitative review of the limited number of existing studies.

5
Although there is a small body of empirical work that assesses the employment effects of various waste
management policies.

13
ENV/WKP(2018)4

The likely impacts of a circular economy transition, when assessed in an ex ante context, depend on
(i) how such a transition is defined, (ii) the structure and assumptions of the model that is used to assess it,
and (iii) the policies that are implemented to enable it.6 This paper addresses each of these issues in three
sections. The remainder of this introduction outlines the scope of this review in terms of the modelling
approaches and definitions of the circular economy that are considered. The second section describes
existing modelling tools in terms of four key characteristics: geographic coverage, sectoral coverage,
linkages with physical material flows, and the mechanisms used to simulate efficiency improvements and
different types of substitution. The third section discusses the types of policies that have been implemented
in existing studies, and summarises the main results in terms of macroeconomic impacts.

1.2 The scope of this review

1.2.1 Circular economy definitions

Recent literature reviews of the circular economy make it clear that it has no single commonly
accepted definition (CIRAIG, 2015; Rizos et al., 2017). There are at least three common views (Bocken et
al., 2016), each of which is summarised in Figure 1. Although each definition involves different processes
and actors, they share a similar outcome: increased resource efficiency or, in other words, the decoupling
of natural resource extraction and use from economic output.

6
In some cases, market forces alone have been enough to stimulate the use of secondary materials, and high
recycling rates have emerged without policy intervention (see Coulomb et al., 2015 and McCarthy, 2018
forthcoming for selected metals). This is limited to a few materials, however, and a broader transition will
require active policy intervention.

14
 ENV/WKP(2018)4

Figure 1. Differing definitions of the circular economy

One core view (closing resource loops in Figure 1) of the circular economy is that it can be defined
relative to a traditional linear economic system, i.e. one where natural resources are extracted, transformed
into materials and products, and eventually disposed of in incineration or landfill facilities. Higher rates of
material circularity, achieved through concepts such as “closing material loops” or “using waste as a
resource”, are central to this vision. Substitution of recycled materials for those derived from virgin
resources, remanufactured goods for their traditional equivalents, and used products for new ones are seen
as the key processes. The main sectors of the economy likely to be involved are therefore: waste
management services, other services (e.g., repair), secondary material manufacturing (e.g., recycling), and
other manufacturing (e.g., remanufacturing).

A second, slightly broader, view of the circular economy (slowing resource loops in Figure 1) stresses
the importance of slower material flows, either within an economy with some degree of material
circularity, or within one that is more linear. Product design is typically highlighted as playing an
important role; products that are designed to be robust and more easily repairable will last longer and slow
the introduction of new natural resources into the economy. Addressing firm incentives for designing
productions with planned obsolescence in mind is seen as a key driver. The main sectors likely to be
involved are therefore various kinds of manufacturing.

The third, and broadest, view of the circular economy (narrowing resource flows in Figure 1) is that it
involves a more efficient use of natural resources, materials, and products within an existing linear system.
The development and diffusion of resource efficient technologies is central to this view, as is a postulated
shift in consumption patterns towards less material intensive goods and services. There is also an idea in
the literature that there is significant “structural” waste in current consumption patterns. Oft-cited examples
include a perceived under-utilisation of assets such as office space and private vehicles, and the high rates

15
ENV/WKP(2018)4

of food waste in many countries. Key drivers that are often highlighted in this view of the circular
economy include investment in R&D and resource-saving technology, an increased awareness in the
external effects of consumption decisions, and the continued emergence of the sharing economy. A much
broader group of sectors are likely to be affected than those in the core view of the circular economy, as
this broad view of the circular economy affects potentially all economic activities, not only those that have
a high material use profile.

This divergence in definitions found in the general circular economy literature is also represented in
the modelling literature considered in this review. Although most modelling assessments to date have
applied the broadest definition (narrowing resource flows), there are also a number of models that attempt
to introduce material circularity. This review therefore includes publications on the circular economy
strictly defined (closing resource loops), and those on resource efficiency more generally (slowing resource
loops and narrowing resource flows). We consider the circular economy as any process that enables the
decoupling of economic output from virgin resource extraction. Figure 2 summarises the set of available
mechanisms in this respect.

Figure 2. Decoupling mechanisms: material circularity vs material efficiency in production vs consumption

Note: Technological change can also facilitate more circular material flows when it results in improved secondary production
technologies.

1.2.2 Modelling approaches

As discussed above, because many aspects of a circular economy transition are “out of sample”,
empirical approaches relying on historic experience have not been widely utilised. Rather, most modelling
studies have used ex ante simulations to assess the likely macroeconomic consequences of any such
transition. There are two main variants.

16
 ENV/WKP(2018)4

The first approach, termed accounting modelling (after Dubois, 2015), involves the development of
scenarios regarding material circularity or technological progress in one or several sectors (e.g., Bastein et
al., 2013; Ellen McArthur Foundation, 2013; Stegeman, 2015, and SITRA, 2015). Scenarios are based on
expert opinion, and are typically described in terms of higher future recycling, remanufacturing, repair, or
re-use rates. These changes are modelled autonomously; that is, they are not driven by the implementation
of a particular policy. The resulting economic benefits, either in terms of the cost savings achieved through
reduced material use, or in terms of job creation, are then estimated. In some cases, the changes in final
demand and in production of the directly affected sectors are used to calculate indirect effects throughout
the rest of the economy using input-output tables. This procedure gives some insight into impact of the
supply shock in other (third party) sectors but, because there is no price mechanism, these impacts don’t
fully reflect economic feedback processes.

The second approach involves the use of economy-wide quantitative models: computable general
equilibrium (CGE) and macro-econometric (ME) models.7 Despite making different assumptions about
agent behaviour, these models share two distinct advantages with respect to accounting models. First, they
both explicitly represent the role that prices play in determining supply and demand for products,
commodities, and ultimately, natural resources. This is important in the context of resource efficiency;
increased output from secondary material sectors may reduce demand for natural resources, but this is
likely to be partially offset by the lower prices that this entails. Second, multi-sectoral models, including all
CGE and some ME models, are based on an underlying social accounting matrix (SAM) that accounts for
economic flows throughout the entire economy. As such, these models can identify the potential
interactions and spill-overs of a policy on sectors and agents other than the ones initially affected.

This review is restricted to studies that use CGE or ME models to assess the macroeconomic impacts
of circular economy and resource efficiency enabling policies. The complexity of the envisaged circular
economy transition, the fact that it will affect large parts of the economy, and the likelihood of rebound
effects mean that such models are the most suitable for such an analysis. Although accounting models can
provide detailed insight into the likely costs and benefits of increased material or product circularity, they
tend to do so for specific products, and without feedbacks associated with changing prices.

7
Gradually, other types of models are also emerging, such as agent-based models and DSGEs. The literature on
applying these to resource efficiency is so scarce, however, that they are excluded from this review.

17
ENV/WKP(2018)4

2. MODELLING THE TRANSITION TO A CIRCULAR ECONOMY: FOUR KEY DIMENSIONS

The studies that are considered in this review are presented in Table 1. The majority of this work has been
undertaken during the last few years, probably in response to growing interest in the circular economy. It is also clear
that this research area is still in its infancy; while there is now a considerable body of expertise in integrated
economic-energy modelling, incorporating material flows into such models raises an entirely new set of issues.
(Pollitt et al., 2010) assessed sixty of the most widely used macroeconomic models, and found that around half had a
strong energy focus, while “consumption of material inputs is largely unexplored within a dynamic macroeconomic
framework”. This section describes four aspects of the CGE and ME models that have been used to assess a circular
economy transition: their geographic, sectoral, and material coverage, and the economic mechanisms that they
include.

Table 1. The studies and models considered in this review

Modelling Group Key Paper Model Name # Regions # Sectors # Materials Circularity?

GWS - SERI Bockermann et al. (2005) PANTA RHEI 1 59 4 (p) N

Hitotsubashi Okushima and Yamashita (2005) ODIN-WR 1 25 10 (s) Y
NIES Masui (2005) AIM 1 41 18 (s) Y
KEI Korea Kang et al. (2006) AIM-INCGE 1 32 19 (s) Y
IAMC China Unpublished AIM-IPAC ? ? N
Wuppertal InstituteDistelkamp et al. (2010) PANTA RHEI 1 59 4 (p) N
TNO TNO (2012) EXIOMOD 27 15 ? N
UCL Ekins et al. (2012) E3ME 30 42 19 (p) N
Cambridge
Cambridge Econometrics (2014) E3ME 34 43 19 (p) N
Econometrics
French Ministry of
Godzinski (2015) Vulcain 1 5 2 (s) Y
Environment

NERA Tuladhar et al. (2015) NewERA 5 17 0 (p) (s) N

Ellen McArthur Bohringer and Rutherford (2015) ? 5 16 0 (p) (s) N
World Bank Bouzaher et al. (2015) ? 1 12 1 (s) N
NIER Sweden Soderman et al. (2016) EMEC 1 26 34 (s) N
Ex'Tax Groothuis et al. (2016 E3ME 28 43 19 (p) N
CSIRO Schandl et al. (2016) GIAM 13 21 ? N
ERC Hartley et al. (2016) SAGE 1 49 13 (s) Y
ICES 19 20 ? N
DYNAMIX Bosello et al. (2016) MEMO 1 10 ? N
MEWA 1 18 ? N
SIMRESS Unpublished GINFORS 39 27 7 (p) N
Meyer et al. (2016) GINFORS 39 27 7 (p) N
POLFREE
Hu et al. (2016) EXIOMOD 26 36 5 (p) N
UCL Winning et al. (2017) ENGAGE Material17 35 1, 1 (p) (s) Y
UNEP IRP UNEP (2017) GTEM 28 21 10 (p) N
WIFO Sommer and Kratena (2017) WIFO DYNK 1 62 4 (s) Y
Note: (p) = linkages with primary materials, (s) = linkages with secondary materials.

18
 ENV/WKP(2018)4

2.1 The geographic coverage in existing studies

The level of disaggregation of countries and regions in the modelling analysis is particularly
important in the context of this paper; the highly uneven distribution of natural resources across countries
means that circular economy enabling policies will probably have quite different impacts in resource-rich
and resource-poor countries. Models with lower geographic disaggregation will tend to group country
types together, resulting in biases in the modelling outcomes. These differentiated impacts make the
linkage to international trade crucial to an accurate assessment of a transition to a circular economy.

Another important consideration relates to how trade linkages are represented. In some models,
economic activity in each region is linked to that in all other regions through bilateral trade flows. This is
the case in some multi-regional models (EXIOMOD and GINFORS for example) that explicitly represent
bilateral international trade. In the context of this paper, this is useful to the extent that intermediate and
final goods from different regions have differing resource intensities; resource efficiency in a particular
country will partially depend on where its imports originate from.

The models considered in this review can be divided into those that consider a single, usually national
level, economy and those that link multiple economies through trade (Figure 3). Geographic coverage in
multi-regional models ranges between 5 (e.g. Tuladhar et al. 2015) and 39 regions (e.g. Meyer et al. 2016).
The multi-region models have a more explicit focus on the circular economy and, more specifically, on the
broadest (narrowing resource flows) definition of it. This is largely a function of data constraints;
modelling material circularity in a multi-regional context would require harmonised cross-country
economic data for waste management, recycling, and secondary production sectors that is largely
unavailable.8

The single-region assessments of resource efficiency or circular economy policies considered in this
review have been undertaken for many OECD countries, including Japan, Korea, Sweden, Germany and
France (Table 1). It is notable that no single-region assessments have been undertaken for countries with
large extractive sectors. Representation of trade flows in single-region models is usually limited to sectoral
trade balances in the SAM; bilateral trade flows are generally lacking. In addition, the concepts of resource
efficiency and the circular economy don’t necessarily appear explicitly; these models have often been
developed with a more specific focus on waste management and energy policies.9 This means that they
have well developed linkages between economic and physical material flows in the post-consumption part
of the economy; material circularity is relatively well represented. Clearly, this is facilitated by the
availability of detailed data describing economic and physical waste flows in at the country level.

