Resource accounting for a circular economy: evidence from a digitalised waste management system

ABSTRACT The transition to a circular economy requires accounting for all resource streams in the economy to facilitate the maintenance of these resources at their highest value and utility for as long as possible. In this paper, we investigate the role of resource accounting tools and practices in driving the transition to a circular economy by measuring resource streams and taking measures to promote circularity in those resource streams. In particular, we focus on (1) the role of digitalised resource accounting systems in enabling the identification and reporting of resource streams, and (2) the effect of using such information to incentivize household recycling behavior to promote more circularity in resource streams. Our empirical setting is the development of a digitalised waste management system in Western Norway. Our research design has three empirical pillars: qualitative interviews of managers in the industry, historical data analyses of waste generation and recycling, and two illustrative large-scale natural quasi-experiments on more than 170,000 households' recycling behavior. Our study illuminates the development of the resource accounting system and demonstrates the effect of using resource accounting information to incentivize households' recycling behavior through experimentation with a “pay-as-you-throw” (PAYT) system. Our findings show the manner in which the resource accounting system enabled measurement of resource streams and the use of that resource information to promote circularity among households. Furthermore, our experiments reveal the effectiveness of using resource accounting data for incentivizing recycling behavior. Thus, our study illuminates the role of accounting in the transition to a circular economy.


Introduction
The transition towards a circular economy is one of the most comprehensive responses to the current global sustainability problem (European Commission, 2020;WBCSD, 2010).The interrelated challenges of climate change and resource scarcity are increasingly seen as outcomes of our linear economy.Such an economy is based on a "take, make, dispose" model that decimates natural resources, is powered by non-renewable energy, and culminates in vast amounts of emissions, waste and pollution (Stahel, 2016).Conversely, a circular economy refers to "a regenerative system in which resource input and waste, emission, and energy leakage are minimised by slowing, closing, and narrowing material and energy loops" (Geissdoerfer et al., 2017, p. 759).A circular economy thus implies identifying and managing all resource streams in a manner that allows for reuse and recycling of all resources.
Our point of departure in this paper is that the transition to a circular economy requires accounting for all resource streams in the economy.How will we move towards a circular economy if we do not know which resources we have, how much of those resources end up in various waste streams, and to which degree we are successful in taking measures to maintain their value over time?Consequently, we need a resource accounting system that enables us to measure the current degree of (un-)circularity and to take measures based on that information that can make our economy more circular.In this paper, we investigate the role of resource accounting tools and practices in driving the transition to a circular economy by measuring resource streams and taking measures to promote circularity in those resource streams.The aim of our paper is thus twofold.First, we analyse the role of digitalised resource accounting systems in enabling the identification and reporting of resource streams.Second, we investigate the effect of using such information to incentivize household recycling behaviour to promote more circularity in resource streams.The first aim relates to measurement and reporting aspects of this accounting problem, while the second aim relates to its performance management aspect. 1he scope of the challenge of going circular is daunting.Recent efforts have been made to quantify the so-called "circularity gap", i.e. waste generation plus stock depletion, minus recovered materials (Aguilar-Hernandez et al., 2019).This is a global measure of all the waste materials that are theoretically available for circularity but are largely neither accounted for nor collected.Aguilar-Hernandez et al. (2019) found that the circularity gaps of countries in Europe and North America range from 1.6 to 2.2 tons per capita, while the global average is approximately half of this.Attempts are also made to estimate our current circularity gaps, by indicating "how circular" the global economy is.The Circularity Gap Reporting Initiative concluded that "[t]he world is now 8.6% circular" (de Wit et al., 2020).While such an estimate needs to be interpreted prudently, it at least provides a rough estimate of how far our current economy is from being circular.More fine-grained attempts at local levels to quantify circularity gaps and efforts to close them are necessary to succeed with such a transition.
We conceive of this as an accounting challenge in the following sense: The resource accounting system in waste management is a form of non-financial accounting that captures the flow of various waste streams from households' consumption and back to new production via recycling.In light of the EU Monitoring Framework for the Circular Economy (European Commission, 2018), this pertains to three goals and their associated indicators in the EU framework: . Indicator 3a-c ("Waste generation"), reflecting the goal of minimising waste generation . Indicator 5a-b ("Overall recycling rates"), reflecting the goal of increasing recycling as part of the transition to a circular economy. .Indicator 6a-f: ("Recycling rates for specific waste streams"), reflecting the goal of recycling key waste streams.
These goals and indicators reflect a subset of the transition to a circular economy, namely accounting for waste reduction, more fine-grained resource sorting and recycling behaviour.The latter is thus merely an important first step.
Accountants can play several roles in using resource accounting systems to make better business decisions for circularity and to incentivize behaviours that promote circularity among customers in waste management systems (households, companies, etc.).Waste management companies have long measured amounts of resource streams, but to a lesser extent turned this data into information for decision-making and business intelligence purposes.In developing such non-financial accounting systems, there is an important role to play for accountants in determining and developing the right kinds of performance indicators, analysing data to generate business intelligence and actionable insights, as well as using such information for various management control purposes.In the context of the circular economy, these roles are crucial for developing information systems about input, throughput and output of various resource streams for increased circularity.
This paper addresses the challenge of designing and using a digitalised resource accounting system to promote circularity in the form of recycling behaviours related to the indicators outlined above.It thus sheds light on the role of accounting in transitioning towards a circular economy.Our empirical setting is the measurement and digitalisation of resource streams in the waste management system in Western Norway and the use of resource accounting information to incentivize households' recycling behaviour.As pointed out by Ciulli et al. (2020), digital platforms are expected to be important in the transition towards circular business practices, and the digital resource accounting system in our study illustrates this.
Our empirical investigation comprises three pillars.First, we conducted qualitative interviews with managers in two waste management companies in Western Norway and supplemented these interviews with documents provided to us by them.This data provided rich, in-depth insight into the development of the resource accounting system.Second, we collected detailed resource accounting information for several different resource streams in nine Norwegian municipalities for the period from 2005 to 2019.Third, we conceived of the implementation of a so-called "pay-as-you-throw" (PAYT) system, whereby households were billed for actual waste generation, as a large-scale natural quasi-experiment in two rounds on more than 170,000 households. 2aken together, these empirical investigations enabled us to demonstrate how resource accounting can play a role in promoting circularity through waste management systems.More specifically, we demonstrate how a digital resource accounting system can be used to conduct behavioural interventions that promote circularity among households, which our natural experiment shows can be effective to promote circularity.Our paper thus provides empirical evidence on both the development of resource accounting tools and practices for promoting circularity in the Norwegian waste management system and the effectiveness of implementing behavioural interventions based on this system to promote household recycling behaviour.We thus contribute to the understanding of the role of a resource accounting system in promoting circularity and the use of such information to induce more recycling in households.These insights have policy implications for regulators trying to promote circularity through national recycling schemes, as well as for the waste management companies that are trying to stimulate household recycling by means of schemes such as PAYT.
The paper proceeds as follows.First, we discuss the role and importance of resource accounting in the transition to a circular economy.In doing so, we shed light on the nature of the circular economy as well as the current state of knowledge on accounting for circularity.Second, we outline our research design.Third, we present our findings.Finally, we discuss our findings and their implications.

