Applied Environmental Economics: Bridging the Divide Between Policy

Applied Environmental Economics: Bridging the divide between policy and theory within the context of recycling, macro-level environmental indicators, and water trading

Thomas Andrew Longden

May 2014

Supervisor: Professor Kevin J. Fox

A thesis submitted for the degree of

Doctor of Philosophy

School of Economics

The University of New South Wales

Acknowledgements

First and foremost, I would like to thank my supervisor, Professor Kevin Fox, for his advice, supervision and general assistance. Kevin has always been extremely approachable and has encouraged me to pursue my own thoughts and research ideas. Above all else, I thank him for encouraging me to simultaneously undertake teaching and study demands, making the PhD process an extremely valuable experience. In addition I must thank Professor Tony Owen, who was my thesis supervisor during my first year and coursework components. Many of the base research areas were established during this time and his encouragement was invaluable.

I wish to gratefully acknowledge the financial support received from the Australian Nuclear Science and Technology Organisation (ANSTO). The views expressed in this thesis do not necessarily reflect those of the agency. My thanks must also go to the School of Economics for additional financial assistance and for providing me ample opportunities to gain teaching experience. I am truly grateful to Nigel Stapledon, Regina Betz, Diane Enahoro, Trevor Stegman and Hazel Bateman for making my teaching a productive and enjoyable experience.

The Graduate Research School and the Postgraduate Research Support Scheme was invaluable in funding my attendance and presentation of two research papers at the 17th Annual Conference of the European Association of Environmental and Resource Economists (EAERE) held at the VU University in Amsterdam, The Netherlands 24 - 27 June 2009. I am also grateful to the Centre for Applied Economic Research (CAER) for organising the Environmental Challenges Workshop in 2008 and FEEM/IEFE for the invitation to present at their joint seminar series in July 2010, in which I had the opportunity to present parts of my thesis to a broad audience. Between these four presentations and the appreciated comments from three anonymous referees, the bulk of the research held within this thesis has been reviewed and subsequently refined.

My sincerely thanks to my PhD and academic colleagues who have made my studies at the University of New South Wales a wonderful experience. Many thanks go to Amy Cheung,

Lisa Lee, Philia Restiani, Zaida Contreras, Joaquin Vespignani, Daniel Bunting, Renuka Sane, Thai Vinh Hguyen, Rushmila Alam, Abu Shonchoy, Lorraine Ivancic, Xuan Nguyen, Iqbal Syed, Natalia Garabato, Ha Nguyen, Phuong Hoang, Adeline Tubb, Murali Radhakrisnan, Milica Kecmanovic, Jonathan Lim, Amani Elnasri, Carmit Schwartz, to name a few. I also thank all professional and technical staff at the School of Economics, Fei Wong, Grace Setiawan, Joanna Woo, Emma Proud, Dominique Motteux, Rebecca Crosby, Tahnee Fogarty, who have always been extremely friendly, helpful and capable.

My two best friends, Greg Kannard and Nick Learmonth, who have put up with me ever since high school need to be thanked greatly As does Kelly Sutherland, who was a very valuable partner throughout a great segment of my PhD studies. I would like to thank all of the extended Longden family, but need to especially thank my father Colin, my grandfather Thomas E, my grandmother Nellie and my great aunt Margret. In recent years, Ines Österle and the whole extended Österle family have become very important since my arrival in Milan.

This thesis would not have been possible without the most important people in my life, my mother Renata, my sister Bianca, and my brother Nicholas. My aunt and uncle, Ilona and Denzil, have provided continual support for which I am indebted. Lastly, I would like to dedicate this thesis to my grandmother, Marie Brezinova, as I am heartbroken that she will be unable to see me graduate.

Thomas Longden

May 2014

Abstract

By focusing upon a range of topics and utilising a range of techniques, this Thesis provides a review of a range of issues related to successful environmental economics policy prescriptions. How and why certain policies are selected will be crucial to making effective policy prescriptions and this is the central focus of this Thesis. Spanning from a review of a policy prescription that was not followed to the development of a model that is applied to the review of a policy proposal; examples of policy failure, policy success, and policy prescriptions are at the core of the discussion throughout. With an appraisal of policy-making on a local level, the case of recycling and the household charge for waste collection is reviewed using data from municipalities in New South Wales, Australia. This analysis shows that without an adequate pricing scheme, different levels of the household charge for waste collection across municipalities have little association with different levels of recycling. An appraisal of intergovernmental agreements (including the Montreal Protocol) follows and finds that evidence points to an induced policy response, rather than the Environmental Kuznets Curve relationship. While declines in emissions cannot be solely attributed to the timing of the targets prescribed, the Montreal Protocol was associated with notable emission reductions. The last area of research covers a range of issues surrounding water trading between agricultural firms. Chapter four introduces a distinctive and innovative agent based model built to provide projections of trading that allow for: - the constraint of transaction costs, - the mixture of firms within the trading scheme, - out of equilibrium market operation, and – a range of crop water demands. This model is then utilised to gauge the impact of the implementation of a ‘Network Trading Scheme’ which has the aim of reducing the impact of transaction costs on otherwise viable trades.

Table of Contents

Chapter 1 – Introduction

1.1 Importance of the research area ...... 10

1.2 Sequence of the thesis ...... 16

Chapter 2 – The Persistence of Flat Fee Waste Collection Charges

2.1 Introduction ...... 19

2.2 Discussion of Literature and a Case Study on Council Recycling Programs ...... 25

2.2.1 Initial Implementation of Council Recycling Programs ...... 25

2.2.2 Focus on Initial Implementation ...... 27

2.2.3 Flat Fee Recycling Programs ...... 29

2.2.4 Recycling at Reasonable Rates ...... 35

2.3 Empirical Analysis ...... 37

2.3.1 Data Analysis ...... 37

2.3.2 Full Dataset Regressions ...... 43

2.3.3 Relationship between Collection Costs & Number of Pickups ...... 48

2.3.4 Two Period Data ...... 50

2.4 Conclusion ...... 54

Chapter 3 – Going forward by looking backwards on the Environmental Kuznets Curve

3.1 Introduction ...... 58

3.2 Literature Review and Background Discussion ...... 64

3.2.1 Humble Beginnings ...... 64

3.2.2 Scarce Attention Given to CFCs ...... 67

3.2.3 Montreal and Kyoto ...... 70

3.3 Empirical Analysis ...... 74

3.3.1 Econometric Foundations ...... 74

3.3.2 Estimation Outline ...... 77

3.3.3 Estimation Results – CFC ...... 81

3.3.4 Estimation Results – CO2 ...... 86

3.3.5 Emission Reductions and Policy ...... 91

3.4 Conclusion ...... 96

Chapter 4 – Water Trading, Diversity and Transaction Costs

4.1 Introduction ...... 99

4.2 Literature Review ...... 105

4.2.1 Applied Water Trading Schemes ...... 105

4.2.2 Transaction Costs ...... 113

4.3 Establishing an Agent Based Water Trading Model ...... 117

4.3.1 Introduction to the Water Trading Model ...... 117

4.3.2 Search Mechanism ...... 126

4.3.3 Baseline Data ...... 128

4.3.4 Decision Variable ...... 133

4.3.5 Agent Based Modeling...... 137

4.3.6 A Summary of the Agent Based Model ...... 140

4.4 Analysis of the Model Results ...... 151

4.4.1 Graphical Analysis ...... 151

4.4.2 Regression Analysis ...... 162

4.5 Conclusion ...... 176

Chapter 4 – Water Trading, Diversity and Transaction Costs

5.1 Introduction ...... 180

5.2 Literature Review ...... 186

5.2.1 Market Design ...... 186

5.2.2 Non-uniform Good ...... 191

5.3 Incorporating Network Trade ...... 193

5.3.1 Network Trade ...... 193

5.3.2 Modelling the Permit and Tax Hybrid ...... 199

5.4 Review of Results ...... 202

5.5 Conclusion ...... 210

Chapter 6 – Conclusions and Plans/Recommendations for Further Research

6.1 Conclusions ...... 212

6.2 Plans and Recommendations for Further Research ...... 216

Chapter 1 – Introduction

1.1 Importance of the research area

Economics has extended away from traditional areas, focusing on extensions of theory that have been driven by the demands of applied policy. While traditional economic theory has and still does deal with closed-form solutions, perfectly rational economic actors and the analysis of equilibria outcomes; the needs of policy have highlighted the usefulness of reviews of alternate methodologies. These alternate methodologies include: - the assessment of auction processes, - the development of game theory, - the development of advanced simulation models, and - the use of human subjects in experimental economics. Examples of these alternate methodologies are noted in Roth (2002). Within ‘The Economist as Engineer:

Game Theory, Experimentation, and Computation as Tools for Design Economics’, Roth discusses the development of an emerging discipline of design economics that does not focus on perfectly rational economic actors and the analysis of equilibrium outcomes. The conviction shown by Roth in the paper is reflective of the divide between the demands of real-world applications and the existing body of research within the discipline of Economics.

As an example of the core issue, upon redesigning the labour market clearinghouse for

American physicians, Roth (2002) noted that the only available theory “that directly applied to the medical market were the counterexamples and they all warned that, in more complicated markets, problems could sometimes arise” (Roth, 2002: 1372).

In a paper that reviews the implementation of marketable permits and emissions charges,

Hahn raises the issue of the ability of Economists to select between a variety of economic instruments for policy prescriptions. In addition, the paper discusses the lack of support given to Economist’s policy prescriptions. Hahn (1989) opens with the statement that “one of the dangers with ivory tower theorizing is that it is easy to lose sight of the actual set of problems which need to be solved, and the range of potential solutions.” (Hahn, 1989: 95) Indeed, while models may be designed and found to have stable solutions, this may not be enough to convince external stakeholders to support or implement the implied policy designs. In relation to this Hahn states that “it is possible to design marketable permit systems which are more efficient and ensure better environmental quality over time, yet these systems have not been implemented.” (Hahn, 1989: 111) An example used by Hahn is the potential reaction of environmentalists who are assumed to show reluctance to support such policy measures due to a fear of such measures being a signal that could be seen as legitimatising pollution.

Indeed, successful policy prescriptions may need to account for stakeholder views and opinions on the theoretical or modelling work that they are based on. Other real-life contingencies may also need to fall in place for policy prescriptions to work. An assessment of successful policy prescriptions also needs to acknowledge that success may also occur for the right and/or wrong reasons.

It is with such examples in mind that the subsequent research has been undertaken, as the ultimate test of theory and policy prescriptions is how well they apply to real-life situations.

In addition, how and why certain policies are selected will be crucial to understanding of the effectiveness of applied environmental economics. By focusing upon a range of topics and utilising a range of techniques, this Thesis provides an interesting discussion of a range of issues related to successful policy prescriptions. The motivating reasons for the selection of certain policies will be crucial to evaluating and designing effective policy prescriptions and this is the central focus of this Thesis. The Thesis spans from a review of a policy prescription that was not followed to the development of a model that captures real-life situations and is then applied to the development of a policy proposal. Examples of policy failure, policy success, and policy prescriptions are at the core of the discussion throughout.

As an example, the first chapter evaluates some of the potential reasons why a certain policy prescription has not been put in place within New South Wales. With an appraisal of policy- making on a local level, Chapter two focuses on the issue of policy inadequacy due to a lack of unit based pricing for household waste collection and reviews institutional factors governing the continued use of flat fee pricing. Chapter three focuses on an appraisal of two intergovernmental agreements and includes an example that has been deemed a success (the

Montreal Protocol) and another where success is unlikely to be widely recognised (the Kyoto

Protocol). Upon comparing the real-world application of intergovernmental agreements the issue of how interests align are found to be important. This chapter also reviews the issue of whether finding an Environmental Kuznets Curve relationship is related to spurious regressions and an induced policy response.

The subsequent chapters then focus on the development of an agent based model to be used in the assessment of water trading that is constrained by different levels of transaction costs and potential market structures that can minimise the impact of this issue. The model utilises an innovative framework that attempts to conform to the discipline of design economics that was discussed in Roth (2002). In addition, an attempt has been made to capture real-life dimensions so as to follow in the spirit of Hahn (1989). For example, the agent based model does not include perfectly rational economic actors, nor does it focus on the analysis of equilibrium outcomes. The model includes: - transaction costs that depend upon distance between traders, - different search mechanisms for agents, and - other assumptions that capture factors that infer a market may not reach equilibrium. Indeed, insufficient trading due to an inability to find trading partners is an issue in water trading that this model aims to investigate. Chapter four focuses on the development of the model to be utilized to assess the success of water trading schemes and Chapter five utilises this model to evaluate the potential of a Network Trading Scheme to foster trade in such trading schemes.

Chapter six concludes the Thesis with an overview of the main results of each of the chapters.

As a prelude to the analysis, the rest of this chapter discusses the key features of each of the consequent chapters and in doing so previews some of the results.

Within the area of recycling, economists have prescribed the adoption of unit pricing for waste collection so as to achieve an efficient rate of household recycling. For an example refer to Callan and Thomas (1999). Upon reviewing data from municipalities in New South

Wales, Australia; Chapter two shows that different levels of the household charge for waste collection across municipalities has little association with different levels of recycling. The

13 typical collection contract that has prevailed involves the establishment of a flat fee collection scheme and the analysis finds that this is likely to occur due to the interaction of a cost minimising government agency with a profit maximising firm. With a cost minimisation motive, local governments tend to adopt basic recycling programs while satisfying household and constituent demand for the provision of an undefined recycling program. As a result, to achieve universal adoption of unit pricing collection schemes, economists may need to advocate the lobbying of state governments to prescribe unit pricing programs or provide information on the asymmetric information within an out-sourced collection contract. This chapter is important within the overall framework of the Thesis as it reviews a policy proposal that has not been undertaken and investigates why this has not occurred.

Chapter three appraises two intergovernmental agreements with the intention of establishing that notable and sustained emission reductions tend to be associated with a concerted policy effort. This is in contrast to the Environmental Kuznets Curve (EKC) which contends that a significant negative relationship exists between high levels of national income and per capita emissions. An additional intention of this chapter is to focus on whether such a relationship persists once a concerted policy effort has been allowed for. While the timing of the Montreal

Protocol coincides with a significant negative decline in CFC consumption between 1992 and

2008, this reduction cannot be solely attributed to the specific targets of the Montreal

Protocol. The decline has also been driven by factors related to the auxiliary explanations for the success of the Montreal Protocol. These auxiliary explanations are: - the existence of a supportive industry group, - pre-existing legislation and commitment within the United

States, - affordable and available substitutes for CFC gases, as well as - acceptance of the underlying scientific (and Nobel prize winning) explanation of the link between CFCs and ozone depletion. This investigation is important within the overall framework of the Thesis as

14 it shows that even in a case where an intergovernmental agreement has been deemed a success; the success may be linked to a myriad of factors, rather than the policy targets alone.

Indeed, as noted previously, policy success may occur for both the right and wrong reasons.

The last area of research in Chapters four and five covers issues surrounding water trading between agricultural firms. Many of the trading schemes that have been developed in the

United States have fallen below expectations with little trading taking place. The amount of trades that take place is important. For example, with grandfathered allocations, if no trades were to occur, this would imply that either the amount of permits initially allocated was efficient or there are significant barriers to trade. As the starting point for these types of trading schemes is often an out-of-equilibrium allocation, a distinctive and innovative agent based model has been developed to review the projections of trading subject to: - transaction costs, - a lack of auctioning, - the mixture of firms within the trading scheme, and - crop water demands. Upon reviewing these constraints, an innovative trading scheme has been developed which consists of a permit/tax hybrid policy mechanism that allows for the reduction of transaction costs via the option of utilising an intermediary. In addition, the scheme incorporates an environmental tax to directly account for the regional impacts of water use on water quality. This area of research is important within the framework of this

Thesis as it aims to focus on issues that are not traditionally incorporated in economic models

(such as geographic distance and out of equilibrium behaviour) while reviewing the ability of a Network Trading Scheme to foster trade and gauge the importance of the inclusion of intermediaries (or representative agents).

While these three streams of research are distinct from each other, the underlying theme linking these topics is the focus on applied environmental economics and the appraisal of policy prescriptions in each area. The interaction of policy and theory is an important underlying theme which tends to highlight theoretical gaps and new areas of research for economists to uncover. As a result, this Thesis is titled ‘Applied Environmental Economics:

Bridging the divide between policy and theory within the context of recycling, macro-level environmental indicators, and water trading’.

1.2 Sequence of the thesis

The next chapter addresses recycling collection contracts and a conflict between the incentives facing local governments and the prescriptions of economists hoping to establish an efficient rate of recycling through unit pricing for household waste collection. In doing so, chapter two reviews a case where these economic policy recommendations have not been adopted. As seen in sub-sections 2.3.2 and 2.3.4, analysis of data from municipalities in New

South Wales shows that different levels of the household charge for waste collection across municipalities has little association with different levels of recycling. Instead, the level of household income and the percentage of the council population who rent a unit or apartment are stronger determinants of the level of recycling across municipalities. Chapter Two is consequently entitled: The Persistence of Flat Fee Waste Collection Charges1 as it investigates the issues surrounding the municipality’s decision to implement recyclable material collection schemes and the underlying factors that result in households providing as much material as they do without a clear price signal.

1 Chapter Two is a modified version of a paper presented on 27 June 2009 during the Waste and Recycling session at the 17th Annual Conference of the European Association of Environmental and Resource Economists (EAERE) held at the VU University in Amsterdam. The author would like to thank all participants and the paper discussant (Johan Eyckmans) for their useful feedback and comments. 16

Chapter Three then reviews contrasts between the Montreal Protocol and the Kyoto Protocol and an evaluation of their policy success. In doing so, there is also a focus on whether an

Environmental Kuznets Curve (EKC) relationship persists once a concerted policy effort has been allowed for. Accordingly, Chapter Three is entitled: Going forward by looking backwards on the Environmental Kuznets Curve2.

The last two chapters focus upon the constraints of trade within a non-point source water permit market. This market is formulated as an ‘offset’ trading program for water quality trading where there is an estimated level of effluent that has been accounted for in setting the cap on water allocations to agricultural firms. Having developed a theoretical model of the agricultural firm’s decision between their primary production activity (growing crops) and selling permits, an agent based model is employed to review the model’s sensitivities using a representative region of the Murray Darling basin in New South Wales, Australia. In developing the model, the price differential between the respective crop price and the water permit price, as well as the water intensity of these crops, are identified as important determinants in the decision to trade permits. In addition to these variables and traditional determinants of trade (such as transaction costs), the diversity of the region (i.e. the mixture of potential buyers and sellers of permits) is important in fostering successful trade. Chapter

Four is accordingly entitled: Out of equilibrium trade, diversity and transaction costs3 and

2 Chapter Three is a modified version of a paper presented on 25 June 2009 during the International Agreements 1 session at the 17th Annual Conference of the European Association of Environmental and Resource Economists (EAERE) held at the VU University in Amsterdam. The author would like to thank all participants and the paper discussant (Gernot Wagner) for their useful feedback and comments. 3 Chapter Four and Five are extensions of a workshop presentation titled Network Trading, Salinity and the Level of Effluent Monitoring within the Design of Non-Point Source Water Trading, presented at the Centre for Applied Economic Research’s third multidisciplinary workshop on salinity mitigation on 8 December 2008. Exerts of this work was also presented at a FEEM/IEFE Joint Seminar held on 8 July 2010 in the FEEM offices 17 provides the details of an agent based model of non-point source agricultural water trade for

Faba Beans, Wheat, Barley, Oats and Canola in the Murray Darling Basin. Within Chapter

Five there is an application of this model to a review of a Network Trading Scheme. Chapter five utilises the model developed in chapter four for a test-bedding of an alternative market design. The title of this Chapter is: Appraisal of a Network Trading Scheme.

Chapter Six then concludes the Thesis by reviewing the conclusions of the research within this work, as well as highlighting plans and recommendations for further research.

in Milan Italy. The seminar was titled: ‘Out of Equilibrium Trade, Network Trading and Transaction Costs – an agent based model of agricultural water trade in the Murray Darling Basin’. The author would like to thank all participants for their useful feedback and comments. 18

Chapter 2 – The Persistence of Flat Fee Waste Collection Charges

2.1 Introduction

While Australia established recycling services comparatively early with respect to many other countries, the country is also one of the biggest per capita producers of waste. The first initiative by a municipality within New South Wales (NSW) to separate recyclable materials from household waste occurred in 1975 and kerbside recycling collection programs started to be introduced in urban areas during the late eighties and early nineties (Planet Ark, 2005).

Within the report titled New South Wales State of the Environment 2009, it was noted that as at the end of 2008, kerbside collections for dry recyclables4 were provided by 119 councils across NSW – a 19% increase since 2000. Within the Sydney Metropolitan Area the kerbside recycling rate has risen from 88.2 kg/person/year in 2000-01 to 106.5 kg/person/year in 2007-

08 (DECC, 2009). The NSW Department of Environment, Climate Change and Water has stated that “despite their good recycling performance, audits show that there is still a large

4 Dry recyclables include newsprint, cardboard, paper, and food and beverage containers. 19 amount, more than 820,000 tonnes, of potentially recyclable material being put in household garbage bins” (NSW DECCW, 2009). While the same report states that “there is clearly an opportunity to improve performance by working with key stakeholders, such as local councils, to encourage householders to recycle more material”, the implementation of a scheme that imposes an economically efficient pricing of waste, such as a unit price for domestic waste collection, was not considered. As this was not considered by the NSW

Department of Environment, Climate Change and Water report, the focus of the chapter is why the persistence of flat fee waste collection charges occurs despite advice from

Economists, such as Callan and Thomas (1999), that a variable pricing scheme would be more efficient.

Over the last two decades municipal solid waste policy reform in the United States has gradually incorporated market instruments to encourage source reduction and waste diversion activities such as recycling and composting. This reform has been driven from an awareness of the external environmental costs of waste disposal, increased land fill charges and the potential revenue gained from selling recyclable materials. One of the market instruments proposed has been the introduction of a variable rate or unit pricing scheme aimed at creating an incremental cost of disposing household waste and increasing the amount of recyclables collected via subsidised (or free) household collection of recyclables. However, a direct alternative to such a program, a flat fee pricing scheme, is still the most prevalent. This is inherently inefficient due to the non-incremental nature and subsidisation of the domestic waste collection charge. Jenkins et al. (2003) notes that “without unit pricing, most communities finance waste disposal via general tax revenues or flat fees ... from the perspective of households, this places a marginal price of zero on waste disposal.” (Jenkins et al. (2003): 295)

In contrast to the growth of unit pricing in the United States, Halvorsen (2012) found that within a 2008 survey focusing on policy incentives, over 60% of Australian respondents reported being within a flat fee waste collection system. It should be noted that this figure

(60%) has not been adjusted for ‘other’ and ‘don’t know’ responses. In addition, the five other fee systems reviewed are individually reported by less than 5% of Australian respondents (Halvorsen, 2012: 22-23). Indeed, volume, weight, frequency or per capita fee systems are rare within Australia.

Unless the subsidisation of domestic waste collection decreases or the cost of disposal increases to the point that annual or monthly council rates are substantially increased, there is little price incentive for households to recycle. Further to this, as the review is focused on the public provision of waste management services, any price mechanism is subdued with a lack of explicit and timely communication of the domestic waste component of the council rate charged per year (Callan and Thomas, 2007: 357). Indeed, “charges buried in the annual rate bill provide no signal about costs and no reward for reducing the quantity of garbage” (Wills,

2007: 33a). Yet recycling occurs even though the negative externalities and social costs of waste disposal (such as odour, groundwater pollution, aesthetics, and greenhouse gas emissions) are not fully accounted for by the household as they are not internalised within the disposal decision other than via domestic waste receptacle size. It is from this that it can be asked – if there is no price incentive to increase recycling, then why do households continue to recycle at a significant rate and what explains the difference between the recycling rates of local municipalities?.

It is this issue that the current chapter focuses on, as while the encouragement of recycling has been a policy directive, providing incentives via the internalisation of the social cost of waste has not been actively sought. The prevalence of outsourcing recyclable material collection has tended to dominate and this chapter contends that the combination of a flat-fee charge for waste collection, relatively high recycling rates and outsourcing of collection occurs due to the interaction of a cost minimising government agency with a profit maximising firm. With a cost minimisation motive, local governments tend to adopt basic recycling programs while satisfying household and constituent demand for the provision of an undefined recycling program. Despite a focus upon efficient resource use pointing toward the implementation of a unit based recycling program, this chapter will show that adverse selection may tend to reinforce the adoption of a basic collection program. A profit maximising collection agent has little incentive to offer a more intensive service. Indeed, collection agents are likely to hedge their bets and rely upon higher than expected recycling rates to accrue additional rent. For example, the recapture of aluminium tends to be high and this recyclate attracting a large sale price and a low conversion loss – refer to table 2.2 for details.

The trend of local governments outsourcing provision of recycling services to private and public trading companies has been established within this chapter through data from the

Australian Bureau of Statistics (refer to the discussion surrounding table 2.1) and within a survey of waste management specialists from a range of NSW councils (refer to sub-section

2.2.2 for further details). Upon reviewing the responses received from council representatives on why a recycling collection program was implemented, refer to sub-section 2.2.2, it was noted that in addition to ‘community and political desire ... it was a sweetener offered by the contractor when Council went out to tender for waste collection services’. During the 2009-

2010 financial year there were 709 private firms and public trading companies operating in waste management/collection within NSW and these organisations received 693.2 million

Australian Dollars from sales of recyclable or recoverable materials while expending 190.4 million Australian Dollars for contracting and subcontracting expenses. In comparison, 153 government organizations expended 317.8 million Australian Dollars on contracting and subcontracting expenses, while only receiving 9.7 million Australian dollars from sales of recyclable or recoverable materials (ABS, 2011).

Section 2.2 commences the paper with a discussion of literature on recycling programs and a presents a case study based on responses from a survey of municipal local governments in the state of New South Wales on the eastern coast of Australia. Sub-section 2.2.1 reviews literature that discusses examples where economic policy recommendations have not been adopted. Following a discussion of the intricacies of household domestic waste and recycling collection (within sub-sections 2.2.2, 2.2.3 and 2.2.4), an analysis focusing on the per capita level of recycling within 76 municipal local governments in the state of New South Wales will be conducted.

To commence the empirical analysis of this chapter, contained within section 2.3, balanced panel data from the 1998/1999 to the 2005/2006 financial year will be introduced in sub- section 2.3.1 and used to conduct the analysis within sub-sections 2.3.2, 2.3.3 and 2.3.4.

Reviewing the question proposed in the previous paragraph, sub-section 2.3.2 will focus on whether recycling per capita is associated with the council rate charged for all council services. In addition, the domestic waste charge component of this council rate is also focused on the effectiveness of a flat fee service charge and whether subsidisation occurs.

Sub-section 2.3.3 will further explore this issue of whether waste collection is subsidised and if economies of scale exist in the provision of waste collection. Sub-section 2.3.4 will then focus on additional factors which are related to recycling levels per capita and by incorporating the 2001 and 2006 Australian census data it is established that medium household income and the profile of housing are determinants of recycling.

As a prelude to the results of the chapter, which are presented in section 2.4, the analysis of data from municipalities in NSW shows that different levels of the household charge for waste collection across municipality services has little association with different levels of recycling. Instead, the level of household income and the percentage of the council population who rent a unit or apartment are stronger determinants of the level of recycling across municipalities.

2.2 Discussion of Literature and a Case Study on Council Recycling Programs

2.2.1 Initial Implementation of Council Recycling Programs Upon noting increased council provision of household recycling collection, Kinnaman and

Fullerton (1999) examine whether economic or non-economic forces are responsible for this trend. Their conclusion is that the increased provision has been due to a mixture of factors and that the probability of a curbside recycling program increases with higher cost savings, increased tipping fees, larger population density, higher levels of environmental interest group membership, and greater university/college education (Kinnaman and Fullerton, 1999:

16). In a subsequent study, Kinnaman (2005) established three explanations for why 8,937 municipalities in the United States have implemented curbside recycling programs. These were found to include an economic motive through the revenue received from the sale of recycled material which in turn subsidises the cost of domestic waste collection. State legislation was also a key explanation and has been enacted to either stimulate or directly mandate recycling programs. The third explanation established was residential demand for voluntary collection programs or constituent pressure to implement a mandatory collection program (Kinnaman, 2005: 1).

With respect to unit pricing programs, Miranda and Aldy (1998) note that at the time of their review, “more than 3400 communities in the US employ some variation of unit pricing”

(Miranda and Aldy, 1998: 79). The validity of these variations of unit pricing may be up for debate, as differing definitions could result in an overestimate of the number of unit based collection programs, irrespective of the move towards implementing some level of incremental cost. For example, programs based on a subscription service have been identified

25 as providing a discontinuous signal and only provide a weak incentive to recycle (Jenkins, et al. 2003: 312). Programs with a bag, tag or sticker approach are expected to be more responsive and act in the direct manner implied by the use of the term ‘unit price’. It has also been noted that with respect to unit pricing programs, “adoption rates nationwide have fallen below expectations, and many cities and towns struggle to gain support from their constituencies” (Callan and Thomas, 1999: 504). While these ‘sub-par’ adoption levels are usually attributed to ignorance of the environmental and economic costs of a subsidised flat fee domestic waste collection, the plausible explanations posed do not explain why some municipalities have adopted unit pricing schemes and others not (Callan and Thomas, 1999:

504). Rather than a broad range of factors, Fullerton and Kinnaman (1996) contend that the initiation of a volume-based recycling program in several communities in the US was due to specific economic or regulatory pressures. To summarise these factors, it was noted that “the average tipping fee paid by garbage collectors to landfills has tripled over a six year period, largely due to rising land prices and new EPA regulation” (Fullerton and Kinnaman, 1996:

971). In section 2.2.2 and 2.2.3 an additional reason for low adoption is proposed – that being that the incentives of the major stakeholders (specifically local governments and collection agents) results in a basic recycling program being preferred.

2.2.2 Focus on Initial Implementation Before continuing, it is imperative that this review explore whether the motivating factors for the implementation of a flat fee recycling program are consistent with those highlighted. As part of the analysis of NSW council data, contact was sought with the councils involved within the study to seek a description of the range of motivating factors for the initial implementation of (flat-fee) kerbside recycling collection within their council. Even though an attempt was made to contact all 76 of the councils in the study, the number of valid responses answering this question was limited to seventeen. While this response rate (22%) was lower than hoped for, the respondents were well placed to answer the question posed as they varied between Waste Education Officers, Waste Project Officers, to Managers of

Recycling and Waste Services from a range of urban and rural localities.

In any case, these responses will be treated as having come from individual interviews conducted online. As noted, the representative from the council was asked for the motivating factors for the initial implementation of household kerbside recycling collection. The explanations received ranged from public demand for a recycling program due to ‘increased awareness of the affect of waste on the environment both within council and the community’ to increased costs of landfill and hence the potential of cost savings from reduced domestic waste management costs. Discussion on the ability to outsource collection and ‘the opportunity to transport recycled material to a regional MRF’ also highlight the influence of private domestic waste and recycling service provision within the area. It must be noted that the responses received were of a high standard and while some responses were straight to the point or highlighted a specific issue influencing the council’s decision, the majority of responses were quite detailed and reflected an enthusiasm in discussing a program the

27 respondent took pride in. An example of such a complete and detailed response is highlighted in the following extended quote.

The motivators for a kerbside recycling service in (detail removed to maintain

anonymity) were the normal ones. Community expectation that we should offer

such a service, reduction of waste to landfill, (an) important step within any

waste management hierarchy, captured materials "easily" recyclable with good

expectation of resource recapture notwithstanding the obvious commercial

volatility of the recycling market, Governmental and legislative drivers,

avoidance of waste levies, green credentials at the political level and benefits of

economies of scale in the manner in which we have entered into a company

partnership with two adjoining Councils to capture the commercial advantages

that go with a service area of (over 100,000) properties.

For the municipalities included in the surveyed group, a variety of policy and economic motivations led to the establishment of kerbside collection programs. The quoted response raises the issue of economies of scale in the collection of waste and such a relationship is confirmed empirically in sub-section 2.3.3. With respect to the empirical analysis it should be noted that the a-priori expectation of the author is that a ‘good’ level of recycling will occur irrespective of a price effect. As noted in the quoted response, there was a similar expectation within the certain municipalities with the adoption of a collection program being coupled with an expectation that “captured materials (were) ‘easily’ recyclable with good expectation of resource recapture”. Also noted was the lack of interest in dealing with the “commercial

28 volatility of the recycling market”. Refer to Oskamp et al. (1991) for a discussion on the range of factors that influence household recycling behaviour.

2.2.3 Flat Fee Recycling Programs The discussion now turns to the decision between implementing flat fee or incremental pricing on domestic non-recyclable waste collection to stimulate recycling behaviour. Within a political sphere, increased recycling and the provision of collection programs are a positive result for a local government. This does not necessarily coincide with the implementation of unit based pricing collection based on the aim of achieving an efficient level of waste disposal and recycling. As noted in sub-section 2.2.1 and 2.2.2, from the council’s point of view, the varying factors which impact upon their desire to implement a scheme range from political feasibility and constituent demand for recycling collection to reduced management costs due to the compensating revenue from the sale of recyclable materials or outsourced collection. In all likelihood, irrespective of the underlying motivation of cost minimisation, councils may not desire to become a seller of recyclable materials as it is outside their speciality and has a large initial outlay of sunken establishment capital costs.

As suggested by the responses from the councils surveyed, there may be an incentive to outsource collection and reduce costs. In discussing outsourcing collection, Beede and

Bloom (1995) note that “profit-seeking firms generally have greater flexibility and incentive than government bureaucracies both to redeploy workers and physical capital quickly in response to changing circumstances and to design cost-cutting innovations” (Beede and

Bloom, 1995: 136). This motive of outsourced provision must not be taken for granted; upon reviewing the responses received from council representatives on why a recycling collection

29 program was implemented, it was noted that in addition to ‘community and political desire ... it was a sweetener offered by the contractor when Council went out to tender for waste collection services’. As previously noted, another response highlighted the desire to capture

“materials ‘easily’ recyclable with good expectation of resource recapture notwithstanding the obvious volatility of the recycling market”.

Reviewing data from 1995 for the US, Walls, Macualey and Anderson (2005) note that “42% of communities use contracts for curbside collection of recyclables, but a slightly lower percentage, 36%, use contracts for waste collection” (Walls, Macualey and Anderson, 2005:

593). This implies that the attractiveness of outsourced contracting for domestic waste and recycling collection is not limited to the NSW municipalities, as within the results of the

International City/County Management Association survey, government provision was at

38% for curbside solid waste collection and 40% for curbside recyclable material collection – with the rest being outsourced (Walls, Macualey and Anderson, 2005: 593). Indeed while there is a high tendency of government owned processing facilities for general waste, only

30% of the communities processing recyclables doing so using a government-owned facility.

This is in comparison to an approximate 56% government ownership of landfills facilities

(Walls, 2005: 207-208).

Table 2.1 – Waste Management Income and Expenses in New South Wales for 2009/10 New South Wales Australia

Government % Private/Public % % Trading Organisations at end June no. 153.00 709.00 Employment at end June no. 2,311.00 8,730.00 Income Income from waste services $m 168.60 17.3 1,953.40 61.6 Income from sales of recyclable or $m 9.70 1.0 693.20 21.8 recoverable material

Other sources of income and income $m 793.50 81.7 526.40 16.6 from energy generated from waste

Total income $m 971.80 3,173.00 Expenses Wages and salaries $m 140.40 17.0 527.90 17.7 Contract and subcontract expenses for $m 317.80 38.5 190.40 6.4 waste management services

Fees for the treatment/processing $m 161.70 19.6 245.80 8.3 and/or disposal of waste

Other expenses $m 206.20 25.0 2,010.10 67.6 Total expenses $m 826.10 2,974.20 Waste received by facilities other than landfills Recovered or reprocessed '000 t 4,010.3 65.0 Disposed to landfill or other final '000 t 1,300.9 21.1 destination

Transferred to other businesses for '000 t 876.7 14.2 recovery/reprocessing

Total '000 t 6,173.7 Source: ABS(2011)

Within Australia, during the 2009-2010 financial year, there were 586 private and public trading firms that were involved in collecting/selling recyclable or recoverable materials with income from sales of such materials having a value of 2,231 million Australian Dollars

(contributing to 26% of these companies total income). For government run enterprises the total sales of materials were 42.5 million Australian Dollars across 248 organisations

(contributing to 1.6% of these organisations total income). Government operators relied on

77% of their income coming from rate charges and private firms relied on income from waste services (59.9%). A notable proportion of these companies were located in NSW (709 waste management services equating to 33.4% of the total waste management services businesses in

Australia). New South Wales accounted for 31% (private and public trading firms) and 23%

(government operators) of income from the sale of recyclable or recoverable material. Table

2.1 utilises data from ABS (2011) to review the aggregate breakdown of income and expenses for waste management organizations within NSW. During the 2009-2010 financial year there were 709 private firms and public trading companies operating within NSW and these organisations received 693.2 million Australian Dollars from sales of recyclable or recoverable materials while expending 190.4 million Australian Dollars for contracting and subcontracting expenses. In comparison, 153 government organizations expended 317.8 million Australian Dollars on contracting and subcontracting expenses, while only receiving

9.7 million Australian dollars from sales of recyclable or recoverable materials (ABS, 2011).

Table 2.1 also reviews the amount of waste that was transferred to other businesses for recovery/reprocessing at the national level. In 2009/10, 876.7 thousand tonnes of waste was transferred to other businesses rather than being sent to landfill. Making up 14.2% of material received by facilities that were not landfill, this at least partly reflects the prevalence of subcontracting and sale/re-sale of recyclable/recoverable material. Table 2.2 reviews the percentage of households in NSW who reported having engaged in recycling and shows a high level of recycling through municipal kerbside recycling. For 2009, the percentage of households who stated that they did not recycle due to ‘not being interested, thinking that is was too much effort or prohibitive cost’ was low but differed across materials. The spread of the lack of recycling is interesting as the most recycled materials tend to be those with the lowest material loss, which may be linked to education, awareness and/or the availability of

32 recycling programs which have a focus on aluminium or steel cans. Indeed, recycling of aluminium cans was established at an early stage with high rates of recovery as early as 1994

(64% - Planet Ark, 2005) and this can be associated with the high commodity price of scrap aluminium, reported to be as high as $2000/t within LGASA (2009).

Table 2.2 – Recycling Trends in New South Wales 2000* 2003* 2006* 2009* Overall Mt Average Material Material Carbon 2007- prices Loss Emissions 2008 for Rates + Saved per Recycling recyclate Ton of per NSW per Material Resident tonne δ Recovered (kg) δ +

Percent of Households Percent Mt C Kilogram $/t NSW Households that Recycle 94.00% 94.80% 97.50% 97.90%

HH Recycling in Capital City 97.70%

HH Recycling in Other Areas 98.20%

Collected via municipal kerbside 89.40% 92.40% recycling

HH that do Total 14.00% not Recycle due to: Not interested / Paper, Cardboard or 15.00% 60% -93% 0.76 - 0.85 61.6 200-220 too much Newspapers effort / cost Glass 8.80% 88% 0.08 25.4 97

Aluminium Cans 5.80% 93% 3.7 0.98 2,000

Steel Cans 5.70% 98% 0.49 2.3 25

Plastic Bottles 12.50% 78% 0.42 5.6 400

Plastic Bags 11.40%

Kitchen or Food Waste 17.80%

Garden Waste 8.80%

Electronic Equipment 1.40%

Motor Oil 1.70%

Source: * ABS(2009), + EPA(2006), δ LGASA(2009) [as denoted in the first row].

Within a review of the structure of local government contracts with respect to recycling,

Walls (2005) noted that contracts “tend to be ‘fixed price’ to the extent that the contractor receives a fixed amount (usually per household) from the government and keeps any cost savings, or incurs losses, that result” (Walls, 2005: 208). Indeed, the “most fundamental reason why markets with imperfect information differ from those in which it is perfect is that actions (including choices) convey information, market participants know this, and it affects their behaviour” (Stiglitz, 2004: 37). This confirms that there is a prevailing moral hazard5 on behalf of firms appropriating a larger share of potential rents while offering a basic rather than intensive incremental collection program.

Hence the incentive to minimise costs and satisfy constituent demand for household recycling collection must also be considered with respect to the asymmetric information6 present within an out-sourcing contract’s development. With costs associated with contract development and monitoring, contracts are likely to be incomplete as all contingencies will not be allowed for.

Walls (2005) also notes that factors around the development of recycling contracts have tended to be more complicated with respect to domestic waste collection, and that this means a risk premium is likely to be paid to collection contractors. As the contracts tend to be fixed priced in terms of cost, “contractors have an incentive to try to reduce costs in order to increase their profits, but are not given direct financial incentive to recycle more” (Walls,

2005: 218). In light of the recycling rate within the table 2.2, the price differential of differing types of recyclables results in additional benefit to the contractor if the recycling rate or recyclables mix is more favourable than expected at initial contract negotiation.

5 Moral hazard is the case where a person does not guard themselves from risk as they are protected from its consequences. 6 Asymmetric information deals with cases where one party has more information whilst making a certain transaction or contract. 34

2.2.4 Recycling at Reasonable Rates Recycling may be expected to occur at reasonable rates due to non-economic factors such as: benevolence, environmental preference, or enjoyment of the activity itself. In discussing why households recycle with no direct price incentive, behaviour within the home needs to be considered with a review of time-use and opportunity cost. Hence the central question, if households are indeed insensitive to the level of price charged and with an effective marginal cost of zero of disposing household waste, then why does recycling occur? Halvorsen (2012) reviews the norms and attitudes that drive recycling and shows that Australian respondents from a 2008 survey noted that the most important reasons for recycling was a desire to contribute to a better environment and their civic duty.

