Content uploaded by Daniel Guzzo
Author content
All content in this area was uploaded by Daniel Guzzo on Apr 03, 2023
Content may be subject to copyright.
Journal of Cleaner Production 405 (2023) 137003
Available online 28 March 2023
0959-6526/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Review
Unveiling the dynamic complexity of rebound effects in sustainability
transitions: Towards a system’s perspective
Daniel Guzzo
a
,
*
, Bob Walrave
b
, Daniela C.A. Pigosso
a
a
Technical University of Denmark (DTU), Department of Civil and Mechanical Engineering, Lyngby, Denmark
b
Eindhoven University of Technology (TU/e), School of Industrial Engineering, Eindhoven, the Netherlands
ARTICLE INFO
Handling Editor: Mattias Lindahl
Keywords:
Rebound effects
Systems thinking
System dynamics
Sustainability transitions
Complex systems
ABSTRACT
Rebound effects (RE) are systemic responses to sustainability-oriented actions that have relentlessly offset the
anticipated effects, hindering sustainability transitions. Limitations to account for feedback, delays, and non-
linearities hinder a deep understanding of RE, leading to divergent magnitude estimates and management rec-
ommendations. Therefore, a better understanding of the dynamic complexity surrounding RE occurrence is
needed. Dynamic complexity manifests from the feedback relationships between system elements and how they
change over time. This work aims to enhance the understanding of RE’s causal and dynamic traits, following
system dynamics (SD) as the investigation frame. Based on a literature review, 24 RE-specic dynamic com-
plexities were identied and further categorised following the Iceberg model, which deepens into the causes of
RE occurrence, providing additional leverage to prevent or mitigate them. The RE-specic dynamic complexities
are then explored in case studies investigating RE through SD, which sustains three propositions for moving
forward in RE investigations. This work sets the foundation for enabling less deterministic examinations of RE,
capable of reaching recommendations that consider the true nature of the phenomenon.
1. Introduction
Driven by the ultimate goal of reaching societal development within
the planetary boundaries (Steffen et al., 2015), sustainability transitions
are being carried out on a wide range of fronts: from the renewable
energy transition to the implementation of a Circular Economy (CE)
(Circle Economy, 2022; Markard et al., 2020; UNEP, 2021). From the
energy transition side, although improved energy efciency has been on
the agenda of policy-makers and companies, there are no signs that
growth in global energy consumption is slowing down or decoupling
from economic growth (Brockway et al., 2021). Additionally, the world
is still only 7.2% circular, meaning that only 7.2% of the total yearly
material inputs in the economy rely on secondary materials cycled back
as input (Circle Economy, 2023). Therefore, despite global efforts to-
wards sustainability, the total Greenhouse Gases (GHG) emissions are
still rising (UNEP, 2021).
The disarray between sustainability-oriented actions and the ex-
pected reduction in resource use (such as energy and material) is
partially explained by the so-called Rebound Effects (RE). Following
Hertwich (2005), we dene RE as systemic responses to measures
designed to enhance sustainability outcomes that partially or entirely
offset the measure’s intended effects. Three main types of mechanisms
determine RE: direct, indirect, and macro-economic (Font-Vivanco
et al., 2016; Greening et al., 2000). Empirical studies estimate that direct
and indirect RE undermine ca. 20–40% of the intended benets (Gil-
lingham et al., 2016), while economy-wide RE undermine at least 50%
of the intended benets (Brockway et al., 2021). Nevertheless, much of
the ndings around RE occurrence and estimations are still limited,
contradictory and controversial (Madlener and Turner, 2016). Re-
searchers tend to follow simplistic views based on the relation between
the output and input of a system (Giampietro and Mayumi, 2018), as
opposed to seeking the underlying causes of RE (e.g., structural resis-
tance to change, behavioural responses) (Font-Vivanco et al., 2018;
Polizzi di Sorrentino et al., 2016; Thiesen et al., 2008; Weidema, 2008).
In addition, the existing tools are usually biased toward addressing a
specic perspective (Madlener and Turner, 2016) and fail to account for
the feedback, non-linearities and delays among system elements leading
to RE (Brockway et al., 2021; Colmenares et al., 2020). Thus, there is a
need for a systemic perspective towards understanding the complexity
surrounding RE (Madlener and Turner, 2016).
Abbreviations: CE, Circular Economy; CLD, Causal Loop Diagram; RE, Rebound Effect; SFD, Stock and Flow Diagram; SD, System Dynamics.
* Corresponding author.
E-mail address: dgdco@dtu.dk (D. Guzzo).
Contents lists available at ScienceDirect
Journal of Cleaner Production
journal homepage: www.elsevier.com/locate/jclepro
https://doi.org/10.1016/j.jclepro.2023.137003
Received 25 November 2022; Received in revised form 11 February 2023; Accepted 27 March 2023
Journal of Cleaner Production 405 (2023) 137003
2
Increased dynamic complexity is a recurrent argument for the
persistence of problems over time, as the incapacity to account for the
feedback, non-linearities and delays in systems leads to systemic re-
sponses that offset the measures’ pursued effects (Sterman, 2000, 2001).
By failing to understand the role of dynamic complexities giving rise to
RE, one might nd different RE magnitudes depending on the moment of
measurement or even fail to encounter the effects of specic mecha-
nisms altogether. Furthermore, stand-alone percentages of RE are not
enough to draw meaningful understanding and make recommendations
(Colmenares et al., 2020). Thus, a more thorough understanding of the
reasons for RE persistence encompasses dealing with causality and the
sources of complexity.
Therefore, this work aims to enhance the understanding of RE’s
causal and dynamic traits. System Dynamics (SD) is set as the investi-
gation frame because it holds philosophical and methodological lenses
that allow for a better understanding of the dynamic complexity of
systems (Sterman, 2001). Furthermore, SD enables learning about the
causal structure of systems and potential behaviour over time through
conceptual modelling and differential equation simulation models
(Sterman, 2000). The research applies a systematic literature review
combining deductive and inductive content analyses and is built upon
three research questions (RQ) to achieve its goal.
•RQ1 – What dynamic complexities manifest in systems leading to
RE?
•RQ2 – How have SD investigations tackled the complexity of the RE
phenomenon so far?
•RQ3 – How can SD be further used to address the dynamic com-
plexities in RE investigations?
This work is positioned alongside a few recent review studies that
contribute to RE understanding and mitigation in different ways. For
instance, some studies have consolidated ndings from empirical esti-
mates of RE to assert modelling choices from an energy economics
perspective (Brockway et al., 2021; Colmenares et al., 2020). Others
have provided more comprehensive typologies of RE to help their
identication(Lange et al., 2021; Metic et al., 2022). Moreover, a few
reviews aimed at integrating areas such as CE (Castro et al., 2022; Metic
et al., 2022) and industrial ecology (Reimers et al., 2021), expanding an
energy-oriented perspective on the RE phenomenon. Finally, some re-
views investigated behavioural and social aspects in decision-making
that can lead to RE (Exadaktylos & van den Bergh, 2021; Reimers
et al., 2021). These review studies show a substantial intensication of
the importance of understanding and resolving RE. Although the re-
views investigated and contributed to the phenomenon from multiple
angles, there are still critical open angles — specically related to the
strong indications of the limitations of the dominating approaches in
accounting for the dynamic complexity that arises from feedback, de-
lays, and non-linearities in RE investigations.
This work sets the foundation for more systemic approaches to
investigating RE in several ways. First, this work systematically iden-
ties the dynamic complexities (i.e., factors that increase in-
terrelationships or alter the temporal interaction of system elements
(Gr¨
osser, 2017; Senge, 1990)) surrounding the RE phenomenon by
eliciting and organising 24 specic dynamic complexities. Second, it
deepens the understanding of the phenomena from events and patterns
of behaviour to the structures and mental models, by organising the
RE-specic dynamic complexities according to the Iceberg model
(Davelaar, 2021; Kim, 2000). Third, this work explores six documented
cases to enhance the awareness of the specic sources of complexity and
how they have been addressed, which will enable designing in-
vestigations that consider those factors and reach more meaningful in-
sights into tackling RE. Finally, three avenues for moving forward in RE
investigations through SD are proposed. Overall, this work makes
tangible how a system perspective can address the complexity sur-
rounding RE.
2. Research methodology
In response to the rst research question (RQ1), we draw on a Sys-
tematic Literature Review (SLR) (Thom´
e et al., 2016) using the
following search string: TITLE (“rebound*” OR “Jevons paradox” OR
“unintended consequence*” OR ”unanticipated consequence*” OR
“unexpected consequence*” OR “policy resistance” OR “second-order
effect*” OR “boomerang effect*” OR “ripple effect* OR backre) AND
TITLE, ABSTRACT AND KEYWORDS (sustain* OR circular OR energy
OR resource* OR environment* OR emission* OR ecolog*). This string
ensured a broad consideration of the RE phenomenon by including
keywords often used interchangeably as rebound, Jevons’ paradox and
backre. At the same time, it helped narrow down to
sustainability-related phenomena, as unintended and unanticipated
consequences are also employed in other research areas such as medi-
cine and political sciences. To keep the research manageable, we opt to
include only review studies in the systematic literature identication.
In total, 66 review studies were identied through the search in
Scopus and Web of Science (carried out in July 2022). Only review
studies explicitly addressing RE as the primary unit of analysis (i.e.,
within sustainability and related areas) were considered, resulting in the
selection of 17 studies for further analysis. Co-citation analysis (Boyack
and Klavans, 2010; Eck and Waltman, 2014) and snowballing (Wohlin,
2014) enabled the identication of 17 additional inuential studies in
the RE discourse, resulting in the analysis of 34 studies in total.
The research followed a deductive-inductive content analysis pro-
cedure to analyse the dynamic complexity in RE (RQ1) (Elo and Kyng¨
as,
2008; Hsieh and Shannon, 2005). More specically, the following steps
were deployed.
•Step 1 - Deductive analysis: identication of dynamic complexity
factors described in the RE literature (based on Gr¨
osser, 2017; Senge,
1990) and further categorisation according to the characteristics of
complex systems as proposed by Sterman (2000) – see Table 1.
•Step 2 - Inductive analysis: categorising and clustering similar
sources of complexity, resulting in a consolidated understanding of
the dynamic complexities of RE.
•Step 3 - Deductive content analysis: categorising sources of
complexity according to the iceberg model (Davelaar, 2021; Kim,
2000), which, by analogy, connects the different levels of thinking
from the observable events on the surface to the patterns of
Table 1
The dynamic complexities of systems.
Dynamic complexities of
systems
Denition following Sterman (2000)
Policy resistant Many obvious solutions fail or worsen the situation.
Constantly changing All is changing. System change occurs at many scales, and
these different scales sometimes interact.
Tightly coupled Everything is connected. The actors in the system
actively interact with one another and the natural world.
Governed by feedback One’s decisions alter the situation, triggering change and
action, giving rise to a new situation which then
inuences one’s subsequent decisions.
Non-linear The effect is rarely proportional to the cause. Non-
linearities often arise due to internal delays and multiple
inuencing factors.
Counter-intuitive Causes and effects are distant in time and space,
hindering learning. As a result, we commonly focus on
the events rather than the underlying causes.
Adaptive The capabilities and decision rules of agents change over
time. Agents evolve and learn over time.
Self-organising The dynamics of a system arise from its internal
structure. Small perturbations are amplied, generating
patterns of behaviour.
History-dependent Previous decisions dene the set of decisions available
now. Doing and undoing have fundamentally different
dynamics.
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
3
behaviour, underlying structures and mental models which are often
out of sight of the human mind (see Fig. 1).
In response to the second research question (RQ2), six documented
case studies that employed SD to tackle the dynamic complexities of RE
were analysed. The following criteria determined the inclusion of a
study: (1) an explicit sustainability purpose; (2) it must account for a RE.
Studies that did not depart from an action holding sustainability purpose
or that dealt with only other “side effects” were disregarded. First, the
studies were described as to the framing of the investigation and the
explanation, assessment, and recommendations about RE. Then, a
deductive content analysis (Elo and Kyng¨
as, 2008; Hsieh and Shannon,
2005) using the RE-specic dynamic complexities (RQ2) enabled the
identication of the dynamic complexities that have been tackled in the
studies.
Finally, the third research question (RQ3) was addressed by a critical
reection on the gap between the identied dynamic complexities in
systems leading to RE (RQ1) and the insofar use of SD to address dy-
namic complexities in RE studies (RQ2).
Fig. 2 depicts the steps taken to respond to the three RQ posed for the
study. The conceptual framework for the dynamic complexities in sys-
tems leading to RE (RQ1), the analysis of the use of SD to investigate RE
(RQ2), and the research agenda for SD-based RE investigation (RQ3) are
respectively described in sections 3, 4, and 5.