Accounting for a circular economy
It is often stated that accounting is the language of business (Weiser, 1966), but will we have to expand its vocabulary, grammar and metaphors for the purposes of the circular economy?This question relates to the nature of performance in a circular economy and what that implies both for reporting of performance and for performance management.Accounting for societal and environmental impactso-called sustainability accounting, CSR reporting, or equivalent phenomenahas become a widespread practice (e.g.Lamberton, 2005;Lehman & Kuruppu, 2017;Parker, 2011).The integration of sustainability reporting with conventional financial reporting in so-called integrated reports is also becoming more common (e.g.Cheng et al., 2014;Gibassier et al., 2018).It is relatively recent, however, that circular economy concepts and indicators have been addressed as accounting topics, and the field of accounting for the circular economy is thus still somewhat in its infancy (e.g.Geng et al., 2012;Haupt et al., 2017).
As pointed out by Geng et al. (2016, p. 68), the transition to a circular economy "can be aided by integrating performance evaluations and accounting approaches for complex systems with multiple social and environmental dimensions", but "current performance indicators may be inadequate" (see also Maas et al., 2016;Mathews & Tan, 2016).For instance, there is a need for more sophisticated measures of resource efficiency (Di Maio et al., 2017).Before we shed light on accounting for the circular economy, however, we should consider the characteristics of such an economy and what the transition towards it implies.
The definition by Geissdoerfer et al. (2017) outlined in the introduction emphasises that a circular economy implies reducing the input of resources into the economy, maintaining the value of those resources for as long as possible, and reducing the leakage of resources in the form of waste generation, pollution and so on (cf.Kirchherr et al., 2017).It is common to distinguish between two main streams in the circular economy so-called biological nutrients, i.e. resources that are designed to re-enter the biosphere safely, and so-called technical nutrients, i.e. resources that are designed to circulate at high quality without re-entering the biosphere (Murray et al., 2017).In waste management, food waste is an example of the former, while plastics are an example of the latter.Importantly, the circular economy is not only about recycling.The most impactful innovations for circularity rather involve waste avoidance, through novel business models that allow for reducing inputs altogether, or for reusing materials, components and products through refurbishing, repair, rental, and so on (cf.Lüdeke-Freund et al., 2019).As pointed out by Frishammar and Parida (2019), transitioning to circular business models therefore require systemic shifts in business mindsets and approaches.
At least two practical challenges are associated with isolating and separating biological and technical resource streams and circulating them over time.First, it requires reducing the throughput of these resources through the economic system.Second, it requires maintaining the value and utility of those resource streams for as long as possible (cf.Franklin-Johnson et al., 2016;Iacovidou et al., 2017).These objectives imply a need to identify what the resource is, to track it through the system and to use that information to allow for repeated reuse of the resource.The latter objective can be achieved in many ways.These include the creation of markets for excess resources (e.g. when one company purchases and uses the "waste" or excess resources of another, such as in industrial symbioses) and the incentivization of households or companies that generate waste to promote recycling behaviour that is one necessary condition for circularity.
While social and environmental accounting practices become increasingly widespread and sophisticated (e.g.Parker, 2011), they only solve part of the problem of moving towards a circular economy.As pointed out by Nadeem et al. (2018), such reporting is a step in the right direction in its development of metrics related to e.g.greenhouse gas emissions, carbon dioxide equivalents and energy usage.At the same time, argues Nadeem et al. (2018), the mere reporting of such outputs (i.e.waste, pollution, and so on) still allows for using the same business models while aiming for incremental ecoefficiency or other improvements (see also Gray, 2006).Nadeem et al. (2018) therefore argues that more research on the needs of accounting tools and practices for the circular economy is needed.Some research efforts have been made to address accounting for the circular economy.Franklin-Johnson et al. (2016) propose a new performance metric indicator for the circular economy and applies it to the case of metals in mobile phones.They argue that metrics should capture contribution to material retention based on the amount of time a resource is kept in use, but that much more research is needed to develop such metrics for measuring the impact of business decisions on the longevity of precious materials.Several prior studies develop and investigate metrics for national levels of material consumption and waste generation (e.g.Pratt et al., 2016;Tisserant et al., 2017).However, these studies provide a macro depiction of resource flows in national economies, but do not provide fine-grained data for instance at the level of cities, companies or households, which can be used to lead resources from one actor to another along or across supply chains (cf.Ciulli et al., 2020).
Finally, existing accounting practices for the circular economy primarily provide accounts of historical footprints.As pointed out by Geng et al. (2016), however, the circular accounting toolbox also needs instruments that encourage reuse and recycling, while discouraging the use of nonrenewable resources.That is, there is a need for insight into how such accounting can be used to stimulate behaviours that promote circularity.Such applications of these accounting numbers by governments, companies and other stakeholders to drive behavioural change are relevant measures of their actual impacts (Geng et al., 2012).As pointed out by Tseng et al. (2018), using data-driven analyses to promote sustainable consumption and production on the part of households and companies, respectively, holds considerable promise (see also Ringenson et al., 2018).Experiments can also be used to investigate the effect of efforts to increase resource usage and efficiency (Chompu-Inwai et al., 2015).
In sum, there are still unanswered questions related to how indicators of circularity and our progress towards circularity can be measured (whether globally, nationally or locally).There is also still scarce insight into the potential use of accounting information to inform and design behavioural interventions that incentivize behaviour that promotes circularity on the part of the customers of waste management companies.Acquiring such insight is important for the circular transition, and our two research aims outlined above aim to contribute to this gap in the literature.

Research design
Our research design is a triangulated multi-method investigation of the development of the digitalised waste management system in Western Norway, which in turn allowed for detailed accounting practices for various resource streams.These accounting practices included both the identification and measurement of relevant resource streams in the waste management system, as well as the use of this accounting information to design and implement behavioural interventions to promote household recycling behaviour.We combine qualitative analyses of the development of the resource accounting system with quantitative analyses of the development of waste streams throughout the period and the effect of behavioural interventions to influence recycling behaviour.Thus, our empirical investigation sheds light on both overarching research aims based on interviews, documents, historical waste data and two natural quasi-experiments.