With the major cost for household members being time, not money, there may be some form of ‘warm glow’ or ‘negative association’ which motivates recycling. A ‘warm glow’ effect refers to a positive association attached to recycling based on some notion of good will, environmental friendliness or contribution to sustainability. A ‘negative association’ effect refers to a negative sentiment with being seen as not acting in a ‘warm glow’ manner, such as being caught disposing of recyclable materials in general waste receptacles by passer-bys or neighbours. This negative association effect is expected to be linked to household income and hence the possibility of residing within a prosperous neighbourhood. While public provision of curbside recycling programs decrease the household’s time and effort devoted to recycling, the provision of such a service itself may also reinforce the notion of recycling as a ‘positive’ behaviour and supports the ‘warm glow’ or a ‘negative association’ effect. It has been noted that “the existence of social norms give rise to extrinsically motivated behaviour, as individuals may recycle to keep up appearance and gain respect in the community, or to express their attitude towards environmental issues” (Halvorsen, 2004: 4).

Within sub-section 2.2.4 the inclusion of a tenure variable is intended to reflect the impact of housing density and the observation that “conventional doctrine has been that recycling activity is likely to be limited in multi-family dwellings (MFDs)” (Ando and Gosselin, 2005:

426). Highlighted in the discussion within Ando and Gosselin (2005) is an United States

Environmental Protection Agency (US EPA) report which seemingly notes that “waste- management experts emphasise the importance of convenience in yielding high waste- diversion rates in MFD recycling programs, with some ‘record-setting’ programs using strategies like doorstep pickup of recyclables in high-rise apartment buildings” (Ando and

Gosselin, 2005: 426). It is expected that the inconvenience of being further away from the council provided receptacles matters. However there are other influencing factors such as whether the household has enough storage space for separate domestic waste and recyclables bins within their property. Indeed, Ando and Gosselin’s (2005) own regression results seem to reflect this irrespective of a limited sample size (15 observations from fourth floor households) and the possibility of correlation between floor location and missing demographic variables (Ando and Gosselin, 2005: 433-436). Ando and Gosselin (2005) admit this upon stating that they “only expect bin distance to have an effect on MFD recycling if by choosing to recycle an item of waste one commits to taking an extra trip to the waste disposal area” (Ando and Gosselin, 2005: 434).

2.3 Empirical Analysis

This section will review the determinants of the average amount of recyclables per residential property across different NSW Local Government Councils. By doing so, this section will review whether current rates of recycling are related to the price of domestic waste or auxiliary (non-economic) reasons. In other words, it focuses, in part, on why reasonable recycling occurs without the imposition of an economic instrument. Section 2.3.1 reviews the data that will be used in sections 2.3.2, 2.3.3 and 2.3.4 as part of regression analysis.

2.3.1 Data Analysis The report titled ‘Comparative Information on NSW Local Government Councils 2005/2006’ was produced by the Department of Local Government which is under the jurisdiction of the

NSW State Government. From the accompanied data published by the NSW Department of

Local Government for the financial periods from 1998-1999 to 2005-2006 a balanced panel dataset of 76 local council areas has been established. The formation of this balanced panel means that the 76 councils are those which have operated under the same name (hence allowing for significant mergers or consolidation) and also recorded above zero recycling in all of the periods reviewed. The resulting council areas, listed within table A1.1 in appendix

A, range from urban localities (such as metropolitan developed, regional town/city and fringe areas) to broader rural localities (such as agricultural and remote areas).

A variety of indicators related to domestic waste management and recycling services have been identified from within the report. The specific indicators that will be reviewed in this analysis are: the average amount of recyclables per residential property (RPerRP), the annual amount of domestic waste per residential property (DWKgPerRP), the average total rate charged per residential assessment (AverRate), the average charge for domestic waste

management services per residential property (AvDWChar), the percentage of outstanding

rates, charges and fees (PerOutst), the cost per service for domestic waste collection

(CostDW), and the level of environmental management and health expenses per capita

(EnManExpPerC). The indicators defined in terms of Australian Dollar value have been

adjusted to their ‘real’ value using the Consumer Price Index for the State’s capital city,

Sydney (ABS (2006) 6410.0). The labels used will continue as specified, but it is important to

note that such an adjustment has been made. In addition to these indicators, there are also

some descriptive variables which include: the Department of Local Councils Group Number,

which is based on the Australian Classification of Local Government and has been used to

create four location based dummy variables [metropolitan/developed urban, regional town or

city (RTownorCity), fringe urban (UrbanFringe), and rural (Rural)], as well as the number of

fulltime equivalent staff (FtEqStaff). Table 2.3 lists these variable definitions.

Table 2.3 – Variable Definitions Variable Variable Definition RPerRP average amount of recyclables per residential property DWKgPerRP annual amount of domestic waste per residential property AverRate average total rate charged per residential assessment AvDWChar average charge for domestic waste management services per residential property PerOutst percentage of outstanding rates, charges and fees CostDW cost per service for domestic waste collection EnManExpPerC level of environmental management and health expenses per capita RTownorCity located in metropolitan/developed urban, regional town or city UrbanFringe located in fringe urban area Rural located in rural area FtEqStaff number of fulltime equivalent staff Note: subscript i refers to the individual Local Council and t refers to time.

The average rate per residential assessment (AverRate) is the primary charge that households

pay a NSW council and covers a range of services including waste management. One

intention of the subsequent regression analysis is to confirm whether this variable has any

significant impact on the amount of recycling observed between the differing council areas.

Caution will be needed upon reviewing the effect of this variable as it may reflect a range of other factors within the municipality such as socioeconomic status, the mixture of residential, farmland and business properties, the level of services provided in the area and the financial/revenue status of the council itself. As a result, a range of other variables have been included as they account for some of these additional factors. The average charge for domestic waste management services per residential property (AvDWChar) represents the relative share of AverRate derived from waste management. This variable will also need careful interpretation to allow for the differences between recycling services provided and any correlation with the factors impacting the overall AverRate indicator. The level of environmental management and health expenses per capita (EnManExpPerC) is included to reflect the intensity of council participation in environmental regulation and may be extended in interpretation to cover the environmental concerns/knowledge of the general population as well as the council itself. Caution will be needed as the health component of this indicator will include factors such as the population and land usage mix.

To allow for the socioeconomic characteristics of the area and any regional specific shocks that may adversely affect household finances the percentage of outstanding rates, charges and fees (PerOutst) has been included within this analysis. Also included is the overall level of domestic waste per residential property (DWKgPerRP) which will provide a basis for the possibility of recycling diversion, and possibly the differing level of consumption between areas reflecting the level and profile of goods consumed. The cost per service of domestic waste collection (CostDW) will allow for the methods used in collecting waste as well as any difficulties in providing the service, such as narrow streets or multiple high density buildings.

The number of domestic pickups per week (DPickupsPW) is included within the analysis via the inclusion of CostDW; as discussed in sub-section 2.2.3.

Figure 2.2 to figure 2.9 review the general movement in the indicators of interest across the

76 council localities from 1998/1999 to 2005/2006. Focusing on Figures 2.5 and 2.6, the relationship between the cost per service of domestic waste collection and the number of domestic pickups per week seem to be interrelated. From the 2002/2003 period, there seems to be a structural break with a corresponding increase in the number of pickups and decrease in the cost per collection. In sub-section 2.2.3, this relationship is examined further and confirms a significant relationship implying the existence of increasing returns to scale from the number of household pickups per week. An alternative explanation of this structural break may be the introduction of the Waste Avoidance and Resource Recovery Act 2001, which resulted “in the abolition of waste boards, the formation of a specialist waste reduction agency (Resource NSW), a requirement for a NSW Waste Strategy and waste reduction targets and the introduction of strengthened extended producer responsibility (EPR) provisions” (NSW Dept. Of Environment and Climate Change, 2007). Hence, this decline in collection cost may also be related to a subsequent introduction of new receptacles within certain council areas and an overall improvement in knowledge/practices.

A general downward trend in the amount of domestic waste collected per residential property is accompanied by an upward trend in the amount of recycling per residential property. While this may seem to infer a direct relationship between these two indicators, multivariate analysis is needed to evaluate this as many of the other indicators also have an upward trend during the same period. Figure 2.9 reviews the correlation between the key variables from

1998/1999 to 2005/2006 with the statistical significance denoted with an asterix after the correlation estimate. The highest correlation of 0.6 is between AvDWChar and the time trend,

40 notable correlation exists between AverRate and AvDWChar (0.55) and also between

AverRate and the time trend (0.44). Variance inflation factors for these parameters have been reviewed and have been found to be between to 1 and 3, expect for the rural dummy with a

VIF of 4. Kennedy (1992) notes that a VIF of greater than ten reflects harmful collinearity – however the author has found cases where 2.5 was deemed a concern and caution should be used when reviewing AverRate and AvDWChar as they are correlated to many of the other explanatory variables.

Ideally, it is preferable to use data on individual households rather than using a cross section of council areas to examine specific factors influencing recycling within councils, as household data would allow us to make direct assertions rather than identifying general trends. However, the inclusion of multiple years of sequential data allows for a review of the difference in recycling rates between councils across key indicators and over time. It is on this basis that the identification of the determinants of recycling without price incentives is sought, as well as the relationship between the level of flat annual rate charges and their domestic waste charge component.

Figure 2.1 to 2.8 – Average of selected indicators across 76 council areas from 1998/1999 to 2005/2006

Figure 2.1 - DWKgPerRP (kg) Figure 2.2 - RPerRP (kg) 120000 500 100000 400 80000 300 60000 40000 200 20000 100 0 0

Figure 2.3 - AverRate ($) Figure 2.4 - AvDWChar ($) 800 250 600 200 150 400 100 200 50 0 0

Figure 2.5 - CostDW ($) Figure 2.6 - DPickupsPW 150 40000 30000 100 20000 50 10000 0 0

Figure 2.7 - FtEqStaff Figure 2.8 - EnManExpPerC ($) 360 30 350 25 340 20 330 15 320 10 310 5 300 0

Figure 2.9 – Correlation between indicators from 1998/1999 to 2005/2006

Recycling per residential property (kg)

1000 0.26* Average 500 Annual Rate ($)

0 400 0.18* 0.55* Average DW Charge 200 Component ($) 0 10 0.18* 0.44* 0.60* Year 5 Dummy Variable (1-9) 0 40 -0.05 -0.34* -0.24* -0.19* Percent of Charges 20 Outstanding (%) 0 1500 0.10 0.36* 0.16* 0.04 -0.22* Full Time 1000 Equiv. 500 Staff (No.) 0 300 0.01 0.17* 0.39* 0.14* -0.08 0.02 Cost Per 200 DW 100 Collection Run ($) 0 5000 -0.08 -0.28* -0.18* -0.15* -0.15* 0.11* -0.06 Domestic Waste per Residential Property (kg) 0 100 -0.00 -0.18* -0.08* 0.28* 0.17* -0.25* -0.10 0.05 EnMH 50 Expenses Per Resident 0 0 1000 20000 500 1000 0 200 400 0 5 10 0 20 40 0 500 1000 15000 100 200 3000 5000

2.3.2 Full Dataset Regressions Based on this dataset the following regression will be applied to examine the relationship between these variables on a multivariate basis:

(2.1)

As the average council rate charged to households may be affected by a high level of correlation to the domestic waste component of this fee, these regressions will be run once as the unrestricted model above and then subsequently as a restricted version where is dropped and the following test conducted, against . This restricted version reviews whether any finding of significance related to this variable in the unrestricted

43 model is due to the charge itself, or a spurious relationship with the level of income within the area. This suspected spurious effect may be due to the ability of the council to charge a higher rate in localities which have a higher level of wealth, as well as any relationship with higher household expenditure, the prevailing level of education or citizenship, environmental preference and warm glow or negative association motives.

The initial motive for directly contacting councils was to collect data on the type of collection and receptacles applied within the respective council area during the period reviewed as this is a crucial variable missing from the regression model. Unfortunately, the valid responses were limited and overall there was not enough information to create an effective variable to be incorporated within a regression analysis. Some of the responses also noted that there was a mixture of different recepticles within the same council area depending upon the rate of replacement and the staggered introduction of new recepticles. However, the information recieved does reflect an expected trend, where recycling collection schemes and the associated recepticals were implemented differently across different regions. It is from this that a range of locality based dummy variables have been included to allow for the differences between receptical size and other household factors related with the difference between metropolitan/developed areas, regional town/city areas, urban fringe areas and rural areas (agricultural or remote). Returning to the responses from the survery on the initial decision to adopt a recycling collection program (refer to sub-section 2.2.2), it can be confirmed that ‘rural areas may have issues with even providing recycling bins as the economics simply isn't there (adding another fleet of trucks costs millions to buy and run)’.

To allow for external effects that are not captured within the balanced panel, fixed effects and random effects regressions have been utilised. The fixed effects regression approach incorporates intercept terms which vary acrossthe individual units and is often referred to as the ‘within’ estimator. As interpreted by Verbeek (2000), this implies that the fixed effects estimator explains ‘what extent differs from’ and does not explain ‘why is different from’

(Verbeek, 2000: 315). The inclusion of the random effect regression approach is motivated by the intention to review overall population characteristics which explain the different rates of recycling between councils rather than explaining the level of recycling within individual local councils.

As time series cross section models are commonly expected to violate Gauss-Markov assumptions such, as panel heteroscedasticity, the Breusch-Pagan/Cook-Weisberg test has provided evidence that heteroscedasticity exists and hence corrected standard errors have been applied using the Huber/White/sandwich estimator of variance upon performing the subsequent RE/FE estimates. Also note that the Wooldridge test for autocorrelation in panel data could not reject the null hypothesis of no first order autocorrelation. Normality has been tested for using the Shapiro-Wilk test and as the null was reject in many cases the following regressions have had boot-strapping applied to gain consistent standard errors. As previously noted, correlation between the parameters of the major variables is presented in the appendix,

Figure 2.9. While there is significant correlation between some of the parameters (such as

AverRate and AverDWRate or AverDWRate and Year), it is limited in the case of the dependent variable and the forthcoming analysis reviews the importance of AverRate versus

AverDWRate within its design.

Reviewing the random effects7 estimations for the unrestricted model with the application of boot-strapping and robust standard errors (table 2.4); the average rate has a significantly positive relationship with the level of recycling per residential property. Irrespective of this, care must be taken in reviewing these results further as there is a high level of correlation between the average rate charged and the urban based intercept. This is evident from a comparison between the unrestricted and restricted model, as removing the average rate variable results in an increase in the intercept estimate and a sharp decrease in the rural indicator estimate. Within the restricted model, the amount of recycling per residential property in urban localities is estimated at 209 kilograms (given the other variables within the regression), whilst regional towns or cities and rural areas have higher levels. This higher level may reflect a lower percentage of households in comparison to the total recyclers

(including industrial or agricultural parties). Hence, it is evident that these dummy variables have proved useful in allowing for unobserved differences between regions based on demographics and possibly receptacle size.

Interestingly, within the restricted model the average domestic charge is slightly significant which reflects some limited evidence of a link between the cost and level of service provision to the recycling rates of different council areas. However the time trend is also significant with a higher confidence and there is likely to be a correlation effect as when the variable is not adjusted for inflation it is significant at a 1% confidence level. The number of fulltime equivalent staff is significant in both the unrestricted and restricted model pointing to some relationship with the overall size of the council.

7 Although I have presented the Fixed Effects Regression results, I will not review them as the focus is on reviewing the differences between the level of recyclables collected by councils. Further to this, upon applying the Bruesch-Pagan LM test for Random Effects, the null hypothesis is rejected and random effects are deemed more suitable. 46

Table 2.4 – Recycling per Residential Property – FE and RE Regression Analysis (n=608, no. councils = 76, no. years = 8) Fixed Effects Random Effects Constant 289.316** 221.685** 74.756 209.394*** (125.78) (92.52) (58.12) (48.64) Regional Town/City 53.510* 26.357 (28.43) (24.69) Fringe Urban -18.560 -19.220 (24.39) (22.09) Rural 99.178 9.265 (47.57) (36.09) Average Rate -0.212 0.324*** (0.29) (0.11) Average DW Char 0.376 0.353 0.405 0.478* (0.40) (0.35) (0.27) (0.31) Year 21.203 14.109** 1.290 11.792** (13.50) (6.36) (7.69) (5.83) Percentage Outst 1.324 1.147 0.474 0.803 (1.97) (2.69) (1.90) (2.25) FTeqS 0.034 0.026 0.085** 0.085*** (0.21) (0.23) (0.04) (0.03) Cost Per Serv -0.145 -0.123 -0.099 -0.139 (0.32) (0.26) (0.21) (0.21) DW per Res Prop 0.045 0.043 0.020 0.024 (0.05) (0.02) (0.04) (0.04) EnMH Exp pCap 0.188 0.199 -0.053 -0.097 (0.92) (1.08) (0.88) (0.81)

R-Squared 0.00 0.03 0.10 0.05 Breusch-Pagan LM Test for RE8 264.07*** 324.28*** P Value: *** - 1% ** - 5% * - 10%

Note: The variables included within this regression are as follows: Dependent variable – RperRP – Amount of recyclables per residential property, Independent variables – Constant – Intercept and the metropolitan urban area dummy variable, RTownorCity – Rural town or city dummy variable, UrbanFringe – Urban fringe dummy variable, Rural – Rural area dummy variable, AverRate – Average total rate charged per residential assessment, AvDWChar – Average charge for domestic waste management services per residential property, Year – Time trend, PerOutst – Percentage of outstanding rates, charges and fees, FTeqS – Number of fulltime equivalent staff, CostDW – Cost per service of domestic waste collection, DWperRP – Overall level of domestic waste per residential property, EnManExp – Environmental management and health expenses per capita.

8 Breusch-Pagan Lagrangian Multiplier Test for RE – null hypothesis: variances of groups are zero. 47

2.3.3 Relationship between Collection Costs & Number of Pickups Upon undertaking a review of the prevalent trends within the full dataset it became obvious that there is a relationship between collection costs and the number of weekly pickups. Using the following model shown in equation 2.2, the estimation of this relationship uses a range of dummy variables representing location, the number of pickups per week squared (to represent a high level of this indicator), the level of domestic waste collected, the average rate charged and the domestic waste component of this charge.

(2.2)

Reviewing the FE/RE results with robust standard errors applied,9 Table 2.5 shows that there is a significant decrease in the cost of collection with higher numbers of domestic pickups per week, pointing to increasing returns from scale. Note that upon reviewing municipal recycling in the United States in 1989 and 1996, Folz (1999) also found evidence of economies of scale with lower average unit costs corresponding with higher volumes of waste recycled. Costs also differ depending upon location and are generally related to the total domestic waste charge, but not the overall average rate charged to rate payers, possibly due to subsidisation.

This analysis was initially included to reinforce an observation made based on plots of the average trend of the key variables within the full dataset (figures 2.2 to 2.9). However, it should also be acknowledged that these results confirm a commonly used assumption, in that since “the average cost per ton of collecting, processing, or disposing of MSW generally

9 As time series cross section models are commonly expected to violate Gauss-Markov assumptions such as panel heteroscedasticity the Breusch-Pagan/Cook-Weisberg test has provided evidence that heteroscedasticity exists and hence corrected standard errors using the Huber/White/sandwich estimator of variance have been applied upon performing the subsequent estimates. Also note that the Wooldridge test for autocorrelation in panel data could not reject the null hypothesis of no first order autocorrelation. 48

varies with the amount of waste being handled, the scale of operations may be crucial to the

selection of cost effective management options” (Beede & Bloom , 1995: 126). Note that a

squared variable, which has a large number as a unit, results in small units for DPickupsPW².

Table 2.5 – Total DW Collection Costs – FE and RE Regression Analysis (n=608, no. councils = 76, no. years = 8) Fixed Effects Random Effects Constant 334.74 1417.459*** -43.874 -39.083 (425.39) (407.39) (401.22) (421.85) Regional Town/City -599.328*** -597.57*** (143.23) (177.74) UrbanFringe 360.068 357.019 (235.56) (269.97) Rural - - 1046.571*** 1045.740*** (279.42) (285.54) DPickupsPW 16.591 21.428* 75.544*** 75.801*** (14.99) (12.64) (10.91) (15.54) DPickupsPW² -0.00007 -0.00008 -0.0003*** -0.0003*** (0.00) (0.00) (0.00) (0.06) TotDWKgColl -0.039*** -0.039** 0.013 0.013 (0.00) (0.02) (0.01) (0.01) AverRate 3.431*** 0.043 (0.82) (0.40) AverDWChar 5.216*** 5.555** 6.253*** 6.268*** (1.83) (2.55) (1.70) (2.09) Year - 74.865* 21.998 22.987*** 32.141*** (32.79) (40.40) (23.44) (31.91)

R-Squared 0.03 0.00 0.76 0.76 Breusch-Pagan LM Test for RE 10 311.03*** 315.92*** P Value: *** - 1% ** - 5% * - 10%

Note: The variables included within this regression are as follows: Dependent variable – CollCosts – Total domestic waste collection costs, Independent variables – Constant – Intercept and the metropolitan urban area dummy variable, RTownorCity – Rural town or city dummy variable, UrbanFringe – Urban fringe dummy variable, Rural – Rural area dummy variable, DPickupsPW – Number of domestic pickups per week, TotDWKgColl – Total kilograms of domestic waste collected, AverRate – Average total rate charged per residential assessment, AvDWChar – Average charge for domestic waste management services per residential property, Year – Time trend.

10 Breusch-Pagan Lagrangian Multiplier Test for RE – null hypothesis: variances of groups are zero. 49

2.3.4 Two Period Data Using Australian Bureau of Statistics Census data from each of the council regions for the closest corresponding period (i.e. 2001 and 2006), a smaller dataset has been obtained with the addition of a variable representing median household income and another representing household tenure type (i.e. the typical type of residence within the area). The addition of median household income is focused on examining the relationship between recycling, the average rate charged and the level of household income. In addition, a variable capturing differences in tenure has been incorporated into the model to represent the percentage of the council population who rent a unit or apartment, as it is unlikely that they receive the council’s rate notice (therefore there is almost no price incentive for these households). As noted previously, ‘charges buried in the annual rate bill’ are expected to be an issue with respect to whether people change behaviour based on cost (Willis, 2007). Further to this renters typically have such charges incorporated in the amount of rent charged. The residents captured within the tenure variable are also likely to have a shared domestic waste bin system implying less constraint on receptacle space. The inclusion of the tenure variable is based on the notion that there will be a negative impact on unit/apartment recycling levels due to larger general waste receptacles and insensitivity to the council rate charged for domestic waste collection. While ideally it is preferable to capture these factors using separate variables, since the focus is upon council level data (as opposed to household level data), these results should be seen to be indicative of a general trend.

To commence the analysis of the two period dataset, the previous regressions will be repeated to confirm whether the relationship found within the full period dataset persists. Reviewing the unrestricted random effects results in table 2.6 shows that there is some evidence that the increase in recycling from the 2000/2001 to the 2005/2006 financial year and the difference

50 across councils can be related to the level of average rate charged. From the restricted model results, this relationship does not hold and is seemingly unrelated to the domestic waste component of the overall charge.

The focus now turns to the impact of the two additional variables on the amount of recyclables collected. From table 2.7, the inclusion of the additional Census data variables, removes any significant relationship between the average rate charged and the average amount of recycling per residential property. Indeed, there is a persistent significant positive relationship between the level of recycling in a council area and median household income. A persistent negative relationship holds for the type of tenure that persists within the area, with estimates of a decrease of between 9.6 kg and 11 kg in the average amount of recyclables per residential property with every 1% increase in the amount of residences that are rental units/apartments within the municipality. This is a 2.3% or 2.6% decrease compared to the

NSW average for each percentage difference in tenure. This implies that recycling rates in areas with a high prevalence of multi-family dwellings are indeed a potential policy and research issue needing further consideration. Hence it can be concluded that household specific factors are driving the differences in recycling across municipalities within New

South Wales. The cost of disposal and the charge per residential property has little to no impact on the differences within a sample of 76 local councils for the period between 1998-

1999 to 2005-2006. The low explanatory power of these regressions does lead to a need for caution as the likelihood of missing variable bias is high.

Table 2.6 – Recycling per Residential Property – FE and RE Regression Analysis (n=152, no. councils = 76, no. years = 2) Fixed Effects Random Effects Constant 204.446 141.956 125.345 251.750* (572.22) (483.82) (140.92) (154.48) Regional Town /City 62.891 36.515 (54.93) (82.24) Fringe Urban 21.291 24.914 (68.07) (79.74) Rural 198.373** 91.913 (98.21) (104.74) Average Rate -0.381 0.383* (1.31) (0.22) Average DW Char -0.161 -0.151 0.412 0.576 (1.50) (1.84) (0.74) (0.84) Year 109.239 41.847 -47.451 11.181 (295.79) (160.78) (96.37) (93.15) Percentage Outst 0.325 -0.315 -1.592 -1.100 (13.00) (12.92) (5.48) (5.87) FTeqS 1.039 0.956 0.113 0.114 (1.32) (1.36) (0.11) (0.09) Cost Per Serv -0.478 -0.473 -0.242 -0.305 (1.40) (1.28) (0.62) (0.62) DW per Res Prop 0.011 -0.003 0.018 0.024 (0.23) (0.18) (0.09) (0.09) EnMH Exp pCap -1.238 -1.322 -0.387 -0.552 (4.67) (4.31) (2.29) (2.15) R-Squared 0.00 0.01 0.06 0.03 Breusch-Pagan LM Test for RE 11 2.08 2.97* P Value: *** - 1% ** - 5% * - 10%

Note: The variables included within this regression are as follows: Dependent variable – RperRP – Amount of recyclables per residential property, Independent variables – Constant – Intercept and the metropolitan urban area dummy variable, RTownorCity – Rural town or city dummy variable, UrbanFringe – Urban fringe dummy variable, Rural – Rural area dummy variable, AverRate – Average total rate charged per residential assessment, AvDWChar – average charge for domestic waste management services per residential property, Year – Time trend, PerOutst – Percentage of outstanding rates, charges and fees, FTeqS – Number of fulltime equivalent staff, CostDW – Cost per service of domestic waste collection, DWperRP – Overall level of domestic waste per residential property, EnManExp – Environmental management and health expenses per capita.

11 Breusch-Pagan Lagrangian Multiplier Test for RE – null hypothesis: variances of groups are zero. 52

Table 2.7 – Recycling per Residential Property – ABS Variables – FE and RE Regression Analysis (n=152, no. councils = 76, no. years = 2) Fixed Effects Random Effects Constant 303.123 206.230 337.715* 411.717** (731.26) (611.19) (189.48) (178.97) Regional Town/City 12.716 -1.762 (78.93) (110.38) Fringe Urban -51.711 -59.296 (76.37) (108.05) Rural 98.777 48.920 (129.32) (142.45) Average Rate -0.586 0.168 (1.24) (0.25) Average DW Charge -0.238 -0.204 0.459 0.526 (1.56) (1.87) (0.76) (0.84) Year 107.140 14.789 -56.983 -38.074 (312.16) (162.87) (99.11) (101.49) Percentage Outst 0.244 -0.621 -2.333 -2.294 (17.22) (12.77) (5.44) (5.78) FTeqS 1.126 0.986 0.128 0.131 (1.39) (1.45) (0.11) (0.10) Cost Per Serv -0.429 -0.439 -0.345 -0.381 (1.50) (1.45) (0.64) (0.63) DW per Res Prop 0.010 -0.008 0.028 0.031 (0.24) (0.19) (0.08) (0.10) EnMH Exp pCap -1.698 -1.700 -0.883 -1.024 (5.12) (4.36) (2.46) (2.34) Median HH Income 765.346 540.45 808.11* 935.307** (1331.11) (980.70) (448.44) (458.40) Tenure -5.457 -5.722 -9.614*** -10.955*** (51.44) (48.41) (3.57) (2.81) R-Squared 0.01 0.01 0.10 0.10 Breusch-Pagan LM Test for RE 12 1.35 1.47 P Value: *** - 1% ** - 5% * - 10%

Note: The variables included within this regression are as follows: Dependent variable – RperRP – Amount of recyclables per residential property, Independent variables – Constant – Intercept and the metropolitan urban area dummy variable, RTownorCity – Rural town or city dummy variable, UrbanFringe – Urban fringe dummy variable, Rural – Rural area dummy variable, AverRate – Average total rate charged per residential assessment, AvDWChar – average charge for domestic waste management services per residential property, Year – Time trend, PerOutst – Percentage of outstanding rates, charges and fees, FTeqS – Number of fulltime equivalent staff, CostDW – Cost per service of domestic waste collection, DWperRP – Overall level of domestic waste per residential property, EnManExp – Environmental management and health expenses per capita, Median HH Income – Median household income within the council area, Tenure – Percentage of the council population renting a unit or apartment.

12 Breusch-Pagan Lagrangian Multiplier Test for RE – null hypothesis: variances of groups are zero. 53

2.4 Conclusion

The ineffectiveness of a flat fee pricing system in impacting recycling behaviour should be of no surprise as the domestic waste collection charge within NSW is typically set without any direct/immediate incremental increase in the size of the charge based on the amount of waste emitted. However, with correlation between recycling rates and either the total annual rate charged for municipal services or the median household income of a region, there is some evidence that recycling rates are positively related with higher incomes. Demand for environmental quality has often been linked to relatively higher incomes and this assumption has been used in the application of hedonic regressions within the valuation of environmental goods. Indeed, “the health and aesthetic implications of low quality air, land, and water make it likely that better-educated households have stronger preference for environmental quality”

(Beede and Bloom, 1995: 132).

The results in table 2.7 also show that recycling is negatively related to the proportion of the region that is comprised of units/apartments which are rentals. A persistent negative relationship holds for the type of tenure that persists within the area, with estimates of a decrease of between 9.6 kg and 11 kg in the average amount of recyclables per residential property with every 1% increase in the share of residences that are rental units/apartments within the municipality. Ando and Gosselin (2005) have previously noted that the

“conventional doctrine has been that recycling activity is likely to be limited in multi-family dwellings” (Ando & Gosselin, 2005: 426).

Part of the relationship between recycling rates and household income has been identified as reflecting the overall level of waste produced from higher consumption, however the 54 inclusion of a variable allowing for the average level of domestic waste per residential property (DWKgperRP) has partially controlled for a consumption effect. The statistical significance of the two variables focused upon in the two period regression points toward a warm glow/negative association effect where households recycle due to their knowledge and pleasure in doing a good deed or concern about not acting in the manner expected by peers.

Lower recycling rates based on the amount of units/apartments within the council area implies that sharing large or multiple waste receptacles means that disposal decisions are not impacted by a constraint on space. Consequently this implies that concerns of dumping garbage in either a neighbours’ bin or illegal dumping are not motivating factors for these households to undertake recycling.

With the bulk of the literature on recycling focusing upon either the motivating factors for undertaking recycling activities, the increase in the provision of household recycling collection programs, and the impacts of unit pricing on recycling rates, there has been little discussion of the local government’s decision between flat rate and unit pricing collection.

With a cost minimisation motive, local governments are likely to adopt basic recycling programs as their implementation coincides with reduced costs from outsourced collection to a profit maximising collection agent and satisfy household and constituent demand for the provision of an undefined recycling program. Despite economists, such as Callan and

Thomas (1999), having recommended the implementation of a unit based recycling program, adverse selection is expected to reinforce the adoption of a basic collection program. This is assumed to be due to the basic collection program satisfying the local governments’ cost minimisation objective and political requirements to constituents, while the profit maximising collection agent has little incentive to offer a more intensive service. Indeed, collection agents

55 may even hedge their bets and rely upon higher than expected recycling rates to accrue additional rent.

The trend of local governments outsourcing provision of recycling services to private and public trading companies has been established within this chapter with data from the

Australian Bureau of Statistics (refer to the discussion surrounding table 2.1) and a survey of waste management specialists from a range of NSW councils (refer to sub-section 2.2.2 for further details). Within the responses received from council representatives within an online survey on why a recycling collection program was implemented, it was noted that in addition to ‘community and political desire ... it was a sweetener offered by the contractor when

Council went out to tender for waste collection services’.

The regression analysis provides some evidence that recycling rates are likely to remain healthy due to warm glow or negative association effects. Therefore without external intervention from a higher level of government, removal of asymmetric information, or an exogenous factor such as a notable increase in tipping fees, the provision of basic flat fee recycling programs will persist. This partly explains why ‘sub-par’ adoption levels are usually attributed to ignorance of the environmental and economic costs of a subsidised flat fee domestic waste collection. The prevailing explanations posed “do not explain why some municipalities have overcome these universal hurdles and adopted unit pricing schemes, while others have not” (Callan and Thomas, 1999: 504). The incentives within the industry point to the status quo of flat fee pricing for waste collection coupled with outsourced provision of recyclable material collection being the preferred arrangement between private companies and local governments.

Chapter 3 – Going forward by looking backwards on the Environmental Kuznets Curve

3.1 Introduction

Having appraised policy-making on a local level within Chapter 2, the focus now turns to the appraisal of an intergovernmental agreement that has been deemed a success (the Montreal

Protocol) and another where success is unlikely to be widely recognised (the Kyoto Protocol).

The review of the relative success of these intergovernmental agreements is driven by the intention to establish that successful emission reductions tend to be associated with a concerted policy effort. This is in contrast to the concept of the Environmental Kuznets Curve

(EKC) which contends that a significant negative relationship exists between high levels of national income and per capita emissions. The second intention of this chapter will be a focus on whether such a relationship persists once a concerted policy effort has been allowed for.

And while the level of development and national income are likely to be linked to an ability to make structural changes and/or implement policy (as reflected in the separation of targets between Annex A and Article 5 countries within the respective intergovernmental

58 agreements), this review finds no evidence of a EKC consistent negative quadratic relationship between income and emissions.

The Montreal Protocol is an intergovernmental agreement that has been deemed to be a success and has been associated with the phase out of a range of ozone-depleting substances.

It has been noted that the treaty process for addressing ozone depletion “fundamentally changed the way certain industries conduct their business, already creating in some countries a complete phase-out of certain classes of chemicals.” (DeSombre, 2001: 49) In addition, a report by the US Environmental Protection Agency notes that “the ozone layer has not grown thinner since 1998 over most of the world, and it appears to be recovering because of reduced emissions of ozone-depleting substances.” (US EPA, 2007: 5) It also notes that “the Antarctic ozone is projected to return to pre-1980 levels by 2060 to 2075.” (US EPA, 2007: 5) In comparison to the Kyoto Protocol, the Montreal Protocol has been ratified by all UN recognised nations and the reasons for success are of interest to both environmental economists and policy makers. And while the success of the Montreal Protocol presents a case where emissions have been reduced by a concerted policy effort, auxiliary explanations for the success of the policy intervention are important. A comparison to the case of the

Kyoto Protocol highlights the importance of: - a supportive industry group, - pre-existing legislation and commitment by a lead nation, - affordable and available substitutes, as well as

- acceptance of the underlying scientific explanation of the link between emissions and a key detrimental impact.

The review of the success of the Montreal Protocol in comparison to the stagnation of the negotiations surrounding the Kyoto Protocol is driven, in part, by the intention to establish

59 that successful emission reductions tend to be associated with a concerted policy effort. This is in contrast to the concept of the Environmental Kuznets Curve (EKC) which contends that a significant negative relationship exists between high levels of national income and per capita emissions. And while the level of development and national income are likely to be linked to an ability to make structural changes and/or the implementation of environmental policy, this paper finds no evidence of an EKC consistent negative quadratic relationship between income and CFC emissions once key considerations, such as biased estimations and policy effort, have been accounted for.

The Environmental Kuznets Curve (EKC) is a positive relationship between a certain level of per capita Gross Domestic Product (GDP) and environmental quality improvements; which has been interpreted as implying that environmental management can take a backseat to attempts to advance GDP growth. Stern (2004) defines the EKC as “a hypothesized relationship between various indicators of environmental degradation and income per capita.”

(Stern, 2004: 1419) The body of literature on the existence of an EKC relationship is an interesting one, especially in light of its original observation being sourced from a paper with no direct intention of examining whether levels of GDP have a direct relationship with environmental quality. Grossman and Krueger (1991) actually set out to review whether reductions in trade barriers would improve or harm environmental quality with a focus upon the ‘Environmental Impacts of a North American Free Trade Agreement’. The subsequent discussion within the paper revolves around concerns that a pollution-haven13 may occur with the (then) impending introduction of the North American Free Trade Agreement (NAFTA). It

13 The pollution haven hypothesis has been described as being the situation where increased demand for environmental quality, “assumed to rise with increased income levels, does not lead to a shift to a cleaner production process in the country where the demand is generated, but rather to a movement of the production process to a location outside of the country” (Rothman, D. (1998):186). 60 was expected that “industry groups in the United States will demand less stringent pollution controls in order to preserve their international competitiveness, so that environmental standards will tend toward the lowest common denominator” (Grossman and Krueger, 1991:

2).

With these foundations, it may be concluded that the Environmental Kuznets Curve (EKC) relationship has been stumbled upon and subsequently interpreted and estimated before a theoretical basis could be established. Indeed, the authors themselves noted that their

“findings must remain tentative until better data became available” (Grossman and Krueger,

1991: 36). Within their follow up paper, ‘Economic Growth and the Environment’, Grossman and Krueger (1995) prioritise a view that any subsequent process leading to improved environmental conditions is not automatic. And while their paper does note that technological substitution and structural transformation are in principal important, “a review of the available evidence on instances of pollution abatement suggests that the strongest link between income and pollution in fact is via an induced policy response” (Grossman and

Krueger, 1995: 372). It is on this basis that skeptism concerning the validity of the EKC relationship has been borne; while there may be a correlation between a country’s level of development and their level of environmental quality, the factors driving this trend are by no means assured and the substitution, structural and preference changes underlying this correlation are sure to be diverse.

If indeed the underlying EKC relationship is an ‘induced policy response’, then an examination of the existence of an EKC relationship between income and the use of

Chlorofluorocarbons (CFCs) is of interest. The Montreal Protocol is a notable

61 intergovernmental agreement and has been deemed successful in reducing harm to the environment from an externality with transboundary implications. Focusing on CFCs and the

Montreal Protocol allows for a simultaneous investigation on whether an EKC consistent relationship exists and whether this relationship may alternatively be explained as an induced policy response with some level of difference based on income (or level of development).

Since the Montreal Protocol is not the only intergovernmental agreement to separate the level of policy response based on income levels and hence level of development, the investigation will be extended to investigate the existence of the EKC relationship within CO2 data and whether any such relationship found is impacted by the intention to ratify the Kyoto Protocol and fulfill the targets implied by this action.

Upon reviewing the Environmental Kuznets Curve (EKC) literature one notable case which has had little attention is the reduction of Chlorofluorocarbons (CFCs) and the implementation of the Montreal Protocol. The focus on the role of policy with respect to the existence of an EKC consistent relationship is based on observations within the literature that while technological substitution and structural transformation are in principal important, pollution abatement tends to occur through an induced policy response (Grossman and

Krueger, 1995: 372). So apart from disentangling the impacts of econometric issues with respect to the appraisal of the validity of a quadratic relationship between national income and environmental quality, the focus on policy allows this study to simultaneously investigate the issues related to the establishment, ratification and eventual success of intergovernmental agreements in achieving reduced international greenhouse gas concentrations and ozone depleting substances.

The structure of the chapter is as follows. A literature review and background discussion is conducted in section 3.2. After an initial review of the ‘humble beginnings’ of the EKC relationship and subsequent concerns over its validity/robustness in the sub-section 3.2.1, the chapter then highlights the ‘scarce attention given to CFCs’ in sub-section 3.2.2. Upon discussing CFCs and CO2, it is important that the impact of the Montreal and Kyoto

Protocols on the differing pollutant trends is addressed and this occurs in sub-section 3.2.3.

Empirical analysis then follows in section 3.3. After a brief discussion of some of the concerns over the EKC relationship’s ‘econometric foundations’ in sub-section 3.3.1, the diagnostic tests and functional forms to be estimated are specified in sub-section 3.3.2.

Before concluding the chapter in section 3.4, the results of the estimations will be reviewed for CFCs in sub-section 3.3.3, followed by the results for CO2 emissions from fossil fuels in sub-section 3.3.4. With the aim of focusing on the issue of policy success and attributable emissions reductions, section 3.3.5 will focus on whether the emission reduction targets of the Montreal Protocol are attributable to the rate of CFC reductions and whether CO2 emission reductions increased at a faster rate in the period after the Kyoto protocol became a binding agreement.

3.2 Literature Review and Background Discussion

3.2.1 Humble Beginnings From humble beginnings the EKC relationship has sparked a large debate which has captured the imagination, praise and scorn of many. It proved so topical that within about a year of appearing within the literature, the relationship was included within the subsequent World

Development Report published by the World Bank in 1992. Differing results across different pollutants and datasets meant that as early as 1994 the discussion extended to focus on reasons why these discrepancies may exist. It is with this that the literature started to review wider considerations of the existence of the EKC relationship. In 2005, Nahman and

Antrobus (2005) described the literature as one being divided between optimists, who strongly support the EKC and interpret it as validating a strategy of growth before all else (or much else), and critics, who suggest that methodological flaws are the reason for the relationship being found and that much more caution is needed when interpreting results showing an EKC consistent relationship (Nahman and Antrobus, 2005: 105). As the EKC relationship has attracted increasing criticism based on the lack of rigidity in much of the econometric underpinning, this review will focus upon the existence of an EKC with respect to policy implementation, while maintaining a sound methodological/econometric basis.