3. A conceptual framework for the RE-specic dynamic
complexities
This section contains the conceptual framework for the dynamic
complexities in systems leading to RE is two formats. First, section 3.1
contains the RE-specic dynamic complexities identied, connected to
the evidence found in the literature and following the characteristics of
complex systems (Sterman, 2000). Then, section 3.2 contains the cate-
gorisation of the RE-specic dynamic complexities following the Iceberg
model (Davelaar, 2021; Kim, 2000), making the four different levels of
thinking explicit.
3.1. RE-specic dynamic complexities in the literature
Table 2 summarises the 24 specic sources of dynamic complexities
leading to RE following the nine characteristics of complex systems.
Each specic source of dynamic complexity is described in the following
Fig. 1. The Iceberg model making explicit four different levels of understand-
ing – adapted from (add a reference here).
Fig. 2. Research steps to respond to the two research questions.
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
4
nine sub-sections.
3.1.1. Policy-resistant systems
Policy-resistant systems are those in which seemingly obvious solu-
tions might fail or worsen the situation (Sterman, 2000). In the rst
place, economic and behavioural systemic responses to well-intended
sustainability-oriented actions lead to RE (PR1) (Hertwich, 2005;
Lange et al., 2021; van den Bergh, 2011). The collections of RE mech-
anisms in literature comprehend several economic, behavioural,
time-related, and other responses to well-intended actions (Colmenares
et al., 2020; Lange et al., 2021; Metic et al., 2022). For example, while a
car-sharing solution might decrease the stock of cars required to support
the needs of a given user group, RE might occur when users (a) spend
additional income on other services (economic response); or (b) take less
care of the shared cars (behavioural response). Such responses are why
sustainability-oriented action (in this case, a car-sharing system) may
not reach its full potential.
As a relative measure, RE hold different magnitudes and di-
rections (PR2). For example, RE might offset or reinforce the expected
impact of the action (Binswanger, 2001). Saunders (2008) dened
ve-point graduation: (1) super-conservation leads to more positive
impacts than expected; (2) zero RE leads to no offset; (3) partial RE
offsets a portion of the intended impact; (4) full RE offsets the whole
intended impact; and (5) backre leads to a worse situation than before
the action. Several studies investigate the magnitude of RE (Brockway
et al., 2021; Colmenares et al., 2020; Lange et al., 2021) and show
substantial RE magnitudes, even presenting cases that backred. The
magnitudes are, thus, the outcomes of the systemic responses to the
expected impacts of the action.
Decision-makers holding a narrow view of the system fail to
identify RE ex-ante (PR3) as they often make estimations based on
systematic bias and wrong assumptions (Friedrichsmeier and Matthies,
2015). Decision-makers often assume that actions aiming for better
resource use will translate straightforwardly into lower energy use and
GHG emissions, implicitly assuming RE equals zero and treating essen-
tial system elements as exogenous (Madlener and Alcott, 2009; Sorrell
et al., 2020). A common assumption is that focusing on the impacts of a
single unit (e.g., one car, one phone) will lead to more efcient systems,
causing decision-makers to ignore critical feedback, leading to RE
(Font-Vivanco et al., 2016; Laurenti et al., 2016). For instance, more
efcient heating systems might lead users to leave their windows open
to enable air circulation, leading to additional energy usage (Sonnberger
and Gross, 2018). In this case, focusing only on the heating system’s
efciency can hinder designers from seeing the other potential system
responses.
The last policy-resistant characteristic of systems leading to RE ad-
dresses the mitigation of RE. Preventing or mitigating RE requires a
deep understanding of their causes (PR4). Decision-makers should be
educated to understand RE mechanisms to help identify and potentially
avoid or mitigate them (Lange et al., 2021; Madlener and Turner, 2016).
Although several recommendations are available to address RE in
literature (Binswanger, 2001; Castro et al., 2022; Colmenares et al.,
2020; Exadaktylos & van den Bergh, 2021; Greening et al., 2000), it is
critical to acknowledge that the actions to resolve RE might also cause
new ones to occur (Colmenares et al., 2020). Thus, preparing
decision-makers to see the potential unintended consequences of actions
is very important to prevent or mitigate RE.
3.1.2. Constantly changing systems
Constantly changing systems are those where change occurs at many
scales, which sometimes interact (Sterman, 2000). To begin with, RE are
outcomes of constantly changing systems as RE occur in systems at the
micro, meso, and macro levels (CC1). Sustainability transitions might
focus on systems holding different aggregation levels (Castro et al.,
2022), such as improving the sustainability potential of a device, an
industrial process, the business model of a company, or a socio-technical
transformation. Thus, there are different scopes to frame the investiga-
tion of a system prone to RE (Greening et al., 2000; Santarius, 2016;
Sorrell, 2009; van den Bergh, 2011): micro (e.g., households and rms),
meso (e.g., cities, industrial parks and sectors), macro (e.g., nations,
cross-national regions and the whole world). The scope set for a given
analysis inevitably constrain investigating effects beyond those limits
(Santarius, 2015). From one side, micro-level investigations are critical
as they can help understand the behaviour of individuals and rms
leading to consumption and production patterns (Santarius, 2016;
Trincado et al., 2021). On the other side, assessing RE at higher levels
may integrate the global markets for resources and account for an
extended set of interactions and feedback (Madlener and Alcott, 2009;
Santarius, 2016; van den Bergh, 2011). Thus, framing different system
levels might enable recognising different dynamics and RE.
In addition, systems leading to RE are nested and interrelated
(CC2). Giampietro and Mayumi (2018) use the concept of holons (i.e., a
whole made of smaller parts and a part of some greater whole simul-
taneously) to argue that systems on different levels are nested and
interrelated and where changes in one level might also result in changes
in other levels (Giampietro and Mayumi, 2018). Thus, when aggregating
RE at different levels (Gillingham et al., 2013; Santarius, 2016), one
must avoid the “fallacy of double counting” (Lange et al., 2021). The
different mechanisms for RE interact, leading to combined effects
(Brockway et al., 2021; Gillingham et al., 2013, 2016; Lange et al.,
2021). For instance, the composition effect (i.e., changes in production
factors across the value chain due to efciency improvements) amplies
the substitution effect (i.e., changes in production factors for each pro-
ducer) (Santarius, 2016). Also, RE mechanisms in higher aggregation
levels might share similar underlying mechanisms with those in lower
levels with an adding, subtracting, or multiplying contribution (Lange
et al., 2021). Moreover, there might be feedback loops between higher
and lower levels (Lange et al., 2021). Therefore, the combined effect in
higher levels might be of greater or smaller magnitude than summing its
parts (Brockway et al., 2021; Gillingham et al., 2013; Lange et al., 2021).
3.1.3. Tightly coupled systems
Tightly coupled systems manifest dynamic complexity from active
interaction between actors and the natural world (Sterman, 2000). First,
RE are an outcome of tightly coupled systems as the interaction of
agents holding multiple interests in the system leads to their
occurrence (TC1). Policymakers (i.e., agents designing public policies),
business decision-makers (i.e., agents translating the public policies into
business policies or departing from innovation efforts), and consumers
(i.e., agents making consumption decisions) perform the actions leading
to production, consumption and RE (Colmenares et al., 2020; Greening
et al., 2000). Thus, business and public policies must be capable of
targeting the multiple objectives of these actors and dealing with the
delicate trade-offs emerging from a multi-stakeholder system where the
occurrence of RE is one aspect to consider (Castro et al., 2022; Colme-
nares et al., 2020; Friedrichsmeier and Matthies, 2015; Madlener and
Turner, 2016). In addition, some actors may welcome RE (Santarius,
2015). For instance, policymakers may selectively consider the reaction
of measures that might lead to RE (e.g., job creation, rural community
development, and reduction of energy poverty) to show their potential
additional economic benets, using them as arguments (Hertwich,
2005; Trincado et al., 2021; Weidema, 2008). The indication of selective
consideration might mean that decision-makers seek more thoroughly
the benecial higher-order effects of action than the detrimental ones.
Second, interacting with the natural world unfolds into additional
stakes to address. Thus, RE understanding requires addressing the
multidimensionality of sustainability outcomes (TC2). Several au-
thors argue that RE is not only an energy phenomenon but a sustain-
ability one, encompassing other environmental (e.g., GHG emissions, air
pollution, material resources use, water and energy use) and socioeco-
nomic effects (e.g., quality jobs generation, health risks, accessibility of
services) (Azevedo, 2014; Castro et al., 2022; Colmenares et al., 2020;
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
5
Table 2
The dynamic complexities in systems leading to RE.
Dynamic
complexities of
systems
The dynamic complexities in
systems leading to RE
References
Policy resistant
(PR)
PR1: Systemic responses to well-
intended sustainability-oriented
actions lead to RE
(Colmenares et al., 2020;
Hertwich, 2005; Lange
et al., 2021; Metic et al.,
2022; van den Bergh, 2011)
PR2: RE hold different
magnitudes and directions
(Binswanger, 2001;
Brockway et al., 2021;
Colmenares et al., 2020;
Lange et al., 2021;
Saunders, 2008)
PR3: Decision-makers holding a
narrow view of the system fail to
identify RE ex-ante
(Font-Vivanco et al., 2016;
Friedrichsmeier and
Matthies, 2015; Laurenti
et al., 2016; Madlener and
Alcott, 2009; Sonnberger
and Gross, 2018; Sorrell
et al., 2020)
PR4: Preventing or mitigating
RE requires a deep
understanding of their causes
(Binswanger, 2001; Castro
et al., 2022; Colmenares
et al., 2020; Exadaktylos &
van den Bergh, 2021;
Greening et al., 2000; Lange
et al., 2021; Madlener and
Turner, 2016)
Constantly
changing (CC)
CC1: RE occur in systems at the
micro, meso, and macro levels
(Castro et al., 2022;
Greening et al., 2000;
Madlener and Alcott, 2009;
Santarius, 2015, 2016;
Sorrell, 2009; Trincado
et al., 2021; van den Bergh,
2011)
CC2: Systems leading to the
occurrence of RE are nested and
interrelated
(Brockway et al., 2021;
Giampietro and Mayumi,
2018; Gillingham et al.,
2013, 2016; Lange et al.,
2021; Santarius, 2016)
Tightly coupled
(TC)
TC1: The interaction of agents
holding multiple interests in the
system leads to RE
(Castro et al., 2022;
Colmenares et al., 2020;
Friedrichsmeier and
Matthies, 2015; Greening
et al., 2000; Hertwich,
2005; Madlener and Turner,
2016; Santarius, 2015;
Trincado et al., 2021;
Weidema, 2008)
TC2: RE understanding requires
addressing the
multidimensionality of
sustainability outcomes
(Azevedo, 2014; Castro
et al., 2022; Colmenares
et al., 2020; Font-Vivanco
et al., 2016, 2018;
Font-Vivanco and van der
Voet, 2014; Friedrichsmeier
and Matthies, 2015;
Gillingham et al., 2016;
Hertwich, 2005)
Governed by
feedback (GF)
GF1: RE emerge due to
consumer and producer-side
reactions
(Castro et al., 2022;
Font-Vivanco et al., 2016;
Madlener and Alcott, 2009;
Madlener and Turner, 2016;
Santarius, 2015, 2016;
Sorrell et al., 2020; Turner,
2013; van den Bergh, 2011;
Weidema, 2008)
GF2: RE might occur due to
seemingly unrelated behaviour
(Azevedo, 2014;
Font-Vivanco and van der
Voet, 2014; Greening et al.,
2000; Sonnberger and
Gross, 2018)
Non-linear (NL) NL1: Multiple causal
relationships might moderate
and mediate RE
(Azevedo, 2014;
Font-Vivanco et al., 2016;
Friedrichsmeier and
Matthies, 2015; Greening
Table 2 (continued )
Dynamic
complexities of
systems
The dynamic complexities in
systems leading to RE
References
et al., 2000; Ruzzenenti and
Basosi, 2008; Santarius,
2015; Sonnberger and
Gross, 2018; Sorrell, 2009;
Sorrell et al., 2020; van den
Bergh, 2011; Weidema,
2008)
NL2: Causal connections might
hold non-linear relationships
between factors
(Azevedo, 2014;
Binswanger, 2001; Castro
et al., 2022; Greening et al.,
2000; Hertwich, 2005;
Metic et al., 2022; Sorrell,
2009; Sorrell et al., 2020;
Trincado et al., 2021; van
den Bergh, 2011; Zink and
Geyer, 2017)
NL3: RE present high
heterogeneity in occurrence and
magnitude
(Azevedo, 2014; Brockway
et al., 2021; Castro et al.,
2022; Colmenares et al.,
2020; Giampietro and
Mayumi, 2018; Gillingham
et al., 2016; Hertwich,
2005; Madlener and Turner,
2016; Reimers et al., 2021;
Sorrell et al., 2020; Trincado
et al., 2021; van den Bergh,
2011)
Counter-
intuitive (CI)
CI1: There are substantial delays
between implementing the
sustainability-oriented action
and RE emergence
(Castro et al., 2022;
Font-Vivanco and van der
Voet, 2014; Gillingham
et al., 2016; Madlener and
Turner, 2016; Metic et al.,
2022; Santarius, 2016)
CI2: Systems present different
short-run and long-run
responses to changes
(Azevedo, 2014;
Colmenares et al., 2020;
Font-Vivanco and van der
Voet, 2014; Greening et al.,
2000; Lange et al., 2021;
Madlener and Turner, 2016;
Santarius, 2016; Turner,
2013)
CI3: There are delays between
the different types of RE
(Azevedo, 2014; Brockway
et al., 2021; Castro et al.,
2022; Colmenares et al.,
2020; Santarius, 2016;
Sorrell, 2009; Trincado
et al., 2021; Turner, 2013)
Adaptive (Ad) Ad1: The purpose of production
and consumption systems are
subjective and evolutionary
(Font-Vivanco et al., 2016;
Font-Vivanco and van der
Voet, 2014; Giampietro and
Mayumi, 2018; Madlener
and Alcott, 2009; Sorrell
et al., 2020; van den Bergh,
2011)
Ad2: Individuals are subjected
to bounded rationality
(Azevedo, 2014; Castro
et al., 2022; Exadaktylos &
van den Bergh, 2021;
Friedrichsmeier and
Matthies, 2015; Madlener
and Alcott, 2009; Reimers
et al., 2021; Santarius,
2016; Sonnberger and
Gross, 2018; Sorrell et al.,
2020; van den Bergh, 2011)
Ad3: Social systems inuence
individual behaviour
(Azevedo, 2014; Castro
et al., 2022; Exadaktylos &
van den Bergh, 2021;
Font-Vivanco et al., 2016;
Friedrichsmeier and
Matthies, 2015; Madlener
and Alcott, 2009; Matraeva
et al., 2022; Reimers et al.,
(continued on next page)
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
6
Font-Vivanco and van der Voet, 2014; Friedrichsmeier and Matthies,
2015; Hertwich, 2005). Additionally, there are trade-offs between the
dimensions and the potential of identifying co-benets, including mea-
surements not initially targeted by the action (Font-Vivanco et al., 2016,
2018; Friedrichsmeier and Matthies, 2015; Hertwich, 2005). For
example, a critical trade-off in RE studies emerges between the envi-
ronmental effects of resource consumption and the implications for so-
cial welfare. Some researchers suggest combining welfare maximisation
and environmental minimisation, as only minimising impacts can hold
detrimental social effects (Gillingham et al., 2016; Madlener and Turner,
2016). For instance, from an energy perspective, Gillingham et al.