Background
We collaborated with the municipality-owned waste management company BIR,3 and its partly-owned subsidiarythe waste-tech start-up Carrot.4These companies manage the digital and physical infrastructure and collection of waste in the entire region and could thus provide us with comprehensive data.Our study comprises rich, complementary data sources on the design, implementation and effects of the resource accounting tools and practices that the companies implemented.
BIR is one of Norway's largest waste disposal companies and handles waste management for more than 356 000 inhabitants in seven municipalities on the Norwegian west coast, of which the largest is Bergenthe second largest city in Norway.In a separate business unit, it also offers waste management for business, and the resale of recyclable materials is among the company's revenue streams.The current municipality-owned company was founded in 1881 and incorporated in 2002.
BIR collects resources from households in three main ways: (1) weekly collections of residual waste and monthly collections of plastic and paper and cardboard from households; (2) a distributed infrastructure of collection points in parking lots and other public spaces, where households can dispose of recyclable resources that are not logged to the single household in the way the resources from the households are; (3) eleven recycling stations, three in Bergen and eight in the neighbouring municipalities, where households can bring all resource streams, both those that are collected at the households and special resource streams such as toxic waste, e-waste and so on.
Carrot is a waste-tech start-up owned in part by BIR and in part by venture capital investors.It has developed a cloud-based digital platform that captures, stores and makes available data about resource streams in the waste management system.As such, Carrot has been formed as the culmination of BIR's two-decade long work on digitising its resource accounting system.This system for data capture and sharing allows for the tracking and distribution of waste, by means of a comprehensive digital and physical infrastructure that includes sensors, underground container solutions, truck-mounted weights and recycling stations.In doing so, the company collects real-time data of all waste transactions and logistical events that are immediately made available in the truck, and in turn stored in a rich database and distributed seamlessly to all the decision-making systems in BIR.

Empirical setting
The broader context of our study was the digitalisation of the waste management system in Western Norway, which was carried out by BIR and Carrot.The main development and implementation took place in the period from 2008 to 2018, but is still very much ongoing.Figure 1 shows a timeline of important stages in the development of the resource accounting system.Our study benefits from how this digitalisation enabled the waste management company to develop a resource accounting system that could measure and collect fine-grained data about various resource streams in real-time, all the way upstream to the single household.The digitalisation implied both investments in physical infrastructure, such as RFID sensors attached to all household waste containers, antennas attached to all waste trucks, underground installments for a waste-suction collection system, and digital infrastructure, such as the cloud-based database developed by Carrot.
In extension of this digitalisation, our study also investigates the introduction of a socalled Pay-As-You-Throw (PAYT) system for households in BIR's market.PAYT systems are still uncommon in most markets (Brown & Johnstone, 2014), although they exist in some European countries (see e.g.Dahlén & Lagerkvist, 2010;Morlok et al., 2017;Reichenbach, 2008) and some states in the US (Folz & Giles, 2002;Van Houtven & Morris, 1999).The implementation of BIR's PAYT system allowed for our experimental investigation.Specifically, the resource accounting information was used to design and implement behavioural interventions that could promote recycling behaviour (cf.Linder et al., 2018;Steg & Vlek, 2009).
The implementation was carried out in two "waves", preceded by a pilot.The pilot was conducted in the municipality of Osterøy gave early insights into the potential effectiveness of PAYT.The system was then implemented in the six other neighbouring municipalities in 2009, due to the relative ease with which existing infrastructure in those municipalities could be adapted to PAYT.In Bergen municipality, an urban area with more than 250 000 inhabitants, implementation required more development and adaptation.Furthermore, there had been some pushback from local political institutions.Consequently, the PAYT system was implemented in Bergen municipality in 2016.
Prior to the PAYT system, all households paid an annual, undifferentiated municipal fee that covered waste management.The PAYT system worked as follows.Households still paid a fixed fee for waste management, but it was substantially lower than before.This amount covered a minimal amount of residual waste.For amounts above the minimum, however, a variable fee was added based on the number of waste collections from the household.The minimum baseline fee was 1933 NOK (approx.222 USD) per month, which included the minimum number of times the waste management trucks picked up residual waste from the household.Consumers were billed for additional pick-ups.The average household has 14 additional pick-ups, which gives a fee of 2398 NOK (approx.276 USD).That is, the average household was billed a 24% higher waste management fee than the minimum.The maximum fee, which includes 40 pickups more than the minimum, is 3263 NOK (approx.375 USD).That is, the maximum fee is 36% higher than the average fee and 69% higher than the minimum fee.
The stated objectives for introducing the PAYT system from BIR's point of view were threefold: . A fairer allocation of fees on households based on actual waste generation and recycling behaviour .Increased sorting and recycling of resources .Lower full cost and increased value creation due to strategic control of resource streams BIR had calculated the waste fee for the PAYT system in 2007, based on data collected in the years in which it tested the new physical and digital infrastructure of the resource accounting system.Figure 2 visualises the change in the waste fee in the years before and after the implementation of PAYT in Bergen, compared to the fee in Oslo and Trondheim, two comparable cities.The lower, fixed fee in Bergen is shown in the solid grey line, while the maximum amount households could pay if they used the maximum amount of collections.Thus, the maximum fee for a household was 3263 NOK (approx.375 USD), which was still lower than the full, fixed fee in the two other cities.

Data and analyses
Our empirical investigation is based on three data sources.First, we conducted numerous qualitative interviews with representatives from the two companies and supplemented these interviews with documents provided to us by the company.From late 2017 to late 2021, we conducted several semi-structured and unstructured interviews with top management (see Table 1).We held several informal meetings with the heads of various business units in BIR: household waste, business waste, IT services, communications and logistics.The qualitative data provided rich information about the digitalisation of the waste management system and the physical and technological infrastructure  that allowed for data capture.After conducting the quantitative analyses, we conducted follow-up interviews with BIR managers to probe and dig deeper into surprising patterns in the findings from the quantitative analyses.
Our second data source was detailed accounting information for each of the resource streams at the level of each municipality.The data ranges from 2005 to 2019; however, data prior to the start of the digitalisation pilot in 2008 are less fine-grained and somewhat less precise and reliable, due to the analogue data collection methods of the period.This rich data material allows for analyses of the development in various resource streams throughout the period, which is central to our first research aim: to investigate the role of the resource accounting system in enabling the reporting of resource streams.In addition, we got hold of historical data of residual waste generation in Bergen municipality all the way back to the initiation of municipal waste collection in 1881.
Our third and related data source was related to the PAYT pilot in 2008 and the implementation of the PAYT system in 2009 and 2016, respectively.We conceive of the implementation of the "pay-as-you-throw" system as two natural quasi-experiments (in two experimental waves).This allows for considering the two waves of implementation as a crossover design, in which Bergen municipality serves as a control group for the implementation (treatment) of the neighbouring municipalities in 2009, while the neighbouring municipalities serve as a control group during the implementation of PAYT in Bergen in 2016.Through analyses of the differences between the trajectories in the two groups, we reveal the effect of using information from the resource accounting system to incentivize recycling behaviour.Importantly, such quasi-experiments do not have random assignment of municipalities to treatments.Therefore, their findings must be interpreted prudently.However, the crossover design does allow for comparing the effects of the interventions across the two groups, albeit in two different time periods.