2.2 The sectoral coverage in existing studies

The sectoral coverage of the models considered in this review varies widely, both in terms of the
number of sectors included, and the number of sectors that are directly relevant for the circular economy.
The SAMs underlying most models typically have between 15 and 40 sectors, although some have as many
as 164 (EXIOBASE), albeit with greater uncertainty regarding the accuracy of the data. 10 Single-region
models generally have greater sectoral disaggregation than multi-regional models (Figure 3). This section

8
The EXIOBASE (version 3.0) dataset that is currently in development is one exception to this.
9
An early example of a national CGE model for waste management is Bartelings et al. (2004).
10
The national accounts data provided by most countries does not have this degree of disaggregation. Separating
a particular economic activity from its parent sector therefore typically requires making various assumptions
and the use of other data sources.

19
ENV/WKP(2018)4

addresses the level of sectoral disaggregation in parts of the economy directly relevant for a circular
economy transition: upstream resource extraction, material transformation and manufacturing, waste
management and recycling, and several service sectors. However, disaggregation of the sectors that are not
directly involved in provision or handling of materials is also important; one of the main mechanisms for
reducing material use is shifting the economy away from sectors that are relatively material intensive to
sectors that are less so.

Figure 3. The models considered in this review by regional and sectoral coverage

Note: The horizontal axis of this chart reflects an article’s publication date of rather than the date of model development.

2.2.1 Extractive sectors

Disaggregation of upstream extractive activities, preferably by resource type, is important for

modelling the circular economy. It allows economic instruments to be applied to specific extractive
activities (e.g., fossil fuel, mineral, fisheries, or forestry extraction etc.), and therefore to better grasp the
economic effects of these instruments further downstream. As discussed in Section 2.3.1, greater sectoral
disaggregation also allows more realistic linkages to be established between economic flows and their
physical equivalents.

All models considered in this review include at least one upstream extractive sector, and many
differentiate between mineral extraction and different types of fossil fuel or biomass extraction.
EXIOBASE has the most disaggregated upstream production structure, differentiating between forestry,
fisheries, 11 agricultural sectors, 4 fossil fuel sectors, and 11 mining sectors. The GTAP series of
databases, upon which a number of the models assessed here are based, also has disaggregated agricultural
and fossil fuel sectors, but does not currently distinguish between extraction of metallic and non-metallic
minerals. Further disaggregation of extractive sectors is possible11, but can be highly time-consuming due
to the data collection required.

11
For example, Winning et al. (2017) disaggregate the GTAP sector ‘Other Mining’ into three separate sectors:
iron ore mining, non-ferrous mining, and other mining.

20
 ENV/WKP(2018)4

2.2.2 Transformation sectors

Transformation sectors – those that refine raw materials into processed commodities such as paper,
timber, refined fuels, processed metals etc. are important for an analysis of the circular economy. The
SAMs underlying most models have a good disaggregation in this area, generally distinguishing between
five or six of the major material classes. The main issue in the context of this review is that, for a given
material, there is generally no distinction between facilities that use virgin natural resources, and those that
use secondary raw materials.

While that is largely irrelevant for materials that are not recyclable (e.g., fossil fuels) or that are
mostly recycled within the sector they are used in (construction minerals), it is critical for materials like
metals, plastics, and paper where secondary production typically utilises a specific technology. If detailed
data on the alternative technologies is available, representing these specific secondary production
technologies is usually done through modelling different production technologies within one sector,
effectively modelling two different ‘sub-sectors’ that provide the same output, but using different
combinations of inputs. Enabling policies such as recycling standards or subsidies for secondary
production are difficult to model unless technologies using secondary materials are explicitly represented.

Only two of the multi-region CGE models considered in this review introduce secondary production
sectors. The ENGAGE-Material model, currently being developed at University College London (Winning
et al. 2017), splits steel production into two additional sectors; one that (primarily) uses virgin mineral ores
as input and one that uses secondary metal scrap. The EXIOBASE SAM makes the same distinction for six
metals, although this version of it is yet to be utilised in the EXIOMOD CGE model. While it is possible to
explicitly represent the specification of alternative material use technologies and their evolution over time,
the implied need of disaggregating technologies and modelling sector specificities and dynamics constrains
the modeller to very few materials (usually metals). There seems to be a clear trade-off between a
comprehensive but not explicit representation of all materials and a detailed sectoral representation, but
necessarily restricted to very few materials.

In addition, there are at least five single-region CGE studies that introduce some sort of secondary
production, albeit in slightly different ways. Okushima and Yamashita (2005) follow the approach used in
ENGAGE-Material and EXIOBASE, explicitly separating out secondary production in seven
transformation sectors, including ferrous and non-ferrous metals, food, paper and pulp, and ceramic, stone
and clay. Masui (2005), Kang et al. (2006), and Godzinski (2015) take a slightly different approach. They
introduce a waste management sector that produces a disposal service along with multiple secondary
commodities12. These then substitute with primary raw materials in downstream manufacturing sectors13.
Hartley et al. (2016) also introduce secondary production, but do so by monetising the recyclable content
of 13 individual waste streams and using this value as an exogenous supply shock for resource availability
in the model. Again, this illustrates that the analysis may be deeper for country-level studies (more
available data sources as well as technology and policy details), but this depth comes at the expense of
linking the materials use with international trade flows and macroeconomic dynamics.

12
For example, Masui (2005) and Kang et al. (2005) introduce a waste management sector for industrial and
municipal waste respectively (with the former defined as waste generated in production activities and the latter
as that generated in household and business activities). These sectors produce 18 commodities which then
become substitutes for the equivalent primary commodities in 12 downstream sectors.
13
See section 2.4.3 for further description on how this substitution is modelled.

21
ENV/WKP(2018)4

2.2.3 Waste management and recycling sectors

Waste management and material recovery activities are poorly represented in the multi-regional
quantitative models that have been used to assess a circular economy transition14. This has hindered the
modelling of a number of relevant enabling policies, including landfill taxes, disposal bans, and recycling
quotas. In many cases, waste related activities are included within a generic service sector that produces
everything from insurance, health, education, and financial services (e.g. ICES and MEMO II models:
Bosello et al. 2016). Some studies distinguish between a waste management activity (comprising collection
and disposal), and a recycling activity (comprising material sorting and secondary production). However,
both are usually aggregated with an array of other activities; waste management with other public services
and recycling with other types of manufacturing. The EXIOBASE SAM has perhaps the best
representation of waste related activities, with individual sectors for waste collection, incineration,
disposal, metal recycling and non-metal recycling, but the quality of the underlying data is unclear. The
MEWA model used in the DYNAMIX project also introduces a dedicated recycling sector, although it is
unclear what the output of this sector is, and how substitutable this output is with that from extractive
sectors.

Several of the single-region CGE models discussed in Section 2.1 do introduce specific waste
management sectors into the SAM (Masui, 2005; Kang et al., 2006; Godzinski, 2015). This sector utilises
the solid waste generated by production and consumption activities15 to produce two outputs: a disposal
service (in some cases split between landfilling and incineration), and one or more secondary raw
materials. As discussed above, these then substitute with primary raw materials in downstream
manufacturing sectors.

2.2.4 Product-life extending activities

Remanufacturing, repair, and trade of second hand goods are clearly also relevant for a circular
economy transition. However, these are not well represented in the models considered here. Currently,
remanufacturing is mostly undertaken by original product manufacturers. In the SAM, the value of
remanufacturing output in a particular sector is therefore aggregated with the value of traditionally
manufactured output. Separating these two components to allow substitution between them is likely to be
prohibitively difficult. Output from repair services is aggregated with a range of other services in most
SAMs. For example, GTAP 8 groups this with a number of other activities including all retail sales and
hotels and restaurants. EXIOBASE provides better disaggregation – the sale, maintenance, and repair of
vehicles is separated from all other retail trade – however, this is still insufficient to model impacts on the
repair sector. These activities are at the core of many soft policies to decrease material use. Since their
accounting in data and representation in models is difficult, the results of most studies have to be handled
carefully.

14
For example, in the GTAP database, waste management is aggregated with other (government) services, such
as education, health and defence.
15
Masui (2005) and Kang et al. (2006) introduce pollution as a factor of production; it represents a transfer from
the representative firm in each sector to the waste management sector. In Godzinski (2015), consumers
demand a waste disposal service in addition to a generic consumption good. This sets up a transfer from
consumers to the waste management sector.

22
 ENV/WKP(2018)4

2.3 Material coverage and linking of economic and physical flows

2.3.1 Primary materials: upstream linking

A central objective of the studies considered in this review is to assess the impact that circular
economy enabling policies may have on natural resource extraction rates. Among other things, this will
depend on the changes in relative prices that these policies stimulate, and the reaction of firms and
households to the new set of prices. As such, any model of the macroeconomic consequences of a circular
economy transition needs to include some accounting of economic flows.

Given that different sectors are likely to expand and contract to different extents during a transition to
a circular economy, and that individual sectors have differing material intensities, these economic flows
also need to be linked with their physical equivalents. Changes in sectoral economic output can then be
translated into aggregate resource extraction terms.

The range of resources or materials that can be considered is constrained only by the availability of
physical extraction data. Distinguishing between a larger number of materials is important because of
differences in: (i) physical abundance – some materials are more scarce than others, (ii) environmental
impacts – extraction, processing, and use of some materials can produce differentially large damages,
(iii) historic decoupling rates – there has been relatively rapid historic decoupling for particular materials,
and (iv) potential future decoupling mechanisms – material circularity, for example, is only relevant for
some materials. In practice, limited sectoral disaggregation of economic value flows can hinder the
accuracy with which physical and value flows can be linked (see below). For this reason, most models
consider between five and ten materials, although models with more disaggregated SAMs have considered
as many as eighty (e.g., Hu et al., 2016).

Economic and physical flows have traditionally been linked using a so-called production approach.16
This typically involves assigning Domestic Material Consumption17 (DMC) to the domestic sector where it
first enters the economic system. Sector specific material intensities can then be estimated by distributing
this extraction through the economic system on the basis of economic value flows represented in the SAM.
The main limitation of this approach is that it does not account for the materials embedded in traded goods
and services. This can lead to potentially misleading results. For example, countries without domestic
resource extraction or processing sectors can appear to consume very few materials, even though there may
be significant volumes embedded in imports. The recent development of environmentally extended multi-
regional input-output (EEMRIO) tables largely addresses this issue (see Giljum et al., 2014; Wiedmann et
al., 2015; Wood et al., 2015). This approach uses bilateral trade data to link intermediate goods and
services imported in one country back to their country of origin and, in turn, to the resources that were used
in their manufacture. By including the bundle of material resources embodied in intermediate inputs and
finished goods in the analysis, this approach allows the calculation of DMC, but also the embedded
materials in imports, as reflected in Raw Materials Consumption (RMC). Ideally, the information on DMC
and RMC is presented next to each other to provide full insights into overall material use. The distinction
between DMC and RMC is also very helpful in the analysis of international trade consequences of the
transition to a circular economy.

16
For fossil fuels, there is relatively ample data available, and data on volume flows can be directly linked to the
respective economic activities.
17
DMC is the physical quantity of domestic extraction of a given resource plus the difference between any
equivalent imports and exports.

23
ENV/WKP(2018)4

There are two main sources of error in mapping physical material flows according to their associated
economic equivalents. The first relates to an implicit assumption about the homogeneity of the output of
each sector represented in the SAM. In practice, many SAMs are highly aggregated; a given production
sector groups multiple products that end up in numerous supply chains. The distribution of material
extraction throughout the economy on the basis of value flows could therefore lead to a contestable
allocation of materials (e.g. Schaffartzik et al., 2014). For example, if, as in several of the models
considered in this review, metallic and non-metallic mining are grouped into a single extractive sector in
the SAM, then domestic extraction of construction minerals will find its way into the same supply chain as
iron, copper, aluminium, and other metal ores18. This issue is often cited as a key motivation for the
development of highly disaggregated MRIOs such as EXIOBASE (Giljum et al., 2016). This type of error
is especially valid for models that take a “supply-side” approach to materials accounting, i.e. for models
that link material use to the extracting sector and then follow the flows through the economy based on
where the extraction sector outputs go. The alternative is to link materials to the demand for the outputs of
the extraction sector (the “demand-side” approach). In this set-up, one can make ad-hoc, but logical,
assumptions on the differences in materials content coming from one aggregated extraction sector; for
instance, the materials flowing into metals processing would comprise metals, whereas the inputs from
extraction to the construction sector would comprise building materials (essentially non-metallic
minerals)19.