Concerns over the methodology applied within reviews of the EKC are not new, as the limitation of a reduced function form specification led Grossman and Krueger (1995) to profess that the functional form does not “even investigate the means by which income changes influence environmental outcomes” (Grossman and Krueger, 1995: 371). In 1997,

Panayotou discussed the implications of using a simple reduced-form approach (without a lengthy theoretical consideration) by comparing it to a ‘black box’. This term is especially

64 relevant to the present discussion of the EKC relationship in that this comparison reflects the view that such an approach “hides more than it reveals since income level is used as a catch- all surrogate variable for all the changes that take place with economic development”

(Panayotou, 1997: 466). Taking wider considerations into account is important, as an explanation of an appropriate EKC relationship is likely to be dynamic with a large multitude of underlying factors depending upon the pollutant, the countries included within the sample and the period reviewed. In addition to reviewing the application of the cerebus paribus principle, a concern with the interpretation of the EKC relationship is that any level of economic activity implies the use/extraction of resources. This resource use/extraction is not consistent with a functional form allowing the dependant variable to decrease to zero without some transfer between pollutants. Indeed, the first law of thermodynamics means that some waste is inevitable and as a result it should be enquired where this waste could be going.

Ultimately, this brings us back to the original Grossman and Krueger (1991) paper, as the transfer of pollution or polluting industries to less developed countries (i.e. the pollution haven hypothesis) has become a common explanation for the EKC relationship being found.

In addition to the pollution haven hypothesis, the new toxins explanation is also of interest to a review of the existence of an EKC relationship and the relationship between CFCs and hydrofluorocarbons (HCFCs). The new toxics scenario notes that as some pollutants are dealt with, other pollutants emerge and results in overall environmental quality stability, rather than reduction. In others words, “while some traditional pollutants might have an inverted U- shape curve, the new pollutants that are replacing them do not” (Stern, 2004: 1428). Indeed whilst the Montreal Protocol is often described as a success, the reduction in CFC emissions can also be seen as a rare, but fortunate case where a direct substitute for the pollutant was available. DeSombre (2001) described this substitutability between inputs and the existence

65 of gases such as HCFCs as a ‘happy coincidence’. Indeed DeSombre notes that within the

United States there was substantial support from industry for the Montreal Protocol and there was evidence of petitioning by major CFC manufacturers (such as DuPont) for ratification of the Montreal Protocol. Indeed the manufacturers that “were creating substitute chemicals would benefit from international regulation and the increased overseas demand for their new products it would bring” (DeSombre, 2001: 57). Typically this is not the case for CO2 emissions from fossil fuels as there is a high dependency on such resources and significant barriers to direct substitution. In addition, CFCs were produced by a relatively small number of manufacturers who could be effectively monitored and were often the producer of both the substitute (HCFCs) and the ‘targeted’ problematic gas (CFCs).

A factor that reinforces the new toxins explanation is that for many pollutants, measurement and the creation of datasets tends to follow health concerns or the actual implementation of environmental policy. This is true of many datasets, including the one used to source data for the current analysis. It has been said of a commonly used database [GEMS - which was used in the original Grossman and Krueger (1991) paper], that its research and data collection has focused “on a few ‘criteria’ pollutants, so-designated because legal statutes have required regulators to specify their damaging characteristics” (Dasgupta, et al. 2002: 150-151). Within the EKC literature itself, there are many and broad classes of emissions that have not been focused upon, especially in the case of toxic pollutants which often cause death, disease or birth defects. Further to this it has been contended by Dasgupta, Laplante, Wang and Wheeler

(2002) that “industrial countries surely must consider the daunting possibility that they are not actually making progress against pollution as their incomes rise, but instead are reducing only a few measured and well-known pollutants while facing new and potentially greater environmental concerns” (Dasgupta, et al. 2002: 149).

3.2.2 Scarce Attention Given to CFCs Even though the EKC relationship has been extensively researched, the relationship between

GDP and CFCs has been scarcely analysed and the data used within the existing studies are often insufficient. Mason and Swanson (2003) note that to their knowledge only three papers have studied the issue of CFCs and an EKC relationship. In 1997, Cole et al intended to extend the previous empirical analyses of the EKC relationship by reviewing a wider range of environmental indicators, including CFCs. Using cross-sectional analysis of data from 1986 and 1990 it was found that the adoption of the Montreal Protocol changed the growth profile of CFCs between these two years. This observation was accompanied by the statement that this result illustrated “the importance of multilateral action for a global air pollutant and tends to confirm that, without such a policy initiative, global air pollutants will increase monotonically with income” (Cole, et al. 1997: 412). Having established this result for CFCs,

Cole et al (1997) proceed to reinforce the view of Grossman and Krueger (1995) that while some developed countries have ‘grown out of’ some pollution problems, “there is nothing inevitable about the relationship between per capita income and environmental quality, as encapsulated in the EKC fitted to historical data” (Cole, et al. 1997: 412).

The conclusion that CFC emissions in the absence of the Montreal Protocol would continue to grow over the foreseeable future due to an excessively high EKC turning point was reinforced by Mason and Swanson (2003). Using an unbalanced panel of CFC production data from 1976 to 1988, Mason and Swanson (2003) find no evidence of an EKC consistent relationship using the traditional functional form specification and an excessively high turning point once a one period lag of CFC production is introduced into the model. While the analysis of Mason and Swanson (2003) does overcome some of the issues from previous analyses (such as cross-sectional data), the period involved limits the analysis to an

67 examination of the impacts of ratification at that point in time and subsequently the paper also forecasts the eventual impact of the Montreal Protocol using the targets set before the introduction of the Beijing Amendments. The time span of data is not the only data issue that can be identified within appraisals of the Montreal Protocol. Upon appraising the widely cited article by Murdoch and Sandler (1997), which reviews reductions in emissions and whether they are associated with non-cooperative Nash behaviour or cooperative behavior, Wagner

(2009) notes that the use of imputed data by Murdoch & Sandler (1997) leads to a spurious result. Specifically Wagner (2009) states that “the qualitative and quantitative evidence that

MS present to support their view relies on largely imputed data from the World Resources

Institute … which overstate emission reductions and appear to induce a spurious positive correlation between income and CFC cutbacks” (Wagner, 2009: 192).

On this basis, the usefulness of the dataset released by The Secretariat for the Vienna

Convention and the Montreal Protocol for the period 1992 to 2008 is evident, 14 as it both coincides with the first stage of the Montreal targets and covers the period of the Beijing amendments, including the period within which the maximum amount of reductions for all signatories was determined. It is on this basis that this chapter will review the existence of an

EKC consistent relationship and the impacts of the Montreal Protocol targets using a balanced CFC consumption dataset for the 67 countries within sub-section 3.3.3. In addition to this, an analysis of an EKC consistent relationship and the impacts of the ratification of the

Kyoto Protocol will also be presented within sub-section 3.3.4 using a balanced CO2 dataset

14 These data have been sourced from the UNEP’s GEO Data Portal which provides data compiled by a large range of original data providers. These data can be accessed via the Data Portal (http://geodata.grid.unep.ch) and is cited with respect to the source (UNEP, The GEO Data Portal). 68 spanning from 1990 to 2007 for 124 countries compiled by the Carbon Dioxide Information

Analysis Centre (CDIAC)15.

15 These data have been sourced from the UNEP’s GEO Data Portal which provides data compiled by a large range of original data providers. These data can be accessed via the Data Portal (http://geodata.grid.unep.ch) and is cited with respect to the source (UNEP, The GEO Data Portal). 69

3.2.3 Montreal and Kyoto The scant attention given to ozone depleting substances (with respect to greenhouse gases) within the EKC literature may be associated with the existence and relative success of the

Montreal Protocol. With CFC levels having been seen to decrease across developed nations,

“by most accounts, the treaty process for addressing ozone depletion is an unqualified success” (DeSombre, 2001: 49). And while the level of ratification and policy action related to the Montreal Protocol has thought to have had an impact on the reduction of CFCs,

DeSombre (2001) notes that the members of industry producing ozone-depleting substances

(ODS) and market forces have played a valuable role. The qualification here is that some of the market forces underlying a reduction in CFCs are seen to have occurred “as a direct result of the way the Protocol process is structured, and others because of serendipity in the way the industry has made or used ozone depleting substances” (DeSombre, 2001: 57). Further explaining this contention, DeSombre notes that “due to what is in part a happy coincidence, and in part well-developed regulatory incentives, some of the main ODS-producing industries were the main innovators of the substitutes used to replace them” (DeSombre, 2001: 57). This differs substantially to the policy process and the debate surrounding the Kyoto Protocol and the control of greenhouse gas emissions.

Reduced emissions and policy success are not the only differences between the Montreal and

Kyoto protocols, as ratification levels and industry support have substantially differed with climate science being scrutinized and debated. While the identification of climate change and its cause has been a subject of debate, by the time that the Montreal Protocol was introduced, the scientists (Rowland and Molina) whom advanced the theory behind the CFC explanation for ozone depletion in 1974 had already been awarded the Nobel Prize in Chemistry for their work. As a result, the risks associated with ozone depletion and their relation to CFC gases

70 was deemed credible and of direct concern to industrialised nations. This broader context is one of the contributing factors of the success of the Montreal Protocol and the support it received from industrialised nations. In contrast, the discussion of the collapse of the climate change negotiations in The Hague in December 2000 in Grubb and Yamin (2001) provides some of the issues that have surrounded the level of ratification of the Kyoto Protocol. In describing “the Protocol’s critics from all shades of the political spectrum” (Grubb and

Yamin, 2001: 262), Grubb and Yamin (2001) list the respective critics as follows.

These critics include the dwindling band of scientific sceptics who claim that the

scientific evidence base is still too weak to justify international action; the

predominantly Northern-based economic and industrial critics who claim that

industrialized countries’ Kyoto targets are too strong, and that international efforts

should focus on a fundamental rewriting of the Protocol to weaken these targets

and/or extend them to developing countries; and idealists who believe that targets

are too weak to be worthwhile (Grubb and Yamin, 2001: 262).

Indeed, based on the rate of ratification and reductions of CFCs many have concluded that the

Montreal Protocol and its predecessor (the Vienna Convention) are the most effective international agreements in existence. While Figure 3.1 shows the level of Montreal ratification to be high, upon comparing it to the Kyoto Protocol (using the data compiled by

The Secretariat for the Vienna Convention and the Montreal Protocol as well as the United

Nations Framework on Climate Change) it had a similar overall level of ratification as at

2004. While the overall level of ratification is important, the profile of the member countries must also be considered. In contrast to the Montreal Protocol, the United States has not ratified the Kyoto Protocol and this directly led to a nervous wait for the agreement to

71 become legally binding due to the requirement for 55 countries accounting for at least 55% of

1990 carbon dioxide emissions for the Protocol to enter into force. With the receipt of the

Russian Federation’s instrument of ratification on November 18 2004, the Executive

Secretary of the Climate Change Secretariat stated that “a period of uncertainty has closed.

Climate change is ready to take its place again at the top of the global agenda” (UNFCCC,

2004).

Figure 3.1 - Level of Adoption of Intergovernmental Agreements (n = 237)

200 180 160 140 120 Ozone 100 Kyoto 80 UNFCCC 60 40 20

1992 1991 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 1990

Focusing on the difficulties of intergovernment agreements and concerns over ratification there is academic debate about whether any international environmental agreement can have a ‘real’ impact in light of free riding and a lack of penalties/enforcement. Barrett (1990) notes that with no world authority able to intervene and enforce the targets/standards set, “there are strong incentives for government not to co-operate, or to defect from an agreement should one be reached” (Barrett, 1990: 69). This focus of individual parties following their own self- interest and private property rights leads to the reason why “the core problem in the first

72 period allocations (apart from the US withdrawal) concerned allocations to the EITs

(Economies in Transition) that have proved excessive” (Grubb, 2003: 186). The unexpected/unaccounted for fall of the USSR has lead to a situation where there was an excess of permits and hence the carbon price was expected to fall close to zero. Indeed,

Grubb (2003) notes that projections of the carbon price since 2001 have plummeted upon the introduction of three factors, these being: “the withdrawal of the US, by far the largest source of potential ‘demand’ in the system; revision of Russian energy projections which greatly increased their projected allowance surplus; and the subsequent Bonn/Marrakech deal on carbon sinks” (Grubb, 2003: 160).

3.3 Empirical Analysis

3.3.1 Econometric Foundations In recent years there has been a substantial literature focusing on the econometric basis of the

EKC relationship and while this chapter will aim to review this relationship using a solid econometric/methodological foundation, it is by no means a complete econometric review of the EKC. The intention is to establish whether an EKC consistent relationship exists for a pollutant where persistent decreases have been noted or whether other factors prevail (such as policy initiatives, intergovernmental agreements/targets, or unobserved country specific factors). The increase in studies using econometric methodologies to test the EKC relationship has been noted by Stern (2004) and can be seen as quite important as while “the

EKC is an essentially empirical phenomenon ... most of the EKC literature is econometrically weak” (Stern, 2004: 1420). Stern (2004) is critical of the nature of many past studies which look for significant coefficient estimates without paying attention to the statistical properties of the data used. The importance of reviewing the existence of the EKC using a robust empirical methodology is highlighted within the statement that “one of the main purposes of doing econometrics is to test which apparent relationships, or “stylised facts”, are valid and which are spurious correlations” (Stern, 2004: 1420).

Indeed Stern was not the first to notice that the lack of explanatory power within substantial

EKC studies meant that “explanations for the coefficient estimates are given ex-post, i.e., they are forced upon the regression results but remain untested” (de Bruyn, 1997: 487). In other words, the formulation of theory after estimation is not as rare as it is treacherous.

Empirical estimations need a theoretical base otherwise the risk of running a spurious regression is quite high, except in cases where the econometric analysis is particularly strong.

As a result, there is an increasingly common consensus that the EKC analysis is not robust, is based purely on prior assumptions and has actually missed some basic steps before estimation should begin. In support of this sentiment it has been suggested that with many of the analyses “choice of the quadratic estimates and their interpretation of these as inverted-U’s would therefore seem to derive more from their prior judgement as to plausibility than from the econometric results, which are indeterminate” (Ekins, 2000: 190).

Amongst the work focusing on the econometric validity of the EKC is the review of Perman and Stern (2003) who in focusing on panel cointegration16 found that the evidence for the

EKC relationship is questionable. Amongst the work applying unit root testing and adjustment for cointegration is Day & Grafton (2003) which also finds little evidence of an

EKC relationship. Decomposition analysis has also been applied within the literature in applications such as Stern (2002) where the issue of income is said not to matter and that there is an overbearing “importance of globally shared, emissions-specific technical change and total factor productivity growth in individual countries in reducing emissions” (Stern,

2002: 217). Also using decomposition analysis and regression on SO2 emission reductions, de Bruyen (1997) found that “the downward sloping part of the EKC can be better explained by reference to environmental policy than to structural change” (de Bruyen, 1997: 499). A recent study focused on CO2 emissions for Canada has utilised a range of estimation methods and finds no EKC relationship, indeed per-capita GDP and per-capita emissions increase monotonically (He & Richard, 2010). Indeed, much of the research completed since 2010 has focused on within country analysis with Shahbaz et al. (2013) for Romania, Fan and Zheng

(2013) for China, Tiwari et al. (2013) for India, and Sephton and Mann (2013) for Spain cited

16 Panel cointegration considers the degree of heterogeneity across the ‘n’ dimension of a sample. 75 as examples. An exception to this is Lin and Liscow (2013) that focuses upon instrumental variables and water pollution.

Having established that policy has been found to be more important than structural changes, a similar result will be established for CO2 and CFCs. Indeed, the aim of this chapter is to review the robustness of any EKC consistent result found using the standard functional form by introducing variables expected to remove any missing variable bias while adjusting for heteroscedasticity and serial correlation. It is with this that diagnostic tests to confirm whether heteroscedasticity and serial correlation are present have been run on the fixed effect and random effect estimations. As expected, due to the nature of panel data as well as the nature of the variables, both heteroscedasticity and serial correlation are found for both pollutants in each of the model specifications described in the following sub-section. Fixed and random effects regression analysis will be applied as it is expected that the reduced form specification of the model requires the allowance for unobserved country specific effects. The fixed effects and random effects estimates will be adjusted for heteroscedasticity and serial correlation on separate incidences, with feasible GLS being applied to examine the impact of allowing for both issues simultaneously.

3.3.2 Estimation Outline Following the standard functional form discussed within the EKC literature17, this analysis will begin with a review of whether such an EKC consistent relationship is present using the datasets complied. Moving from the standard EKC specification for CFCs (equation 3.1), the review will then examine whether any EKC consistent relationship found is robust enough to persist upon introducing key variables expected to explain the level and trend of CFC consumption during the sample period. Starting with the level of HCFC consumption (eq.

3.2), as HCFC gases are a commonly identified substitute for CFC gases, this chapter will then focus on the impacts of the Montreal Protocol’s targets for CFC consumption/production reduction (eq. 3.3), and allow for the few countries within the sample which have hesitated in ratifying the Protocol (eq. 3.4). Equation 3.3 and 3.4 also contain the Non Article5 time trend variable which allows for the separation of level of emissions with the change in emissions for these countries overtime.

(3.1)

(3.2)

(3.3)

(3.4)

A similar analysis will then be conducted for CO2 emissions, starting with the same specification for testing an EKC consistent relationship (eq. 3.5). Having established this basis, the robustness of any EKC consistent relationship will be tested with the inclusion of

17 Typically the standard EKC regression model is commonly specified as:

, with the turning point income specified as: .

77 variables indicating CO2 levels, Kyoto Ratification and UNFCCC participation. With the addition of the level and trend of CO2 emissions for Annex A countries, the estimates are expected to reflect the underlying justification for setting targets for these countries alone (eq.

3.6). Having established the basis and justification for targets being set for Annex A countries, the analysis will then evaluate the impacts of the respective views of the validity of the Kyoto Protocol and the expectation of the eventual/actual introduction of binding targets as reflected in the relationship between emissions and the number of years of Kyoto ratification and/or UNFCCC participation (eq. 3.7). Separate time trends for Annex 1 countries and the number of years a country has been a UNFCCC signatory have been included in equations 3.6 and 3.7. The accumulative number of years the Kyoto protocol has been ratified by each country specified is denoted as for country i, period t.

(3.5)

(3.6)

(3.7)

The introduction of this model specification has been based on the concerns during the sample period reviewed (1990-2009), over whether the Kyoto Protocol would reach the prescribed requirements for binding legality. Under this uncertainty, it can be expected that the countries ratifying relatively early are likely to be the countries determined to take action in line with the intentions of the Kyoto Protocol and the UNFCCC (rather than taking a ‘wait and see approach’). Additionally upon ratifying the Protocol, these countries also have a direct incentive to take earlier action to meet their prescribed target for the first phase (2008-

2012) as they have been based on their respective 1990 baseline emission level. As discussed within the earlier comparison between the Montreal Protocol and the Kyoto Protocol, these policy and motivational aspects are important contrasting factors which need to be taken into account upon discussing emission reductions, especially upon focusing on CFCs and CO2.

Subsequently sub-section 3.3.5 directly focuses on these issues with the estimation of an additional specification.

All of the equation specifications in this chapter (which are based on the same basic origins of equation 3.1) will be estimated using data from the Secretariat for the Vienna Convention and the Montreal Protocol (CFCs) and the Carbon Dioxide Information Analysis Centre

(CO2). Apart from including a more extensive time dimension (14 to 16 years), the balanced datasets compiled have a range of countries (with differing levels of CFCs/CO2 and GDP per capita) within the respective samples of 67 and 124 countries, respectively. Table B1.1 and

B1.2 within appendix B list the countries included within these samples and also denote those belonging to the Non-Article 5 and Annex A groupings.18 Dealing with growth paths of countries with differing levels of development implies that a diverse mix of countries within the analysis is important. Indeed, some past research has investigated the EKC using panels of data with only a few countries and some have even been limited to OECD countries – with the results often being interpreted as having direct applicability to non-OECD countries. In the case of CFCs, only 12 of the 67 countries are labelled as Non-Article 5 countries. In the case of CO2, only 28 of the 124 countries are labelled as Annex A countries. This mix is important as apart from levels of development, these distinctions also reflect differing levels of policy prescriptions and targets.

18 Non-Article 5 countries are those which have been allocated targets under the first phase of the Kyoto Protocol and tend to be classified as developed countries by the World Bank. In other words, these are those countries who that not granted a reprieve from country specific targets in the first phase of Kyoto. 79

3.3.3 Estimation Results – CFC Within the heteroscedasticity robust (het robust) fixed effect and random effect estimation

(fe/re) results shown in table 3.1 there is some evidence of an EKC consistent relationship under both the fixed effect estimation techniques using the functional form specified in equation 3.1 and 3.2. However, upon including the Montreal Protocol target variables specified in equation 3.3 (Non-Article 5 and Non-A5 Time Trend) this EKC relationship is replaced with significant evidence of a policy induced decline in CFCs by Non-Article 5 countries above the decreases occurring over time by all of the countries in the sample. These results show that in addition to an overall decrease in the consumption of CFC gases during the time period and based on exogenous factors within individual countries19, Non-Article 5 countries had significantly higher levels of CFC emissions, decreasing by approximately

0.8% per year during the relevant phase out period.

It should be noted that the targets implemented by the Montreal Protocol mandate both the production and consumption of CFCs. In interpreting the results, it needs to be remembered that they apply to the consumption of CFC gases and hence will include the consumption of

CFCs from imported goods by the respective Non-Article 5 and Article 5 countries. This is beneficial as any review of the production of CFC gases would need to consider concerns of

‘pollution havens’ and the export of emissions which have been noted as a potential factor behind results showing an EKC consistent relationship. In light of these considerations, the results shown within table 3.1 reflect the influence of the factors impacting upon end user emissions as the consumption of CFCs has been calculated by taking national production of

CFCs, adding imports, and subtracting exports, destroyed quantities and feedstock uses of individual CFCs. Upon allowing for autoregressive order one AR(1) disturbances, the results

19 Indeed a negative trend is expected as action on reducing CFC consumption has existed since the banning of nonessential aerosols in the USA, Canada, Norway and Sweden in 1978 (Auffhamer et al (2005): 379). 81 in table 3.1 are largely replicated within table 3.2 with similar policy results shown. The estimates show an EKC consistent relationship being replaced by a statistically significant decrease in CFC consumption within Non-Article 5 countries of approximately 0.4% or 0.7% per year depending upon whether fixed effects or random effects are applied.

While allowances have been made for heteroscedasticity (het) and serial correlation (AR) separately, both of these factors can be simultaneously controlled using feasible generalised least squares (FGLS). Allowing for heteroscedasticity and an AR(1) process, the results from these FGLS estimations are shown in table 3.3. These results do not have the fixed/random effect model specification applied, so specification bias and differences with the previous results are potentially present. Of interest within these FGLS results is a comparison of the het adjusted estimates with the het/AR adjusted estimates which mainly differ upon reviewing the statistical significance of the respective coefficient estimates. The discrepancy reflected is consistent with an observation made by Wooldridge (2008) while discussing the simultaneous occurrence of both heteroscedasticity and serial correlation; “much of the time serial correlation is viewed as the most important problem, because it usually has a larger impact on standard errors and the efficiency of estimators than does heteroscedasticity”

(Wooldridge, 2008: 440). Focusing on the results, the policy variables show a significant decrease in CFC consumption within Non-Article 5 countries of approximately 0.7% or 0.8% per year depending on whether het and AR have been controlled for simultaneously.

However, within these results there is no significant difference between the level of consumption of Non-Article 5 and Article 5 countries. An EKC consistent result is also not found for the FGLS het and AR joint-adjusted results, casting doubt on the relationship’s validity with respect to CFCs with and without the impact of the Montreal Protocol.

Table 3.1 – CFC per capita – Fixed/Random Effects (1992-2008) lgCFCpc – FE lgCFCpc – RE lgCFCpc – FE lgCFCpc – RE lgCFCpc - FE lgCFCpc – RE lgCFCpc - FE lgCFCpc – RE Constant -48.785*** -22.893*** -36.260** -15.175* 22.748 -7.350 31.186* -7.729 (18.88) (7.88) (18.71) (7.99) (17.77) (5.55) (18.00) (5.58) lgGDPpc 5.983** 1.433 4.800* 1.089 -4.069 -1.153 -5.903** -1.143 (2.94) (1.45) (2.90) (1.44) (2.75) (1.00) (2.83) (1.01) lgGDPpcsq -0.260** -0.069 -0.213* -0.061 0.071 0.063 0.157 0.063 (0.12) (0.07) (0.12) (0.06) (0.11) (0.05) (0.12) (0.05) Time Trend -0.528*** -0.516*** -0.566*** -0.551*** -0.308*** -0.416*** -0.330*** -0.423*** (0.05) (0.03) (0.05) (0.03) (0.05) (0.03) (0.05) (0.03) lgHCFCpc 0.229*** 0.211*** 0.146*** 0.147*** 0.142*** 0.147*** (0.04) (0.04) (0.04) (0.04) (0.04) (0.04) Non-Article5 - -0.764 - -0.475 - (0.96) - (0.99) Non-A5 TimeT -0.818*** -0.791*** -0.805*** -0.784*** (0.06) (5.55) (0.06) (0.06) Ozone Tre. 1.877*** 0.596 (0.70) (0.48)

n 1139 1139 1139 1139 1139 1139 1139 1139 i 67 67 67 67 67 67 67 67 R² 0.13 0.19 0.14 0.18 0.16 0.49 0.17 0.49 20 1654.20*** 1707.57*** 553.93*** 541.88*** 3.44 7.96* 6.69 13.19** Equation (3.1) (3.2) (3.3) (3.4) P Value: *** - 1% ** - 5% * - 10%

Note: The variables included within this regression are as follows: Dependent variable – CFC per capita – Amount of CFC emissions per capita, Independent variables – Constant – Intercept, lGDPpc – log of GDP per capita, lGDPpcsq – log of GDP per capita squared, Time Trend – time trend for 1992-2008, lgHCFCpc – log of HCFC emissions per capita, Non- Article5 – Dummy variable for Non-Article 5 countries, Non-A5 TimeT – Time trend for Non-Article 5 countries only, Ozone Tre. – Ratified an Ozone Treaty (zero until year of ratification).

20 Breusch and Pagan Lagrangian multiplier test for random effects – null hypothesis: Var(ai) = 0 (random effects inappropriate).

Table 3.2 – CFC per capita – Fixed/Random Effects with AR(1) disturbances (1992/1993-2008) lgCFCpc – FE lgCFCpc – RE lgCFCpc – FE lgCFCpc – RE lgCFCpc - FE lgCFCpc – RE lgCFCpc - FE lgCFCpc – RE Constant -21.769** -17.229** -20.734** -15.342* -6.285 -10.694* -3.775 -11.037** (9.61) (7.99) (9.72) (19.45) (10.92) (5.69) (11.07) (5.72) lgGDPpc 4.174 0.371 4.084 0.350 1.175 -0.923 0.694 -0.883 (5.45) (1.473) (5.40) (1.48) (5.02) (1.04) (5.06) (1.04) lgGDPpcsq -0.303 -0.020 -0.298 -0.021 -0.162 0.055 -0.142 0.05 (0.26) (0.07) (0.26) (0.07) (0.23) (0.05) (0.23) (0.05) Time Trend -0.656*** -0.573*** -0.657*** -0.584*** -0.548*** -0.463*** -0.546*** -0.467*** (0.11) (0.04) (0.11) (0.04) (0.10) (0.04) (0.10) (0.04) lgHCFCpc 0.034 0.063** 0.035 0.067** 0.037 0.067** (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) Non-Article5 - -0.462 - -0.322 - (1.17) - (1.19) Non-A5 TimeT -0.386*** -0.725*** -0.390*** -0.721*** (0.14) (0.10) (0.14) (0.10) Ozone Tre. 0.547 0.314 (0.96) (0.51)

n 1072 1139 1072 1139 1072 1139 1072 1139 i 67 67 67 67 67 67 67 67 R² 0.07 0.19 0.07 0.19 0.13 0.49 0.13 0.49 21 18.31 21.09*** 31.00*** 28.80*** Equation (3.1) (3.2) (3.3) (3.4) P Value: *** - 1% ** - 5% * - 10%

21 Hausman specification test – null hypothesis: the individual effects are uncorrelated with the other regressors in the model. 84

Table 3.3 – CFC per capita – FGLS with het and AR(1) adjustments (1992-2008) het adjusted results het and AR(1) adjusted results lgCFCpc lgCFCpc lgCFCpc lgCFCpc lgCFCpc lgCFCpc lgCFCpc lgCFCpc Constant -14.967*** -13.951*** -8.413*** -8.475*** -14.233*** -19.630*** -11.599* -11.793* (0.65) (1.24) (1.72) (1.85) (3.26) (7.89) (6.74) (6.75) lgGDPpc -0.056 -0.055 -1.278*** -1.293*** -0.414 0.812 -1.380 -1.564 (0.13) (0.24) (0.36) (0.36) (0.60) (1.57) (1.36) (1.37) lgGDPpcsq -0.002 -0.004 0.073*** 0.075*** 0.009 -0.041 0.083 0.099 (0.01) (0.01) (0.02) (0.02) (0.02) (0.08) (0.07) (0.07) Time Trend -0.516*** -0.510*** -0.388*** -0.386*** -0.536*** -0.564*** -0.467*** -0.492*** (0.01) (0.01) (0.03) (0.03) (0.07) (0.04) (0.07) (0.07) lgHCFCpc 0.040*** 0.109*** 0.106*** 0.016 -0.005 -0.002 (0.01) (0.02) (0.02) (0.01) (0.02) (0.02) Non-Article5 -0.250 -0.392 0.793 0.972 (0.47) (0.48) (1.10) (1.06) Non-A5 TimeT -0.824*** -0.831*** -0.661*** -0.659*** (0.04) (0.04) (0.09) (0.09) Ozone Tre. -0.119 1.284** (0.28) (0.63)

N 1139 1139 1139 1139 1139 1139 1139 1139 I 67 67 67 67 67 67 67 67

2494.82*** 3544.56*** 10036.80*** 9765.14*** 66.59*** 241.37*** 633.51*** 674.50*** Equation (3.1) (3.2) (3.3) (3.4) P Value: *** - 1% ** - 5% * - 10%

3.3.4 Estimation Results – CO2 The analysis now turns to an examination of an EKC consistent relationship and the impacts of the ratification of the Kyoto Protocol using CO2 data spanning from 1990 to 2007 compiled by the Carbon Dioxide Information Analysis Centre (CDIAC). Within table 3.4 (het robust results) it can be noted that there is evidence of a significant EKC relationship upon applying the random effects estimation procedure to equation 3.5. Upon the introduction of the Annex A variables (with the estimation of equation 3.6) it can be noted that during the full sample period these countries had a significantly higher rate of emissions (approximately higher than non-Annex A countries by 1.32%) which decreased by approximately 0.02% per year, having controlled for country specific effects. These results reflect the justification for separating the policy making and target setting into two groups as the Annex A countries were noted to have higher emissions due to their level of development. Indeed, Grubb (2003) notes that higher per capita emissions in the industrialized countries are “one of the reasons why industrialized countries accepted the responsibility for leading climate change efforts in the UNFCCC and subsequent Kyoto negotiations: unless they can control their own high emissions there is little prospect of controlling emissions from developing countries that start from a very much lower base” (Grubb, 2003: 144).

The decrease of only 0.02% per year reflects a lack of action in reducing emissions during this period and reinforces a statement made in a 1997 United Nations Framework Convention on Climate Change (UNFCCC) press release outlining the negotiated targets accompanying the Protocol. Using projected emission statistics for the year 2000, the UNFCCC noted that even though industrialised nations have been postulated to reduce their collective GHG emissions by 5.2%, “the total reductions required by the Protocol will actually be about 10%; this is because many industrialised countries have not succeeded in meeting their earlier non- binding aim of returning their emissions to 1990 levels by the year 2000” (UNFCCC, 1997).

Upon estimating equation 3.7 and hence adding the amount of years of Kyoto ratification and

UNFCCC participation into the model there is no significant decrease of CO2 emissions based on the number of years of ratification and participation. A lack of significance is unsurprising due to the differing views on the validity of the Kyoto Protocol and the expectation of the eventual/actual introduction of binding targets.

Within table 3.5 the results allow for AR(1) disturbances and again find no evidence of an

EKC relationship. With the introduction of the Annex A variables it can be noted that during the full period reviewed these countries had a significantly higher rate of emissions22 which decreased by approximately 0.03% per year, having controlled for country specific effects.

Table 3.6 presents results for FGLS (adjusted for heteroscedasticity and serial correlation simultaneously), shows little similarity with the fixed and random effects results leading to concerns of misspecification bias. Hence priority will be given to the previous fe/re results within table 3.5 as these results allow for country specific effects and have been adjusted for serial correlation.

22 This is noted to be at least 1.36% above the country specific differences and the emissions of non-Annex countries – reflected in an intercept of -0.87. 87

Table 3.4 – CO2 per capita – Fixed/Random Effects (1990-2007) lgCO2pc – FE lgCO2pc – RE lgCO2pc – FE lgCO2pc – RE lgCO2pc - FE lgCO2pc – RE Constant -23.242*** -20.787*** -21.059*** -18.367*** -21.222*** -18.404*** (0.94) (0.84) (1.01) (0.88) (1.02) (0.88) lgGDPpc 0.561*** 0.419*** 0.364*** 0.201*** 0.374*** 0.203*** (0.08) (0.08) (0.09) (0.08) (0.09) (0.08) lgGDPpcsq -0.004** -0.003* -0.000 0.001 -0.000 0.001 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Time Trend -0.009*** -0.002** -0.004 0.006* 0.005 0.012* (0.00) (0.00) (0.00) (0.00) (0.01) (0.01) Kyoto Rat. -0.021 -0.019 -0.002 -0.004 (0.02) (0.02) (0.03) (0.03) UNFCCC Part. 0.020 0.017 0.004 0.005 (0.02) (0.02) (0.02) (0.02) Annex A - 1.315*** - 1.320*** - (0.31) - (0.31) Annex A TimeT -0.016*** -0.018*** -0.016*** -0.018*** (0.00) (0.00) (0.00) (0.00) Yrs Kyoto Rat. -0.005 -0.004 (0.01) (0.01) Yrs UNFCCC -0.009 -0.007 (0.01) (0.01) n 2232 2232 2232 2232 2232 2232 i 124 124 124 124 124 124 R² 0.22 0.22 0.21 0.34 0.21 0.34 23 17628.07*** 17217.10*** 17072.76*** Hausman 33.02*** 42.40*** 44.12*** Equation (3.5) (3.6) (3.7) P Value: *** - 1% ** - 5% * - 10% Note: The variables included within this regression are as follows: Dependent variable – CO2 per capita – Amount of CO2 emissions per capita, Independent variables – Constant – Intercept, lGDPpc – log of GDP per capita, lGDPpcsq – log of GDP per capita squared, Time Trend – time trend for 1990-2007, Kyoto Rat. – Ratified the Kyoto Protocol (zero until year of ratification), UNFCCC Part. – Participate in the UNFCCC, (zero until year of commencement), Annez A – Dummy variable for Annex A countries, Annex A TimeT – Time trend for Annex A countries only, Yrs Kyoto Rat. – Years since Kyoto Protocol was ratified (used to distinguish between long-term and recent ratification), Yrs UNFCCC – Years of UNFCCC participation (used to distinguish between long-term and recent participation).

23 Breusch and Pagan Lagrangian multiplier test for random effects – null hypothesis: Var(ai) = 0 (random effects inappropriate). 88

Table 3.5 – CO2 per capita – Fixed/Random Effects with AR(1) disturbances (1990/1991-2007) lgCO2pc – FE lgCO2pc – RE lgCO2pc – FE lgCO2pc – RE lgCO2pc - FE lgCO2pc – RE Constant -15.488*** -17.308*** -14.737*** -16.052*** -15.040*** -16.067*** (0.23) (1.11) (0.24) (1.11) (0.25) (1.11) lgGDPpc -0.022 0.137 -0.104 0.031 -0.098 0.030 (0.15) (0.11) (0.15) (0.11) (0.16) (0.11) lgGDPpcsq 0.005 0.002 0.008* 0.003 0.007* 0.03 (0.00) (0.00) (0.00) (0.00) (0.01) (0.03) Time Trend -0.003 0.002 0.003 0.011*** 0.034 0.016** (0.01) (0.00) (0.01) (0.00) (0.02) (0.01) Kyoto Rat. -0.010 -0.012 -0.005 -0.008 (0.02) (0.02) (0.02) (0.02) UNFCCC Part. 0.007 0.011 0.015 0.010 (0.01) (0.02) (0.02) (0.02) Annex A - 1.680*** - 1.671*** - (0.31) - (0.31) Annex A TimeT -0.022*** -0.021*** -0.022** -0.021*** (0.24) (0.01) (0.01) (0.01) Yrs Kyoto Rat. -0.010 -0.003 (0.01) (0.01) Yrs UNFCCC -0.021 -0.006 (0.02) (0.01)

n 2108 2232 2108 2232 2108 2232 i 124 124 124 124 124 124 R² 0.24 0.23 0.22 0.38 0.21 0.38 24 5.41 - - Equation (3.5) (3.6) (3.7) P Value: *** - 1% ** - 5% * - 10% Note: The variables included within this regression are as follows: Dependent variable – CO2 per capita – Amount of CO2 emissions per capita, Independent variables – Constant – Intercept, lGDPpc – log of GDP per capita, lGDPpcsq – log of GDP per capita squared, Time Trend – time trend for 1990-2007, Kyoto Rat. – Ratified the Kyoto Protocol (zero until year of ratification), UNFCCC Part. – Participate in the UNFCCC, (zero until year of commencement), Annez A – Dummy variable for Annex A countries, Annex A TimeT – Time trend for Annex A countries only, Yrs Kyoto Rat. – Years since Kyoto Protocol was ratified (used to distinguish between long-term and recent ratification), Yrs UNFCCC – Years of UNFCCC participation (used to distinguish between long-term and recent participation).

24 Hausman specification test – null hypothesis: the individual effects are uncorrelated with the other regressors in the model. 89

Table 3.6 – CO2 per capita – FGLS with het and AR(1) adjustments (1990-2007) het adjusted results het and AR(1) adjusted results lgCO2pc lgCO2pc lgCO2pc lgCO2pc lgCO2pc lgCO2pc Constant -12.685*** -13.470*** -13.4307*** -15.187*** -11.788*** -11.083 (0.18) (0.15) (0.19) (2.01) (2.16) (2.64) lgGDPpc -0.286*** -0.154*** -0.155*** -0.040 -0.366 -0.459* (0.02) (0.02) (0.02) (0.20) (0.23) (0.28) lgGDPpcsq 0.011*** 0.006*** 0.006*** 0.005 0.012** 0.015** (0.00) (0.00) (0.00) (0.01) (0.01) (0.01) Time Trend 0.005*** -0.013*** -0.034*** -0.002 0.010* 0.014 (0.00) (0.00) (0.00) (0.01) (0.01) (0.01) Kyoto Rat. 0.211*** -0.004 -0.015 -0.001 (0.02) (0.02) (0.03) (0.03) UNFCCC Part. 0.161*** 0.217*** 0.008 -0.004 (0.02) (0.02) (0.02) (0.02) Annex A ------Annex A TimeT ------Yrs Kyoto Rat. 0.081*** -0.028** (0.01) (0.01) Yrs UNFCCC 0.020*** -0.008 (0.01) (0.02)

n 2232 2232 2232 2232 2232 2232 i 124 124 124 124 124 124

3767.57*** 2439.70*** 2228.58*** 37.69*** 42.68*** 37.98*** Equation (3.5) (3.6) (3.7) P Value: *** - 1% ** - 5% * - 10% Note: The variables included within this regression are as follows: Dependent variable – CO2 per capita – Amount of CO2 emissions per capita, Independent variables – Constant – Intercept, lGDPpc – log of GDP per capita, lGDPpcsq – log of GDP per capita squared, Time Trend – time trend for 1990-2007, Kyoto Rat. – Ratified the Kyoto Protocol (zero until year of ratification), UNFCCC Part. – Participate in the UNFCCC, (zero until year of commencement), Annez A – Dummy variable for Annex A countries, Annex A TimeT – Time trend for Annex A countries only, Yrs Kyoto Rat. – Years since Kyoto Protocol was ratified (used to distinguish between long-term and recent ratification), Yrs UNFCCC – Years of UNFCCC participation (used to distinguish between long-term and recent participation).

3.3.5 Emission Reductions and Policy Having established that there is little to no evidence of an EKC consistent relationship within the CFC and CO2 datasets employed in sub-sections 3.3.3 and 3.3.4, the analysis will now further develop the discussion on emission reductions with respect to intergovernmental agreements. Within sub-section 3.3.4 policy impacts were left at an aggregate level – separated between Article 5 and non-Article 5 countries. In this section there is a concerted effort to disentangle the specific policy target periods and define the stages within which emission reductions were to be met. Table 3.7 shows the results from a modified analysis which applies the three regression methods to the equation specification shown in equation

3.8. The major change to this equation is the removal of the Article 5 related variables specified in equation 3.3 and their replacement with dummy variables representing the timing of the different levels of legislated emission targets.