Results
In this section, we present our results related to our two research aims: to investigate (1) the role of digitalised resource accounting systems in enabling the identification and reporting of resource streams, and (2) the effect of using such information to incentivize household recycling behaviour to promote more circularity in resource streams.We address our first research aim in two parts.First, we account for the development of new resource accounting system was designed.Second, we provide insight into the new kinds of data this digitised resource accounting system could generate, and in doing so, we account for trends and developments in the waste generation in the period.Next, we turn to our second research aim, for which we investigate how this data made it possible to use this accounting information to promote recycling behaviour.We thus reveal the effects of this behavioural intervention both for the pilot in 2008 and the two experimental waves in 2009 and 2016.

Historical overview
The digitalisation of the waste management system and the development of the resource accounting system took place in response to the increasing amounts of residual waste in the late twentieth century.As pointed out by a manager in BIR, this period reflects the development of "Norway going from being a relatively poor country to its present state as a modern and wealthy economy that is boosted by an oil economy."Figure 3 shows waste generation in Bergen since 1881.The graphs visualise residual waste (in blue colour) and paper and plastic (in orange colour), respectively, in kilograms per person annually.
In 1881, residual waste amounted to approximately 10 kg per person.For a long time, it was stable, but a manager in BIR argued that "[a]fter 1960, waste generation was influenced by new waste streams like packaging of processed food as well as the spread of plastic."The amounts of waste exploded after 1970roughly coinciding with the oil boom in the Norwegian economy.This increase continued in the two subsequent decades and peaked at more than 250 kg per person in the late 1980s.According to a manager in BIR, "[t]his was also the period when the recycling of glass, and later, paper and cardboard, began in Bergen." In a similar way as the oil boom led to an increase in waste generation, financial crises tend to reduce waste generation (cf.Namlis & Komilis, 2019).This can be seen in the figure in sudden decreases with subsequent increases, for instance during the financial crisis of 2008 (which roughly coincides with the first wave of PAYT introduction) and the Asian financial crisis of the late 1990s.After 1990, BIR started doing separate collection of plastic, which led to a stabilisation in the development of residual waste.However, as shown by the average of more than 200 kg per person of paper and plastic in the decades after 1990, total waste generation (including other resource streams) continued to increase.Total waste, including all waste streams, started approaching 500 kg per person in the 2000s, but as pointed out by a manager in BIR, "[t]his is still quite low if you compare for instance to the US, where it was more than 900 kg per person around 2015." Importantly, the increase in resource streams sorted for recycling implied that residual waste generation had stabilised.

Towards a new resource accounting system for circularity
The development of a modern, digital resource accounting system was developed against this backdrop.This process started in the late 1990s with a simple logistical puzzlemanagers in BIR wanted a better overview of their physical infrastructure.They knew how many waste containers they had bought, but they did not know how many were still in operation, where they were, and so on.This led to the investment in RFID sensors to be attached to all waste and recycling containers.This digital infrastructure became a backbone in what would become a fully digitised waste management system.As pointed out by a manager in BIR: "There can be no circular economy without digitalisationin order to enable traceability, documentation of the quality of resource streams as well as their origin." The digital technology was combined with physical weight scales on the waste collection trucks, which allowed for mapping the actual content in each container directly to each household.A manager in BIR made the observation that the need for digital and physical infrastructures to increase control, traceability andultimatelymore circularity in resource streams was widespread in the industry."The problem, however", he pointed out, "was that all waste management companies are trying to solve the same problem, but all of them in their own way.But standardization had to be key!"Additional ambitions were to simplify time-consuming tasks and to ensure that the transformation of data into information was immediate."This was central to making information timeliera key goal of the digitalization", argued a manager in Carrot.He further pointed out: It was important for BIR that any digital solution needed to work across the different solutions in their waste management system.This involves different infrastructures like to vacuum suction system in the city, the various types of containers used at different locations, as well as regular waste truck pick-up services.
Furthermore, a manager in BIR argued: We had goals related both to performance management and to enterprise reporting.It is central that reporting is in "real-time", but also that subsequent data mining and deeper analysis is possible.This is crucial for using data to further develop the services of the company.
When the digital infrastructure was taking shape, the idea for a PAYT system emerged.At the turn of the millennium, an internal task force had started drafting what such a dynamic fee system would look like.Thus, the question of which information about household waste generation and recycling behaviour would be needed, quickly became prominent."We were sure that incentives mattered, and that using them in a dynamic fee system with feedback to users, could lead households to change their recycling behaviors", argued a manager in BIR.To test whether such a system was feasible, top management in BIR decided that RFID sensors should be installed on all waste containers in one of the smaller municipalities, Fusa (approx.population 3800) in 2002-2003.At the same time, a data system that could capture and systematically treat this data was put in place.In doing so, BIR also laid the groundwork for a PAYT pilot."We had to experiment at a smaller scale to test the technology and the backend solutions", argued a manager in BIR.The installation and testing of the physical and digital infrastructure on this small scale revealed to top management that designing and implementing such a resource accounting system was feasible.In 2004, the decision was therefore made to roll out the resource accounting system in all the nine municipalities and to move towards a PAYT system throughout its household waste market.
The installation of the RFID sensors throughout the municipalities was implemented from 2004 and 2007, which reflects the comprehensiveness of the digitalisation of a waste management system and enabling fine-grained data capture and treatment of the resource streams therein.In total, more than 190,000 sensors have been installed in BIR's system since 2004, allowing for a rich flow of resource data flowing into BIR's database in real-time.As one manager in BIR pointed out, "[w]e had often had to rely on data from months back in order to make business decisions that really required up-to-date data."The next step of the implementation was a more comprehensive test of the PAYT system.As accounted for above, the pilot in Osterøy was the first, full-fledged test of the combination of the resource accounting system and a PAYT model based on this information (see section 4.4.1 below for results from the pilot).After it proved successful, the implementation in all municipalities except Bergen followed suit in 2009 (see section 4.4.2 for results).
While BIR prepared for implementing the system in Bergen, several investments were made in infrastructure.An important innovation was an underground suction-based waste management system in central Bergen, which made traditional waste containers redundant.Instead, residents in the urban areas used individual RFID tags to unlock delivery points on the street, in which they inserted bags of waste and recycling.The system scanned and identified the type and amount of waste, which in turn could be used for billing purposes when the PAYT system was eventually introduced (see section 4.4.2 for results).Thus, another innovation in the resource accounting system was developeda technology that allowed for the same detailed data capture for this kind of waste and recycling collection as what had been done with traditional containers.This innovation has since become internationally recognised and award-winning innovation by the International Federation of Municipal Engineering in Kansas City, USA.
The concept of Carrot took shape in this period in the early to mid-2010s, as BIR conceived of the concept BossID (translates as WasteID).The idea was a detailed and fine-grained mapping of data on all household and business waste generation and recycling behaviour, which could be used for billing purposes and to introduce incentives and other behavioural interventions that could promote waste avoidance and recycling behaviour.The new company launched its cloud-based digital platform for collecting, treating, sharing and analysing resource data in real-time.As a manager in Carrot pointed out, "[o]ur solutions are based on transparency, standardization, automatization and sharing of data."This system was in turn integrated into BIR's digital infrastructure in the late 2010s.BIR had arrived at a resource accounting system that could help it drive circular solutions in its waste management practices.