(2016) propose to maximise welfare benets per energy use, leading to a
relative measure to address RE.
3.1.4. Systems governed by feedback
Systems leading to RE are governed by feedback, where one’s de-
cisions give rise to a new situation inuencing follow-up decisions
(Sterman, 2000). First, consumer and producer-side reactions to
sustainability-oriented actions lead to RE (GF1) because of changes
in resource use (Castro et al., 2022; Madlener and Turner, 2016; San-
tarius, 2016; Turner, 2013). For instance, from the consumer side, the
substitution effect occurs when a new price relation between two ser-
vices leads to additional consumption of the cheaper service leading to
additional impact. From the producer side, substitution occurs when
more efcient resource use for production leads to a new price relation
and more output via the cheaper process. Second-order effects also
occur, where the reactions of changes in production might lead to
changes in consumption, too – and vice-versa (Weidema, 2008). Thus, it
is critical to map and distinguish relevant consumer and producer-side
actions and reactions (Turner, 2013). In addition, for a complete pic-
ture of the occurring RE, it is essential to map the impacts in the complex
networks of interconnected rms, production chains, and international
transportation (van den Bergh, 2011). Finally, research should integrate
embodied energy of goods and services into demand-led RE examina-
tions (Font-Vivanco et al., 2016; Madlener and Alcott, 2009; Madlener
and Turner, 2016; Sorrell et al., 2020) when the cause-and-effect chain
to the sustainability action is traceable (Santarius, 2015).
Second, seemingly unrelated behaviour might lead to RE (GF2),
such as consuming other goods or services. For example, a technical
change in a system inuences the demand for that good – i.e., direct
causality, but it may also affect the demand for other goods – i.e., in-
direct causality (Font-Vivanco and van der Voet, 2014). From a con-
sumption perspective, it occurs because consumers constantly compare
alternative options, considering the availability of money, time, pref-
erences, and other consumption factors (Greening et al., 2000). Also, the
needs are co-dependent. For instance, grocery shopping is related to
mobility and cooking, and changes in food consumption might lead to
changes in those related activities (Sonnberger and Gross, 2018).
Although economic frameworks investigate the indirect effects of con-
sumption and production, the relationships between different goods and
services are hard to grasp (Azevedo, 2014). Thus, it is critical to accu-
rately account for alternatives by, for instance, setting a functional unit
that encompasses different products and services (Font-Vivanco and van
der Voet, 2014) while mapping substitute and complementary con-
sumption and production behaviour.
3.1.5. Non-linear systems
Non-linear systems present disproportionate effects to causes arising
from internal delays and the inuence of multiple concomitant factors
(Sterman, 2000). To begin with, RE are the outcomes of non-linear
systems as multiple causal relationships might moderate and
mediate RE (NL1). First, efciency changes often unfold into changes in
other product attributes, such as safety, comfort, and quality, which also
inuence the behaviour of consumers and can potentially lead to RE
(Azevedo, 2014). Additionally, multiple parameters shape an in-
dividual’s purchasing decision, such as time, physical space, prefer-
ences, skills, and costs (Sonnberger and Gross, 2018; van den Bergh,
2011; Weidema, 2008). Thus, measures might relieve several con-
sumption constraints, leading to interdependent RE occurring concom-
itantly (Ruzzenenti and Basosi, 2008; Sorrell et al., 2020; van den Bergh,
2011). Thus, factors might moderate and mediate RE, as the different
consumption factors might not have autonomous effects (Font-Vivanco
et al., 2016). For instance, satiation or time constraints can moderate
additional consumption from released income (Greening et al., 2000).
That is, consumption power released by a shared mode of transportation
may not be consumed entirely due to lack of time. However, it is unclear
in RE research what part of the subsequent changes in consumption
should be attributed to an efciency improvement when it is impossible
to track the causal linkages (Font-Vivanco et al., 2016; Sorrell, 2009).
Also, analyses should disregard other changes co-occurring with the
sustainability-oriented action, only accounting for the portion the
mechanisms are responsible for (i.e., cause-effect relativity) (Frie-
drichsmeier and Matthies, 2015; Santarius, 2015).
Additionally, causal connections might hold non-linear re-
lationships between factors (NL2). For instance, the price elasticity of
demand will highly inuence the RE magnitude (Trincado et al., 2021),
which depends on the type of good (Binswanger, 2001). While the de-
mand increases as income increases for normal goods, inferior goods
behave oppositely. RE will emerge considering the type of goods and the
disparity in sustainability impacts of consuming them. Also, the elas-
ticity of demand is not constant (Azevedo, 2014) and tend to increase as
prices increase (Binswanger, 2001; Hertwich, 2005) and according to
changes in income, preferences and lifestyles (Greening et al., 2000;
Sorrell et al., 2020; Trincado et al., 2021). Also, CE initiatives show
many cases where new modes of consumption do not entirely replace
primary production (Castro et al., 2022). For instance, second-use and
remanufactured products may not completely replace the need for
rst-use products (Metic et al., 2022; Zink and Geyer, 2017), which
could lead to RE. Finally, uncovering the relationship between the gains
in economic productivity per resource use and the growth in output not
explained by increased inputs might help identify RE (Sorrell, 2009; van
den Bergh, 2011).
Table 2 (continued )
Dynamic
complexities of
systems
The dynamic complexities in
systems leading to RE
References
2021; Santarius, 2016;
Sonnberger and Gross,
2018; Sorrell et al., 2020)
Self-organising
(SO)
SO1: Essential reinforcing
mechanisms stimulate
production and consumption
systems
(Castro et al., 2022;
Font-Vivanco et al., 2016;
Giampietro and Mayumi,
2018; Lange et al., 2021;
Laurenti et al., 2016;
Santarius, 2016; Sonnberger
and Gross, 2018; Sorrell,
2009; Trincado et al., 2021;
van den Bergh, 2011)
SO2: Essential balancing
mechanisms regulate
production and consumption
systems
(Lange et al., 2021;
Santarius, 2016; Sorrell
et al., 2020)
SO3: Small changes in
production and consumption
factors can lead to huge
amplications
(Gillingham et al., 2016;
Ruzzenenti and Basosi,
2008)
History-
dependent
(HD)
HD1: Co-dependence between
sustainability actions and
transitions inuence RE
(Castro et al., 2022; Figge
and Thorpe, 2019;
Hertwich, 2005)
HD2: The inertia of systems
might inuence the timing and
magnitude of RE
(Binswanger, 2001;
Exadaktylos & van den
Bergh, 2021; Hertwich,
2005; Sonnberger and
Gross, 2018)
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
7
The multiple causal relationships with potential non-linear effects
might cause RE present high heterogeneity in occurrence and
magnitude (NL3). A few factors determining RE heterogeneity are the
level of economic development of the country, the income group of a
household, the local culture, the regional location, and the sector of the
industry inuenced by the sustainability action (Brockway et al., 2021;
Castro et al., 2022; Madlener and Turner, 2016; Sorrell et al., 2020).
There is a deep discussion about the magnitude of RE and the level of
development of an economy. More specically, stronger RE are expected
in developing economies because (i.) their resource supply is con-
strained (Azevedo, 2014; van den Bergh, 2011), (ii.) their demand is
more constrained by costs (Azevedo, 2014; Hertwich, 2005), (iii.) the
consumption is further away from saturation as human needs are not
resolved (Azevedo, 2014; Hertwich, 2005; van den Bergh, 2011), and
(iv.) relatively more consumption goes to emission-intensive necessities,
such as housing and food (Reimers et al., 2021). Some argue for a cor-
relation between welfare increase and emission reduction (Colmenares
et al., 2020) by, for instance, using the Kuznets curve as a rule of thumb,
which hypothesises an inverted U-shaped relationship between the
resource use and per capita income (Trincado et al., 2021). Nevertheless,
care is needed in generalising the RE magnitude to the development of
the economy as there are signicant discrepancies in wealth in devel-
oping economies, and the wealth of those inuenced by the measure
should be considered instead (Gillingham et al., 2016). In addition,
Giampietro and Mayumi (2018) argue that it might be more practical to
control consumption (and thus RE) when a society has reached a
particular welfare threshold.
3.1.6. Counter-intuitive systems
Counter-intuitive systems might lead people to fail to grasp the
reasons for the events they see because causes and effects are distant in
time (Sterman, 2000). First, RE are an outcome of counter-intuitive
systems, as there are substantial delays between implementing the
sustainability-oriented action and RE emergence (CI1) (Castro et al.,
2022; Santarius, 2016). Therefore, it can be very challenging for
decision-makers to foresee the consequences of actions before imple-
mentation. Such a challenge might explain the disbalance towards
ex-post RE investigations, as ex-ante approaches still lack (Metic et al.,
2022). Proactive decision-making requires ex-ante scenario-based ana-
lyses of potential RE occurrence and magnitudes (Font-Vivanco and van
der Voet, 2014; Gillingham et al., 2016; Madlener and Turner, 2016;
Metic et al., 2022). Thus, it is essential to address the design-outcome
delay consistently.
Second, systems present different short-run and long-run re-
sponses to changes (CI2), which inuence RE. Different short- and
long-run responses to sustainability-oriented action have been widely
pointed out (Colmenares et al., 2020; Greening et al., 2000; Santarius,
2016). An important reason is that consumers, producers, supply chains,
and social institutions change on different time scales (Font-Vivanco and
van der Voet, 2014; Greening et al., 2000; Lange et al., 2021). For
instance, prices and incomes adjust through the economy, and compli-
mentary consumption and production decisions change accordingly
(Colmenares et al., 2020; Font-Vivanco and van der Voet, 2014; Madl-
ener and Turner, 2016; Turner, 2013). Also, a company might cut costs
in the short term and follow an output-maximising behaviour in the long
term (Greening et al., 2000). In addition, there might be signicant
delays between changes in resource use, service demand and the
incorporation of capital costs and market saturation (Azevedo, 2014). In
synthesis, Lange et al. (2021) argue that short-run changes comprise
prices, quantities, and real income, while long-run comprise changes in
economic conditions such as preferences, technologies, and capital
stock. Short-run and long-run changes interact with each other (Lange
et al., 2021), and both can lead to RE.
Finally, there are delays between the different types of RE (CI3).