The second source error relates to an assumption about the homogeneity of prices for a given output.
It is not uncommon for different economic activities to pay different prices for the same product. One
example relates to corn for food versus corn for feedstock. In this situation, distributing material flows on
the basis of aggregated value flows that consist of multiple sub-products with different qualities and prices
may also result in biased sectoral material intensities. To the extent that price differentials stem from
differentiation in tax rates, net-of-taxes value flows can be used and are not biased.

2.3.2 Downstream linking

Linking economic and physical flows further downstream is also important, particularly for modelling
material circularity. As discussed in Section 2.1, there are a handful of (mostly single-region) studies that
introduce secondary production, and thereby allow for substitution between the outputs of primary and
secondary material sectors. An important consideration in this context is the representation of supply of
secondary materials. Output from secondary sectors is in reality constrained by the availability of waste; it
isn’t possible, or even desirable, to prematurely recycle the in-use stock of capital and consumer goods in
order to increase the availability of secondary materials. This is distinct to the situation for primary
materials where, at least in the medium term, additional demand can be satisfied through the expansion of
upstream extractive capacity.

This issue has been overlooked in existing work. For example, the POLFREE project introduced a
standard of 70% recycled content20 for all metals (Meyer et al. 2016). Given that the current recycled
content in global steel, aluminium, and copper supply is currently around 20% (McCarthy, 2018

18
The same issue also exists in more disaggregated SAMs. For example, non-ferrous metals such as copper,
aluminium, and the suite of rare earth elements follow quite distinct supply chains. That said, in many cases,
these resources are assigned to a single economic sector: either undifferentiated mining or non-ferrous metal
processing.
19
Furthermore, the iron used in construction effectively comes from the steel sector, rather than being taken
directly from iron ore mining.
20
In this context, recycled content refers to the proportion of total metal supply that originates from secondary
facilities.

24
 ENV/WKP(2018)4

forthcoming), this would require a massive increase in secondary output, even in a relatively optimistic
scenario where total metal production remains constant. Whether there would be sufficient scrap available
to support this expansion is quite unclear.

Addressing this issue requires linking the economic flows in waste management and secondary
production sectors with their physical equivalents. Six of the studies considered in this review, do this to
some extent (Okushima and Yamashita, 2005; Masui, 2005’; Kang et al. 2006; Godzinski, 2015; Hartley et
al., 2016; Soderman et al., 2016). It is notable that each of these models is both single-region and static;
linking downstream economic and physical flows in a multi-regional dynamic context would likely be
prohibitively difficult. For one, this would require internally consistent data on waste generation and
recycling rates across different materials and countries; no such dataset is currently available. Further, for
dynamic models, this would necessitate the existence of a stock-flow model that describes the evolution of
waste generation as a function of historic capital investment and contemporary consumption21.

2.4 Decoupling mechanisms in existing studies

There are three key mechanisms that can result in decoupling of economic output from natural
resource use. Technological change leading to improved material productivity or, “producing more with
less” as this is more commonly referred to, allows a particular activity to produce additional economic
value without increasing material inputs (or substituting from materials inputs to other inputs). Slowing
resource loops with the introduction of more durable goods can have a similar effect; products that are
designed to last longer will slow the introduction of new materials to economy without necessarily
decreasing its size in value terms, at least if the higher quality of more durable goods translates into higher
prices. Substitution – either between natural resource inputs and other factors of production, primary and
secondary materials, new and remanufactured goods, or differentially materially intensive goods and
services – can also have the same result. In the case that different materials are substitutes in production
(for example between different building materials in the construction sector), different mechanisms need to
be considered in conjunction, especially by looking at whether technological change affects the relative
productivity of one material versus the other. This section discusses the extent to which each of these three
mechanisms is incorporated into existing models.

2.4.1 Representing technological change

Almost all existing models of a circular economy transition model technological change
exogenously22. Technological change consists of two distinct items: (i) innovation of new technologies that
were not previously available to any producer, and (ii) diffusion of existing technologies to new users. The
former improves the frontier of materials efficient production, while the latter reduces the gap between the
average production techniques used and the frontier.

Material saving innovations tend to appear as exogenous “manna from heaven”, rather than resulting
endogenously from particular policy changes. In practice, this entails making autonomous changes to
certain production function parameters, either in specific sectors or across the supply side of the economy.
This approach is well summarised in Tuladhar et al. (2015), which states that “production functions
typically contain parameters that govern the overall efficiency of production …” and “… increasing the
parameter for overall productivity implies that the same amount of output can be produced with a

21
Fraunhofer ISI are developing such a model for copper flows in Germany.
22
Two models – the MEMO II and MEWA models used in the DYNAMIX project – do incorporate a
mechanism allowing for endogenous technical change, but these are highly stylised models and not large-scale
CGE or ME models.

25
ENV/WKP(2018)4

proportionally smaller utilisation of every input”. The implication is that innovation is costless, and will
somehow materialise as a by-product of the policy implementation.

Implementing innovation exogenously is not a problem for no-new-policies baseline projections, and
is strongly preferable to modelling approaches that ignore technological change completely. Baseline
technology projections will reflect the most plausible consequences of existing trends and existing policies,
in the absence of new policies. It is more problematic, however, for policy scenarios. Models that
incorporate exogenous innovation ultimately make some assumption about the achievable rate of
additional material productivity growth that is triggered at no net cost by the policies that are implemented.
Such an approach denies the effort required to achieve the associated innovation. Beliefs about the rate of
‘free’ productivity growth triggered by the circular economy enabling policies differ widely across the
studies considered in this review; whether or not they are realistic is discussed in more detail in
Section 3.2.1.

One rationale for assuming zero costs for innovation stems from the idea that the most cost-effective
new technologies will deliver benefits that outweigh the costs; this feature is prominent for example in the
McKinsey marginal abatement cost curves for greenhouse gases (McKinsey, 2009). Thus, all new
technologies that have a benefit-cost ratio larger than one can be implemented without net costs. Caveats to
this reasoning are however significant. First, it ignores that an up-front cost to develop the innovative
technology must be borne, possibly by an economic agent that is not the same as the one that reaps the
future benefits. Future benefits may also not always be tangible or transferable. Secondly, negative-cost
options as identified in the literature often miss important hidden costs or other barriers to implementation
that prevent the uptake of the technology.

A related issue concerns the linkages between the assumed efficiency gains and their underlying
drivers. It is well established that R&D activity is an important long run determinant of technological
change and that, in turn, R&D expenditure is determined largely by market incentives and various aspects
of the regulatory environment. Further, the diffusion of a given innovative technology, once it has
emerged, will generally not be instantaneous, whether due to lack of knowledge on the potential new
technology, or property rights that restrict diffusion, or other reasons. Most of the models considered in this
review do not appear to represent these mechanisms adequately; productivity improvements emerge from
nowhere and diffuse throughout the economy at no apparent cost.

Apart from induced innovation, virtually all models capture technology choice and diffusion of new
technologies in some way. In most cases, technology choice boils down to changing the combination of
inputs in production, i.e. substitution (see Section 2.4.3). Essentially, firms can respond to increasing prices
for a given factor by investing in existing technologies that use that factor more efficiently, potentially at
higher overall costs, or at least at higher overall costs at the initial set of relative prices. However, in some
cases, diffusion of new technologies is represented as a pure efficiency improvement, i.e. the reduction of
one input of production per unit of output, without increased use of other inputs and represents a cost
reduction. Ideally, all available technologies should be fully specified. But given a lack of detailed data on
technologies that are not yet mainstream, and in some cases the wide range of technology options that
exist, many models simplify technology choice to a continuous function of changing shares of different
input factors, especially a substitution between resource use and capital, as described below. That said,
regional cost curves describing material “abatement” possibilities that have a firm basis in data are still
lacking, and there is currently very little information available on this subject.23

23
Specification of such material efficiency cost curves could build upon existing cost curves for emissions and
energy efficiency.

26
 ENV/WKP(2018)4

2.4.2 Representing longer lived products

The emergence of longer lived, more durable products is often highlighted as a key element of a
circular economy transition. Products that are designed to last longer will remain in use for longer, thereby
slowing the introduction of new materials into the economy24. This process is not explicitly represented in
the studies considered here; CGE and ME models are based on representations of economic flows, they
include very little stock accounting25. In practice, slower material loops could be modelled in two ways.
First, one could exogenously decrease demand for the hypothetical longer lived product; in this setting, a
new, more robustly designed product that lasts twice as long leads to a reduction in the annual demand for
the product and thus lowers total sales value by half. This would thus result in decreased revenue in the
affected sector, which may not be an accurate representation of reality. It may be that products with longer
lifetimes fetch higher prices; this would, at least partially, negate lower sales volumes. The second
approach is therefore that the longevity improvement is entirely captured in the price of the product; i.e. the
new product with a lifetime that is twice as large sells for double the price and total sales of the product
remains the same. In this case, the total demand in value terms is unaffected by the slower material loops,
and the policy has no noticeable direct effect on the economy, although there will be indirect effects caused
by the reduced demand for materials. It is clear that both approaches are caricatures of reality; robust
approaches that have a more realistic representation are, however, not yet available.

2.4.3 Representing substitution

Dynamic analysis of materials use and economic activity requires an assessment of the evolution of
production technologies over time. The simplest (and crudest) approach is to use a fixed Social Accounting
Matrix and vary only the levels of production; this Leontief approach is common in input-output analysis.
Dynamic CGE models in contrast specify substitution elasticities to accommodate shifts in input
requirements over time. Three distinct supply-side substitution possibilities are relevant in the context of a
circular economy transition. These are: (i) substitution between a particular material and other production
inputs such as capital and labour, (ii) between different types of materials, and (iii) between primary
materials (derived from natural resources) and secondary materials (derived from waste). In addition,
opportunities for substitution also exist on the demand side of the economy. Depending on relative prices
of consumption goods, households will have incentives to substitute away from material intensive goods
and services, either to other goods and services that effectively fulfil the same consumption use (e.g.
changing transportation modes to go on holidays) or to other consumption uses (e.g. reducing the number
of holidays in favour of local activities). The modelled effects of changes in relative resource and material
prices, resulting from the introduction of economic instruments for example, depends considerably on how
these possibilities are represented in a given model.

The models considered in this review generally do not represent the physical flows of materials as an
explicit factor of production, but only represent the value flows from extraction sectors to use sectors.26
Material inputs into the economy therefore originate in various extractive sectors, and usually substitute
with other intermediate goods and services according to a constant elasticity of substitution (CES)

24
This may also have potentially negative environmental effects. For example, extending the in-use period of a
particular product may not be desirable in situations where the energy (or other material) efficiency of new
equivalents is rapidly increasing.
25
Developing physical accounts of materials stocks is a necessary first step before stock accounting can be
integrated in economic models; such developments are still in their infancy, however.
26
For example, the steel sector is modelled to use “metallic mineral mining products” in production, but not iron
ore. This is in contrast to some resources – fossil fuels in energy production and land in agriculture for
example – that are usually captured with explicit volume indicators.

27
ENV/WKP(2018)4

production function that reflects a stylised summary of production technology choices that firms have.
Output in each sector is then typically produced by combining the resulting intermediate goods and
services composite with a value added plus energy composite. The position of the extraction inputs in the
production sectors needs to be specified with care, as this reflects the ease with which firms can substitute
away from material use by switching to other inputs. With the exception of some materials – fossil fuels in
energy production and feed inputs in livestock production for example – most models have not adequately
differentiated inputs from the extraction sectors to realistically represent material substitution possibilities.
Further, many models implicitly assume no substitution possibilities at all (i.e. all intermediate inputs are
required in fixed proportions in production in a CES function with the substitution elasticity equal to zero).
This may resemble a plausible lack of substitution possibilities, or a lack of data to specify more
differentiated production processes.