(3.8)

Table 3.7 shows that for the Non-Article 5 countries the stages of having a 75% and 100% reduction target in CFC emissions (denoted in equation 3.8 as

) are insignificant and that there were no notable trend of emissions reductions above that associated with the time trend. This lack of defined action can be seen until after the policy stipulates no allowable CFC emissions (as represented by a significant reduction in the Non-A5 – Post variable) and can be reconciled with claims that emission reductions were already occurring independent of the Montreal protocol. This pre-existing tendency is reflected in a consistently significant and negative relationship with the time trend implying decreases in per capita emissions of between 0.4% and 0.6% per annum from

1992 to 2008. These results can be interpreted as the following: notable reductions in CFCs can directly be associated with the period that the Montreal protocol was in force (but not the timing of the reduction targets) while the low rate of emission reductions that did occur in

Non-Article 5 countries cannot be associated with the stipulated reductions set by the Kyoto targets. This result coincides with the earlier discussion of the success of the Montreal protocol being associated with already present industrial factors that can be seen in an existing decreasing relationship over time and as reflected in the constant (which is partially associated with the non-Article 5 75% reduction target period of the Montreal targets). With respect to Article 5 countries, the results show that higher rates of CFC use persisted throughout the initial phases of the Montreal targets, these being above the overall CFC decline seen in all countries and associated with the time trend and the constant - which in part represents the decreases associated with the initial period of Non-Article 5 targets. The rationale for separating targets was intended to allow less developed nations more time to adjust to the policy and this indeed was utilized by these nations. With respect to the Article 5 specific variables, the results show that decreases in CFCs were associated with the overall time trend, rather than the specific targets set. Note that the current data only reaches the penultimate target period for Article 5 countries and the success of reductions towards zero use cannot be determined using this data.

Turning our focus to the Kyoto protocol, equation 3.7 has been adapted to include the period since the Kyoto protocol became binding and a variable which represents whether the protocol has been ratified and the level of emission reductions this implies due to that respective country’s target (as reflected in TargetandRat equation 3.9). Table 3.8 allows us to focus on the Post Binding period variable which shows that there is a no significant difference in the emission rate of CO2 by countries who are subject to the binding targets.

There was also no evidence of a difference in emissions based on the target set for the respective countries and this adds uncertainty to whether emission rates have reduced specifically due to the Kyoto protocol or due to other underlying factors. One consideration is that the European Union (which makes up a large proportion of the countries within the

Annex 1 category) had enacted strong policies before the post binding period and that associated emission reductions are likely to be part of the significant and negative time trend representing the period from 1990 to 2007. In addition to this are factors such as continued inaction by a range of countries, the doubt over continued climate change policy, and that the data set does not cover the initial emissions reduction target period which is 2008 to 2012.

(3.9)

Table 3.7 – CFC per capita – Policy Focus (1992-2008) FE RE FE – AR(1) RE – AR(1) FGLS het FGLS het AR(1) Constant 8.065 -9.462* -19.167* -12.455** -7.627*** -11.553*** (17.32) (5.41) (11.12) (5.75) (1.65) (4.05) lgGDPpc -2.059 -0.565 2.489 -0.467 -1.184*** -1.262 (2.70) (0.98) (4.75) (1.05) (0.34) (0.89) lgGDPpcsq 0.019 0.035 -0.167 0.032 0.069*** 0.086* (0.11) (0.04) (0.22) (0.05) (0.02) (0.05) Time Trend -0.436*** -0.494*** -0.675*** -0.527*** -0.476*** -0.552*** (0.06) (0.03) (0.10) (0.05) (0.02) (0.04) lgHCFCpc 0.184*** 0.184*** 0.051 0.084*** 0.132*** 0.048*** (0.04) (0.03) (0.03) (0.03) (0.01) (0.01) Non-A5 – 75% ------Non-A5 – 100% -0.668 -0.718 0.448 -1.395* -0.845* -0.931 (1.00) (0.88) (1.16) (0.78) (0.49) (0.73) Non-A5 – Post -9.086*** -9.124*** -4.342*** -7.877*** -9.796*** -7.642*** (0.85) (0.64) (1.57) (0.77) (0.29) (0.72) A5 – Base 0.956 0.014 -0.307 0.116 -0.695* 0.857 (0.70) (0.50) (0.99) (0.57) (0.37) (0.64) A5 – Freeze 2.402*** 1.406** 0.595 1.011 0.420 1.477** (0.79) (0.61) (1.04) (0.63) (0.44) (0.69) A5 – 50% 3.188*** 2.282*** 1.844* 1.960*** 1.181*** 2.238*** (0.72) (0.52) (1.02) (0.60) (0.37) (0.73) A5 – 85% 2.931*** 1.949*** 2.163** 1.809*** 1.367*** 3.067*** (0.82) (0.64) (1.12) (0.71) (0.47) (0.82) A5 – 100% -0.147 -1.224* 0.533 -0.276 -2.087*** 0.997 (0.85) (0.68) (1.19) (0.79) (0.48) (0.88)

n 1139 1139 1139 1139 1139 1139 i 67 67 67 67 67 67 R² 0.27 0.52 0.26 0.51

647.04*** Hausman 9.91 36.09*** Wald Chi² 3728.65*** 718.98*** P Value: *** - 1% ** - 5% * - 10% Note: The variables included within this regression are as follows: Dependent variable – CFC per capita – Amount of CFC emissions per capita, Independent variables – Constant – Intercept, lGDPpc – log of GDP per capita, lGDPpcsq – log of GDP per capita squared, Time Trend – time trend for 1992-2008, lgHCFCpc – log of HCFC emissions per capita, Non-A5 – 75% - Dummy variable for Non-Article 5 countries during the period within which there was a 75% reduction target, Non-A5 – 100% - Dummy variable for Non-Article 5 countries during the period within which there was a 100% reduction target, Non-A5 – Post - Dummy variable for Non-Article 5 countries during the period within which there were no CFC emissions allowed, A5 – Base - Dummy variable for Article 5 countries during the period with no emission targets, A5 – Freeze - Dummy variable for Non-Article 5 countries during the period within which emissions were to show no growth, A5 – 50%/85%/100% - Dummy variable for Non-Article 5 countries during the period within which there was a 50%/85%/100% reduction target.

Table 3.8 – CO2 per capita – Policy Focus (1990-2007)

FE RE FE – AR(1) RE – AR(1) Constant -21.187*** -18.338*** -15.035*** -16.069*** (1.03) (0.89) (0.25) (1.11) lgGDPpc 0.370*** 0.196** -0.099 0.030 (0.09) (0.08) (0.16) (0.11) lgGDPpcsq -0.000 0.001 0.007* 0.003 (0.00) (0.00) (0.00) (0.00) Time Trend 0.006 0.013* 0.034 0.016** (0.01) (0.01) (0.02) (0.01) Kyoto Rat. -0.012 -0.014 -0.010 -0.011 (0.03) (0.03) (0.02) (0.02) UNFCCC Part. 0.004 0.005 0.016 0.010 (0.02) (0.02) (0.02) (0.02) Annex A - 1.332*** - 1.671*** - (0.31) - (0.31) Annex A TimeT -0.019*** -0.021*** -0.023** -0.021*** (0.01) (0.01) (0.01) (0.01) Yrs Kyoto Rat. -0.005 -0.004 -0.010 -0.002 (0.01) (0.01) (0.01) (0.01) Yrs UNFCCC -0.010 -0.007 -0.021 -0.006 (0.01) (0.01) (0.02) (0.01) Post Binding 0.001 0.001 -0.007 -0.011 (0.05) (0.05) (0.03) (0.03) Target and Rat -0.001 -0.001 -0.003 -0.002 (0.01) (0.01) (0.00) (0.01)

n 2232 2232 2108 2232 i 124 124 124 124 R² 0.21 0.36 0.21 0.38

499.25*** 220.78*** Hausman 44.03*** 1.40 P Value: *** - 1% ** - 5% * - 10%

Note: The variables included within this regression are as follows: Dependent variable – CO2 per capita – Amount of CO2 emissions per capita, Independent variables – Constant – Intercept, lGDPpc – log of GDP per capita, lGDPpcsq – log of GDP per capita squared, Time Trend – time trend for 1990-2007, Kyoto Rat. – Ratified the Kyoto Protocol (zero until year of ratification), UNFCCC Part. – Participate in the UNFCCC, (zero until year of commencement), Annez A – Dummy variable for Annex A countries, Annex A TimeT – Time trend for Annex A countries only, Yrs Kyoto Rat. – Years since Kyoto Protocol was ratified (used to distinguish between long-term and recent ratification), Yrs UNFCCC – Years of UNFCCC participation (used to distinguish between long-term and recent participation), Post Binding – Period within which the Kyoto Protocol became binding, Target and Rat – Percentage target of emission cuts if the country ratified the Kyoto Protocol. 3.4 Conclusion

From the humble beginnings of a review of the ‘Environmental Impacts of a North American

Free Trade Agreement’, the EKC relationship has been the source of a plethora of papers. In recent times this literature has increasingly become critical of the EKC, especially with respect to the fragility and limitations of a reduced functional form. Indeed, the literature has noted many possible explanations for a reduced form relationship between GDP per capita and emissions. For CFCs within the current sample (1992-2008) there is a significant policy induced negative relationship for Non-Article 5 countries subject to targets under the

Montreal Protocol. This negative relationship is noted to be up to 0.8% per year above the existing decline for all countries within the same sample. A significantly different level of

CFC consumption between Article 5 and Non-Article 5 countries persists with no indication of an EKC consistent relationship between GDP per capita and CFCs once this policy impact has been allowed for. The confirmation of this result is important as CFCs have had little attention within the EKC literature. In cases where such a relationship has been explained the data underpinning the analysis has been found to be insufficient. Using the dataset of the

Secretariat for the Vienna Convention and the Montreal Protocol has also allowed for a review on whether reductions in CFCs can be attributed to the timing and the levels set within the Montreal Protocol.

Results within sub-section 3.3.5 show that while significant decreases in CFC consumption occurred in the 100% reduction phase of the Non-Article 5 targets, the significant negative decline in CFC consumption between 1992 and 2008 cannot be directly attributed to the targets of the Montreal Protocol. This decline is likely to be driven by the auxiliary explanations for the success of the Montreal Protocol – where upon emission reductions occurred during the reduction phases but were not directly linked to the targets specified. The auxiliary explanations for the success of the policy intervention were the existence of a supportive industry group, pre-existing legislation and commitment in the United States, affordable and available substitutes, as well as acceptance of the underlying scientific (and

Nobel prize winning) explanation of the link between CFCs and ozone depletion.

In the case of CO2 and the Kyoto Protocol, when an EKC consistent relationship did occur it was replaced by evidence that Annex A countries have decreased their per capita CO2 fossil fuel emissions by 0.02% per year once country specific effects were allowed for. As the sample period is between 1990 and 2007, the CO2 analysis is limited to a review of the impacts of a country’s ratification of the Protocol and an incentive for early action to fulfil the targets implied by this action. The lack of a relationship between ratification of the Kyoto

Protocol and the level of targets set for the first phase are likely to be described by the range of issues that have been discussed in sub-section 3.2.3. In addition to doubts over the establishment of binding commitments, the data applied is from before the first phase of binding targets and the analysis is limited to establishing whether early action has been conducted. Such initiatives (as represented in the pre-first phase establishment of the EU trading scheme) are likely to be captured within the statistically significant decrease in CO2 emissions of 0.02% per capita for each year between 1990 and 2007.

Irrespective of the contrasts between the Montreal Protocol and the Kyoto Protocol, and the underlying substitution possibilities for CFCs and CO2, the success of policy should be evaluated after the full period of enforcement has passed. Applicable to the case of Kyoto and

Non-Annex 5 members of the Montreal Protocol, many contingencies can occur and the

97 success of both Protocols are ongoing issues. Further work will be needed to determine the success of the Kyoto Protocol, but initial signs are not encouraging. As significant doubt has also been cast on the existence of an EKC consistent relationship for CO2 and CFCs, future work should focus on the effectiveness of induced policy responses.

Chapter 4 – Water Trading, Diversity and Transaction Costs

4.1 Introduction

Having reviewed the success of an implemented policy, the discussion now focuses on the need for economic theory to bridge divides between applied theory and applied policy to ensure that successful policy-making is more common. The development of market design in economics is one sign that there is a divide that needs to be filled and new techniques have been identified amongst the tools used to do so. Chapter 4 aims at developing a model that can be utilized to assess the success of water trading schemes which are subject to transaction costs which differ based on the distance between trading partners (due to information, search process, communication and negotiation issues). Chapter 5 then applies the model developed in this current chapter by introducing a framework where such transaction costs can be reduced within a policy framework.

These two chapters are also linked due to the issues surrounding the implementation of an

‘offset’ trading program for water quality trading where there is an estimated level of effluent that has been accounted for in setting the cap on water allocations to agricultural firms.

Hence, a focus on non-point source agricultural water use implies a discussion on the level of effluent associated with the use of water in the production of crops. Within chapter 4, there is an assumption that the level of implied effluent remains satisfactory and is maintained under the overall restriction implied by cap on water quantity. Water quality trading and effluent will be discussed in the following to chapters as having a direct connection to each other.

Chapter 5 will introduces a network trading scheme which aims at accounting for changes in effluent across regions via a trading tax and also reducing transaction costs through the introduction of intermediaries.

It is with this in mind that this chapter establishes a unique agent based model of non-point source water trading developed to focus on an agricultural firm’s decision between allocating resources to their traditional production process or by engaging in the water market to sell permits. After establishing a theoretical model designed to simulate the potential impact of transaction costs which vary based on the distance of trading partners, model simulations are conducted using data which represents a segment of the Murray-Darling River basin within south-east Australia. Each agricultural firm is assumed to grow only one crop (either Faba,

Wheat, Barley, Oats or Canola) and water quantity trading occurs on the basis of an ‘offset’ style of trading based on an estimated level of effluent entrenched in every unit of water consumed by a non-point source polluter25. Essential to the decision between growing crops and selling permits, the main determinants of trade has been identified as the amount of profit gained from each activity. The underlying determinants of profit gained from these activities are identified as the crop price and the revenue gained from agricultural activities (subject to

25 Irrespective of these caveats, the model developed in this chapter is based on predicted loads. Chapter 5 will review the issue of monitoring of effluent, while this chapter will focus on the development of a model to review the decision to engage in a water trading system and the impacts of transaction costs. 100 the cost of inputs and production process), as well as the prevailing water price (subject to the water intensity of the crop and the transaction costs borne within the trading process).

The model will incorporate arbitrarily set magnitudes of transaction costs which accumulate based on the distance between the buyer and seller attempting to make a trade. This distance based relationship is applied with the rationale of representing the difficulties of search and information with respect to inter-regional trading. This is one of the key motivations for establishing an agent based model to deal with the issues surrounding permit markets. The second is the effect of constraints on the ability of trade to achieve an efficient allocation of permits in the case where the allocation has been set by other means. And while an auction may be used to allocate environmental tradable permits, in reality, political feasibility often results in some grandfathered allocation. A grandfathered allocation is the distribution of permits based on previous levels of access to the resource or historical emissions. With the assumption of efficient market trading and the Coase theorem, a grandfathered allocation is said not to matter with respect to establishing equilibrium with an efficient distribution.

While the Coase Theorem implies that bargaining with no transaction costs should result in an allocation of resources that are independent of the property rights, this may not hold in reality. For example, Kahneman, Knetsch and Thaler (1990) find that upon employing market experiments focusing on bilateral bargaining, an endowment effect is likely to occur where there is a reluctance to trade a good after ownership is achieved (Kahneman, Knetsch and

Thaler, 1990: 1339 – 1341). Upon highlighting ‘transferable pollution permits’ as a case where the endowment effect is expected to occur, due to the nature of acquisition often being due to ‘historical accident or fortuitous circumstances’, the issue of transaction costs are also

101 raised. The difficulty of separating the influence of endowment effects and transaction costs as a cause of low trading volumes is noted and the experiment designed aims at minimising the effect of transaction costs.

With respect to transaction costs, Stavins (1995) found that the presence of transaction costs implies that the initial distribution of permits may predetermine whether the resulting allocation will be efficient and equitable. With respect to this initial distribution, Stavins

(1995) notes that “unless transaction costs are prohibitive, tradable permits will retain an information advantage over command-and-control approaches and taxes” (Stavins, 1995:

145). The presence of transaction costs takes on added significance in light of concerns of the necessity of constituent appeasement which inevitably involves free distribution (Stavins,

1995: 146). And while it has been observed by Montgomery (1972) that there is an independence between reaching equilibrium and the initial allocation of licences, Stavins

(1995) specifically notes that this is not necessarily the case and that transaction costs imply a reduction in the amount of discretion for policy makers and reduced feasibility of a tradable permit system (Stavins, 1995: 143). Using experiments, Cason and Gangadharan (2003) test the transaction cost model of Stavins (1995). Using two functions for the marginal rate of transaction costs (increasing and decreasing) they concluded that prices and final allocations do deviate from the zero transaction cost competitive equilibrium and that a deadweight loss should be included in the overall cost of a trading program (Cason and Gangadharan, 2003:

163).

After the firm has made the decision to enter the market via the initial auction or allocation process, there are factors occurring in this post-allocation period which determine the overall

102 market's success. A review of out of equilibrium behaviour is important to the implications of grandfathered allocations and successful policy making. The efficient distribution will be found as long as there are no constraints to trade – yet constraints to trade often exist. Hence, the development of an agent based model is focused on the allocation of permits, dis- equilibrium trading and factors which inhibit trading. While the review may be extended to include initial auctions, at this point the focus is on a discussion of what occurs after the initial/early distribution has been settled upon.

Within this chapter, the focus is to build an agent based model which can be applied to review the extent to which transaction costs reduce trading volumes. With no allowance for endowment effects, the results of this chapter include finding that transaction costs are significant and decrease the probability of successful trades. In addition to this, an emergent phenomena of the agent based model is that there is a declining incremental impact for each dollar of transaction cost imposed (as a function of distance). The investigation of the impact and magnitude of transaction costs is important, as while “the magnitudes of transaction costs associated with environmental and natural resource policies are demonstrably important, few studies to date have attempted to actually quantify transaction costs” (Cann et al. 2005: 528).

The diversity within the trading region is also found to be important to the successful trading of permits, as are considerations of crop water intensity, the price at which crops can be sold, and the type of search mechanism employed.

The chapter is structured as follows. Section 4.2 contains a literature review on a water trading schemes (sub-section 4.2.1) and transaction costs (sub-section 4.2.2). Section 4.3 then introduces the agent based model (sub-section 4.3.1) and addresses a range of factors that

103 should be considered upon evaluating the model. Section 4.4 then reviews the results of simulation runs using the model and does this both graphically (sub-section 4.4.1) and by applying multivariate regression analysis (sub-section 4.4.2). Section 4.5 then presents the conclusions of the chapter.

104

4.2 Literature Review

4.2.1 Applied Water Trading Schemes Encouraged by the development of trading programs for a range of pollutants, market based approaches have been encouraged and adopted in the area of water quantity. Within the

United States this can be seen in the sanctioning of trading solutions for water pollution problems by the EPA in the early 1980s and the release of a ‘Policy Statement on Effluent

Trading in Watersheds’ in 1996 (Jarvie & Solomon, 1998: 136). The promotion of water markets has occurred in Australia since the 1970s and individual states started to allow the transferring of water entitlements via market mechanisms as early as the late 1980s (Qureshi, et al. 2009: 1642). And while there has been substantial development in the respective markets within the US and Australia, many cases of successful trading have had a reliance on existing relationships (intra-firm) or the establishment of intermediaries such as water exchanges.

The range of programs with an aim of increasing the water quality within watersheds have had varied success in achieving the aims of reducing effluent and doing so at least cost.

Indeed, the application of tradable permits to policing water quality is proving immensely difficult with many schemes in the United States reporting little or no trade (Morgan and

Wolverton, 2005: 21). With so many cases of limited trading it has become imperative that the limitations and capabilities of these trading systems are examined. While air quality permits have also had some teething problems, water provides an interesting application of the complex incentive problems involved in setting up a trading scheme. Some of the water specific issues include the difficulties in monitoring non-point source pollutants such as

105 agricultural run-off, as well as the historical policies and perceptions which treat water as a free resource with little regulatory constraint.

The issue of managing non-point sources within a water quality trading system has been seen as being imperative by Schary and Fisher-Vanden (2004) as they are likely to provide opportunities for reducing effluent reduction at a lower cost than point sources. And while more opportunities for cost-effective reductions are a key component of a successful trading program, challenges remain with respect to “the achievement of the environmental goal at least cost and without creating new obstacles and steep transaction costs for both regulators and stakeholders” (Schary and Fisher-Vanden, 2004: 281). Stavins (1995) noted that in addition to implementation and policy action advancing beyond the understanding of some fundamental design issues, the “claims made for the cost-effectiveness of tradable-permit systems have often exceeded what can reasonably be anticipated” (Stavins, 1995: 133). It is subsequently recommended that the appraisal of cost effectiveness should occur with applications of comparisons to actual trading programs or ‘reasonably constrained’ theoretical permit programs.

Several authors have noted distinct contradictions between the advantages and disadvantages of water quality trading with respect to non-point source effluent. While Schary and Fisher-

Vanden (2004) discuss the potential gains of cost effective non-point source reductions, they also note that the traditional offset model has had “questionable success in achieving the intended environmental goal at least cost” (Schary and Fisher-Vanden, 2004: 282). The issue of monitoring effluent from non-point sources is key to much of the discussion surrounding effluent trading with respect to agricultural run-off. However, interest in non-point source

106 trading has remained with effluent trading being appropriate in cases where the pollutant is associated with an accumulated load. Woodward (2003) summaries the situation by stating that while effluent trading is “most appropriate from a physical perspective … it is most difficult from a regulatory perspective” (Woodward, 2003: 236). Due to the issue of monitoring it is typical that trading schemes involving non-point sources quantify effluent levels with respect to an estimated effluent level for the associated applications with effluent reductions monitored on the basis of the implementation of ‘best management practices’

(Woodward, 2003: 236).

With respect to the issue of effluent trading and water quantity trading, this chapter’s focus is on water quantity trade between non-point sources with an implied water effluent level used to set the cap based on ‘best management practices’. Efficiency improvements and effluent reductions are not allowed for, as the subsequent model assumes that firms have all applied the best management practices and that there is an insignificant difference in effluent loads between firms growing the same type of crop. This is an important qualification as monitoring and enforcement are essential to the success of non-point source effluent policies.

In point to non-point effluent trading systems the issue of monitoring non-point source effluent typically applies ‘best management practices’ to set the expected runoff for the amount of effluent reduced and a trading ratio between the major types of effluent sources.

Typically these trading ratios are set at a relatively aggregate level. As an example Ribaudo,

Horan & Smith (1999) note that “the ratio at which nonpoint expected runoff allowances can be converted to point-source emissions allowances is 2:1 for the Dillon Reservoir, and 3:1 for cropland and 2:1 for livestock for Tar-Pamlico” (Ribaudo, Horan & Smith, 1999: 55). Central to this issue is the constraint that “current monitoring efforts of nonpoint-source pollution are incapable of attributing changes to water quality to the actions of a specific polluter”

107

(Ribaudo, Horan & Smith, 1999: 20). This is an overbearing concern for policy and will be covered further in Chapter 5. Chapter 5 will review the issue of monitoring effluent, while this chapter will focus on the development of a model to understand the decision to engage in a water trading system where the cap has been set based on the predicted effluent loads and the subsequent impacts of transaction costs.

While the offset approach has prevailed in the case of non-point pollution, there are a range of market structures that have been applied to trading water quality and property rights. The variety of approaches have been categorised by Woodward and Kaiser (2002) as: exchanges, bilateral negotiations, clearing houses, and sole source offsets. Exchanges have a unique market-clearing price and a public forum which assists the trade of permits at the prevailing market price. The successful operation of such a market requires a standardised good and this requirement often leads to water quality trading with the application of trading ratios – which increase the market’s complexity. Bilateral negotiations are common when there is diversity of sellers (as opposed to situations such as exchanges which have a public and impersonal forum) and for cases where non-standardised goods prevail. With a focus on establishing a model based on the decision to buy and sell permits, this chapter focuses on bilateral negotiation as trade is assumed to be driven by a process where agents seek to engage with other agents on a one-to-one basis.26 This approach allows for an individual assessment by the regulator to ensure that environmental standards do not deteriorate. Woodward and Kaiser

(2002) note that even though bilateral negotiations tend to have ‘relatively high’ transaction costs, due to their applicability to thin markets and low initial establishment costs, this structure is commonly applied.

26 Note that Chapter 5 will move away from this focus on bilateral negotiation and look at the potential role of intermediaries on the trading market. 108

Water quality clearinghouses are a typical case of an intermediary within the market which eliminates direct and personal contractual links between buyers and sellers. Within New

South Wales and Victoria, the majority of trades are of a temporary nature through public water exchanges (Qureshi et al. 2009: 1643). And while exchanges such as ‘Watermove’ have been attributed to increases in water trading, wider trading is limited as “the volume of net inter-regional trade (‘trade-out’ minus ‘trade-in’) in seasonal allocations remains small and varies across irrigation districts from year to year” (Qureshi et al. 2009: 1644). The last type of market structure defined by Woodward and Kaiser (2002) are sole-source offsets – these are typically cases where pollution reductions are achieved off site or via a nonstandard means. The advantage of this approach is that it allows for flexibility in reductions, but may not create incentives for pollution reductions on site. These tend to be special cases, which apply outside of the common regulatory framework and are typically one-off situations.

In 2005 Morgan and Wolverton reviewed the progress of ‘Water Quality Trading in the

United States’ on behalf of the National Centre for Environmental Economics within the US

Environmental Protection Agency. Irrespective of the range of market structures available, it was found that out of the 11 offset/trading programs where trading has taken place, “four programs have had only one trade, one program has had two trades, and two programs have had three trades since inception” (Morgan and Wolverton, 2005: 21). This is a definite concern, as either the allocation was optimal from the initial allocation and there were no changes in the region to prompt trading, or significant barriers to trading have prevailed. In contrast to these programs and as a possible point of optimism, there have been a few programs with between 20 and approximately 60 trades, which have been deemed successful.

109

From this mixed history, it seems that water trading is at a difficult stage of development which must be investigated before it is concluded that the limitations of such trading programs outweigh the possible usefulness of such a policy instrument. Indeed, at some point the debate may need to address whether these historical experiences imply that the concept of nonpoint source water trading without intermediary assistance is inherently flawed. The reason why these limitations should be overcome has been effectively described by Grafton et. al. (2009), where upon the authors note that “markets help mitigate water scarcity because they allow users with higher marginal values in use to purchase or lease water rights from those who have lower marginal values and, thereby, increase the aggregate benefits of water use” Grafton, et. al. 2009: 14). Indeed, it is this diversity of marginal values and the revelation of relative water values that this model attempts to capture.

There are physical and institutional characteristics of water trading systems which can account for part of the ineffectiveness of trading programs within the US. Due to the natural flow of water in a watershed, the environmental impacts of water pollution is variable depending upon the point of discharge, the type of discharge and the natural make up of the ecological and hydrological environment. This has subsequently resulted in programs which have been deemed too complex and inherently difficult to operate and monitor. In addition to this the number of participants is constrained to the size of the watershed and thus this may result in a ‘thin’ market with limits on the number of traders available and the possibility that traders can manipulate prices (Woodward, 2003: 236). In the case of non-point source pollution, the establishment of trading systems is complicated by an interspersed discharge of pollutants; an example of this is agricultural runoff of pesticides, fertilizers and salt. Whilst it is a major contributor to water pollution (effluent and salinity based), it also remains largely unregulated due to the difficulty of monitoring nonpoint source pollution. There is also the

110 problem of designing appropriate permit trading programs when damages are associated with accumulated pollutant loads such as nutrients (Woodward, 2003: 236). As a result, trading programs have primarily been of the point-to-point or point-to-nonpoint type, and it has been found that in 2006 there were no programs in the US with a substantial nonpoint-to-nonpoint trading basis (Nguyen et.al. 2006: 12). Due to constraints on monitoring non-point source effluent, point to non-point trading systems tend to restrict the amount of allowances traded based on the expected runoff to be reduced by non-point polluters and applies a trading ratio based on expected effluent differences or regional impact.

An early case of effluent trading was piloted on the Fox River in 1981 based on point-to- point source biochemical oxygen demand (BOD) concentration trading and resulted in only one trade taking place in 1995. Amongst the physical and institutional reasons identified for the lack in trade were regulatory constraints imposed on the market. These regulatory constraints include the prohibition of trades on cost saving grounds, the need to submit trades for approval by the regulatory authority, the prevention of trades from the same facility and the prevention of temporary trades (DEFRA UK Website). In addition to this, uncertainty of property rights, strategic behaviour and prohibitive transaction costs have also been highlighted. With the diversity of trading programs has come a diversity of reasons identified for the lack of trade performed. With a review of the Lake Dillon Trading Program in

Colorado (America’s first point-to-nonpoint trading program) set up in 1984, Woodward

(2003) identified that the reason for a lack in trade was a lack in demand due to an unforeseen decrease in emissions amongst sewage plants (from minor plant modifications) and counterproductive restrictions such as a prohibition of point to point trading and credit banking (Woodward, 2003: 239). No trading took place in the program until 1999 when an expanded development project created a demand for phosphorous credits.

111

The Morgan and Wolverton (2005) review provides a synthesis of the range of reasons identified as to why trades have not occurred. The most prevalent of these reasons is that point source restrictions have simply been met through other means either due to insufficient limits or alternative initiatives. Other reasons highlighted by Morgan and Wolverton (2005) for the lack in trade were the high cost of trades, regulatory obstacles, difficulties in the identification of sellers, uncertainty of trading rules, uncertain weather, small number of potential traders, contention over the issuing of water rights, and misaligned incentives with potential beneficiaries of trading not being the participants facing limits (Morgan and

Wolverton, 2005: 24). As part of a further review of the reasons why trades might not be occurring, Morgan and Wolverton (2005) also reviewed the largest implementation challenges for the same eleven trading programs. The most prevalent challenges were the difficulties in identifying participants and the difficulties of calculating credits for nonpoint source activities. The other challenges highlighted by the review were the setting of trading rules, market and price uncertainties, a lack of regulatory drivers, changes in thinking around water flows/quality, contention over the use of water rights, low levels of participation (when programs were not compulsory), a small market base, and staff shortages (Morgan and

Wolverton, 2005: 25).

112

4.2.2 Transaction Costs Having discussed a range of experiences with respect to water trading schemes and the reason why there is need for analysis, we can now review a major issue that has been identified for the problems that have occurred when trading was put into practice – transaction costs.

Fundamental to the application of an agent based model in this context is the random search process, the distance based transaction costs and the heterogeneity of the environment.

Heterogeneity of the environment in this application focuses on the mixture of firms subject to the trading scheme, as well as their regional dispersion. Accordingly the issue of transaction costs will be discussed in this section, section 4.3.1 will introduce the model, section 4.3.2 will further explain the search mechanism to be used in the agent based model, and section 4.3.3 will then start the application of the model to the agent based simulations by describing the data on which these simulations are based.

Following on from the work of Coase (1937) and Williamson (1985) transaction costs have been identified as a determinant of industrial structure and the establishment of the firm itself.

With respect to environmental policy, transaction costs are a commonly cited reason for market inefficiency and water trading is no different in this respect. Indeed the additional costs borne whilst planning/carrying out a trade will be an essential component of our study of the difficulties in creating a market for permits. The various components which may impose a cost include the difficulties of finding a trading partner, and of verifying the amount of credits involved, as well as the negotiation and enforcement of the trade itself (Abdalla et. al. 2007: 120). Thus there are three potential sources of transaction costs within pollution trading, these being searching and information gathering, bargaining and decision making, as well as monitoring and enforcement (Stavins, 1995: 134). The importance of monitoring and enforcement typically impacts a market with the introduction of uncertain completion of

113 trades as well as the imposition of administration costs. In addition to the physical difficulties with nonpoint source pollutants, difficulties with transaction costs in trading permit programs have been heightened by the complexity of the trading schemes themselves with unclear and complicated rules (especially in the credit certification process and monitoring) (Abdalla et. al. 2007: 120).

The requisite of reducing transaction costs highlights the need to undertake market design to ensure the availability of information, reduce regulatory uncertainty, and avoid regulatory barriers. Some of the design features that have been identified are the pre-approval of trades, brokerage services and the inclusion of future markets. The government/regulator performing the additional role of a broker has been one suggested method of reducing transaction costs

(whilst increasing administration costs). With the aim of providing information on potential buyers and sellers, this aims to reduce the costs associated with searching and information gathering (Policy Research Initiative, 2006: 25). While this review will not directly focus on issues of monitoring and enforcement, the importance of transaction costs in our review is highlighted by expectations that due to a watershed’s physical characteristics, transaction costs attributed with non-point source (NPS) pollution are higher than that involving point source (PS) pollution. Specifically, “the transaction costs of finding a trading partner are higher because NPS are widely distributed across a watershed, and each source can generate only small numbers of credits in comparison with the larger demand of PS credit buyers”

(Abdalla et. al. 2007: 120).

In discussing the establishment of the transaction cost explanation for the existence of firms

[or ‘governance structures’ (Williamson (1975)], Nooteboom (1993) links the size of firms to

114 the amount of transaction costs. While size of the firm will not be directly incorporated into the current model, individual agricultural firms do tend to be small to medium firms with some degree of asset specificity. With the assumption of bounded rationality (as imposed by the current model), stages of contact limit opportunities of search and the “wider sets of contingencies and actions of the larger firm require a greater capacity for the collection and processing of information” (Nooteboom, 1993: 289). Jacobides and Winter (2005) note the importance of ‘vertical scope’ which may drive institutions to abandon expansion across markets in favour for vertical integration due to the potential for hold-ups and opportunistic behavior. Key to understanding the importance of vertical scope are the transactional and capability conditions which determine firm decisions. Upon developing a theoretical framework to explain how capabilities of firms co-evolve with transaction costs and hence determines firm choices. Indeed, Jacobides and Winter (2005) discuss how capabilities of firms will impact upon outsourcing and that extensions of their research may yield insights related to an appeal in Coase (1992) to better understand the ‘institutional structure of production’. And while such insights are out of the scope of this research, the discussion of intermediaries within Chapter 5 is linked to this discussion of outsourcing and the type of market actions that may emerge within an industry.

A key assumption within this review and model is that transaction costs differ depending on distance between the traders and that the market is typically out of equilibrium due to imperfect pricing of water and permit allocation. Previous models have focused on the issue of transaction costs before, such as that of Kohn (1991) where using an arbitrarily set magnitude of transaction costs, their impact on the optimal level of pollution control is reviewed with the magnitude of costs increasing with the level of control. More typically transaction costs are set as a function of exchange, as in Stavins (1993,1995). Within this

115 analysis, transaction costs borne from each exchange are assumed to be dependent upon distance as this is a proxy for time spent communicating, any supplementary communication costs, a search process and related uncertainty. It should be noted that threshold effects and learning have not been incorporated into the model. Modelling transaction costs based on distance is an approach that has been applied by Venkat & Wakeland (2006), however their focus was on quantifying a constraint in the form of “a distance-based transaction cost that traders must pay in order to transport goods to each other” (Venkat & Wakeland, 2006: 2).

With respect to this constraint it is found that in their artificial world key parameters change when the initial allocation of goods is changed and that there is an emergence of trade networks within the economy (Venkat & Wakeland, 2006: 8).

A range of studies have looked into the issue of the level of transaction costs involved in eliminating externalities and found that this is affected by the number and diversity of agents.

One such case is Kozloff et al (1992) who upon reviewing the acquisition of cropping rights and the reduction of non-point source effluent find that cropland retirement programs should focus on watersheds with the greater heterogeneity across physical or economic variables as they should yield the greater cost-effectiveness. Thus the composition and profile of the region matters to the ability of making a successful trade and this is something that will be examined via the introduction of a diversity variable into the agent based model to allow for the relative diversity of the immediate neighbours of each firm within the model. Within the model, transaction costs are calculated per permit, so in addition to distance, the overall cost the trade is subject to is also dependent upon the quantity of permits being traded. This simplifies the analysis, as well as allowing for the difficulties of trying to trade a large amount of permits, due to the potential mismatch with the amounts of permits sought per trade.

116

4.3 Establishing an Agent Based Water Trading Model

4.3.1 Introduction to the Water Trading Model Having established this chapter’s motivation for reviewing the impact of transaction costs on an individual firm’s decision and ability to trade water, the focus now turns to establishing a decision framework which can be implemented in an agent based model. Central to this model is the agricultural firm’s decision between allocating resources to their traditional production process or engaging in the water market to gain or sell permits. Each agricultural firm is assumed to grow only one crop (either Faba beans, Wheat, Barley, Oats or Canola) and water quantity trading occurs on the basis of an ‘offset’ style of trading based on an estimated level of effluent entrenched in every unit of water consumed by a non-point source polluter. Due to monitoring issues related to non-point sources, permits within this market for agricultural firms are for a given quantity of water, linked to a forecasted level of effluent upon which the total cap is set.

The representative firm within the water trading model maximises profit (which is a function of the revenue from crop and permit sales) given auxiliary costs borne in producing these two products. The decision of how the firm shall prioritise its resources and efforts are predominantly based on the price of the crop they produce and the permits held via the expected revenue from these two sources. A natural extension of this decision process is the possibility of a long term predilection toward switching crops or adopting newly developed efficient irrigation technology, given the price of crops, permits, as well as transaction costs.

117

This model’s agricultural firm is assumed to produce goods on a 'produce to order' basis. It is assumed that the firm’s entire crop is eventually bought at the agreed terms based on the exogenous crop price with the decision of allocating resources between selling crops and selling permits being made before the trading/production period. The assumption that all crops will be bought based on the regional price does not hold for water permits. This is because the crop price is assumed to be set at a market clearing price, whereas uncertainty exists within the permit market.

Auctioning implies some intermediary intervention and with no such intervention, trading after the initial stages typically occurs via a process of bilateral negotiation. Hence trading and the related search process are driven by the individual firm’s objective of gaining permits to match the intended seeding and production of the projected crop output using the forecasted amount of irrigation sourced water. This means that it is the buyers who commence the search process to seek opportunities to make trades, partially based on the assumption that they have an incentive to gain the best price based on speculative investment or crop sowing needs. This search process is important as it allows for heterogeneity in the calculation of transaction costs based on distance and a focus on the spatial composition of the firms subject to the trading scheme. Both of these are key factors behind the application of agent based modelling to simulating the behavior defined in the following model.

Within this analysis the trade of water has been restricted to being permanent, as the modelling of the perceived risks or timing related to temporary trades would be prohibitive.

In addition, as our focus is on simulating the impact of restrictions of trade, permanent trades have been given priority as substantial trading of temporary rights assisted by intermediaries

118 has been noted within the Murray Darling basin – this is discussed in section 4.6. Temporary trading is an important consideration as these types of trades have dominated in practice, however within our model the decision of the production process and the requirements for water is made at the start of each individual period (linked to production of a given crop quantity). The impact of weather and soil composition is reflected in the amount of water demanded for irrigation which differs between crops and regions as reflected in the water intensity variable ( ) (which is specified within equation 4.2). Weather has not been directly modelled, due to the assumption that expectations of rainfall are not significantly different from historical trends. Realistically, uncertainty of rainfall will be a major factor in the decision to temporarily buy permits for many firms due to imperfect knowledge and the risks of incorrect forecasts. Hence this review has not included the issue of water rights security and reliability of flows. Nor has the need for any regulatory approvals been included.

Trading occurs at the beginning of the crop growing cycle for winter crops and is sufficient for the whole period based on the assumption above. With sections of the Murray Darling

Basin having gone through a period of drought over the seven years before 2011, there has been an increase in water trading and in practice such exogenous components must be taken into account.

As noted, the firm within this model maximises profit, which is a function of revenue from crop and permit sales, given auxiliary costs borne in producing these two products. This primary objective is defined within equation 4.1 and is dependent upon equations 4.2, 4.3 and

4.4. Crop quantity is determined by the amount of water input ( ) and the interaction with the amount of land available ( ), which is then impacted by a water intensity variable ( ) which allows for the water demands of different crop types. Note that an i

119 subscript makes this water intensity variable also differ across firms based on factors such as irrigation technology, soil type and regional rainfall. Using the exogenous world/regional price of the respective crop ( ), the determination of crop revenue is defined in equation

4.2. Permit revenue, shown in equation 4.3, is simply the permit price multiplied by the change in the amount of permits (i.e. licenses, ) in each period. Equation 4.4 notes the costs associated with the production of crops and the sale of permits. The costs of selling permits includes the transaction costs of making the trade ( ). Note that it has been assumed that the firm is operating with productive efficiency and the necessary irrigation infrastructure also exists. The auxiliary costs associated with the production of crops ( ) are the costs of the water input and productive inputs (such as irrigation technologies, seeds, labour, harvesting equipment, and fertilizer).

(4.1)

(4.2)

(4.3)

(4.4)

(4.5)

Within equation 4.5, , denotes the decision indicator which reflects the firm’s preference between producing crops and selling permits based on the expected revenue gained from each type of activity. Since eventual production and transaction costs are highly uncertain, the decision indicator’s composition is limited to a comparison of expected revenue streams.