Resource accounting in the BIR area from 2005 to 2019
The data BIR collected through its resource accounting system in the period since 2005 provides insight into the development of all resource streams since 2005.This has enabled increasingly fine-grained measurement of those resource streams and the ability to trace them back to separate households.The separate collection of these resource streams and the increasing quality of the data thereof has also allowed BIR to use this information for business purposes.This not only relates to the use of that information for incentivizing household recycling behaviour, as detailed in the natural quasi-experiment in the subsequent section, but also possibilities for reselling resources to other companies that can recycle or otherwise use these resources again.Thus, the resource accounting system increasingly enables BIR to promote circularity."Going forward, more waste streams will be considered valuable resources, and secondary markets will emergefor which our system also can provide an important digital infrastructure", argued a manager in Carrot.
Figure 4 shows the amounts collected from households in the period, measured in kg per capita.We distinguish between resources collected in Bergen and in the combined neighbouring municipalities respectively, for the following resource categories: (1) residual waste, (2) paper and cardboard, (3) plastic, and (4) glass and metals.Note that the accounting numbers for glass and metals only go back to 2007, while the numbers for plastic only go back to 2009.
Simply inspecting the graphs reveal trends that were indicative already in the historic visualisation in Figure 2. Let us consider each of them in turn.Residual waste generation is on a downward trend, which in the early 2000s was likely in part due to a general emphasis on the need for recycling in Norwegian society and in part due to the improvement of the recycling infrastructure in the period."Waste management data really provides insight into societal and economic changes in a country", argued a manager in BIR.As noted in the discussion of the historic data in section 4.1.,waste generation correlates with changes in GDP, which means that financial crises tend to lead to reductions in waste generation.This is evident in the residual waste graphs, as the financial crisis of 2008 leads to a reduction in residual waste generation prior to the first wave of PAYT in 2009.This exacerbates the effect of PAYT, which explains the very comprehensive drop in residual waste generation in this period.Importantly, however, the drop in residual waste is larger in the surrounding municipalities where PAYT was introduced, compared to Bergen, where PAYT had not yet been introduced.
Paper and cardboard are also on a slight downward trend in the period.According to several managers in BIR, the increasing digitalisation of society has led to a reduction in paper and cardboard consumption.This is particularly true for paper, while cardboard is in high demand due to increasing e-commerce."We see a slight increase in brown cardboard", a manager in BIR pointed out, "but generally paper and cardboard are on the decrease."Recent historical data on resource streams in waste management similarly show reductions in paper and cardboard recycling.For instance, the United States Environmental Protection Agency reported a 21% reduction in paper-based waste from 2000 to 2012 (EPA, 2012).The converse is true for the two other resource streams shown in Figure 4. Recycling of plastic and glass and metals is moderately increasing through the period.For these resource streams, and glass and metals in particular, we should note that the amounts per household per month are quite low, which means that there is substantial fluctuation in amounts from month to month.The trends for the different resource streams are qualitatively similar in the different municipalities, but as we will inspect in the subsequent section, there are important differences in the trajectories between municipalities that relate to the PAYT implementation.

Using accounting information for behavioural interventions in the PAYT system
In the following sections, we account for the results of the PAYT implementation in 2008, 2009 and 2016.Our results give insight into waste generation and recycling behaviour in the BIR area in the period.The nine municipalities are Bergen, Askøy, Fusa, Kvam, Os, Osterøy, Samnanger, Sund and Vaksdal.The eight non-urban municipalities neighbouring the city of Bergen are from small to moderate size (by Norwegian standards; total population 5.4 million).The smallest is Fusa (approx.population of 4000) and the largest is Askøy (approx.population of 29,000).Table 2 gives an overview of the population of Bergen municipality and the combined population of the neighbouring municipalities at several key time points of our study period (see Table A, Panel A in Appendix for an overview of populations in each of the municipalities).
In Table A, Panel B-D in Appendix, we provide an overview of relevant socio-demographic variables (age, gender, income) at the municipal level for all years in the period (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018).The table shows that income levels per capita are somewhat higher in the neighbouring municipalities than in Bergen, and that this is particularly the case for the three suburban municipalities closest to Bergen -Askøy, Sund and Os.The average age is slightly lower in Bergen than in the neighbouring municipalities, and again, the more rural municipalities have higher average ages.With regard to gender, the percentage of female inhabitants is slightly higher in Bergen than in all the neighbouring municipalities.
For all analyses, we report resource amounts per capita per month.Using monthly rather than annual data is necessary to get a larger number of observations to allow for statistical tests.However, reported numbers are therefore small, especially in recycling resource streams.This has two reasons: (1) we report monthly resource amount numbers and (2) we report per capita rather than per household numbers.As noted in footnote 2 above, the average household had 2.07 inhabitants, which implies that reported per capita numbers are approximately half of what the resource amount would be in an average household (e.g. if we report 2 kgs per capita per month, this would translate to 4 kgs per household per month and to 48 kgs per household per year).

The Osterøy pilot in 2008
The pilot was conducted in Osterøy in 2008 to test the feasibility of the PAYT system on a smaller scale and only for residual waste.The other neighbouring municipalities were not yet subject to the system.We selected Kvam municipality as a comparison, as it resembles Osterøy in population (approximately 7000) and other characteristics.This allows us to compare the development in residual waste generation prior to and after the pilot between the treatment group (Osterøy) and the control group (Kvam).In order to minimise the impact of natural fluctuations from year to year, we use a moving average of the two years prior to the pilot (2006)(2007) and compare it to the year of the pilot (2008) to compare the effects.The results are, however, qualitatively similar when we use only the year prior to and the year of the pilot (Figure 5).
While the average waste generation per month in both municipalities decreased in the period (ΔMean Osterøy = −3.36kg/capita/month; ΔMean Kvam = −0.47kg/capita/month), an independent samples t-test showed that the reduction was significantly larger in Osterøy than in Kvam (t (22) = −2,34, p = 0.014; one-tailed5 ).In relative terms, residual waste was reduced by 21.2 percentage points in Osterøy compared to a reduction of only 2.4 percentage points in Kvam.