2.4.3.1 Substitution between materials and capital and labour

In the type of model described in the previous paragraph, there is limited scope for substitution
between materials and capital and labour; value added is combined with intermediates (of which materials
are a part), often in fixed proportions. One way to address this involves modifying the nesting structure of
the production function to allow particular materials to explicitly substitute for capital or labour.27 This
approach is common in the energy modelling literature where an energy bundle is allowed to substitute for
capital, usually with an elasticity of substitution around 0.5. Similarly, at least the more advanced models
tend to represent fertiliser inputs in agriculture in a dedicated nesting structure. One obvious issue in the
resources context is the identification of the best nesting structure (i.e. what are the direct substitutes?), and
of the equivalent elasticity parameter. Very little research has addressed this issue.

2.4.3.2 Substitution between different materials

A set-up with a detailed and flexible production function provides some scope for substitution
between different materials. For example, to the extent that timber and cement production are
differentiated in the SAM, the construction sector could effectively substitute between the two. Similarly,
there may be scope for substitutability between plastics and metals in certain manufacturing sectors. This
may be important in situations where varying tax rates are applied to different materials on the basis that
their production results in differential environmental damage. Capturing such substitution possibilities is
often hampered by a lack of detail in the representation of different sectoral inputs; any material use that
concerns switches between materials that are represented in the economic model as coming from the same
sector, e.g. different non-metallic minerals used in construction, is not reflected in a change in the
economic flows in the model, and requires either an ad-hoc adjustment of the different material uses within
the sector or a further disaggregation of the sector involved. Furthermore, there is again very little data
describing the degree of substitutability between materials in different applications, and most models
assume fixed input coefficients at this level, implying that such substitution possibilities are ignored.

2.4.3.3 Substitution between primary and secondary materials

Three main approaches have been used to represent substitution between primary and secondary
materials. The first involves representing available alternative technologies explicitly, producing the same
output as the current technology, but with different input combinations. This is the approach used in some
models for representing steel production, differentiating between primary iron ore and secondary scrap.

27
More advanced models include a sector-specific generic “natural resource” input in production that can
substitute with other production inputs.

28
 ENV/WKP(2018)4

This approach gives a lot of control over specifying the differences between the production processes, but
requires detailed information on these differences.

The second approach involves making exogenous changes to the production functions of selected
material intensive sectors such that the substitutability between primary materials and own sector output
increases. For example, Meyer et al (2016) state that the “technical coefficients are adjusted for the
manufacturing of metals, non-metallic minerals and paper. Less input is needed from the mining or forestry
sector and more input is needed from the own sector”. The underlying idea is that this mechanism
simulates higher recycling rates by allowing manufacturers to substitute their output (future waste in other
words) for materials derived from natural resources. While at first glance unintuitive, it can provide an
accurate representation of materials re-use and recycling within a sector. One difficulty involves the
attribution of value to implicit waste flows.

The third approach involves the separation of one or more dedicated waste management or secondary
material production sectors in the underlying SAM (see Sections 2.2.2 and 2.2.3, and Table 2). These
produce secondary raw materials28 which then substitute for their primary equivalents in downstream
manufacturing sectors. Different assumptions have been made about the ease of this substitution (Table 2);
some studies assume a fixed proportions production technology (Masui, 2005; Kang et al. 2006) whereas
others allow an elasticity of substitution as high as 2 or 3 (Godzinski, 2015). In reality, the actual ease of
substitution between primary and secondary materials will differ considerably as a function of the material
considered and the application in which it is used. More research is therefore needed to find plausible
values of this elasticity for different materials in different sectors; uncertainty analysis represents a logical
way forward in the meanwhile. An additional consideration in this approach relates to the representation of
secondary material supply; this may serve to constrain the potential expansion of the secondary sector (see
Section 2.3.2 for additional discussion).

Table 2. Elasticities of substitution between primary and secondary material in existing modelling

Study Model Year Type ES (prim-sec)

Masui (2005); Kang et al. (2006) AIM 2005 Secondary sectors introduced 0
Okushima and Yamashita (2005) ODIN-WR 2005 Secondary sectors introduced 0.3
Exogenous technology
Distelkamp et al. (2010) and PANTA RHEI
2010 modification, favouring 1
Meyer et al. (2016) GINFORS
own sector inputs
2 (aluminium)
Godzinski (2015) Godzinski 2015 Secondary sectors introduced
3 (steel)
Winning et al. (2017) ENGAGE-Materials 2017 Secondary sectors introduced 0.1 - 0.3 (steel)

Note: Masui (2005), Kang et al. (2006), Okushima and Yamashita (2005), Distelkamp et al. (2010), and Meyer et al. (2016) use the
same elasticity of substitution across the entire range of secondary materials that they consider.

2.4.3.5 Demand side substitution between different goods and services

All the models considered in this review allow for demand side substitution that can be stimulated by
changes in relative prices resulting from circular economy policies. Consumers are able to substitute away
from material intensive goods and services as they become relatively expensive. For example, consumption
of services – travel, education, entertainment – could increasingly substitute for consumption of goods –
vehicles, clothing, household electronics – if material taxes were to increase significantly. Clearly, the

28
And in some cases a waste disposal service.

29
ENV/WKP(2018)4

model parameter describing the responsiveness of consumers to price changes is critical; the overall cost of
a circular economy transition will be considerably higher if consumers do not consider that viable
substitution possibilities are available (or if these are not fully represented in the model).

2.4.3.4 Demand side substitution between goods and services that fulfil the same use

Circular economy policies can also trigger demand side substitution between goods and services that
effectively fulfil the same use. Examples include: (i) shifting transport modes – public transport may
become increasingly competitive with private transport, (ii) the digitalisation of entertainment – literature
and music can be consumed online without ownership of the physical products (books, records, or CDs)
themselves, and (iii) the so-called sharing economy – under-utilised accommodation, office space, and
vehicles can be leased rather than owned. The overall impact of both digitalisation and the sharing
economy on material extraction and use remains unclear. Digitalisation may decrease demand for some
materials, but demand for others – the metals used in computer servers for example – may increase.
Similarly, the sharing economy may increase the utilisation of certain goods without necessarily reducing
demand for materials – if shorter product lifetimes result in faster turnover for example. These demand
substitution possibilities are not always well represented in existing models. In particular, the sharing
economy, with its focus on long-lived under-utilised assets and consumer to consumer transactions, does
not fit well with the structure of most macroeconomic models. One possible approach is to autonomously
reduce demand for new goods, although it is not always clear that the sharing economy will have such an
effect.

30
 ENV/WKP(2018)4

3. ASSESSING THE ASSUMPTIONS AND RESULTS OF EXISTING STUDIES

3.1 Business as usual projections

Assessing the impacts of circular economy and resource efficiency enabling policies requires some
sort of counterfactual scenario: how will patterns of economic activity and material use evolve under a
business as usual scenario? In the models considered in this review29, this information is contained in
baseline scenarios that reflect certain assumptions about future population growth, productivity growth,
consumption patterns, and the relationship between economic output and material use. Each baseline
reflects a different “storyline” about the possible evolution of the main drivers of economic and resource
flows across countries and sectors. Policy scenarios are then constructed as a set of specific assumptions on
certain elements of the system, e.g. the introduction of a tax on the input of specific inputs in production
and using the model to then find the associated changes in the economic system and related material use.
The consequences of modelled scenarios involving the implementation of circular economy enabling
policies can then be determined through a comparison of the policy scenario with to the appropriate
baseline.

Two issues require particular attention in the context of baseline development for models of a circular
economy transition. First, many discussions on the circular economy emphasise the importance of wide
ranging structural changes in production and consumption patterns. To some extent, these are already
taking place; one example relates to the recent emergence of peer to peer sharing platforms. These are
changing the way that individuals consume, with potentially significant impacts for both the distribution of
income and the flow of materials. This, along with other structural shifts – an increasing preference for
services and experiences relative to goods for example – are important considerations in the context of
what a business as usual scenario might look like. There is uncertainty in how these trends will play out,
and this provides some rationale for the development of multiple visions of future “states of the world”.
Most existing models of the circular economy introduce a single baseline, and the underlying assumptions
regarding sectoral shifts are not always made explicit.

The second issue relates to what rate of decoupling between economic output and material use, both at
the sectoral and at the macro level, should be incorporated into baseline scenarios. History provides some
insight here, at least on the macro level. Today, the global economy generates around 30% more economic
value per unit of resource use than it did in 1980 (OECD, 2015b). This equates to an average decoupling
rate of around 0.8% per annum (p.a.), but with considerable variation across materials and across time.
Average decoupling rates between 1980 and 2010 were fastest for biomass (1.8% p.a.) and fossil fuels
(1.5% p.a.), slower for metallic minerals (0.2% p.a.), and non-existent for industrial minerals (-0.3% p.a.).
Decoupling rates also appear to have slowed during recent years; they were notably slower for fossil fuel
resources between 2000 and 2010, while recoupling actually took place for both metallic and construction
and industrial minerals during this period (Figure 4). These data are not always consistent with baseline
assumptions in existing models; several assume business as usual decoupling rates of 1.5% to 2% p.a.
across the economy (Meyer et al., 2016 and Hu et al., 2016), others assume no decoupling at all
(UNEP, 2017).

29
Only the dynamic models; static models generally take the counterfactual scenario to be represented by
economic and physical flows recorded in the base year of the SAM, although in a few cases a projection of the
SAM for a future year is used as basis for the analysis.

31
ENV/WKP(2018)4

The underlying rate of material productivity growth incorporated in baseline scenarios will result in
some decoupling as transformation and manufacturing sectors use resource inputs more efficiently, and as
the share of different sectoral economic activities changes over time and across regions. In other words, a
shift in economic activity from very resource-efficient production sectors in one region to more materials-
intensive production in another region will, reduce the average global materials intensity of the economy.
The question is whether this change in economic activity should then translate into decoupling on a “one to
one” basis. In other words, are there factors other than changes in the economic system that could lead to
changes in the relationship between economic output and natural resource use? One possible example
relates to recycling rates; an increase in the proportion of secondary materials in overall material inputs30
will, all else equal, generate more decoupling than that associated with economic efficiency alone.
Increased product utilisation associated with continued expansion of the sharing economy could have a
similar effect31. Again, existing studies do not always explicitly discuss the assumptions that are made for
these issues.

Figure 4. Evolution of material productivity at the global level for major material categories

Note: Material productivity = economic output per unit of materials extraction. Source: OECD (2015b) and SERI (2017).

3.2 Policy coverage

The modelled macroeconomic impacts of a transition to a more circular economy depend on the
modelling framework used for the assessment, and on the type of policies that are implemented to enable
the transition. Existing modelling focusses heavily on three main policy areas (Table 3): policies that drive
technological change and efficiency improvements, policies that drive various kinds of substitution through

30
Perhaps due, for example, to fluctuating waste supply associated with the decommissioning of the evolving
infrastructure stock.
31
Again, historic data provides some insight. Multi-factor productivity growth at the global level averaged
around 1% between 1990 and 2015 (OECD, 2015c). Decoupling rates for energy and biomass were faster
than that, while those for metallic and construction minerals were considerably slower. That either implies
differential rates of productivity growth across sectors, or the existence of other processes not captured in
productivity calculations.

32
 ENV/WKP(2018)4

changes in the relative prices of natural resources, and policies (usually relating to the introduction of
product information requirements and public education schemes) that drive changes in consumption
patterns.

There is a clear divergence between the measures that are commonly modelled, and those that are
emphasised by policy makers interested in promoting a circular economy transition. In terms of the latter,
policies such as disposal bans, recycling quantity standards, extended producer responsibility (EPR)
schemes, eco-design standards, and green public procurement are often cited (e.g., see EU Circular
Economy Action Plan). Implementing the latter policies in a CGE or ME framework can be challenging.
Some of them are so-called “soft” policies; their impact on prices are difficult to establish. Modelling the
effects of others would require the introduction of waste management or recycling sectors, which, as
discussed in Section 2.2, can be hampered by data issues. In contrast, technological change and resource
taxes have been assessed in energy models for many years; they can be relatively quickly adapted to
models of the circular economy.