With the production costs of crops being stable within the model, expected revenue is

120 assumed to be a reliable proxy for profit as agricultural costs are assumed to be sunk (or heavily committed) at the beginning of the period. Profit from the sale of water permits are assumed to closely match the revenue with the impact of transaction costs impacting the permit price. Using the decision indicator, the respective firms (or agents) within the model then decide on the amount of permits used for growing crops (their primary production activity) and selling permits (their newly developed activity in which they are amateurs). This allows for a direct comparison of the marginal product of water in crop production and the prevailing price of water.

(4.6)

(4.7)

Having identified the decision variable, it is necessary to define how the expectations of revenue are formed. This is captured within equations 4.6 and 4.7, where it can be seen that expectations are formed based on the previous period’s values of the respective variables.

These expected revenue streams have no probability linked to them and there is no learning within the current version of the model. Expected crop revenue is assured with the assumption of the production of goods occurring on a 'produce to order' basis. The expected permit revenue is calculated in a straight-forward manner and there is no learning effect with respect to setting expectations or the operation of the search mechanism. Future work will incorporate a more complex composition of expected revenue and allow for learning within the search process. In addition modification of this decision indicator for an explicit incorporation of costs may arise due to the theoretical context.

121

Within the current configuration, costs are implicitly included within this decision indicator.

With respect to the firm’s calculation of the expected permit revenue, the incorporation of cost in periods other than period one is due to the inclusion of opportunity cost and transaction costs within the representative permit price (minpermp). This representative permit price is the minimum permit price of the successful trades within the period and hence is the market clearing price. The incorporation of cost within the calculation of the expected crop revenue also occurs with the assumption of an efficient market price for crops (with the application within this model of a ‘world’ crop price) under the assumption of productive efficiency. Note that within the expected crop revenue calculation, there is the condition that

cannot surpass the level of crop which the land area of the farm and the water intensity of the crop can sustain.

Equation 4.8 denotes the amount of permits/licenses put towards sale ( ) and equation 4.9 denotes the amount of permits used for growing crops ( ). These amounts are calculated as a ratio of the individual’s overall permit revenue and total expected revenue.

(4.8)

(4.9)

(when = 1) (4.10)

The permit reserve price ( ) denoted in equation 4.10 is the price of water at which the agent is indifferent between growing crops and selling permits. Effectively this is the price at which = 1. This leads to the ultimate decision to trade in equation 4.11, where trade occurs

122 as long as the agent’s permit buy price ( ), set to their home patch’s permit reserve price, is greater than the prospective trading partner’s patch permit reserve price ( ), subject to the overall amount of transaction costs ( ). The overall amount of transaction costs is calculated as the transaction cost level ( ) multiplied by the distance between the agent’s trading patch (farm) and its own patch (farm) ( ). The determinants of emissions include water use, crop water intensity ( ), water delivery system, crop growing technique, fertiliser and insecticide use, as well as the geological and hydrological profile of the farm/region.

However as noted previously within the development of the model, emissions have been incorporated into this model via an estimate based on expected water use and a reduction in the overall cap. Emissions are an output of the firm, but are assumed to be controlled due to the assumption of productive efficiency and ‘best management’ practices.

Trade if (4.11)

Figure 4.1 is a graphical representation of the firm’s decision between growing crops or selling permits and this novel approach allows us to review the discrepancies between crops with varying water intensities. Starting from the top right quadrant, the permit market, it can be seen that the permit price and quantity of water available to the firm is set via the intersection of the marginal cost curve ( and the demand curve ( ). The quantity of water available can then be mapped to the crop market via the bottom right quadrant by way of an intersection with the respective crop’s water intensity curve ( ). The crop market within the bottom left quadrant is simply the interaction of the quantity of crop grown by the farm with the fixed world/regional price of the respective crop ( ). The last quadrant

(top left quadrant) shows the relative price differential between the world/regional price of the respective crop and the permit price ( ). Of direct interest to the decision indicator is

123 the result that the top right and bottom left quadrants show the respective revenue amounts from the production of crops as opposed to the sale of permits. Note that within the permit market there is a transaction cost curve for each individual’s trade ( ), which may effectively inflate their permit reserve price above the market clearing price and cause the trading of permits to be reduced or possibly even cease completely.

124

Figure 4.1 – Graphical Representation – Decision between Growing Crops and Selling Permits

MCit

Pet

125

4.3.2 Search Mechanism Due to assumptions that the establishment of water trading schemes begins with a grandfathered allocation and with no intermediary structures, a review of how traders are expected to find a trading partner is brought to the forefront. In essence, the issue dealt with here is the modelling of a procurement process with transaction costs being dependent upon the distance between traders as a proxy for computational and communication complexities.

Bilateral negotiations of this type have been associated with low establishment costs, suitability to thin markets, the ability to establish incentives for self-regulation and relatively high transaction costs (Woodward & Kaiser (2002: 377).

The complexities of bilateral negotiations are assumed to exist based on previous reviews of trading systems and the profile of potential participants being sole proprietors who know their primary business very well, but for whom trading permits is a foreign concept (especially without the use of intermediaries). The intuition here is that trading programs have worked in stimulating trades in the case of pollutants such as sulphur dioxide (SO2) and carbon dioxide

(CO2) due to their application to medium to large scale firms who have the necessary experience and the availability of resources (such as lawyers or compliance auditors). These large scale firms also often have the flexibility to adapt to the policy, while smaller scale firms tend to be exempt (e.g. cap on firm size in EU Carbon Emission Trading Scheme). As noted by Nooteboom (1993), “the demand for greater capacity of information processing in the larger firm is met by a greater ability to identify, collect and absorb relevant external information, because of the possibility to employ specialized staff in different functional areas” (Nooteboom, 1993: 289).

126

In this model, the search process for trading partners occurs via a random search of one of the neighbouring farms or via a random search of any of the farms across the four regions. The search mechanism itself is driven by an individual firm’s objective of gaining permits to match the intended seeding and thus the projected crop output. Hence potential sellers seek trades based on the respective reserve price based on a search mechanism which only allows for one attempt per trading period. For the sake of simplicity only one attempt at trading in each period has been allowed (rather than allowing multiple bids), with the distinction implying a certain level of ability or learning. This is another area where this model can be extended; however modelling on this basis involves differing levels of intelligence. A related issue also not modelled here is the possibility of potential traders becoming impatient or disenchanted with the permit market in cases where repeated attempts at trade fail. The endowment effect discussed by Kahneman, Knetsch and Thaler (1990) has also not been incorporated into the model.

127

4.3.3 Baseline Data The Murray Darling Basin spans 1,059,000 square kilometres (14% of Australia's land area) and 84% of this land has been identified as being owned by businesses operating within the agricultural industry (ABS, 2008: 1). As this river system is a vital part of Australia’s agricultural industry, the Murray Darling Basin is the obvious choice to apply a model of agricultural firms within an Australian context. In addition to this, due to low average and intermittent rainfall and general water scarcity issues, irrigation and water trading have prevailed in large parts of the river system. Qureshi et al (2009) note that the promotion of water markets has occurred in Australia since the 1970s and that some states have allowed the transferring of water entitlements using markets since the late 1980s (Qureshi et al. 2009:

1642). Previously water entitlements had only been tied to land use and transferability was only achieved through the sale of land.

In 2007-08 the volume of permanent trades within the New South Wales irrigation corporations (Murray Irrigation Ltd., Murrumbidgee Irrigation Ltd., Coleambally Irrigation

Corp. Ltd., Jemalong Irrigation Corp. Ltd., and Western Murray Irrigation Ltd.) was

119,783ML across a total of 193 trades (National Water Commission, 2008: 94). It should be noted that these trades dwarf the volume of entitlements transferred between of the respective irrigation districts with a total of 2277ML traded during an in-determinant number of trades

(National Water Commission, 2008: 96). This relatively small volume of transfers out of the region may partially be attributed to the irrigation corporations charging exit fees up to as high as 80% of the market value of the entitlement (ACCC, 2006: 45). The importance of distinguishing between permanent and temporary trades has been identified before and is a common feature when discussing trading within the Murray Darling basin. Peterson et al

(2004) notes that while annual permanent trades are commonly low (with less than 1% of

128 total entitlements identified), temporary trades have emerged at levels as high as 10 – 20% of allocations. A range of explanations have been used to explain the differences within the trade of permanent and temporary entitlements. Transaction costs, endowment effects, skeptism or the insecurity of property rights and even gaming have been mentioned in the literature.

With respect to the overall amount of trades, it has been noted that the volume of trade has increased by 50% from 2004-05 to 2007-08 (Grafton et. al. 2009: 18). This has been attributed to drought conditions (which persisted for over seven years) and the existing separation of allowances into categories of high security and general security. Unprecedented low allocations to those holding high security entitlements have meant that “to make up the shortfall those irrigators with high marginal values of water have entered the water market to secure water, that, in the past, they would have received as seasonal allocations” (Grafton et. al. 2009: 18).

The baseline data used within the model has been sourced from two key sources which have been combined to create a hypothetical representation of the Murray Darling basin for four important natural resource management (NRM) regions and five crops. Based on data from the Australian Bureau of Statistics (ABS)27 the number of firms growing each crop and the associated average hectares per farm has been set for each of the five crops in each region.

The regions include Lower Murray Darling (LMD), Murray (Murr), Murrumbidgee

(Murrum), and Lachlan (Lac) NRM regions and the selection of crops are Faba beans, Wheat,

Barley, Oats and Canola. Example data is set out in Table 4.1 and compared to key ABS data

27 Australian Bureau of Statistics (2010a) ‘71210DO001_200809 - Agricultural Commodities, Australia, 2008- 09’. Australian Bureau of Statistics (2010b) ‘46180DO001_200809 - Water Use on Australian Farms, 2008-09’. 129 with the national distribution of NRM regions mapped within Figure 4.2. The regions and crops focused on have been selected on the basis of forming a scenario with the most comprehensive intersection of the ABS statistics and the New South Wales Department of

Industry and Investment Gross Margin Budgets for a range of winter crops using irrigation technologies [refer to NSW Dept of Industry and Investment references (2010) to (2010c)].

Figure 4.2 – Natural Resource Management Regions in Australia

Sourced from http://www.environment.gov.au/biodiversity/threatened/nrm-regions-map.html

130

Table 4.1 – Individual Farm Baseline Scenario and ABS data No. Hectare ML/ha t/ha ML/t Crop Initial ABS ABS Firms per Firm Price/t Water Data - Data - within Price/ML ML/ha No. model Firms LMD - Faba 1 6.77 4.5 5.0 0.9 320.0 16.1 5.4 1 LMD - Wheat 6 70.70 2.5 3.5 0.7 175.0 16.1 5.4 6 LMD - Barley 2 36.75 2.5 4.0 0.6 160.0 16.1 5.4 2 LMD - Oats 1 10.27 2.5 3.0 0.8 150.0 16.1 5.4 1 LMD - Canola 1 33.31 2.5 1.5 1.7 400.0 16.1 5.4 1 Murray - Faba 1 6.77 4.5 5.0 0.9 320.0 16.1 2.7 1 Murray - Wheat 123 70.70 2.5 3.5 0.7 175.0 16.1 2.7 123 Murray - Barley 60 36.75 2.5 4.0 0.6 160.0 16.1 2.7 60 Murray - Oats 54 10.27 2.5 3.0 0.8 150.0 16.1 2.7 55 Murray - Canola 44 33.31 2.5 1.5 1.7 400.0 16.1 2.7 45 Murrum - Faba 2 14.84 4.5 5.0 0.9 320.0 28.9 4.5 2 Murrum - Wheat 266 109.69 2.0 4.0 0.5 175.0 28.9 4.5 266 Murrum - Barley 155 50.38 2.5 4.5 0.6 160.0 28.9 4.5 155 Murrum - Oats 162 18.01 2.5 3.0 0.8 150.0 28.9 4.5 162 Murrum - Canola 83 48.51 3.5 3.0 1.2 400.0 28.9 4.5 83 Lachlan - Faba 1 14.84 4.5 5.0 0.9 320.0 19.7 3.7 1 Lachlan - Wheat 142 68.52 4.0 5.0 0.8 175.0 19.7 3.7 142 Lachlan - Barley 68 32.30 2.5 4.0 0.6 160.0 19.7 3.7 68 Lachlan - Oats 89 8.60 2.5 3.0 0.8 150.0 19.7 3.7 90 Lachlan - Canola 35 21.55 3.0 2.5 1.2 400.0 19.7 3.7 36

As the data used to build a workable agent based model of agricultural behaviour has been

sourced at the Natural Resource Management Region level it is acknowledged that the

specific numerical results of this study are only representative of the region and industries

analysed. This is reflected in table 4.1 with differing estimates of the average water used

(ML/ha) within the region. The current model estimates tends to under-estimate the ABS

average due to allowances for different crops and the assumption of productive efficiency

with no endogenous impact from rainfall that is different to average patterns. Upon

comparing Figure 4.2 to Figure 4.3 it can be seen that a substantial component of the area

reviewed across these four NRM regions has had between 5 and 10% trading intensity28

during the 2007-2008 period.

Figure 4.3 – Water Entitlement Trading (2007-08)

Sourced from National Water Commission (2008)

28 Trading intensity has been defined as the total volume of trades (within, out and into the region) as a percentage of the overall allocation of water. 4.3.4 Decision Variable The decision variable drives much of the behavior within the model and allows for complete heterogeneity between traders as the decision variable is updated each period for every agent based on the prevailing individual variable values. Defined in equation 4.5 as the relative difference between expected income, the determination of the variable is driven by the world crop price, water intensity, prevailing water price and the upper limit of full production capacity. Based on the prevailing water price in period zero, and shown in figure 4.4, faba bean growers in LMD have an initial decision variable score of 22.15 implying that they prefer growing crops than selling permits by a factor of approximately 22:1. This means that for faba bean growers to be indifferent between growing crops and selling all of their permits, the price of water needs to increase from $16.05/ML to $350/ML. Figures 4.4 to 4.8 set out the decision variable scores for all regions and all crops. Upon reviewing the graphs a higher price when the decision score is set to one (Price Dec=1) should be interpreted as higher level of preference towards growing that crop. Regional differences persist as do mixed results upon reviewing water intensity and crop price. With respect to water intensity, Table 4.1 shows that wheat grown in the Murrumbidgee uses the least amount of water amongst the regions and crops reviewed, yet it does not have the highest decision variable score. Overall,

Oat producers are identified as the firms where there is the largest potential preference towards selling part of their water entitlements. This is driven by the lower price of oats on the world market. Faba beans tend to be the most likely to continue producing at a high level

(subject to regional variations), with Canola in the Lachlan region also likely to prevail.

133

Figure 4.4 – Faba – Permit Price and Decision Variable

400 25

350 20

300

250 15 Price

200 Price Dec=1 $/ML Decision 150 10 Decision Score Decision Decision=1 100 5 50

0 0 LMD Murray Murrum Lachlan

Note: Price Dec = 1 and Decision = 1 is the case where an agent would be indifferent between the sources of income since costs have been sunk at the beginning of the period in question. Figure 4.5 – Wheat – Permit Price and Decision Variable

400 25

350 20

300

250 15 Price

200 Price Dec=1 $/ML Decision 150 10 Decision Score Decision=1 100 5 50

0 0 LMD Murray Murrum Lachlan

Note: Price Dec = 1 and Decision = 1 is the case where an agent would be indifferent between the sources of income since costs have been sunk at the beginning of the period in question.

134

Figure 4.6 – Barley – Permit Price and Decision Variable

400 25

350 20

300

250 15 Price

200 Price Dec=1 $/ML Decision 150 10 Decision Score Decision Decision=1 100 5 50

0 0 LMD Murray Murrum Lachlan

Note: Price Dec = 1 and Decision = 1 is the case where an agent would be indifferent between the sources of income since costs have been sunk at the beginning of the period in question.

Figure 4.7 – Oats – Permit Price and Decision Variable

400 25

350 20

300

250 15 Price

200 Price Dec=1 $/ML Decision 150 10 Decision Score Decision Decision=1 100 5 50

0 0 LMD Murray Murrum Lachlan

Note: Price Dec = 1 and Decision = 1 is the case where an agent would be indifferent between the sources of income since costs have been sunk at the beginning of the period in question.

135

Figure 4.8 – Canola – Permit Price and Decision Variable

400 25

350 20

300

250 15 Price

200 Price Dec=1 $/ML Decision 150 10 Decision Score Decision=1 100 5 50

0 0 LMD Murray Murrum Lachlan

Note: Price Dec = 1 and Decision = 1 is the case where an agent would be indifferent between the sources of income since costs have been sunk at the beginning of the period in question.

136

4.3.5 Agent Based Modelling The development of the underlying methodology of computational models with multiple interacting agents has been attributed to von Neumann’s work from the 1940s on cellular automata. In the 1970s Thomas Shelling developed an agent based model (ABM) using pennies and dimes on a chess board and based on certain simple rules found that while neighbours tolerated difference, the population tended to segregate into groups (Schelling,

1971: 143). Axelrod (1984) applied ABM to show why ‘Tit-for-Tat’ strategies can evolve to be the dominant strategy within repeated Prisoner’s Dilemma tournaments.

With respect to land use change and agriculture the use of ABM is relatively new. Most of the literature on ABM trading has been focused on verifying the results of game theoretical and auction based experimental research through the application of multiple simulations.

With respect to land use Matthews et al (2007) provides a review of applications built on the attention received due to the possibility of “replacing transition probabilities or differential equations at one level (e.g. populations) with decision rules of entities at a lower level (i.e. individuals or groups of individuals) along with appropriate feedbacks” (Matthews et al.

2007: 1448). Within the review of land use based ABM applications, Matthews et al. (2007) identified the study of Lansing and Kremer (1993) as the first of such studies. Their focus was on the practical working of the Balinese water temple networks and how these resulted in emergent properties that show “that the structure of water temple networks could have developed through a process of spontaneous self-organisation, rather than deliberate planning by royal engineers or other planners” (Lansing and Kremer, 1993: 112). Key to the result is the application of an agent based focus as it allows them to progressively shift their focus from the individual farmer and field to the level of the subak and water temple.

137

A notable Australian application of ABM to water trading is the development of the Tindall

Aquifer Water Trading Model where the core element of the ABM is “the bidding behavior of horticultural agents within a water market in which agents may buy and sell water access entitlements from each other” (Smajgl et al. 2009: 192). Based on experimental data sought through field work conducted on 20% of the population, Smajgl et al (2009) reviews the impacts of different bidding behavior in an open market, closed market, and a non-market situation. Results show a large variation in results based on the burden of water restrictions between new and established applicants. In addition the results show that as long as there are no issues with respect to monitoring and enforcement, the maintenance of minimum environmental flows through capping water licenses can occur at a higher level of profit over the 22 year simulation period (Smajgl et al. 2009: 199-200). Berger (2001) applied agent based modelling to technology diffusion and resource changes within agriculture using a region within Chile. Differences in diffusion speed and saturation level are compared, with bandwagon effects and pressure for traditional farmers to leave the agricultural industry found. Berger et al (2007) applied agent based simulation techniques to building a model for micro-watersheds with a focus on the impacts of technical change and informal rental markets on household income and water use efficiency. The model consists of multiple sub- models with layers representing: communication networks, land and water transactions, land use, farm holdings, ownership, soil properties, and water flow. The motivation for using multiple layers is the complexity of users and uses, as well as capturing social phenomena and income distribution.

138

Irrespective of the current model’s demand for a spatial framework to analyse distance based transaction costs and the diversity of trading regions, there is also the opportunity of using such a model to enrich theory via the investigation of emergent behavior. The importance of emergent behavior has been seen in other disciplines with the relationship between physics and engineering as well as the relationship between biology, medicine and surgery (Roth,

2002: 1342). Roth (2002: 1342) gives the example of designing a suspension bridge which needs to extend upon an elegant, yet simple, theoretical model consisting of gravity and perfectly rigid beams with consideration of metallurgy, soil mechanics, seismic activity, as well as the forces of wind and water. The relevance to the current situation is that “many questions concerning these complications can’t be answered analytically, but must be explored using physical or computational models” (Roth, 2002: 1342). A consequence of this is that, “engineering is often less elegant than the simple underlying physics, but it allows bridges designed on the same basic model to be built longer and stronger over time, as the complexities and how to deal with them become better understood” (Roth, 2002: 1342). This is the primary advantage of using simulation in reviewing water trading, as it allows analysis of market systems and behaviour that is simply too complex or un-uniformed to analyse using traditional, closed-form techniques. The results within section 4.11 reflect this with the establishment of a declining incremental effect per dollar of distance based transaction costs.

Within the framework applied within this chapter, distance based transaction costs, diversity of traders and spatial arrangements have driven the modeler towards agent based modeling.

139

4.3.6 A Summary of the Agent Based Model An agent based model has been applied in this case as it facilitates the consideration of some aspects typically ignored in analytical models, including variability among individuals, spatial heterogeneity and differing individual behaviour upon adapting to changing internal and external environments. While the agent based approach is promising it has been acknowledged that with an increase in the complexity of structure, agent based models tend to be “more difficult to analyse, understand and communicate than traditional analytical models” (Grimm et al. 2006: 116). Typically this is due to a lack of mathematical language, such as that employed in analytical models, and a lack of standard format such as a table of parameter values, typically employed in an empirical analysis. This section will describe the agent based model employed within this chapter in the manner stipulated by Grimm et al.

(2006), followed by a graphical review of simulations in section 4.10, and a review in section

4.11 of the results using a probit estimation of the occurrence of a successful trade and the elasticities of the key variables used within that regression. The Overview, Design concepts, and Details (ODD) protocol stipulated by Grimm et al (2006) and summarised in table 4.2 has been proposed to overcome the difficulties in communicating the design of an agent based model and will be employed to communicate a model still novel within economic literature. Accordingly, this section has been arranged under the six headings shown in the far right column of table 4.2.

Table 4.2 – ODD protocol Purpose Overview State variables Process overview and scheduling Design Concepts Design concepts Initialisation Details Simulation Experiments

140

Purpose

The purpose of the model is to understand how a range of factors affect the success of a tradable permit scheme for water use with an implied level of effluent across five different crops. The level of transaction costs will be modified, as will the type of search mechanism, to highlight any sensititivities within the model. Key issues driving the application to an agent based model are the modelling of transaction costs based on distance between traders, the subsequent importance of the diversity of firms within the trading scheme regions, and the potential for emergent behavior/results to be confirmed. In Chapter Five, this base model will then be used to highlight the benefits of the Network Trading Model.

State variables

Figure 4.11 shows the basic design of the farm and its trader (or farmer). These are represented by a triangle situated within a square and hence these base level variables start with a trader upon a patch which represents the farm they operate. Using water, designated by the white tipped arrows, and productive inputs (such as seeds, fertilisers, land and machinery), designated by the black tipped arrows, the firm has the possibility of producing one of two types of productive outcomes: - excess water via permits sold to other firms and - the primary good of production (faba beans, wheat, barley, oats and canola). This relationship between water and productive inputs in producing crops has been built into the model via the application of NSW Dept of Primary Industries (DPI) Gross Margin Budgets. As defined by the NSW DPI the gross margin budgets are “intended to provide a guide to the relative profitability of similar enterprises and an indication of management operations involved in different enterprises” (NSW Dept. of Primary Industries, 2007). These budgets are calculated using: crop yields for the region, the forecasted commodity price, the current input prices,

141 and technical information provided by district agronomists. As a result the model constrains the ability of crop production depending on the land area available and the suggested amount of water to grow a corresponding amount of crop in that river basin. The Gross Margin

Budget scenarios are specific to each crop and each river basin, allowing for a range of variables, including the costs of: harvesting, irrigation, crop insurance, weed control, sowing and other auxiliary costs.

Interacting with the trader (the triangle in figure 4.11) and patch variables (the square in figure 4.11 – representing the farm area) are auxiliary variables (represented by the arrows) which range from the crop price and the permit price, to the amount of transaction costs. A full list of the variables used can be found in table 4.3, and figures 4.9 to 4.10 show how the basic variables (patches and agents) are visually and analytically arranged to allow for the spatial heterogeneity implied by this spatially explicit model. Figures 4.9 and 4.10 use the following base colours to represent the regions: Lower Murray Darling as shades of green,

Murray as shades of red, Murrumbidgee as shades of blue, and Lachlan as shades of grey.

The distribution of the different crops within that region is represented by dispersions from the base colour. While the numbers of the crop producers is constant based on the representative data employed, the location of these different crops is set randomly per simulation run and this effect is partially captured within the diversity variable. Within figure

4.10 we can see the agents (black triangles) ready to embark on a trade search during period three, with the graphical outputs updated on the software interface so monitoring and preliminary analysis of key variables can be conducted as the simulations are completed.

142

Table 4.3 – List of Model Variables Variables

Patch – LMD – Faba Patch – MURRUM – Canola Corner

Patch – LMD – Wheat Patch – LAC – Faba Permits Bought

Patch – LMD – Barley Patch – LAC – Wheat Permits Sold

Patch – LMD – Oats Patch – LAC – Barley Successful Trade

Patch – LMD – Canola Patch – LAC – Oats Agent Distance from Patch

Patch – MUR – Faba Patch – LAC – Canola Transaction Cost

Patch – MUR – Wheat Decision New Permit Allotment

Patch – MUR – Barley Permits For Sale Agent

Patch – MUR – Oats Permits For Use Hectare

Patch – MUR – Canola Permit Reserve Price Initial Permit Allocation

Patch – MURRUM – Faba Permit Buy Price Initial Permit Price

Patch – MURRUM – Wheat Representative Price Crop Price

Patch – MURRUM – Barley Diversity Initial Crop Output (t)

Patch – MURRUM – Oats Edge Expected Permit Revenue

Expected Crop Revenue

143

Figure 4.9 – NetLogo Program – Time Period Zero – Patches

File Edit Tools Zoom Tabs Help

Interface I 1 I Prnro~o roc I

~ view upda:es I + lr, .q Button Settings... Edit Delete Add · I I normal speed I I I ' Icontinuous 1·1 ~ l ~ ~~. ticl:6 1000 ilci rt:ii:i r~ 1 r~r~~ ri a9ggg~ c ~~ ~! iii~ : lggi !'rlr'll iili: :~ "5 i!~~~ :;l 0

i I 0 "' 0 Time 10 L 0 n TIMo 1n ~ ~ fiiilf Clear --.:

144

Figure 4.10 – NetLogo Program – Time Period Three – Agents

~ view updates + I••"' Button •l I I S>tt;ngs... Delete Add ~- '-""---....J. normal speed Icontinuous l•l 1

Min Permit Price ~- 171 Q. Q. :E

Time Period Successful Trades 3 363 Time 10 Crop Output

377 '------I~ transcost o I c "c Min Permit Price := 163.79355 - o- 0 Time 10 IIperiods+ / 10 Decision Variable 22.1

Num Successful Trades :6 > 1000 c :~ :; :;l 0

1"' -

145

Figure 4.11 – Agent in Focus

146

Process overview and scheduling

Within an agent based model there are a range of individual processes which the agents run independently, such as the movement of traders (to represent the process of searching for a trading partner), reactions to external disturbance events, and the management of other variables (such as the patches which represent the farm the trader operates). These processes have been summarised within the pseudo-code shown within figure 4.12. This pseudo-code outlines the skeleton of the code of the model and notes the order of the process and the stages in which a selection of the state variables are updated. The pseudo-code has been provided for ease of interpretation. Note that the NetLogo code is available from the author upon request.

Figure 4.12 – Pseudo Code of Agent Based Model To Set up • Clear all • Set up the variables and graphical output • Set up plots on program interface To Go • Each agent makes a production/sale decision • Set up and move each agent as part of trading process • Allow any feasible trades to occur and update data accordingly • Move agent back to their patch and update the plots on the interface • Set up next trading period • Run subsequent trading periods by repeating the above ‘To Go’ commands

Design concepts

Emergence: Emergence considers which phenomena actually emerge from individual traits and which phenomena are imposed by the design structure. Within this model, many of the

147 key variables are calculated subject to certain rules and contingencies. The truly emergent phenomena tend to be those which depend upon the randomness within the model, such as the results that are dependent upon the occurrence of trades with a random trading partner based on the respective calculated variables. Hence, even though many of the variables are calculated using the decision-making process described in figure 4.1, the actual transaction of trades depends upon the patch they attempt to trade with and the associated reserve price of that patch calculated using its own respective variable values. Hence key variables that are subject to emergent behavior are those that depend upon the spatial dimension of the model, such as transaction costs and diversity.

Objectives: As an economic actor, the key objective of the trader is to maximise the revenue created from producing their primary productive good (their crop) and from selling any excess permits not expected to be required during the production of the primary good. Since eventual production and transaction costs are highly uncertain, the decision indicator’s composition is limited to a comparison of expected revenue streams. With the production costs of crops being stable within the model, expected revenue is assumed to be a reliable proxy for profit. Profit from the sale of water permits are assumed to closely match the revenue with the impact of transaction costs impacting the permit price. This is reflected within the decision variable specified within equation 4.5 which summaries the decision- making process described in figure 4.1.

Prediction: At the beginning of each trading period the agent needs to use an expected permit price to calculate the decision variable. In the initial period, this is the water price specified within the starting values based on the pre-existing water-pricing scheme. In subsequent periods it is the previous period’s minimum price at which a successful trade took place and hence represents the market clearing price. The crop price is the regional/world crop price and remains constant throughout the period as the agricultural firms and hence the overall

148 agricultural area is seen to be a price taker. Hence the assumption is that the amount of crop produced is too small to influence the crop price based on the regional crop production changes which occur within the model. This process is reflected in equations 4.6 and 4.7.

Sensing: The range of variables that the agent is aware of is quite broad (as shown in table

4.3), however each agent is only aware of the values of their own variable up until the agent actually moves to another patch. Upon moving to the new patch, the agent discovers their potential trading partner’s reserve price and makes a trade accordingly. This information is revealed only if they are the first agent on that patch to make the inquiry. The initialisation of this trade process happens at random in cases where more than one trader moves to the same patch.

Interactions: The only interaction amongst individuals is the movement between patches and the subsequent information revelation discussed in the ‘Sensing’ section. These movements occur either to a neighbour patch in a random direction or to any patch within the grid at random. These movements are representative of the decision making process and the search for suitable trading partners.

Observations: Data is collected at the end of each trading period and exported to another program for graphical review and regression analysis. This only occurs once all traders have returned to their original patch and updated the data according to the trades made during that trading period. The program then updates the graphical interface by turning the patch white and updating the graphical outputs on the software’s interface.

Initialisations

At the beginning of the trading simulation (i.e. period zero) the environment is set up to conform to the four different regions, with the five crops distributed at random amongst the

149 patches in each region. The spatial distribution of the patches are random as the individual farm data sourced is only representative and does not provide any information on the specific spatial distribution of region itself. With each simulation the spatial distribution is varied and the effect of this is partially captured within the diversity variable listed in table 4.3 and applied in the regression analysis to allow represent the diversity changes in each run of the model.

Simulation Experiments

There are two main variables that are modified as part of sensitivity testing and hence requiring separate simulation experiments. The first is varying the level of transaction costs per grid square which separates the agent and their own patch. The cost is set per grid as distance is used as a proxy for the difficulties in searching and information gathering. With the level of transaction costs ranging from $0 to $10 per grid square, a simulation has been run for each unit between these arbitrarily selected extremes. The second variable varied is the trade search mechanism which differs between restricting the search mechanism to a random neighbour patch and expanding the search to any patch on the grid at random. In modifying these two variables we are left with four basic replications of the model given the existence and non-existence of the range of transaction cost values and the two types of search mechanisms. Note that the model has also been extended in Chapter Five using simulations based on differing transaction costs and tax levels.

150

4.4 Analysis of the Model Results

4.4.1 Graphical Analysis Having described the model, it is now time to review the results of the output of the simulations in graphical form based on a series of replications using certain values of key variables. Figure 4.13 to 4.16 show the sensitivity of key variables to different levels of transaction costs ($0, $1, $5, and $10 per patch) under the Neighbour trading process (i.e. trade with a random neighbour patch). Figure 4.17 to 4.18 show the sensitivity of key variables to different levels of transaction costs ($0, $1, $5, and $10 per patch) under the

Random trading process (i.e. trade with a random patch anywhere within the overall grid).

Upon reviewing figures 4.13 to 4.16, we can see that the there is consistency in all the variables, subject to some shocks, the most significant of which can be seen affecting the decision variable. In this case, figure 4.14, the minimum price at which a trade occurred was at a very low price and proved to be a minor shock leading to a slight increase in crop output due to the reduced price of water. Average crop output falls due to speculation on the market and the assumption that some part of their water entitlements may not be used as the agents are attempting to gain revenue on the market. However, a failure to find trading partners and the decision to sow only a certain percentage of available land becomes a negative impact of such speculative decisions.

The initial representative permit price is persistently $16.05 and then varies around the

$100/ML level depending upon the extent of transaction costs. Wheat dominates trading as it is the most common type of crop grown in all regions. Wheat decreases from an average of

151 almost 400 tonnes to between 200 and 300 tonnes in all neighbour trading simulations, with average production in barley and canola staying relatively stable and production of oats decreasing. These trends reflect the decision scores set out in figure 4.4 to 4.8, with oats and wheat having the lower scores and hence being more sensitivity to higher water prices leading to a higher likelihood of becoming indifferent between growing crops and selling water permits.

With further review of the trends in average output we can see that there is a lagged relationship where the amount of output tends to decrease when the permit price has increased, and increase when the permit price has decreased. The decision variable tends to decrease in all cases, due to the permit price changing based on the underlying sales of permits and the level of crop output reacting to the change in the price of a key input, water.

Figures 4.13 to 4.16 show that the number of successful trades tends to decline as the level of transaction costs increase. Upon reviewing figures 4.17 to 4.20 we can see the impact of distance based transaction costs. Figures 4.17 to 4.20 show very similar trends in all key variables, expect for the stronger negative relationship between transaction costs and the amount of trades fulfilled. As well as greater decreases in the average output of crops, high levels of transaction costs tend to have a higher permit price, restrict water trade and add notable speculative selling to the market.

While brief, this graphical review of the results produced by multiple model simulations has proven useful in that it establishes that there is consistency across the results and the simulation scenarios. It is important to confirm that separate runs of the model provide similar results, subject to the underlying variables driving the model. However in addition to

152 a review of these trends, this analysis reviews the results of multiple scenario runs using probit regression which will be employed to establish the influence of all key variables upon the probability of a successful trade taking place. It is within this regression analysis

(contained within section 4.11) that we will explore the differences across regions, the marginal impact of transaction costs, the impact of being on the edge of the trading space, as well as the diversity of respective crop producers within the trading regions.

153

Figure 4.13 – Key Variable Graphical Output – Neighbour Trading – Trans Cost = $0

Number of Successful Trades

300

200

100 No. TradesNo.

0 1 2 3 4 5 6 7 8 9 10 11

Minimum Permit Price

300

200

$/ML 100

0 1 2 3 4 5 6 7 8 9 10 11

Average Crop Output

400

300 200

100 Tonnes/Firm 25 0 1 2 3 4 5 6 7 8 9 10 11 20

15 Average Decision

10 20

Decision Decision Score 5 10

0 Score Decision 0 1 2 3 1 4 2 3 54 5 6 6 7 8 97 10 11 8 9 10 11

Faba Wheat Barley Oats Canola

154

Figure 4.14 – Key Variable Graphical Output – Neighbour Trading – Trans Cost = $1

Number of Successful Trades

300

200

100 No.Trades

0 1 2 3 4 5 6 7 8 9 10 11

Minimum Permit Price

300

200

$/ML 100

0 1 2 3 4 5 6 7 8 9 10 11

Average Crop Output

400

300 200

100 Tonnes/Firm 25 0 1 2 3 4 5 6 7 8 9 10 11 20 15 Average Decision 40 10 30

Decision Decision Score 5 20

0 10 Decision Score Decision 0 1 2 3 1 4 2 3 5 4 5 6 6 7 8 79 10 11 8 9 10 11

Faba Wheat Barley Oats Canola

155

Figure 4.15 – Key Variable Graphical Output – Neighbour Trading – Trans Cost = $5

Number of Successful Trades

300

200

100 No.Trades

0 1 2 3 4 5 6 7 8 9 10 11

Minimum Permit Price

300

200

$/ML 100

0 1 2 3 4 5 6 7 8 9 10 11

Average Crop Output

400

300 200

100 Tonnes/Firm 25 0 1 2 3 4 5 6 7 8 9 10 11 20 15 Average Decision 25 10 20 15

Decision Decision Score 5 10

0 5 Decision Score Decision 0 1 2 3 1 4 2 3 54 5 6 6 7 8 79 10 11 8 9 10 11

Faba Wheat Barley Oats Canola

156

Figure 4.16 – Key Variable Graphical Output – Neighbour Trading – Trans Cost = $10

Number of Successful Trades

300

200

100 No.Trades

0 1 2 3 4 5 6 7 8 9 10 11

Minimum Permit Price

300

200

$/ML 100

0 1 2 3 4 5 6 7 8 9 10 11

Average Crop Output

400

300 200

100 Tonnes/Firm 25 0 1 2 3 4 5 6 7 8 9 10 11 20 15 Average Decision 25 10 20 15

Decision Decision Score 5 10

0 5 Decision Score Decision 0 1 2 3 1 4 2 3 54 5 6 6 7 8 97 10 11 8 9 10 11

Faba Wheat Barley Oats Canola

157

Figure 4.17 – Key Variable Graphical Output – Random Trading – Trans Cost = $0

Number of Successful Trades

300

200

100 No. TradesNo.

0 1 2 3 4 5 6 7 8 9 10 11

Minimum Permit Price

300

200

$/ML 100

0 1 2 3 4 5 6 7 8 9 10 11

Average Crop Output

400

300 200

100 Tonnes/Firm 25 0 1 2 3 4 5 6 7 8 9 10 11 20 15 Average Decision

10 30 20

Decision Decision Score 5 10

0 Score Decision 0 1 2 3 1 4 2 3 54 5 6 6 7 8 97 10 11 8 9 10 11

Faba Wheat Barley Oats Canola

158

Figure 4.18 – Key Variable Graphical Output – Random Trading – Trans Cost = $1

Number of Successful Trades

300

200

100 No.Trades

0 1 2 3 4 5 6 7 8 9 10 11

Minimum Permit Price

300

200

$/ML 100

0 1 2 3 4 5 6 7 8 9 10 11

Average Crop Output

400

300 200

100 Tonnes/Firm 25 0 1 2 3 4 5 6 7 8 9 10 11 20 15 Average Decision

10 30 20

Decision Decision Score 5 10

0 Score Decision 0 1 2 3 1 4 2 3 5 4 5 6 6 7 8 79 10 11 8 9 10 11

Faba Wheat Barley Oats Canola

159

Figure 4.19 – Key Variable Graphical Output – Random Trading – Trans Cost = $5

Number of Successful Trades

300

200

100 No.Trades

0 1 2 3 4 5 6 7 8 9 10 11

Minimum Permit Price

300

200

$/ML 100

0 1 2 3 4 5 6 7 8 9 10 11

Average Crop Output

400

300 200

100 Tonnes/Firm 25 0 1 2 3 4 5 6 7 8 9 10 11 20 15 Average Decision

10 30 20

Decision Decision Score 5 10

0 Score Decision 0 1 2 3 1 4 2 3 54 5 6 6 7 8 97 10 11 8 9 10 11

Faba Wheat Barley Oats Canola

160

Figure 4.20 – Key Variable Graphical Output – Random Trading – Trans Cost = $10

Number of Successful Trades

300

200

100 No.Trades

0 1 2 3 4 5 6 7 8 9 10 11

Minimum Permit Price

300

200

$/ML 100

0 1 2 3 4 5 6 7 8 9 10 11

Average Crop Output

400

300 200

100 Tonnes/Firm 25 0 1 2 3 4 5 6 7 8 9 10 11 20 15 Average Decision

10 30 20

Decision Decision Score 5 10

0 Score Decision 0 1 2 3 1 4 2 3 5 4 5 6 6 7 8 79 10 11 8 9 10 11

Faba Wheat Barley Oats Canola

161

4.4.2 Regression Analysis Due to the unfamiliarity of agent based models and associated difficulties in the communication of their results [as identified by Grimm, V. et al (2006) and discussed in section 4.9] this section utilises probit regressions and the related elasticities to summarise the results and the trends seen within the model. Probit regression analysis has been used as it allows for a consistent framework for reviewing results where the level of the dependent variable may differ by scales that are difficult to comprehend without further information.

The interpretation of the results is simplified by considering how the key variables included in the agent based model impact the probability of a successful trade occurring for each individual firm.

The level of heterogeneity within the model from applying a random search process is also consistent with utilising a regression analysis as the error term is likely to be random. And while the variables within the model are related in that they assist in defining how many permits the agents intend to trade, the overall occurrence of a trade depends on the respective reserve prices of the random trading partner and the level of transactions costs, which differ, based on the distance between these randomly allocated trading partners. In addition, this heterogeneity means that the dependant variable is not directly determined by the variables that are driving behavior within the model and that it is expected that sufficient random variance exists. Correlation between explanatory variables will exist, however these results should be interpreted as being indicative of general trends.