The effect of the PAYT system: two waves of implementation
In 2009, PAYT was implemented in the eight neighbouring municipalities, but not in Bergen.Thus, in the following analyses, we compare the change in collected resource streams for the neighbouring municipalities that were subject to the PAYT system (treatment group) to Bergen municipality (control group).This allows us to compare the development in waste generation prior to and after the implementation in the two groups.We used a moving average of the three years prior to the implementation (2006)(2007)(2008) as a baseline and compared it to the three years after implementation (2009)(2010)(2011) to compare the effects.As in the pilot, the results are qualitatively similar when we used only the year prior to and the year of the implementation.As noted above, the implementation of PAYT in the neighbouring municipalities took place during the financial crisis that had started in 2008, which needs to be taken into consideration when inspecting the results.
Figure 6 shows the comparison between the treatment group (the neighbouring municipalities; 2009 population = 72,738) and the control group (Bergen municipality; 2009 population = 252,051) for residual waste.Analyses revealed that the reduction in monthly waste generation was larger in the municipalities where the PAYT system was introduced (ΔM neighboring = −3.12kg per capita/month) than in the control group (ΔM Bergen = −2.60 kg per capita/month).The difference is statistically significant only at the 10% level (t (70) = −1,314, p = 0.097; one-tailed).In relative terms, residual waste was reduced by 17.9 percentage points in the treatment group compared to a reduction of 15.2 percentage points in the control group.When inspecting the overall development in residual waste generation in Figure 3 above, we can see a quite substantial drop in residual waste generation in 2009 across all municipalities, which is in line with the comparable trajectories of the treatment group and the control group, respectively.
We now turn to the development in the other resource streams: paper and cardboard and glass and metals.Since resource accounting for plastic only began in 2009, we only report results for this resource stream for the second experimental wave below.As shown in Figure 7, household recycling of paper and cardboard fell in both groups (ΔM Bergen = −1.72 kg per capita per month and ΔM neighboring = −1.46kg per capita per month).A ttest shows that difference in reduction between the two groups was marginally significant (t (70) = 1.954, p = 0.055; two-tailed).Paper and cardboard was reduced by 27.2 percentage points in the treatment group compared to a reduction of 28.9 percentage points in the control group.
For glass and metals, household recycling increased for both groups (ΔM neighboring = 0.08 kg per capita per month; ΔM Bergen = 0.02 kg per capita per month).However, a t-test shows that the difference between the two groups was insignificant (t (70) = 0.889, p = 0.189; one-tailed).In relative terms, collection of glass and metals increased by 14.2 percentage points in the treatment group compared to 2.3 percentage points in the control group.While there is no baseline data for plastic prior to PAYT, a simple comparison of the amounts of recycled plastic after the implementation show that collected kgs per capita per month is more than twice as large in the treatment group as in the control group (M neighboring = 0.60 kg per capita per month and M Bergen = 0.29 kg per capita per month).
The second experimental wave took place in 2016.For these analyses, we compare the change in collected resource streams for Bergen municipality, which was now subject to the PAYT system (the treatment group; 2016 population = 277,391) to the neighbouring municipalities (the control group; 2016 population = 81,973).Importantly, these analyses differ from the first experimental wave in the sense that the neighbouring municipalities are not a true control group, as they were already subject to the PAYT system at this point.However, this still allows for difference-in-differences analyses since the relevant baseline here is that one would not expect a similar reduction in these municipalities as one would in Bergen, when the latter is subject to the PAYT system for the first time.
Again, we compare the development in residual waste generation and recycling prior to and after the implementation in the two groups.We again used a moving average of the three years prior to the implementation (2013-2015) as a baseline and compared it to the three years after implementation (2016)(2017)(2018).Again, the results are qualitatively similar when we used only the year prior to and the year of the implementation.Figure 8 shows the comparison between the treatment group (Bergen municipality) and the control group (the neighbouring municipalities) for all residual waste.Analyses revealed significantly larger reduction in waste generation (t(70) = −1.935,p = 0.029; one-tailed) in the treatment group, Bergen (ΔM Bergen = −2.24kg/capita/month), than in the control group, i.e. the neighbouring municipalities (ΔM neighboring = −1.37 kg/ capita/month).Residual waste was reduced by 13.9 percentage points in the treatment group compared to a reduction of 9.0 percentage points in the control group.We again turn to the development in the other resource streams, which now includes plastic in addition to paper and cardboard and glass and metals.As shown in Figure 9, household recycling of paper and cardboard fell in both groups (ΔM Bergen = −0.43kg per capita per month and ΔM neighboring = −0.66kg per capita per month).A t-test shows that the decrease was significantly larger in the control group (t(70) = 2.447, p = 0.017; twotailed).Paper and cardboard was reduced by 13.3 percentage points in the treatment group compared to a reduction of 20.0 percentage points in the control group.
For glass and metals, however, household recycling increased in both groups, similarly as in the first wave in 2009.The increase was significantly larger (t (70) = 2.256, p = 0.014; one-tailed) for Bergen (ΔM Bergen = 1.92 kg per capita) than for the control group (ΔM neighboring = 0.52 kg per capita).Glass and metals recycling increased by 16.6 percentage points in the treatment group compared to an increase of 8.7 percentage points in the control group.
For plastic, recycling fell in the control group (ΔM neighboring = −0.05kg per capita per month), while it increased in the treatment group (ΔM Bergen = 0.10 kg per capita).The difference between the two groups is statistically significant (t (70) = 9.564, p < 0.0001; one-tailed).Plastic recycling increased by 27.3 percentage points in the treatment group compared to a decrease of 6.7 percentage points in the control group.