33
ENV/WKP(2018)4
32
Table 3. Summary of policy coverage in selected studies

Masui Distelkamp Ekins Cambridge Godzinski Schandl Soderman UNEP Bosello Hu Meyer
(2005) et al. (2010) et al. (2012) Economics (2015) et al. et al. (2016) (2017) et al. et al. et al.
(2014) (2016) (2016) (2016) (2016)

Landfill taxation

Carbon tax

Economic Per-unit waste disposal tariff

Instruments Material consumption tax

Differentiated VAT rate

Targeted subsidies

Labelling: % raw material inputs

Labelling: recyclability/repairability
Information
Public education programs
Based
Collaborative platforms

Certification scheme: secondary inputs

EPR

Eco Ecodesign requirement: durability

Design Ecodesign requirement: repairability

Ecodesign requirement: recyclability

Recycling rate standard (on EOL-RR)

Final disposal quota

Other
Reform of end of waste rules
Regulation
Waste shipments: proper enforcement

Sharing Economy regulatory framework

Green public procurement

Public
Targeted public R&D
Provision
Services e.g., separated collection

32
Only two of the studies shown in the table implement a carbon tax. This does not reflect any perception that carbon taxes are unimportant for a circular
economy transition. Rather, some studies have chosen to focus on other instruments given the rich literature on carbon taxes in climate change literature.

34
 ENV/WKP(2018)4

3.2.1 Innovation, investment, and technological change

Technological change is the most frequently modelled “policy” in existing models. It also seems to be
the factor that drives modelling outcomes to the greatest extent; studies that assume high rates of
productivity growth induced by the circular economy transition also tend to be those that find the largest
positive GDP changes in a circular economy transition.

As highlighted in Section 2.4.1, almost all models considered in this review implement productivity
improvements from innovation exogenously; they appear as “manna from heaven” rather than resulting
endogenously from certain changes in policy. This is problematic for two main reasons. First, the reality of
the modelled outcomes is clearly questionable if they are based on naive assumptions about the rate and
cost of productivity gains. Second, this approach only provides limited insights for policymakers:
productivity gains clearly result in faster output growth, but what policies are required to realise the
emergence and dispersion of innovation in the first place?

In practice, exogenous technological change is implemented through the modification of certain

parameters of the production function, either in particularly material intensive sectors, or universally across
the supply side of the economy, while endogenous technology choice is implemented through allowing
supply-side substitution possibilities.33 In a number of studies, scenarios involving technological changes
resulting in improvements in material productivity of 2% p.a. above baseline are specified (Table 4). This
is more than twice the rate of aggregate decoupling that was observed at the global level between 1980 and
2010, and has led some authors to criticise this work on the basis that such changes would be largely
unprecedented (e.g., Lenzen et al., 2016).

The potentially available improvements in resource productivity used in the modelling literature
appear to originate from two main sources. In several projects, bottom up estimates of sector specific
efficiency gains are undertaken as a complement to macroeconomic modelling (e.g., TNO, 2012; Pfaff and
Sartorius, 2015; DYNAMIX, 2016). These usually involve detailed assessments of specific alternative
technologies, and therefore remain quite rare. In other cases, estimates of potentially available productivity
gains are either assumed, or taken from third party studies. This often involves borrowing from the energy
literature where the possibilities and constraints to efficiency gains are much better documented than for
other materials. One frequently cited publication in this context is Factor 5 (Von Weizsacker et al., 2009),
which outlines innovations that would lead to five-fold improvements in energy productivity in the
construction, steel, cement, agriculture, and transport sectors. It is unclear whether such estimates have any
relevance for other resources; energy productivity has historically grown significantly more rapidly than
that for metallic or construction minerals. Furthermore, the incentives to improve efficiency in use may
also be lower for materials that can be reused or recycled than for materials that can only be used once.

In addition to questions about the achievable rate of material productivity gains are uncertainties
about the associated costs. The treatment of the R&D and capital investment costs required for the
invention and dispersion of material saving innovations is unclear in most studies. In many cases, the

33
The rationale for making autonomous changes to production functions varies across studies. In some cases,
this approach is taken simply because representing other decoupling mechanisms can be difficult in a
macroeconomic framework (e.g. Bohringer and Rutherford, 2015; Tuladhar et al., 2015). Other studies provide
some underlying “storyline” for productivity improvements. One example involves addressing information
failures; Distelkamp et al. (2010) state that “producers do not use the best practice technology concerning
resource consumption because they do not know all the alternatives they have”. Another explanation involves
the adoption of material efficient technologies that are supposedly, at worst, cost neutral (e.g. TNO, 2012 and
UNEP, 2017). The underlying assumption is that firms would naturally adopt such technologies if existing
barriers were removed.

35
ENV/WKP(2018)4

absence of the policies required to drive technological change mean that there are no costs at all. This is
acknowledged explicitly in several cases. UNEP (2017) state that their “modelling has not fully accounted
for costs related to either resource efficiency policies or the innovation that will undoubtedly be required to
achieve the increases in resource efficiency assumed by the modelling”. Similarly, CGE modelling
undertaken for the Ellen McArthur Foundation (Bohringer and Rutherford, 2015) states that “exogenous
productivity gains, …, towards a circular economy are not traded off against the resource inputs to
facilitate the specific technological change nor the opportunity cost of choosing a different technological
future”. Finally, Schandl (2016) state that, “the material efficiency measures have been derived purely
following the logic of a physical economy and in the absence of economic considerations such as the level
of investment and changes in price”.

Table 4. Assumptions regarding average material productivity improvements in selected studies

Annual decoupling rates

Material
Baseline Scenario
All 0.8%
Biomass 1.8%
Historically
Fossil fuels 1.5% N/A
observed rates
C&I minerals -0.3%
Metal minerals 0.2%
All 0.9% 2-3%
Cambridge Biomass 0.9% ?
Econometrics Fossil Fuels ? ?
(2014) C&I minerals -0.2% ?
Metal minerals 0.3% ?
Bohringer and
All ~2% 2%
Rutherford (2015)

Schandl et al. (2016) All 1.5% 2-3%

Wu et al. (2016) All 2.0% 0.7%

UNEP (2017) All -0.04% 0.7%

Note: Historic rates are at the global level and derived from SERI material extraction data and OECD GDP data. Scenario decoupling
rates are above and beyond those contained in baselines.

There are a small number of studies that do attempt to cost exogenous material saving technological
change. Assumed improvements in material productivity are paid for through the diversion of public
resources to: (i) R&D activity (MEWA: Bosello et al., 2016), (ii) consulting services for firms (e.g.,
Distelkamp et al., 2010; Meyer et al., 2016), or (iii) investments in resource efficient capital stock
(Cambridge Econometrics, 2014). That said, it is apparent that the cost curves describing material
“abatement” possibilities are not at all well constrained. Cambridge Econometrics (2014) use data from the
energy literature as a starting point; EUR 31.4bn of annual investment is required to achieve each 1%
reduction in energy consumption in the EU. This implies that investments scale up linearly, with no decline
in the marginal productivity of additional units of investment. Other studies borrow data from detailed
engineering models in specific sectors. Meyer et al. (2016) highlights work undertaken in the German

36
 ENV/WKP(2018)4

manufacturing sector (Fischer et al., 2004 and Little et al., 2005); a 10 to 20% permanent improvement in
material efficiency is possible at a cost equivalent of one year of the resulting resource savings. Bosello et
al. (2016) generate a cumulative 10% improvement in material efficiency between 2010 and 2050 with
additional annual public R&D investment equivalent to 0.4% of GDP.

3.2.2 Material taxes and subsidies

Many of the studies considered in this review also introduce economic instruments to promote a shift
away from the use of natural resources, raw materials, or goods produced in materially intensive sectors.
Material taxes are the most frequently used instrument, but subsidies are also used in several studies to
stimulate production of secondary materials. Taxes are implemented in various ways. Some studies assume
that they could feasibly be introduced in all regions of the model, whereas others assume that sufficient
political capital will only exist in resource importing countries and regions. The treatment of the revenues
raised by these taxes also differs, but what almost all studies have in common is that they implement taxes
and subsidies under the assumption of fixed total government expenditures. Some studies model an
environmental tax reform whereby labour tax rates are lowered such that overall revenue neutrality is
achieved (e.g., Cambridge Econometrics, 2014). Others recycle revenues directly back to households in the
form of lump sum transfers (e.g,. UNEP, 2017), or use them to fund investments in public R&D (e.g.,
Bosello et al., 2016).

Material taxes are also implemented at different points in the value chain, but not on the volume of
extraction itself. The resource royalties and mineral taxes utilised in many extractive economies are not
therefore modelled directly. In most cases, taxes are instead levied on the output of specific upstream
extractive sectors. For example, Bosello et al. (2016) state that their materials tax is, “implemented as a tax
on the sales of timber and of mined materials to all other sectors”. Some studies also implement material
taxes further downstream. For example, Bosello et al. (2016), Hu et al. (2016), and Meyer et al. (2016)
introduce a tax on the output of the meat processing sector. Hu et al. (2016) attempt to introduce
consumption taxes based on the embodied materials (in RMC terms) contained in finished goods and
services, however these could not be practically implemented34.

Increasing prices should trigger substitution away from material use and provide incentives for the
development of material saving technologies. However, these mechanisms are not always well represented
in the models considered in this review. With respect to the former, and as discussed in Section 2.4.3, the
nesting structure of production functions in many models and the absence of specific secondary sectors
mean that important substitution possibilities are limited. With respect to the latter, only two of the models
considered in this review incorporate a mechanism for endogenous technological change, whereby the rate
and direction of technological progress is a function of relative prices. Taken together, this may mean that
material taxes have a muted effect relative to more realistic “real-world” scenarios where rising resource
prices stimulate substitution of secondary for primary materials and the development of material saving
innovations. On the other hand, most models tend to ignore specific barriers to changes in behaviour and
transaction costs that would imply an overestimation of the effectiveness of policies. The net effect is
therefore unclear.

34
The authors state that “an RMC based tax requires a lot of computation power, because it requires calculating
footprints using an input-output model. Every year the footprints change”. They therefore chose to “implement
the tax at the extraction level to achieve the same effect”.

37
ENV/WKP(2018)4

3.2.3 Soft policies and exogenous changes in consumption patterns

Many recent studies assume exogenous changes in the preferences of future consumers (Bosello et al.,
2016; Hu et al., 2016; Meyer et al., 2016; UNEP, 2017). In some cases, these appear out of nowhere
without any underlying explanation. In others, they are motivated by various so-called “soft” policies;
public education campaigns or product labelling schemes. The overall effect is the same, consumers
increasingly prefer less material intensive goods and services: vegetable based diets to meat based ones,
repaired and remanufactured goods to their new equivalents, and services and experiences to physical
goods more generally.

Assumptions about changes in material intensity of consumer preferences can theoretically drive
modelling results to a significant extent.35 By promoting a transition from high to low material intensity
sectors, they may have limited first order effects on aggregate economic growth while resulting in
potentially large reductions in natural resource extraction. That said, such an outcome depends entirely on
the feasibility of the underlying assumptions about changing future preferences and associated
consumption patterns. Furthermore, by changing preferences, the scenario analysis assumes that there are
no welfare costs associated with these changes; in reality, people change their behaviour not because
they’re forced to, but because they prefer it. These caveats should be clearly presented.

3.3 Economic consequences projected by the existing studies

The impact of circular economy and resource efficiency enabling policies are presented across various
metrics in existing studies. In terms of macroeconomic indicators, almost all studies considered in this
review report changes in GDP. Fewer studies present changes in sectoral value added, domestic trade
balances, employment, or household income inequality. Similarly, changes in aggregate resource
extraction and use are often reported, but are not necessarily split into various resource categories (metallic
minerals, energy carriers etc.). The following section focusses on the results of existing studies in terms of
GDP and aggregate resource extraction, the metrics for which data is most readily available. This analysis
comes with the caveat that the results of individual studies are not directly comparable, since the modelling
assumptions as well as the characteristics of the implemented policies differ.