The regression results have been arranged in the following order to allow for a range of simulation scenarios within the model. Firstly, the regression model with no transaction

162 costs and the basic neighbour search mechanism is reviewed in table 4.5. The second table of results (table 4.6) then modifies the regression results by reviewing the model, subject to the alternate random search mechanism. These two tables are then repeated with respect to a range of arbitrarily set transaction costs levels (tables 4.7 and 4.8). In converting the probit results to elasticities, the starting values listed in table 4.4 have been applied. Typically they are the average of the respective indicator for each region, except for the area and crop dummy variables, the time trend (set equal to one and hence making the reference period the first period), the representative permit price variable (set to the average representative permit price across all regions in the standard model), and the level of transaction costs (set equal to the marginal effect per dollar, with a base of $1 for each of the levels). Note that within the next section, transcostpsq = $1 is short hand for the level of transaction costs per square (or patch) being set to $1. The terms edge and corner refer to patches which are on the boundary of the overall grid and allows for the reduced opportunities of trade within the neighbour search mechanism.

163

Table 4.4 – Marginal Effects Base Values X – X – X – X – Variable Variable Definition Units LMD MUR MURR LAC FLMD Intercept for Faba Producers Growers in the LMD DV 1 0 0 0 FMur Intercept for Faba Producers Growers in the Mur DV 0 1 0 0 FMurrum Intercept for Faba Producers Growers in the Murrum DV 0 0 1 0 FLac Intercept for Faba Producers Growers in the Lac DV 0 0 0 1 WLMD Intercept for Wheat Producers in the LMD DV 1 0 0 0 WMur Intercept for Wheat Producers in the Mur DV 0 1 0 0 WMurrum Intercept for Wheat Producers in the Murrum DV 0 0 1 0 WLac Intercept for Wheat Producers in the Lac DV 0 0 0 1 BLMD Intercept for Barley Producers in the LMD DV 1 0 0 0 BMur Intercept for Barley Producers in the Mur DV 0 1 0 0 BMurrum Intercept for Barley Producers in the Murrum DV 0 0 1 0 BLac Intercept for Barley Producers in the Lac DV 0 0 0 1 OLMD Intercept for Oat Producerss in the LMD DV 1 0 0 0 OMur Intercept for Oat Producerss in the Mur DV 0 1 0 0 OMurrum Intercept for Oat Producerss in the Murrum DV 0 0 1 0 OLac Intercept for Oat Producerss in the Lac DV 0 0 0 1 CLMD Intercept for Canola Producers in the LMD DV 1 0 0 0 CMur Intercept for Canola Producers in the Mur DV 0 1 0 0 CMurrum Intercept for Canola Producers in the Murrum DV 0 0 1 0 CLac Intercept for Canola Producers in the Lac DV 0 0 0 1 t_faba Timetrend for Faba Producers Growers Yrs 1 1 1 1 t_wheat Timetrend for Wheat Producers Yrs 1 1 1 1 t_barley Timetrend for Barley Producers Yrs 1 1 1 1 t_oats Timetrend for Oat Producers Yrs 1 1 1 1 t_canola Timetrend for Canola Producers Yrs 1 1 1 1 Output_faba Output Level of Faba t 33.8568 33.8568 74.1991 74.1991 Output_wheat Output Level of Wheat t 247.459 247.459 438.752 342.589 Output_barley Output Level of Barley t 147.009 147.009 226.709 129.182 Output_oats Output Level of Oats t 30.8117 30.8117 54.0289 25.8125 Output_canola Output Level of Canola t 49.9581 49.9581 145.537 53.8652 Decision_faba Decision Variable of Faba Producers α 22.15 22.15 12.32 18.01 Decision_wheat Decision Variable of Wheat Producers α 15.26 15.26 12.13 11.71 Decision_barley Decision Variable of Barley Producers α 15.95 15.95 9.98 12.97 Decision_oats Decision Variable of Oat Producers α 11.21 11.21 6.24 9.12 Decision_canola Decision Variable of Canola Producers α 14.95 14.95 11.88 16.89 Min_Pprice Prevailing Permit Price Aus $ 16.05 16.05 28.85 19.74 Reserve_Pprice Permit Reserve Price Aus $ 5.0414 6.2823 5.4034 6.2409 Diversitypos Positive Diversity Aus $ 28.0084 35.1934 79.4365 36.9729 Diversityneg Negative Diversity Aus $ -29.493 -15.8104 -51.4591 -35.4591 Transcost=$1psq Transaction Cost - $1 per square Aus $ 1 1 1 1 Transcost=$2psq Transaction Cost - $2 per square Aus $ 1 1 1 1 Transcost=$3psq Transaction Cost - $3 per square Aus $ 1 1 1 1 Transcost=$4psq Transaction Cost - $4 per square Aus $ 1 1 1 1 Transcost=$5psq Transaction Cost - $5 per square Aus $ 1 1 1 1 Transcost=$6psq Transaction Cost - $6 per square Aus $ 1 1 1 1 Transcost=$7psq Transaction Cost - $7 per square Aus $ 1 1 1 1 Transcost=$8psq Transaction Cost - $8 per square Aus $ 1 1 1 1 Transcost=$9psq Transaction Cost - $9 per square Aus $ 1 1 1 1 Transcost=$10psq Transaction Cost - $10 per square Aus $ 1 1 1 1 Edge Agent is on the Edge of the Area DV 0 0 0 0 Corner Agent is in the Corner of the Area DV 0 0 0 0

164

Within table 4.5, the probability of a successful trade for each respective agent using the neighbour search method with no transaction costs applied is between 1% and 0.002% depending upon the region. Note that as the region specific calculations are elasticities, the changes in probabilities from these estimates are in percentage terms and not basis points (or percentage points). The basic trend within table 4.5 is that the probability of a successful trade remains relatively constant across time periods, and hence almost all of the time trend variable estimates are insignificant. Output had no significant impact upon the probability of a success of trade; except for barley which has a significant negative relationship. This negative relationship is consistent with expectations as an increase in output will lower the amount of permits that are available for sale. The regional differences in the output estimates reflect the regional differences across the initial decision variable scores shown in figure 4.6.

Upon focusing on the decision variable, there is an indicator of the proportion of permits being put up for sale which is partially based on the price differential for each crop. In this scenario the decision variable is insignificant, and it seems that successful trades tend to be related to the reserve price of the trader – where a 1% increase in the trader’s reserve price leads to a decrease in the probability of a successful trade by between 0.7 and 0.9%. In addition to this variable, successful trade is also driven by the influence of the respective diversity variables which are based on the difference between the neighbourhood29 average reserve price and the agent’s own reserve price. In the formulisation of the diversity variable there is a separation of the positive and negative scores to allow for any directional influence.

The relative diversity within the region was a significant determinant of the probability of a successful trade, irrespective of the representative permit price and the decision variables. An

29 Neighbourhood in this case has been defined as the set of patches which share a border with the respective patch in question. 165 increase of one percent in the positive disparity between the average reserve price of its neighbour and the reserve price of the trader (i.e. having a lower reserve price than the neighbourhood average), led to between a 1.4 and 5 percentage decrease in the respective base probability of a successful trade occurring. This is compared with between a 0.3 and a

1.3 percentage increase in the respective base probability of a successful trade occurring with a one percent decrease in the disparity between the neighbour average reserve price and the reserve price of the trader. This implies that the mixture of potential trading partners within the region has a significant impact upon the success of the overall trading program, due to the impact on the probability of a successful trade occurring irrespective of the given representative permit price and the decision variable (which represents the balance between expected crop and permit revenue).

166

Table 4.5 –Probability of a Successful Trade – Neighbour Search – No Trans Costs (n=156,816) Probit Elasticity – Elasticity – Elasticity – Elasticity – Probit Elasticity – Elasticity – Elasticity – Elasticity – MRM MMUR ML MM MRM MMUR ML MM FLMD -0.4676 -1.6084 t_Barley 0.0124 0.0428 0.0333 0.0546 0.0368 (1.02) (3.59) (0.01) (0.04) (0.03) (0.04) (0.30) FMur 0.2075 0.5549 t_Oats 0.0252* 0.0867* 0.0674* 0.1107* 0.0747* (1.03) (2.72) (0.01) (0.05) (0.04) (0.06) (0.04) FMurrum -0.5775 -2.5353 t_Canola 0.0198* 0.0682 0.0531 0.0871 0.0588 (1.43) (6.24) (0.01) (0.05) (0.04) (0.06) (0.04) FLac 0.1299 0.3848 Output_faba -0.0062 -0.7223 -0.5616 -2.0202 -1.3637 (1.67) (4.92) (0.02) (2.42) (1.87) (6.89) (4.57) WLMD -0.5140*** -1.7680** Output_wheat -0.0001 -0.1057 -0.0822 -0.2391 -0.1260 (0.20) (0.90) (0.00) (0.23) (0.18) (0.53) (0.28) WMurr -0.3489*** -0.9332*** Output_barley -0.0015** -0.7323* -0.5693** -1.4411** -0.5543** (0.10) (0.35) (0.00) (0.40) (0.30) (0.71) (0.28) WMurrum -0.3372*** -1.4803** Output_oats -0.0026 -0.2760 -0.2145 -0.6174 -0.1991 (0.13) (0.63) (0.00) (0.40) (0.31) (0.89) (0.28) WLac -0.3743*** -1.1091*** Output_canola 0.0016 0.2680 0.2084 0.9962 0.2489 (0.11) (0.39) (0.00) (0.31) (0.24) (1.12) (0.29) BLMD -0.4441 -1.528 Decision_faba 0.0050 0.3841 0.2987 0.2727 0.2691 (0.35) (1.40) (0.04) (2.62) (2.03) (1.88) (1.84) BMurr -0.3019** -0.8076* Decision_wheat 0.0035 0.1856 0.1443 0.1883 0.1227 (0.14) (0.42) (0.01) (0.31) (0.24) (0.32) (0.21) BMurrum -0.3805** -1.6704** Decision_barley 0.0052 0.2826 0.2197 0.2257 0.1980 (0.16) (0.73) (0.01) (0.43) (0.33) (0.34) (0.30) BLac -0.2474* -0.7332* Decision_oats 0.0061 0.2364 0.1838 0.1679 0.1657 (0.13) (0.43) (0.01) (0.53) (0.42) (0.39) (0.38) OLMD -0.7675 -2.6400 Decision_canola -0.0010 -0.0507 -0.0394 -0.0514 -0.0494 (0.54) (2.31) (0.01) (0.41) (0.32) (0.42) (0.40) OMurr -0.6512*** -1.7418*** Min_Pprice -0.0003 -0.0149 -0.0116 -0.0342 -0.0158 (0.17) (0.60) (0.00) (0.02) (0.02) (0.05) (0.02) OMurrum -0.5976*** -2.6234*** Reserve_Pprice -0.0003*** -0.0685** -0.0533** -0.0874** -0.0590** (0.17) (0.83) (0.00) (0.03) (0.03) (0.04) (0.03) OLac -0.7485*** -2.2182*** Diversity Pos -0.0144*** -1.3870*** -1.3550*** -5.0201*** -1.5773*** (0.16) (0.64) (0.00) (0.37) (0.30) (0.60) (0.29) CLMD -0.6130 -2.1087 Diversity Neg -0.0058*** 0.5917*** 0.2466*** 1.3174*** 0.6128*** (0.45) (1.86) (0.00) (-0.17) (-0.06) (-0.18) (-0.12) CMurr -0.6966*** -1.8631*** Edge 0.0832** - - - (0.14) (0.58) (0.04) CMurrum -0.8817*** -3.8705*** Corner 0.0053 - - - (0.24) (1.16) (0.21)

CLac -0.6755*** -2.0018*** 3270.03*** (0.16) (0.63) y 0.0008 0.0097 0.00002 0.0040 t_Faba -0.0020 -0.0070 -0.0055 -0.0089 -0.0060 (0.07) (0.26) (0.20) (0.33) (0.22) t_Wheat -0.0081 -0.0280 -0.0218 -0.0357 -0.0241 (0.01) (0.25) (0.02) (0.03) (0.02) P Value: *** - 1% ** - 5% * - 10% With the introduction of a random cross-regional search process in table 4.6, there is a relative decrease in the expected probability of successful trades in all regions, except for the

Murrumbidgee region. In this case, the minimum permit price has a significant but small effect on the probability of trades. The reserve price of sellers is no longer significant, showing the importance of the prevailing permit price, rather than the neighbours likelihood to sell (due to a lower reserve). The decision variables are now significantly different from zero for firms which grow faba, wheat and barley. A 1% increase in the decision score leads to a positive change in the probability of successful of trades which is quite substantial in the case of faba beans. Indeed a 1% increase leads to between a 3% and 7% change in the probability of a successful trade depending on the region. This is a counter intuitive result and may be due to a correlation with changes in the output levels as suggested by the graphical analysis in section 4.10. With respect to output, an increase in the production of oats has been attributed to a decrease in the probability of a successful trade. Note that the interpretation of the diversity variables are not as straight-forward as in the case of the neighbourhood searching scenario. Under a random search process, agents may trade with any other trader in the active environment, subject to no transaction costs. Under the random trading process, the diversity variables do not have such a straight forward interpretation; however their statistical significance and relative decrease in estimate size leads us to the conclusion that the extreme values underpinning the calculation of these indicators may still be impacting upon the ultimate result. In this case, with the extension of trading opportunities, the diversity variables should be cautiously interpreted and it can be concluded that in this case they tend to be picking up outliers with extremely high or low reserve prices. While the introduction of a random cross-regional search mechanism has decreased the probability of successful trades occurring, the introduction of transaction costs (dependent upon the distance between trading partners) into the model is expected to impact the level of successful trades greatly. Table 4.6 – Probability of a Successful Trade – Random Search – No Trans Costs (n=156,816) Probit Elasticity – Elasticity – Elasticity – Elasticity – Probit Elasticity – Elasticity – Elasticity – Elasticity – MRM MMUR ML MM MRM MMUR ML MM FLMD -2.8370*** -10.1269** t_Barley 0.0173* 0.0617* 0.0720* 0.0523* 0.0608* (1.00) (4.76) (0.01) (0.04) (0.04) (0.03) (0.03) FMur -3.0124 -12.5481** t_Oats 0.0026 0.0092 0.0108 0.0078 0.0091 (0.29) (5.26) (0.01) (0.05) (0.06) (0.04) (0.05) FMurrum -2.5310 -7.6643 t_Canola 0.0026 0.0093 0.0108 0.0079 0.0091 (1.89) (5.61) (0.01) (0.04) (0.05) (0.04) (0.04) FLac -2.8054 -9.8667 Output_faba 0.0163 1.9684 2.2971 3.6596 4.2505 (1.81) (6.26) (0.03) (3.69) (4.28) (6.5380) (7.59) WLMD -0.8678*** -3.0976*** Output_wheat -0.0000 -0.0099 -0.0115 -0.0149 -0.0135 (0.20) (1.19) (0.00) (0.25) (0.29) (0.37) (0.34) WMurr -0.8732*** -3.6373*** Output_barley -0.0011 -0.5558 -0.6486 -0.7271 -0.4812 (0.11) (0.74) (0.00) (0.39) (0.43) (0.49) (0.32) WMurrum -0.5113*** -1.5483*** Output_oats -0.0080** -0.8780** -1.0246** -1.3060** -0.7247** (0.14) (0.53) (0.00) (0.45) (0.49) (0.65) (0.35) WLac -0.9623*** -3.3844*** Output_canola 0.0020 0.3612 0.4215 0.8926 0.3837 (0.11) (0.71) (0.00) (0.35) (0.40) (0.84) (0.37) BLMD -0.8676*** -3.0969* Decision_faba 0.0821** 6.4930** 7.5770** 3.0637* 5.2018** (0.34) (1.65) (0.04) (3.11) (3.43) (1.65) (2.65) BMurr -0.8802*** -3.6664*** Decision_wheat 0.0145** 0.7919** 0.9241** 0.5340** 0.5987** (0.14) (0.83) (0.01) (0.39) (0.41) (0.25) (0.28) BMurrum -0.4946*** -1.4976*** Decision_barley 0.0298*** 1.6975*** 1.9809*** 0.9010*** 1.3601*** (0.16) (0.56) (0.01) (0.62) (0.60) (0.29) (0.43) BLac -0.8049*** -2.8310*** Decision_oats 0.0030 0.1217 -0.1420 0.0575 0.0975 (0.13) (0.69) (0.01) (0.57) (0.66) (0.27) (0.46) OLMD -0.4076 -1.4550 Decision_canola 0.0006 0.0326 0.0381 0.0220 0.0363 (0.49) (1.94) (0.01) (0.44) (0.51) (0.30) (0.49) OMurr -0.9531*** -3.9701*** Min_Pprice 0.0012*** 0.0675** 0.0787** 0.1029** 0.0817** (0.16) (0.95) (0.00) (0.03) (0.03) (0.05) (0.04) OMurrum -0.5187*** -1.5706*** Reserve_Pprice -0.002 -0.0442 -0.0516 -0.0375 -0.0436 (0.17) (0.60) (0.00) (0.03) (0.03) (0.02) (0.03) OLac -0.8671*** -3.0496*** Diversity Pos -0.0097*** -0.9668*** -1.4176*** -2.3260*** -1.2574*** (0.15) (0.78) (0.00) (0.26) (0.23) (0.37) (0.22) CLMD -1.1757*** -4.1967* Diversity Neg -0.0041*** 0.4290*** 0.2684*** 0.6350*** 0.5082*** (0.47) (2.37) (0.00) (-0.12) (-0.05) (-0.12) (-0.10) CMurr -0.9419*** -3.9233*** Edge 0.0367 - - - - (0.14) (0.88) (0.04) CMurrum -0.6561*** -1.9866** Corner 0.1714 - - - - (0.26) (0.85) (0.21)

CLac -0.8391*** -2.9513*** 3555.71*** (0.16) (0.79) t_Faba 0.1554** 0.5545* 0.6471* 0.4704* 0.5464** y 0.0005 0.00004 0.0032 0.0006 (0.08) (0.32) (0.35) (0.29) (0.32) t_Wheat 0.0075 0.0268 0.0313 0.0227 0.0264 (0.01) (0.03) (0.03) (0.02) (0.03) 169

The introduction of transaction costs into the basic model with neighbourhood trading, as seen within table 4.7, shows that the basic time trend has consistently become significant across all crop except for faba beans. The relative changes in output are significant across all crops except for faba, with the probability of a successful trade decreasing with an increase in production. Again the reserve price of traders is significant, but has a notably smaller numerical effect. The diversity variables remain significant, showing that diversity in potential trading partners is important to the overall successfulness of trade. In addition to the variables that remain from the first neighbourhood scenario, the introduction of transaction cost variables sees a notable decrease in the overall probability of successful trade – as summarized in table 4.9. Reviewing the transaction cost estimates within Table 4.7 reveals that a dollar increase in cost per trade has a declining incremental effect upon the probability of successful trade. As summarized in figure 4.21, an emergent relationship from the model is that the impact of the initial dollar of each transaction cost level tends to be the strongest with a consistent, convex and declining incremental effect upon the probability of successful trade.

Regional differences in the level of impact exists, with an increased impact of transactions costs on the probability of a successful trade in the Murrumbidgee region – which is the largest region modeled and hence has more traders/observations.

Table 4.7 –Probability of a Successful Trade – Neighbour Search – Trans Costs (n=156,816) Probit LMD MUR MURRUM LAC Probit LMD MUR MURRUM LAC FLMD -0.7605*** -2.8590*** Output_barley -0.0013*** -0.7197*** -0.7087*** -1.2967*** -0.6044*** (0.29) (1.14) (0.00) (0.13) (0.12) (0.35) (0.10) FMur -0.4950* -1.8322* Output_oats -0.0036*** -0.4122*** -0.4060*** -0.8446*** -0.3300*** (0.29) (1.09) (0.00) (0.14) (0.14) (0.29) (0.11) FMurrum -0.9108* -4.0000* Output_canola 0.0020*** -0.3731*** -0.3674*** -1.2700*** -0.3845*** (0.48) (2.11) (0.00) (0.11) (0.10) (0.35) (0.11) FLac -0.9874** -3.5467* Decision_faba 0.0106 0.8820 0.8685 0.5731 0.6853 (0.51) (1.85) (0.01) (0.81) (0.79) (0.54) (0.63) WLMD -0.5753*** -2.1625*** Decision_wheat -0.0002 -0.0130 -0.0128 -0.0121 -0.0095 (0.06) (0.32) (0.00) (0.07) (0.07) (0.07) (0.05) WMurr -0.5503*** -2.0369*** Decision_barley 0.0041** 0.2448** 0.2411** 0.1790** 0.1903** (0.03) (0.16) (0.00) (0.11) (0.10) (0.08) (0.08) WMurrum -0.3432*** -1.5072*** Decision_oats -0.0026 -0.1096 -0.1079 -0.0712 -0.0852 (0.04) (0.20) (0.00) (0.14) (0.14) (0.09) (0.11) WLac -0.4680*** -1.6811*** Decision_canola 0.0008 0.0441 0.0434 0.0409 0.0476 (0.03) (0.15) (0.00) (0.11) (0.11) (0.10) (0.12) BLMD -0.1715* -0.6446** Min_Pprice 0.0001 0.0065 0.0064 0.014 0.0076 (0.10) (0.40) (0.00) (0.01) (0.01) (0.01) (0.01) BMurr -0.2912*** -1.0777*** Reserve_Pprice -0.0003*** -0.0677*** -0.0666*** -0.0790*** -0.0646*** (0.04) (0.17) (0.00) (0.01) (0.01) (0.01) (0.01) BMurrum -0.4136*** -1.8162*** Diversity Pos -0.0154*** -1.6244*** -2.0099*** -5.3825*** -2.0492*** (0.05) (0.23) (0.00) (0.13) (0.10) (0.20) (0.10) BLac -0.2949*** -1.0592*** Diversity Neg -0.0064*** 0.7075*** 0.3735*** 1.4421*** 0.8128*** (0.04) (0.16) (0.00) (-0.06) (-0.02) (-0.06) (-0.04) OLMD -0.6351*** -2.3872*** Transcost=$1psq -0.0837*** -0.3144*** -0.3096*** -0.3673*** -0.3005*** (0.17) (0.76) (0.02) (0.07) (0.07) (0.08) (0.07) OMurr -0.6660*** -2.4651*** Transcost=$2psq -0.0437*** -0.1644*** -0.1619*** -0.1921*** -0.1571*** (0.05) (0.24) (0.01) (0.04) (0.03) (0.04) (0.03) OMurrum -0.4993*** -2.1929*** Transcost=$3psq -0.0361*** -0.1358*** -0.1337*** -0.1587*** -0.1298*** (0.05) (0.25) (0.01) (0.02) (0.02) (0.03) (0.02) OLac -0.6431*** -2.3102*** Transcost=$4psq -0.0284*** -0.1066*** -0.1050*** -0.1246*** -0.1019*** (0.05) (0.21) (0.00) (0.02) (0.02) (0.02) (0.02) CLMD -0.7756*** -2.9153*** Transcost=$5psq -0.0235*** -0.0883*** -0.0870*** -0.1032*** -0.0844*** (0.15) (0.69) (0.00) (0.02) (0.01) (0.02) (0.01) CMurr -0.6566*** -2.4303*** Transcost=$6psq -0.0232*** -0.0874*** -0.0860*** -0.1021*** -0.0835*** (0.04) (0.21) (0.00) (0.01) (0.01) (0.01) (0.01) CMurrum -0.3853*** -1.6921*** Transcost=$7psq -0.0234*** -0.0879*** -0.0866*** -0.1027*** -0.0840*** (0.07) (0.33) (0.00) (0.01) (0.01) (0.01) (0.01) CLac -0.4041*** -1.4514*** Transcost=$8psq -0.0234*** -0.0881*** -0.0868*** -0.1030*** -0.0842*** (0.05) (0.19) (0.00) (0.01) (0.01) (0.01) (0.01) t_Faba 0.0032 0.0122 0.0120 0.0143 0.0117 Transcost=$9psq -0.0214*** -0.0804*** -0.0792*** -0.0940*** -0.0769*** (0.02) (0.08) (0.08) (0.10) (0.08) (0.00) (0.01) (0.01) (0.01) (0.01) t_Wheat 0.0155*** 0.0582*** 0.0573*** 0.0680*** 0.0556*** Transcost=$10psq -0.0206*** -0.0773*** -0.0762*** -0.0904*** -0.0739*** (0.00) (0.01) (0.01) (0.01) (0.01) (0.00) (0.01) (0.01) (0.01) (0.01) t_Barley 0.0111*** 0.0418*** 0.0412*** 0.0488*** 0.0400*** Edge 0.0765*** - - - - (0.00) (0.01) (0.01) (0.01) (0.01) (0.01) t_Oats 0.0232*** 0.0873*** 0.0860*** 0.1020*** 0.0835*** Corner 0.0375 - - - - (0.00) (0.02) (0.02) (0.02) (0.02) (0.07) t_Canola 0.0225*** 0.0844*** 0.0831*** 0.0986*** 0.0807*** (0.00) (0.02) (0.01) (0.02) (0.01) Output_faba 0.0057 0.7204 0.7094 1.8445 1.5087 38698.37*** (0.01) (0.93) (0.91) (2.35) (1.93) Output_wheat -0.0006*** -0.5564*** -0.5479*** -1.1525*** -0.7361*** y 0.0002 0.0003 0.00002 0.0004 (0.00) (0.09) (0.08) (0.17) (0.11) 171

Figure 4.21 – Incremental Negative Dollar Effect on Probability of Successful Trade – Neighbour Search

0.40

0.35

0.30

0.25

0.20

0.15

0.10

0.05

% Change% in ProbabilitySuccof Trade 0.00 1 2 3 4 5 6 7 8 9 10 $ per Patch

LMD Mur Murrum Lac

Note: this figure should be interpreted as showing the negative impact of each incremental dollar of transaction cost on the probability of a successful trade.

From table 4.8, it can be noted that while a few variables have lost statistical significance, the majority of those that we have been focusing upon remain significant. The decision variables are now insignificant, except for the case of barley. The reserve price is now a determinant in the random search case; however its numerical impact has been eroded due to the impact of transaction costs. The diversity variables are again significant, however it must be stressed that under the random search process they are simply highlighting extreme values of the reservation price. Under this search process, the transaction costs are still significant and since these indicators are based on an incremental dollar value, as well as the distance between trading partners, numerically significant too. The same declining incremental effect persists, implying that even at a $1 per square level of transaction costs there is a notable impact upon the probability of a successful trade taking place. As summarized in figure 4.22, an emergent relationship from the model is that the impact of the initial dollar of each transaction cost imposition during the trading process tends to be the strongest with a consistent declining incremental effect upon the probability of successful trade. Regional differences in the level of impact exists, but is now reversed in comparison to the neighbourhood search mechanism. In comparison to the neighbourhood trading scenario, the impact of transaction costs is significantly higher and decrease at a slower rate due to the variability in potential distances allowed for within the random search scenario.

173

Table 4.8 –Probability of a Successful Trade – Random Search – Trans Costs (n=156,816) Probit LMD MUR MURRUM LAC Probit LMD MUR MURRUM LAC FLMD -0.7697*** -4.6097*** Output_barley -0.0003 0.2406 0.2619 0.2876 0.2239 (0.30) (1.85) (0.00) (0.26) (0.28) (0.31) (0.24) FMur -1.0196*** -6.6460*** Output_oats -0.0144*** -2.1107*** -2.2973*** -2.8691*** -1.8725*** (0.30) (2.05) (0.00) (0.43) (0.44) (0.58) (0.36) FMurrum -0.5362 -2.4895 Output_canola 0.0025*** 0.7359*** 0.8010*** 1.6620*** 0.8403*** (0.46) (2.13) (0.00) (0.20) (0.21) (0.44) (0.22) FLac -1.1686*** -7.4115*** Decision_faba 0.0207* 2.7489* 2.9920* 1.1852* 2.3669* (0.46) (2.93) (0.01) (1.63) (1.77) (0.73) (1.43) WLMD -0.8825*** -5.2849*** Decision_wheat 0.0025 0.2317 0.2522 0.1428 0.1883 (0.10) (0.77) (0.00) (0.21) (0.23) (0.13) (0.17) WMurr -0.8590*** -5.5994*** Decision_barley 0.0086*** 0.8257*** 0.8987*** 0.4005*** 0.7110*** (0.04) (0.35) (0.00) (0.31) (0.33) (0.15) (0.26) WMurrum -0.2435*** -1.1303*** Decision_oats 0.0059 0.3934 0.4282 0.1698 0.3389 (0.05) (0.25) (0.01) (0.51) (0.56) (0.22) (0.44) WLac -0.8645*** -5.4824*** Decision_canola 0.0012 0.1077 0.1173 0.0664 0.1289 (0.04) (0.35) (0.00) (0.28) (0.30) (0.17) (0.33) BLMD -0.9388*** -5.6224*** Min_Pprice 0.0001 0.0082 0.0089 0.0114 0.0107 (0.15) (1.12) (0.00) (0.02) (0.02) (0.02) (0.02) BMurr -0.9189*** -5.9896*** Reserve_Pprice -0.0001* -0.0222* -0.0242* -0.0172* -0.0236* (0.06) (0.44) (0.00) (0.01) (0.01) (0.01) (0.01) BMurrum -0.4692*** -2.1782*** Diversity Pos -0.0103*** -1.7322*** -2.3691*** -3.8085*** -2.4216*** (0.07) (0.32) (0.00) (0.14) (0.11) (0.18) (0.11) BLac -0.8890*** -5.6384*** Diversity Neg -0.0065*** 1.1471*** 0.6693*** 1.5515*** 1.4605*** (0.05) (0.41) (0.00) (-0.09) (-0.03) (-0.07) (-0.07) OLMD -0.4567* -2.7352* Transcost=$1psq -0.2381*** -1.4257*** -1.5518*** -1.1052*** -1.5098*** (0.25) (1.63) (0.02) (0.15) (0.13) (0.09) (0.13) OMurr -0.9914*** -6.4623*** Transcost=$2psq -0.1885*** -1.1287*** -1.2285*** -0.8750*** -1.1953*** (0.08) (0.61) (0.01) (0.10) (0.07) (0.05) (0.07) OMurrum -0.4286*** -1.9895*** Transcost=$3psq -0.1843*** -1.1034*** -1.2010*** -0.8554*** -1.1685*** (0.09) (0.40) (0.01) (0.09) (0.06) (0.04) (0.06) OLac -0.8774*** -5.5644*** Transcost=$4psq -0.1891*** -1.1327*** -1.2329*** -0.8781*** -1.1995*** (0.08) (0.54) (0.01) (0.09) (0.05) (0.04) (0.05) CLMD -1.4039*** -8.4079*** Transcost=$5psq -0.1853*** -1.1098*** -1.2080*** -0.8604*** -1.1753*** (0.29) (2.21) (0.00) (0.09) (0.05) (0.04) (0.05) CMurr -1.0422*** -6.7936*** Transcost=$6psq -0.1754*** -1.0505*** -1.1434*** -0.8143*** -1.1124*** (0.06) (0.46) (0.00) (0.08) (0.05) (0.04) (0.04) CMurrum -0.5982*** -2.7772** Transcost=$7psq -0.1670*** -1.0000*** -1.0884*** -0.7752*** -1.0590*** (0.09) (0.42) (0.00) (0.08) (0.04) (0.03) (0.04) CLac -0.8075*** -5.1209*** Transcost=$8psq -0.1616*** -0.9677*** -1.0533*** -0.7502*** -1.0248*** (0.06) (0.44) (0.00) (0.08) (0.04) (0.03) (0.04) t_Faba 0.0537** 0.3213** 0.3497** 0.2491** 0.3402** Transcost=$9psq -0.1533*** -0.9179*** -0.9991*** -0.7115*** -0.9720*** (0.03) (0.16) (0.17) (0.12) (0.17) (0.00) (0.07) (0.04) (0.03) (0.04) t_Wheat 0.0006 0.0034 0.0037 0.0026 0.0036 Transcost=$10psq -0.1452*** -0.8694*** -0.9463*** -0.6740*** -0.9207*** (0.00) (0.02) (0.02) (0.01) (0.02) (0.00) (0.07) (0.04) (0.03) (0.04) t_Barley 0.0059 0.0351 0.0382 0.0272 0.0371 Edge -0.1166*** - - - - (0.00) (0.02) (0.03) (0.02) (0.03) (0.02) t_Oats 0.0034 0.0206 0.0224 0.0159 0.0218 Corner -0.1513 - - - - (0.01) (0.04) (0.05) (0.03) (0.05) (0.11) t_Canola 0.0067 0.0399 0.0434 0.0309 0.0422 (0.01) (0.03) (0.03) (0.02) (0.03) Output_faba -0.0019 -0.3812 -0.4149 -0.6476 -0.8846 59026.90*** (0.01) (1.44) (1.57) (2.46) (3.36) Output_wheat -0.0001 -0.1375 -0.1497 -0.1890 -0.2017 y 2.839e-09 9.553e-11 4.590e-06 3.047e-10 (0.00) (0.16) (0.17) (0.22) (0.23) 174

Figure 4.22 – Incremental Negative Dollar Effect on Probability of Successful Trade – Random Search

1.8

1.6

1.4

1.2

0.8

0.6

0.4

0.2 % % Change Probabilityin of Succ Trade 0 1 2 3 4 5 6 7 8 9 10 $ per Patch

LMD Mur Murrum Lac

Note: this figure should be interpreted as showing the negative impact of each incremental dollar of transaction cost on the probability of a successful trade.

Table 4.9 – Probability of a Successful Trade – Summary Table MRM MMUR ML MM Neigh 0.00080 0.00970 0.00002 0.00400 Random 0.00050 0.00004 0.00320 0.00060 Neigh Trans 0.00020 0.00030 0.00002 0.00040 Random Trans 2.839e-09 9.553e-11 4.590e-06 3.047e-10

4.5 Conclusion

The primary contribution of this chapter is the establishment of a unique theoretical model of bilateral negotiation based water trading, with distance dependant transaction costs and spatial considerations allowing for local diversity amongst trading partners. The establishment of this model is important as it allows for the evaluation of the potential success of trading programs within differing regions, with different distributions of firms growing certain crops. This model notes the development of decision and diversity variables which underlie a unique modelling of individual firm agricultural and water use behaviour. In applying representative data for the Murray-Darling Basin in New South Wales, Australia, it has been confirmed that the mixture of firms matters within the trading system. Thin markets may prevail and constraints on trading are likely to increase concerns on the number of traders with whom trading may occur. Irrespective of the diversity mix, the impact of transaction costs are persistent and will prevent trades from occurring even with respect to the firms that are most likely to trade.

Transaction costs within this review differ based on the distance between the respective trade participants involved in the exchange. This modelling approach has been applied to reflect difficulties in searching and gathering information with respect to conducting inter-regional trading between firms with no pre-existing relationship and under the condition of no intermediary assistance. The incremental impact of transaction costs on the probability of successful trade is another interesting result as it shows that the decline in the probability of successful trades is largest with the first dollar imposed. The results imply that regional trading (as reflected in the neighbourhood trading scenario and figure 4.21) has lower transaction costs, which substantially decreases in a convex manner after the first dollar of

176 cost is imposed. Wider trading across regions (represented by the random trading simulation scenario and figure 4.22) has a substantial impact upon the probability of successful trade, which tends to persist after an initial decrease from the first dollar imposed. The importance of establishing changes in transaction costs is an important contribution as it has been acknowledged that while “the magnitudes of transaction costs associated with environmental and natural resource policies are demonstrably important, few studies to date have attempted to actually quantify transaction costs” (McCann et al. 2005: 528). Typically the measurement of transaction costs is quite difficult, and results tend to be in the form of a percentage of overall costs. Studies such as Stavins (2005) and Montero (1998) typically apply a range of functional forms related to marginal transaction costs. Hence results establishing how transaction costs develop and what types of functional forms persist are important in establishing solutions to minimize such costs.

In addition to results related to the diversity of potential trading partners and transaction costs, this chapter has also established a working model of non-point water trade with interpretations on crop output based on different water prices. As reflected and modelled in the decision variable, the diverse underlying drivers for crop production are captured. These drivers include the crop price, water price and the available land. Within the Murray Darling region of Australia, it is found that the establishment of a higher price of water will tend to impact wheat and oat production most significantly, while the production of faba beans, canola and barley are likely to remain stable.

In recent years there has been an increase in models developed for the purposes of designing tradable permit schemes which reduce the influence of transaction costs. An example of such

177 a model is that of Collentine (2005) which proposes a tradable discharge permit system with three interrelated markets: one to identify marginal abatement costs, a second which is a buyer’s market where permits are priced at the marginal abatement costs set in the initial market, and a third market where permits can be exchanged between other firms (Collentine,

2005: 47-48). This is the focus with which Chapter Five takes, as in addition to these results, this model has been developed with the intention of assisting in the design of markets via the review of a network trading scheme designed to overcome transaction costs and integrate externalities into the market.

178

179

Chapter 5 – Appraisal of a Network Trading Scheme

5.1 Introduction

Having reviewed policy inadequacy in chapter two, policy success in chapter three, and built a model to investigate issues that constrain a market-based policy instrument in chapter four, the focus is now on whether market design can lessen the impacts of transaction costs on a tradable permit scheme. In this chapter, the focus is on the ability of a Network Trading

Scheme to foster trade of water entitlements (set via a cap determined by the expected effluent levels) and review the inclusion of intermediaries (or representative agents). The model developed within chapter 4 will now be altered with traders having a choice between bilateral negotiation and seeking the assistance of an intermediary (or representative agent).

Whether the impact of transaction costs is reduced depends on the successfulness of bilateral negotiation and the fee charged by the intermediary. This fee is endogenous and based on the amount of transaction costs incurred by the representative agent, as well as any inter-regional taxes. The inclusion of inter-regional trading taxes will be aimed at addressing the allocation of property rights, allowing for greater protection of sensitive areas/regions within the river system.

180

In developing the model within chapter four, a discussion of the insufficiency of policies related to non-point sources and a prevailing lack of trade was presented. This chapter uses the same model and as such, the same discussion applies. However, as the focus is now on transaction costs and localised effluent concerns, a discussion on monitoring is now relevant.

During the development of a theory dealing with agricultural runoff as an externality, Griffin and Bromley (1982) indentify three distinct problems associated with agricultural runoff. The first is that runoff removes sediment, nutrients and chemicals – resulting in a loss of resources. The second is that a farmer may mine the soil resource of its nutrients at a rate that is not socially optimal. And lastly, a spatial externality occurs with the runoff of sediment, nutrients and chemicals contaminating water bodies such as rivers, lakes, coastal waters

(surface water) and underground aquifers (ground water). Pollution from runoff typically occurs when there is a nutrient surplus from excessive fertiliser use and high-density livestock operations (Carpenter et al., 1998). Nevertheless, it can also occur via soil runoff and the occurrence of salinity.

The characteristics of this runoff and contamination of surface water via drainage ditches, streams and underground channels has been met by a policy response that has been labelled as ‘slow’ and ‘relatively ineffective’. In addition, while politics has played its part, nonpoint source pollution is by definition “diffuse, prone to discharge in pulses, and difficult to pin on any single pollutant event or source” (Dowd et al. 2008: 152). Upon discussing a policy literature that is ‘generally small’, Dowd et al (2008) note five economic instruments that may be applied to non-point source pollution. These are an input tax, an ambient tax/subsidy, government financial assistance, tradable permits, and liability rules & performance bonds.

181

Due to uncertainty concerning the level of emissions, tradable permit programs such as the

Tar-Pamlico program have set trading ratios between point source and non-point source market participants. In this case, the ratios were set above 1:1 as “non-point source loadings are less predictable over time and space and more random and less reliable than point sources” (Hoag and Hughes, 1997: 256). And while the difficulty of monitoring non-point source effluent has tended to prevail, Stephenson, Norris and Shabman (1998) argue that similar measurement issues occur in the case of point sources as well. Key reasons why non- point monitoring should not prove too difficult is that point sources also tend to have measurement issues (due to rainfall and the need for probability distributions to be applied to

‘measured’ loads), and that incentives may be established so that monitoring is borne by the water user. It is concluded that with the right incentives, a market for nonpoint source credits will stimulate innovation and lower cost methods of measurement to assure regulators of the effluent reductions achieved. Within this model, effluent will be linked to water rights in the same manner as chapter four – a cap set for water use based on the effluent implied from

‘best management’ practices. However, effluent will be partially accounted for, as regional taxes will incorporate a feedback mechanism to directly link observed effluent to the use of water within that region.

In smaller water basins, trading programs with trading ratios between point and nonpoint sources tend to be infeasible. In such cases, trades between nonpoint source polluters may be possible either within or across water basins. In the case of nonpoint and nonpoint source trade, water quality concerns can shift to areas where permits prevail due to the allocation of permits following higher profitability of agriculture types (Dowd et al., 2008). Indeed,

‘pollution hotspots’ may develop with unintended regional effects and environmental concerns.

182

As this chapter also uses the Murray Darling scenario established in chapter four, an important policy issue within the region should also be discussed – i.e. the buyback of water allocations. Irrespective of permanent water trades having been allowed within New South

Wales since 1989 and that intra-firm trades or those facilitated by an intermediary have prevailed, there is a perception that the policy progress within the Murray Darling Basin is disappointing. With a range of Federal and State government policies and investments going into the region, it has been noted irrespective of “all of the effort, (that) there is considerable underachievement as a result of poorly informed investment decisions and persistent institutional weaknesses” (Lee and Ancev, 2009: 20). Amongst the areas where criticism has been made, one notable issue is that of buy-backs. In recent years, Australian governments have purchased water entitlements, however this is likely to be economically inefficient. As noted by Grafton and Hussey (2006), buy-backs suffer from two flaws: the current price of water does not reflect the value of keeping water in non-use, and any purchases are conditional on users undertaking water saving measures. Indeed, the “policy should be changed to ensure the greatest possible amount of water is returned as environmental flows per dollar of expenditure” (Grafton and Hussey, 2006: 23).