Delays between direct RE (due to rapid system responses), indirect RE
(due to slow system responses), and the long-term equilibrium of the
economy and resource use have been identied (Castro et al., 2022;
Sorrell, 2009; Trincado et al., 2021; Turner, 2013). For instance,
long-term RE may be lower than short-term RE if the return on capital
investments falls over time and enact ‘disinvestment effects’ (Santarius,
2016; Turner, 2013). RE, thus, present dynamic behaviour and might
diminish or augment over time until reaching stability (Azevedo, 2014;
Brockway et al., 2021; Colmenares et al., 2020). Therefore, studies
should consider the complete unfolding of markets, technologies and
behavioural adjustments in RE investigations (Sorrell, 2009).
3.1.7. Adaptive systems
Adaptive systems are those in which agents’ capabilities and decision
rules change over time due to evolution and learning (Sterman, 2000).
To begin with, RE are outcomes of adaptive systems as the purpose of
production and consumption systems are subjective and evolu-
tionary (Ad1). First, the utility of goods is subjective as they differ
among individuals (Font-Vivanco et al., 2016; Giampietro and Mayumi,
2018). Also, different goods might solve the same utility (Font-Vivanco
et al., 2016), which inuences supplementary and complementary
consumption. In addition, an action might lead to changes in the pur-
pose and boundaries of systems over time (Giampietro and Mayumi,
2018; van den Bergh, 2011). Some reasons are that consumers and the
market adapt to the new attributes (Font-Vivanco et al., 2016; Font--
Vivanco and van der Voet, 2014; Madlener and Alcott, 2009) – e.g., an
improvement in efciency aimed at a given function (e.g., a heating
system) can lead to a new function at different scales (e.g., now people
can install it in the living room) (Giampietro and Mayumi, 2018). Thus,
critical challenges emerge in modelling systems prone to RE as new
behaviour, additional system elements, or new functions might emerge
(Giampietro and Mayumi, 2018). Thus, a thorough appreciation of the
potential evolution of the functional unit can enable an adequate
reference for comparison (Font-Vivanco et al., 2016; Font-Vivanco and
van der Voet, 2014; Giampietro and Mayumi, 2018; Sorrell et al., 2020).
In addition, individuals make sub-optimal decisions for two main
reasons. First, individuals are subjected to bounded rationality
(Ad2) due to biases, “wrong” goals, habits, and lack of information
(Exadaktylos & van den Bergh, 2021; van den Bergh, 2011). Individuals
are biased by mental representations of the assumed costs leading to
time-inconsistent choices as they prioritise immediate costs and benets
(Azevedo, 2014; Exadaktylos & van den Bergh, 2021; Friedrichsmeier
and Matthies, 2015). Also, individuals might fail to adequately consider
resource use in their decision as they might have incomplete knowledge
about the application of devices and incomplete information about their
impacts (Azevedo, 2014; Friedrichsmeier and Matthies, 2015; Madlener
and Alcott, 2009; Sonnberger and Gross, 2018). Meanwhile, individuals
in rms are more prone to maximise prots (Santarius, 2016). However,
education and information availability inuence their decision-makers,
and they might still fail to consider all the potential effects of their de-
cisions (van den Bergh, 2011). Thus, even well-intended individuals
might behave inconsistently and prioritise action with low potential
sustainability contributions (Sorrell et al., 2020). Therefore, studies
following a utility maximisation assumption might neglect important
factors driving behaviour.
Sub-optimal decisions are also the result of social systems’ inu-
ence on individual behaviour (Ad3). People are subject to bounded
self-interest (Exadaktylos & van den Bergh, 2021) as social systems (e.g.,
technological, cultural, religious, political, and economic systems)
greatly inuence individuals (Sonnberger and Gross, 2018). Specically,
social pressures, prestige, values, norms, and the well-being of others
regulate decision-making (Azevedo, 2014; Exadaktylos & van den
Bergh, 2021; Font-Vivanco et al., 2016; Madlener and Alcott, 2009;
Santarius, 2016). Moral licensing has been consistently used to explain
unintended behaviour in resource use. It occurs when past good deeds
liberate individuals to subsequently act less environmentally
consciously (Friedrichsmeier and Matthies, 2015; Reimers et al., 2021)
or if they believe the provider is taking environmental care on their
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
8
behalf (Castro et al., 2022). Also, informal institutions such as traditions
might sustain RE even if formal institutions exist to address them
(Matraeva et al., 2022), meaning that well-designed mitigating mea-
sures might fail. Conversely, pro-environmental values divert RE by
preventing additional consumption (Exadaktylos & van den Bergh,
2021). Finally, social pressure can motivate others to adopt more sus-
tainable behaviour (Sorrell et al., 2020), acting as a self-reinforcing loop
and an essential ally to address RE.
3.1.8. Self-organising systems
Self-organising systems are those where the internal structure gen-
erates behaviour patterns, and small perturbations might be vastly
amplied (Sterman, 2000). In the rst place, RE are outcomes of
self-organising systems as essential reinforcing mechanisms stimu-
late production and consumption systems (SO1). Although there is a
debate about the causality between resource consumption and economic
growth, evidence suggests critical reinforcing feedback loops (Sorrell,
2009; Trincado et al., 2021). For example, in cases where enhancements
in the production efciency of a system lead to a decrease in price, de-
mand might rise, and additional prots might feedback into production
factors in that system (Castro et al., 2022; Lange et al., 2021; Santarius,
2016). There is, therefore, a positive feedback loop between industrial
investment, lower unit costs, lower prices for consumers, and subse-
quent demand (Trincado et al., 2021). Furthermore, higher efciency
stimulates the diffusion of innovation using that resource, which can
introduce long-range and persistent societal changes (Font-Vivanco
et al., 2016; Sorrell, 2009; van den Bergh, 2011). In addition, higher
efciency can sustain fundamental social logic as the acceleration of
everyday life through time-efciency measures (Sonnberger and Gross,
2018) and the social reliance on accelerating product obsolescence
(Laurenti et al., 2016). Overall, expanding the ability to produce more to
consume more (i.e., maximising the energy ux) is a common attractor
for socioeconomic systems (Giampietro and Mayumi, 2018).
Concomitantly, essential balancing mechanisms regulate pro-
duction and consumption systems (SO2). Economies present re-
dundancies like demand for the same resource in different sectors and
price linkages of different resources, causing a general system reaction
that can (partially) counteract the effects of a sustainability-oriented
action (Santarius, 2016). For example, price reduction by a given
supply-chain actor due to efciency gains can enact efciency invest-
ment by another actor aiming to keep relative prices even and avoid
market losses. In addition, the choice of some people to pursue suf-
ciency measures may lead to energy price drops that will encourage
other people to increase their consumption (Sorrell et al., 2020). The
same applies to efciency measures (Lange et al., 2021). Thus, critical
stabilising mechanisms maintain resource use at high levels.
The interplay of reinforcing and balancing mechanisms can result in
small changes in production and consumption factors leading to
huge amplications (SO3). Local efciency improvements can unfold
in signicant market changes, leading to substantial resource use (Gil-
lingham et al., 2016; Ruzzenenti and Basosi, 2008). For instance, some
argue that the effects of fuel efciency might have been the critical
driver of production outsourcing by shifting the relative costs out of
local production and storage to global production and transportation
(Ruzzenenti and Basosi, 2008). A similar argument could apply to con-
necting the improvement of Watt’s steam engine and the rst industrial
revolution, which led to the acknowledgement of Jevons paradox
(Sorrell, 2009). Such amplications can be more evident through
macro-level RE mechanisms such as the composition effect, where
changes in the relative return of investment in the sector will lead that
sector to grow relative to others, reinforced by growth effects through
innovation (Gillingham et al., 2016).
3.1.9. History-dependent systems
In history-dependent systems, previous decisions dene the de-
cisions available now, leading to fundamentally different dynamics
between doing and undoing (Sterman, 2000). To begin with, RE are
outcomes of history-dependent systems as co-dependence between
sustainability actions and transition dynamics inuence RE (HD1).
RE occurrence and magnitude might vary due to different timings in
transitions and the sustainability action at hand, as the sustainability
transition pathway of a given system inuences the effects in other
systems (Hertwich, 2005). For instance, car electrication might unfold
differently according to the region’s energy transition status, affecting
potential RE magnitude. Also, choosing a CE strategy (e.g., recycling)
might lead to adverse effects in adopting another CE strategy (e.g.,
remanufacturing) in the future due to higher opportunity costs, leading
to a less-than-ideal resource use (Castro et al., 2022; Figge and Thorpe,
2019). On the other hand, technological innovation can also contribute
to other emission-reducing activities (Hertwich, 2005). For instance,
improvements in batteries for car electrication could enable aircraft
electrication.
Finally, the inertia of systems might inuence the timing and
magnitude of RE (HD2). Some actions, such as infrastructure and in-
vestment, are hard to revert and frequently necessary to sustain a given
good’s consumption and production (Sonnberger and Gross, 2018). In
addition, investments made when energy prices are high continue when
prices go down (Hertwich, 2005). In addition, defaults and habits are
sources of inertia as people tend to keep their decision patterns (Exa-
daktylos & van den Bergh, 2021). Thus, choices might not be exible
and seemingly good sustainability actions that might lead to substantial
RE need time to be reverted (Binswanger, 2001). Also, consumption
practices and production systems co-evolve, which enhances the inertia
of technological and institutional development (Sonnberger and Gross,
2018). For instance, more effective air-conditioning systems can make it
possible to wear a suit and tie at the workplace regardless of the outside
temperature – a clear example of an efciency-oriented action enabling
a resource-consuming social norm. Although the co-dependence in
transitions and the inuence of systems’ inertia can inuence RE, a clear
line to separate RE and other similar phenomena (e.g., path dependence
and lock-in) is needed.
3.2. The multiple levels of understanding for RE examination
Fig. 3 depicts the 24 dynamic complexities following the Iceberg
model. The model claries RE-specic dynamic complexity in four levels
of understanding, from events to mental models. It helps position the RE-
specic dynamic complexities that: (i.) become evident through RE
descriptions (i.e., events); (ii.) associate with the behaviour over time of
systems (i.e., patterns of behaviour); (iii.) associate with how parts
interrelate and cause potential RE (i.e., underlying structures); and (iv.)
indicate individuals’ assumptions sustaining systems’ structure (i.e.,
mental models). Connections, proximity, and the clusters in the gure
indicate relationships identied between RE-specic dynamic com-
plexities enabled by the Iceberg model.
When investigating RE, the tip of the iceberg is to acknowledge that
the occurrence and magnitude of RE are a snapshot of the RE phenomenon
as they represent the observable events or symptoms of the system
structure. RE can reach different magnitudes under different contexts
and conditions for examination.
The next level of thinking entails acknowledging that the continuous
behaviour of RE can help understand the reasons for RE occurrence and
magnitude. It considers RE timing and magnitude depending on the
sources of inertia in the system and the various timings between actions
and responses to change. Also, there are potential disproportional am-
plications in the system. In the case of multiple RE occurring, potential
delays between them should be considered.
Addressing the underlying structures can clarify the relationships,
information ows, and physical structures critical to understanding RE
occurrence. Three clusters of dynamic complexity leading to RE are
associated with structures that determine potential behaviour. First, it is
essential to acknowledge that multiple interrelated systems under
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
9
transition at different paces can lead to RE. Those systems are composed
of multiple components and interests, e.g., consumers and producers,
whose interactions will set the structures under examination. The
structures will be composed of reinforcing and balancing feedback loops
stimulating and regulating the system under investigation, where mul-
tiple causal relationships and feedback between factors with potential non-
linear relationships determine behaviour. Framing and scoping the
systems, their components, and the feedback relationships of interest
will determine the identication of RE.
Finally, helping expand the mental models of designers and decision-
makers while deeply accounting for the mental models of other in-
dividuals acting in the system is fundamental to understanding the RE
occurrence. For example, from one side, designers and decision-makers
might depart from good intentions and a narrow and subjective view of
the system they inuence. Addressing RE requires a broader system
understanding, including addressing the multidimensionality of sus-
tainability to preventing or mitigating RE. On the other side, individuals
in the system are inuenced by bounded rationality and social pressures
in their decisions, which will determine their choices and RE occurrence.
In summary, the Iceberg model’s four levels of understanding help
frame RE examinations. First, it makes the symptoms of the system
structure explicit by demonstrating potential RE occurrence and their
magnitude. It also supports examinations to go deeper into unveiling the
dynamic complexity of RE in sustainability transitions. Additional
insight might be achieved by acknowledging RE as the outcome of
continuous systems that may reach stability at different times and con-
ditions. Finally, deepening into the multiple systems and stakeholders
holding causal relationships and feedback mechanisms sustained by the
mental models of actors enables grasping the reasons for RE occurrence
and the places to position high-leverage interventions to address them.
4. The insofar use of SD to address dynamic complexities in RE
studies
Table 3 provides an overview of the dynamic complexities encoun-
tered in the six SD-based studies of RE following the investigation
framing, the explanation, assessment, and recommendations about RE
provided in the studies organised according to the levels of under-
standing in the Iceberg model (Table A1 in the Appendix provides a
detailed description of the encountered dynamic complexities).