Additional indicators would be worthwhile, among which environmentally-adjusted or “green” GDP,

which none of the studies reported. Adjusting GDP for changes in natural capital stocks and ecosystem
services has been advocated, for instance in (Boyd, 2007), or as a part of the debate around green growth
(Stiglitz, Sen and Fitoussi, 2008; Reilly, 2012). Although measuring green GDP was not widely attempted
in an ex-ante modelling framework, it could be considered in future assessments. Beyond natural capital
accounting, assumptions concerning the role that natural capital plays in supporting economic growth can
influence model results. Most models do not represent natural resources endowment, or the waste
assimilation function or ecosystem services that they provide. Key barriers to representing feedbacks on
future economic growth following a change in the availability or quality of these resources36 include
uncertainties about: (i) resource endowments, (ii) the responsiveness of natural systems to perturbations,
and (iii) the monetary value of environmental damages. That said, several recent OECD publications have
assessed the costs of inaction on urban air pollution and greenhouse gas emissions, and found that they
could reach 1% and 2% of GDP by mid-century respectively (OECD, 2016c and OECD, 2017b).

35
In addition, all sorts of policies can influence consumer behaviour with unchanged underlying preferences by
changing relative prices or regulating specific behaviour.
36
Such as would be likely to occur if ongoing resource extraction results in the depletion of near-surface low-
cost mineral deposits, or if capital flows are diverted away from productive investments in order to finance the
provision of the ecosystem services currently available at no private cost.

38
 ENV/WKP(2018)4

3.3.1 Macroeconomic impacts

With respect to the baseline scenarios used, all of the studies considered in this review generate lower
natural resource extraction through the implementation of circular economy policy instruments. This is
generally achieved in a process of relative decoupling; future aggregate economic growth is either
unaffected (i.e., continues as in business as usual), or shows, often significant, improvements (Figure 5).
Only two studies conclude that a circular economy transition could have a significant depressing effect on
economic growth; Schandl et al. (2016) find that economic output would be 1.6% below baseline by 2050
while Hu et al. (2016) find that it would 0.6% below baseline. A number of other studies find reductions in
GDP of less than 0.1%.

These results provide policymakers with some cause for optimism. They suggest that a transition to a
(broadly defined) circular economy might take place without significant negative impacts on aggregate
economic growth and employment. That said, there are three important caveats to this conclusion. First, the
results of ex ante modelling assessments are not predictions of the future. Rather, they represent one
possible evolution of the modelled (endogenous) variables that is consistent with the assumptions
contained in the model, and about the enabling policies themselves (for example, the assumptions
regarding the achievable rate and cost of technological change). Results need to be interpreted with this in
mind (see 3.3.3 and 3.3.4 for additional discussion). Second, that natural resource extraction can apparently
be reduced without adverse effects on economic output suggests that there is significant inefficiency in the
current economic system. What these inefficiencies are, and whether they can be practically addressed, is
not always discussed, but should be in future work. Third, what is positive for the aggregate economy may
not necessarily be so for all of the parts. Any transition to a circular economy will likely lead to particular
countries, sectors, and types of skills doing better than others.

Possible distributional effects are important; concerns about the potential “losers” can hold back a
transition in the first place. One consistent finding in existing modelling is decreased activity in upstream
extractive sectors – mining, oil and gas, agriculture, fishing, and forestry – and material transformation
sectors – metal smelting and fuel refining. Countries specialising in these activities, and workers employed
in them, are likely to emerge worse off from a circular economy transition.37 For example, in their Global
Cooperation scenario, Meyer et al. (2016) find that Russia (-27.7%), Brazil (-16.5%), and Canada (-7%)
would experience significantly lower GDP relative to business as usual to 2050 (global GDP increases by
5.2% in this scenario). Other sectors may expand when circular economy enabling policies are
implemented. Modelling shows that manufacturing activity often increases as technological change
proceeds, waste management and recycling activities grow as manufacturers substitute secondary for
primary materials, and the services sector expands as consumers substitute services for goods.

37
Social and labour market policies can be considered to alleviate the worst impacts on local communities that
rely heavily on these shrinking economic activities.

39
ENV/WKP(2018)4

Figure 5. Headline modelling results in the studies considered in this review: GDP with respect to baseline

Note: Geographic coverage – national, regional, global – refers to the area that results were reported for. Some models have global
coverage but only report results for a particular region.

3.3.2 Impacts on resource extraction

The results of existing studies suggest that the implementation of circular economy policies can
reduce resource extraction by up to 80% relative to business as usual (Figure 6). They also highlight the
existence of a trade-off, at least in the time horizons considered, between higher rates of economic growth
on one hand, and lower resource extraction on the other. As shown in Figure 6, a number of assessments
indicate that circular economy policies (almost exclusively those relating to productivity improvements)
can boost economic growth, but without any significant reduction in resource extraction. This is probably
not surprising; productivity improvements are critical for growth in the long-run, but can also trigger
strong rebound effects as household incomes grow. A number of other assessments shown in Figure 6 find
that circular economy policies can generate large reductions in resource extraction, but without any
significant boost to economic growth. This is probably largely due to additional costs that are imposed on
economies due to the material taxes implemented in these studies.

40
 ENV/WKP(2018)4

Figure 6. Headline modelling results in the studies considered in this review: GDP and resource extraction
with respect to baseline

3.3.3 Assumptions about the implemented policies can determine results

Four of the studies considered in this review find that a circular economy transition could result in
GDP gains in excess of 5% by 2050 (Distelkamp et al. 2010; Bohringer and Rutherford, 2015; Bosello et
al. 2016; Meyer et al. 2016). With the exception of Bosello et al. (2016), which represents technological
change endogenously, each of these studies introduce exogenous changes to either production technologies
or consumer preferences. The underlying assumptions about the achievable rates and costs of these
changes appear to drive modelling outcomes to a significant extent. For example, Bosello et al. (2016) state
that “while the effect of increasing R&D expenditure is notoriously difficult to assess, this instrument was
judged as being a necessary precondition for the success of all others”. More generally, assumptions
concerning the rate and cost of technological and preference changes are not always well presented and
sensitivity analysis of changes in key parameters are often not undertaken. These factors, taken together,
make it difficult to assess the robustness of the modelled macroeconomic gains.

Many studies also introduce material taxes to promote a shift away from the use of natural resources,
raw materials, or goods produced in materially intensive sectors. The size of these taxes, and the treatment
of the resulting tax revenues, differ considerably across studies and often appear to drive diverging results.
In their central scenario, Cambridge Econometrics (2014) introduces raw materials taxation and recycle the
revenues back to consumers through lower labour taxes. This results in a small positive GDP impact, but
which the authors emphasise is conditional on the revenue recycling assumption. Without this, “the net
positive GDP impacts are much smaller and become negative over time”. Other studies use revenues
originating from material or environmental taxation to finance green R&D programs. One example is
provided by Bouzaher et al. (2015), which implements taxes on emissions (PM10 and CO2), solid waste,
and waste water, and recycles them in two main ways. In the first, the resulting tax revenues are used to
reduce corporate taxation elsewhere in the economy; this results in a large (~14%) reduction of GDP below

41
ENV/WKP(2018)4

baseline. In the second approach, these revenues are used to finance green capital investment and R&D;
the resulting technological change results in a small increase in GDP (~2%) above baseline.38

Whether a particular enabling policy is implemented individually or as one element in a broader

policy mix also influences results, as optimal reactions to a policy mix differ from those to the sum of
individual policies. In particular, the technological change assumed in many models drives material
productivity improvements, but the resulting reduction in material extraction is often, at least partially,
offset by rebound effects. Several studies therefore emphasise the importance of coupling R&D policies
with those that increase the relative prices of natural resources and primary materials. Bosello et al. (2016)
state that “although supporting material efficiency R&D might seem the “optimal” policy to foster absolute
decoupling, it should be accompanied by further regulation or incentives limiting material use or
promoting dematerialised services”.

3.3.4 Assumptions in the models themselves also determine results

Assumptions inherent to the models themselves can also drive modelling results. This is well
illustrated in studies that use different models to assess the impact of a particular circular economy policies
or policy mixes. For example, Bosello et al. (2016) introduce a material tax on the output of the major
extractive sectors, and find quite different results across the models used. Implementation of the tax in the
ICES model generated a 5% reduction in GDP by 2050 relative to business as usual whereas
implementation in the MEMO II and MEWA models generated increases of 2% and 6% respectively
(Figure 7). One key difference between the models is that the latter two include a mechanism for
endogenous technical change; the incentives created by increasing material prices can stimulate the
development of material saving technologies. The resulting productivity and GDP growth therefore occurs
in MEMO II and MEWA, but not in ICES39.

Figure 7. Effects of a materials tax in the ICES, MEMO II, and MEWA models

38
There are also some other differences between the scenarios that may affect the GDP impacts, but the authors
suggest the recycling mechanism is the dominant factor here.
39
Material tax revenues are also treated differently across these models. They are rebated in a lump-sum transfer
to households in ICES and recycled through reduced labour taxes in MEWA. MEMO II rebates 50% of tax
revenues and recycles the remaining 50%. Some of the variation in modelling results is likely to reflect this.

42
 ENV/WKP(2018)4

A second example is provided by the POLFREE project, which utilised two models – EXIOMOD
CGE and GINFORS ME – to assess similar circular economy policy mixes. Strongly contrasting results
emerged from each model. In the GINFORS simulations, under their Global Cooperation scenario40, highly
ambitious climate and resource extraction targets were achieved with an associated 5.2% increase in world
GDP by 2050. When implemented in EXIOMOD, the same set of policies were unable to achieve the
desired environmental targets, and resulted in small decline in GDP (-0.6%) relative to baseline. Meyer et
al. (2016) highlights underlying differences in the theoretical foundations of the two models as a key
reason for this divergence. The Keynesian assumptions embedded in GINFORS – a demand driven
economy with underutilised capacity and supply clearing the market – can generate strong multiplier
effects. Essentially, new investment, either originating directly from public spending or indirectly from
additional private saving associated with efficiency gains, spurs job creation, resulting in higher aggregate
income, additional consumption, and ultimately higher aggregate output. This mechanism is a key reason
why Walz (2011) concludes that “macro-econometric models tend to assess the effects of environmental
protection policy slightly less pessimistically than the equilibrium models. This is particularly distinct in
the more Keynesian oriented models…”.

One key difference is that some models tend to assume new investments do not crowd out
investments (nor employment) in other sectors, while others assume full crowding out. In the latter case, a
change in investment from a highly productive but polluting sector to investments in material savings will
lead to a negative effect on the macro economy, while in the models without crowd-out the new investment
by definition boost income. Whether such assumptions are more valid in the longer run, when there is
ultimately a finite supply of productive labour and capital, is questionable. The effects of this assumption
are stronger on GDP than on consumption, at least in models than maintain an income constraint
(consumption plus savings equal income, and overall savings equal investments), implying that it is more
useful to look at consumption impacts than GDP impacts.

40
Which involves a broad set of economic and regulatory instruments implemented across each of the regions in
the model.

43
ENV/WKP(2018)4

4. CONSIDERATIONS FOR FUTURE MACROECONOMIC ASSESSMENTS

Ex-ante, economy-wide, quantitative dynamic models appear to be the tool best suited for assessing
the likely macroeconomic consequences of a circular economy transition. Ex-ante modelling analysis is
more appropriate than ex-post empirical analysis because many aspects of a circular economy transition
are “out of sample”; the policy mixes that are typically suggested have not been widely implemented
historically.41 Economy-wide modelling appears indispensable because a circular economy transition will
involve spill-over and interaction effects between sectors, thereby leading to structural shifts across the
economy across sectors and regions. The quantitative nature of these tools helps paint a comprehensive
picture of all the complex systemic interactions brought about by the circular economy transition, and
clarifies which mechanisms lead to significant changes in the socioeconomic system. Dynamic modelling
is preferable as the transition to a circular economy will take place in parallel with other major
socioeconomic trends such as digitalisation and automation, and future resource use will significantly
differ from current patterns.