Buybacks have occurred due to concerns with environmental flows; however, problems persist with this approach, such as being costly and imprecise. An issue raised by Connell and

Grafton (2008) is that a buyback intended to recover water flow will depend on whether

‘sleepers’ and ‘dozers’ are included in the calculations as un-used entitlements must be accounted for. Indeed, one issue for regulators applying market-based instruments such as trading permit schemes is the loss of control and uncertainty on the effect of the policy. Using

183 the tax-permit hybrid model outlined within this Network Trading Scheme, there is the benefit of targeting environmental quality and environmental flows within certain regions, whilst allowing the market to decide which agricultural firms have the lowest marginal value of permits and hence should give up at least part of their allocations.

Carey, Sunding and Zilberman (2002) review the ‘Transaction costs and trading behaviour in an immature water market’ with the observation that within the Westlands Water market in

California’s Central Valley approximately 75% of trade occurred as an internal transfer within networks of affiliated farms. Their contention was that transaction costs for market transactions provide a strong incentive to trade internally where the transaction costs are lower (Carey et al. 2002: 743). These transaction costs were based on “the ability of a large network to use a ‘representative farm’ to conduct market trades and then redistribute the water through internal trades” (Carey et al. 2002: 748). This ability to trade within a network has been captured within the Network Trading Scheme, with the focus of reducing transaction costs by trying to reduce the number of times the transaction cost is incurred.

Within the network built around the framework specified, the intention is that localised areas use a ‘representative trader’ to cover any surplus needed within the individual trading sectors.

The chapter is structured as follows. Section 5.2 contains a literature review on important issues relevant to this chapter. Sub-section 5.2.1 reviews the literature concerning market design and the need for policy prescriptions to be tested before implementation. Sub-section

5.2.2 looks at the difficulties in establishing a market for a non-uniform (or un-standardised) good. Section 5.3 reviews the experiment that is conducted within this chapter. Sub-section

5.3.1 introduces the Network Trading Scheme and discusses the individual components of the

184 trading scheme. Section 5.3.2 discusses how the permit and tax hybrid system underlying the

Network Trading Scheme has been integrated within the agent based model developed in chapter 4. Section 5.4 reviews regression and graphical results from the revised agent based model. Section 6 then concludes the chapter.

185

5.2 Literature Review

5.2.1 Market Design Market design is vital to economic policy and economic theory in that it has helped economics to be at the forefront of many areas, such as the environment and energy use.

Further to this it has increased the demands and expectations of economic theory by fuelling the need for insights into areas which have either not been focused on or have been overlooked by mainstream theory. The need for such insights has promoted the evolutionary development of market structures and has fuelled the use of new, innovative techniques such as agent based simulation and experimental economics. Using theory and practical experience, agent based simulation allows us to build a microeconomic model which can replicate market wide experiences. Whilst this technique has not gained widespread utilisation, an advantage is that it allows us to review the impact of dynamic actions with the possibility of sensitivity testing.

Thus in addition to reviewing the implementation of policy, a review of the way in which economists view markets may be necessary due to the difficult questions thrown up by market design. In line with this sentiment it has been suggested that while “an economist can carefully craft the problem specification, ignoring irrelevant features, abstracting away details, approximating details and solutions where appropriate ... The AI researcher has no such luxury” (Boutilier et al. 1997: 3). A review of market design should also incorporate or replicate the structures that have developed in the real world. People often do not take part in commerce through auctions and most trade occurs with at least some intermediary involvement. Indeed Rolfe (2012) noted that “the establishment of water markets and property rights often impacts on the design of these institutions, or necessitates the creation of

186 new ones”, albeit at a cost that may be substantial (Rolfe, 2012: 208). These institutions may include the promotion of the use of brokers or increased supply of information, possibly leading to the evolution of new market structures (such as online trading). Challen (2000) notes that transaction costs are a dynamic issue and relate to continuing institutional change and that a case of institutional reforms for regulating water resource will need to be flexible due to imperfect knowledge on the ecology of a river system and the environmental consequences of water use.

Like many other disciplines, economic research and policy making finds itself in a situation where humans can significantly influence their external environment. With regard to economics, the design of economic institutions has been undertaken by a range of actors/interest groups such as entrepreneurs, managers, legislators, regulators, lawyers, judges, politicians and computer programmers (Roth, 2002: 1341). In addition, economists themselves have often been placed in a situation where they have been asked to design an economic institution or market. Indeed Marks (2006) identifies five examples of designed markets which have become common place: simulated stock markets, emission markets, auctions for electro-magnetic spectrum, electricity markets, and on-line e-commerce markets

(Marks, 2006: 1341). LeBaron (2006) which attempts to review how trading actual occurs, notes that the examination of market design and microeconomic structures is suitable as it allows for the production of a large amount of data and sensitivity testing in a heterogeneous and adaptive environment.

It is from developments such as these that Roth (2002) noted “an emerging discipline of design economics, the part of economics intended to further the design and maintenance of

187 markets and other economic institutions” (Roth, 2002: 1341). Whilst traditional economic theory has dealt with closed-form solutions, perfectly rational economic actors and analysis of equilibria outcomes, the need/possibility of economic design has highlighted the usefulness of new avenues of research, such as experimental economics. Upon redesigning the labour market clearinghouse for American physicians, Roth (2002) discovered the only available theory “that directly applied to the medical market were the counterexamples and they all warned that, in more complicated markets, problems could sometimes arise” (Roth,

2002: 1372). The traditional research approach leads to a situation where an economist might be viewed as searching for their dropped car keys under the closest street light (irrespective of its distance from the car), rather than in the darkness around the car. It is from this that we can compare these new research techniques, including simulation, with a torch as “the advantage of using simulation techniques is that they provide us with light where the analytical techniques cast little or none” (Marks, 2006: 1343). Specifically, Marks (2006) highlights four reasons why market design has moved away from the traditional closed-form equilibrium analysis. These being: tractability, out-of-equilibrium behaviour, bounded rationality, and model learning (Marks, 2006: 1354).

In employing an agent based model, this chapter will focus on tractability and out-of- equilibrium behaviour by employing a ‘potentially fruitful computational development’ labelled as Agent-based Computational Economics (ACE) and defined by Tesfatsion (2006) as “the computational study of economic processes modelled as dynamic systems of interacting agents” (Tesfatsion, 2006: 835). Whilst the concept of a Walrasian equilibrium and auctioneer has proved to be a useful theoretical concept, it was never designed to address how trade actually takes place in real-world economies (Tesfatsion, 2006: 834). This has resulted in a focus on ‘procurement processes’ where the consumer/supplier identifies the

188 goods to buy and sell, as well as the volume and price. Specific to our discussion, potential trading partners must be identified, offers must be prepared and evaluated, transaction and payment processing must occur. It is from this that an agent based model is employed to examine the specific design issues behind the success of a tradable permit scheme for water use, with an implied level of effluent across five different crops with a focus on expectation formation and transaction costs. Indeed, there will be extensive areas of potential extension to the model as Tesfatsion (2006) notes “the modeler must now come to grips with challenging issues such as asymmetric information, strategic interaction, expectation formation on the basis of limited information, mutual learning, social norms, transaction costs, externalities, market power, predation, collusion, and the possibility of coordination failure (convergence to a Pareto-dominated equilibrium)” (Tesfatsion, 2006: 835). In addition to an examination of these issues, the development of this agent based model will also have the intent of subsequent synthesis of the benefits of sensitivity analysis and replication within theoretical models. Marks (2006) reiterates this in that it is only “once we understand through analysis how the elements of the phenomenon of concern work together, (that) we can ask the question of how to improve its operation: how better to design it” (Marks, 2006: 1343). Note that while chapter four was aimed at understanding the evolution of distance based transaction costs, this chapter reviews a potential market design that adjusts for the disruptive impact of transaction costs.

There are varieties of market structures that exist in ‘non-designer’ markets for traditional goods and it would be useful for a designer of any new market to examine the historical evolution of ‘non-designer’ markets. There are lessons to be learnt from the adaptations that have evolved in markets such as commodity exchanges, grocery stores, brokers, agents and the creation of online auction markets. Indeed some innovation has occurred in the case of

189 water quality trading programs with Woodward and Kaiser (2002) reviewing four different market structures (including exchanges, bilateral negotiations, clearinghouses, and sole- source offsets) currently used in the United States. Within their review, Woodward & Kaiser

(2002) note that these market structures have developed due to the differing physical nature of the pollutant, polluter characteristics, and regulatory constraints. As a result, there is no one structure that will suit all situations and each structure has its own advantages and disadvantages. Different market structures will suit different pollutants, different polluter profiles, different watersheds, as well as different regulators (Woodward & Kaiser, 2002:

367).

The recent experience of water trading programs having limited trades taking place highlights the importance of an economists’ role in reviewing the design factors contingent in establishing a new market. Indeed, “appreciating the strengths and limitations of alternative market structure is important because, by and large, the form that a pollution market takes does not arise naturally as a consequence of market pressures, but is instead a result of decisions made by agencies and legislators” (Woodward and Kaiser, 2002: 367). Hence, economists should assist policy makers with reliable information to base these decisions on.

It is rare for a market to match the textbook ideal and hence the basic exchange market will not be suitable in all cases. Within chapter four, these provisions were taken into account whilst developing a theoretical model of agricultural firms and their decision between primary production and selling water entitlements.

190

5.2.2 Non-uniform Good Concerning the importance of traditional exchange markets, the uniformity of a good is critical for it to be traded on an exchange as it implies low information costs and little need for revelation of market participants. A decision to treat non-point source related water quality as a uniform good, such as the decision made in the case of suphur dioxide and the

US EPA SO2 trading market, is inappropriate due to the importance of the location of emission source, difficulties in establishing the quality of the good, and subsequent regulatory constraints. Further to this, nonpoint pollutants do not lend themselves to uniformity and overall it would be very difficult to implement successfully (Woodward & Kaiser, 2002: 374-

376).

The most common market structure for a water trading program has been identified by

Woodward and Kaiser (2002) to be bilateral negotiation. The example used to contextualise bilateral negotiation with its diversity of sellers and its variability of the good traded is the

‘market for used cars sold by private parties’. Within such a market, high information, contracting and enforcement costs are unavoidable. Irrespective of these high transaction costs, bilateral negotiation allows for a greater degree of regulatory analysis with the potential for case-by-case assessment of trades. The extent of monitoring is discretionary and there is a trade off with the functioning of the market. In the case of water trading, this trade off has even reached the stage where the need for regulatory approval can be attributed with the failure of a trading scheme. For example, within the Fox River trading scheme, the review process was known to take up to six months for an approval and this is quite unsurprisingly seen as a reason for the lack in generated trades (Woodward & Kaiser, 2002: 376-377).

191

Following the used car market example of Woodward & Kaiser (2002), the situation where a buyer can go to a second-hand car dealer can be effectively treated as a case of an intermediary assisting with trade. The intermediary effectively adds a degree of uniformity and thus decreases contracting and information costs. While an intermediary can take many forms, a water quality clearinghouse “differs from a broker in a bilateral market in that clearinghouses eliminate all contractual or regulatory links between sellers and buyers so that parties interact only with the intermediary” (Woodward & Kaiser, 2002: 377). While transaction costs are reduced using this market structure, there are the costs of establishing and managing the clearinghouse, as well as an efficiency loss since price is no longer equal to marginal cost. Another potential disadvantage of a clearinghouse is that it is only appropriate if the pollutant can be made uniform, is available to a large number of potential traders, and if the regulator/agent assumes the risks associated with the role of an intermediary (Woodward

& Kaiser, 2002: 377-379). Whilst trading programs have had a limit to the amount of structures that have been applied to them, “if properly designed, these programs have the potential to be an important part of water quality control in the nation’s changing regulatory environment” (Woodward & Kaiser, 2002: 380).

Thus, the discussion now focuses on a possible market structure that may provide further direction for policy makers and economists. The eventual structure may be a hybrid of existing market structures as “in practice, two or more market structures may function side by side” (Woodward & Kaiser, 2002: 380). The benefits of implementing tax and non-tax economic instruments has been raised by Grafton and Devlin (1996) who while focusing on rent capture note that complementary aspects may be combined and “result in a policy that focuses on the strengths of each type of instrument while side-stepping their weaknesses”

(Grafton and Devlin, 1996: 283).

192

5.3 Incorporating Network Trade

5.3.1 Network Trade Figure 5.1 shows the basic design of a Network Trading Scheme, which can also be described as a tax-permit hybrid scheme. In an attempt to reduce transaction costs, a representative agent or external intermediary is appointed to assist in the facilitation of trades within their respective trading area. Transaction costs are further reduced with the representative agent making trades with the other representative agents to fulfill any surplus or deficit. The basic idea, other than reducing the costs of searching for trading partners is that making one trade and dividing it between subscriptors will significantly reduce the overall transaction costs incurred, especially if these costs are a function of distance. This process is represented within figure 5.1 by the five nodes within each trading area, four being normal agricultural firms and the central node being the representative agent. These nodes are connected based on potential trading links between the internal firms, and the representative agents being linked to each firm within their trading area as well as the representative agents from the other trading areas.

Implementing such a system is not a completely original idea as the notion of intermediaries and clearinghouses have persisted. However, if economic theory notes that intermediaries are needed to assist trades then designing markets to account for this from the offset is important.

The natural entry of intermediaries may take a notable amount of time and result in trading schemes with little or no trade (situations such as those outlined in chapter 4). One advantage of this specific trading system is that if you base an inter-regional tax system on one or more environmental indictors, there is the ability to have greater influence on the environmental impacts of trading within the system. This is important as while the overall cap is important,

193 the spatial distribution of allocations may also lead to issues within the hydrological and biological system. In addition to a stronger link between the permit system and environmental objectives, this system may also assist in the issue of monitoring effluent and water use. By setting a differential tax upon a region, the regulator has the ability to grant tax concessions to firms that gain exemptions (upon proof of best practice behaviour) and it also provides an incentive for peers to monitor their neighbours’ behavior which may impact adversely or favourably upon the environmental factors on which the tax has been based.

The effect with respect to improved monitoring is intended to be similar to that of ‘group compliance permits’ such as those applied in the case of the Neuse River in North Carolina, as described in Shabman and Stephenson (2007). In the case of the Neuse program, when the group compliance permit is exceeded, then a fee/penalty is imposed to pay for load reductions from outside the group. Compliance was a notable success with only one case of the cap being exceeded (Sharman and Stephenson, 2007). In the Network Trading Scheme, this effect and the incentive for self-monitoring is borne by the granting of tax exemptions as well as periodic revisions to match the quality of water within the region (based on observed effluent changes).

194

Figure 5.1 – Basic Outline of a Network Trading Scheme

The major contribution of this research will be showing that the Network Trading Scheme promotes trade by reducing transaction costs. However, the model also incorporates a tax system to demonstrate the concept and examine whether any impact can be noted. Using the respective kilogram per hectare amount of Nitrogen and Phosphorus net exports for the four

Natural Resource Management regions (Lachlan, Murrumbidgee, Murray and the Lower

Murray Darling), a tax rate differential has been set for the four regions modelled using the

195

Australian Agriculture Assessment. The differential has been set with Lachlan having a relative value of four, with Murrumbidgee having a value of five, and the Murray and Lower

Murray regions having a value of three. As there is uncertainty concerning the necessary level of taxes, this will be tested for sensitivity with values starting at 0.03%, 0.04% and

0.05% respectively, and then being multiplied by successive values from zero to ten in successive simulations. One potential issue that may prevent the usefulness of different taxes between regions are cases where poor environmental quality is the norm and consistent across all areas. Indeed, for the twenty-three river valleys of the Murray Darling Basin a recent independent audit of environmental health found that twenty of these regions were in poor or very poor health (Davies et. al. 2008). With the appropriate level of tax set, this would effectively only promote interregional trading from three regions into any of the other twenty regions.

With respect to these taxes, their usefulness can be highlighted with the emergence of environmental indices such as those used within the 2006 US EPA ‘Wadeable Streams

Assessment: A Collaborative Survey of the Nation's Streams’. Using an index of Biotic

Integrity, the resulting macro-invertebrate index is based on a range of indicators representing biological conditions and aquatic indicators of stress. The development of these indexes and hence setting any subsequent tax differentials will be difficult. As noted by Bellenger and

Herlihy (2009) the facets of importance may be interrelated and highly correlated, this is expected as “each of the metrics represents a different component to macro-invertebrate population health, and that these metrics are likely influenced by similar natural features such as water temperature, pH level, or stream flow” (Bellenger and Herlihy, 2009: 2217). In the case of agricultural firms, the link between such an indicator and chemicals such as phosphorous, nitrogen, and chloride are important. Indeed, Bellenger and Herlihy (2009) note

196 that including such factors into a directional output distance function “could be applied to a variety of environmental policy areas, including conservation site selection, baseline monitoring, and non-market valuation” (Bellenger and Herlihy, 2009: 2222). This is important within this context, as these applied areas of focus (conservation, monitoring and valuation) are exactly those which the inclusion of a tax system seeks to address. Using a permit-tax hybrid also reinforces the notion that monitoring environmental quality needs to occur at a level that makes sense with respect to nonpoint source pollutants/impacts (i.e. the level at which the pollutants/impacts are observable) and incorporates other specific policy objectives into the permit system.

To further emphasise the benefits of applying a Network Trading Scheme, we can look to early work where an important theoretical foundation of a market in licenses was established, such as Montgomery’s paper Markets in Licenses and Efficient Pollution Control Programs.

Specifically it was found that “the market in licenses has an equilibrium which achieves externally given standards of environmental quality at least cost to the regulated industries”

(Montgomery, 1972: 396). This theoretical foundation has been very important for the validation of the use of permits; however a few refinements have been made in recent years.

One of the conditions/features of Montgomery’s theoretical foundation was that polluters will be “required to hold a portfolio of licenses covering all recent monitoring points”

(Montgomery, 1972: 396). This is obviously a heavy requirement, especially in cases where location matters. The need to overcome heavy burdens on the license/permit holder has been acknowledged more recently with Hung and Shaw (2005) proposing the use of a trading ratio system “to specifically incorporate location effect into TDP (tradable discharge permit) trading for non-uniformly mixed pollutants in water” (Hung and Shaw, 2005: 85). The computational requirements are reduced using such a system as it implies that “if a discharger

197 wishes to increase his effluent, he only needs to buy TDPs of the same or upstream zones to offset this increase, and does not need to buy TDPs from all zones that are affected by his effluent” (Hung and Shaw, 2005: 85). A permit-tax hybrid system such as the Network

Trading Scheme may also be used to reduce such complexity while also controlling both uniformly and non-uniformly mixing pollutants. Note that in this chapter, the application of trading ratios is relevant to a discussion of restricting trade between regions.

198

5.3.2 Modelling the Permit and Tax Hybrid Figure 5.2 displays the framework upon which the Network Trading Scheme has been modelled as an agent based model. Using the lowest square as our reference point for an eventual trade with the square furthest to the right, there are two ways that the trade can be performed. Following the black arrow, the trader can trade directly with a trading partner from outside of their trading area, but this trade will incur the full level of transaction costs and inter-regional tax. Alternatively, the trader can follow the grey arrow and trade internally within their own trading region using the assistance of their representative agent (yellow polygon). In this case the transaction costs will be decreased via the intermediation of the representative agent and no tax will be applied. If this trade cannot be met internally, then the trading agent can facilitate an interregional trade subject to a fee, which incorporates the transaction costs and tax which the representative agent incurs (as represented by the green arrow).

In modifying the trading model developed in chapter 4, traders now have a decision between trading without assistance or with assistance. These options are determined at random so that the trading partners in the assisted and unassisted case are not the same. Each agent has a grey and black arrow which reflects the options in figure 5.2 and incorporates the decision between the options based on the revenues offered. This is important as it means that the result is not predetermined and no one option is always preferred. For example, while the use of an intermediary means that the burden of transaction costs and inter-regional taxes are shared, an unassisted trade may still be preferred due to a fortuitous arrangement with a local trading partner. The major revision between the Netlogo code for the Network Trading

Scheme and the code from Chapter Four which models the random search trading process is the process of having three separate traders acting in a similar manner to that within figure

199

5.2. Figure 5.3 shows the pseudo-code which accompanies the Netlogo code that is available from the author upon request.

Figure 5.2 – Modelling the Network Trading Scheme

200

Figure 5.3 – Pseudo Code of Agent Based Model To Set up • Clear all • Set up the variables and graphical output • Set up plots on program interface

To Go • Each agent makes a production/sale decision • Set up and move each black agent as part of trading process • Set up and move each grey agent as part of trading process • If black agent revenue is less than grey revenue, remove black agent • If grey agent revenue is less than black agent revenue (or if grey agent was unsuccessful at making at trade), set agent colour to green and then seek a trade within another region via the representative agent • Allow any feasible trades to occur and update data accordingly • Move agent back to their patch and update the plots on the interface • Set up next trading period • Run subsequent trading periods by repeating the above ‘To Go’ commands

201

5.4 Review of Results

Table 5.2 presents the results of the transaction cost simulation with a notable decrease in the probability of a successful trade in comparison to the no transaction cost simulation and a similar decreasing incremental effect as that found in chapter 4. (Note that table 5.4 compares the probabilities across scenarios and figure 5.6 shows the incremental effect of transaction cost levels.) The decision variables for wheat, barley and canola are now significant and show the expected relationship. This being that as the decision score increases (hence the level of indifference between primary production and selling permits widens – refer to equation 4.5 in chapter four) the probability of a successful trade decreases. Regional differences persist and the largest effect is attributed to faba beans (even though the coefficient estimates are insignificant). While the decision variables coefficient estimate for faba is insignificant, the relationship is numerically high which is to be expected as in chapter four faba bean production was forecasted to remain indifferent at the highest level of water prices. An increase in an agent’s reserve price tends to have a negative impact on trade, as does an increase in the price of permits. The result with respect to the price of permits is difficult to interpret, but it should be noted that this might be due to a relationship between the price of permits and transaction costs and/or the decision variables.

Figure 5.4 summaries the impact of the intermediary on trade under the transaction cost scenario. At the levels of transaction costs reviewed, intermediary trades occur more often and have a tendency to persist as the level increases. With respect to the no transaction cost case, the number of successful trades conducted independently of the Network Trading

Scheme (which involves intermediary assistance) varies between 92 and 127 trades, with intermediaries facilitating between 157 and 244 trades. With the transaction cost level set at

202

$1 per unit of distance, the number of successful trades decreases to between 60 and 98 in the independent case and between 91 and 172 in the assisted case. Figure 5.4 reviews one more level of transaction cost, that being $5 per unit of distance, with 39 trades in the independent case decreasing to 0 in period eleven. The case of intermediary assistance starts with only 6 trades, but then increases to 82 trades in period nine. The aggregate amount of trades conducted via an intermediary coincides with a rate of 2 to 1 in comparison to independent trades, for the zero and $1 transaction cost levels. This coincides with 2373 compared to 1210 trades across the eleven periods in the zero transaction cost case and 1651 compared to 845 trades in the $1 transaction cost case. This ratio increases to 8 to 1 in the $5 transaction cost case with 693 assisted trades being conducted across the eleven periods in comparison to 91 unassisted trades.

203

Table 5.1 –Probability of a Successful Trade – Network Trading – Random Search – No Trans Costs (n=156,816) Probit LMD MUR MURRUM LAC Probit LMD MUR MURRUM LAC FLMD -2.5512* -9.8121 Output_barley 0.0000 0.0219 0.0215 0.0260 0.0177 (1.45) (6.54) (0.00) (0.40) (0.39) (0.48) (0.32) FMur -3.1778** -12.01* Output_oats -0.0035 -0.4168 -0.4096 -0.5625 -0.3202 (1.49) (6.43) (0.00) (0.39) (0.37) (0.52) (0.29) FMurrum -5.7200** -16.9298** Output_canola -0.0029 -0.5596 -0.5499 -1.2545 -0.5533 (2.95) (8.34) (0.00) (0.40) (0.37) (0.86) (0.37) FLac -6.0013* -21.17* Decision_faba 0.0175 1.4913 1.4654 0.6383 1.1120 (3.16) (12.23) (0.05) (3.65) (3.59) (1.64) (2.76) WLMD -0.6816*** -2.6216** Decision_wheat -0.0042 -0.2483 -0.2440 -0.1519 -0.1747 (0.20) (1.23) (0.01) (0.32) (0.31) (0.19) (0.22) WMurr -0.5081*** -1.9203*** Decision_barley -0.0076 -0.4667 -0.4586 -0.2247 -0.3480 (0.09) (0.57) (0.01) (0.49) (0.47) (0.23) (0.35) WMurrum -0.2888*** -0.8548** Decision_oats 0.0168 0.7235 0.7109 0.3099 0.5398 (0.12) (0.41) (0.01) (0.59) (0.56) (0.25) (0.43) WLac -0.5060*** -1.7847*** Decision_canola -0.0159* -0.9118 -0.8960* -0.5576* -0.9447* (0.09) (0.53) (0.01) (0.58) (0.53) (0.32) (0.56) BLMD -0.9297** -3.5756* Min_Pprice 0.0000 -0.0015 -0.0015 -0.0021 -0.0017 (0.33) (1.90) (0.00) (0.02) (0.02) (0.03) (0.02) BMurr -0.5969*** -2.2557*** Reserve_Pprice -0.0001 -0.0319 -0.0313 -0.0245 -0.0292 (0.12) (0.70) (0.00) (0.04) (0.04) (0.03) (0.03) BMurrum -0.5071*** -1.5008*** Diversity Pos -0.0094*** -1.0167*** -1.2554*** -2.2191*** -1.2308*** (0.14) (0.53) (0.00) (0.31) (0.28) (0.45) (0.27) BLac -0.6312*** -2.2264*** Diversity Neg -0.0045*** 0.5105*** 0.2689*** 0.6855*** 0.5629*** (0.11) (0.65) (0.00) (-0.17) (-0.07) (-0.16) (-0.13) OLMD -0.4471 -1.7196 Transcost=$1psq (0.49) (2.14) OMurr -0.6460*** -2.4413*** Transcost=$2psq (0.14) (0.78) OMurrum -0.6511*** -1.9271*** Transcost=$3psq (0.15) (0.61) OLac -0.9496*** -3.3494*** Transcost=$4psq (0.13) (0.89) CLMD -1.2800** -4.9228 Transcost=$5psq (0.57) (3.18) CMurr -0.7608*** -2.8752*** Transcost=$6psq (0.13) (0.84) CMurrum -0.0480 -0.1422 Transcost=$7psq (0.23) (0.67) CLac -0.5119*** -1.8056*** Transcost=$8psq (0.14) (0.66) t_Faba -0.1043 -0.4012 -0.3942 -0.3088 -0.3679 Transcost=$9psq (0.07) (0.31) (0.29) (0.20) (0.26) t_Wheat -0.0054 -0.0209 -0.0205 -0.0161 -0.0191 Transcost=$10psq (0.01) (0.03) (0.03) (0.02) (0.03) t_Barley 0.0016 0.0062 0.0061 0.0048 0.0057 Edge 0.0240 - - - - (0.01) (0.04) (0.04) (0.03) (0.03) (0.04) t_Oats -0.0005 -0.0020 -0.0019 -0.0015 -0.0018 Corner -0.1417 - - - - (0.01) (0.04) (0.04) (0.03) (0.04) (0.25) t_Canola 0.0063 0.0244 0.0239 0.0188 0.0223 (0.01) (0.05) (0.05) (0.04) (0.04) Output_faba 0.0802* 10.4466 10.2652* 17.6184** 20.9953* 3588.68*** (0.04) (6.56) (6.11) (8.81) (12.27) Output_wheat -0.0004 -0.3421 -0.3361 -0.4667 -0.4343 y 0.0002 0.0002 0.0040 0.0006 (0.00) (0.30) (0.29) (0.40) (0.38) P Value: *** - 1% ** - 5% * - 10% 204

Table 5.2 –Probability of a Successful Trade – Network Trading – Random Search – Trans Costs (n=156,816) Probit LMD MUR MURRUM LAC Probit LMD MUR MURRUM LAC FLMD 0.0858 0.3101 Output_barley -0.0007** 0.3696** 0.5823** 0.8063** -0.5756** (0.27) (0.97) (0.00) (0.17) (0.26) (0.35) (0.25) FMur -0.2162 -1.2312 Output_oats -0.0036* -0.4015* -0.6327* -0.9962* -0.5963* (0.26) (1.48) (0.00) (0.23) (0.35) (0.56) (0.33) FMurrum -0.0403 -0.2059 Output_canola 0.0003 -0.0604 0.0951 0.2487 0.1154 (0.42) (2.17) (0.00) (0.13) (0.20) (0.52) (0.24) FLac -0.4656 -2.9835 Decision_faba -0.0156 -1.2515 -1.9719 -0.9848 -1.8038 (0.45) (2.87) (0.02) (1.32) (2.02) (0.99) (1.82) WLMD -0.0320 -0.1157 Decision_wheat -0.0089*** -0.4929*** -0.7766*** -0.5543*** -0.6705*** (0.07) (0.27) (0.00) (0.16) (0.23) (0.17) (0.20) WMurr -0.5354*** -3.0495*** Decision_barley -0.0088** -0.5085** -0.8013** -0.4502** -0.7330** (0.04) (0.30) (0.00) (0.24) (0.36) (0.20) (0.33) WMurrum -0.4011*** -2.0516*** Decision_oats 0.0125 -0.5065 0.7981 0.3989 0.7305 (0.05) (0.27) (0.01) (0.36) (0.57) (0.29) (0.52) WLac -0.7227*** -4.6309*** Decision_canola -0.0120*** -0.6466*** -1.0188*** -0.7270*** -1.2950*** (0.04) (0.36) (0.00) (0.24) (0.35) (0.25) (0.45) BLMD 0.0010 0.0037 Min_Pprice -0.0007*** -0.0422*** -0.0665*** -0.1074*** -0.0920*** (0.12) (0.42) (0.00) (0.01) (0.01) (0.02) (0.02) BMurr -0.5880*** -3.3491*** Reserve_Pprice -0.0002*** -0.0251*** -0.0396*** -0.0355*** -0.0445*** (0.05) (0.37) (0.00) (0.01) (0.01) (0.01) (0.02) BMurrum -0.5247*** -2.6833*** Diversity Pos -0.0109*** -1.1072*** -2.1920*** -4.4427*** -2.5908*** (0.06) (0.34) (0.00) (0.13) (0.10) (0.24) (0.14) BLac -0.7688*** -4.9263*** Diversity Neg -0.0059*** 0.6252*** 0.5281*** 1.5434*** 1.3325*** (0.05) (0.43) (0.00) (-0.08) (-0.04) (-0.09) (-0.08) OLMD -0.3100 -1.1207 Transcost=$1psq -0.3141*** -1.1354*** -1.7889*** -1.6064*** -2.0126*** (0.22) (0.88) (0.02) (0.16) (0.16) (0.13) (0.16) OMurr -0.6352*** -3.6176*** Transcost=$2psq -0.2940*** -1.0627*** -1.6745*** -1.5036*** -1.8839*** (0.07) (0.47) (0.01) (0.13) (0.12) (0.09) (0.12) OMurrum -0.5293*** -2.7070*** Transcost=$3psq -0.2552*** -0.9224*** -1.4533*** -1.3050*** -1.6350*** (0.07) (0.40) (0.01) (0.11) (0.10) (0.08) (0.09) OLac -0.9170*** -5.8761*** Transcost=$4psq -0.2238*** -0.8091*** -1.2749*** -1.1448*** -1.4343*** (0.07) (0.54) (0.01) (0.10) (0.08) (0.07) (0.08) CLMD -0.1961 -0.7087 Transcost=$5psq -0.2047*** -0.7399*** -1.1658*** -1.0469*** -1.3116*** (0.18) (0.68) (0.01) (0.09) (0.07) (0.06) (0.07) CMurr -0.4876*** -2.7772*** Transcost=$6psq -0.1894*** -0.6846*** -1.0787*** -0.9686*** -1.2136*** (0.05) (0.36) (0.00) (0.08) (0.07) (0.05) (0.07) CMurrum -0.3277*** -1.6760*** Transcost=$7psq -0.1766*** -0.6385*** -1.0061*** -0.9034*** -1.1319*** (0.08) (0.43) (0.00) (0.08) (0.06) (0.05) (0.06) CLac -0.5045*** -3.2324*** Transcost=$8psq -0.1645*** -0.5945*** -0.9367*** -0.8412*** -1.0539*** (0.06) (0.41) (0.00) (0.07) (0.06) (0.04) (0.06) t_Faba -0.0116 -0.0420 -0.0662 -0.0594 -0.0745 Transcost=$9psq -0.1526*** -0.5515*** -0.8689*** -0.7803*** -0.9776*** (0.02) (0.09) (0.13) (0.12) (0.15) (0.00) (0.07) (0.05) (0.04) (0.05) t_Wheat -0.0050* -0.0179* -0.0282* -0.0253* -0.0317* Transcost=$10psq -0.1423*** -0.5144*** -0.8104*** -0.7277*** -0.9118*** (0.00) (0.01) (0.02) (0.02) (0.02) (0.00) (0.06) (0.05) (0.04) (0.05) t_Barley -0.0043 -0.0156 -0.0245 -0.0220 -0.0276 Edge -0.0497*** - - - - (0.00) (0.02) (0.02) (0.02) (0.03) (0.02) t_Oats 0.0050 0.0180 0.0283 0.0254 0.0318 Corner -0.0916 - - - - (0.01) (0.02) (0.04) (0.03) (0.04) (0.10) t_Canola -0.0085* -0.0307* -0.0483* -0.0434** -0.0544* (0.01) (0.02) (0.03) (0.02) (0.03) Output_faba -0.0002 -0.0202 -0.0319 -0.0627 -0.0786 64029.20*** (0.01) (0.86) (1.36) (2.68) (3.35) Output_wheat -0.0002 -0.1793 0.2825 0.4497* 0.4399* y 0.0004 1.65e-08 4.21e-07 1.99e-10 (0.00) (0.11) (0.17) (0.28) (0.27) P Value: *** - 1% ** - 5% * - 10% 205

Table 5.3 introduces the simulation scenario focusing on the level of inter-regional taxes.

Under this simulation scenario, the coefficient estimates for the decision variable are consistent with expectations (and the results previously reviewed in table 5.2). Output is now a significant determinant with an increase in crop production being consistent with a reduction in the likelihood of successful trade. The level of tax imposed has a significant impact on trade at all levels tested, with a relatively constant impact across incremental levels. Table 5.4 shows the prevailing probabilities of successful trade across all the scenarios, with the tax scenarios related to a higher probability of trade than in the transaction cost simulation in most regions (not LMD). Figure 5.5 shows that the aggregate amount of trade within the tax scenario simulations remains relatively high, even in the case of an inter-regional tax rate of between 15 and 25% percent (implied by a tax rate level of 5). To conclude the review of the tax scenario simulation, Figure 5.6 presents the breakdown of trade between intermediaries and via independent bilateral negotiation. Trading through an intermediary tends to dominate with a persistent ratio of intermediary trade in comparison to independent trade of 2 to 1 across the aggregate amount of trade for all eleven periods, across all three tax levels. Hence, across all tax levels reviewed, intermediary trades occurred at a rate of 2 to 1 in comparison to independent trades. Figure 5.7 explains this consistency with a flat and regionally consistent pattern occurring in the case of taxes, which does not hold in the case of transaction costs. The stability shown in figure 5.7 also implies that taxes are notably less distortionary than transaction costs (based on the levels assumed within the analysis)

206

Table 5.3 –Probability of a Successful Trade – Network Trading – Random Search – Tax Levels (n=156,816) Probit LMD MUR MURRUM LAC Probit LMD MUR MURRUM LAC FLMD 0.1303 0.6840 Output_barley -0.0013*** -1.0015*** -0.9845*** -1.3963*** -0.8242*** (0.24) (1.25) (0.00) (0.20) (0.19) (0.27) (0.16) FMur 0.0672 -0.3467 Output_oats -0.0105*** -1.7028*** -1.6740*** -2.6995*** -1.3361*** (0.24) (1.23) (0.00) (0.28) (0.26) (0.42) (0.20) FMurrum 0.6510* 3.0884* Output_canola -0.0047*** -1.2396*** -1.2186*** -3.2647*** -1.2518*** (0.34) (1.64) (0.00) (0.18) (0.16) (0.42) (0.16) FLac 0.6842* 3.3630* Decision_faba 0.0096 1.1134 1.0945 0.5599 0.8479 (0.37) (1.80) (0.01) (1.59) (1.56) (0.81) (1.22) WLMD -0.4145*** -2.1752*** Decision_wheat -0.0092*** -0.7399*** -0.7274*** -0.5317*** -0.5318*** (0.07) (0.45) (0.00) (0.19) (0.18) (0.13) (0.13) WMurr -0.4519*** -2.3314*** Decision_barley -0.0063** -0.5229** -0.5141** -0.2958** -0.3982** (0.03) (0.21) (0.00) (0.27) (0.26) (0.15) (0.20) WMurrum 0.0407 0.1930 Decision_oats 0.0049 0.2903 0.2854 0.1461 0.2212 (0.04) (0.18) (0.01) (0.40) (0.40) (0.20) (0.31) WLac -0.3948*** -1.9405*** Decision_canola -0.0182*** -1.4253*** -1.4011*** -1.0240*** -1.5081*** (0.03) (0.19) (0.00) (0.30) (0.28) (0.20) (0.30) BLMD -0.5361*** -2.8136*** Min_Pprice -0.0009*** -0.0767*** -0.0754*** -0.1246*** -0.0883*** (0.12) (0.72) (0.00) (0.01) (0.01) (0.01) (0.01) BMurr -0.4427*** -2.2836*** Reserve_Pprice -0.0003*** -0.0718*** -0.0706*** -0.0649*** -0.0673*** (0.04) (0.25) (0.00) (0.02) (0.01) (0.01) (0.01) BMurrum -0.1620*** -0.7684*** Diversity Pos -0.0102*** -1.4949*** -1.8466*** -3.8331*** -1.8482*** (0.05) (0.22) (0.00) (0.13) (0.11) (0.19) (0.10) BLac -0.4313*** -2.1196*** Diversity Neg -0.0053*** 0.8121*** 0.4280*** 1.2810*** 0.9144*** (0.04) (0.23) (0.00) (-0.07) (-0.03) (-0.07) (-0.06) OLMD -0.8795*** -4.6156*** Taxlevel1 -0.0880*** -0.4619*** -0.4541*** -0.4176*** -0.4326*** (0.25) (1.54) (0.02) (0.10) (0.07) (0.09) (0.09) OMurr -0.6307*** -3.2537*** Taxlevel2 -0.1021*** -0.5356*** -0.5266*** -0.4843*** -0.5017*** (0.05) (0.34) (0.01) (0.07) (0.06) (0.05) (0.05) OMurrum -0.3747*** -1.7776*** Taxlevel3 -0.0838*** -0.4399*** -0.4325*** -0.3977*** -0.4120*** (0.05) (0.27) (0.01) (0.05) (0.04) (0.04) (0.04) OLac -0.7472*** -3.6726*** Taxlevel4 -0.0868*** -0.4555*** -0.4478*** -0.4118*** -0.4267*** (0.05) (0.32) (0.00) (0.04) (0.04) (0.03) (0.03) CLMD -0.3927** -2.0610** Taxlevel5 -0.0961*** -0.5046*** -0.4960*** -0.4562*** -0.4726*** (0.17) (0.96) (0.00) (0.05) (0.03) (0.03) (0.03) CMurr -0.3977*** -2.0519*** Taxlevel6 -0.0998*** -0.5238*** -0.5149*** -0.4736*** -0.4906*** (0.04) (0.26) (0.00) (0.05) (0.03) (0.03) (0.03) CMurrum -0.3695*** 1.7529*** Taxlevel7 -0.1040*** -0.5460*** -0.5368*** -0.4936*** -0.5114*** (0.07) (0.31) (0.00) (0.05) (0.03) (0.03) (0.03) CLac -0.0344 -0.1691 Taxlevel8 -0.1033*** -0.5420*** -0.5328*** -0.4900*** -0.5076*** (0.04) (0.22) (0.00) (0.05) (0.03) (0.03) (0.03) t_Faba -0.0154 -0.0810 -0.0796 -0.0732 -0.0759 Taxlevel9 -0.1033*** -0.5421*** -0.5329*** -0.4901*** -0.5077*** (0.02) (0.12) (0.12) (0.11) (0.11) (0.00) (0.05) (0.03) (0.02) (0.03) t_Wheat 0.0029 0.0152 0.0149 0.0137 0.0142 Taxlevel10 -0.1080*** -0.5669*** -0.5573*** -0.5125*** -0.5309*** (0.00) (0.01) (0.01) (0.01) (0.01) (0.00) (0.05) (0.03) (0.03) (0.03) t_Barley 0.0011 0.0059 0.0058 0.0053 0.0055 Edge 0.0005 - - - - (0.01) (0.02) (0.02) (0.02) (0.02) (0.02) t_Oats 0.0012 0.0063 0.0062*** 0.0057 0.0059 Corner 0.0392 - - - - (0.01) (0.02) (0.02) (0.02) (0.02) (0.08) t_Canola -0.0074* -0.0386* -0.0380* -0.0349* -0.0362* (0.00) (0.02) (0.02) (0.02) (0.02) Output_faba -0.0171*** -3.0451*** -2.9935*** -6.0333*** -6.2503*** 56781.99*** (0.01) (1.08) (1.05) (2.23) (2.26) Output_wheat -0.0010*** -1.276*** -1.2542*** -2.0450*** -1.6542*** y 2.06e-07 3.32e-07 2.79e-06 1.19e-06 (0.00) (0.16) (0.14) (0.23) (0.19) P Value: *** - 1% ** - 5% * - 10% 207

Table 5.4 – Probability of a Successful Trade – Summary Table LMD MUR MURRUM LAC

Random No Trans 0.0002 0.0002 0.0040 0.0006 Random Trans 0.0004 1.65e-08 4.21e-07 1.99e-10 Taxes No Trans 2.06e-07 3.32e-07 2.79e-06 1.19e-06

Figure 5.4 – Independent Trades versus Intermediary Trades – Transaction Costs

300

250

200

150

100

50 Number of Successful Number Successful of Trades 0 1 2 3 4 5 6 7 8 9 10 11

Time Periods

Indep - Trans 0 Inter - Trans 0 Indep - Trans 1 Inter - Trans 1 Indep - Trans 5 Inter - Trans 5 Note: Indep denotes Independent Trades and Inter denotes Trades Facilitated by an Intermediary

Figure 5.5 – Trading With Regional Taxes

400 350 300 250 200 150 100

50 Number of Successful Number Successful of Trades 0 1 2 3 4 5 6 7 8 9 10 11

Time Periods

TaxLevel0 TaxLevel1 TaxLevel2 TaxLevel3 TaxLevel4 TaxLevel5

Figure 5.6 – Independent Trades versus Intermediary Trades – Tax Levels

300

250

200

150

100

50 Number of Successful Number Successful of Trades 0 1 2 3 4 5 6 7 8 9 10 11

Time Periods

Indep - Taxlevel0 Inter - Taxlevel0 Indep - Taxlevel1 Inter - Taxlevel1 Indep - Taxlevel5 Inter - Taxlevel5 Note: Indep denotes Independent Trades and Inter denotes Trades Facilitated by an Intermediary

Figure 5.7 – Incremental Negative Dollar Effect on Probability of Successful Trade – Tax and Transaction Cost Levels

2.5

1.5

0.5

0 % % Change Probabilityin of Succ Trade 1 2 3 4 5 6 7 8 9 10 Incremental Factor ($ per patch for transcost and taxlevel for taxes)

LMD - Transcost Mur - Transcost Murrum - Transcost Lac - Transcost LMD - Taxes Mur - Taxes Murrum - Taxes Lac - Taxes

Note: this figure should be interpreted as showing th e negative impact of each incremental dollar of transaction cost on the probability of a successful trade.