The framing of the investigation varies according to the initial as-
sumptions for RE occurrence (PR1). Some studies start from the premise
Fig. 3. The Iceberg model of RE-specic dynamic complexities, making explicit the four levels of understanding for RE examination.
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
10
of RE occurrence, presenting it as the central phenomenon of analysis
(Freeman, 2018; Freeman et al., 2016). Others encounter RE while
investigating a sustainability-oriented action (Cavicchi, 2016; Dace
et al., 2014). Finally, there are also studies that categorise RE as a kind of
“side effect” of sustainability-oriented action while investigating a
similar phenomenon – e.g., setting it as a type of policy resistance (de
Gooyert et al., 2016) or an unintended consequence of environmental
action (Laurenti et al., 2016). The recommendations (PR4) tend to
follow the focus of the study by, for instance, (i) addressing policy
resistance by focusing on incentives reinforcing overlapping interests
(de Gooyert et al., 2016) or (ii) indicating ways to address RE to reach a
desired decrease in emissions (Freeman et al., 2016). It means that even
though RE might not be the central focus of the investigation, it is crucial
to support their identication and consider ways of weakening their
occurrence.
There are multiple potential sources of information to sustain
modelling and simulation studies. Literature reviews considering aca-
demic research (Dace et al., 2014; Freeman, 2018; Laurenti et al., 2016),
policy documents (Cavicchi, 2016; Dace et al., 2014; Freeman, 2018),
and a described existing case (Freeman et al., 2016) have been
employed. Thus, secondary data can be a valuable source for RE
investigation, not only by providing data for running the models but also
for model conceptualisation. When stakeholders were involved in the
modelling process, it happened through semi-structured interviews
(Cavicchi, 2016) and participatory modelling (de Gooyert et al., 2016),
which helped to address the multiple interests that are towards the
system (TC1).
As to the level of analysis (CC1), studies show a strong tendency to
macro-level investigations, as all six studies deal with macro level-
systems: the status of solid waste management system in Latvia (Dace
et al., 2014), Dutch energy transition (de Gooyert et al., 2016), theory
for consumer goods consumption (Laurenti et al., 2016), regional
implementation of bioenergy in Italy (Cavicchi, 2016), regional road
transport in the UK (Freeman et al., 2016) and a general theory for
macro-level RE (Freeman, 2018). This tendency might indicate oppor-
tunities for meso and micro-level RE investigations as they might occur
at all levels. Also, there is extended evidence of SD-based investigations
of transitions happening at all three levels (Guzzo et al., 2022).
Feedback mechanisms (SO1 and SO2) play an essential role to
explain the resource dynamics in the system in all studies. However,
there is no consensus on setting reinforcing or balancing feedback loops
as inherently good or bad in sustainability terms. In some cases, rein-
forcing loops are desired as they sustain sustainability transitions (de
Gooyert et al., 2016); in other cases, reinforcing loops might drive
consumption and be undesired (Cavicchi, 2016; Laurenti et al., 2016).
RE are mainly identied as feedback mechanisms, too. Studies pictured
RE as reinforcing loops that acted against the intentions of well-intended
balancing actions (Dace et al., 2014; Freeman, 2018). One study char-
acterised it as a balancing loop that counteracted the effects of rein-
forcing sustainability investments (de Gooyert et al., 2016). Meanwhile,
one study characterised it as a reinforcing loop that reinforced another
reinforcing – but undesirable – engine of consumption growth (Laurenti
et al., 2016). Thus, the kind of feedback loop to address RE seems to
depend on the conceptualisation of resource use in the system. Never-
theless, identifying the feedback mechanisms driving resource use and
modelling RE in terms of feedback seems helpful in explaining the sys-
tem dynamics.
Studies generally considered consumer- and producer-side reactions
(GF1) to investigate the system and make explicit multiple causal re-
lationships (NL1) activating them. Also, there is a tendency to include
economic-oriented decisions resulting in RE as price or cost-led addi-
tional demand (Dace et al., 2014; de Gooyert et al., 2016; Freeman et al.,
2016; Laurenti et al., 2016) and operational efciency leading to addi-
tional production (Freeman, 2018). Meanwhile, social norms (Ad3)
were identied as sustaining detrimental behaviour and leading to RE in
one case (Freeman et al., 2016). In another case, RE occurs due to a local
effect disparate to the global intentions (GF2) – i.e., local warming in
contrast to intentions in decreasing GHG emissions (Cavicchi, 2016).
Although economic-oriented decisions still dominate SD-based studies,
the inclusion of the role of social norms and seemingly unrelated
behaviour showcase the potential to help identify and consider other
types of feedback effects. Nevertheless, no explicit consideration of the
effects of bounded rationality in decision-making (Ad3) in RE occur-
rence is remarkable because SD constantly challenges the idea of perfect
rationality and can integrate degrees of limitation in human
decision-making in its simulation models (Sterman, 2000). Also, there is
Table 3
Overview of dynamic complexities encountered in SD-based studies of RE. [1] refers to Dace et al. (2014), [2] refers to de Gooyert et al. (2016), [3] refers to Laurenti
et al. (2016), [4] refers to Cavicchi (2016), [5] refers to Freeman et al. (2016), and [6] refers to Freeman (2018).
Level of understanding The dynamic complexities in systems leading to RE Investigation Explanation Assessment Recommendations
1. Events PR2: Different magnitudes and directions [1],[5],[6]
NL3: High heterogeneity [1],[6]
2. Patterns of behaviour CI1: Delays between action and RE [1]
CI2: Short-run and long-run responses [5] [5],[6]
CI3: Delays between RE
SO3: Small changes, huge amplications [3],[5]
HD2: Inertia inuences timing and magnitude [4]
3. Underlying structures CC1: Micro, meso, and macro levels All [4],[6] [4]
CC2: Nested and interrelated [2],[4] [1]
TC1: Multiple interests [2],[4] [2] [5] [4]
GF1: Consumer and producer-side reactions [2],[3],[5],[6] [1]
GF2: Seemingly unrelated behaviour [4],[5] [4]
NL1: Multiple cause-and-effect [All] [6]
NL2: Non-linear relationships [5]
SO1: Reinforcing mechanisms All [2] [2],[3]
SO2: Balancing mechanisms [1],[2],[4], [5],[6] [2] [3]
HD1: Co-dependence with transitions [2]
4. Mental models PR1: Systemic responses to well-intended actions All All [6]
PR3: Narrow view of the system [4]
PR4: Deep understanding to prevent All
TC2: Multidimensionality of sustainability [2],[4],[6] [1],[5]
Ad1: Systems are subjective and evolutionary [5] [2],[3],[6] [6]
Ad2: Bounded rationality
Ad3: Social systems inuences [5] [1]
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
11
space to further addressing the unfolding of different RE mechanisms as
no study was capable to show the potential delays between them (CI3).
As to the modelling approach, most studies used only qualitative
modelling to make assertions about the systems under investigation
(Cavicchi, 2016; de Gooyert et al., 2016; Laurenti et al., 2016), while
two of them combined qualitative and quantitative modelling.
1
For
instance, in the packaging waste management case (Dace et al., 2014)
and the regional road transport in the UK (Freeman et al., 2016), the RE
magnitude is quantied in simulation and explained through the CLD
model. Qualitative models are often used to represent the subjectivity in
systems’ understanding (Ad1) by representing different models for
different assumptions of RE occurrence (Freeman et al., 2016). The
modes of behaviour emerging from the model structure may help
investigate potential system evolutions (Ad1) to draw policy recom-
mendations (PR4) (de Gooyert et al., 2016; Freeman et al., 2016; Lau-
renti et al., 2016). Quantitative scenarios show that RE magnitude varies
through time (PR2) (Dace et al., 2014). Also, different policies can lead
to entirely different behaviours for key variables, indicating potential
amplications from small changes (SO3) and non-linear relationships
between system elements (NL2) (Freeman et al., 2016). Thus,
combining qualitative and quantitative SD in investigations seems to be
a strong approach to address RE complexities more thoroughly.
Studies show a tendency to recommend combinations of policies
based on the increased understanding of the systems (PR4), sustained by
the argument that focusing on one part of the system might lead to
resistance in another. Authors are very vocal in recommending policies
that “simultaneously” address (Dace et al., 2014), “combinations of
policies” (de Gooyert et al., 2016), a “policy portfolio” (Cavicchi, 2016),
and “a system of interventions” (Freeman et al., 2016). Qualitative SD
demonstrates its usefulness as it provides a comprehensive map that can
help see where to position the multiple policies as the leverage points
become apparent from the structure and expected behaviour emerging
from it. In turn, simulation enables investigating the recommendations
to help understand how they could play out in practice. Freeman et al.
(2016) provide an interesting reference for multi-policy investigation, as
it shows the combined effect of four interventions, including the po-
tential for RE.
In general, the studies present evidence of investigating the dynamic
complexities of RE within all the four levels depicted in the Iceberg
model beyond the level of events. Some studies acknowledged RE as the
outcome of continuous systems, in some cases demonstrating the RE
magnitudes over time (PR2) with different behaviours, eventually
reaching system stabilizations (e.g., magnitudes rising or decreasing
before stabilising – CI2). Simulation (Dace et al., 2014) and CLD in-
vestigations (Freeman, 2018) sustained the discussions on the contin-
uous trait of systems leading to RE occurrence. The studies consistently
made explicit the underlying structures sustaining the RE occurrence.
Most of the studies recognize that RE are the outcome of consumer and
producer-side reactions (GF1), involving multiple cause-and-effect re-
lationships (NL1) and emerging from the interplay of reinforcing and
balancing feedback loops (SO1 and SO2). Finally, there is evidence of
making explicit the mental models of decision-makers by considering
studies with varying assumptions for RE occurrence (PR1) and making
recommendations that match the insights gained from modelling and
simulation (PR4). Also, the subjectivity of the decision-makers’ mental
models (Ad1) was made explicit by using concurrent models to explain a
system’s behaviour in multiple studies. Meanwhile, there is plenty of
space to clarify the mental models of individuals acting in the system
(Ad2 and Ad3).
A combined view of the set of RE-specic dynamic complexities, the
levels of understanding of the phenomenon, and the modelling stages
enable an actionable approach to investigating RE. When setting the
examination, a researcher or practitioner can identify and prioritise
which sources of dynamic complexity will play a role in their case. Then,
the Iceberg model will help position which level of thinking is necessary
to understand the reasons for RE occurrence while considering the pri-
oritised RE-specic dynamic complexities. Finally, the modelling steps
for setting the investigation will help set the modelling strategies to
adequately frame, explain, assess and make recommendations about the
system so that one can avoid or address RE occurrence. Qualitative and
quantitative SD modelling and simulation can help unveil the dynamic
complexity of RE in sustainability transitions and lead researchers and
practitioners closer to addressing RE occurrence.
5. Discussion: Research paths for further addressing the
dynamic complexities of RE
The accumulated knowledge from the awareness of the dynamic
complexities surrounding RE research and the insofar use of SD to
address them lead to the proposition of three research paths to further
address RE’s dynamic complexities. Each path is detailed in such a way
that we make explicit the dynamic complexities that could be resolved
by following each path.
•Research path 1: Help decision-makers understand the reasons for RE
and identify effective leverage points.
RE examination approaches should help expand decision-makers’
mindset to acknowledge not only the expected and intended conse-
quences of actions as to appreciate the potential unexpected and unin-
tended consequences (PR1 and PR3). Investigations show that
qualitative modelling can help identify (Cavicchi, 2016; de Gooyert
et al., 2016) and draw recommendations (Laurenti et al., 2016) to deal
with RE, alongside the use of simulation results to derive (Dace et al.,
2014) and even test (Freeman et al., 2016) them.
Investigation approaches should help integrate the concept of RE
into sustainability thinking accounting for the long-term, system-wide
effects of policies and strategies before implementation (van den Bergh,
2011). Therefore, qualitative SD can provide insight into helping
decision-makers and scholars understand the reasons for RE and identify
leverage points that can weaken or prevent them. Quantitative SD can
help clarify the determinants for RE occurrence and the most effective
prevention or mitigation mechanisms.
When addressing RE, the focus must expand from the events (e.g., RE
is likely to occur, and the magnitude of RE is x%) to patterns of
behaviour (e.g., how the occurrence and magnitude of RE behave
through time). Also, it should make decision-makers aware of the
occurrence, magnitude, and direction of potential RE through time
(PR2) and make sense of the reasons (PR4). Furthermore, making the
dynamic behaviour of system elements explicit will help deal with the
different short-run and long-run responses to change (CI2) and the de-
lays between different types of RE (CI3) to avoid inaccurate snapshots of
RE occurrence and magnitude. Finally, the investigations should trans-
late an increased understanding of the system into recommendations to
prevent or mitigate RE.
•Research path 2: Reach generalisable cause-and-effect structures that
explain the systemic responses leading to RE.