Discussion
Can accounting for resources facilitate our transition to a circular economy, and can the information captured in such accounting systems provide levers for influencing the behaviours of households and companies in ways that promote circularity?By digitising resource streams and building a resource accounting system that can collect, analyse and use resource accounting information, perhaps the transition towards a circular economy can be catalysed?Our study shows that developing tools and practices for resource accounting can indeed enable the identification of degrees of circularity, as well as designing and taking measures to promote circularity through recycling behaviour.Our findings from the qualitative interviews shed light on the development of a resource accounting system, the data on waste generation and recycling it has generated historically and the interventions put in place to catalyse household recycling behaviour.Thus, our findings contribute to the understanding of accounting for the circular economy and go some way in filling the gaps identified in our review of the literature above.
Residual waste generation per capita has risen twentyfold since BIR began measuring it in 1881, but since 1985, the number has flattened, and partly decreased.This is due in part to the advent of a convenient recycling system both by means of collection from households and from infrastructure to which households can deliver recycling.However, it is also likely that the downward trend is a consequence of greater awareness around the problem of pollution and waste generation, which leads to a combination of waste avoidance (cf.digitalisation and reduction in paper consumption), reuse and other forms of resource upcycling, such as at-home composting, and so on (see e.g.DiGiacomo et al., 2018).
Our more fine-grained data from 2005 to 2019 reveals that residual waste generation is steadily decreasing, with a converse increase in household recycling of both plastic and glass and metals, respectively.The same trend does not hold for paper and cardboarda resource stream for which recycling actually decreased.As noted above, this deviation from the overall patterns is largely a consequence of digitalisation and other developments that have led to the reduction of paper consumption.The patterns of development through the period are relatively similar for urban Bergen and the more rural neighbouring municipalities.
When we take simultaneously into account the data from the resource accounting from 2005 and 2019 and the results from the PAYT intervention, an important insight emerges.The underlying trend of downward-sloping waste generation in the period, i.e. that households are apparently becoming better both at avoiding waste and at recycling, seems somewhat reinforced by behavioural interventions in the form of the PAYT system.Our results from the PAYT intervention show that in turn using the resource accounting information to incentivize households for further recycling behaviour can strengthen this transition.In both experimental waves, implementing PAYT is associated with an increase in recycling behaviours.This does not hold for all resource streams consistently in the different waves of the experiment, but results are indicative of such an effect.
Thus, measuring and accounting for historical and current resource streams is a first and necessary step for measuring and promoting circularity.However, when this information is used to inform the design and implementation of incentive structures like the PAYT system, it can alter the behaviours of the households whose waste generation and recycling behaviour the resource accounting system captures.
An important question related to our investigation is: how good are the measures in BIR's resource accounting system as indicators of circularity?As pointed out by Haupt et al. (2017), many current indicators are not suitable performance indicators for a circular economy, and collection rates neither give adequate information about available quantities of secondary resources produced nor information about their final destinations.This arguably also holds for the indicators reported in this study, in the sense that it mostly captures developments in various resource streams handled by BIR.Instead, circular performance indicators should measure reductions in input, increases in the longevity of throughput (cf.Franklin-Johnson et al., 2016) as well as the degree to which resource streams are redistributed into new uses in various supply chains where they have valuethus making these outputs into new input (cf.Mathews & Tan, 2016). 6hat said, the resource accounting system investigated in this paper goes some way in moving waste management practices in this direction.The measurement of the various resource streams and identification of behavioural levers that can be pushed to improve material recycling are important first steps in leveraging digital technologies for measuring and using resource data to drive change (cf.Ringenson et al., 2018).As pointed out by Geng et al. (2016, p. 223), the effectiveness and impact of circularity indicators hinges exactly on the degree to which they are integrated into methodologies for decision making and policy setting in companies and other institutions.Our study of resource accounting and the PAYT intervention responds to the call by Geng et al. (2016) for more insight into the diffusion, usefulness, and evolution of circularity indicators.Furthermore, as emphasised by Nadeem et al. (2018), such indicators are also central to understanding drivers of internal and external cost in resource accounting and the complexity of circular-economic processes.Maas et al. (2016) pointed out the need for more knowledge on the integration and interplay of accounting, management control and reporting approaches and specifically pointed out the following challenge: to empirically examine how companies actually collect, analyze, use and internally communicate sustainability information, what tools they apply to do so, how the processes between company-internal actors are organized and how different sustainability accounting, management control and internal communication and reporting methods play together.(Maas et al., 2016, p. 245) Our study provides rich insight into exactly this challenge.We demonstrate the comprehensiveness of conceiving of, designing and building the physical and digital infrastructure for such a resource accounting system.Our findings also demonstrate the potential for using resource accounting information to incentivize pro-environmental behaviours in households and beyond.
While our resource accounting data ends in 2019, BIR continues its efforts to expand the system and make the resource streams and data thereof substantially more comprehensive.At the time of writing, BIR and Carrot are conducting pilots for several other resource streams, the first of which is food and other organic waste.Carrot has developed tools for capturing data on organic waste streams and the two companies are together working on developing supply chains for the resale of organic waste for upcycling purposes.As the system accounts for more and more resource streams and the companies simultaneously identify buyers for such excess resources prior to them becoming waste, the potential for using the resource accounting system to drive real circularity becomes larger.This is also relevant for the call by Franklin-Johnson et al. (2016) on connecting insights into the longevity of resources and the economic return of doing so.As pointed out by Tseng et al. (2018), data-driven analyses that can produce reliable information for use in industrial networks can enable more efficient resource usage.Thus, the parts of the resource accounting system investigated in this paper are only the beginning and as such a template of the accounting practices yet to be realised.
In the interviews with managers from BIR and Carrot, their views on the current and future developments of the resource accounting system also became evident.One BIR manager argued: The system has now allowed for capturing and structuring data in a way that makes data standardized.This is key to allow for more sophisticated solutions -AI-driven solutions, a more comprehensive set of sensors and digital technologies that allow for more comprehensive and customized services to customers.This will also make it possible to provide households and other users with mobile access to data that can inform their behavior, while on the part of the waste management company, processes like planning, maintenance and business intelligence will become faster and easier.
As these developments occur, the opportunities for using the resource accounting system to shape the behaviours of both internal and external stakeholders will become even greater.
Our findings also have implications for practice.First, our natural experiments show that PAYT systems are potentially effective in driving recycling behaviour.As shown in the simple comparison with other larger Norwegian cities, using a PAYT system allows for pricing that can make waste management fees lower for households while being costeffective for the waste management company.Second, our study sheds light on the importance of experimentation in the process of developing a resource accounting system for the circular economy.The considerable scope of this challenge, and the relative scarcity of consensus on what are the best indicators, methods of measurement and analysis, as well as implications for decision making and policy, trial-and-error will have to play an important role.The systematic and controlled use of collected resource information to design and execute pilots (cf. the PAYT pilot in 2008) holds great promise.Experimenting with companies elsewhere in the supply chain that can make use of the resources captured in the resource accounting system is also likely to be necessary, as these companies are simultaneously experimenting on becoming more circular.