209

5.5 Conclusion

Extending upon the agent based model developed in chapter four, chapter five reviews the impacts of introducing a Network Trading Scheme on the probability of successful trades with the incorporation of a tax-permit hybrid that allows further incorporation of supplementary environmental outcomes. The underlying motivation for the development of this model, and the inclusion of representative agents, is to examine the importance of an intermediary in reducing transaction costs and facilitating trades. If found to be important, trading designs incorporating an intermediary would be appropriate in policy prescriptions.

Such prescriptions are important as transaction costs have a notable and significant impact upon the probability of successful trade (as established in chapter four). The alternative of waiting for intermediaries to independently enter the market, may potentially result in a similar situation as that within the US, where little to no trades have occurred for a substantial period.

Within the Network Trading Scheme structure, the trader now has a decision between using an intermediary and acting independently. The trading partners linked in each option differ allowing for heterogeneity in the model and the success of localised trading or fortuitous matches over the use of an intermediary. However, as the level of distance based transaction costs increases, trades facilitated by the intermediary tend to prevail. This lends support to the requirement of intermediaries and the establishment of a market structure such as that presented in this chapter. In addition to minimising the impact of transaction costs, the same choice between assisted and unassisted trading is also subject to an inter-regional tax based on environmental indicators. This inter-regional tax allows the policymaker flexibility to modify incentives in face of localised problems and these taxes are significantly less

210 distortionary than even the lowest level of transaction costs implemented in this analysis.

Matched to the level of effluent observed these taxes may also promote self-regulation as the tax can be modified between periods based on regional effluent levels or upon proof of ‘best practice’ behaviour. With further refinements it is hoped that an agent based model such as the one applied in chapter four and five will be able to directly compare policy options and facilitate market design. As demonstrated here, applied policy recommendations in areas such as non-point source water trading would greatly benefit from such a contribution.

211

Chapter 6 – Conclusions and Plans/Recommendations for Further Research

6.1 Conclusions

This Thesis has focused on a range of issues that are related to making prescriptions for policy and, hence, working within applied environmental economics. As the ultimate test of theory is how well it applies to real-life situations, a crucial element of evaluating the usefulness of such theory is their ability to inform practical policy needs. How and why certain policies are selected will be crucial to making policy prescriptions and understanding the effectiveness of applied environmental economics. By focusing upon a range of topics and utilising a range of techniques, this Thesis provides an interesting discussion of a range of issues related to successful, and unsuccessful, applied environmental economics policy prescriptions.

As a starting point for this analysis, the first two chapters evaluate existing policies and the lessons that can be learnt from two separate cases, these being: - local domestic recycling programs, and - Intergovernmental agreements. Within Chapter two, the persistence of flat

212 fee pricing for domestic waste is a case of unsuccessful policy prescriptions as economists, such as Callan and Thomas (1999), have recommended that unit based collection be implemented. Chapter three then reviews differences between two Intergovernmental agreements on reducing emissions, one of which has been deemed to be a success. Chapters four and five then focus on an example of making a policy prescription with the development of a model that can be used for the assessment of policy issues and recommends a market structure that can minimise the impact of distortion from transaction costs.

To conclude this Thesis, this chapter reviews the inferences that can be sought from each chapter in the order that they were presented. This is then followed by a discussion of potential extensions that may be the basis of further research.

Chapter two focuses on the ineffectiveness of a flat fee pricing system in increasing the amount of recyclable material provided for collection. There is evidence that recycling rates are positively related to higher incomes and negatively related to the proportion of the region that is comprised of units/apartments which are rentals. Demand for environmental quality has often been linked to relatively higher incomes and this analysis has allowed for an income/consumption effect. As a result, the two period regression points towards a warm glow/negative association effect where households recycle due to the pleasure of doing a good deed or concern about not acting in the manner expected by peers. Additionally, lower recycling rates based on the amount of units/apartments within the council area implies that sharing large or multiple waste receptacles means that disposal decisions are not impacted by a constraint on space.

213

With the bulk of the literature on recycling focusing upon the motivating factors for undertaking recycling activities, there has been little discussion of the local government’s decision between flat rate and unit pricing collection. Despite the policy recommendation of economists for the implementation of a unit based recycling program so as to achieve efficient resource use, refer to Callan and Thomas (1999), this analysis finds that adverse selection may reinforce the adoption of a basic collection program. This is due to the basic collection program satisfying the local governments’ cost minimisation and political requirements, while any profit maximising collection agent has little incentive to offer a more intensive service. Indeed, Australian Bureau of Statistics figures show that the waste management industry tends to have government parties relying upon a high level of outsourcing to private and public-trading firms who dominate the creation and capture of revenue from recycled materials.

Chapter three finds that for CFCs within the period reviewed (1992-2008) there was a significant policy induced emission reduction in Non-Article 5 countries which were subject to targets under the Montreal Protocol. The reduction of CFCs is noted to be up to 0.8% per year above the existing decline for all countries within the same sample. However, while the

Montreal Protocol has been deemed effective, the results within section 3.3.5 show that a significant negative decline in CFC consumption between 1992 and 2008 cannot be directly attributed to the timing of the targets of the Montreal Protocol. This decline is likely to be driven by the auxiliary explanations for the success of the Montreal Protocol. These auxiliary explanations have been identified as being: - the existence of a supportive industry group, - pre-existing legislation and commitment within the United States, - affordable and available

214 substitutes, as well as - the acceptance of the underlying scientific (and Nobel prize winning) explanation of the link between CFCs and ozone depletion. Upon reviewing the emission reductions related to the Montreal and Kyoto Protocols, the existence of an Environmental

Kuznets Curve for CFCs and CO2 is doubtful as there is little evidence of such a relationship when policies are accounted for.

The primary contribution of chapter four is the establishment of a unique agent based model of bilateral negotiation based water trading. This model has been designed to include distance dependent transaction costs and spatial considerations that allow for local diversity amongst trading partners. The establishment of this model is important as it allows for the evaluation of the potential success of trading programs within differing regions with different distributions of firms growing certain crops. The incremental impact of the transaction costs on the probability of successful trade is an interesting result as the model shows that the decline in the probability of successful trades is largest with the first dollar of transaction cost imposed. The results imply that regional trading (as reflected in the neighbourhood trading scenario and figure 4.22) has lower transaction costs, which decrease in a convex manner after the first dollar of cost is imposed. Wider trading across regions (represented by the random trading simulation scenario and figure 4.23) has a substantial impact upon the probability of successful trade; however transaction costs also have a notable impact.

Extending upon the agent based model developed in chapter four, chapter five reviews the impacts of introducing a Network Trading Scheme where the trader now has a decision between using an intermediary and acting independently. The trading partners linked in each option differ and the success of localised trading or fortuitous matches competes with the use

215 of an intermediary. As the level of transaction costs increases, trades facilitated by the intermediary tend to prevail. This reinforces the importance of intermediaries and supports the establishment of a market structure in line with that presented in this chapter. An inter- regional tax implemented as part of the Network Trading Scheme also allows the policymaker flexibility to modify incentives depending upon the existence of localised problems. These taxes are significantly less distortionary than even the lowest transaction cost implemented within this analysis.

6.2 Plans and Recommendations for Further Research

Having established these results within the Thesis, it is appropriate to conclude with a discussion of the plans and recommendations for further research derived from this work.

Research tends to be a continuous process, where refinements, review and reincarnation are a necessary part of the process. The analysis within chapter two could be extended with a replication of the survey to local councils on the original motivations for the establishment of recycling programs. In addition, the development of a game theoretical structure for why flat- fee pricing tends to persist could be built on the conditions that such a collection program satisfies the local governments’ objective of cost minimisation and political requirements to constituents, while the profit maximising collection agent has little incentive to offer a more intensive service. With respect to the regression analysis, a review of the impacts of non- economic factors upon household recycling (such as rental apartments and household income) at an individual household level rather than at the council level would be preferable.

However, the author is currently unaware of such a dataset within New South Wales, nor

Australia.

216

It is unlikely that the appraisal of the success of the Kyoto Protocol will necessitate an intensive analysis due to inaction by many parties. Australia meeting its Kyoto target may be an example a case where inaction could be deemed both a success and failure. The major issue at this point in time tends to revolve around what will happen with policy action in the future. Eom et al. (2014) notes that the period between 2030 and 2050 is important for ambitious mitigation strategies and that a range of Integrated Assessment Models show that this is the period with the most rapid shift to low greenhouse gas emitting technology.

The lessons from the Montreal Protocol are mixed as the analysis within this Thesis points towards both success in reducing emissions and the importance of non-policy specific considerations. However what is clear is that intergovernmental agreements have had success in the past. With respect to the Environmental Kuznets Curve, many criticisms have been levelled at the concept, and as a result, much of the gloss has been removed from this GDP-

Emission relationship. Whilst the construction of an alternate model to validate an EKC relationship may be the aim of some, it will be very difficult to build a wide ranging model with applicability to multiple emission types. Focusing on the econometric side of the EKC relationship, testing for panel cointegration and panel unit roots was not conducted within this

Thesis as the time dimension of much of the data is limited; however this concern will be resolved as the years roll by. In any case, future research should focus on attempting to derive efficient policy induced responses, such as the attempt in chapter five of this Thesis.

The water trading model established in this Thesis is a unique one; however it can be refined with the inclusion of more crops, testing for sensitivities surrounding the cost of producing crops with the implementation of alternate cost functions, the addition of crop switching and

217 crop rotation, the ability to find efficiency improvements in the use of water, as well as utilizing firm specific data rather than representative data. The analysis has also assumed that productive efficiency is achieved by all firms, however, in reality there will be differences between firms in this respect. This assumption is important as differing efficiency of water use will result in different levels of permits for sale. Such efficiency improvements are important as a similar level of crop production may occur with less water used. The addition of exogenous seasonal weather changes will also be interesting, especially for the Murray

Darling Basin, where a drought persisted for many years and this has been labelled as the

Millennium Drought, running from 1997-2009.

As the Network Trading Scheme involves the inclusion of additional environmental targets via the inclusion of taxes, modelling the impact of water use on water quality would be very beneficial as it may allow for the approximation of the level of taxes to meet the goals assigned to each trading region. Irrespective of modelling environmental quality (which would be a challenging task), simply experimenting with hypothetical scenarios and testing the sensitivities of the necessary disparity or level of taxes that achieve the ideal trading patterns would be a considerable contribution to the literature.

In closing, the Author wishes to note that while it is hoped that this thesis provides significant contributions to important parts of the environmental economics literature, there is no doubt that there remain many challenges and issues to occupy a research agenda far into the future.

218

Appendix A – Appendix to Chapter 2

Table A1.1 – List of Councils Included Within the Dataset Ashfield, The Council of the Municipality of Inverell Shire Council Uralla Shire Council Wagga Wagga City Ballina Shire Council Junee Shire Council Council Bankstown City Council Kempsey Shire Council Walcha Council Kiama, The Council of the Bega Valley Shire Council Municipality of Warringah Council Berrigan Shire Council Kogarah Municipal Council Waverley Council Blacktown City Council Ku - ring - gai Council Weddin Shire Council Willoughby City Blayney Shire Council Lane Cove Municipal Council Council Wingecarribee Shire Blue Mountains City Council Leichhardt Municipal Council Council Wollondilly Shire Bombala Council Liverpool City Council Counci l Botany Bay, The Council of Wollongong City the City of Maitland City Council Council Woollahra Municipal Broken Hill City Council Manly Council Council Byron Shire Council Marrickville Council Young Shire Council Camden Council Moree Plains Shire Council Canterbury City Council Mosman Municipal Council Cessnock City Council Muswellbrook Shire Council Coffs Harbour City Council Nambucca Shire Council Cootamundra Shire Council Newcastle City Council Cowra Shire Council North Sydney Council Eurobodalla Shire Council Orange City Council Gloucester Shire Council Parramatta City Council Gosford City Council Penrith City Council Greater Taree City Council Pittwater Council Great Lakes Council Port Stephens Council Griffith City Council Randwick City Council Gundagai Shire Council Rockdal e City Council Gunnedah Shire Council Shellharbour City Council Guyra Shire Council Shoalhaven City Council Hawkesbury City Council Singleton Shire Council Holroyd City Council Strathfield Municipal Council Hornsby, The Council of the Shire of Sutherland Shire Council Hunters Hill, The Council of the Municipality of Tenterfield Shire Council Hurstville City Council Tweed Shire Council

220

Appendix B – Appendix to Chapter 3

221

Table B1.1 – CFC per capita – Countries within Sample (1992-2008) n = 67 Antigua and Barbuda El Salvador Mali Saudi Arabia Argentina Gambia Mexico Seychelles Australia* Ghana Nepal Sierra Leone Bangladesh Guatemala New Zealand* Solomon Islands Belarus* Guinea Nicaragua South Africa Botswana Guinea-Bissau Niger Sri Lanka Brazil Iceland* Nigeria Switzerland* Burkina Faso India Norway* Thailand Cameroon Indonesia Pakistan Trinidad and Tobago Canada* Iran Panama Tunisia Cape Verde Israel* Papua New Guinea Turkey Chile Jamaica Paraguay Uganda China Japan* Peru Ukraine* Croatia Jordan Philippines United States of America* Dominican Republic Kenya Republic of Korea Uruguay Ecuador Kyrgyzstan Russian Federation* Venezuela Malawi Rwanda Malaysia * denotes Non-Article 5 countries (n = 12)

Table B1.2 – CO2 per capita – Countries within Sample (1990-2004) n = 124 Albania Cape Verde France* Jamaica Netherlands* Samoa Turkey* Algeria Chad Gabon Japan* New Zealand* Saudi Arabia Uganda Angola Chile Gambia Jordan Nicaragua Senegal United Kingdom* Antigua/Barbuda China Germany* Kenya Niger Seychelles U R of Tanzania Argentina Comoros Ghana Lao Nigeria Sierra Leone United States* Australia* Congo Greece* Lebanon Norway* Singapore Uruguay Austria* Cote d'Ivoire Grenada Luxembourg* Pakistan Solomon Islands Vanuatu Bangladesh Cyprus Guatemala Madagascar Panama South Africa Venezuela Belgium* D R of the Congo Guinea Malawi Papua New Spain* Viet Nam Guinea Belize Denmark* Guinea-Bissau Malaysia Paraguay Sri Lanka Yemen Benin Dominica Guyana Mali Peru Sudan Zambia Bolivia Dominican Honduras Malta Philippines Swaziland Republic Botswana Ecuador Hungary* Mauritania Poland* Sweden* Brazil Egypt Iceland* Mauritius Portugal* Switzerland* Bulgaria* El Salvador India Mexico Romania* SAR Burkina Faso Ethiopia Indonesia Mongolia Rwanda Thailand Burundi Finland* Iran Morocco Saint Lucia Togo Canada* Ireland* Mozambique Saint Vincent Tonga Israel Namibia Trinidad/Tobago Italy* Nepal Tunisia * denotes Annex A countries (n = 28)

223

Appendix D – Appendix to Chapter 5

225

Table D1 – List of Variables for Regressions in Chapter Five Label Definition Variable Name Definition FLMD Intercept for Faba Producers Growers in the LMD t_Oats Timetrend for Oat Producers FMur Intercept for Faba Producers Growers in the Mur t_Canola Timetrend for Canola Producers FMurrum Intercept for Faba Producers Growers in the Murrum Output_faba Output Level of Faba FLac Intercept for Faba Producers Growers in the Lac Output_wheat Output Level of Wheat WLMD Intercept for Wheat Producers in the LMD Output_barley Output Level of Barley WMurr Intercept for Wheat Producers in the Mur Output_oats Output Level of Oats WMurrum Intercept for Wheat Producers in the Murrum Output_canola Output Level of Canola WLac Intercept for Wheat Producers in the Lac Decision_faba Decision Variable of Faba Producers BLMD Intercept for Barley Producers in the LMD Decision_wheat Decision Variable of Wheat Producers BMurr Intercept for Barley Producers in the Mur Decision_barley Decision Variable of Barley Producers BMurrum Intercept for Barley Producers in the Murrum Decision_oats Decision Variable of Oat Producers BLac Intercept for Barley Producers in the Lac Decision_canola Decision Variable of Canola Producers OLMD Intercept for Oat Producerss in the LMD Min_Pprice Prevailing Permit Price OMurr Intercept for Oat Producerss in the Mur Reserve_Pprice Permit Reserve Price OMurrum Intercept for Oat Producerss in the Murrum Diversity Pos Positive Diversity OLac Intercept for Oat Producerss in the Lac Diversity Neg Negative Diversity CLMD Intercept for Canola Producers in the LMD Transcost=$1psq-$10psq Transaction Cost - $1 to $10 per square CMurr Intercept for Canola Producers in the Mur Taxlevel1-10 Tax Level - 1 to 10 CMurrum Intercept for Canola Producers in the Murrum Edge Agent is on the Edge of the Area CLac Intercept for Canola Producers in the Lac Corner Agent is in the Corner of the Area t_Faba Timetrend for Faba Producers Growers t_Wheat Timetrend for Wheat Producers t_Barley Timetrend for Barley Producers

226

Bibliography

Abdalla, C., Borisova, T., Parker, D., and Saacke Blunk, K. (2007) ‘Water Quality Credit

Trading and Agriculture: Recognizing the Challenges and Policy Issues Ahead’, Choices,

Vol.22, No. 2: 117-124.

ACCC (2006) A regime for the calculation and implementation of exit, access and

termination fees charged by irrigation water delivery businesses in the southern

Murray–Darling Basin. Advice to the Australian, New South Wales, South Australian

and Victorian Governments.

Ando, A. and Gosselin, A. (2005) ‘Recycling in Multifamily Dwellings: Does Convenience

Matter?’, Economic Inquiry, Vol. 43, No. 2: 426-438.

Auffhamer, M., Morzuch, B., and Stranlund, J. (2005) ‘Production of Chloroflurocarbons in

Anticipation of the Montreal Protocol’, Environmental and Resource Economics, 30,

377-391.

Australian Bureau of Statistics (2008) ‘4610.0.55.007 - Water and the Murray-Darling Basin

- A Statistical Profile, 2000-01 to 2005-06’.

227

Australian Bureau of Statistics (2009) ‘4602055002DO001_200903 Environmental Issues:

Waste Management and Transport Use’.

Australian Bureau of Statistics (2010) ‘6401.0 - Consumer Price Index’, Australia, Jun 2010.

Australian Bureau of Statistics (2010a) ‘71210DO001_200809 - Agricultural Commodities,

Australia, 2008-09’.

Australian Bureau of Statistics (2010b) ‘46180DO001_200809 - Water Use on Australian

Farms, 2008-09’.

Australian Bureau of Statistics (2011) ‘86980DO001_200910 Waste Management Services,

Australia, 2009-10’.

Australian Government Department of the Environment, Water, Heritage and the Arts (2005)

Natural Resource Management (NRM) Region Boundaries (formerly known as Natural

Heritage Trust II (NHT2) Region Boundaries), Spatial Data and Meta Data.

Australian Natural Resources Atlas (2007) Australian Agriculture Assessment 2001 -

Appendix 1 — River basin budgets - River nutrients.

Axelrod, R. (1984), The Evolution of Cooperation, Basic Books.

Barrett, S. (1990) ‘The problem of global environmental protection’, Oxford Review of

Economic Policy, Vol. 6 No 1: 68-79.

Beede, D. and Bloom, D. (1995) ‘The Economics of Municipal Solid Waste’, The World

Bank Research Observer, Vol. 10, No. 2: 113-150.

Bellenger, M.J. and Herliny, A.T. (2009) ‘An Economic Approach to Environmental

Indices’, Ecological Economics, Vol. 68: 2216-2223.

228

Berger, T., 2001. ‘Agent-Based Models Applied to Agriculture: A Simulation Tool for

Technology Diffusion, Resource Use Changes and Policy Analysis’. Agricultural

Economics, 25 (2/3), 245-260.

Berger, T., R. Birner, J. Diaz, N. McCarthy & H. Wittmer. (2007). ‘Capturing the complexity

of water uses and water users within amulti-agent framework’. Water Resources

Management 20:129–148.

Boutilier, C., Shoham, Y., and Wellman, M. (1997) ‘Economic Principles of Multi-Agent

Systems’, Artificial Intelligence, Vol. 94, No. 1-2: 1-6.

Callan, S. and Thomas, J. (1999) ‘Adopting a Unit Pricing System for Municipal Solid

Waste: Policy and Socio-Economic Determinants’, Environmental and Resource

Economics, Vol. 14: 503-518.

Callan, S. and Thomas, J. (2007) Environmental Economics and Management: Theory,

Policy and Applications, Cincinnati, OH, South-Western College Publications.

Carey, J.M., Sunding, D. and Zilberman, D. (2002). ‘Transactions Cost and Trading Behavior

in an Immature Water Market’. Environment and Development Economics, 7: 733-750.

Carpenter, S.R., Caraco, N.F., Correll, D.L., Howarth, R.W., Sharpley, A.N. and Smith, V.H.

(1998) ‘Nonpoint pollution of surface waters with phosphorus and nitrogen’. Ecological

Applications 8(3), 559-568

Cason, T. N., and Gangadharan, L. (2003) ‘Transactions Costs in Tradable Permit Markets:

An Experimental Study of Pollution Market Designs’. Journal of Regulatory Economics

23(2): 145–165.

229

Challen, R. (2000) ‘Institutions, Transaction Costs and Environmental Policy’, Edward Elgar

Publishing.

Coase, R. (1937). ‘The Nature of the Firm’. Economica 4 (16): 386–405.

Coase, R. (1992) ‘The Institutional Structure of Production’, American Economic Review,

82, 4: 713-719.

Cole, M., Rayner, A., and Bates, J. (1997) ‘The environmental Kuznets curve: an empirical

analysis’, Environment and Development Economics, 2, 401-416.

Collentine, D. (2005). ‘Phase-In of Nonpoint Sources in a Transferable Discharge Permit

System for Water Quality Mangement: Setting Permit Prices’. [Towards More Integrated

Water Resources Management]. Ambio, 34(7), 573-578.

Connell, D., Grafton, R, Q, 2008, ‘Planning for Water Security in the Murray-Darling Basin’,

Public Policy, Vol 3, No 1, pp 67-86.

Dasgupta, S., Laplante, B., Wang, H., and Wheeler, D. (2002) ‘Confronting the

environmental Kuznets Curve’, Journal of Economic Perspectives, 16, 147–168.

Davies, P.E., Harris, J.H., Hillman, T.J. & Walker, K.F. (2008) ‘Sustainable Rivers Audit

Report 1: a report on the ecological health of rivers in the Murray–Darling Basin, 2004–

2007’, Report prepared by the Independent Sustainable Rivers Audit Group for the

Murray–Darling Basin Ministerial Council, Murray–Darling Basin Commission,

Canberra. de Bruyn, S. (1997). ‘Explaining the Environmental Kuznets Curve: Structural Change and

International Agreements in Reducing Sulphur Emissions’, Environment and

Development Economics, 2, 485-503.

230

DeCanio, S. (2003) ‘Economic Analysis, Environmental Policy, and Intergenerational Justice

in the Reagan Administration The Case of the Montreal Protocol’ International

Environmental Agreements: Politics, Law and Economics, vol. 3(4), pages 299-321,

December.

Department of Environment and Climate Change - DECC (2009) ‘New South Wales State of

the Environment 2009’.

Dept for Environment, Food and Rural Affairs, UK (2009) ‘Annex 3: International

Experience, Economic Instruments for Water Pollution’,

(http://www.defra.gov.uk/environment/water/quality/econinst1/eiwp09.htmU- accessed 6

March 2009).

DeSombre, E. R. (2001) ‘The Experience of the Montreal Protocol: Particularly Remarkable

and Remarkably Particular’. UCLA Journal of Environmental Law and Policy, 19 (1):

49–81.

Devlin R.A. and R.Q. Grafton (1996). ‘Marketable emission permits: efficiency profitability

and substitutability.’ Canadian Journal of Economics, XXIX Special Issue, S260-64.

Dowd, B.M., Press, D., & Los Huertos, M. (2008). ‘Agricultural non-point source water

pollution policy: the case of California‘s Central Coast’. Agriculture, Ecosystems &

Envronment, 128, 151-161.

Ekins, P. (2000). Economic growth and environmental sustainability: the prospects of green

growth. Routledge, New York.

Environmental Protection Agency (2006) Solid Waste Management and Greenhouse Gases A

Life-Cycle Assessment of Emissions and Sinks. Chapter 3 – Source Reduction and

Recycling.

231

Fan, C., & Zheng, X. (2013). An empirical study of the environmental kuznets curve in

sichuan province, china. Environment and Pollution, 2(3): 107-115.

Folz, D. (1999) ‘Municipal Recycling Performance: A Public Sector Environmental Success

Story’, Public Administration Review, Vol. 59, No. 4: 336-345.

Fullerton, D. and Kinnaman, T. (1996) ‘Household Responses to Pricing Garbage by the

Bag’, American Economic Review, Vol. 86, No. 4: 971-984.

Geoscience (1997) Australia's River Basins 1997, Spatial Data and Meta Data.

(HUhttp://www.ga.gov.au/meta/ANZCW0703005427.htmlUH - accessed 21 Nov 2008)

Grafton, R. Q., Landry, C., Libecap, G. D., and O’Brien, J. R. (2009) ‘Markets Water

Markets: Australia’s Murray-Darling Basin and the US Southwest’, International Centre

for Economic Research, Working Paper No. 15/2009.

Grafton, R.Q. and K. Hussey. ‘Buying Back the Murray: At What Price?’ Australasian

Journal of Environmental Management 14(2): 74-81, 2007.

Griffin, R. and Bromley, D.W. (1982) ‘Agricultural Runoff as a Nonpoint Externality’.

American Journal of Agricultural Economics 64: 547-52.

Grimm, V. et. al. (2006) ‘A Standard Protocol for Describing Individual-Based and Agent-

Based Models’, Ecological Modelling, Vol. 198, No. 1-2: 115-126

Grossman, G. M., and Krueger, A. B. (1991) ‘Environmental impacts of a North American

Free Trade Agreement’, National Bureau of Economic Research Working Paper 3914,

NBER, Cambridge MA.

Grossman, G. M., and Krueger, A. B. (1995) ‘Economic growth and the environment’,

Quarterly Journal of Economics, 110, 353–377.

232

Grubb, M. (2003) ‘The Economics of the Kyoto Protocol’ World Economics, vol. 4(3), pages

143-189.

Grubb, M. and Yamin, F. (2001) ‘Climatic Collapse at The Hague: What Happened, Why,

and Where Do We Go From Here?’, International Affairs, 77, 261-276.

Hahn, R. (1989) ‘Economic Prescriptions for Environmental Problems: How the Patient

Followed the Doctor's Orders’, Journal of Economic Perspectives, Vol. 3, No. 2: 95–114.

Halvorsen, B. (2004) ‘Effects of norms, warm-glow and time use on household recycling’,

Statistics Norway – Discussion Paper, No. 389: 1-31.

Halvorsen, B. (2012) ‘Effects of norms, warm-glow and time use on household recycling: An

international comparison’, Resources, Conservation and Recycling, 67: 18-26.

He, J. and Richard, P. (2010) ‘Environmental Kuznets Curve for CO2 in Canada’, Ecological

Economics, 69: 1083-1093.

Hoag, D. L., and Hughes-Popp, J.S. (1997). ‘Theory and Practice of Pollution Credit Trading

in Water Quality Management’ Review of Agricultural Economics 19(2),.

Hung, M. & Shaw, D. (2005) ‘A trading-ratio system for trading water pollution discharge

permits’ Journal of Environmental Economics and Management, vol. 49(1), pages 83-

102.

Jacobides, M. and Winter, S. (2005) ‘The Co-Evolution of Capabilities and Transaction

Costs: Explaining the Institutional Structure of Production’, Strategic Mgmt. Journal, 26:

395-413.

Jarvie, M., Solomon, B.D., (1998) ‘Point-nonpoint effluent trading in watersheds: a review

and critique’. Environ. Impact Assess. Rev. 18 (2), 135–157

233

Jenkins, R., Martinez, S., Palmer, K. and Podolsky, M. (2003) ‘The Determinants of

Household Recycling: A Material-Specific Analysis of Recycling Program Features and

Unit Pricing’, Journal of Environmental Economics and Management, Vol. 45: 294-318.

Kahneman, Daniel, Knetsch, J. and Thaler, R. (1990) ‘Experimental Tests of the Endowment

Effect and the Coase Theorem’ Journal of Political Economy, 98, 1325-1348.

Kennedy, P. (1992) ‘A Guide to Econometrics’. Blackwell. Third Edition.

Kinnaman, T. (2005) ‘Why do Municipalities Recycle?’, Topics in Economic Analysis and

Policy, Vol. 5, No.1: 1294.

Kinnaman, T. and Fullterton, D. (1999) ‘The Economics of Residential Solid Waste

Management’, National Bureau of Economic Research Working Paper 7326: 1-44.

Kohn, R. E., (1991) ‘Porter’s Combination Tax and Subsidy for Controlling Pollution’

Journal of Environmental Systems, Vol. 20(3).

Kozloff, K., Taff, S.J. and Wang, Y. (1992). ‘Microtargeting the acquisition of cropping

rights to reduce nonpoint source water pollution’. Water Resources Research 28(3).

Lansing, J. and Kremer, J. (1993) ‘Emergent properties of Balinese water temples’. American

Anthropologist 95 (1), 97–114.

Lavee, D. (2007) ‘Is Municipal Solid Waste Recycling Economically Efficient?’,

Environmental Management, Vol. 40: 926-943.

LeBaron, B. (2006) ‘Agent-based Computational Finance’, Handbook of Computational

Economics, in: Leigh Tesfatsion & Kenneth L. Judd (ed.), Handbook of Computational

Economics, edition 1, volume 2, chapter 24, pages 1187-1233 Elsevier.

234

Lee, L. and Ancev, T. (2009) ‘Two Decades of Murray-Darling Water Management: A River

of Funding, a Trickle of Achievement’, Agenda, Vol. 16, No.1.

Lin, C. and Liscow Z. D. (2013) Endogeneity in the Environmental Kuznets Curve: An

Instrumental Variables Approach, Am. J. Agr. Econ., 95 (2): 268-274.

Local Government Association of South Australia (2009) Analysis of Markets for Council

Collected Recyclables. (Prepared by Geoff Johnston of OneWorld Environmental

Solutions)

Marks, R. (2006) ‘Market Design Using Agent-Based Models’, found within Tesfatsion, L.

and Judd, K. (2006) Handbook of Computational Economics, Vol. 2: Agent-Based

Computational Economics.

Mason, R. and Swanson, T. (2003) ‘A Kuznets curve analysis of ozone-depleting substances

and the impact of the Montreal Protocol’, Oxford Economic Papers, 55, 1-24.

Matthews, R., Gilbert, N., Roach, A., Polhill, J., Gotts, N. (2007) ‘Agent-based land-use

models: a review of applications’. Landscape Ecology 22, 1447–1459.

McCann, L., Colby, B., Easter, K., Kasterine, A. and Kuperan, K. (2005) ‘Transaction cost

measurement for evaluating environmental policies’. Ecological economics, 52.

Miranda, M. and Aldy, J. (1998) ‘Unit Pricing of Residential Municipal Solid Waste: Lessons

from Nine Case Study Communities’, Journal of Environmental Management, Vol. 52:

79-93.

Montero, J. (1998) ‘Marketable pollution permits with uncertainty and transaction costs’

Resource and Energy Economics, vol. 20(1), pages 27-50, March.

235

Montgomery W.D. (1972) ‘Markets in licenses and efficient pollution control programs’.

Journal of Economic Theory 5, pp. 395-418

Morgan, C. and Wolverton, A. (2005) ‘Water Quality Trading in the United States’, EPA

Working Paper #05-07, June.

Murdoch, J. C. & Sandler, T. (1997) ‘The voluntary provision of a pure public good: The case

of reduced CFC emissions and the Montreal Protocol’ Journal of Public Economics, vol.

63(3), pages 331-349.

Nahman, A. and Antrobus, G. (2005) ‘The environmental Kuznets curve: a literature survey’,

South African Journal of Economics, 73, 105-120.

National Water Commission (2008) ‘Australian Water Markets Report 2007-2008’

New South Wales Department of Industry and Investment (2010) ‘Central Gross Margin

Budgets – Winter’.

New South Wales Department of Industry and Investment (2010a) ‘Murray Gross Margin

Budgets - Winter’.

New South Wales Department of Industry and Investment (2010b) ‘Murrumbidgee Gross

Margin Budgets - Winter’.

New South Wales Department of Industry and Investment (2010c) ‘Murrumbidgee Valley

Gross Margin Budgets - Winter’.

New South Wales Department of Industry and Investment (2010d) ‘Winter crop variety

sowing guide 2010’.

New South Wales Department of Primary Industries (2007) (accessed on 1 Oct 2009 –

http://www.dpi.nsw.gov.au/agriculture/farm-business/budgets/about/intro.html)

236

Nguyen, T. N., et al. (2006) ‘A Guide to Market-Based Approaches to Water Quality’,

(HUhttp://edu.nutrientnet.org/UH - accessed 6 March 2009).

Nooteboom, B. (1993) ‘An analysis of specificity in transaction cost economics’.

Organization Studies 14, 3, 443–451.

NSW DECCW(2009) New South Wales State of the Environment 2009 (accessed via:

http://www.environment.nsw.gov.au/soe/soe2009/chapter3/print_chp_3.4.htm)

NSW Dept. of Environmental and Climate Change (2007) Review of the Waste Avoidance

and Resource Recovery Act 2001

(http://www.environment.nsw.gov.au/education/SPD_WARR2001_Review.htm)

NSW Dept. of Local Govt. (2006) Comparative Information on NSW Local Government

Councils 2004/2005.

NSW Dept. of Local Govt. Comparative Information on New South Wales Local Government

Councils 1994/95 - 2003/04: Excel Data File.

(www.dlg.nsw.gov.au/dlg/dlghome/documents/comparatives/0405data.xls)

Oskamp, S., Harringon, M. J., Edwards, T. C., Sherwood, D. L., Okuda, S. M., & Swanson,

D. C. (1991) ‘Factors influencing household recycling behavior’, Environment and

Behavior., 23(4): 494-519.

Panayotou, T. (1997) ‘Demystifying the environmental Kuznets curve: Turning a black box

into a policy tool’, Environment and Development Economics, 2, 465–484.

Perman, R. and Stern, D. (2003) ‘Evidence from panel unit root and cointegration tests that

the Environmental Kuznets Curve does not exist’, Australian Journal of Agricultural and

Resource Economics, 47, 325-347.

237

Peterson D, Dwyer G., Appels D. & Fry J.M. (2004) ‘Modelling water trade in the southern

Murray–Darling Basin’. Productivity Commission, Staff Working Paper, Nov-04,

Melbourne.

Planet Ark (2005) ‘Ten Years of Recycling - The Good, The Bad & The Ugly’. A Planet Ark

Recycling Report - November 2005. (accessed via:

http://recyclingnearyou.com.au/documents/05nrw_gbugly_report-1.pdf)

Policy Research Initiative (2006) Can Water Quality Trading help to address agricultural

sources of pollution in Canada?, Final report, May.

Qureshi, M.E., Shi, T., Qureshi, S., Proctor, W. and Kirby, M. (2009) ‘Removing Barriers to

Facilitate Efficient Water Markets in the Murray Darling Basin – A Case Study from

Australia’, CSIRO Working Paper Series 2009-02, Canberra.

Rasmusen, E. (2007) Games and Information: An Introduction to Game Theory, 4th Ed.,

Blackwell Publishing.

Ribaudo, M. O., Horan, R. and Smith, M. (1999). ‘Economics of Water Quality Protection

from Nonpoint Sources: Theory and Practice.’ AER Pub. No. 782, U.S. Department of

Agriculture/Economic Research Service.

Rolfe, J. (2012) ‘Water Trading and Market Design’, within Crase, L. ‘Water Policy in

Australia: The Impact of Change and Uncertainty’, Routledge.

Roth, A. (2002) ‘The Economist as Engineer: Game Theory, Experimentation, and

Computation as Tools for Design Economics’, Econometrica, Vol. 70, No. 4: 1341-1378.

Schary, C. and Fisher-Vanden, K. 2004, ‘Applying the Acid Rain Program’s Cap and Trade

Approach to Water Quality Trading’ Journal of Environmental Practice, 6(4), 1-15.

238

Schelling, T. C. (1971). ‘Dynamic Models of Segregation’ Journal of Mathematical

Sociology, 1(2), pp. 143–186.

Sephton, P. and Mann, J. (2013) ‘Further evidence of an Environmental Kuznets Curve in

Spain’, Energy Economics, Volume 36, March 2013: 177-181.

Shabman, L. and Stephenson, K. (2007) ‘Achieving Nutrient Water Quality Goals: Bringing

Market-like Principles to Water Quality Management’ Journal of American Water

Resources Association 43 (4): 1076-1089.

Shahbaz, M., Mutascu, M. and Azim, P. (2013) ‘Environmental Kuznets curve in Romania

and the role of energy consumption’, Renewable and Sustainable Energy Reviews,

Volume 18: 165–173.

Smajgl, A., Carlin, G., House, A., Pambudhi, F., Butler, J., Bohensky, E., et al. (2009)

‘Simulating Pathways to Sustainability in Indonesia: Design Document for Agent-based

Model SimPaSI East Kalimantan’. CSIRO Sustainable Ecosystems, Townsville.

Stavins, R.N. (1993) ‘Transaction Costs and the Performance of Markets for Pollution

Control’ Resources For the Future, Discussion Paper QE93-16.

Stavins, R.N. (1995) ‘Transaction Costs and Tradeable Permits’ Journal of Environmental

Economics and Management 29: 133-48.

Stephenson, K., Norris, P. and Shabman, L. (1998) ‘Watershed-Based Effluent Trading: The

Nonpoint Source Challenge’ Contemporary Economic Policy, Western Economic

Association International, vol. 16(4), pages 412-421, October.

Stern, D. (2004) ‘The rise and fall of the environmental Kuznets curve’, World Development,

32, 1419-1439.

239

Stiglitz, J. (2004) ‘Information and the Change in the Paradigm in Economics’ within

Szenberg, M. and Ramrattan, L. (2004) New Frontiers in Economics, Cambridge Uni

Press.

Tesfatsion, L. (2006) ‘Agent-Based computational Economics: A Constructive Approach to

Economic Theory’, found within Tesfatsion, L. and Judd, K. (2006) Handbook of

Computational Economics, Vol. 2: Agent-Based Computational Economics.

Tiwari, A. K., Shahbaz, M. and Hye, Q. (2013) ‘The environmental Kuznets curve and the

role of coal consumption in India: Cointegration and causality analysis in an open

economy’, Renewable and Sustainable Energy Reviews, Volume 18: 519-527.