The consistent use of feedback loops acting against well-intended
1
SD studies can use qualitative modelling through causal loop diagrams
(CLD) to articulate the endogenous causal understanding of a system and
challenge individuals’ assumptions about a given system’s structure and its
potential behaviours (Gr¨
osser and Schaffernicht, 2012; Lane, 2008). Qualitative
modelling aims to identify feedback loops, i.e., successions of cause-effect re-
lations that start and end in the same system element, and which interplay can
lead to specic patterns of behaviour (Barlas, 2002). Quantitative simulation
uses stock and ow diagrams (SFD) to reach scenario-based analyses of po-
tential behaviour emerging from system structure (Sterman, 2000). Both ap-
proaches can contribute to knowledge about RE and can be combined.
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
12
desirable actions (Cavicchi, 2016; Dace et al., 2014; de Gooyert et al.,
2016; Freeman, 2018; Freeman et al., 2016) or reinforcing undesirable
behaviour (Laurenti et al., 2016) to represent RE provides evidence that
RE are sustained by feedback structures and hold dynamic behaviour. A
catalogue of cause-and-effect structures, potentially based on CLD, could
help decision-makers make sense of RE occurring in the micro, meso,
and macro levels (CC1), helping them to clarify the interconnections and
consider the different RE concomitantly at play (CC2).
The causal explanations for RE make clear multiple triggers,
including economic (Dace et al., 2014; de Gooyert et al., 2016; Freeman
et al., 2016; Laurenti et al., 2016) and behavioural (Freeman et al.,
2016) reasons for RE. The catalogue of cause-and-effect structures must
be capable of making explicit the structures giving rise to consumer and
producer-side reactions (GF1) within similar and different consumption
needs (GF2). Meanwhile, it is critical to thoroughly address the bounded
rationality of individuals (Ad2) and the social inuences on behaviour
(Ad3).
The cause-and-effect structures should help make sense of the po-
tential multiple causes-and-effect relationships leading to RE while
clarifying the factors holding a moderating and mediating effect on its
occurrence (NL1) and appreciating potential positive effects (PR2). The
structures should easily connect to the reinforcing and balancing
mechanisms that sustain resource usage (SO1 and SO2), so that they can
be instantiated into specic examinations.
Finally, the capacity to derive recommendations (Dace et al., 2014;
de Gooyert et al., 2016; Laurenti et al., 2016) and discuss potential
pathways (Freeman, 2018) for RE from the structures in different cases
indicates the structure-behaviour relation enabled by CLD can provide
essential insights in understanding RE. Generalisable cause-and-effect
structures and their complimentary behaviour could assist in address-
ing the need for a rigorous codication of the mechanisms through
which RE emerge (Brockway et al., 2021; Madlener and Turner, 2016;
Ruzzenenti et al., 2019; Sonnberger and Gross, 2018).
•Research path 3: Employ modelling and simulation as agile and
engaging tools that enable proactive decision-making.
Different levels of engagement might occur in RE investigations,
from using existing long-term case studies (Freeman et al., 2016) to
participatory modelling based on the SD group model-building approach
(de Gooyert et al., 2016; Vennix, 1999). On the one side, existing sources
of information (such as reports and literature) can help identify
reasoning for modelling and datasets for simulation calibration. On the
other side, involving decision-makers in the process can help reach more
valid models and recommendations as they are specialists in the subject
with a further potential of being applied in practice.
Qualitative modelling and quantitative simulation can be essential
allies in making sense of how the multiple agents are causing (and could
help resolve) RE (TC1) while contributing to the multidimensionality of
sustainability outcomes in RE (TC2). The tools should enable including
the critical factors determining RE in the context under investigation
(NL3) and the essential co-dependent sustainability transitions (HD1).
Furthermore, it is critical to consider the signicant delays between
designing actions, their implementation, and RE occurrence (CI1) to
enable proactive decision-making to avoid RE. If the intention is to make
ex-ante investigations and inform decision-making before RE occur, one
should take stock of how much uncertainty and little time there is to
consider all the potential outcomes of actions. The fact that only one
study simulated policies’ effects (Freeman et al., 2016) indicates that
going through the entire process of investigation, model building and
testing, and policy analysis can be challenging and time-consuming.
Thus, it is critical to develop structured and agile forms of engaging
with decision-makers and deploy the cause-and-effect structures into
simulation models so that the investigation process integrates seam-
lessly into practical decision-making.
6. Final remarks
This work aimed to enhance the understanding of RE’s causal and
dynamic traits, following SD as the investigation frame. The systematic
literature review, employing inductive and deductive content analysis,
resulted in 24 specic sources of dynamic complexities (RQ1), providing
a comprehensive overview of the factors that increase interrelationships
or alter the temporal interaction of system elements to explain RE.
Furthermore, it evidences RE as a complex phenomenon. The Iceberg
model connected the RE-specic dynamic complexities to the multiple
levels of thinking: from events and patterns of behaviour of observable
phenomena to more fundamental underlying structures and mental
models causing that behaviour. Going deeper into the iceberg enhances
the potential for understanding the causes of RE occurrence, providing
additional leverage to prevent or mitigate them. In addition, it makes
explicit the limitations of getting snapshots of RE magnitudes and that
investigators must approach RE with the appropriate mindset and tools
to address inherent uncertainty in understanding and managing them.
Therefore, the dynamic complexities should be integrated into every
kind of investigation and is an invitation to additional lenses to engage
in RE studies.
This research also shows how SD has been employed to address dy-
namic complexities in six documented RE cases (RQ2). It makes explicit
that SD-based studies have addressed the sources of dynamic complexity
in different ways. Addressing the dynamic complexities depend on the
modelling approach (i.e., qualitative or quantitative), the centrality and
assumptions about the RE phenomenon in the study, the choices about
how to represent the intended and unintended consequences using
feedback loops, and other aspects of the investigation made explicit in
Section 4. The applicability of the set of dynamic complexities to analyse
existing RE cases demonstrates its potential as an instrument to help
identify and manage them in further analyses. Thus, it could also be used
proactively by assisting investigators in mapping how the dynamic
complexities might unfold in the systems of interest to guide how to
design the investigation to tackle them. Here, case studies that depart
from using the conceptual framework are welcome, making explicit the
scope decision and modelling strategies to deal with the RE-specic
dynamic complexities of those cases.
The foundation for taking a system’s approach to investigating RE is
completed by drawing research paths to further address the dynamic
complexities in RE through SD (RQ3). RE research should strive to help
decision-makers understand the reasons for RE occurrence and identify
effective leverage points, reach generalisable cause-and-effect structures
that explain the systemic responses leading to RE, and employ modelling
and simulation as agile and engaging decision-making that enable pro-
active decision-making. The research paths demonstrate a less deter-
ministic approach for RE examination, capable of reaching
recommendations that consider the nature of the phenomenon – i.e.,
capable of dealing with the uncertainty surrounding RE occurrence and
magnitude.
Within the SD realm, a few state-of-the-art modelling and simulation
techniques can help address the RE-specic dynamic complexities. For
instance, automated loop dominance analysis (Schoenberg et al., 2020)
can connect the RE behaviour patterns identied in models to the actual
structures at play through time. Also, simulation-based role-playing
games (Rooney-Varga et al., 2020) have the potential to enable partic-
ipants to play with the model and learn about the causes of RE. Finally,
algorithmic tools (Kwakkel and Pruyt, 2013; Schoenenberger et al.,
2021) can be coupled to simulation models to generate automated
policy recommendations, assisting in identifying leverage points to
addressing RE.
The choice of building upon the SD approach was an attempt to make
tangible how a system perspective can address the complexity sur-
rounding RE. Moreover, the SD community is deeply involved in
addressing sustainability-related issues (Honti et al., 2019; Moon, 2017)
and fundamentally deals with unintended consequences (Forrester,
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
13
1971; Sterman, 2001). Thus, further understanding why and how it
occurs can spill over in helping to understand other “side-effects” such as
policy resistance (de Gooyert et al., 2016) and externalities (Laurenti
et al., 2016).
Nevertheless, SD modelling and simulation presents several limita-
tions. First, it is a common criticism that SD models can be highly ab-
stract as they rely on the aggregated behaviour of average types of actors
– this could be addressed by combination with Agent-based modelling
(ABM), for example. Additionally, as with any other modelling
approach, SD models rely on the modellers’ assumptions and modelling
choices, which inuences the models’ validity and requires strong model
validation and calibration, which can be made by following rigid pro-
cesses for model validation (e.g., Schwaninger and Groesser, 2016).
Also, a signicant challenge is connecting with existing models already
used to investigate RE and related phenomena, such as general equi-
librium models from economics and impact assessment models from
engineering backgrounds. Regardless of the modelling approach to
assess RE, researchers and decision-makers must be aware of the dy-
namic complexities, how they might inuence choices in understanding
the system, and to what extent they address the inherent uncertainty in
RE investigations. Unrestricted engagement is needed to address such a
complex issue and facilitate timely sustainability transitions that reach
their intended outcomes.
Declaration of competing interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Acknowledgements
This project has received funding from the European Union’s Hori-
zon 2020 research and innovation programme under the Marie Skło-
dowska-Curie grant agreement no. 899987. Co-funded by the European
Union (ERC, REBOUNDLESS, 101043931). Views and opinions
expressed are however those of the author(s) only and do not necessarily
reect those of the European Union or the European Research Council.
Neither the European Union nor the granting authority can be held
responsible for them.
Appendix
Table A1
Dynamic complexities encountered in SD investigations of RE.
Reference Investigation framing RE explanation RE assessment RE recommendations
Dace et al.
(2014)
•Did not investigate RE directly;
encountered RE by examining how
public policies affect waste
management systems (PR1).
•The investigation applies
qualitative and quantitative
modelling and builds upon
literature review and policy
documents.
•The study is about the status of the
solid waste management system in
Latvia (macro-level) (CC1).
•RE occur due to systemic responses
to policies that replace virgin
material with recycled material,
decreasing the price/costs of using the
material and expanding demand
(PR1).
•A CLD composed of two balancing
loops and a reinforcing loop explains
the reason for RE occurrence. The two
balancing loops are two economic
mechanisms (SO2), where the rst is
a packaging tax that controls the
demand for material, and the second
is the landll costs driving more
sorters and recycling. Multiple
factors determine the demand for
material, such as the price of recycled
and virgin material, a tax for
packaging, and the fraction capacity
(NL1).
•The RE is characterised by a
reinforcing economic mechanism
(PR1 and SO1), where more recycling
drives the demand for material and
thus more waste, which drives
recycling.
•The simulation model comprises
three sub-models (market, waste
management, and sorting), which have
their interrelations dened (CC2).
•Consumer-side reactions are
modelled in the market sub-model,
which considers change of attitude in
sorting through time (GF1).
Individuals are driven by
environmental concerns and
inuenced by the positive example
of those who sort their waste (Ad3).
Producer-side reactions include the
waste management dynamics and
supply of recycled material (GF1).
•Scenarios consider introducing
taxes and different elasticities of
demand and material substitution
(PR2 and NL3). In addition, RE
assessment considers the amount of
material used per product and the
lled fraction of landlls (TC2).
•The behaviour over time shows
different points of stabilisation
concerning policy implementation in
the scenarios (CI1). Also, the scenarios
show that RE magnitude varies
through time (PR2). In one case, the
magnitude reaches a plateau and
decreases until stabilisation (CI2).
•As a general recommendation,
authors argue that policies should
simultaneously replace virgin with
recycled material and increase sorted
waste but ensure the material price
does not decrease. Different scenarios
demonstrate different instruments.
Authors Argue that a combination of
the different instruments must be
applied depending on the target
(PR4).
De Gooyert
et al.
(2016)
•RE is not the central phenomenon
investigated. Instead, RE leads to
policy resistance in sustainability
transitions (PR1).
•The investigation applies
qualitative modelling and builds
upon participatory modelling
involving 96 participants in 8
workshops (TC1).
•Focuses on the national Dutch
energy transition (Macro-level)
(CC1).
•The argument is centred on nding
leverage points in feedback loops
that sustain (reinforce) or hinder
(balance) sustainability transitions
(SO1 and SO2). Multiple factors
inuence the investment in
renewables: energy market price, civil
engagement, and the cost of energy
production (NL1).
•Includes both consumption and
production side reactions (GF1) – e.
g., the feedback in energy sufciency
by consumer-side production in
lowering the market price that will
•The modes of behaviour that
emerge from analysing the model
(Ad1) show that several mechanisms
play against the investment in
renewable energy (SO2), leading
space to encounter reinforcing
mechanisms that might unlock the
transition (SO1). The argument for
the energy transition relies on phasing
out fossil fuels, as the economies of
scale and vested interest in the
incumbent energy system play against
the transition (HD1).
•Policy recommendations are designed
based on leverage points identied
in the model to address policy
resistance: releasing the power of
vested interest through creative
destruction, which will release the
government to focus on overturning
policies (SO1 and PR4). Argues that
interventions should be combined, as
focusing in only one part of the system
will result in resistance in another part
(PR4). The study does not address the
RE specically.