Two specific categories pertaining to this class of models – macro-econometric (ME) and computable
general equilibrium (CGE) models – have been the main focus of this review. These are only very recently
being employed more widely in the context of a circular economy transition; well over half of the known
literature has been published since 2015. The main message from the studies reviewed in this paper is that
lower rates of resource extraction and use can be achieved with associated increases in aggregate economic
output. However, this review identified the issue that the robustness of this key conclusion crucially
depends on assumptions about the enabling policies implemented, and about mechanisms in the models
themselves. In this respect, there remains considerable room for improvement. The following sections offer
four concrete recommendations for future modelling assessments.

Specific recommendations on the choice of the regional aggregation, incl. single-country versus
global modelling, cannot be made, as there are merits to having a variety of in-depth national studies in
combination with global assessments that provide the international context. Modelling at different regional
scales can thus nicely complement and inform each other.

Similarly, it is hard to draw conclusions on which policy instruments need to be modelled. A wide
variety of instruments have been included in existing studies (see Table 3), although even these do not
include all instruments contemplated by policy makers. It is clear that the ‘soft policies’, such as extended
producer responsibility, green public procurement, eco-design and labelling, are important from a policy
perspective, while also being the hardest to adequately represent in modelling exercises due to a lack of
data on the costs and effectiveness of these policies and their often local and heterogeneous nature.
Nonetheless, future modelling studies would do well to study broad policy packages and not limit
themselves to taxes and subsidies, as that would provide a partial and biased view on the potential
effectiveness and costs of the transition to a circular economy.

41
This does not imply that empirical analysis on this topic should not be pursued. Empirical studies, that will
probably at least initially be of a more local nature, can complement and eventually strengthen modelling
exercises.

44
 ENV/WKP(2018)4

4.1 Baseline development

There is considerable uncertainty surrounding the evolution of natural resource extraction and use in
the coming decades. In a baseline development without new policies, there are still several mechanisms
that will influence material use that are fundamentally uncertain. One uncertainty relates to the base year
use of resources; although there are different databases, they often contain considerably different numbers.
Second, a crucial source of uncertainty concerns the evolution of economic activity over time and the
associated changes in economic structures over time. Finally, the uncertainty on the rate of decoupling
resource use from economic activity, which relies on the evolution of alternative production technologies,
is a critical assumption that affects future resource use.

Given the strong links between economic activity and resource use, the evolution of economic
structures and its impact on natural resource use is vital for projecting future resource use. The baselines
used in some of the existing models assume decoupling rates that are considerably higher than those
observed historically. Historic trends give some indication of what may be possible in the future, but these
trends are insufficient by themselves to assess the likelihood and possible impacts of structural changes in
production and consumption patterns. The importance of these changes is stressed in many discussions of
the circular economy. To some extent, they are playing out already with the emergence of the sharing
economy and product as service business models, as well as through major socioeconomic trends such as
globalisation. Other factors, such as how consumer preferences are likely to evolve in the absence of new
policies, have important implications for resource use but are surrounded by uncertainty. Future
assessments would benefit from clearly identifying the driving assumptions behind future resource use to
assess the plausibility of these baseline developments.

A possible but more complex route to understand the range of plausible projections of future resource
use and the circularity of the economy could be the creation of multiple baseline scenarios, with a range of
assumptions about the evolution of key drivers and the impacts of structural change. This would allow the
macroeconomic consequences of a circular economy transition to be assessed in different future states of
the world; it is clear that projections of future resource use will substantially differ across each of these.
Such scenario analysis can take the form of a systematic sensitivity analysis where one major driver at a
time is varied, or more complex sets of assumptions that together provide a consistent world view (a
storyline). An existing example of the latter methodology is the design of the Shared Socioeconomic
Pathways (SSP scenarios) developed in the climate change community, where the deep drivers of climate
change mitigation and adaptation serve as basis for creating five contrasting storylines to assess the
consequences of policy. However, this route involves an extensive analysis of the produced scenarios and
may not improve clarity of the policy insights.

Presenting multiple baselines has at least two major advantages. First, the large underlying
uncertainties in projecting future materials use are clearly laid out. Secondly, when analysing the costs and
benefits of specific policies across multiple baseline scenarios, no-regret options that are worthwhile
regardless of which scenario materialises can be identified. Moreover, robust policy interventions can be
identified that work well in all scenarios. More advanced modelling techniques will also allow the
identification of hedging strategies, i.e. the specification of policies that maximise the expected net benefits
of action, but these are difficult to implement in large-scale models.

To understand the consequences of domestic consumption patterns and resource policies at a global
scale, information is needed on how economic activities are internationally linked in global value chains
and bilateral trade flows. Building on such information, a footprint type of analysis can be made by looking
at the embedded materials use in imported goods and services, i.e. moving from Domestic Materials
Consumption to Raw Materials Consumption. This is however very data and computationally intensive, as
discussed in Section 2.3. Furthermore, in a global analysis overall materials consumption is identical.

45
ENV/WKP(2018)4

4.2 Modelling technological change

The direction and speed of technological change will be a critical driver of any future resource
decoupling. More efficient technologies for upstream processes will mean that a greater proportion of
metal can be extracted from ores, and goods could be built with less material (e.g. cars). Innovative new
product designs will provide equal or higher levels of functionality without additional material inputs. The
emergence of new business models will extend the in-use lifetime of products, others will increase their
utilisation rates. Further downstream, the development of improved material sorting technology will allow
for more efficient separation of different materials in the waste stream.

In almost all of the studies considered in this review, modelled policy scenarios assume a significantly
higher rate of productivity growth than in the respective baselines. Overall technological change is
typically implemented exogenously – there are no linkages with the underlying drivers of innovation.
Furthermore, in many cases, these productivity gains appear for free; the costs of developing and adopting
resource efficient technologies are ignored. Finally, where technological change is costed, it is often done
so linearly: there is no decrease in marginal productivity gains with each unit of additional investment.
Some studies do, however, explicitly represent competition between existing and future technologies, and
substitution towards more efficient technologies. Technology-specific modelling gives a richer description
of the technological dynamics, but comes at the price of an increased need in data (and is hindered by the
fact that innovations are difficult to predict).

Future macroeconomic assessments of a circular economy transition would benefit from addressing
both technology choice and technological change, even if a detailed description of endogenous innovation
is currently out of reach for large-scale models. First, assumptions about the potentially available rates42 of
future resource productivity growth need to be considered. For already existing technologies, the
development of bottom-up technology models would allow the resource saving potential associated with
their widespread adoption to be better constrained. In addition, any assumptions about the emergence of
new technologies – those that are currently unknown – should be clearly presented with caveats. Second,
the costs associated with the development and adoption of resource efficient technologies should be
incorporated into modelling frameworks. These are likely to have a strong influence on technology
adoption decisions. They will also have consequences for other parts of the economy; investments in
resource efficiency may divert funds that would otherwise be used elsewhere. Only when such costs are
considered will the macroeconomic consequences of circular economy enabling policies reflect the net
gains from the policies. The main hurdle here is the valuation of those costs, which ultimately require
detailed, and often time consuming, bottom up technology modelling to establish in detail.

4.3 Representing substitution

Substitution away from natural resource inputs is a key element in most conceptions of a circular
economy transition. Indeed, the main purpose of the resource or material taxes that are often discussed in
this context is to stimulate this substitution. Three main supply side substitution possibilities exist: (i)
between a particular material and other production inputs such as capital and labour, (ii) between different
types of materials, and (iii) between primary materials (derived from natural resources) and secondary
materials (derived from waste). Substitution opportunities also exist on the demand side of the economy.
Firms and households can substitute away from material intensive goods and services, either by consuming
a different bundle of products or by consuming products that fulfil the same use, but in a less material
intensive way. Any future substitution will be driven primarily by changes in relative prices and by the

42
This will likely feature some divergence across different resource types; assigning rates derived from the
energy literature to other resources will probably introduce significant biases.

46
 ENV/WKP(2018)4

ease with which these inputs can substitute for each other. This suggests that models that capture the
sectoral interactions and responses to changes in relative prices, such as CGE or ME models, are
particularly well-suited to study the economic effects of circular economy enabling policies.

Many of the models considered in this review do not allow for one or more of the above substitution
possibilities. Substitution between natural resources and capital or labour is largely precluded by the
nesting structure of most production functions; resource inputs can substitute with other intermediate
inputs, but not (directly) with various components of value added. Similarly, substitution between primary
and secondary materials is precluded by the absence of secondary sectors in most SAMs.

Future macroeconomic assessments of the circular economy would benefit from addressing both of
these issues. The energy modelling literature offers some possible direction in the former case; materials
could be allowed to trade off with capital and labour in much the same way that energy does in energy-
focussed models. In the latter case, substitution between primary and secondary materials could be
achieved in a variety of ways, but ideally requires the introduction of a secondary production sector (for
each material of interest) into the SAM. A number of existing single-region models do this. The challenge
to do so is the collation of comprehensive data for assessing the potential for substitution. This difficulty is
magnified in a multi-region framework given the differences in economic and biophysical circumstances in
different countries.

A further complication in adequately representing substitution between virgin and secondary materials
is the need to specify supply constraints for secondary materials. Most models do not have any capacity
constraint on the secondary sector, which implies an overestimation of the potential substitution effect. In
some models, this is overcome by using CES functions with a constant elasticity that inherently assume
that substitution becomes more difficult once the share of secondary materials increases. This should
ideally be coupled with constraints on the supply of secondary materials. The lack of full stock-flow
accounting (as highlighted in Section 2) hampers an endogenous representation of the supply of secondary
materials over time, and therefore the most advanced models apply ad-hoc rules on secondary material
supply.

Representing these substitution possibilities in a detailed manner in models will thus require a large
volume of data for the various, materials, material uses, sectors regions. This data is not yet available43. An
alternative modelling approach (especially when data is scarce) is the use of CES functions to represent the
different inputs (materials, capital, energy, and labour) associated with different technologies. One initial
approach may be to use sensitivity analysis to frame the range of likely outcomes.

4.4 Modelling “soft” enabling policies

There is a clear divergence between the circular economy enabling policies emphasised in policy
forums and those implemented in modelling assessments. General discussions about how to enable a
circular economy transition tend to highlight an extensive list of policy measures, many of which focus on
promoting changes in consumer behaviour (so-called soft policies). This is in contrast to the studies
considered in this review; most tend to focus on a relatively narrow policy mix consisting mainly of
technological change and upstream materials taxes. Many soft policies – those that do not directly affect
prices – are significantly under-represented. Important examples include extended producer responsibility
(EPR) schemes, public education campaigns, eco-design standards, and green public procurement.

43
Particularly given that the appropriate elasticity will vary across different resources and materials, and across
the different applications they are used in.

47
ENV/WKP(2018)4

Future macroeconomic assessments of a circular economy transition would benefit from considering a
broader set of instruments. If nothing else, this would provide policymakers with insights into the cost
effectiveness of meeting resource efficiency targets with different policy mixes. Implementing soft policies
in a large-scale modelling framework will clearly be challenging but, as highlighted by Dubois (2015),
could be done through exogenous changes to production and utility function parameters. For example, the
effect of eco-design standards could be modelled by modifying the production function of manufacturing
sectors such that raw material inputs are used more efficiently (longer lived products reduce demand for
material inputs). Similarly, a future shift in consumer behaviour from ownership to access, perhaps due to
some public education campaign, could be represented by modifying the utility function of the
representative consumer such that demand for services increases.44 Doing so however is extremely difficult
since the effectiveness of soft policies is not easily measured and thus the translation of such measures in
physical and economic terms rests on many assumptions (many of them subjective and dependent on the
policy analyst). In all cases, the motivation and assumptions behind these modifications should be clearly
stated.

44
For example, in the case of car sharing, the assumption is that the underlying preferences for mobility do not
change because of the policy, but the expressed preference for specific commodities to satisfy this need for
mobility do; thus, the modelled preferences for commodities will shift, thereby generating a change in
behaviour.

48
 ENV/WKP(2018)4

BIBLIOGRAPHY

Bartelings, H. et al. (2004), ‘Modelling market distortions in an applied general equilibrium framework:
the case of flat-fee pricing in the waste market’, in: J. van den Bergh and M. Janssen (eds),
‘Economics of industrial ecology: materials, structural change, and spatial scales’, MIT Press,
Cambridge (MA).