Limitations
There are several limitations to the data upon which our empirical study builds.First, during the period the digital resource accounting system was developed and implemented, the scope and quality of collected data steadily increased.As shown in the studies above, some resource fractions (e.g.plastic) are not available for the early years of PAYT, while others (e.g.food waste) are only in the process of being integrated now.This limits the scope and completeness of our analyses.Second, during the time of our empirical investigation, the resource accounting data sets were not yet integrated with data on household socio-demographic characteristics such as type of household, gender, and so on.We refer to the account of socio-demographic variables in section 4.4 above (see also Appendix).This rendered us unable to include control variables for investigating moderating and heterogeneous effects of the PAYT intervention, which should be explored in future studies.
A final limitation is related to the fact that our investigations are all conducted within a single geographic region.The degree to which our findings can be generalised thus need to be considered prudently.All experimental investigation in the field necessarily takes place in a specific setting, with the strength of proximity to actual behaviour and the limitation of idiosyncratic characteristics of the setting (cf.Al-Ubaydli & List, 2015).Furthermore, according to List (2020), the external validity claims of a behavioural study of this nature is related to whether or not there are differences in the preferences, beliefs and constraints of actors in the sample compared with those in the target population.The Norwegian recycling system is comparable to many other recycling systems in modern, industrialised economies.Research shows that the inclination to recycle increases with higher income and educational levels (Milford et al., 2015), which are both high in Norway.In addition, the financial incentives in the PAYT system places constraints on households that may be different from some countries.Thus, the findings are likely to be more generalisable to other countries characterised by high income and educational level.Seeing how the study investigates the effects of introducing financial incentives through the PAYT system, the findings are arguably generalisable also to settings without such incentives currently.We have made efforts to shed light on these characteristics of the setting in a manner that allows for critical inquiry into conditions that may limit the generalizability of our findings.

Conclusion
If we look back at the present time 30 years from now and conclude that we managed to achieve a circular economy, resource accounting will most certainly have played a key role in making it happen.Resources never become waste if we can account for them, retain the data about them and make decisions about how to prolong the life of those resources based on that information.This was the mindset captured in the quote by Anders Waage Nilsen from Carrot in the introduction to the paper.As revealed through our qualitative investigations, it was also the underlying philosophy of BIR's design and implementation of a comprehensive, digitally supported and data-driven resource accounting system for various resource streams in its waste management system.During two decades of experimentation and development, BIR has laid the groundwork for a resource accounting system that can facilitate circularity.
Importantly, waste management alone will never be sufficient to create a circular economy, as circularity is much more than recycling.A truly circular approach to waste reduction starts with behaviours that avoid waste altogether, while our study addresses the part of waste reduction that can be influenced by the waste management companies, i.e. promoting recycling behaviour among their constituents.The plans of BIR are, however, broadening the scope of the role waste management companies can play.This involves not only including many other recyclable resource streams into waste management, but also the expansion of waste management upstream into the business markets they serve.This can enable the creation of resource accounting and resource streams that avoid turning the resources into waste in the first place, by enabling the transportation of various resource streams from where they are in excess to where they are needed.Examples include transporting tons of plastic waste in the basement of a shopping mall to a plastic recycling factory, or transporting food waste in the kitchen of a large cafeteria to a farm that upcycles the food waste by feeding it to larvae (see Liverød, 2019).Hence, waste management companies' abilities to drive the diffusion of circular practices in households and beyond can increase substantially.Here, accountants can clearly play an important role in developing and using these resource accounting systems and the data they generate to create actionable insights to promote circularity.
For this reason, this study only shows the embryonic contours of what a future resource accounting system for the circular economy might look like.Much more research is needed to investigate the needed characteristics of such systems, what are the best indicators for circularity, how to connect resource accounting for households and companies, and so on.Our study thus lays the groundwork for further empirical investigation of such questions.
There are several exciting avenues for future research.First, regarding the promotion of circularity, indicators are needed for measuring the effects of various efforts for waste avoidance, in addition to measures of material sorting and recycling once resources have entered the waste management system.The scope of such waste avoidance behaviours is very broad and is linked to transformations of business models, through which individuals and households consume products as services, consume products through refills, and so on.To fully capture the transition to the circular economy, such improvements in circularity also need to be measured.Second, there is a need for further insight into the economic performance dimensions of increased circularity.As discussed, there are increasing market opportunities for prolonging the lifespan of products, components and materials.The economic upside of such measures by waste management companies and others is a highly important topic for empirical investigation.
Finally, there is plenty of scope for further investigation of behavioural interventions to stimulate behaviours that promote circularity based on resource accounting information.A promising field of research is to investigate behavioural interventions at the level of business waste generation and recycling.Both due to the vast amounts of resource streams in this area, the relative lack of measures in place, and the potential scalability of such solutions, this area holds large promise.Examples include using conventional financial incentives, the use of gamification technologies and social interventions such as using peer information to drive behaviour.

Funding
We gratefully acknowledge the financial support from the Norwegian Research Council [grant number 299378] and the Norwegian Retailers ' Environment Fund [grant number 1124294].

Figure 1 .
Figure 1.A timeline of the development of resource accounting in Bergen and neighbouring municipalities.

Figure 2 .
Figure 2. Comparison of waste collection fees in Norway's three largest cities, including the maximum amount (variable + fixed) in Bergen (amounts in NOK).Note: The unit of measurement is kg/person.Blue colour represents residual waste, while orange colour represents paper and plastic (combined).

Figure 3 .
Figure 3. Waste generation in Bergen municipality 1881-2015 (prior to the PAYT implementation); based on data compiled and generated by the Head of R&D in BIR.Note: The unit of measurement is kg/person.Blue color represents residual waste, while orange color represents paper and plastic (combined).

Figure 4 .
Figure 4. Collected waste and recycling in Bergen and neighbouring municipalities from 2005 to 2019 (in kg per capita).

Figure 5 .
Figure 5. Effects of the experimental pilot in Osterøy (treatment group) on residual waste per capita, compared to Kvam (control group).

Figure 6 .
Figure 6.Effects of the PAYT implementation on residual waste per capita/month first experimental wave comparing the neighbouring municipalities (treatment group) and Bergen municipality (control group).

Figure 7 .
Figure 7. Effects of the PAYT implementation on recycling streams (kg per capita/month), first experimental wave comparing the neighbouring municipalities (treatment group) and Bergen municipality (control group).

Figure 8 .
Figure 8. Effects of the PAYT implementation on residual waste per capita, second experimental wave comparing Bergen municipality (treatment group) and the neighbouring municipalities (control group).

Figure 9 .
Figure 9. Effects of the PAYT implementation on recycling streams (kg per capita), second experimental wave comparing Bergen municipality (treatment group) and the neighbouring municipalities (control group).

Table 2 .
Population included in the study at various time points in the period.