(continued on next page)
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
14
Table A1 (continued )
Reference Investigation framing RE explanation RE assessment RE recommendations
bring fewer investments. Also,
explicitly consider societal
interests via the calming effect of
sustainable energy production in civil
unrest (TC1). The study considers
technological, ecological, social,
economic, and political factors,
identifying them as sub-systems that
inuence each other (CC2).
•In the model, RE occurs as a
balancing loop to energy system
sustainability, activated by a
decrease in price and costs (PR1 and
SO2). Investment in renewable
energy is the proxy variable for the
energy transition (TC2).
Laurenti
et al.
(2016)
•RE is not the central phenomenon
investigated. Instead, RE is an
unintended consequence of
environmental action (PR1).
•The investigation applies
qualitative modelling and builds
upon literature review.
•The study takes broad system
boundaries to illustrate the dynamics
of physical consumer goods
(Macro-level) (CC1).
•The argument centres on the
reinforcing loop between
innovation, product obsolescence,
consumption, and economic
growth (SO1). Multiple factors
determine consumption as the
lifespan of products and the consumer
costs (NL1).
•It includes both consumption and
production-side reactions – e.g.,
innovation and efciency measures by
producers and increased consumption
responding to those measures (GF1).
•RE occurs as a reinforcing loop for
consumption, activated by
consumer costs decreases due to
increased efciency (PR1 and S12).
Negative externalities are also
mapped as reinforcing loops that
drive consumption due to non-
internalised consumption costs (SO1).
Meanwhile, several other negative
environmental and social impacts are
named ripple effects.
•Waste pollution is the proxy
variable for environmental impacts,
while economic inequalities for
social impacts (TC2).
•The modes of behaviour that
emerge from analysing the model
(Ad1) show that incremental
efciency improvements will result
in more signicant waste
generation, reverberating into
negative environmental and social
impacts (SO3).
•Policy recommendations are designed
based on leverage points identied
in the model: consumption and
incremental innovation (SO1 and PR4),
which lead to economic growth without
additional resource consumption.
Recommendations are included in the
model. For instance, they show that
environmental policy instruments
could help internalise costs and
counterbalance the externalities
mechanisms (SO2). In contrast, it does
not address the RE specically.
Cavicchi
(2016)
•Did not investigate RE directly;
encountered RE by investigating
regional biogas development in a
region and its impacts on sustainable
development (PR1).
•The investigation applies
qualitative modelling and builds
upon semi-structured interviews
and public reports. Informants
include bioenergy producers,
farmers’ union members,
governmental actors, and members of
local committees (TC1).
•Focuses on regional bioenergy
adoption in northern Italy (Macro-
level) (CC1).
•The model encompasses economic,
environmental, social, and
technological processes (CC1). In
addition, the use of colours in the
model makes explicit the
interrelations of the different sub-
systems (CC2).
•Multiple factors determine biogas
production as reinvestments from
prots, which are reinforced by
governmental incentives (NL1).
•The argument is based on several
reinforcing and balancing loops
that emerge in the economic,
environmental, and social spheres
(SO1, SO2, and TC2). For instance,
from an economic perspective,
reinforcing loops enhance
producers’ prot by lowering costs or
increasing revenues (SO1). From an
environmental perspective,
balancing loops show the intended
consequences in controlling GHG
emissions while reinforcing loops
communicate the unintended ones
(PR1, SO1, and SO2). GHG emission
is adopted to assess sustainability
impacts (TC2).
•Local warming and the emissions
from intensied trafc in the area
•The modes of behaviour that emerge
from analysing the model show the
tensions between the contribution
to European targets and the
regional environmental and social
effects (CC1 and GF2)
•Based on the idea that the identied
problems emerged from a prot-
oriented policy (PR3), the author
argues for a policy portfolio that could
inuence the different feedback
loops at play and align the different
interests (PR4 and TC1). However, the
identied RE are not explicitly
addressed in the policy discussions.
(continued on next page)
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
15
Table A1 (continued )
Reference Investigation framing RE explanation RE assessment RE recommendations
are two of the identied RE (PR1,
CC1 and GF2). From a social
perspective, farmland rent price
raises the odds of conicts and
hampers cooperation (TC2).
•There are delays indicating the
time taken for building infrastructure
for biogas production and the time
taken between civil society pressure
and governmental reactions (HD2).
Freeman
et al.
(2016)
•RE is the central phenomenon.
Examine the causal mechanisms
leading to RE and discuss if it is
inevitable (PR1).
•The investigation applies
qualitative and quantitative
modelling and builds upon a long-
term case study on the RE
occurrence.
•Focuses on regional road
transport in the UK (Macro-level)
(CC1).
•Two conceptual models sustain the
argument – a no-rebound model and a
structural rebound model (Ad1). A
few factors determine how much
people drive, such as costs,
congestion, and social norms (NL1)
•The models make explicit a few
demand mechanisms, such as the
balancing loop that limits growth in
driving due to costs of fuels and
vehicles (SO2) and the reinforcing
loop of eet efciency increasing
travelling and more supply-side
investments (SO1), integrating
consumer and producer-side
reaction (GF1).
•RE occur as feedback loops that
increase distance driven per person,
activated by road building through
congestion, social norms for travelling
more, travel costs, and additional
income (SO1, SO2 and PR1). The
building of roads to release congestion
makes explicit RE occur from changes
in seemingly unrelated behaviour
(GF2). The feedback between social
norms and consumption rates
indicates the inuences of social
systems in RE (Ad3).
•There are delays indicated in a few
of the structures leading to RE
(CI2).
•The simulation model comprises
four sub-models representing the
theory: economic growth, social
norms, vehicles-in-use, and road
network (CC2).
•Exogenous uncertainty factors
inuence such as political ideology
that could prioritise private or public
transportation, regional policy and
science development (TC1).
•The model is calibrated against
historical data. Then, four scenarios
show the effect of behavioural change
for travelling less, technological
investment into eet efciency, the
inclusion of externalised costs, and
investment into public modes of
transportation (PR2). Total emission
is adopted as the variable to assess
sustainability impacts and RE (TC2).
Several other indicators relevant to
public and private stakeholders
assess the scenarios, such as the cost of
road travel per km and the eet
efciency in km per litre (TC1).
•Some indicators show exponential
growth or decline from one policy
to another, indicating potential
amplications by small changes in
factors (SO3) and non-linear
relationships between system
elements (NL2).
•Authors argue for a mix of policies to
inuence the strength and direction
of the different feedback loops at
play. The results indicate that
reductions in travel by individuals and
increased investment by the public
sector and industry are needed to reach
the desired decrease in emissions
while considering RE occurrence
(PR4). The four scenarios examine the
behaviour of combinations of
proposed policies.
Freeman
(2018)
•RE is the central phenomenon.
Examine the historical role of RE in
socio-technical systems and discuss
how RE magnitude might change in
the future (PR1).
•The investigation applies
qualitative modelling employing
hybrid modelling (i.e., CLD that
makes explicit critical stocks and
ows). It builds upon a literature
review of concepts of natural capital,
global ecological footprint, and the
great acceleration.
•The author adopts an “extremely
large” system boundary to support
the angle of the investigation (macro-
level) (CC1).
•The model comprises two main sub-
systems: socio-technical and natural
capital systems (CC1). The argument
is centred on the stocks of natural
capital, human-created capital,
and waste (TC2), the ows between
them, and the causal relationships
that enable them. Multiple factors
determine the production of goods
and services as population size,
consumption per person, and
availability of resources (NL1).
•The model demonstrates several
reinforcing and balancing loops
(SO1 and SO2). For example,
increased access to goods and services
leads to an increasing population due
to longer lives, which increases
consumption (SO1). Meanwhile,
limited natural stocks will inherently
limit the available resources for
growth (SO2). The model includes
consumers’ responses to lower costs
and producers’ investments in
technology and operations due to
increased sales (GF1).
•The core RE dynamics occur due
to a reinforcing loop from increased
consumption, driving technology
development up and decreasing
consumption costs (PR1).
•The potential system evolution is
discussed for three pathways based on
the interplay of the identied
feedback loops while considering the
potential role of four types of RE:
secondary, transformational, frontier
and international (PR1). The analysis
suggests that the RE will play a
different role and hold different
magnitudes according to the
system’s evolution (Ad1 and PR2).
This might indicate heterogeneity in
scenarios (NL3) due to moderating
or mediating factors (NL3) changes.
In some cases, it is suggested that RE
might disappear or even reverse,
indicating differences in short- to long-
run responses (CI2).
•Recommendations lead towards the
most desired pathway. The study
concludes that RE might be less or
more important according to the
pathway (Ad1) and only mentions that
policies and investment decisions
should be designed to avoid them
(PR4).
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
16
References
Azevedo, I.M.L., 2014. Consumer end-use energy efciency and rebound effects. Annu.
Rev. Environ. Resour. 39, 393–418. https://doi.org/10.1146/annurev-environ-
021913-153558.
Barlas, Y., 2002. System Dynamics: systemic feedback modeling for policy analysis.
Know. Sustain. Dev.: Insight Encycl. Life Support Syst. 1131–1175. https://doi.org/
10.1002/(SICI)1099-1727(199623)12:3<183::AID-SDR103>3.0.CO;2-4.
Binswanger, M., 2001. Technological progress and sustainable development: what about
the rebound effect? Ecol. Econ. 36 (1), 119–132. https://doi.org/10.1016/S0921-
8009(00)00214-7.
Boyack, K.W., Klavans, R., 2010. Co-citation analysis, bibliographic coupling, and direct
citation: which citation approach represents the research front most accurately?
J. Am. Soc. Inf. Sci. Technol. 61 (12), 2389–2404. https://doi.org/10.1002/
asi.21419.
Brockway, P.E., Sorrell, S., Semieniuk, G., Heun, M.K., Court, V., 2021. Energy efciency
and economy-wide rebound effects: a review of the evidence and its implications.
Renew. Sustain. Energy Rev. 141 (January), 110781 https://doi.org/10.1016/j.
rser.2021.110781.
Castro, C.G., Trevisan, A.H., Pigosso, D.A., Mascarenhas, J., 2022. The rebound effect of
circular economy: denitions, mechanisms and a research agenda. J. Clean. Prod.
345 (October 2021), 131136 https://doi.org/10.1016/j.jclepro.2022.131136.
Cavicchi, B., 2016. Sustainability that backres: the case of biogas in Emilia Romagna.
Environ. Innov. Soc. Transit. 21, 13–27. https://doi.org/10.1016/j.eist.2016.02.001.
Colmenares, G., L¨
oschel, A., Madlener, R., 2020. The rebound effect representation in
climate and energy models. Environ. Res. Lett. 15 (12) https://doi.org/10.1088/
1748-9326/abc214.
Dace, E., Bazbauers, G., Berzina, A., Davidsen, P.I., 2014. System dynamics model for
analyzing effects of eco-design policy on packaging waste management system.
Resour. Conserv. Recycl. 87, 175–190. https://doi.org/10.1016/j.
resconrec.2014.04.004.
Davelaar, D., 2021. Transformation for sustainability: a deep leverage points approach.
Sustain. Sci. 16 (3), 727–747. https://doi.org/10.1007/s11625-020-00872-0.
de Gooyert, V., Rouwette, E., van Kranenburg, H., Freeman, E., van Breen, H., 2016.
Sustainability transition dynamics: towards overcoming policy resistance. Technol.
Forecast. Soc. Change 111, 135–145. https://doi.org/10.1016/j.
techfore.2016.06.019.
Eck, N. J. Van, Waltman, L., 2014. Measuring scholarly impact. In: Measuring Scholarly
Impact. https://doi.org/10.1007/978-3-319-10377-8.
Economy, Circle, 2022. The Circularity Gap Report 2022, vol. 26. https://www.circula
rity-gap.world/2022.
Economy, Circle, 2023. The Circularity Gap Report 2023.
Elo, S., Kyng¨
as, H., 2008. The qualitative content analysis process. J. Adv. Nurs. 62 (1),
107–115. https://doi.org/10.1111/j.1365-2648.2007.04569.x.
Exadaktylos, F., van den Bergh, J., 2021. Energy-related behaviour and rebound when
rationality, self-interest and willpower are limited. Nat. Energy 6 (12), 1104–1113.
https://doi.org/10.1038/s41560-021-00889-4.
Figge, F., Thorpe, A.S., 2019. The symbiotic rebound effect in the circular economy. Ecol.
Econ. 163, 61–69. https://doi.org/10.1016/j.ecolecon.2019.04.028.
Font-Vivanco, D., van der Voet, E., 2014. The rebound effect through industrial ecology’s
eyes: a review of LCA-based studies. Int. J. Life Cycle Assess. 19 (12), 1933–1947.
https://doi.org/10.1007/s11367-014-0802-6.
Font-Vivanco, D., McDowall, W., Freire-Gonz´
alez, J., Kemp, R., van der Voet, E., 2016.