Bastein, T. et al. (2013), Opportunities for a Circular Economy in the Netherlands, The Netherlands
Organisation for Applied Scientific Research (TNO),
www.government.nl/documents/reports/2013/10/04/opportunities-for-a-circular-economy-in-the-
netherlands.

Bocken, N. et al. (2016), ‘Product design and business model strategies for a circular economy’, Journal of
Industrial and Production Engineering, 33(5), pp.308-320.

Bockermann, A. et al. (2004), ‘Modelling sustainability: comparing an econometric (PANTA RHEI) and a
systems dynamics model (SuE)’, Journal of Policy Modelling, 27, pp. 189-2010.

Boyd J. (2006), Nonmarket benefits of nature: What should be counted in green GDP?, Ecological
Economics, Volume 61, Issue 4, 2007, Pages 716-723, http://dx.doi.org/10.1016/j.ecolecon.2006.06.016.

Bohringer, C. and T. Rutherford, (2015), The Circular Economy – An Economic Impact Assessment,
Report to SUN-IZA, https://sunstiftungsfonds.files.wordpress.com/2015/06/report-circular-
economy.pdf.

Bosello, F. et al. (2016), Report on Economic Quantitative Ex-Ante Assessment of DYNAMIX Policy Mixes,
DYNAMIX Deliverable 6.2, http://dynamix-project.eu/quantitative-economic-ex-ante-assessment-
dynamix-policy-mixes.html. Milano: Fondazione Eni Enrico Mattei.

Bouzaher, A. et al. (2014), ‘How to go green: a general equilibrium investigation of environmental policies
for sustained growth with an application to Turkey’s economy’, Letters in Spatial and Resource
Sciences, 8(1), pp.49-76.

Cambridge Econometrics (2014), Study on modelling of the economic and environmental impacts of raw
material consumption, European Commission Technical Report 2014-2478.

CIRAIG (2015), Circular Economy: A Critical Literature Review of Concepts,

www.ciraig.org/en/v.php?id=396&locale=en&year=2015&type=2.

Coulomb, R., et al. (2015), "Critical Minerals Today and in 2030: An Analysis for OECD Countries",
OECD Environment Working Papers, No. 91, OECD Publishing, Paris,
http://dx.doi.org/10.1787/5jrtknwm5hr5-en.

Distelkamp, M. et al. (2010), Quantitative and Qualitative Effects of a forced Resource Efficiency Strategy.
Wuppertal Institute.

Dubois, M. (2015), Economic Benefits of the Circular Economy – A Critical Review of the Literature and
Opportunities for CGE Modelling, document for official use ENV/EPOC(2015)14, OECD, Paris.

49
ENV/WKP(2018)4

Ellen McArthur Foundation (2013), Towards the Circular Economy, Volume 1,

www.ellenmacarthurfoundation.org/assets/downloads/publications/Ellen-MacArthur-Foundation-
Towards-the-Circular-Economy-vol.1.pdf.

FMEAE (2015), Germany makes resource efficiency a priority of its G7 Presidency, Press Release from
German Federal Ministry for Economic Affairs and Energy,
www.bmwi.de/Redaktion/EN/Pressemitteilungen/2015/20150312-deutschland-macht-
ressourceneffizienz-zu-einem-schwerpunkt-seiner-g7-praesidentschaft.html.

G7 (2016), Toyama Framework on Material Cycles, www.mofa.go.jp/files/000159928.pdf.

Giljum, S. et al. (2014), ‘Material Footprint Assessment in a Global Input-Output Framework’, Journal of
Industrial Ecology, 19(5), pp.792-804.

Godzinski, A. (2015). Circular economy and energy: a computable general equilibrium approach. French
Ministry of Environment, https://afse2015.sciencesconf.org/61878/document.

Griffith, M. (2013), "Measures Supporting Minerals and Primary Metals Extraction and Processing: Case
Study: Australia", OECD Trade and Environment Working Papers, No. 2013/03, OECD Publishing,
Paris, http://dx.doi.org/10.1787/5k43n12wlnr0-en.

Hartley, F. et al. (2016), The general equilibrium impacts of monetising all waste streams in South Africa,
Energy Research Centre, University of Cape Town.

Hu, J. et al. (2016), Report about integrated scenario interpretation EXIOMOD / LPJmL results,
POLFREE Deliverable D3.7B.

International Energy Agency (IEA) (2015), World Energy Outlook 2015, OECD Publishing, Paris,
http://dx.doi.org/10.1787/weo-2015-en.

Jaffe, A. and R. Stavins, (1994), ‘The energy-efficiency gap: what does it mean?’ Energy Policy, 22(10),
pp.804-810.

Kang, S. et al. (2006), National CGE modeling for Resource Circular Economy, Korea Environmental
Institute.

Lenzen, M. et al. (2016), ‘Reply to Schandl et al., 2016, JCLEPRO and Hatfield-Dodds et al. 2015, Nature:
How challenging is decoupling for Australia?’ Journal of Cleaner Production, 139, pp.796-798.

Masui, T. (2005), ‘Policy evaluations under environmental constraints using a computable general
equilibrium model’, European Journal of Operational Research, 166(3), pp.843-855.

Mathews, J. and H. Tan, (2016), ‘Circular economy: Lessons from China’, Nature, 531(7595), pp.440-442.

McCarthy, A. (2018, forthcoming), “Mapping Primary and Secondary Metal Production”, OECD
Environment Working Papers, OECD Publishing, Paris, http://dx.doi.org/10.1787/19970900.

McKinsey (2009), Pathways to a low-carbon economy: version 2 of the global greenhouse gas abatement
cost curve, McKinsey & Company.

Meyer, B. et al. (2016), Report about integrated scenario interpretation GINFORS / LPJmL results,
POLFREE Deliverable D3.7A.

50
 ENV/WKP(2018)4

MIE (2016), A Circular Economy in the Netherlands by 2050, Netherlands Ministry of Instructure and
Environment, www.government.nl/documents/leaflets/2016/09/22/a-circular-economy-in-the-
netherlands-by-2050.

MoE (2013), Fundamental Plan for Establishing a Sound Material-Cycle Society, Japan Ministry of
Environment, www.env.go.jp/en/recycle/smcs/3rd-f_plan.pdf.

Nuss, P. and M. Eckelman, (2014), ‘Life Cycle Assessment of Metals: A Scientific Synthesis’, PLoS ONE
9(7).

OECD (2017a), "Fisheries: Fisheries support stimates (Edition 2017)", OECD Agriculture Statistics
(database), http://dx.doi.org/10.1787/62e7ec95-en.

OECD (2017b), Investing in Climate, Investing in Growth, OECD Publishing, Paris,

http://dx.doi.org/10.1787/9789264273528-en.

OECD (2016a), Policy Guidance on Resource Efficiency, OECD Publishing, Paris,

http://dx.doi.org/10.1787/9789264257344-en.

OECD (2016b), Agricultural Policy Monitoring and Evaluation 2016, OECD Publishing, Paris,
http://dx.doi.org/10.1787/agr_pol-2016-en.

OECD (2016c), The Economic Consequences of Outdoor Air Pollution, OECD Publishing, Paris,
http://dx.doi.org/10.1787/9789264257474-en.

OECD (2015a), OECD Companion to the Inventory of Support Measures for Fossil Fuels 2015, OECD
Publishing, Paris, http://dx.doi.org/10.1787/9789264239616-en.

OECD (2015b), Material Resources, Productivity and the Environment, OECD Green Growth Studies,
OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264190504-en.

OECD (2015c), The Future of Productivity, OECD Publishing, Paris,

http://dx.doi.org/10.1787/9789264248533-en.

OECD (2012), OECD Environmental Outlook to 2050: The Consequences of Inaction, OECD Publishing,
Paris, http://dx.doi.org/10.1787/9789264122246-en.

Okushima, S. and H. Yamashita, (2005), ‘A General Equilibrium Analysis of Waste Management Policy in
Japan’, Hitotsubashi Journal of Economics, 46(1), pp. 111-134.

Pfaff, M. and C. Sartorius, (2015), ‘Economy-wide rebound effects for non-energetic raw materials’,
Ecological Economics, 118, pp.132-139.

Reilly JM, Green growth and the efficient use of natural resources, Energy Economics, Volume 34, 2012,
Pages S85-S93, http://dx.doi.org/10.1016/j.eneco.2012.08.033.

Rizos, V. et al. (2017), The Circular Economy A review of definitions, processes and impacts, CEPS
Research Report, 2017(08), www.ceps.eu/publications/circular-economy-review-definitions-
processes-and-impacts.

Schaffartzik, A. et al. (2014), Environmentally Extended Input-Output Analysis, Social Ecology Working
Paper 154, Alpen-Adria University, www.aau.at/wp-content/uploads/2016/11/working-paper-154-
web.pdf.

51
ENV/WKP(2018)4

Schandl, H. et al. (2016), ‘Decoupling global environmental pressure and economic growth: scenarios for
energy use, materials use and carbon emissions’, Journal of Cleaner Production, 132, pp.45-56.

SERI, (2017), Data portal, www.materialflows.net/data/data-download/.

SITRA (2015), The Opportunities of a Circular Economy for Finland, SITRA Studies 100.

Söderman, L. et al. (2016), ‘Integrated Economic and Environmental Assessment of Waste Policy
Instruments’, Sustainability, 8(5), p.411.

Sommer, M. and Kratena, K, (2017), ‘Recycling of Steel, Aluminum, Paper and Glass and Their Energy-
Economic Scope, http://nachhaltigeswirtschaften-soef.de/sites/default/files/Meyer.pdf.

Stegeman, H, (2015), The Potential of the Circular Economy, Rabobank Special Publication,
https://economics.rabobank.com/publications/2015/july/the-potential-of-the-circular-economy/.

Stiglitz J., Sen A. and Fitoussi J.P. (2008), Report by the Commission on the Measurement of Economic
Performance and Social Progress, "Commission on the Measurement of Economic Performance and
Social Progress".

TNO (2012), Potential approaches for modelling resource efficiency related to buildings and
infrastructure. Reflections on a hybrid set-up, European Commission report
ENV.F.l/ETU/2011/0044.

Tuladhar, S. et al. (2015), An Economic Analysis of The Circular Economy, NERA Consulting,
www.gtap.agecon.purdue.edu/resources/download/8300.pdf.

United Nations Environment Programme (UNEP) (2017), Resource Efficiency: Potential and Economic
Implications. A report of the International Resource Panel, UNEP.

US EPA (2015), U.S. EPA Sustainable Materials Management Program Strategic Plan, United States
Environmental Protection Agency, www.epa.gov/sites/production/files/2016-
03/documents/smm_strategic_plan_october_2015.pdf.

Von Weizsacker, E. et al. (2009), ‘Factor Five’, Routledge, ISBN 9781844075911.

Walz, R. (2011), ‘Employment and structural impacts of material efficiency strategies: results from five
case studies’, Journal of Cleaner Production, 19, pp.805 – 815.

Wiedmann, T. et al. (2013), ‘The material footprint of nations’, Proceedings of the National Academy of
Sciences, 112(20), pp.6271-6276.

Wijkman, A. and K. Skanberg, (2016), The Circular Economy and Benefits for Society, The Club of Rome,
www.clubofrome.org/wp-content/uploads/2016/03/The-Circular-Economy-and-Benefits-for-
Society.pdf.

Winning, M. et al (2017), Towards a circular economy: insights based on the development of the global
ENGAGE-materials model and evidence for the iron and steel industry, International Economics and
Economic Policy, pp. 1-25.

Wood, R. et al. (2014), Global Sustainability Accounting—Developing EXIOBASE for Multi-Regional

Footprint Analysis, Sustainability, 7(1), pp.138-163.

52
 ENV/WKP(2018)4

World Steel Association (2012), Sustainable steel: At the core of a green economy,
www.worldsteel.org/en/dam/jcr:5b246502-df29-4d8b-92bb-afb2dc27ed4f/Sustainable-steel-at-the-
core-of-a-green-economy.pdf.

53