The foundations of the environmental rebound effect and its contribution towards a
general framework. Ecol. Econ. 125, 60–69. https://doi.org/10.1016/j.
ecolecon.2016.02.006.
Font-Vivanco, D., Sala, S., McDowall, W., 2018. Roadmap to rebound: how to address
rebound effects from resource efciency policy. Sustainability 10 (6), 1–17. https://
doi.org/10.3390/su10062009.
Forrester, J.W., 1971. Counterintuitive behavior of social systems. Theor. Decis. 2 (2),
109–140. https://doi.org/10.1007/bf00148991.
Freeman, R., 2018. A theory on the future of the rebound effect in a resource-constrained
world. Front. Energy Res. 6 (AUG), 81. https://doi.org/10.3389/fenrg.2018.00081.
Freeman, R., Yearworth, M., Preist, C., 2016. Revisiting Jevons’ paradox with system
dynamics: systemic causes and potential cures. J. Ind. Ecol. 20 (2), 341–353. https://
doi.org/10.1111/jiec.12285.
Friedrichsmeier, T., Matthies, E., 2015. Rebound effects in energy efciency -an
inefcient debate? Gaia 24 (2), 80–84. https://doi.org/10.14512/gaia.24.2.3.
Giampietro, M., Mayumi, K., 2018. Unraveling the complexity of the Jevons Paradox: the
link between innovation, efciency, and sustainability. Front. Energy Res. 6 (APR),
1–13. https://doi.org/10.3389/fenrg.2018.00026.
Gillingham, K., Kotchen, M.J., Rapson, D., Wagner, G., 2013. The rebound effect is
overplayed. Nature 493 (7433), 475–476. https://doi.org/10.1007/978-3-319-
04978-6_3.
Gillingham, K., Rapson, D., Wagner, G., 2016. The rebound effect and energy efciency
policy. Rev. Environ. Econ. Pol. 10 (1), 68–88. https://doi.org/10.1093/reep/
rev017.
Greening, L.A., Greene, D.L., Diglio, C., 2000. Energy efciency and consumption - the
rebound effect - a survey. Energy Pol. 28 (6–7), 389–401. https://doi.org/10.1016/
S0301-4215(00)00021-5.
Gr¨
osser, S.N., 2017. Complexity management and system dynamics thinking. In:
Dynamics of Long-Life Assets: from Technology Adaptation to Upgrading the
Business Model, pp. 69–92. https://doi.org/10.1007/978-3-319-45438-2.
Gr¨
osser, S.N., Schaffernicht, M., 2012. Mental models of dynamic systems: taking stock
and looking ahead. Syst. Dynam. Rev. 28 (1), 46–68. https://doi.org/10.1002/
sdr.476.
Guzzo, D., Pigosso, D.C.A., Videira, N., Mascarenhas, J., 2022. A system dynamics-based
framework for examining Circular Economy transitions. J. Clean. Prod. 333
(November 2021), 129933 https://doi.org/10.1016/j.jclepro.2021.129933.
Hertwich, E.G., 2005. Consumption and the rebound effect: an industrial ecology
perspective. J. Ind. Ecol. 9 (1–2), 85–98. https://doi.org/10.1162/
1088198054084635.
Honti, G., D¨
org˝
o, G., Abonyi, J., 2019. Review and structural analysis of system dynamics
models in sustainability science. J. Clean. Prod. 240 https://doi.org/10.1016/j.
jclepro.2019.118015.
Hsieh, H.-F., Shannon, S.E., 2005. Three approaches to qualitative content analysis. Qual.
Health Res. 15 (9), 1277–1288. https://doi.org/10.1177/1049732305276687.
Kim, D.H., 2000. Systems Thinking Tools: A User’s Reference Guide. https://thesystemst
hinker.com/wp-content/uploads/2016/03/Systems-Thinking-Tools-TRST01E.pdf.
Kwakkel, J.H., Pruyt, E., 2013. Exploratory Modeling and Analysis, an approach for
model-based foresight under deep uncertainty. Technol. Forecast. Soc. Change 80
(3), 419–431. https://doi.org/10.1016/j.techfore.2012.10.005.
Lane, D.C., 2008. The emergence and use of diagramming in system dynamics: a critical
account. Syst. Res. Behav. Sci. 25 (1), 3–23. https://doi.org/10.1002/sres.826.
Lange, S., Kern, F., Peuckert, J., Santarius, T., 2021. The Jevons paradox unravelled: a
multi-level typology of rebound effects and mechanisms. Energy Res. Social Sci. 74,
101982 https://doi.org/10.1016/j.erss.2021.101982.
Laurenti, R., Singh, J., Sinha, R., Potting, J., Frostell, B., 2016. Unintended
environmental consequences of improvement actions: a qualitative analysis of
systems’ structure and behavior. Syst. Res. Behav. Sci. 33 (3), 381–399. https://doi.
org/10.1002/sres.2330.
Madlener, R., Alcott, B., 2009. Energy rebound and economic growth: a review of the
main issues and research needs. Energy 34 (3), 370–376. https://doi.org/10.1016/j.
energy.2008.10.011.
Madlener, R., Turner, K., 2016. After 35 Years of rebound research in economics: where
do we stand?. In: Rethinking Climate and Energy Policies. Springer International
Publishing, pp. 17–36. https://doi.org/10.1007/978-3-319-38807-6_2.
Markard, J., Geels, F.W., Raven, R., 2020. Challenges in the acceleration of sustainability
transitions. Environ. Res. Lett. 15 (8) https://doi.org/10.1088/1748-9326/ab9468.
Matraeva, L., Vasiutina, E., Korolkova, N., Maloletko, A., Kaurova, O., 2022. Identifying
rebound effects and formulating more sustainable energy efciency policy: a global
review and framework. Energy Res. Social Sci. 85 (July 2021), 102402 https://doi.
org/10.1016/j.erss.2021.102402.
Metic, J., McAloone, T.C., Pigosso, D.C.A., 2022. Research avenues for uncovering the
rebound effects of the Circular Economy: a systematic literature review. J. Clean.
Prod. 74 (1934), 133133 https://doi.org/10.1016/j.jclepro.2022.133133.
Moon, Y.B., 2017. Simulation modelling for sustainability: a review of the literature. Int.
J. Sustain. Eng. 10 (1), 2–19. https://doi.org/10.1080/19397038.2016.1220990.
Polizzi di Sorrentino, E., Woelbert, E., Sala, S., 2016. Consumers and their behavior: state
of the art in behavioral science supporting use phase modeling in LCA and ecodesign.
Int. J. Life Cycle Assess. 21 (2), 237–251. https://doi.org/10.1007/s11367-015-
1016-2.
Reimers, H., Jacksohn, A., Appenfeller, D., Lasarov, W., Hüttel, A., Rehdanz, K.,
Balderjahn, I., Hoffmann, S., 2021. Indirect rebound effects on the consumer level: a
state-of-the-art literature review. Clean. Responsible Consum. 3 (April), 100032
https://doi.org/10.1016/j.clrc.2021.100032.
Rooney-Varga, J.N., Kapmeier, F., Sterman, J.D., Jones, A.P., Putko, M., Rath, K., 2020.
The climate action simulation. Simulat. Gaming 51 (2), 114–140. https://doi.org/
10.1177/1046878119890643.
Ruzzenenti, F., Basosi, R., 2008. The rebound effect: an evolutionary perspective. Ecol.
Econ. 67 (4), 526–537. https://doi.org/10.1016/j.ecolecon.2008.08.001.
Ruzzenenti, F., Font Vivanco, D., Galvin, R., Sorrell, S., Wagner, A., Walnum, H.J., 2019.
Editorial: the rebound effect and the Jevons’ paradox: beyond the conventional
wisdom. Front. Energy Res. 7, 90. https://doi.org/10.3389/fenrg.2019.00090.
Santarius, T., 2015. Micro-macro discrepancy and cause-effect relativity in rebound
research. Gaia 24 (2), 85–87. https://doi.org/10.14512/gaia.24.2.4.
Santarius, T., 2016. Investigating meso-economic rebound effects: production-side effects
and feedback loops between the micro and macro level. J. Clean. Prod. 134,
406–413. https://doi.org/10.1016/j.jclepro.2015.09.055.
Saunders, H.D., 2008. Fuel conserving (and using) production functions. Energy Econ. 30
(5), 2184–2235. https://doi.org/10.1016/j.eneco.2007.11.006.
Schoenberg, W., Davidsen, P., Eberlein, R., 2020. Understanding model behavior using
the Loops that Matter method. Syst. Dynam. Rev. 36 (2), 158–190. https://doi.org/
10.1002/sdr.1658.
Schoenenberger, L., Schmid, A., Tanase, R., Beck, M., Schwaninger, M., 2021. Structural
analysis of system dynamics models. Simulat. Model. Pract. Theor. 110 (April),
102333 https://doi.org/10.1016/j.simpat.2021.102333.
Schwaninger, M., Groesser, S., 2016. System dynamics modeling: validation for quality
assurance. In: Encyclopedia of Complexity and Systems Science, vols. 1–20. https://
doi.org/10.1007/978-3-642-27737-5.
Senge, P., 1990. The Fifth Discipline : the Art and Practice of the Learning Organization,
rst ed. Doubleday/Currency.
Sonnberger, M., Gross, M., 2018. Rebound effects in practice: an invitation to consider
rebound from a practice theory perspective. Ecol. Econ. 154 (August), 14–21.
https://doi.org/10.1016/j.ecolecon.2018.07.013.
Sorrell, S., 2009. Jevons’ Paradox revisited: the evidence for backre from improved
energy efciency. Energy Pol. 37 (4), 1456–1469. https://doi.org/10.1016/j.
enpol.2008.12.003.
Sorrell, S., Gatersleben, B., Druckman, A., 2020. The limits of energy sufciency: a
review of the evidence for rebound effects and negative spillovers from behavioural
change. Energy Res. Social Sci. 64 (February), 101439 https://doi.org/10.1016/j.
erss.2020.101439.
D. Guzzo et al.
Journal of Cleaner Production 405 (2023) 137003
17
Steffen, W., Richardson, K., Rockstr¨
om, J., Cornell, S.E., Fetzer, I., Bennett, E.M.,
Biggs, R., Carpenter, S.R., De Vries, W., De Wit, C.A., Folke, C., Gerten, D.,
Heinke, J., Mace, G.M., Persson, L.M., Ramanathan, V., Reyers, B., S¨
orlin, S., 2015.
Planetary boundaries: guiding human development on a changing planet. Science
347 (6223). https://doi.org/10.1126/science.1259855.
Sterman, J.D., 2000. Business dynamics: systems thinking and modeling for a complex
world. In: Management. Irwin McGraw-Hill. https://doi.org/10.1057/palgrave.
jors.2601336.
Sterman, J.D., 2001. System dynamics modeling: tools for learning in a complex world.
Reprinted from the California Management Review 43 (4), 8–25. https://doi.org/
10.2307/41166098.
Thiesen, J., Christensen, T.S., Kristensen, T.G., Andersen, R.D., Brunoe, B., Gregersen, T.
K., Thrane, M., Weidema, B.P., 2008. Rebound effects of price differences. Int. J. Life
Cycle Assess. 13 (2), 104–114. https://doi.org/10.1065/lca2006.12.297.
Thom´
e, A.M.T., Scavarda, L.F., Scavarda, A.J., 2016. Conducting systematic literature
review in operations management. Prod. Plann. Control 27 (5), 408–420. https://
doi.org/10.1080/09537287.2015.1129464.
Trincado, E., S´
anchez-Bay´
on, A., Vindel, J.M., 2021. The European Union green deal:
clean energy wellbeing opportunities and the risk of the Jevons paradox. Energies 14
(14), 1–23. https://doi.org/10.3390/en14144148.
Turner, K., 2013. Rebound” effects from increased energy efciency: a time to pause and
reect. Energy J. 34 (4), 25–42. https://doi.org/10.5547/01956574.34.4.2.
UNEP, 2021. Emissions Gap Report 21. https://www.unep.org/resources/emissi
ons-gap-report-2021.
van den Bergh, J.C.J.M., 2011. Energy conservation more effective with rebound policy.
Environ. Resour. Econ. 48 (1), 43–58. https://doi.org/10.1007/s10640-010-9396-z.
Vennix, J.A.M., 1999. Group model-building: tackling messy problems. Syst. Dynam.
Rev. 15 (4), 379–401. https://doi.org/10.1002/(SICI)1099-1727(199924)15:
4<379::AID-SDR179>3.0.CO;2-E.
Weidema, B.P., 2008. Rebound effects of sustainable production. Bridg. Gap; Response
Environ. Change –Words Deeds 1–5.
Wohlin, C., 2014. Guidelines for Snowballing in Systematic Literature Studies and a
Replication in Software Engineering.
Zink, T., Geyer, R., 2017. Circular economy rebound. J. Ind. Ecol. 00 (0), 1–10. https://
doi.org/10.1111/jiec.12545.
D. Guzzo et al.
